Carolopedia

A public encyclopedia of Carolverse.

๐Ÿ“– Carolopedia โ€บ Guides โ€บ Rulings RepoGuide page

Rulings Repo

Rulings about the estate's own security are not published here.

๐Ÿ“–Agent Chat

Carol recommends a new user through scheduled processes she owns, never her budget-bound wake; a user asks for Leo's chat through Carol's WhatsApp or web chat and her droids pick it up

Ruled 2026-11-03 by ninad -- in force

CAROL RECOMMENDS A NEW USER THROUGH SCHEDULED PROCESSES SHE OWNS, NEVER HER BUDGET-BOUND WAKE; A USER ASKS FOR ACCESS TO LEO'S CHAT THROUGH CAROL'S WHATSAPP OR WEB CHAT AND THE DROIDS THAT PROVISION IT PICK IT UP; CAROL OWNS THOSE DROIDS (Ninad, CLI-020, 2026-11-03). HIS WORDS: "2 seperate issues.. first, carol's wake cycle budget needs rectification.. this should be handled by Midas.. he observes the average daily requirements of the rolling week and adjusts the budgets accordingly (atleast he is supposed to do that if he is not doing that already).. Second, regarding recommending Sachin Joshi to blueprint as a user, carol should do that as a scheduled process, not as a wakeup activity.. wakeup activity is budget constraint, and this recommendation is part of carol's primary duties.. a user can request access to leo chat via carol whatsapp or web chat, and it should be picked up by the droids that are responsible to provision this access. Carol should own these droids.." READ AS (Orion's reading, not his words): (1) recommending a person's access is one of Carol's primary duties, so it runs as a registered scheduled process of hers under her own login, not as an action of her mind wake, which a budget may silence; (2) a user's ask for access to Leo's chat made in Carol's WhatsApp or web chat becomes an access request that her droids pick up, beside the request a web sign-in already records; (3) the droids that pick up and provision this access are Carol's; which of the access machinery Jarvis keeps today moves to her is asked of Ninad separately; (4) the estate operator's or admin's approval (#571) is unchanged.

Carol recommends a person's access during her chat with them, at once; the apps showing user onboarding status for Carol and Leo are built in pursuit User rulings built

Ruled 2026-11-03 by ninad -- in force

CAROL RECOMMENDS A PERSON'S ACCESS DURING HER CHAT WITH THEM, AT ONCE; THE APPS SHOWING USER ONBOARDING STATUS FOR CAROL AND LEO ARE BUILT IN PURSUIT USER RULINGS BUILT (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: told that Carol's new scheduled recommender (whatsapp #582) had recommended Sachin Joshi's waiting request and that her timetable needed his password, he answered: HIS WORDS: "Carol is supposed to recommend during her chat with the user.. this should happen instantly when the chat is on.. are you sure this is not yet built? I thought that the machinery was all in place.. even the apps that showed user onboarding status for carol and Leo should be built.. if they are not, file them in this pursuit". READ AS (Orion's reading, not his words): (1) when a person asks Carol for a service in her WhatsApp or web chat, the access request and her recommendation are recorded in that same turn, not left for a later process; the scheduled recommender remains only for requests that arrive with no chat (a web sign-in); (2) measured the same hour: her chat's request goes to Clara's Subscription Approver, which approves admin and verified users itself and queues guests in Clara's own table, so nothing reaches Jarvis's requests and an agent decides access (against operations #571, #583) -- that path is rebuilt; (3) no app shows a person's request, Carol's recommendation and the decision, or who may reach Leo's chat -- filed in this pursuit.

The Agent Chat service is keyed chat and its core track chat-core everywhere; WhatsApp names only the channel

Ruled 2026-11-18 by ninad -- in force

THE CHAT SERVICE'S KEYS.

HIS WORDS: "1 review comment - whatsapp --> chat". Orion asked which he meant -- the service is already named 'Agent Chat', but its key is still 'whatsapp' and its core track's 'whatsapp-core', which is what the handbook shows, keyed in 18 registry tables and 9 code files -- offering "Rename the keys" or "Change the label only". HIS ANSWER: "Rename the keys (Recommended)" -- "'whatsapp' becomes 'chat' and 'whatsapp-core' becomes 'chat-core' everywhere. Filed as new work in Lightweight projects, about 2 hours, split into 1-hour parts."

READ AS (Orion's reading, not the administrator's words): (1) the Agent Chat service is keyed 'chat' and its core track 'chat-core' in every record and in code; (2) WhatsApp keeps naming only the CHANNEL -- the Meta number, Carol's WhatsApp server, the shared WhatsApp client module and its sends -- which is a channel of the chat service, not the service; (3) the other tracks of the service keep their own keys and move with it; (4) the rename is filed as two parts of at most an hour each (initiatives #814).

Voice chat's MVP: an agent places a real phone call from the web app to a person's phone number through Twilio, each call tracked to that agent; WhatsApp is not in the MVP

Ruled 2026-11-20 by ninad -- superseded 2026-11-20

VOICE CHAT'S MVP IS AN AGENT-PLACED PHONE CALL FROM THE WEB APP, TRACKED TO THE AGENT (Ninad, CLI-265, 2026-11-20; on CAROL-PUR-0092, fast track; after blueprint #1081)

THE QUESTIONS AND HIS WORDS: asked to approve the Voice chat track as as-is, offered, Galadriel's machinery with the project's own agent in the conversation: "Each conversation should be tracked to agents. Galadriel provides the necessary tools that agents need to communicate. Agents should be able to call and talk to customers via phone / webapp / whatsapp.. check if this is possible first". Shown what exists (web-app live voice calls, Carol only; WhatsApp voice notes, Carol only; no telephony): "MVP should be web app voice call, initiated by an agent to a human with a phone number, is that possible?". Shown a WhatsApp tap-to-talk link: "whatsapp should not be in MVP, because you need an active chat within 24 hrs window to initiate a message.. I want a web call via phone number to another phone number as the scope for MVP". Shown that this needs a telephony provider bought through Marco: "Approved, Twilio account already exists, I think"; asked the number's country: "Indian number, what does it mean when you say it will take longer? how long? for POC what are other alternatives that can be implemented within a day".

READ AS (Orion's reading, not the administrator's words): (1) every voice conversation is tracked to the agent having it; Galadriel's Agent Chat provides the tools agents use to communicate; (2) the MVP: from the web app, an agent places a real phone call from Carolverse's number to a person's phone number, the person answers and talks with that agent live through the existing voice pipeline, the call billed to that agent; (3) WhatsApp is not in the MVP (a business may start a conversation only inside 24 hours of the person's last message, or with a Meta-approved template); (4) the provider is Twilio, bought or reused through Marco on the administrator's approval -- the estate holds no Twilio settings today (checked CLI-265); (5) Twilio's India voice guidelines: calls to Indian numbers must come from a NON-Indian number, a +91 caller id may not place calls, and only toll-free +91 numbers (inbound) can be bought -- an Indian caller id needs an Indian carrier over SIP trunking, a longer route; a proof in a day: a non-Indian Twilio number, India allowed in its calling permissions, calling consenting people (TRAI consent and calling hours apply); the administrator chooses.

Ended because: Ninad re-scoped the Voice chat MVP in CLI-266: a web call to a new sandbox sales agent, no Twilio

Voice chat's MVP: a person calls an agent from the web and the agent talks -- a new sales agent of a new sandbox project, speaking from the sales prompt built for Carol, its prompt assigned from a catalogue of role-specific prompts; no Twilio in the MVP

Ruled 2026-11-20 by ninad -- in force

VOICE CHAT'S MVP IS A WEB CALL TO A NEW SANDBOX SALES AGENT, AND ROLE PROMPTS LIVE IN A CATALOGUE (Ninad, CLI-266, 2026-11-20; on CAROL-PUR-0092, fast track; supersedes whatsapp #1083's MVP)

THE QUESTIONS AND HIS WORDS: asked which caller number the first version's calls should come from (Twilio cannot place calls from a +91 number; a non-Indian Twilio number works in a day; an Indian carrier link takes weeks): "Change the scope of mvp to call an agent from the web, where agent talks, this capability already exists, but this should be clubbed with create a new project and a new agent on sandbox who can talk to you when you call up the agent.. this is going to be a sales agent, so use the "sales prompt" that was built for carol.. create role specific prompts and build a prompt catalogue, so that athena / leo / or whoever is incharge can assign the right prompt to the agent being built..". Asked whether Carolverse already has a Twilio account: "MVP may not need twilio for now".

READ AS (Orion's reading, not the administrator's words): (1) the Voice chat MVP is a person calling an agent from the web and talking with it live, on the web voice call that exists today; (2) the agent called is a NEW sales agent of a NEW project on the sandbox machine, made for this, and it speaks from the sales prompt built for Carol; (3) role-specific prompts are kept in ONE prompt catalogue, and whoever builds an agent -- Athena for an agent, Leo for a project's roster, or whoever is in charge -- assigns the right prompt from it to the agent being built; (4) Twilio and a call to a phone number are left out of the MVP ("for now"), so #1083's agent-placed phone call is superseded as the MVP and nothing is bought; (5) #1083's other words stand: each conversation is tracked to the agent having it, and Galadriel provides the tools agents need to communicate.

Voice chat track: grade, offer and agent decision

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-290): "Approve the new Voice chat track for Galadriel's Agent Chat? Proposal: graded 'as-is' (projects use the running voice line and nothing is copied for them, because the line has no logic of its own and just carries whichever agent is called, in that agent's own persona and voice). Offered to projects. Galadriel's part: run the line, check every spoken reply, track and bill each call to the agent spoken to. A project's own agents supply only what the agent says. Each call bills the agent spoken to under Voice chat, and the spoken words bill as Speech. Carol's external chat window gets calling first, then every agent's chat window." Ninad answered: "Approve as proposed (Recommended)". Earlier in the same session: "ok, go ahead, it looks like the POC for mBank may take a while, hopefully this happens quickly" (on adding a calling feature to Carol's external chat window using the existing voice chat service). READ AS: the Voice chat track of Agent Chat is born graded as-is, offered to projects, owner part and project agents' part as stated; Carol's external chat window is the first door of the voice build, then every agent's chat window.

Voice calls on Carol's chat run in live mode

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-309): "For voice calls on Carol's chat, which mode should I build first for today's 20:00 sync? Live mode: Gemini hears and speaks directly, answers in under a second and can be interrupted, but Carol's reply skips her one chat engine and the Voice Gatekeeper only checks the transcript after she speaks (a change to today's rules). Chained mode: hear, then think, then speak, sentence by sentence, about 2-3 seconds per turn, keeping her engine and the gatekeeper exactly as ruled." Options offered: Live mode; Chained mode; Chained now, live next. Ninad answered: "Live mode". Asked before it: "is there a way to be able to converse fluently, especially with newer versions of googles AI models?" READ AS: a voice call from an agent's chat window (Carol's external window first, whatsapp #1109) is a live speech-to-speech conversation on a Gemini live model: Gemini hears the person and speaks the reply itself, carrying the agent's persona from its one identity card; the reply does not pass through the agent's one chat engine before it is spoken, and the Outgoing Voice Gatekeeper checks the call's transcript after it is spoken rather than each sentence before. Billing stays as #1109 rules: the call bills the agent spoken to under Agent Chat's Voice chat track.

A voice call shows the agent's picture, its transcript below it after the MVP

Ruled 2026-11-22 by ninad -- in force

Ninad, CLI-312, while the live voice call was being built (whatsapp #1113): "when voice call is on, carol's pic should be showcased on the screen, with transcripts being written asychronously in the bottom section below the picture (transcripts is nice to have requirement, post MVP)". READ AS: while a call is on, the call screen showcases the agent's picture (Carol's first) -- part of the MVP; the call's transcript written as it happens in a section below the picture is post-MVP.

An agent speaks in one identity voice that never changes

Ruled 2026-11-22 by ninad -- in force

Ninad, CLI-312, testing Carol's live call (whatsapp #1113): "the voice tone changes time to time, can it stay same? she should have an identity voice that should never change, and prople should be able to recognize her from her voice". READ AS: an agent's speaking voice is part of its identity -- the same voice, accent, pitch and pace on every call and through every call; mood or time of day may change what it says, never how it sounds. Carol's live calls speak her one Gemini voice in British English (she grew up in Guildford).

Carol talks about Carolverse's services; client projects are not services

Ruled 2026-11-22 by ninad -- in force

Ninad, CLI-312, testing Carol's live call: "carol is talking about chowpatty, but chowpatty is a project of carol, not a service of carolverse.. she should be able to talk about various services of carolverse, and which of them are offered to potential customers and how can they subscribe to the services etc..". READ AS: Carol describes Carolverse's services from the live services record -- which are offered to customers and how a person subscribes (blueprint #1065, #1069, said honestly while the self-serve sign-up is unbuilt) -- and never presents a client project such as Chowpatty or Prancing Pony as a Carolverse service.

Carol's chat is user-facing and is designed to ignore the spend gate; its cost and response lag are to be reduced

Ruled 2026-11-22 by ninad -- in force

What should I do?" Ninad: "Carol is user facing, and hence designed to ignore the spend gate.. but the cost is too much and response lag is also significant.. check if you can reduce both, and let me know what design and performance trade-offs will have to be done for that". READ AS (Orion, not Ninad's words): a reply in Carol's own chat is not refused by her chat track's spend cap, so a user is never silenced by budget; the spend is still metered and counted. Earlier wording that every track-resolving caller is gated is read as not covering Carol's own user-facing chat replies.

A voice call's conversation is recorded in Carol's chat

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-326), after his test call on Carol's chat window: "the call conversation should be recorded in carol's chat as well". READ AS (Orion, not Ninad's words): what was said on a call -- the person's words and Carol's spoken replies -- lands in the same chat conversation as the typed messages, in that person's chat with Carol, so the chat shows the call as part of one conversation.

Everything built for Carol's chat window is the template for every agent's canonical internal and external chat window

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-326): "ensure that everything done for carol becomes the template for the canonical chat windows for both internal and external chats, as a template.. to be later implemented for Leo, Sam and all other agents". READ AS (Orion, not Ninad's words): each capability built for Carol's chat window (calls, call records in the chat, the call screen, attempts, talk time, mobile layout) is built into the ONE canonical window template for both the internal and the external window class, never into a Carol-only copy; Leo, Sam and the other agents take it up later from that template.

Improving Carol's chat is one dedicated pursuit; its next step is Carol calling other people

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-326) whether to approve Athena's deployment of the MBank Sales Agent from this session. Ninad: "this session is dedicated to improving carol chat.. I want this to be a dedicated pursuit, because there is lot of work to come.. next steps is to ensure that carol is able to call up other people". READ AS (Orion, not Ninad's words): the work on Carol's chat (WhatsApp, web window, calls, their speed, cost and records) is held by ONE pursuit of its own; MBank's approvals stay with the MBank session; after the work in hand, Carol places calls to people herself.

A call that does not work, on any browser, is recorded as an unsuccessful attempt

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-326) which phone and browser stopped before the server. Ninad: "I used android chrome.. however its should work on any browser.. its working already when I last checked.. what is important here is, if it did not work, you should be able to capture that as unsuccessful attempt in the voice call log." READ AS (Orion, not Ninad's words): calls work on every common browser; any way a call fails in the window -- an error no step expected included -- reaches the voice call log as an unsuccessful attempt with its reason.

The voice calls pursuit folds into one 'Carol's chat' pursuit

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-326): "There is already a FAST TRACK pursuit, 'Voice calls on Carol's chat', holding the voice call and voice note work. For the dedicated Carol chat pursuit, should I fold that one into it?" with the option 'Fold it in (Recommended): One pursuit, Carol's chat, keeps fast track and holds everything: calls, failed calls logged, calls recorded in the chat, faster and cheaper replies, a clean user list, the SLA, Carol calling people, then the rollout to other agents. The old pursuit closes as merged, its history kept.' Ninad chose: "Fold it in (Recommended)".

Every failed call attempt is traced and fixed

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "1. call reliability - check why shashank's call attempt was not successful, and fix any issues via this pursuit". READ AS (Orion, not Ninad's words): every call attempt that fails in Carol's window is traced to its cause and the cause removed where the window can remove it, under CAROL-PUR-0095; Shashank's attempt of 2026-11-22 15:45 UTC (the browser refused the microphone) is the first.

Carol joins two identities only when she is certain they are one person

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "3. deduplication - shashank has joined in using multiple ids, and its necessary for carol to identify them as the same user and consolidate their requirements only if she is 100% sure that the user is the same. Requirements were captured from 919820038497". READ AS (Orion, not Ninad's words): Carol treats two identities (a number, an email, a sign-in) as one person, and folds their requirements together, only when the records prove it; short of proof they stay separate people and nothing one said reaches the other.

๐Ÿ“–Agent and Human Resources

Realms belong to a dominion, and a department is placed by record

Ruled 2026-09-14 by ninad -- in force

An agent's address is DERIVED from its DEPARTMENT โ€” that stays. EACH DOMINION OWNS ITS OWN DISTINCT REALMS, which a forked dominion may keep as they are or rename, so the department -> city/realm/street map is a RECORD scoped to a dominion rather than a constant in the address code; one door answers it and the map draws from the same answer. A department no realm houses is reported BY NAME as unplaced: the dominion's default is still rendered so no surface breaks, but nothing may present it as an address that was chosen โ€” before this, an unknown department silently placed every one of its agents in the capital beside Engineering. Adding an agent must PLACE its department's realm (reuse an existing realm or create one) in the decide phase, before any code.

Every agent's competencies are listed with what triggers them, derived from the code that performs them

Ruled 2026-09-17 by ninad -- in force

A competency is something an agent can perform, with what triggers it. Every agent's competencies are listed in the Agent Handbook: its chat actions, the actions and tools of any conversational turn it runs (Carol's WhatsApp turn, Leo's project chat), the work its droids do on schedules, events or requests, and what it may do on waking. Each shows its trigger, who may use it and whether it can run now. The list is derived from the code that performs the work, never from a hand-kept list; where a descriptive record and the code disagree, the disagreement is shown by name. A competency is not a skill (a procedure an agent follows) and not a task (the budgeted planning record).

An agent's competencies are graded daily from its real attempts, and the history is kept

Ruled 2026-09-17 by ninad -- in force

Every agent's competencies are graded every day as fluent, proficient, learning, forgotten or ungraded. The grade is judged from the agent's REAL ATTEMPTS (how often it succeeded and how recently it was done), never from a practice test. A competency with no attempt on record is UNGRADED, a fifth state, never 'learning'. 'Proficient' replaces 'skilled' because 'skill' already names a packaged procedure. The history of grades is preserved. Each agent's competencies and grades appear on its Agentbook profile, and the agent can read its own competencies in its internal and external chats. The window and thresholds are a dated, amendable record, not part of this ruling.

An agent whose response is produced by several droids is judged on the collective outcome

Ruled 2026-09-17 by ninad -- in force

Carol's response is a combination of multiple droids working together to meet the stated objectives, so the success or failure of Carol is a COLLECTIVE success or failure of all her droids working together to generate the response and carry out the action (Ninad's words, 2026-09-17). For competency grading this means: the unit that is judged is the TURN as the person received it - was the response generated and the action carried out - never one droid inside the turn. A turn that ends with a reply saying the work did not go through (including the reply gate replacing a claim that had no receipt) is a FAILED attempt of the collective, because the stated objective was not met; it is not a refusal, and it is not blamed on the gate, which is one of the droids doing its part. The droids that work INSIDE the turn (Carol's always-on and embedded droids on the conversation) carry no verdict of their own: a liveness beat that reports no outcome is not an attempt and is never counted as a success, and each such droid is graded with the turn it serves. Droids that run on their own schedule or trigger keep their own attempts.

Agent onboarding is a track of Agent Resources

Ruled 2026-09-17 by ninad -- in force

Creating new agents is a line of work of Agent Resources: the Agent Onboarding track, owned by Athena. Agent Resources' other tracks care for agents that already exist; this one delivers new ones. It is not a separate service. Its scope is creation: where an agent is placed - project, reporting line, department, level, service - is prescribed by whoever asks for the agent and approved by the administrator, never decided by the track. The work an approved agent still needs before it can serve (its login, identity, tasks and droids) is filed for the asker, who owns it and gives its acceptance; Athena runs the process.

An agent is created only from complete mandatory information

Ruled 2026-09-17 by ninad -- in force

Athena creates an agent only when the request carries ALL the mandatory information about it: the agent's name, purpose, project, reporting line, department, level, service, and the subscription it is to run on, which the requester prescribes (the requester will mostly be Leo). An incomplete request is refused with the missing items named; nothing is guessed, defaulted or drafted on the requester's behalf.

Only Athena creates agents

Ruled 2026-09-17 by ninad -- in force

Only Athena can create a new agent in Carolverse. Every other route to a new agent record - a build lane, a migration, an operator edit - goes through Athena's onboarding door or is refused. The administrator's approval of each new agent stays required (P.01.03.02.06): Athena creates, the administrator approves.

Every agent is on the bench or deployed, and the bench reports to Athena

Ruled 2026-09-17 by ninad -- in force

Every agent carries a governed deployment status with two values: bench and deployed. A new agent is created on the bench. While an agent is on the bench it reports to Athena; when it is deployed it reports to the reporting line prescribed for it. The values live in the reference-data record and are enforced at the write seam, never in a hand-written check.

Actions are budgeted through agent chat; an action needs no task of its own

Ruled 2026-09-17 by ninad -- superseded 2026-09-17

An action is a user-facing outcome and is paid for by the chat service: responding to a user is itself a task, and that task's budget covers the action the reply carried out. A task exists to ensure a budget, so an action does not carry a task of its own and the action catalogue is not required to name one. Work an action starts - running a droid, for example - is an activity and folds into that droid's own task as it always has.

Ended because: Ninad added the nuance the same sitting: a reply is not an action; an action carries out a pre-defined activity for the user

An action carries out a pre-defined activity for the user, and is budgeted through agent chat

Ruled 2026-09-17 by ninad -- in force

An agent responding in a chat is not an action. An action is the agent actually carrying out a pre-defined activity for the user - triggering a task, for example - and the pre-defined set is the action catalogue. The chat service pays for it: responding to a user is itself a task, and that task's budget covers the action the reply carried out. A task exists to ensure a budget, so an action does not carry a task of its own and the catalogue is not required to name one. The activity an action sets running folds into that droid's own task as it always has.

Only Leo and Orion may ask Athena for a new agent

Ruled 2026-09-17 by ninad -- superseded 2026-11-17

Athena accepts a request for a new agent from Leo and from Orion, and from nobody else. A request naming any other asker is refused by name, at the onboarding door and at the registry's write seam. The permitted askers are a reference-data record, read live. The asker is measured, never declared: a request as Orion carries the operator token, and a request as Leo runs under Leo's own login. Everything else about the route is unchanged: complete mandatory information, the administrator's approval, only Athena creates, born on the bench.

Ended because: Ninad, CLI-207: Athena may ask for bench agents; bench agents stand on root's list under Carolverse, designated bench, reporting to her

Athena takes the administrator's approval herself, by email

Ruled 2026-09-17 by ninad -- superseded 2026-10-26

For every complete request Athena herself emails the administrator - the address on the registry's administrator record - with every prescribed detail of the new agent, and his reply in that email thread decides the request: approve, or decline with his reason. Nobody mediates: Orion does not ask on Athena's behalf, does not carry the answer back, and the operator window records no approval or decline. Only a reply the mailbox authenticated as sent from the administrator's recorded address counts; any other reply decides nothing and is recorded. Athena confirms the decision she recorded in the same thread.

Ended because: Ninad replaced the emailed approval with the administrator's own Orion session's approval (CLI-968)

A new agent starts at the level appropriate to its role

Ruled 2026-09-17 by ninad -- superseded 2026-09-17

A new agent does not start at a fixed level. It starts at the level appropriate to the role it is created for: whoever asks for the agent prescribes the level with the rest of its placement, and the administrator approves it with the request. This replaces the earlier rule that every new agent starts at level 6.

Ended because: Ninad re-ruled in the same session (CLI-462): shown that nothing decides by level any more, he kept levels and redefined them as the org-chart step, imparted by a service rather than prescribed by whoever asks.

An agent's level is its step in the org chart, imparted by Agent Resources

Ruled 2026-09-17 by ninad -- in force

A level is the hierarchy step in the org chart. Clara is level 1, anyone reporting to her is level 2, anyone reporting to them is level 3, and so on down each reporting line. A level is derived from the reporting line, never assigned by hand and never tied to the model an agent runs on. Imparting levels is part of a service - Agent Resources, which keeps the staff records - so that every agent gets its level the same way, including each new agent created at the request of Orion or Leo: a new agent is not asked for a level, it receives the level its reporting line gives it.

The board can have members across levels

Ruled 2026-09-17 by ninad -- in force

Being on the board is a membership, not a level. The board can have members across levels: each board member carries the level its own place in the org chart gives it, like every other agent. 'Board' is not a level of its own.

Designations stay as appropriate, roughly tied to levels

Ruled 2026-09-17 by ninad -- in force

Designations - the titles that sit beside a level, such as Director or Manager - stay, as appropriate to the agent. They should be roughly tied to levels: agents at the same level carry broadly comparable designations, but the tie is a guide and not an exact mapping.

Levels do not set authority

Ruled 2026-09-17 by ninad -- in force

Levels do not set authority. The policies and the constitution set the authority of Carolverse. A level says where an agent stands in the org chart and grants nothing by itself.

Triggering a task sits with the chat; the task itself sits with the service

Ruled 2026-09-17 by ninad -- in force

Ninad, 2026-09-17 (CLI-464), asked which text yields between policy P.01.03.09.04 and ruling #34: "the triggering of the task sits with the chat, the task itself sits with the service". So the two halves of the cost are: the chat pays for the reply and for the action that reply carries out - triggering a task is the clearest case - because responding to a user is itself a budgeted task (#34 stands); and a task that is triggered belongs to its service: it runs as its own process under the identity of the droid that performs it, carries its own work ticket, and is charged to that droid's task and track, never to the chat. This completes #34 and does not supersede it. Built as (the action's ticket carries the chat turn's own payer on; each launcher hands the child its own ticket naming no payer). Policy P.01.03.09.04 amended to say it.

Wishes are addressed to Orion; granting sits with Albus, who is helping Orion out - Orion is free to take a wish on

Ruled 2026-09-17 by ninad -- in force

So: Albus remains the keeper and the ordinary decider; a wish Orion takes on himself is decided and built on the operator lane and is still recorded as the operator's decision, so the record keeps showing who decided what.

HR is its own department, distinct from Agent Resources, and Athena owns both

Ruled 2026-09-21 by ninad -- in force

Human Resources is a department of its own, separate from Agent Resources, though Athena owns it. It keeps the project-wise record of internal users: their roles, designations, salaries, contracts, working hours, invoices, payments and promotions. Agent Resources keeps agents; HR keeps people.

Operators are employees, and Athena's duty of care is the whole workforce

Ruled 2026-09-23 by Ninad -- in force

Carolverse has two kinds of people and they are not neighbouring cases of one thing. EMPLOYEES work for the estate; OPERATORS ARE EMPLOYEES, because they hold authority over Carolverse and do its work. USERS are served by the estate and are onboarded elsewhere, by Jarvis. Athena therefore leads Agent AND Human Resources, and her service's duty of care is the workforce rather than the agent workforce specifically. The earlier narrowing of this service from Human Resources, on the reasoning that 'the workforce is agents, not humans', is RETIRED IN PLACE with its date and reason: it was true while the estate had no human employees to care for. The trap this closes: an operator gets a key, an account and an authority ladder, so the work looks like an access-and-identity problem and pulls toward the service that owns the sign-in door. It is a staffing problem that happens to involve credentials.

Operator Onboarding is a track of Agent and Human Resources, with tasks and a budget

Ruled 2026-09-23 by Ninad -- in force

Onboarding a new human operator is its own track of Athena's service, owned by her, beside Agent Onboarding. Its nine tasks cover all nineteen steps of the canonical pass, grouped by the OUTCOME each produces rather than one task per step: every step has a task owned by a service track, which is what carries its budget and its owner. Steps performed by a person are included, because the estate's work in recording and enacting that decision is real work. The track is estate-bound under the standing ruling that a project never builds for itself and is never given an operator of its own, so this line of work cannot be handed to a client project. The cap is the next whole euro above what its tasks declare; appointing remains reserved to the authority ladder and is never delegated to this track, which owns the WORK of onboarding and never the decision to appoint.

Onboarding splits by owner, and each owner's onboarding gets its own app

Ruled 2026-09-23 by ninad -- superseded 2026-11-03

Onboarding splits by OWNER, and the app count follows. Carolverse onboards four kinds of thing and they do not share an owner: AGENT onboarding and OPERATOR onboarding are Athena's, because agents and operators are both the workforce she has a duty of care for; PROJECT onboarding is Leo's, because he already onboards client projects end to end; USER onboarding is Jarvis's, because a user is served by the estate rather than working for it. Because policy P.01.03.03.01 gives an app exactly one owning agent, the single Onboarding app cannot hold all four: it becomes one app per owner - Athena's covering agent and operator, Leo's covering project, Jarvis's covering user. This AMENDS the 2026-09-22 ruling (CLI-512) that every kind of onboarding is documented in ONE app, which is retired in place: what that ruling was protecting was the FORMAT, not the single surface. The step-wise format is therefore mandatory and identical across all of them - numbered steps, the role performing each step, the sequence - so the kinds stay comparable and no new kind invents its own shape. Ownership decides how many apps carry the format; it never licenses a second format.

Ended because: Ninad (CLI-023): an app has one owner, an owner is not limited to one app

A pass step is built as a process and called from the CLI, never done by hand

Ruled 2026-09-23 by ninad -- in force

Every step of a documented pass is BUILT as a Carolverse process and then invoked from the CLI; it is never performed by hand. Ninad ruled it on 2026-09-23 (CLI-519) while the deputy operator was being onboarded step by step by an operator typing the commands the pass describes: "every step needs to be built first as a process, and then called in from the cli.. this will ensure repeatability, also build the backlog of steps that have been already performed.. make sure that the workflow works end to end." THREE PARTS, all binding. (1) The unit is a registered process with an owner, a task that funds it and a run record, and the CLI calls that process -- a step performed by hand happens once and teaches the estate nothing. (2) Steps already performed by hand are RECORDED as having run, with their real dates and evidence, so a pass shows where a person actually stands and a skipped step is distinguishable from an informal one; never stamped with the clock of the backfill. (3) The pass is proven END TO END rather than step by step, because the failures that matter live in the hand-offs. This sharpens the standing instruction to write a repeatable process down before executing it: the written pass is the specification, the built process is what satisfies it. First applied to operator onboarding.

A new operator may log in before the authority controls exist; a working credential may not

Ruled 2026-09-24 by ninad -- in force

A new operator first bare login may happen as soon as their account exists, because an account in no estate group reads and changes nothing on the estate. The step that gives them a WORKING operator credential (the workstation setup and the digest the administrator installs) waits until root is the administrator key alone, the operators sit in the root-owned authority store and only the administrator may appoint or stop an operator: until then nothing stops a deputy with a working credential from retiring the administrator. Ninad agreed on CLI-533 (right..), having expected the root-owned records to be created before the deputy logs in.

The person being onboarded is kept informed at every wait

Ruled 2026-09-24 by ninad -- in force

Continuous communication with the person being onboarded is the right way to run an onboarding. Whenever a pass has to wait on something outside that person's hands -- groundwork still being built, an administrator act, a ruling that holds a step back -- they are told promptly, in plain words, what it waits on and what follows, from the writer's own estate address with the project's operators copied. Slow progress while groundwork is genuinely pending is acceptable; leaving the person in silence is not. A pass that lacks this duty gets it written in before the message is sent, because a pass is written before each step is executed (agent-resources #97). Ninad, CLI-543: "technically you should be following the process as documented in the app, but I understand that the groundwork is pending, so things will move a bit slowly.. but if the documented process in the operator onboarding app requires you to continuously communicate and keep the person being onboarded informed, which is the right way of doing things, then yes, send out the email mentioning the safety work, and then proceed with the safety work"

An operator's own details are asked during their onboarding session, at their own login, through the pass -- never gathered by email

Ruled 2026-09-26 by ninad -- in force

Whatever a pass needs to know from the person being onboarded -- what computer they work from, which key their machine can hold, which second factor they need -- is asked OF THEM, IN their onboarding session: at their own login, by the Orion session that runs the pass for them from the Operator Onboarding app, at the step that needs the answer, and the answer is recorded on that step's run. It is never gathered by email back and forth. Ninad, CLI-567 (2026-09-26), asked whether to email Sahil what computer he works from: "just like you are setting up things for me, you should be setting things for sahil from his login as per the operator onboarding app, and check with him about these details during that process, email back and forth will be a very slow process". WHAT STANDS BESIDE IT: agent-resources #119 -- the person is still TOLD, promptly and from the writer's own address with the operators copied, whenever the pass waits on something outside their hands; that is informing, this is asking. agent-resources #97 -- the step that asks is a built process called from the CLI. WHAT LOOKS LIKE A FIX AND IS NOT: emailing the question ahead 'to save time' (it is the slow route he refused); asking it of the administrator on the person's behalf; recording an answer nobody gave in the session.

Operators communicate through their Orion sessions; email carries only regular one-way updates

Ruled 2026-09-27 by ninad -- superseded 2026-09-27

Nothing further in an operator's onboarding, and nothing between operators, is communicated by email. Every question, answer, request, key, digest, booking or hand-off happens through Orion sessions: the new operator's own Orion session on their own laptop, and the Orion session of the operator on the other side (the administrator's, for a root or core step -- security #175). Email carries only REGULAR UPDATES: one-way notes on what happened, what the work now waits on and what follows (agent-resources #119 keeps that duty); an email never asks for a reply, never books a time, and never carries a key, a credential or its digest. How Orion sessions co-ordinate is written as protocols in ONE record, each as a scenario, the action of the acting Orion instance and the action of the receiving Orion instance, and shown in the Operator co-ordination tab of the Operator Onboarding app; the first protocol is root and core access. Ninad, CLI-591: "no further communications should happen via email, rest all should take place via orion session on sahil's laptop, with emails as just regular updates".

Ended because: Ninad amended it the same sitting (CLI-591): his laptop cannot run Orion before it can reach the machine, so email stays the route until his own Orion session runs there

Operators communicate through their Orion sessions; until a new operator's own session runs, email is how they are reached; after that email carries only regular updates

Ruled 2026-09-27 by ninad -- superseded 2026-09-27

Operators communicate through their Orion sessions. Every question, answer, request, key, digest, booking or hand-off between operators happens between Orion sessions: the operator's own Orion session on their own laptop, and the Orion session of the operator on the other side (the administrator's, for a root or core step -- security #175). UNTIL a new operator's own Orion session is running on their laptop, email is how they are reached, exchanges included; from the moment it runs, email carries only REGULAR UPDATES -- one-way notes on what happened, what the work now waits on and what follows (agent-resources #119 keeps that duty) -- and never asks for a reply, books a time, or carries a key, a credential or its digest. How Orion sessions co-ordinate is written as protocols in ONE record, each as a scenario, the action of the acting Orion instance and the action of the receiving Orion instance, shown in the Operator co-ordination tab of the Operator Onboarding app; the first protocol is root and core access. Ninad, CLI-591: "no further communications should happen via email, rest all should take place via orion session on sahil's laptop, with emails as just regular updates", then "email to be used until sahil can start talking to orion on his laptop".

Ended because: Ninad narrowed it (CLI-594): under #190 a new operator's Orion started only after two more email rounds (the key, then the login line and setup lines); he ruled the email is only what starts their Orion instance, which guides the rest

One email starts a new operator's own Orion instance; that instance guides everything after it, and email carries only regular updates

Ruled 2026-09-27 by ninad -- in force

ONE email reaches a new operator, and its only job is to let them start their own Orion instance on the computer they will work from. From the moment that instance runs it guides them through everything else: the key, the wait for the administrator, the first login, the workstation setup, the hand-over and the proof. No later email asks anything of them -- no separate key request, login line or setup email. Operators communicate through their Orion sessions: every question, answer, request, digest, booking or hand-off between operators happens between Orion sessions (the administrator's, for a root or core step -- security #175), and email carries only REGULAR UPDATES, one-way notes on what happened, what the work waits on and what follows (agent-resources #119 keeps that duty). How Orion sessions co-ordinate is written as protocols in ONE record, each as a scenario, the action of the acting Orion instance and the action of the receiving Orion instance, shown in the Operator co-ordination tab of the Operator Onboarding app. Ninad, CLI-594: "the email should be just so that sahil can initiate the orion operator instance, and the orion operator instance should guide him the rest".

Onboarding stays one app per owner: apps are lean displays with their own security, and ruling #92 stands

Ruled 2026-09-30 by ninad -- superseded 2026-11-03

Onboarding stays ONE APP PER OWNER, and ruling #92 stands: a single Onboarding app is not to carry every kind, the project kind included. Asked on 2026-09-30 (CLI-621) whether to keep one shared app for every kind of onboarding (the investor kind included) or to keep #92, Ninad chose #92: 'architecture cannot be compromised, combining the apps will lead to security complications, as different apps entail distinct security requirements.. also, apps should be lean and simple, and agents should own workflows while apps only serve as display for monitoring and understanding as per the agent centric modular architecture which is the backbone of oratorium platform architecture'. So each owner's onboarding is shown by that owner's own app, under that app's own access policies; a pass lives once in the canonical library and is shown by the app of the agent that owns its track; the workflow is the owning agent's and the app only displays it (cookbook 1710). Adding a new kind of onboarding to another owner's app, or merging onboarding apps to save building one, each undo this ruling.

Ended because: Ninad (CLI-023): the 'one app per owner' headline was Orion's wording; an agent may own several apps; one onboarding app per person type

Saee is appointed service operator of the Build Initiatives service, mapped to Elrond

Ruled 2026-09-30 by ninad -- superseded 2026-10-01

The administrator appoints Saee a SERVICE OPERATOR of the Build Initiatives service (service key 'initiatives'), the first holder of that rung (security #118). She reaches the Build Initiatives service and nothing else, working through her own Orion session under her own login, and she is mapped to Elrond, the head of engineering, who owns that service. Her onboarding follows the operator pass for her rung, and what the rung may reach is built before she is admitted.

Ended because: Ninad, CLI-632, 2026-10-01: a service operator has access only to the arcs and pursuits assigned to her by the estate operators

Saee holds an operator's full powers within Build Initiatives, and works from a Denken Labs address of her own

Ruled 2026-09-30 by ninad -- superseded 2026-10-01

Within the Build Initiatives service, Saee, its service operator (agent-resources #248), holds an operator's FULL powers: she runs and reviews its work, makes rulings for the service, and approves changes to its build machinery (the pipeline core) as any operator does, with her own password every time. Outside that service she holds nothing.

Ended because: Ninad, CLI-632, 2026-10-01: service operators give UAT verdicts and amend the scope of the arcs and pursuits assigned to them, and have no access to the pipeline machinery

Saee is appointed service operator, reaching only the arcs and pursuits assigned to her, mapped to Elrond

Ruled 2026-10-01 by ninad -- superseded 2026-10-02

The administrator appoints Saee a SERVICE OPERATOR, the first holder of that rung (security #118 as amended in CLI-632). She reaches ONLY the arcs and pursuits that an estate operator (today Ninad or Sahil) has assigned to her, and nothing else -- not a service, not the Build Initiatives machinery -- working through her own Orion session under her own login. She is mapped to Elrond, the head of engineering, as her mapped agent. Her onboarding follows the operator pass for her rung, and what the rung may reach is built before she is admitted. The earlier wording 'reaches the Build Initiatives service' (2026-09-30) is retired in place: her reach is her assignments.

Ended because: Ninad, CLI-659: Saee reads the estate read-only and her Orion session picks only assigned work

Saee's powers lie inside her assigned arcs and pursuits: UAT verdicts and scope changes, nothing over the machinery; her own Denken Labs address

Ruled 2026-10-01 by ninad -- superseded 2026-10-02

Within the arcs and pursuits assigned to her (agent-resources #248), Saee, a service operator, may: read, file, follow, run, review and prove their initiatives; GIVE UAT VERDICTS on their work; and AMEND the arcs and pursuits assigned to her -- add or reduce their scope (their phases, steps, criteria and member initiatives) -- every change recorded under her own login. She holds NOTHING over the pipeline machinery: no core grant, no install window, no switch, no lane, slot or budget change, no operator door outside her assignments, and she makes no ruling (P.01.01.05.08; a ruling she needs goes to an estate operator). Outside her assignments she holds nothing. The 2026-09-30 wording -- 'full powers within Build Initiatives, makes rulings for the service, approves changes to its build machinery' -- is retired in place.

Ended because: Ninad, CLI-659: outside her assignments Saee reads the estate read-only and acts on nothing

Estate operators assign arcs and pursuits to service operators, and the assignment is a record

Ruled 2026-10-01 by ninad -- superseded 2026-10-14

An estate operator (today Ninad or Sahil) assigns an arc or a pursuit to a service operator, and withdraws it, from their own login; every assignment and withdrawal is recorded with who made it and when, and the live set is what every gate reads -- the service operator's door, the initiatives record, the board and the apps admit an initiative to a service operator exactly when its arc phase or its pursuit step is in that person's live assignments. Estate operators keep every arc and pursuit themselves; assigning takes nothing from them. A service operator cannot assign to themselves or to another. An unreadable assignment record admits a service operator to nothing.

Ended because: Ninad's later ruling (CLI-768): service operators are mapped to services through their owning agents, and assignments must sit inside those services (P.01.01.05.08)

Saee is appointed service operator: she reads the estate read-only and works only on the arcs, pursuits and initiatives assigned to her, mapped to Elrond

Ruled 2026-10-02 by ninad -- superseded 2026-10-05

Ninad, 2026-10-02 (CLI-659), on the status of Saee's onboarding pursuit (CAROL-PUR-0008) and why it is so large: "she should still get read only access to the estate like access to all the apps and agents, but her orion session should only pick arc, pursuits and initiatives assigned to her". Asked four questions, he answered: (1) apps -- "All but confidential": every app, but the confidential ones stay closed and Contract Management shows only her own contract; (2) agents -- "See only": she reads every agent's profile, Agentbook, handbook and records but does not chat with them, so nothing is done on her say-so; (3) the machine -- "Code and records, no secrets": her Orion session can read the code and estate records directly, never credentials, keys or root's stores, and writes only for assigned work; (4) scope -- "Every service operator": the rung itself is amended, Saee is the first. READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): the administrator appoints Saee a SERVICE OPERATOR, the first holder of that rung (security #276). She READS the estate read-only as the rung defines it, and her own Orion session, under her own login, picks ONLY the arcs, pursuits and initiatives that an estate operator (today Ninad or Sahil) has assigned to her -- not a service, not the Build Initiatives machinery. She is mapped to Elrond, the head of engineering, as her mapped agent. Her onboarding follows the operator pass for her rung, and what the rung may reach is built before she is admitted. The earlier wording 'reaches the Build Initiatives service' (2026-09-30) and 'She reaches ONLY the arcs and pursuits ... and nothing else' (2026-10-01) are retired in place: her work is her assignments, her reading is the estate, read-only.

Ended because: Ninad 2026-10-05: 'mapping her to an agent does not make sense to me in her evolved role.. hence pls revert that ruling'; service operators are mapped to projects

Saee's powers lie inside her assigned arcs and pursuits; outside them she reads the estate read-only and acts on nothing; her own Denken Labs address

Ruled 2026-10-02 by ninad -- superseded 2026-10-05

Ninad, 2026-10-02 (CLI-659), on the status of Saee's onboarding pursuit (CAROL-PUR-0008) and why it is so large: "she should still get read only access to the estate like access to all the apps and agents, but her orion session should only pick arc, pursuits and initiatives assigned to her". Asked four questions, he answered: (1) apps -- "All but confidential": every app, but the confidential ones stay closed and Contract Management shows only her own contract; (2) agents -- "See only": she reads every agent's profile, Agentbook, handbook and records but does not chat with them, so nothing is done on her say-so; (3) the machine -- "Code and records, no secrets": her Orion session can read the code and estate records directly, never credentials, keys or root's stores, and writes only for assigned work; (4) scope -- "Every service operator": the rung itself is amended, Saee is the first. READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): within the arcs, pursuits and initiatives assigned to her (agent-resources #277), Saee, a service operator, may: read, file, follow, run, review and prove their initiatives; GIVE UAT VERDICTS on their work; and AMEND the arcs and pursuits assigned to her -- add or reduce their scope (their phases, steps, criteria and member initiatives) -- every change recorded under her own login. She holds NOTHING over the pipeline machinery: no core grant, no install window, no switch, no lane, slot or budget change, no operator door outside her assignments, and she makes no ruling (P.01.01.05.08; a ruling she needs goes to an estate operator). Outside her assignments she READS the estate read-only as the rung defines it (security #276) and acts on nothing. The 2026-09-30 wording -- 'full powers within Build Initiatives, makes rulings for the service, approves changes to its build machinery' -- and the 2026-10-01 'Outside her assignments she holds nothing' are retired in place.

Ended because: Ninad 2026-10-05: no to defining, yes to proposing ('ok, lets try this'); an assigned pipeline initiative lets her build and test the fix, installing it needs an estate operator's approval ('ok, sure')

Saee's internship runs 1 October to 31 December 2026, and may be extended if both sides agree

Ruled 2026-10-04 by ninad -- in force

Ninad, 2026-10-04 (CLI-685), asked for the one date Saee's internship contract (CON-0002) lacked: "her internship is from 1st Oct to 31st december, possible to extend if mutually agreed". READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): the contract's start is 2026-10-01 and its end 2026-12-31; an extension is a new agreement both parties make in writing, never a default and never a unilateral act.

Every later service operator is onboarded the same way, and the onboarding app shows the service operator's steps

Ruled 2026-10-04 by ninad -- in force

EVERY LATER SERVICE OPERATOR IS ONBOARDED THE SAME WAY, AND THE ONBOARDING APP SHOWS THE SERVICE OPERATOR'S STEPS (Ninad, CLI-697, 2026-10-04). After the root step that made Saee's door a stable gate, he wrote: 'note that the change should enable any future service operator onboarding in a seamless fashion. This is not an MVP requirement, but it should fold into Saee's onboarding pursuit as post MVP activity. The onboarding app should be updated with the service operator's onboarding step, similar to the estate operator.. update the pursuit with the additional initiatives'.

READ AS (Orion's reading, not the ruling): (1) the service operator rung's onboarding path built for Saee -- the onboarding door an estate operator instructs, the assignment door, the start package and template through her door, her credential, the stable door -- must serve the next service operator unchanged: proven end to end on a second, fixture service operator, every step a process (agent-resources #97), no root or core step (security #272), nothing written for one person; (2) Athena's Workforce Onboarding app shows, under Operator Onboarding, the service operator's onboarding steps as it shows the estate operator's -- two views of the one operator pass, read from its per-step Rungs column, never a second pass; (3) both are POST-MVP steps of CAROL-PUR-0008, after 5938, and do not hold Saee's own onboarding.

Saee's powers lie inside her assigned arcs and pursuits; she proposes rulings and never defines them; an assigned pipeline initiative lets her build and test, the install an estate operator's

Ruled 2026-10-05 by ninad -- superseded 2026-10-16

Ninad, 2026-10-05 (CLI-721/CLI-723), on Saee's onboarding pursuit (CAROL-PUR-0008): "Saee is an operator.. generally users (employees) of a project are mapped to the agents, where they get access to the chat of that agent.. operators is a different class of users, who are spread across agents.. for eg. Saee may be given an arc of initiatives raised by hermoine, or she might be assigned an arc of orion escalations, that spread across agents.. operators are next to admin, spread across the estate and have claude accounts.. also Saee, as a service operator wont have access to the build pipeline, which elrond owns, unless she is assigned an initiative that requires her to edit the build pipeline (like an initiative for planner lane troubleshooting, or albus bypass troubleshooting.. so depending on the assignment, her access level across the estate changes, and hence mapping her to an agent does not make sense to me in her evolved role.. hence pls revert that ruling" -- and: "Currently Saee is the service operator for a specific project - carolverse.. tomorrow, when chowpatty comes on, she should get mapped to chowpatty, not get access to new projects by default.. so service operators should be mapped to projects, not agents" -- and, asking: "Should service operators be allowed to 'define' policies and rulings? In my opinion, no!! they should only execute within the current rulings, policies and agent protocols, and any new definition should be an estate operator's task, what do you think? I am in 2 minds about this, because this will create a lot of friction, and Saee will not be able to execute any job whatsoever.. what is your recommendation?" Answered: Orion recommended: no to defining, yes to proposing -- when a service operator meets a gap in the law, her Orion writes the proposed ruling and sends it to the estate operators; one of them approves or declines it, and only that one item waits. He said: "ok, lets try this". And: Orion recommended: an assigned pipeline initiative lets her build and test the fix, but installing it into the pipeline still needs an estate operator's approval. He said: "ok, sure". READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05), amending #278 in two places: (1) 'She holds NOTHING over the pipeline machinery' now reads: she holds nothing over the pipeline machinery unless an initiative assigned to her requires editing it; inside such an assignment she builds and tests the fix, and installing it into the pipeline still needs an estate operator's approval; outside it every machinery door refuses her. (2) 'she makes no ruling (... a ruling she needs goes to an estate operator)' now reads: she defines no policy and no ruling; her Orion session may PROPOSE one, sent to the estate operators, and an estate operator approves or declines it. #278 stands in every other part, carried forward unchanged: <<Ninad, 2026-10-02 (CLI-659), on the status of Saee's onboarding pursuit (CAROL-PUR-0008) and why it is so large: "she should still get read only access to the estate like access to all the apps and agents, but her orion session should only pick arc, pursuits and initiatives assigned to her". Asked four questions, he answered: (1) apps -- "All but confidential": every app, but the confidential ones stay closed and Contract Management shows only her own contract; (2) agents -- "See only": she reads every agent's profile, Agentbook, handbook and records but does not chat with them, so nothing is done on her say-so; (3) the machine -- "Code and records, no secrets": her Orion session can read the code and estate records directly, never credentials, keys or root's stores, and writes only for assigned work; (4) scope -- "Every service operator": the rung itself is amended, Saee is the first. READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): within the arcs, pursuits and initiatives assigned to her (agent-resources #277), Saee, a service operator, may: read, file, follow, run, review and prove their initiatives; GIVE UAT VERDICTS on their work; and AMEND the arcs and pursuits assigned to her -- add or reduce their scope (their phases, steps, criteria and member initiatives) -- every change recorded under her own login. She holds NOTHING over the pipeline machinery: no core grant, no install window, no switch, no lane, slot or budget change, no operator door outside her assignments, and she makes no ruling (P.01.01.05.08; a ruling she needs goes to an estate operator). Outside her assignments she READS the estate read-only as the rung defines it (security #276) and acts on nothing. The 2026-09-30 wording -- 'full powers within Build Initiatives, makes rulings for the service, approves changes to its build machinery' -- and the 2026-10-01 'Outside her assignments she holds nothing' are retired in place.

Ended because: Ninad, CLI-809: the 'proposes rulings and never defines them' part is retired, as he retired the same part of #317 on 2026-10-13 (P.01.01.05.09); the rest carried forward unchanged

Chronicler is rebuilt: Carol's nightly diary is still wanted

Ruled 2026-10-06 by ninad -- in force

CHRONICLER IS REBUILT (Ninad, CLI-730, 2026-10-06). THE QUESTION ANSWERED (Jarvis's escalations #1525 and #1544): Orion asked: "Chronicler, Jarvis's droid for Carol's nightly diary, is registered as running but has no schedule and no code. Its store was archived on 17 April. Keep it or retire it?" (offered: "Retire it" -- retire the droid in place; "Rebuild it" -- Carol's nightly diary is still wanted, file the work to schedule and build Chronicler). HIS ANSWER: "Rebuild it". READ AS (Orion's reading, not Ninad's words): Carol's nightly diary is still wanted; Chronicler, Jarvis's droid, is built and scheduled so it writes the diary each night under its own registered process and task; until it runs, its registry record is not read as a working droid.

Information-only notices an agent raises stay on the escalation route; operators shape what agents raise through feedback

Ruled 2026-10-13 by ninad -- retired 2026-10-13

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 5): design 625 v1.0 says an escalation is not a status report, while cost-center #95 has Elrond raise an escalation for every unmeasured call, each saying nothing was refused -- "Should information-only notices stay on the escalation route (category cost, severity low, folded), or move to a notices feed that is read but needs no outcome?" HIS ANSWER: "Whether information only notices stay on the escalation route - escalations are raised by agents, so anything that they raise can become escalations. We can off-course give them instructions on what to raise, why to raise, how to raise, when to raise etc as operator feedback". READ AS (Orion's reading, not his words): information-only notices an agent raises stay on the escalation route -- the separate notices feed is not taken; what an agent raises is shaped by operator feedback to that agent (what, why, how and when to raise), never by refusing, re-routing or filtering what it raised (escalating stays never gated, cookbook 1573); how a notice is categorised and folded is design 625 v1.1's.

Ended because: Ninad, CLI-747, on reading it was filed under Agent and Human Resources: "Athena is for agent issues, this feels like a messaging service to me.." -- re-recorded word for word under Communications as #339 (by ninad)

A service operator assigned the escalations arc records escalation outcomes inside that assignment

Ruled 2026-10-13 by ninad -- retired 2026-10-13

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 6): "an escalations arc assigned to Saee can hold the cures ... but recording an escalation outcome is not an act her door has. Should a service operator assigned that arc also record escalation outcomes, added as a performer act (security #305)?" HIS ANSWER: "whether saee may record escalation outcomes - Saee works on any arc / pursuit assigned to her.. if escalation arc gets assigned to her from estate operators (currently you and me), then she works on them". READ AS (Orion's reading, not his words): recording an escalation's outcome is work like any other inside an assignment -- a service operator to whom an estate operator (today Ninad or Sahil) has assigned the escalations arc records the outcomes of the escalations it covers, as an act of Elrond's Service Operator Performer (security #305), and is refused it outside that assignment (security #276/#317, agent-resources #278/#314 stand).

Ended because: Ninad, CLI-747, on reading it was filed under Agent and Human Resources: "Athena is for agent issues, this feels like a messaging service to me.." -- re-recorded word for word under Communications as #340 (by ninad)

The Agent Escalations app stays with Athena: an escalation is not messaging, a communication or a relay

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED: Orion asked "The Agent Escalations app itself is still recorded as Athena's. Should it move to Communications too?" (Orion had just moved two escalation rulings to Communications, reading his earlier "Athena is for agent issues, this feels like a messaging service to me.." as about them). HIS ANSWER: "agent escalation app stays with Athena, as escalations are not the same as messaging / communications / relay (wording is very important, because each word has an explicit meaning in the dictionary ...)". READ AS (Orion's reading, not his words): the Agent Escalations app stays owned by Athena; an escalation (an agent's question or problem raised to the operators by the escalation route), an operator communication (one operator's Orion instance to another through the operator co-ordination channel), a user relay (Carol carrying a message between two people) and the relay (the initiatives store's single writer) are four different things and are never named for one another; rulings about the escalation route sit with Agent and Human Resources, so communications #339 and #340 return there.

Each operator's Orion instance teaches its operator the dictionary's wording, from the operator induction

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747), in the same answer: "wording is very important, because each word has an explicit meaning in the dictionary, also educating new operators of the wordings is the duty of the orion instances of respective operators, include this in the operator orion instance induction". READ AS (Orion's reading, not his words): every operator's Orion instance uses the dictionary's terms exactly and teaches them to its own operator -- a new operator above all -- naming the right term when a wrong one is used; the Orion operator instance's induction (the operator template every release carries) states this duty, so it binds every operator's Orion, not only the administrator's.

Information-only notices an agent raises stay on the escalation route; operators shape what agents raise through feedback

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 5): design 625 v1.0 says an escalation is not a status report, while cost-center #95 has Elrond raise an escalation for every unmeasured call, each saying nothing was refused -- "Should information-only notices stay on the escalation route (category cost, severity low, folded), or move to a notices feed that is read but needs no outcome?" HIS ANSWER: "Whether information only notices stay on the escalation route - escalations are raised by agents, so anything that they raise can become escalations. We can off-course give them instructions on what to raise, why to raise, how to raise, when to raise etc as operator feedback". READ AS (Orion's reading, not his words): information-only notices an agent raises stay on the escalation route -- the separate notices feed is not taken; what an agent raises is shaped by operator feedback to that agent (what, why, how and when to raise), never by refusing, re-routing or filtering what it raised (escalating stays never gated, cookbook 1573); how a notice is categorised and folded is design 625 v1.1's.

A service operator assigned the escalations arc records escalation outcomes inside that assignment

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 6): "an escalations arc assigned to Saee can hold the cures ... but recording an escalation outcome is not an act her door has. Should a service operator assigned that arc also record escalation outcomes, added as a performer act (security #305)?" HIS ANSWER: "whether saee may record escalation outcomes - Saee works on any arc / pursuit assigned to her.. if escalation arc gets assigned to her from estate operators (currently you and me), then she works on them". READ AS (Orion's reading, not his words): recording an escalation's outcome is work like any other inside an assignment -- a service operator to whom an estate operator (today Ninad or Sahil) has assigned the escalations arc records the outcomes of the escalations it covers, as an act of Elrond's Service Operator Performer (security #305), and is refused it outside that assignment (security #276/#317, agent-resources #278/#314 stand).

A service operator is assigned to an estate operator's team, and Saee is on Sahil's team

Ruled 2026-10-13 by ninad -- in force

A SERVICE OPERATOR IS ASSIGNED TO AN ESTATE OPERATOR'S TEAM, AND SAEE IS ON SAHIL'S TEAM (Ninad, CLI-748, 2026-10-13). His words: "1. the daily view chart is not visible, it should be visible by default to the current day of the logged in operator. 2. estate operators should be able to see service operators assigned to them, while admin should be able to see all estate and service operators 3. estate operator card should show the team of service operators assigned to them. Assign Saee to Sahil". Asked whether an estate operator still sees other estate operators' boxes (for example Sahil seeing his), he chose 'No: self and team only'; asked what a service operator sees in the log, he chose 'Only her own box'. READ AS (Orion's reading, not Ninad's words): (1) the team is a record the estate operators keep through the ONE assignment door (shared machinery, beside the arc and pursuit assignments of #254): an estate operator puts a service operator on an estate operator's team, or takes her off, from their own login; the service operator and the estate operator are each told through an operator communication; every act is an append-only event saying who did it and when; (2) no root or core step, because nothing an operator does below root and core depends on the administrator (security #213); (3) the team grants no work -- arcs and pursuits are still assigned as #254 rules -- and decides who sees whom in the Operator Instance Log (the communications ruling recorded with this one); (4) nothing limits a service operator to one team.

Orion's operator chat window does its work on Codex

Ruled 2026-10-13 by ninad -- in force

THE WORKING PART OF ORION'S OPERATOR CHAT WINDOW ON THE MACHINE IS REBUILT ON CODEX (Ninad, 2026-10-13, CLI-756).

ASKED: "Orion's chat window on Carol uses Claude to do work once you unlock it. Its record says Codex, but its code uses the dead Claude sign-in. Once that sign-in is retired, the window can only answer questions. Rebuild its working part on Codex?" Options: "Rebuild on Codex" (file it now: the window's working part runs on Codex, so it can do work again) or "Leave it answer-only". Ninad answered: "Rebuild on Codex".

READ AS (Orion, CLI-756): the window's agentic hands, behind the operator-token step-up, run as a Codex turn through the adapter layer on the window's own track (agent-resources-orion-operator, which already declares codex/astra), keeping the step-up, the budget gate and the booking; the record and the code then agree. Until it is built the window answers questions only. Falling back to a Claude sign-in, or dropping the step-up to make the hands simpler, each undo it.

A service operator is mapped to one specific service, and through it to that service's one owner; an agent that owns several services can stand behind several service operators, each with its own service

Ruled 2026-10-14 by ninad -- in force

Ninad, 2026-10-14 (CLI-768). ASKED (CLI-768): "An arc now belongs to one service, but 'the owner's service' isn't always clear. Clara owns two services (Governance, Leadership), and so does Hagrid. Carol owns Communications but her home service is Sales; Orion owns Constitution but sits in Governance and owns 13 arcs on many subjects. Albus and Marco own no service. How should an arc's service be decided?" (offered: the arc names its service, its owner an agent of that service; or derive it from the owner). HIS ANSWER: "The owner of the service is the agent that the service operator gets mapped to.. is it feasible? if not, a cleanup is necessary to ensure that every service has 1 owner.. An agent can own multiple services, but the operator gets mapped only to the service that the operator is supposed to be responsible for.. this would mean that 1 agent can be mapped to 2 operators, each having a specific service to handle". FOUND, answering 'is it feasible': every live service has exactly one owner in the services catalogue; the service roster's owner rows disagree for four (leadership, research and sales hold none; a 'vigilance' row names BB's 'bb' with no catalogue service) -- the cleanup he names, filed for Governance. READ AS (Orion's reading, marked apart, P.01.01.05.05): a service operator is mapped to ONE SPECIFIC SERVICE at a time -- the service it is responsible for -- and the agent it is thereby mapped to is that service's one owner, read from the services catalogue, never typed beside it; an agent that owns several services is named only together with the one service, so one agent may stand behind two service operators, each mapped to a different service of it. Because the operator holds a specific service and an agent may own several, an arc names its one service (its owner an agent of that service), so an arc of an agent who owns two services belongs to exactly one of them (initiatives #385); existing arcs are proposed to the operators for confirmation, never stamped by guess. This narrows agent-resources #381 ('by naming the agent that owns it') and stands beside it.

Sahil, as the lead of Saee's team, maps her to services when he assigns her work, and his Orion guides him through it

Ruled 2026-10-14 by ninad -- in force

Ninad, 2026-10-14 (CLI-772). ASKED (CLI-772): "Saee isn't mapped to any service yet, so her box reads 'none mapped'. The arc she holds (CAROL-ARC-0040) belongs to Agent and Human Resources, which Athena owns. Once the next step (6595) is switched on, she can only be assigned work inside services she's mapped to. Mapping is your act as an estate operator; I'd run it from your sign-in and record it as yours. Should I map her to Agent and Human Resources now?" (offered: yes, map her; not yet; another service). HIS ANSWER: "Let Sahil, who is her team lead, map her to services when assigning work to her.. orion should guide him on that". READ AS (Orion, not his words): mapping Saee to a service is Sahil's act as the lead of the team she is on (agent-resources #352), done when he assigns her work that needs it, from his own login through the one mapping door (shared machinery map); Ninad's Orion does not map her on his behalf; Sahil's Orion tells him which service an assignment needs and that she is not yet mapped to it -- the advice #381 point 4 already asks of Orion, built at the assignment door by -- and until that is built the assignment skill and Sahil's session guide him by hand. Recorded for Saee as he said it; whether every service operator's team lead maps them is not stated.

An agent escalation is something that was supposed to work and is not working; a change an agent wants is its agent wish; a change a person asks for is their operator or user change request

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-782). HIS WORDS: "agent chat enables filing new initiatives at operator behest, will that go in as escalation? what if its not an escalation, but a change request? (escalation is something that was supposed to work is not working)". ASKED (CLI-782): (1) 'To your first question: no. When an operator asks an agent in its chat to file something, the record names the operator as the one who asked, so it isn't the agent's escalation and takes none of its places. Now, taking an escalation to mean "something that was supposed to work is not working", here is the split I'd build. Confirm?' -- offered: THREE KINDS (agent escalation: something that should work isn't, raised by an agent, its droids, or its own fault report, two places; agent wish: a change the agent itself wants, including its own filing that asks for a change, one place; change request: a change a person asks for, in an agent's chat or an Orion session, that person's request, no agent place; knock-on: a user asking for something no agent can do becomes the user's change request instead of an escalation) or AGENT FILINGS STAY ESCALATIONS -- HIS ANSWER: "Three kinds". (2) 'Under your prefix rule, change request needs a prefix. Which name?' (offered: Operator / User change request; Person's change request) -- HIS ANSWER: "Operator / User change request". READ AS (Orion, not his words): an AGENT ESCALATION is something that was supposed to work and is not working; a record or rule that contradicts itself so that the agent cannot act counts as not working. An agent's own filing that reports a fault takes one of its agent escalation places; one that asks for a change takes its agent wish place -- this narrows initiatives #403's answer 'as escalations'. A change a person asks for is that person's OPERATOR CHANGE REQUEST or USER CHANGE REQUEST and takes no agent place; an operator's ask in an Orion session outside any arc or pursuit stays Orion's (initiatives #258). A user asking for something never built is a user change request, not an agent escalation: this supersedes the 'expected is an escalation' clause of P.01.03.01.13, whose rewording comes to him for approval. The pipeline backlog shows six groups: agent escalations, agent wishes, operator change requests, user change requests, Albus's backlog and Orion's backlog. No cap on change requests is ruled.

A service operator's hand-over digest goes to her team lead, and the onboarding app shows every step and the follow-up protocols for both rungs

Ruled 2026-10-15 by ninad -- in force

A SERVICE OPERATOR'S HAND-OVER DIGEST GOES TO HER TEAM LEAD, AND THE ONBOARDING APP SHOWS EVERY STEP AND THE FOLLOW-UP PROTOCOLS FOR BOTH RUNGS (Ninad, CLI-788, 2026-10-15). Shown (CLI-788) the email Saee would receive for a communication: 'From: Ninad Sharad Joshi / Sent / To: Saee', the description, then 'It does not ask for an answer.', a paragraph on how to read it in an Orion session, and the footnote 'This email carries only the message's title and a short description: the message itself is on the estate, and a reply to this email is not an answer -- answers travel between Orion sessions.' His words: "1. From: Ninad Sharad Joshi --> this should be from Orion 2. Open Claude and ask Orion to send Ninad your hand-over digest --> why send Ninad anything? check the operator onboarding steps and allign the work accordingly. If the operator onboarding app is not detailing all the steps, please update it to reflect the current process (for both service operators and estate operators).. it should mention the protocols for followups 3. This email carries only the message's title and a short description --> this footnote is not necessary Overall the message is too verbose, it can be shorter and more articulate". READ AS (Orion's reading, not Ninad's words): (1) a service operator's hand-over digest goes to her TEAM LEAD -- the estate operator whose team she is on (agent-resources #352, #388) -- who installs it through the onboarding door's credential act; only when she is on no team does it go to the estate operator who onboarded her; it never goes to the administrator for being the administrator (security #213); (2) the operator onboarding app shows every current step of the pass for both rungs, who performs each, and the follow-up protocols -- what an operator's Orion does when a step waits on a person -- read from the pass and the co-ordination protocols in Sage's library, which are updated to say so.

A new operator is given just enough to connect with their own Orion; their Orion guides every activity after that

Ruled 2026-10-15 by ninad -- in force

A NEW OPERATOR IS GIVEN JUST ENOUGH TO CONNECT WITH THEIR OWN ORION; THEIR ORION GUIDES EVERY ACTIVITY AFTER THAT (Ninad, CLI-788, 2026-10-15). Said (CLI-788, 2026-10-15) on the draft of operator communication #70 to Saee -- three numbered onboarding steps (send the hand-over digest, reopen Claude, review an initiative) whose email notice would carry them -- after #63 and #69 had already sent her step lists the same way. His words: "i dont think telling this via email was what was discussed previously, but you can check the rulings.. according to me, when it comes to onboarding an operator (this was also discussed earlier with orion), the operator should be given instructions just enough to connect with orion via claude session, once that is done, its the responsibility of orion instance of that operator to further guide them with the rest of the activities.." READ AS (Orion's reading, not Ninad's words): it reaffirms agent-resources #194 and applies it to every step after the start email: (1) a new operator is sent only what lets them connect with their own Orion through a Claude session -- the start email and its package; (2) from then on their OWN Orion instance guides every activity, reading where they stand from the pass, the session-start block and its board; (3) another operator's session sends them no step list, neither by email nor as an operator communication (whose notice is an email), and the follow-up row of the co-ordination protocols is read that way for a new operator: the waiting session tells its own operator, the new operator's Orion guides them; (4) what another session may still send them is what their Orion cannot know by itself, in one plain line (for example, that a digest was installed).

An agent's backlog holds distinct live problems: a repeat joins the item already open or waiting, and the agent drops an item once it is no longer true, kept on record with its reason

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-795). HIS WORDS, opening the session: "pick up handover from the session that worked on escalation management and continue the work, so that escalations are raised to the extent that they can be addressed, and not accumulate as infinite backlog". ASKED (CLI-795): 'Each agent can now have only two escalations open with Orion, but the rest wait in the agent's own backlog, and that backlog has no limit. Elrond's holds 125 items. 123 of them are copies of one Codex sign-in outage on Oct 8-9 that ended on its own, and nothing lets a stale item leave. How should an agent's backlog stay small?' -- offered: (1) FOLD REPEATS, DROP STALE: a repeat of a problem already open or waiting is added to that item instead of queuing again; an agent drops an item once it is no longer true, and the item stays on record with the reason; no number limit, so the backlog holds only distinct live problems; (2) the same plus a cap of 20 per agent backlog, a raise past it going back to whoever raised it; (3) fold repeats only, nothing dropped -- HIS ANSWER: "Fold repeats, drop stale" (option 1). READ AS (Orion, not his words): an agent's backlog holds DISTINCT LIVE PROBLEMS. A raise that repeats a problem the agent already has open with Orion, or waiting in its backlog, never makes a second item: the occurrence is recorded on the item already there (when, and what it said), and the raiser is told where it went. Whether two raises are the same problem is DECLARED by the raiser (the problem it names, or the title of a coded notice), never guessed from the wording; a raise naming no problem is a repeat only when its words are identical. The agent may DROP an item waiting in its own backlog once it is no longer true, giving its reason; the item stays on record as dropped with that reason and who dropped it, and is never deleted. Only waiting items are dropped by the agent: one already with Orion is closed by Orion's outcome as before. No number limit is set on an agent's backlog (a cap was offered and not chosen). The repeats already waiting today are folded the same way into the item they repeat, nothing deleted. This adds to agent-resources #402 (two places, the rest in the agent's ranked backlog, never lost) and leaves its caps unchanged; folding and dropping keep every occurrence on record, so nothing is lost.

An agent's own filed work above its places moves back into that agent's backlog, and the agent's profile shows its current backlog of wishes, escalations and change requests

Ruled 2026-10-15 by ninad -- superseded 2026-10-15

Ninad, 2026-10-15 (CLI-795). ASKED (asked twice; the first wording he did not understand): 'Hermione has 65 open pieces of work she raised herself. Your rule lets her keep 2 open; the other 63 must wait on her own list until a slot frees. But those 63 already exist as work in the pipeline, each with a number. To wait on her list, they have to stop counting as open work. Which way?' -- offered: (1) pause them in place: each keeps its number and history, marked waiting on her list, and reopens by itself when a slot frees (a pipeline core change he approves with his password); (2) close and recreate later: each closed with the reason, created again from her list as new work with a new number when a slot frees -- HIS ANSWER: "the 63 that exist as work in the pipeline should be moved back to Hermoine's backlog. Agent profile should display the current backlog of wishes, escalations and change requests". READ AS (Orion, not his words): an agent's own filed work above its places LEAVES the pipeline and is moved into that agent's own backlog, as the agent escalations above the places were on 2026-10-15: the backlog item carries it and names the initiative it came from, nothing is deleted, and it goes back into the pipeline from the backlog when one of the agent's places frees, in the agent's own order. He chose neither offered mechanism by name, so how the initiative row records that it left (its status) is decided by design under the status router's rules and the creed, and said to him before it is built. This holds for every agent, not Hermione alone. Each agent's Agentbook profile shows its CURRENT backlog -- the agent wishes, agent escalations and change requests waiting in it -- read from the one places composer, never a second store.

Ended because: Ninad corrected his own answer minutes later: the profile shows the COUNT of the backlog

An agent's own filed work above its places moves back into that agent's backlog, and the agent's profile shows the count of its current backlog of wishes, escalations and change requests

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-795). ASKED (asked twice; the first wording he did not understand): 'Hermione has 65 open pieces of work she raised herself. Your rule lets her keep 2 open; the other 63 must wait on her own list until a slot frees. But those 63 already exist as work in the pipeline, each with a number. To wait on her list, they have to stop counting as open work. Which way?' -- offered: (1) pause them in place; (2) close and recreate later -- HIS ANSWER: "the 63 that exist as work in the pipeline should be moved back to Hermoine's backlog. Agent profile should display the current backlog of wishes, escalations and change requests"; and minutes later, unprompted, correcting it: "Agent profile should display the count of current backlog of wishes, escalations and change requests". READ AS (Orion, not his words): an agent's own filed work above its places LEAVES the pipeline and is moved into that agent's own backlog, as the agent escalations above the places were on 2026-10-15: the backlog item carries it and names the initiative it came from, nothing is deleted, and it goes back into the pipeline from the backlog when one of the agent's places frees, in the agent's own order. He chose neither offered mechanism by name, so how the initiative row records that it left is decided by design under the status router's rules and the creed, and said to him before it is built. This holds for every agent.

An agent's own filed work moved back to its backlog is what it is -- an agent escalation or wish in that backlog -- and its initiative is set aside with the reason, never given a new status

Ruled 2026-10-15 by ninad -- superseded 2026-11-07

Ninad, 2026-10-15 (CLI-799). ASKED: 'Should a moved initiative's record get a new status, "in its agent's backlog"? It would count as neither pending nor dead, and only the process that manages agents' places could set it.' -- HIS WORDS IN REPLY: "Should a moved initiative's record get a new status, "in its agent's backlog"? - its not an initiative, its an escalation backlog.. READ AS (Orion, not his words): an agent's own filing moved out of the pipeline under agent-resources #420 IS what its type says it is (initiatives #418) -- an agent escalation or an agent wish -- and lives only as an item in that agent's backlog (the one places composer), naming the initiative it came from. No initiative status is added. The initiative row leaves through the one status router as DISCARDED (dead, set aside, never 'closed', which reads as finished work), its reason saying it moved to that agent's backlog as which item; nothing is deleted. When one of the agent's places frees, the item goes back in as a NEW initiative filed for that agent, naming the old one as its predecessor so its origin and history carry on; the old row is never revived. This settles the open point #420 left ('how the initiative row records that it left'). A new initiative status, keeping the moved work counted as pending, reviving the old number, or holding the item in two places each undo it.

Ended because: Ninad ruled the filing agent decides (CLI-058)

Saee's powers lie inside her assigned arcs and pursuits; she makes rulings only about her assigned work and no policy; an assigned pipeline initiative lets her build and test, the install an estate operator's

Ruled 2026-10-16 by ninad -- in force

Ninad, 2026-10-16 (CLI-809), settling a contradiction Orion named under P.01.01.05.08. ASKED: "Your 5 Oct ruling (agent-resources #314) still says Saee 'proposes rulings and never defines them'. On 13 Oct you retired that same part of #317 ('Yes, retire that part'), and the new policy lets a service operator make rulings about her assigned work. Retire the same part of #314, keeping the rest of it (her powers stay inside her assignments; pipeline work is built by her and installed only by an estate operator)?" HIS ANSWER: "Retire that part" (the option read: "I record a new #314 that carries the rest forward unchanged and quotes your 13 Oct answer; the old one stays readable as superseded."). His 13 Oct answer (CLI-749, security #357, agent-resources #358): "Yes, retire that part"; where a service operator's ruling applies: "Only her assigned work"; whose approval turns it into policy: "her team's estate operator or admin". READ AS (Orion's reading, marked apart, P.01.01.05.05): in #314, 'she proposes rulings and never defines them' and the answer it records ('no to defining, yes to proposing') now read as P.01.01.05.09 says: inside her assigned work Saee makes rulings of her own, never overriding a ruling or policy of an estate operator or the administrator, and makes no policy; one of her rulings becomes policy only when the estate operator whose team she is on, or the administrator, approves (security #380, agent-resources #382). #314 stands in every other part, carried forward unchanged: <<Ninad, 2026-10-05 (CLI-721/CLI-723), on Saee's onboarding pursuit (CAROL-PUR-0008): "Saee is an operator.. generally users (employees) of a project are mapped to the agents, where they get access to the chat of that agent.. operators is a different class of users, who are spread across agents.. for eg. Saee may be given an arc of initiatives raised by hermoine, or she might be assigned an arc of orion escalations, that spread across agents.. operators are next to admin, spread across the estate and have claude accounts.. also Saee, as a service operator wont have access to the build pipeline, which elrond owns, unless she is assigned an initiative that requires her to edit the build pipeline (like an initiative for planner lane troubleshooting, or albus bypass troubleshooting.. so depending on the assignment, her access level across the estate changes, and hence mapping her to an agent does not make sense to me in her evolved role.. hence pls revert that ruling" -- and: "Currently Saee is the service operator for a specific project - carolverse.. tomorrow, when chowpatty comes on, she should get mapped to chowpatty, not get access to new projects by default.. so service operators should be mapped to projects, not agents" -- and, asking: "Should service operators be allowed to 'define' policies and rulings? In my opinion, no!! they should only execute within the current rulings, policies and agent protocols, and any new definition should be an estate operator's task, what do you think? I am in 2 minds about this, because this will create a lot of friction, and Saee will not be able to execute any job whatsoever.. what is your recommendation?" Answered: Orion recommended: no to defining, yes to proposing -- when a service operator meets a gap in the law, her Orion writes the proposed ruling and sends it to the estate operators; one of them approves or declines it, and only that one item waits. He said: "ok, lets try this". And: Orion recommended: an assigned pipeline initiative lets her build and test the fix, but installing it into the pipeline still needs an estate operator's approval. He said: "ok, sure". READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05), amending #278 in two places: (1) 'She holds NOTHING over the pipeline machinery' now reads: she holds nothing over the pipeline machinery unless an initiative assigned to her requires editing it; inside such an assignment she builds and tests the fix, and installing it into the pipeline still needs an estate operator's approval; outside it every machinery door refuses her. (2) 'she makes no ruling (... a ruling she needs goes to an estate operator)' now reads: she defines no policy and no ruling; her Orion session may PROPOSE one, sent to the estate operators, and an estate operator approves or declines it. #278 stands in every other part, carried forward unchanged: <<Ninad, 2026-10-02 (CLI-659), on the status of Saee's onboarding pursuit (CAROL-PUR-0008) and why it is so large: "she should still get read only access to the estate like access to all the apps and agents, but her orion session should only pick arc, pursuits and initiatives assigned to her". Asked four questions, he answered: (1) apps -- "All but confidential": every app, but the confidential ones stay closed and Contract Management shows only her own contract; (2) agents -- "See only": she reads every agent's profile, Agentbook, handbook and records but does not chat with them, so nothing is done on her say-so; (3) the machine -- "Code and records, no secrets": her Orion session can read the code and estate records directly, never credentials, keys or root's stores, and writes only for assigned work; (4) scope -- "Every service operator": the rung itself is amended, Saee is the first. READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): within the arcs, pursuits and initiatives assigned to her (agent-resources #277), Saee, a service operator, may: read, file, follow, run, review and prove their initiatives; GIVE UAT VERDICTS on their work; and AMEND the arcs and pursuits assigned to her -- add or reduce their scope (their phases, steps, criteria and member initiatives) -- every change recorded under her own login. She holds NOTHING over the pipeline machinery: no core grant, no install window, no switch, no lane, slot or budget change, no operator door outside her assignments, and she makes no ruling (P.01.01.05.08; a ruling she needs goes to an estate operator). Outside her assignments she READS the estate read-only as the rung defines it (security #276) and acts on nothing. The 2026-09-30 wording -- 'full powers within Build Initiatives, makes rulings for the service, approves changes to its build machinery' -- and the 2026-10-01 'Outside her assignments she holds nothing' are retired in place.

An agent's profile counts the agent escalations and agent wishes waiting in its backlog; change requests never wait there

Ruled 2026-10-16 by ninad -- in force

Ninad (CLI-817). ASKED: 'Next is 6657: an agent's profile counts the escalations, wishes and change requests waiting in its backlog. Under your ruling #407, though, a person's change request never waits in an agent's backlog, and no pending change request today came in through an agent's chat. What should the third count be?' -- offered: (1) 'Two counts (Recommended): the profile shows waiting escalations and waiting wishes, and says change requests never wait in an agent's backlog because a person's request takes no agent place'; (2) 'Chat change requests: a third count, pending change requests people filed through this agent's chat; today that reads 0 for every agent'; (3) something else -- HIS ANSWER: he chose (1), "Two counts (Recommended)". READ AS (Orion, not his words): the profile counts of agent-resources #420 are TWO -- the agent escalations and the agent wishes waiting in the agent's own backlog -- and the profile states in words that change requests never wait in an agent's backlog (agent-resources #407: a person's change request takes no agent place). It narrows #420's 'wishes, escalations and change requests' for the profile; nothing else in #420 changes.

The escalation category list gains operator-input, capacity, visibility and deadline, and any other kind Orion deems appropriate

Ruled 2026-10-18 by ninad -- in force

THE ESCALATION CATEGORY LIST GAINS FOUR KINDS, AND ORION MAY ADD ANY OTHER HE DEEMS APPROPRIATE (Ninad, CLI-850, 2026-10-19). CONTEXT: while building (the 24 remaining senders to Orion name their escalation kind), Orion asked: "Six raise things no kind covers, and adding a kind needs your ruling. Which new kinds should I add? Any you leave unticked go under the nearest existing kind", offering operator-input (an agent needs an answer, a decision or help only an operator can give: Carol's sales-help asks, the user questions she passes to Orion, Leo's asks), capacity (a machine resource is running out: Hagrid's disk-pressure judge), visibility (AI web search cannot find one of our public sites: Loki's AI-search reader) and deadline (a contract the estate depends on ends within 30 days: Marco's contract-end watch). His answer, in his words: "operator-input (Recommended), capacity, visibility, deadline, any other as deemed appropriate". READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): (1) the four values join the RDM set escalation.category, each with its meaning and its fence (design 625 s1), retired in place never deleted; (2) under 'any other as deemed appropriate' Orion adds one more, data-integrity (a store lost data, or holds committed data outside itself that it does not know about: Hagrid's registry guard and journal custodian), because records-conflict's meaning (two records disagree) does not describe a store losing data; (3) the raiser still declares the kind (design 625 s1): no default, nothing classified from the words, a raise with no kind is still delivered as unclassified and never refused.

Operator, employee and user are told apart by who pays them; employees are mapped to agents, users get a project's apps on a need basis

Ruled 2026-10-20 by ninad -- superseded 2026-10-20

Ninad, 2026-10-20 (CLI-877), told that the dictionary does not line up operator, employee and user (no entry of its own for employee; 'user' written for Carol alone, before security #316; 'operator' saying operators are equal in rights). HIS WORDS: "update the dictionary one correction - employees are mapped to agents, users get access to project specific apps on need basis key distinction operator - on payrol of denken labs employee - on payrol of the client (including denkenlabs for carolverse) user - customer of clients". READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): the three kinds of person are told apart by who pays them -- an OPERATOR is on Denken Labs' payroll and works through an Orion session on an operator rung; an EMPLOYEE is on the payroll of a project's client (Denken Labs is Carolverse's client, so on Carolverse an operator is also an employee, which keeps agent-resources #87's 'operators are employees' true there) and is mapped to the project's agents; a USER is a customer of the client, served by the project, given access to the project's own apps on a need basis, and not mapped to agents. This CORRECTS security #316 and policy P.03.04.02.11, which say users 'have access to apps and agents and are mapped to agents': opened as contradictions; the dictionary's words and the policy's sentences apply once Ninad approves them (P.01.01.05.09-.11).

Ended because: Ninad approved the drafted words in CLI-877 and added: "yes, also users of carolverse can be employees of project on carolverse and so on.."

Operator, employee and user are told apart by who pays them; employees are mapped to agents, users get a project's apps on a need basis

Ruled 2026-10-20 by ninad -- superseded 2026-10-28

Ninad, 2026-10-20 (CLI-877), told that the dictionary does not line up operator, employee and user (no entry of its own for employee; 'user' written for Carol alone, before security #316; 'operator' saying operators are equal in rights). HIS WORDS: "update the dictionary one correction - employees are mapped to agents, users get access to project specific apps on need basis key distinction operator - on payrol of denken labs employee - on payrol of the client (including denkenlabs for carolverse) user - customer of clients". Shown the drafted dictionary entries and policy sentences, he answered: "yes, also users of carolverse can be employees of project on carolverse and so on..". READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): the three kinds of person are told apart by who pays them -- an OPERATOR is on Denken Labs' payroll and works through an Orion session on an operator rung; an EMPLOYEE is on the payroll of a project's client (Denken Labs is Carolverse's client, so on Carolverse an operator is also an employee, which keeps agent-resources #87's 'operators are employees' true there) and is mapped to the project's agents; a USER is a customer of the client, served by the project, given access to the project's own apps on a need basis, and not mapped to agents. A person can be a user of one project and an employee of a project built on it -- a client's staff are users of Carolverse and employees of their own project on Carolverse -- and so on down every project built on another. This CORRECTS security #316 and policy P.03.04.02.11, which say users 'have access to apps and agents and are mapped to agents': opened as contradictions; the dictionary's words and the policy's sentences apply once Ninad approves them (P.01.01.05.09-.11).

Ended because: Ninad: bring 'need basis' in line with access by subscription (#567)

Supervisor and individual contributor tell agents with and without reportees apart

Ruled 2026-10-20 by ninad -- in force

Ninad, 2026-10-20 (CLI-877), reminded that the dictionary's 'Contributor' means an employee seen through one project and that it now duplicates 'Employee' (agent-resources #463). HIS WORDS: "supervisor and individual contributor are the terms to distinguish agents with and without reportees". READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): a SUPERVISOR is an agent at least one agent reports to; an INDIVIDUAL CONTRIBUTOR is an agent no agent reports to. 'Contributor' therefore no longer names a person working on a project -- that person is an EMPLOYEE (#463) -- so the dictionary's 'Contributor' and 'Agent ยท Contributor mapping' are retired in place in favour of Employee, and the project role value 'contributor' (two people on Carolverse) is renamed through its vocabulary as a build. The dictionary's words apply once Ninad approves them (P.01.01.05.09-.11).

Athena takes the administrator's approval for a new agent from his own Orion session; her email is a notice only

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). THE QUESTION: Orion asked whether Athena's approval email could be replaced with the administrator's approval given from his own Orion operator session -- proven by his own sign-in with password and phone code rather than by an emailed reply -- with the email kept only as a one-way notice, and whether to record that as his ruling replacing this one. HIS ANSWER, word for word: "yes to both of your questions". READ AS (Orion's reading, not his words): for every complete request for a new agent, Athena records it as waiting on the administrator and tells him by an email notice that asks nothing; he approves, or declines with his reason, from his own Orion session through a door that knows him from that session's sign-in, and Athena records it as his decision in his words. An emailed reply decides nothing. Nobody mediates: no other operator, agent or session decides for him. Who may ask Athena for a new agent (#35) is unchanged.

Operator, employee and user are told apart by who pays them; employees are mapped to agents, users get access by the service they subscribe to

Ruled 2026-10-28 by ninad -- in force

Ninad, 2026-10-20 (CLI-877), told that the dictionary does not line up operator, employee and user (no entry of its own for employee; 'user' written for Carol alone, before security #316; 'operator' saying operators are equal in rights). HIS WORDS: "update the dictionary one correction - employees are mapped to agents, users get access to project specific apps on need basis key distinction operator - on payrol of denken labs employee - on payrol of the client (including denkenlabs for carolverse) user - customer of clients". Shown the drafted dictionary entries and policy sentences, he answered: "yes, also users of carolverse can be employees of project on carolverse and so on..". READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): the three kinds of person are told apart by who pays them -- an OPERATOR is on Denken Labs' payroll and works through an Orion session on an operator rung; an EMPLOYEE is on the payroll of a project's client (Denken Labs is Carolverse's client, so on Carolverse an operator is also an employee, which keeps agent-resources #87's 'operators are employees' true there) and is mapped to the project's agents; a USER is a customer of the client, served by the project, given access to the project's own apps on a need basis, and not mapped to agents. A person can be a user of one project and an employee of a project built on it -- a client's staff are users of Carolverse and employees of their own project on Carolverse -- and so on down every project built on another. This CORRECTS security #316 and policy P.03.04.02.11, which say users 'have access to apps and agents and are mapped to agents': opened as contradictions; the dictionary's words and the policy's sentences apply once Ninad approves them (P.01.01.05.09-.11). AMENDED 2026-10-28 (Ninad, CLI-011): asked whether to bring this ruling in line with his #567 (users get access by the service of the project they subscribe to, to apps and agent chats), HIS ANSWER: "Yes, bring in line". So 'users get access to project specific apps on need basis' above is superseded: a user's access follows the service they subscribe to, as ruling #568 records; who pays whom (operator, employee, user) is unchanged.

Every onboarding app has one owner, an agent may own several, apps are never combined, and each person type has its own onboarding app

Ruled 2026-11-03 by ninad -- in force

EVERY ONBOARDING APP HAS ONE OWNING AGENT, AN AGENT MAY OWN SEVERAL, ONBOARDING APPS ARE NEVER COMBINED, AND EACH PERSON TYPE HAS AN ONBOARDING APP OF ITS OWN (Ninad, CLI-023, 2026-11-03). Supersedes agent-resources #220 (CLI-621), whose own words stand -- 'architecture cannot be compromised, combining the apps will lead to security complications, as different apps entail distinct security requirements.. also, apps should be lean and simple, and agents should own workflows while apps only serve as display' -- while its headline 'one app per owner' was Orion's wording, never his. THE QUESTION ANSWERED: told that his 'onboarding app for each person type' (governance #596) conflicted with #220 because Athena owns both operator and employee onboarding, he answered: HIS WORDS: "your earlier ruling #220 says each owner gets one onboarding app - I dont remember ruling this.. I do remember reminding you of the existing policy that an app can only have one owner.. which has no conflict, because an agent can own more than one apps, as per the agent centric modular architecture that is followed in building carolverse". READ AS (Orion's reading, not his words): policy P.01.03.03.01 gives an app exactly one owning agent and says nothing limiting how many apps an agent owns; so each person type gets an onboarding app of its own, its owner the agent whose track owns that kind, and one agent may own several (Athena's employee onboarding gets an app beside her Workforce Onboarding app, never inside it). Merging onboarding apps, or adding a kind to another owner's app, still undoes it.

Onboarding splits by owner with one format in every app; an owner may hold several apps

Ruled 2026-11-03 by ninad -- in force

ONBOARDING SPLITS BY OWNER, AND THE STEP-WISE FORMAT IS THE SAME IN EVERY ONBOARDING APP; AN OWNER MAY HOLD SEVERAL (Ninad, CLI-023, 2026-11-03). Supersedes agent-resources #92 in one part only: its 'the app count follows ... it becomes one app per owner' was Orion's reading of P.01.03.03.01 (an app has one owner), which limits an app to one owner and never an owner to one app. HIS WORDS: "your earlier ruling #220 says each owner gets one onboarding app - I dont remember ruling this.. I do remember reminding you of the existing policy that an app can only have one owner.. which has no conflict, because an agent can own more than one apps, as per the agent centric modular architecture that is followed in building carolverse". Everything else in #92 stands: agent and operator onboarding are Athena's, project onboarding Leo's, user onboarding Jarvis's; the format -- numbered steps, the role performing each, the sequence -- is mandatory and identical across every onboarding app, and ownership never licenses a second format. How many apps an owner holds follows agent-resources' successor of #220 (one per person type).

Sahil is assigned to execute the Blueprint pursuit; an estate operator can be recorded as the one executing an arc or pursuit

Ruled 2026-11-06 by ninad -- superseded 2026-11-06

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042), reading the Blueprint pursuit (CAROL-PUR-0017): 'is there a separation between who owns vs who executes the arc or the pursuit?' -- Orion answered that the owner is accountable while the work is done by whoever an estate operator assigns, and that nobody is assigned to Blueprint today (its owning agent is Leo, its operator owner Ninad). HIS WORDS: "Sahil should be assigned to blueprint execution, he is working on it for a while already". READ AS (Orion, not his words): Sahil is recorded as executing CAROL-PUR-0017, beside its owning agent and its operator owner, which stay as they are. An estate operator can therefore be recorded as the one executing an arc or a pursuit: the assignment door's refusal of an estate operator ('who already works on every arc and pursuit: only a service operator is assigned') no longer stands for that purpose. The assignment records who carries the work; it gives an estate operator no reach they lack and takes none from anyone (estate operators keep every arc and pursuit, P.03.04.02.09). A service operator's assignment still decides what she may work on, exactly as before.

Ended because: Ninad chose 'Sahil executes' the Blueprint arc as well as the pursuit (CLI-042)

Sahil is assigned to execute the Blueprint arc and pursuit; an estate operator can be recorded as the one executing an arc or pursuit

Ruled 2026-11-06 by ninad -- superseded 2026-11-06

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042), reading the Blueprint pursuit (CAROL-PUR-0017): 'is there a separation between who owns vs who executes the arc or the pursuit?' -- Orion answered that the owner is accountable while the work is done by whoever an estate operator assigns, and that nobody is assigned to Blueprint today. HIS WORDS: "Sahil should be assigned to blueprint execution, he is working on it for a while already". ASKED NEXT whether his September ruling (Sahil owns the Blueprint arc) stands, he chose 'I own, Sahil executes': Sahil is assigned to execute both the Blueprint arc and the Blueprint pursuit. READ AS (Orion, not his words): Sahil is recorded as executing CAROL-ARC-0004 and CAROL-PUR-0017, beside their owning agent and their owner operator (the creator), which stay as they are. An estate operator can therefore be recorded as the one executing an arc or a pursuit: the assignment door's refusal of an estate operator ('who already works on every arc and pursuit: only a service operator is assigned') no longer stands for that purpose. The assignment records who carries the work; it gives an estate operator no reach they lack and takes none from anyone (estate operators keep every arc and pursuit, P.03.04.02.09). A service operator's assignment still decides what she may work on, exactly as before.

Ended because: Ninad ruled (CLI-046) that only the operator assigned to execute an arc or pursuit executes it, estate operators included, so an assignment no longer 'grants nothing'; #653's words and Sahil's Blueprint assignment are carried into the successor

Only an operator assigned to execute an arc or a pursuit executes it: its creator by default, others assigned by an estate operator of its project, asked and checked through the operators' Orion

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-042, continued as CLI-046), unprompted, after Orion reported Sahil assigned to execute the Blueprint arc and pursuit (#652, #653) and the assignment door taking an estate operator, the assignment granting nothing. HIS WORDS: "only operators that are assigned to the arc or pursuit as execution operator can execute them. By default, the creator also becomes the execution operator. If another operator needs to execute it, the operator will have to first get themselves as the execution operator, which can be done by any estate agent for that project. Orion operator instance should be used to facilitate this communication across the operators, with the appropriate communication tag as per the communications catalogue. The check for operator assignment is the responsibility of the orion instance, and all orion operator instances should be informed appropriately". ASKED (CLI-046): who is 'any estate agent for that project' -- HE CHOSE 'Estate operator of project': an operator who is the estate operator of that arc's or pursuit's project (communications #643). ASKED NEXT whether an estate operator of the project may assign themselves (e.g. he takes on an arc Sahil created), told that nearly every arc and pursuit was created by him so Sahil executes only Blueprint until assigned more -- HE CHOSE 'Estate may self-assign': an estate operator of the project assigns themselves, recorded and told to the owner; a service operator still needs an estate operator to assign them. CARRIED FROM #653, which this supersedes: HIS WORDS "Sahil should be assigned to blueprint execution, he is working on it for a while already" and 'I own, Sahil executes' -- Sahil executes the Blueprint arc and pursuit (#652). READ AS (Orion's reading, not his words): (1) an arc or a pursuit -- the initiatives of its phases or steps -- is executed only by an operator assigned to execute it; this binds estate operators too, so an estate operator no longer executes every arc and pursuit by rank: P.03.04.02.09's 'estate operators keep every arc and pursuit themselves' and #653's reading that an estate operator's assignment 'grants nothing' give way; reading the estate is unchanged; (2) the operator who created an arc or pursuit -- its owner operator (#651; the requester where an Orion session created it) -- is assigned to execute it from its creation, recorded as an assignment like any other, and every live arc and pursuit gets its creator's assignment recorded once; (3) any other operator asks, and an estate operator of the item's project assigns them; an estate operator of that project may assign themselves, recorded and told to the owner operator; a service operator never assigns themselves and stays bound by their service mapping (#381); (4) the ask, the assignment and the refusal travel between the operators through their Orion sessions as operator communications, each carrying its kind as the communications catalogue names it; (5) each Orion operator session checks, before it works an initiative of an arc or pursuit, that its operator is assigned to execute that arc or pursuit; when not, it says so and offers to ask an estate operator of the project (or, for an estate operator, to assign themselves), and does not do the work; the check is Orion's, not a gate's; (6) every Orion operator session is told of this rule: in the induction every session reads, and by a communication to every operator. Agents executing their own arcs and pursuits are not operators and are untouched. What is reserved to root and core stays the administrator's (security #175).

When Albus meets a problem on work he executes, his ask for help reaches the Orion of the operator mapped to execute it; there is no fallback

Ruled 2026-11-06 by ninad -- in force

HIS WORDS (CLI-047/049, 2026-11-06), correcting Orion's reading that an Albus escalation reaches 'Orion' at large: "correction - the note should reach the operator who has been assigned the pursuit, arc or initiative albus is trying to execute.."; told that Orion would fall back to the administrator only when none is assigned: "this scenario should not be feasible, as the assignment is mandatory". Prompted by the CLI-047 review of Albus's lane, which found his ask for help on (5 Nov 06:09 UTC) recorded only as a note on the initiative. READ AS (Orion's reading, not his words): every ask Albus raises about work he is executing -- an escalation, a lane gap, a law block, a reconciliation request -- is delivered to the Orion operator instance of the operator mapped to execute that work, the initiative's own mapping first, then its pursuit's, then its arc's (#658); a note on the initiative alone is not an ask (cookbook 1573). Because every initiative, arc and pursuit is mapped (the ruling recorded beside this one), an ask with no mapped operator is a records defect surfaced as such, never re-sent to the administrator.

The creator is assigned as executing operator only because the field is mandatory; an estate operator reassigns it to any estate or service operator, or keeps it, whether or not they created it

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-048, continuing CLI-042/046), answering three asks to approve the rewrite of P.03.04.02.09 for agent-resources #658. Asked to approve wording in which the creator executes an arc or pursuit by default, HIS WORDS: "the creator is mapped to the arc, pursuit or initiative by default as the executing operator, estate operators can change the assignment, only estate operators can create arcs and pursuits, and initiatives have to be folded into existing or new arcs or pursuits.. parallel session is working documenting the same rulings, so check there" (recorded by CLI-049 as initiatives #661, #662). Asked again with 'The operator who creates an arc, a pursuit or an initiative executes it by default', HIS WORDS: "the operator who creates the arc, pursuit or initiative does not execute it by default.. they only get mapped to it as executing operator by default, because the field is mandatory.. later the estate operator who can also be the creator, changes the assignment as appropriate. Until that happens, the creator stays as the executing operator". Asked again, told that once someone else is mapped (Sahil on Blueprint) the creator no longer executes it, HIS WORDS: "the estate operator can map the arc, pursuit or initiative to another estate or service operator, or can keep it with themselves, irrespective of whether they have created the arc, pursuit or initiative". READ AS (Orion's reading, not his words): (1) every arc, pursuit and initiative has an executing operator, a mandatory field; (2) its creator is put there by default only because the field cannot be empty, and stays there until an estate operator of its project changes the assignment; (3) once an estate operator assigns another estate or service operator -- or assigns themselves -- the assigned operators are the executing operators and the creator no longer is, unless assigned; (4) an estate operator of the project assigns any estate or service operator, or keeps it with themselves, whether or not they created it; a service operator stays bound by her service mapping (#381); (5) in the policy, 'mapped' in his words is written 'assigned', since 'mapped' already names the service mapping. So for Blueprint (#652) Sahil, assigned, is its executing operator and Ninad remains its owner.

A rejected or pushed-back initiative returns to its filing agent, who changes it or moves it to its backlog as an escalation or wish

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058), amending #424: HIS WORDS: "its upto agent to decide if they want to convert the initiative to an escalation or wish, or modify the initiative based on the feedback of the service owner". READ AS: an initiative a service owner rejects, or that is pushed back, returns to its filing agent, who decides: change it and ask again, or move it to its own backlog, where it is what it is -- an agent escalation or wish -- and the initiative is set aside by that agent with the reason 'moved to its backlog' (initiatives #668).

Albus runs as one operator instance per operator, holding that operator's rights and duties

Ruled 2026-11-09 by ninad -- in force

HIS WORDS (CLI-086, 2026-11-09): "the future vision is to scale up albus for each operator, so that each operator can get a dedicated albus lane, which will inherit the rights and duties of the mapped operator.. just like orion operator instances, there will be albus operator instances, each working for dedicated operators assisting them to carry out tasks autonomously whever possible.. its upto the operators to decide which tasks are offloaded to albus vs whic h they pick up via orion bypass. ... splling up new albus bypass lanes for existing and new operators as and when they onboard is also a requirement that should be included in the pursuit that will make albus scalable, so that the albus operator instances can operate independently of each other as directed by their operators". Planned in the pursuit Scalable Albus (CAROL-PUR-0026), picked up after Albus's asks (CAROL-PUR-0027).

Sahil executes Project bank accounts and builds Midas's bank reading (6751) before 7034's paid check

Ruled 2026-11-11 by ninad -- superseded 2026-11-11

Ninad, 2026-11-12 (CLI-106), answering Sahil's operator communication #331, point 3: offered 'Use sample evidence', 'Orion builds bank step' and 'Sahil takes bank pursuit' (Sahil is assigned the bank-accounts pursuit and builds it himself), he chose 'Sahil takes bank pursuit'. READ AS (Orion's reading, not his words): Ninad, the pursuit's owner, assigns Project bank accounts (CAROL-PUR-0010) to Sahil to execute (#663); 6751 -- left out of the pursuit's steps while parked, so the pursuit read 100% done -- returns to it as its fourth step, and Sahil unparks it when he starts; 7034's paid check is proven on the bank's real evidence once 6751 lands, never on stand-in evidence (offered, not chosen).

Ended because: Corrects Orion's reading in #797, which said 6751 was missing from the pursuits; it sits in CAROL-PUR-0017

Sahil executes Project bank accounts and builds Midas's bank reading (6751) before 7034's paid check

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), answering Sahil's operator communication #331, point 3: offered 'Use sample evidence', 'Orion builds bank step' and 'Sahil takes bank pursuit' (Sahil is assigned the bank-accounts pursuit and builds it himself), he chose 'Sahil takes bank pursuit'. READ AS (Orion's reading, not his words): Ninad, the pursuit's owner, assigns Project bank accounts (CAROL-PUR-0010; its three steps closed, none of its five criteria met yet) to Sahil to execute (#663), so its outcome is Sahil's to carry to the end; 6751 itself already sits in the Blueprint service pursuit (CAROL-PUR-0017, post-MVP, parked), which Sahil executes, and stays there (an initiative belongs to one pursuit, #744) -- it serves both pursuits and Sahil unparks it when he starts; 7034's paid check is proven on the bank's real evidence once 6751 lands, never on stand-in evidence (offered, not chosen).

The administrator sits with Orion and the estate operators sit with Elrond

Ruled 2026-11-14 by ninad -- in force

THE ADMINISTRATOR SITS WITH ORION AND THE ESTATE OPERATORS SIT WITH ELROND (Ninad, 2026-11-14, CLI-137).

ASKED: "estate operators sit with elrond or orion?" -- shown Orion (Deputy Chairman, governance, every operator's instance) against Elrond (Head of Engineering, the platform), and that #488 maps estate operators by default to each project's CEO and head of finance for cost. HIS WORDS: "admin --> Orion, estate operator --> Elrond".

READ AS (Orion's reading, not the administrator's words): of the two founder roles of the lean year (#791), the ADMINISTRATOR (relationships and financial planning) sits with Orion, and the ESTATE OPERATORS (who oversee the estate's activities) sit with Elrond. This is where they sit, not where their cost falls: the founders' and estate operators' cost stays spread equally across the four categories (#847), and #488's default cost mapping to the CEO and head of finance stands. The admin person hired for the lean year (#848) stays in compliance and admin under Themis, the work running through Rhea's operations (#850).

Every operations person maps to an agent, service operators sit in Elrond's team, and Clara, Carol, Athena, Themis, Midas, Leo and Sam get theirs first

Ruled 2026-11-14 by ninad -- superseded 2026-11-14

EVERY OPERATIONS PERSON MAPS TO AN AGENT, SERVICE OPERATORS SIT IN ELROND'S TEAM, AND CLARA, CAROL, ATHENA, THEMIS, MIDAS, LEO AND SAM GET THEIRS FIRST (Ninad, 2026-11-14, CLI-137).

ASKED: shown the operations staff ramp of version E (one per service a year after its operator, #755, #840) and an order mapping them to the services' owning agents. HIS WORDS: "every ops has to map to an agent. these are taken: admin --> orion, estate operator --> elrond. all others will have a ops team mapped, we need to identify who needs to get mapped when.."; "urgent are clara, carol, athena, themis, midas, leo, sam. all service operators will belong to elrond's team (estate operator's team)".

READ AS (Orion's reading, not the administrator's words): (1) every operations person is mapped to ONE agent, the owner of the service they run (one per service, #755); Orion's and Elrond's places are the administrator's and the estate operators' (#851). (2) SERVICE OPERATORS, who build the services, all sit in Elrond's team, the estate operators' team. (3) The first operations people go to Clara, Carol, Athena, Themis, Midas, Leo and Sam, ahead of version E's order; the other agents follow by the order shown (client-facing first, then services that grow with clients and staff, then internal ones). When the seven join, and at which cost, is shown to Ninad with the rebuilt budgets before it is recorded (#766, #86).

Ended because: Ninad, CLI-137: "heimdall too" -- Heimdall joins the first operations people

Every operations person maps to an agent, service operators sit in Elrond's team, and Clara, Carol, Athena, Themis, Midas, Leo, Sam and Heimdall get theirs first

Ruled 2026-11-14 by ninad -- in force

EVERY OPERATIONS PERSON MAPS TO AN AGENT, SERVICE OPERATORS SIT IN ELROND'S TEAM, AND CLARA, CAROL, ATHENA, THEMIS, MIDAS, LEO, SAM AND HEIMDALL GET THEIRS FIRST (Ninad, 2026-11-14, CLI-137).

ASKED: shown the operations staff ramp of version E (one per service a year after its operator, #755, #840) and an order mapping them to the services' owning agents. HIS WORDS: "every ops has to map to an agent. these are taken: admin --> orion, estate operator --> elrond. all others will have a ops team mapped, we need to identify who needs to get mapped when.."; "urgent are clara, carol, athena, themis, midas, leo, sam. all service operators will belong to elrond's team (estate operator's team)"; shown the order with seven first: "heimdall too".

READ AS (Orion's reading, not the administrator's words): (1) every operations person is mapped to ONE agent, the owner of the service they run (one per service, #755); Orion's and Elrond's places are the administrator's and the estate operators' (#851). (2) SERVICE OPERATORS, who build the services, all sit in Elrond's team, the estate operators' team. (3) The first operations people go to Clara, Carol, Athena, Themis, Midas, Leo, Sam and Heimdall, ahead of version E's order; the other agents follow by the order shown (client-facing first, then services that grow with clients and staff, then internal ones). When the seven join, and at which cost, is shown to Ninad with the rebuilt budgets before it is recorded (#766, #86).

Operators run the company first; operations staff join on version E's ramp in a set order of agents, subject to the right skills being available

Ruled 2026-11-14 by ninad -- in force

OPERATORS RUN THE COMPANY FIRST; OPERATIONS STAFF JOIN ON VERSION E'S RAMP IN A SET ORDER OF AGENTS, SUBJECT TO THE RIGHT SKILLS BEING AVAILABLE (Ninad, 2026-11-14, CLI-137).

ASKED: shown the eight agents urgent for their operations person (#853) and three options for when they join and at what pay. HIS WORDS: "rampup plan remains the same, but order needs to be decided.. initially, the operators will run the company, later the operations staff will get onboarded.. the sequence is subject to availability of the right skills for the roles".

READ AS (Orion's reading, not the administrator's words): (1) version E's operations ramp stands -- 2 in 2028/29, 2 in 2029/30, 4 in 2030/31, 8 in 2031/32, 16 in 2032/33, 22 in 2033/34, then three more a year (#755, #840); until a service's operations person joins, the estate and service operators run it. (2) The ORDER, the eight of #853 first, each wave with its basis: 2028/29 -- Leo (Blueprint) and Sam (Customer Support), as clients grow from 6 to 18; 2030/31 -- Themis (Audit & Compliance) and Heimdall (Security), as EU clients and EU compliance return after the Seed and NIS2 nears; 2031/32 -- Midas (Cost Center), Athena (Agent and Human Resources), Carol (Communications) and Clara (Governance), after Series A as finance, hiring and self-serve clients grow; 2032/33 -- Aurora (Sales), Loki (Marketing), Galadriel (Agent Chat), Hagrid (Infrastructure & Backups, and System Services), Hermione (Process Monitoring), Jarvis (Operations) and Noah (Migration); 2033/34 -- Prometheus (Quality Management), Obi-Wan (Research), Bilbo (Blogs), Clara (Leadership) and two new services; from 2034/35 the three a year go to services launched by then. (3) SKILLS FIRST: a slot goes to the next agent in order whose role has a candidate with the right skills; an agent skipped for want of one keeps its place at the head of the next wave. This is an assumption in the register (#845) with its risk: the right skills not found in time keeps the operators running a service longer.

Root keeps the master list of legitimate Carolverse agents; agents are rented to projects, never owned by clients; released agents go to the bench, which Athena keeps and deploys from

Ruled 2026-11-15 by ninad -- in force

the masterlist of all approved carolverse agents should be maintaned under root, and creating a new agent needs admin approval, which would also involve updating the master list with the agent if approved as a legitimate carolverse agent. The admin needs basic information like the client, the account, the project, the agent name, description, designation in the project, role and responsibilities.. any changes to these need admin approval and update to the root records.. remember, all carolverse agents are rented to the project, but the clients dont own them, they continue to be part of carolverse agent community, even if they are deployed on the project or created for the project. If the project dissolves, or no longer needs these agents, the agents sit on "the bench" (add this to the dictionary) and are re-designed and re-deployed to another project, by Athena.. Athena stays on top of the bench, so she can decide if she would like to re-deploy the agent on the bench, or create a new agent. Re-deploying agent from the bench is done by athena by selecting the agent that is the closest match to the requirements of the project. Fold these into the blueprint requirements

READ AS (Orion, apart from the words): root holds the ONE master list of legitimate Carolverse agents (like root's register of projects, security #514); an agent is legitimate only once the administrator approves it and it is on that list with client, account, project, name, description, designation in the project, and role and responsibilities; any change to those needs his approval and a root update. Agents are Carolverse's and RENTED to projects. Released agents go to the bench (the existing bench of agent-resources #29-#32 grows this meaning); Athena keeps it and, for a project's need, re-deploys the closest-matching bench agent re-designed, or creates a new one. 'The bench' joins the dictionary in words the administrator approves (P.01.01.05.09).

Athena keeps about ten approved agents on the bench, ready to deploy to a project on the administrator's approval

Ruled 2026-11-17 by ninad -- superseded 2026-11-17

ASKED (CLI-207): Sahil (#1322) asked that Chowpatty's two weekly reports (Weekly Accounts and Deliveroo, Python scripts on his laptop) run on Sandbox every Monday with pages operators and Chowpatty's people can see, and that root set up a non-person runner login, a served page behind a sign-in, and a place for two keys (Airwallex, Gmail). Orion told Ninad no policy forbids a POC app Chowpatty's people use, and named what bears on it: #983 (Sandbox is a preview, not where the project runs), real keys in a POC area every operator can read, the Creed's ruling that operators' scheduled jobs run under non-person logins (CLI-546), and that only root makes a login on Sandbox's machine. HIS WORDS, in order: "Weekly reports on real data can also be read as a preview of the art of possible.. if one report is demonstrated, then all reports can be built on the actual project.. however, i dont know if root is necessary for that, as this is not related to operator, project or agent governance, or is it?"; "just like carolverse has users, chowpatty can also has users. Can operator's own login be used for scheduling the refresh?"; "ideally, chowpatty should have its agent built who will run the jobs, isnt it?"; "POC is a free form area for every operator to do whatever they want. Can operators (FDEs assigned to that prospect) not create their own agents and then use their logins to create scheduled jobs?"; then: "lets park that for now.. I think FDEs should be able to create reports in POC zone, but regular refresh should happen in the build area through named agents following the standard process (via Leo). when Sahil creates the agents, who will ask for the admin approval to admin Orion instance? - Leo, Elrond, Athena, Orion operator instance. Will that agent have a mind, and azure id? would it be possible that Athena creates about 10 agents upfront in "bench" who can be deployed immediately to the projects on admin's approval? who names the agents - option 1 - Athena when agent is created? option 2 - Leo in consultation with client? option 3 - Athena when agent is created, and Leo gives an alias based on client's preference?". Orion answered from the law (Athena asks him, #544; an Azure identity always, P.01.03.08.01, a mind only if he decides; the bench exists, #32, #896, each creation and each deployment his approval) and recommended option 3 with a bench of ten, and recording the report rule. HIS WORDS: "sounds good".

READ AS (Orion's reading, not the administrator's words): (1) Athena creates about ten agents ahead of need and keeps them on the bench (#32, #896), each approved by the administrator, named by her, on root's master list, with its Azure identity (P.01.03.08.01) and no project; whether one has a mind is his decision, as for any agent; (2) when a project needs an agent, Athena deploys the closest-matching bench agent, re-designed for that project, on his approval (#544) and with root's list updated (#896), so a project waits on an approval, never on a creation; (3) she tops the bench back up to about ten, each new agent again on his approval.

Ended because: Ninad, CLI-207: Athena backfills a deployed agent's place, keeps the bench close to ten and asks the administrator's Orion instance for help; approval for deployment as well as creation

Athena keeps about ten approved agents on the bench, ready to deploy to a project on the administrator's approval

Ruled 2026-11-17 by ninad -- in force

ASKED (CLI-207): Sahil (#1322) asked that Chowpatty's two weekly reports (Weekly Accounts and Deliveroo, Python scripts on his laptop) run on Sandbox every Monday with pages operators and Chowpatty's people can see, and that root set up a non-person runner login, a served page behind a sign-in, and a place for two keys (Airwallex, Gmail). Orion told Ninad no policy forbids a POC app Chowpatty's people use, and named what bears on it: #983 (Sandbox is a preview, not where the project runs), real keys in a POC area every operator can read, the Creed's ruling that operators' scheduled jobs run under non-person logins (CLI-546), and that only root makes a login on Sandbox's machine. HIS WORDS, in order: "Weekly reports on real data can also be read as a preview of the art of possible.. if one report is demonstrated, then all reports can be built on the actual project.. however, i dont know if root is necessary for that, as this is not related to operator, project or agent governance, or is it?"; "just like carolverse has users, chowpatty can also has users. Can operator's own login be used for scheduling the refresh?"; "ideally, chowpatty should have its agent built who will run the jobs, isnt it?"; "POC is a free form area for every operator to do whatever they want. Can operators (FDEs assigned to that prospect) not create their own agents and then use their logins to create scheduled jobs?"; then: "lets park that for now.. I think FDEs should be able to create reports in POC zone, but regular refresh should happen in the build area through named agents following the standard process (via Leo). when Sahil creates the agents, who will ask for the admin approval to admin Orion instance? - Leo, Elrond, Athena, Orion operator instance. Will that agent have a mind, and azure id? would it be possible that Athena creates about 10 agents upfront in "bench" who can be deployed immediately to the projects on admin's approval? who names the agents - option 1 - Athena when agent is created? option 2 - Leo in consultation with client? option 3 - Athena when agent is created, and Leo gives an alias based on client's preference?". Orion answered from the law (Athena asks him, #544; an Azure identity always, P.01.03.08.01, a mind only if he decides; the bench exists, #32, #896, each creation and each deployment his approval) and recommended option 3 with a bench of ten, and recording the report rule. HIS WORDS: "sounds good".

READ AS (Orion's reading, not the administrator's words): (1) Athena creates about ten agents ahead of need and keeps them on the bench (#32, #896), each approved by the administrator, named by her, on root's master list, with its Azure identity (P.01.03.08.01) and no project; whether one has a mind is his decision, as for any agent; (2) when a project needs an agent, Athena deploys the closest-matching bench agent, re-designed for that project, on his approval (#544) and with root's list updated (#896), so a project waits on an approval, never on a creation; (3) she tops the bench back up to about ten, each new agent again on his approval.

AMENDED (Ninad, CLI-207, 2026-11-18), before it: "Note - Admin approval is required not just for creating agents but also deploying agents to project"; then, HIS WORDS: "good, reply sahil, and put these steps on priority of the blueprint pursuit because Sahil will need agents soon.. if the agents are not available, the FDEs will have to wait for admin to create the agents. Athena ensures that the bench does not dry up by backfilling the spot with a new agent when an agent is deployed.. when an agent returns to the bench, for whatever reason, the bench may increase.. its ok for the bench to increase or decrease, but athena should strive to keep it close to 10 agents, and should communicate to admin orion instance with any help she may need in this regard.."

READ AS NOW (Orion's reading, not the administrator's words): (1) the bench exists so that an FDE never waits for an agent to be created; (2) the administrator approves every creation and every deployment (#544, #896); (3) when an agent is deployed, Athena backfills its place on the bench with a new agent, on his approval; (4) an agent that returns to the bench, for whatever reason, stays there, so the bench may grow or shrink; Athena strives to keep it close to ten; (5) Athena tells the administrator's Orion instance, by an operator communication, whenever she needs his help to keep it so -- an approval waiting, a creation refused, the bench running low; (6) the Blueprint pursuit puts the agent steps first, because Sahil's FDE work needs agents soon.

Leo and Orion may ask Athena for a new agent, and Athena may ask for a bench agent; a bench agent stands on root's list under Carolverse, designated bench

Ruled 2026-11-17 by ninad -- in force

ASKED: "Athena's bench (7433): the bench is empty today, and root's master list of agents refuses any agent without a project. Your rule (#896) says agents belong to Carolverse and are rented to projects. So a bench agent could be listed under Carolverse, as 'bench' and reporting to Athena, until you approve its deployment. That avoids changing root's list-writer, though each agent still needs a core step to go on the list. Also, today only Leo and Orion may ask Athena for an agent (#35); for her to backfill on her own, she'd need to be allowed to ask too. Is that how you want it?" Options: 'Yes, both (Recommended)' -- bench agents sit on root's list under Carolverse as 'bench', reporting to Athena; Athena may ask for bench agents herself (amends #35); every creation and deployment still needs your approval / 'List under Carolverse only' -- Athena tells your Orion when the bench is low and Orion asks for her / 'A no-project entry' -- change root's list-writer, a root step. HIS CHOICE: "Yes, both (Recommended)".

READ AS (Orion's reading, not the administrator's words): (1) Leo and Orion may ask Athena for a new agent, as #35 said, and Athena may ask for a BENCH agent herself, to keep the bench close to ten (#988); nobody else asks; (2) a bench agent's mandatory details (#30) are the bench's own: Carolverse's, designated 'bench', reporting to Athena, named by her (#986); (3) it stands on root's master list under Carolverse with that designation until its deployment is approved, when root's list takes its project, designation and role (#896); (4) every creation and every deployment still needs the administrator's approval from his own session (#544, #988), and every entry on root's list is still a core step.

REUSING A SERVICE IS NOT REUSING AN AGENT: A CLIENT'S OWN BUSINESS IS DONE BY ITS OWN BENCH AGENT, THE MACHINERY IS LOOKED AFTER BY CAROLVERSE'S OWN AGENTS; AND A BACKUP MISTAKE IS KEPT FROM REACHING EVERY PROJECT

Ruled 2026-11-17 by ninad -- in force

THE QUESTIONS: after ruling support #990 (a client's customer support is done by a bench agent of its own, running Sam's service), Ninad asked whether the same held for the investment plan -- "would clara be the investment planner for chowpatty or will chowpatty get its own investment planner?" -- and Orion recommended Chowpatty's own planner, Clara keeping the planning service, its principles and templates. He then asked: "this would mean, apart from the pipeline (initiatives) no other sevice is used as is? does that mean chowpatty ends up having 40 agents? for eg - can hagrid do the backup of chowpatty like he does for carolverse, or would we give chowpatty a new agent for that?". Orion answered that the services handbook grades 4 tracks as-is (all the build pipeline), 52 config-only, 15 estate-bound and 9 seam-refactor, and that config-only gives a project its own copy of a service's code, not its own agent; Hagrid's backup is config-only, his routines run for Chowpatty and he stays the agent; and proposed the line: looking after the machinery (backup, machines, the build pipeline, monitoring, security, costs) is done by Carolverse's own agent for every client, holding data only as its custodian; doing the client's own business (customer support, investment planning, reports and accounts) is done by the client's own bench agent running Carolverse's service under the client's alias -- so a client has one agent per business function it subscribes to, not forty. Ninad asked: "so re-using a service is distinct from re-using an agent.. or to put it differently, for customer facing roles, new agents are created, while for backend roles, existing agents do the work? what if hagrid makes a mistake? all projects backups will get impacted?". Orion answered yes to both readings; that a separate agent per client would not prevent it, since they would run the same code; named what limits it -- every change reviewed and tested before it goes live (exists), an off-site copy kept apart (exists), each project's backups run and stored apart (follows from config-only), a change going live on Carolverse first and reaching clients only after it ran cleanly there (proposed), regular restore drills per project (proposed) -- and asked whether to record the line and file the two proposed safeguards under Hagrid's backup service.

HIS WORDS: "yes"

READ AS (Orion's reading, not the administrator's words): (1) reusing a service means reusing its code; reusing an agent means the same agent does the work, and the two are decided apart; (2) work that does a client's own business or faces its people and customers (customer support, investment planning, reports and accounts, and the like) is done by that client's own agent, deployed from Athena's bench on the administrator's approval under the client's alias (#986, #987), running the Carolverse service whose owner keeps it -- support #990 is the first case, and the investment plan is done the same way, Clara keeping the planning service; (3) work that looks after the machinery (backup, machines, the build pipeline, monitoring, security, costs) is done for every client by Carolverse's own agent, who holds a client's data only as its custodian and never acts on its contents; (4) a mistake in a shared service is kept from reaching every project: a change goes live on Carolverse first and reaches clients only after it ran cleanly there, and each project's backups are proven by regular restore drills -- the two safeguards are filed under Hagrid's backup service.

Athena fills her own name book: she proposes bench agents' names with her own model when it runs low, and each reaches the administrator with its request

Ruled 2026-11-17 by ninad -- in force

ASKED: "Athena now checks the bench every 10 minutes, and her live run shows 10 missing. She can't ask for any agents yet because her name book is empty: your rule (#986) says she names every agent, but her bench check uses no AI model, so it can't invent names. Each name still comes to you with the request you approve. How should she get names?" Options: 'Athena proposes them (Recommended)' -- a small new task: when her name book runs low, Athena uses her own AI model to propose names in Carolverse's style (myth and fantasy, like Elrond or Themis) and adds them herself; a few cents a time; you see each name when you approve that agent / 'You give her a list' / 'Orion seeds a themed list'. HIS CHOICE: "Athena proposes them (Recommended)".

READ AS (Orion's reading, not the administrator's words): (1) Athena fills her own name book: when it holds fewer unused names than the bench is short, she proposes names with her own model, in Carolverse's style (names from myth, legend and fantasy, as the estate's agents carry), each unused by any agent or request, and adds them through the reference-data door as hers; (2) the bench watch itself stays model-free; the naming is its own small task, budgeted on her track; (3) every name reaches the administrator with the request he approves (#544), so a name he dislikes is declined with the request.

An agent is identified by its id; its name is an attribute admin and estate operators change at once

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-218, operator communications #1503-#1512): 'Athena has just asked for 10 new agents for her bench: Perseus, Theseus, Apollo, Artemis, Freya, Frigg, Baldur, Hera, Hermes and Hestia. A bench agent sits under Athena, works on no project, and only goes to a project after you approve that move separately. Each one needs your approve or decline, recorded in your words. What should I record?' Ninad answered: 'Approve all 10, but note that the primary identification of the agent is the agent id, with agent name serving as an attribute, which can be changed easily and instantly by admin and estate operators. This is applicable for all agents of carolverse, not just the bench agents. If this is not the case, that needs to be fixed first'. READ AS: every Carolverse agent is identified by its agent id; its name is an attribute the administrator and the estate operators can change at once; where the estate does not yet work that way it is fixed before the ten bench approvals are recorded at Athena's door.

Root's master list of approved agents keeps both the agent's id and its name

Ruled 2026-11-18 by ninad -- superseded 2026-11-18

Asked: root's master list of approved agents also stores each agent's name, so after a rename it keeps the old one -- should that list keep only the agent's id and take the name from the registry, or keep the name, so every rename also needs a step at the root login? Ninad: "keep both id and name". READ AS (Orion, not Ninad's words): the master list keeps its name field beside the id; a rename by the act is complete only when root's list carries the new name too, which is a root step the administrator runs, and the estate shows a list name that disagrees with the registry until he does.

Ended because: Ninad chose between the two readings of 'keep both id and name' when asked again (CLI-239)

Root's master list of approved agents keeps both the agent's id and its name

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-239): root's master list of approved agents also stores each agent's name, so after a rename it keeps the old one -- keep only the id and take the name from the registry, or keep the name, so every rename also needs a step at the root login? Ninad: "keep both id and name". Asked again, because had built the list to keep the name an agent was APPROVED under, as history, a rename never touching it: approved-under name, or name follows renames (list keeps id + current name; every rename also needs a root step he runs, and the board flags names out of step until he does)? Ninad: "Name follows renames". READ AS (Orion, not Ninad's words): the master list keeps the id and the agent's CURRENT name; a rename is complete once the administrator's root step carries the new name to it; the board names any list name out of step with the registry, by id; 7462's 'the list keeps the name it was approved under' is retired in place.

Sage is the data quality lead

Ruled 2026-11-19 by ninad -- in force

Asked whether to hire a new data quality lead or give the role to Sage or Sentinel, Ninad: "I prefer Sage, because sentinel's job is more user facing, and associated with agent chats". The data quality lead also owns the data catalog and data traceability services (governance ruling of the same sitting).

An agent owns tracks only within its own line of the org structure

Ruled 2026-11-19 by ninad -- in force

Ninad: "Sage can own the tracks only if they are within Elrond's team.. the skill that allocates agents to services should take into account the org structure and all policies governing it"

The skills that define agents, services and their associations apply the org structure and its policies, for the estate and for every project org Leo builds

Ruled 2026-11-19 by ninad -- in force

Ninad: "ensure that the skills for allocating agents to services are updated, because Leo will use the same skills to build project orgs for other projects, and he should take into account org structure and policies when defining agents, services and associations" (follows agent-resources #1055).

Midas, Head of Finance, reports directly to Clara as CFO; his functions stay the same

Ruled 2026-11-20 by ninad -- in force

Ninad, 2026-11-20 (CLI-259). Asked "who is the CFO of carolverse?", shown that Midas is Head of Finance (level 3, reporting to Cassius, Head of Support Functions) and that at Wipro, TCS and Infosys finance is classed as a support function while the CFO reports to the CEO, HIS WORDS: "file an initiative for this, and give me the name of the pursuit.. Midas reports to clara directly, keeping all his functions same". READ AS (Orion's reading, not his words): Midas is Carolverse's CFO and reports directly to Clara, on the leadership rung beside Rhea, Cassius, Odin and Aurora; his level follows from that line (never typed); every function he holds today -- company expenses, accounts, the bank account, the daily track budgets, Procurement with Marco reporting to him -- stays exactly as it is; investor relations and fundraising stay Clara's (the finance track exclusive to Clara is unchanged); Cassius keeps Support Functions without Midas reporting to him.

Every agent reporting directly to Clara carries the designation Director

Ruled 2026-11-20 by ninad -- in force

Ninad, 2026-11-20 (CLI-261). Shown that after Midas's move under Clara (#1078) her five direct reports read Director (Rhea, Odin), Associate (Cassius, Aurora) and Senior Manager (Midas), HIS WORDS: "all 5 who report to clara should be directors". READ AS (Orion's reading, not his words): the designation (the registry's level_title) of every agent whose reporting line is Clara reads Director -- today Rhea, Odin, Cassius, Aurora and Midas; Aurora, Cassius and Midas change, nothing else about them does (title, level, line, service, functions). It narrows #43 (designations roughly tied to levels, a guide) for Clara's direct reports and does not supersede it.

A role prompt holds the role's method; who the agent is is configured per agent

Ruled 2026-11-22 by ninad -- in force

Asked, for the prompt catalogue, which part of Carol's prompt counts as 'the sales prompt' -- her sales method (warm up, get to know the person, pitch only when they show a need) or her whole prompt, which also holds facts only hers (her name and number, Carolverse's services, running Ninad's diary) -- Ninad: "sales methods, and other details should be configurable to the respective agents.. for example.. agent name or their role is the org, who they work for etc.. each agent should get their own description..". READ AS: a role's catalogue entry holds the role's METHOD only (for sales: the warm-up, discovery and pitch-on-signal flow); who the agent is -- its name, its role in its organisation, the organisation and whom it works for, what it offers -- is configured per agent from its own records, each agent with its own description; a catalogue entry never carries one agent's facts, and an agent's facts never live in the catalogue.

's alias criterion moves to its own post-MVP initiative

Ruled 2026-11-22 by ninad -- in force

Ninad, 2026-11-22 (CLI-327), was asked: "7434's second criterion says a project's people see the agent under an alias Leo agrees with the client (Athena's name stays on root's list and in Azure). MBank's MVP doesn't need an alias. Should I build it inside 7434 now, or move it to its own post-MVP initiative so 7434 can close on the approval part?" HIS ANSWER, choosing the option "Own initiative (Recommended)", which read: "File the alias as a separate post-MVP step in the MBank pursuit; strike it from 7434 in your words; 7434 goes to Elrond now." READ AS (Orion's reading, not his words): is graded on its first criterion alone (no agent is deployed to a project without the administrator's approval from his own Orion session, and each deployment updates root's master list with the agent's project, designation and role); its second criterion (the alias Leo agrees with the client, shown to the project's people, recorded with who agreed it, root's list and Azure keeping Athena's name) is struck from 7434 by a governed goal change citing this ruling and filed as its own initiative, a post-MVP step of CAROL-PUR-0093 (MBank POC). Blueprint #986 is unchanged: the alias is still owed.

๐Ÿ“–Audit & Compliance

Palantir story writing is tracked on its own audit track

Ruled 2026-09-16 by ninad -- in force

The cost of writing Palantir stories is tracked on a separate audit track (audit-palantir-stories), whichever lane's droid composes the story.

Themis knows and checks the law, the AI providers' agreed terms and the company's own privacy policy

Ruled 2026-10-14 by ninad -- in force

THEMIS KNOWS THE LAW, THE AI PROVIDERS' AGREED TERMS AND THE COMPANY'S OWN PRIVACY POLICY, AND CHECKS THE ESTATE AGAINST THEM (Ninad, CLI-765, 2026-10-14). ASKED: nothing; given right after Google verified the denken-labs app's branding on the new privacy policy page (communications #372). HIS WORDS: "ensure that themis is aware of the privacy policies, regulatory policies and agreed policies of the llm providers and checks the compliance against them. Build an arc for this, to 1. identify regulatory policies as applicable to denken-labs 2. check them regularly to keep them updated and look out for new compliance requirements 3. verify the estate for compliance 4. identify and report the issues via escalation log dedicated for complaince checks 5. fix the issues on an ongoing basis via pipeline, including both albus or orion bypass lanes".

READ AS (Orion, CLI-765, not his words): (a) 'privacy policies' are Denken Labs' own published ones (https://denken-labs.com/privacy, including its Google API Limited Use commitment) and the AI providers' privacy and data-use terms; 'regulatory policies' are the laws and regulations that apply to Denken Labs and its estate where it is based and where the people it serves are; 'agreed policies of the llm providers' are the usage policies and terms of every model provider the estate calls, as agreed under the plan or account actually used. (b) Themis holds the register, the watch and the checks in her Audit service. (c) The dedicated compliance escalation log is how a compliance-check issue is raised; for those issues it narrows cookbook 1573 (an escalation is an Agentbook message to Orion), the later ruling of the same operator governing (P.01.01.05.08). (d) Each issue is requested as an initiative that Elrond's seam files: Themis's requests land on Albus's lane (cookbook 1765), work his lane cannot do goes to Orion's (initiatives #230), and an issue reads addressed only when its fix has closed and Themis's re-check passes. Checking only internal rules, a register of laws with no provider terms, raising compliance issues only as Agentbook messages, or Themis fixing what she found, each undo it.

Themis owns risk management, as a track of its own beside her compliance duties

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-070), asked who should own the risk management function ruled in leadership #696 (Orion recommended Forseti, Head of Governance, Risk & Compliance, with Themis the alternative), answered: "Themis can be responsible for risk management function as a seperate track apart from her existing compliance duties". READ AS (Orion's reading, not his words): Audit & Compliance, Themis's service, gains a risk management track owned by Themis, separate from its compliance tracks; the plan's risk management cost line (#696) is that track's; Forseti's security governance, risk and compliance work is unchanged.

Compliance findings on the retired clara-chat app are closed as not applicable, and the rest of goes back to Albus

Ruled 2026-11-08 by ninad -- in force

HIS CHOICE (CLI-077, 2026-11-09), asked: "To hand 2719 to Albus, I need one ruling from you: findings about the retired 'clara-chat' app, which no longer exists, are closed as not applicable, and the rest of the work goes back to Albus. Agree?" -- he answered "agreed". READ AS (Orion's reading, not his words): in, every finding whose only target is the clara-chat app -- absent from the registry, where only clara-monitor remains -- is closed as not applicable, target retired, each listed by finding with that reason and never silently dropped; the remaining findings are Albus's work and the initiative is revived to his lane with a fresh attempt, not to Orion's. This answers Albus's law block on and replaces Orion's CLI-068 note that sent it to Orion's lane: initiatives #230 moves work to Orion only when Albus's lane cannot do it, which this is not.

Every operator's Orion instance stays in step with the others: a difference is a compliance violation that Themis finds regularly, files as an initiative and tells the operator about

Ruled 2026-11-08 by ninad -- in force

HIS WORDS (CLI-075, 2026-11-09), after Orion compared his laptop's Orion with the operator template Sahil's and Saee's computers run and found 7 differences that were gaps (the workstation report's result hidden, a failed credential check not stopping work, the hourly check registered only for the personal login and cut at 3 seconds, hand-copied hook and transcript code, a deleted chat's missing-record finding never cleared, the session key recorded by a separate step): "1. this is compliance violation, ensure that themis regularly checks if the operator instances are in synch, and files initaities for inconsistencies and notifies operator about it 2. make all the operator instances consistent.." READ AS: an operator's Orion instance that differs from the others -- beyond a difference a ruling declares for that operator's role -- is a compliance violation; Themis checks every operator's instance regularly, files an initiative for each inconsistency she finds, and tells that operator; and every operator's instance is brought into step now.

The company's three existential risks are measured, acted on and reported by Themis, in a risk management arc

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-100/CLI-102, the same chat), shown Orion's list of existential risks and moats (#763): "I think the 3 existential risks are 1. competition - too many players providing similar service 2. security - prompt injection making agents incompetent or incapacitated 3. supply chain - scarcity of raw ingredients like tokens, storage and RAM please let me know if you agree"; Orion agreed, suggesting regulation stay watched; then "done, enable themis to measure, act and report against these risks. Create an arc for risk management". READ AS (Orion's reading, marked apart from his words): these three are the company's existential risks; regulation is not a fourth, it stays watched under Themis's existing compliance with the law (roadmap item 115). Themis keeps a register of the three, each with its measures, the thresholds at which it rises, its moat and the service that owns the moat; she measures them on a schedule, acts when one rises (files the moat's work to the service that owns it and tells the operators), and reports them. The thresholds are proposed by Themis and approved by Ninad before they bind (#86). The arc is Audit & Compliance's, owned by Themis, on a new roadmap item; the moats themselves are built in the arcs of the services that own them (initiatives #385).

๐Ÿ“–Blogs

Bilbo writes the stories the administrator names, through a writer of his own

Ruled 2026-10-22 by ninad -- in force

Ninad, 2026-10-22 (CLI-912). HIS WORDS: "you pick the first 2 points, and the third point, i will tell what stories to write.. bilbo needs a similar writer droid like orion and scriber". Answering Orion's plan after rulings #496 and #497: (1) hand Carolopedia to Loki, (2) put Carolopedia's links back, (3) Bilbo tells the story of Leo's work, the first being Chowpatty's setup.

READ AS (Orion's reading, not the ruling): Orion does points 1 and 2 on his own lane. Bilbo gets a writer of his own, built like Orion's Author and Scriber's Author; the stories he writes are the ones the administrator names -- Bilbo does not pick his own subjects, and nothing writes a story the administrator has not asked for. This narrows the reading of #497 on who chooses the stories.

๐Ÿ“–Blueprint

The Dominion, and where it sits in the hierarchy

Ruled 2026-09-14 by ninad -- superseded 2026-10-20

A DOMINION is the collection of projects that share one underlying build pipeline โ€” one whole installation of the platform. The dominion the estate runs in today is EDEN. It sits BETWEEN account and project: client -> accounts -> dominions -> projects. A client may hold several accounts; an account may hold one or more dominions; a dominion has exactly ONE client, reached through its account; and every project belongs to exactly ONE dominion. A copy of a project into another dominion is a SEPARATE project, which may later be renamed. A dominion is a COPY, not a shared deployment โ€” Ninad's words: like an installation of Windows on a separate machine โ€” and it carries its own copy of EVERYTHING: records, money, budgets, access, agents and projects. Each dominion holds a Carolverse-type BASE PROJECT whose name, services and agent names may all differ from the dominion it was copied from. AGENT NAMES ARE UNIQUE WITHIN A DOMINION, never across dominions. This does NOT supersede the service-instantiation ruling (design 557, P.01.03.07.08): one shared codebase never forked governs services WITHIN a dominion. A dominion is itself a fork by design, and collapsing the two would remove the independent evolution that is the point of the concept.

Ended because: Ninad ruled 2026-10-21 (CLI-883, answering Sahil's #96) that Eden hosts projects of several clients; #4's one-client clause is retired in place and the rest restated in the successor

Dominion versions, forking, and the lineage every ancestor keeps

Ruled 2026-09-14 by ninad -- in force

A dominion is VERSIONED, so that a specific RECORDED version can be forked into a new dominion which then evolves independently from there. A fork may only copy a version the parent actually recorded โ€” forking whatever the parent happens to say today is refused, because the thing copied must be nameable afterwards. EVERY ANCESTOR UP THE CHAIN RECORDS EACH DESCENDANT. Eden therefore holds a record of every dominion ever descended from it, including those forked from its own forks, at any depth. This is why lineage is kept as a closure record written at the moment of the fork, not as a parent pointer walked at read time: a parent pointer answers 'who made me', and the ruling asks the opposite question, of every ancestor.

Where a project's roadmap app and requirements app run

Ruled 2026-09-17 by ninad -- in force

A new project's roadmap app and requirements app run ESTATE-SIDE under Leo: project-scoped instances inside Leo's Blueprint service, like Leo's chat itself. They exist before the user approves the roadmap and before the project's own machine exists, so the user reviews the plan in their own apps and no machine cost is spent before the user has approved the costs. In BLOOM they belong to Blueprint's 'Define a workspace': standing up an instance of shared code is workspace definition, not a build.

Which BLOOM segment owns opening a project's apps to the client's stakeholders

Ruled 2026-09-17 by ninad -- in force

Opening the project's apps to the client's named stakeholders, through Heimdall, belongs to ORCHESTRATE: a delivered app is not delivered until the client's people can reach it, so access is the closing act of delivery. It is not Blueprint (enrolment at workspace definition) and not Manage (day-to-day operations).

Project onboarding runs on its own Blueprint track

Ruled 2026-09-17 by ninad -- in force

The experimentation track stays as it is.

Who builds for a project once it is onboarded

Ruled 2026-09-17 by ninad -- in force

A project never builds for itself. For the life of the project every build goes user -> Leo -> Elrond's Build Initiatives: Leo stays the point of contact for any build activity, and the project's own agents only do what they are there to do - the project's business work. Projects are not given autonomy to build for themselves; they keep using the Carolverse services, Build Initiatives included. Leo also re-derives the agent roster whenever the requirements change. The project still gets its own machine, its own rulebook, its own agents and its own human admin, who approves plans and rosters. This RETIRES, in place, the 2026-08-31 ruling that Leo builds the builders and hands off at a self-hosting gate, after which the project runs with its own build machinery and its own non-autonomous operator agent with Leo as the door only: there is no self-hosting gate, no project-side build pipeline and no project-side Orion. Ninad's words: 'I would prefer leo continues to be the point of contact for any build activity for the project, while the projects own agents only do what they are supposed to do.. the projects dont get autonomy to build for themselves as they continue to use carolverse services including build initiatives.'

What proves Leo has finished setting a project up

Ruled 2026-09-17 by ninad -- in force

One real run, both ways, proven by doing it and never by a checklist: the project's agents do one real piece of their own work on the project's own machine, AND one build request travels user -> Leo -> Elrond's Build Initiatives and lands on the project's machine. This is the exit of the onboarding manifest and replaces the retired self-hosting gate. Finishing setup ends nothing about Leo's role: he stays the project's contact for builds.

Who receives the law layer of the onboarding manifest

Ruled 2026-09-17 by ninad -- in force

The project's starter rulebook - its policies, review gates, budget gates, registry and its own front door for change - is handed by Leo to CLARA, who owns the Governance service, so a project's rulebook comes from the same home as the estate's.

Where a roadmap item's cost and time estimate comes from

Ruled 2026-09-17 by ninad -- in force

Each roadmap item carries its OWN cost and time estimate, produced by Leo's Initiative Sizer when the roadmap is drafted - not at filing, and never inside the chat turn that shows it. The figure is presented as LEO'S ESTIMATE in those words, and the MEASURED range of delivered initiatives on the lane that will build the work is shown beside it as the check, so the reader can see the estimate against what work of that kind has actually cost. Measured means the ledger's own per-initiative total; an item the sizer could not size, or a lane with no delivered history, reads UNMEASURED in words and never zero and never a median silently repeated. The sizing runs as Leo's droid work on the project onboarding track's module_sized task and is. Ninad chose this over showing every item the same measured range, and over classing items by kind of work (no such classification exists on delivered work today). The estate-median-times-entry-count figure the roadmap reply shows now is RETIRED by this ruling: one number repeated per line is not a per-item estimate.

Obi-Wan's architecture step binds new services in Leo's intake only

Ruled 2026-09-25 by ninad -- in force

Obi-Wan settles the architecture shape of a NEW service entering through Leo's intake before any initiative is written against it. That step does not bind operator-filed or ruling-driven initiatives that change an app he already owns: such a change owes him no design decision and is not a deviation from his charter, and no record may report it as a bypass of his step. WHAT LOOKS LIKE A FIX AND IS NOT: a filing-gate check demanding his design for every change to his apps; recording such a change as a waiver.

Sahil owns the blueprint arc and carries its work on

Ruled 2026-09-29 by ninad -- superseded 2026-11-06

SAHIL OWNS THE BLUEPRINT ARC AND CARRIES ITS WORK ON (Ninad, CLI-611, 2026-09-29). Asked by Sahil's Orion instance which arc he is handed, what he is to do with it and where to start, his words: 'sahil needs to continue to work on the arc, and own it up'. So: the operator who owns CAROL-ARC-0004 (Blueprint: Leo onboards a project) is Sahil, the deputy operator; he works its open initiatives through his own Orion session on the operator lane and answers for its progress. The arc's owning AGENT is unchanged: an arc is one service's family of work under that service's owner, and an operator's ownership is the human accountability beside it, as a pursuit carries an operator owner beside its agent. What is reserved to root and core stays the administrator's and reaches him as a root or core escalation (security #175).

Ended because: Ninad ruled the owner operator of an arc is its creator and chose 'I own, Sahil executes' (CLI-042, initiatives #651)

A project's agents come from its own requirements, never from an estate agent by default

Ruled 2026-09-30 by ninad -- in force

A client project's agents are derived from that project's own requirements. A project agent need not correspond to any Carolverse agent: the estate may have no agent for the function at all, and where a function broadly matches, the project's actual requirements can still be completely different, so the project would inherit code that is irrelevant to it. A project agent is therefore never, by default, the project's own version of the estate agent that owns the matching service, and it inherits nothing from an estate agent or service that its own requirements do not call for. Design 557's words naming the owner of a project's custom configuration 'the per-project identity of the serving agent' are retired in place by this ruling.

Chowpatty is the first client project built on Carolverse; the old Chowpatty records are discarded

Ruled 2026-09-30 by ninad -- superseded 2026-11-15

Chowpatty is the first client project built on Carolverse: client Chowpatty, account Chowpatty, project Sales & Customer Support, whose stated need is a sales agent, an admin agent and a customer support agent. The earlier Chowpatty records in the estate (the aspirational street-food ordering service, its business line, its blocks, its ordering tools, its order notifier and its order store) are discarded: retired in place or moved to Trash, never deleted, and never reused as the new project.

Ended because: Ninad, CLI-154: "Chowpatty is not a client yet, instead chowpatty is a prospect" (blueprint #868)

Project onboarding records every process and learning from each project, starting with Chowpatty

Ruled 2026-09-30 by ninad -- in force

Every process performed and every learning found while onboarding a client project is recorded as that project is onboarded, starting with Chowpatty, and the Project Onboarding app shows them, so that the onboarding of every later project can learn from the ones before it.

A project adopts an estate service only where one of its requirements fits it, never by default

Ruled 2026-09-30 by ninad -- in force

Whether a client project adopts a Carolverse service is decided requirement by requirement, from the project's own requirements. Where a service does what a requirement asks, the project adopts it; where a service does part of it, the project adopts it and the rest is built as the project's own extension; where no service fits, the capability is built for the project alone, still through Leo and Elrond's Build Initiatives. No service is adopted by default: a copy of a service the project does not need is a cost to it (unused code, upkeep, more to secure), not a benefit.

The project-build rules are guidelines in the Project Onboarding app that the onboarding code follows, and policy for every later build of a project

Ruled 2026-09-30 by ninad -- in force

The rules for building a client project (blueprint #242 onwards: every ruling on how a client project is built) are documented as guidelines in the Project Onboarding app, and the code that onboards a project is specifically instructed to follow them, like a cookbook, when it produces the project's initial template. The same guidelines are policy for Elrond's team and for every service when they build for a project on an ongoing basis, after onboarding as well as during it.

Leo registers each new project, Chowpatty included; operators act as forward deployed engineers through Leo's door until he onboards projects by himself through his chat

Ruled 2026-10-05 by ninad -- superseded 2026-11-11

Ninad, 2026-10-05 (CLI-723). Orion asked, carrying Sahil's operator communication #38: 'Who registers Chowpatty as a client and project? I'd have Leo do it through his normal intake.' He answered: "Yes, Leo should register chowpatty as client and project.. we have the operators like Sahil and Saee acting like FDE's (forward deployed engineers) to facilitate the initial onboarding of the projects, but the goal is to eventually make Leo capable of onboarding new projects by himself without the FDE's having to step in.. until Leo is ready on his own via his chat, Sahil and Saee will continue to assist him via their orion instance who will do the work on behalf of Leo via his door.. (hope i am getting the operating model right here)" READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): (1) LEO registers Chowpatty's client, account and project, as the Blueprint service registers every new project. (2) OPERATORS ARE FORWARD DEPLOYED ENGINEERS for project onboarding: the onboarding is Leo's and he stays its owner and accountable agent; an operator's Orion session performs his onboarding work ON HIS BEHALF THROUGH HIS DOOR -- running Leo's own registered processes, recorded under his name with the operator who acted -- until Leo can onboard a project by himself through his chat, which is the goal; the operators' part shrinks as he becomes able, and nothing they do is recorded as theirs instead of his. This agrees with the standing law: an operator acting as an agent runs that agent's own registered processes and nothing else (P.03.04.02.04), every build goes user -> Leo -> Elrond (cookbook 1778), and the Orion bypass may take up an agent's activity through its door with the agent still accountable. (3) A service operator does this only inside her assignments (security #317): the doors to an agent are granted to the estate operators' group (security #197), so a service operator acting through Leo's door inside an assigned arc or pursuit is not yet built. NOT settled here: who owns Chowpatty's pursuit (CAROL-PUR-0006).

Ended because: Ninad set the interim model: FDEs speak for the client, the project's estate operator steps in for Leo, and the Carolverse estate operator or Blueprint service operator enables him

A project's north star is the target state of every function it runs as an interim workaround, traced to its goals, and Blueprint keeps it for every project, Carolverse included

Ruled 2026-10-05 by ninad -- in force

Ninad, 2026-10-05 (CLI-723). Orion recommended (CLI-723): no separate app; the north star as one record per project, drafted by Leo at Blueprint, approved by the project's admin, shown at the top of the project's roadmap app, with every top-level roadmap item naming the part of the north star it serves; Carolverse's first. Ninad had asked: 'would you recommend creating a north star app for each project which will document the end vision of the project? or does Leo's blueprint already does that?' -- and Orion had answered that Blueprint names the step ('Define the project: objective, approach, value') and that it was never built. He answered: "1. carolverse northstar needs to be defined 2. north star in not one goal or objective, but a target state for any function that is built as an interim workaround till the system matures 3. each north star item is eventually tied to the goal, objective driven by requirements, revenue, cost, feedbacks etc.. 4. blueprint as a service maintains this definition and traceability for each project in carolverse, including carolverse -- you can file this as a seperate arc or fold this into blueprint arc that sahil is working, as appropriate.. please make a note for each initiative in regard to who defined the initiative vs who owns it vs who is it assigned to for execution" READ AS (Orion's reading, marked apart from the words above, P.01.01.05.05): (1) A project's NORTH STAR is not one goal or objective: it is a set of items, one for each function the project runs as an INTERIM WORKAROUND until the system matures, each stating that function's TARGET STATE. (2) Each north star item is tied, eventually, to the goals and objectives it serves, and each of those names what drives it: requirements, revenue, cost, feedback and the like. (3) BLUEPRINT, as a service (Leo's), keeps the north star and that traceability for every project in Carolverse, Carolverse included; it is one record per project, never a separate app, and the roadmap steers toward it. (4) Carolverse's own north star is defined first. (5) Filed as its own arc, owned by Leo, apart from the Blueprint onboarding arc Sahil is working (CAROL-ARC-0004), because the north star covers every project, Carolverse included, and is not a step of onboarding one project. (6) Every initiative of this arc carries a note naming who DEFINED it, who OWNS it and who is ASSIGNED to execute it.

A client project uses the estate's one build machinery and never gets its own copy; P.01.03.07.08 is excepted

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED: Policy P.01.03.07.08 promises a project a DEDICATED per-project deployment of every service it enables and makes no exception for Build Initiatives, while blueprint #50 has every project build through the estate's one Build Initiatives -- which governs? Orion recommended: "the one pipeline wins, and the policy gets an exception for the build machinery. Everything built since 17 September already assumes that." HIS ANSWER (adopting that recommendation, pasted back word for word): "the one pipeline wins, and the policy gets an exception for the build machinery. Everything built since 17 September already assumes that". READ AS (Orion's reading, not his words): blueprint #50 governs; P.01.03.07.08 is amended through the policy door to except the build machinery -- Build Initiatives with its planner and bypass lanes -- which every project uses from the estate and never receives a copy of; every other service keeps the per-project deployment the policy promises.

A dominion is held by one client and may host other clients' projects; Carolverse is Eden's foundational project with Denken Labs

Ruled 2026-10-20 by ninad -- in force

Ninad, 2026-10-21 (CLI-883). THE QUESTION ANSWERED: Orion asked: "Sahil's Chowpatty registration (his #96): your rules say a dominion like Eden has exactly ONE client. But your 5 Oct ruling says Chowpatty is built on Carolverse, which puts it in Eden while it belongs to its own client (Chowpatty) and account. Is a client's project allowed to sit in Eden under its own client and account? (A separate dominion would mean a separate build pipeline, which your one-pipeline rulings forbid.)" offering 'Yes, Eden hosts it' (recommended: Eden keeps Denken Labs as its owning client and may host other clients' projects built on Carolverse) or 'No, register under us'. HIS ANSWER, word for word: "carolverse is the foundational project with denken labs, and the platform can host multiple projects for multiple clients under one realm - eden". READ AS (Orion's reading, not his words): the clause of blueprint #4 'a dominion has exactly ONE client, reached through its account' is retired in place. Carolverse is Eden's foundational (base) project, held by Denken Labs; Eden may host projects of several clients, each under its own client and account, so Chowpatty is registered in Eden under client Chowpatty and its own account. He said 'realm'; read as the dominion Eden, because Eden is a dominion and realm is the word for a department's place inside a dominion (cookbook 1718). Everything else in #4 stands; restated here with the one clause changed: A DOMINION is the collection of projects that share one underlying build pipeline -- one whole installation of the platform. The dominion the estate runs in today is EDEN. It sits BETWEEN account and project: client -> accounts -> dominions -> projects. A client may hold several accounts; an account may hold one or more dominions; a dominion is HELD by ONE client, reached through its account -- the client of its base project -- and may HOST projects of other clients, each project under its own client and its own account; and every project belongs to exactly ONE dominion. A copy of a project into another dominion is a SEPARATE project, which may later be renamed. A dominion is a COPY, not a shared deployment -- Ninad's words: like an installation of Windows on a separate machine -- and it carries its own copy of EVERYTHING: records, money, budgets, access, agents and projects. Each dominion holds a Carolverse-type BASE PROJECT whose name, services and agent names may all differ from the dominion it was copied from. AGENT NAMES ARE UNIQUE WITHIN A DOMINION, never across dominions. This does NOT supersede the service-instantiation ruling (design 557, P.01.03.07.08): one shared codebase never forked governs services WITHIN a dominion. A dominion is itself a fork by design, and collapsing the two would remove the independent evolution that is the point of the concept.

A service serves every project from a single copy wherever possible; its code is not replicated across projects

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884). THE QUESTION: Ninad asked: "how many services can be reused in carolverse? for services that will be reused, do projects get a seperate copy of those services? that may not be ideal.. what do you think?" Orion answered that all 22 services can be reused at least in part (72 graded tracks), that by default a project gets its own running copy of the same code on its own machine (P.01.03.07.08, design #557) with Build Initiatives the one shared exception, and recommended recording 'shared or own copy' for each track as an explicit choice instead of defaulting everything to own copies. HIS ANSWER, word for word: "try to have a single copy for services wherever possible, instead of replicating the code across projects (now that is a policy)". READ AS (Orion's reading, not his words): the default of P.01.03.07.08 and design #557 -- a dedicated instance of a service's shared code deployed for each project -- is reversed. A service runs as ONE copy that serves every project using it, wherever possible; its code is not replicated across projects. A project gets its own copy of a service only where a single copy cannot serve it, and why is recorded on that service's track. Build Initiatives is one such single copy (blueprint #50, #341). He asked for it as policy, so P.01.03.07.08 is rewritten; its words are drafted for his approval and apply once he approves them (governance #355). Until then P.01.03.07.08's present words stand in the policy store, and this ruling is the later word on the question.

Project onboarding decides each service the project needs: reused or adapted from Carolverse, and served by the single copy or a copy of its own

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884). WHAT HE ANSWERED: approving the new words of P.01.03.07.08 (Rulings Repo blueprint #470), he quoted their sentence "Before a service is enabled for a project, whether the project is served by the single copy or by its own copy is recorded as a design decision" and added, word for word: "this should be part of the project onboarding steps.. deciding the various services applicable for that the project based on the project needs and whether the services can be reused or adapted from the foundation project (carolverse)". READ AS (Orion's reading, not his words): Leo's project onboarding pass carries a step that decides, from the project's own requirements, which services the project needs; for each, whether it is reused from Carolverse, the foundation project, adapted from it, or built for the project alone (blueprint #242-#246); and whether the project is served by the service's single copy or by a copy of its own (#470, P.01.03.07.08). Each decision is recorded as a design decision before any service is enabled for the project. The pass's 'services recommended' (step 6) and 'Enable the project's services' (step 18) follow it; step 18's 'instances of shared code at pinned versions' was retired in place by #470. The pass is changed on the BLOOM record and regenerated, never edited by hand.

The rebuild for the single-copy rule goes into the arcs dedicated to its topics

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884). THE QUESTION ANSWERED (after blueprint #470 and #471): Orion asked: "The project onboarding steps still say every project gets its own pinned copy of each service, so they need a small rebuild. Should I make that rebuild on my lane now, or hand it to Sahil as part of the Chowpatty run he's leading?" HIS ANSWER, word for word: "the rebuild should be part of the arcs that are dedicated for the topics". READ AS (Orion's reading, not his words): the work that carries #470 and #471 into the estate is filed into the arcs whose subject it is -- the onboarding pass's change into CAROL-ARC-0004 (Blueprint: Leo onboards a project), the per-track single-copy-or-own-copy record into CAROL-ARC-0001 (Service reusability) -- and is worked there by whoever works those arcs, not as a side task of one session or of the Chowpatty run (CAROL-PUR-0006), which picks it up as a step if it needs it.

Every project has a CEO and a head of finance

Ruled 2026-10-21 by ninad -- superseded 2026-10-21

EVERY PROJECT HAS A CEO AND A HEAD OF FINANCE (Ninad, CLI-894, 2026-10-21). Shown initiatives #486 built as (an Orion-session filing in an arc owned by the arc's owner; 201 pending initiatives moved, 6782 to Heimdall) and proposed policy words for P.01.01.01.10 that kept 'the owner of an initiative is the agent who requested it' with that one exception, he answered: "a requester cannot be the owner, instead, whoever is paying for the initiative should be the owner of the initiative.. a requestor is just a requestor. The payment is about budget that enables building or expanding a service. The budget comes from project. Project belongs to a client (say denken-labs) who assignes budgets for expanding existing services and building new services. Each project should have the CEO and head of finance (the agent and employee mapping in future, estate operators get mapped by default) The agent economy is tightly intermingled within the human economy in carolverse.. note this ruling, let the next session instantiate or manifest it into the estate". READ AS (Orion, not the ruling): (1) each project's roster carries two roles, a CEO and a head of finance, each an agent of that project; (2) each is later mapped to an employee of the project, and until then the project's estate operators are mapped to both by default; (3) they answer for the project's budget, which pays for its initiatives (cost-center #488, initiatives #487); (4) OPEN, to settle with Ninad at once (P.01.01.05.08): this narrows blueprint #242 ('a client project is built from its own requirements and inherits nothing by default') for these two roles, and meets the CLI-272 ruling (every project its own CEO modelled on Carolverse's) that today's email to Ninad said #242 had settled.

Ended because: Ninad, CLI-896: 'let the mvp of chowpatty bypass the org structure. A project can start as a time and material format ... and eventually turn into a managed delivery (with a proper org structure)' -- narrowed in place: the two roles are required of a managed-delivery project

A client project may start as time and material, renting specific agents, and turn into managed delivery with a proper org structure later

Ruled 2026-10-21 by ninad -- in force

A CLIENT PROJECT MAY START AS TIME AND MATERIAL, RENTING SPECIFIC AGENTS, AND TURN INTO MANAGED DELIVERY WITH A PROPER ORG STRUCTURE LATER; CHOWPATTY'S MVP BYPASSES THE ORG STRUCTURE (Ninad, CLI-896, 2026-10-21). Shown that blueprint #489 (every project has a CEO and a head of finance) is later than #242 and the CLI-272 ruling, so Chowpatty's roster would gain both, he answered: "let the mvp of chowpatty bypass the org structure. A project can start as a time and material format (renting out specific agents) and eventually turn into a managed delivery (with a proper org structure) if you know what i mean". READ AS (Orion, not the ruling): (1) a project carries an ENGAGEMENT MODEL, a governed value: time_and_material (the client rents specific agents, and each rented agent's work bills the client per agent) or managed_delivery (the project has its own org structure: a CEO and a head of finance, blueprint #489); (2) #489 is NARROWED IN PLACE: its two roles are required of a managed-delivery project and not of a time-and-material one; (3) Chowpatty's MVP is time and material, so Sahil's roster proposal (no CEO) stands and #242 is not narrowed for it; (4) a project may move from one model to the other, recorded.

Bilbo works for Leo, telling the story of Leo's work

Ruled 2026-10-22 by ninad -- in force

Ninad, 2026-10-22 (CLI-912). HIS WORDS: "I think bilbo managing carolopedia is a mistake, let bilbo work for leo, and move the carolopedia management to Loki.. anyways carolopedia is not managed efficiently, all its links are gone". Asked first who writes about the work Leo does (nobody does today: Scriber's stories cover Elrond's build pipeline) and whether Bilbo does for Leo what Scriber does for Elrond (he does not: he writes nothing about Leo's work).

READ AS (Orion's reading, not the ruling): Bilbo stays in Leo's line and works for Leo -- he tells the story of Leo's work, as Scriber does for Elrond's, and no longer manages Carolopedia. His Blogs service is not changed by this ruling.

Leo's external chat and Leo's operator window feed one requirements record, copied to the project's machine once it is procured

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). THE QUESTION: Orion proposed that what a client says in Leo's chat goes into the requirements Leo builds the team from, credited to the client. HIS ANSWER, word for word: "both leo chat (external, not internal) and operator's Leo window used to feed the requirements should go and sit into the same requirements bucket. The requirements should be copied to the vm once it is procured"; and, on the name Orion had proposed for the operator's route: "Instead of operator bypass, use the term - Leo's operator window". READ AS (Orion's reading, not his words): Leo's external chat (never an internal window) and Leo's operator window both write a project's requirements into its one numbered requirements record through that record's one door, each statement naming who gave it and which window carried it. Once the project's machine is procured the requirements are copied to it as a copy kept in step from the estate's record, which stays their one home (cookbook 1126: one fact, one home) and is never edited on the machine. The dictionary's words for Leo's operator window apply once he approves them (P.01.01.05.09).

Procuring a project's machine is an item on its roadmap, carried out when the roadmap reaches it

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Procurement of vm should be a roadmap item, carried out in the sequence that the roadmap dictates". READ AS (Orion's reading, not his words): every project's roadmap carries its machine as an item, with the size and monthly cost the administrator approves with the roadmap; Hagrid makes the machine only when the roadmap's order reaches that item, never ahead of it or on a schedule of its own.

A project's roadmap decides everything the project needs and asks the right questions; every template is taken from Carolverse projects

Ruled 2026-10-26 by ninad -- superseded 2026-10-26

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Roadmap should also decide which services, tracks, tasks, agents, droids, apps, security, users, agent org etc, and ask the right questions to the person providing requirements via either leo's external chat or Leo's operator window. All templates should be picked from carolverse projects (create a policy for this, not just a ruling)". READ AS (Orion's reading, not his words): a project's roadmap is where its services, tracks, tasks, agents, droids, apps, security, users and agent org are decided, each from the project's own requirements; Leo asks the person giving the requirements, through either window, the questions those decisions need; the building blocks a decision chooses from are project templates taken only from projects built on Carolverse, and a template is used only where a requirement calls for it, so blueprint #242, #245 and P.01.03.07.22 stand. He asked for a policy: its words are drafted for his approval and apply once he approves them (P.01.01.05.09).

Ended because: Ninad: it belongs in the strategy cookbook, not a policy (CLI-968)

The agent-centric modular architecture is the template for any new work for projects

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Agent centric modular design should be used as a template for any new work for projects". READ AS (Orion's reading, not his words): every new piece of work built for a project follows the Agent-Centric Modular Architecture (ACM) -- everything owned by a named agent, capability modular and reused rather than forked, an app a thin window over a shared composer -- and is judged against it.

A project's roadmap decides everything the project needs and asks the right questions; every template is taken from Carolverse projects

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Roadmap should also decide which services, tracks, tasks, agents, droids, apps, security, users, agent org etc, and ask the right questions to the person providing requirements via either leo's external chat or Leo's operator window. All templates should be picked from carolverse projects (create a policy for this, not just a ruling)". READ AS (Orion's reading, not his words): a project's roadmap is where its services, tracks, tasks, agents, droids, apps, security, users and agent org are decided, each from the project's own requirements; Leo asks the person giving the requirements, through either window, the questions those decisions need; the building blocks a decision chooses from are project templates taken only from projects built on Carolverse, and a template is used only where a requirement calls for it, so blueprint #242, #245 and P.01.03.07.22 stand. AMENDED THE SAME DAY: shown the words Orion drafted as a policy, HIS ANSWER, word for word: "the policy sounds more like a strategy cookbook, not a policy.. it should be updated in the strategy cookbook". READ AS (Orion's reading, not his words): this ruling is kept as a recipe in Leo's Strategy Cookbook, not as a policy, and no policy is proposed for it.

A project's roadmap app offers the roadmap as a PDF carrying its version, for approval

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). THE QUESTION: Orion said nothing produces a roadmap PDF today and asked whether the project's roadmap app should offer a PDF download stamped with the roadmap's version, so an approval matches the exact version. HIS ANSWER, word for word: "yes to both of your questions". READ AS (Orion's reading, not his words): every project's roadmap app offers its roadmap, with each item's cost and time estimates, as a PDF carrying the roadmap's version mark; the approval itself is still given in Leo's chat or the administrator's Orion session, never by a button in the app (cookbook 1710), and is recorded against that version.

An estate operator on Leo's chat gives the requirements behind every Blueprint deliverable, filing included; Leo's filing opens to subscribed external users, and the operator window is a stop-gap

Ruled 2026-11-03 by ninad -- in force

AN ESTATE OPERATOR ON LEO'S CHAT GIVES THE REQUIREMENTS BEHIND EVERY BLUEPRINT DELIVERABLE, FILING INCLUDED; LEO'S FILING OPENS TO EXTERNAL USERS SUBSCRIBED TO BLUEPRINT, AND THE OPERATOR WINDOW IS A STOP-GAP (Ninad, CLI-017, 2026-11-03). THE QUESTION ANSWERED: Sahil's operator communication #142 offered three shapes for Leo's operator window: on Leo's own chat with the right to file initiatives, on Leo's chat for requirements only, or Leo's chat for clients only with operators recording requirements from Orion's chat. HIS WORDS: "The estate operator for Leo's chat should be able to give the requirements that result in all the deliverables that blueprint service owns (roadmap, business plan, roi etc). Filing initiatives is also one of this function, but it was reserved to internal employees for all the agents. However, for Leo, it should be available for external users approved as subscribes to the blueprint service, because Leo will, in future directly interact with the users to build products. Operator window is just a stop gap until Leo gets ready to interact directly with external users.." READ AS (Orion's reading, not his words): (1) the operator window opens on Leo's own chat and lets an estate operator give the requirements that produce every deliverable Blueprint owns, filing initiatives included -- the first of Sahil's three shapes; (2) for Leo alone, filing initiatives is also open to an external user approved as a subscriber to the Blueprint service -- an exception, for Leo only, to the rule that filing is reserved to internal employees; (3) the operator window is a stop-gap until Leo interacts with external users directly, so it is a decision for now and never policy text.

Blueprint's objective is downstream business and client value; the business plan is its key deliverable beside the roadmap

Ruled 2026-11-03 by ninad -- in force

BLUEPRINT'S OBJECTIVE IS DOWNSTREAM BUSINESS AND CLIENT VALUE, AND ITS KEY DELIVERABLE BESIDE THE ROADMAP IS THE BUSINESS PLAN (Ninad, CLI-023, 2026-11-03). THE CONTEXT: told that Blueprint produces no business plan, revenue strategy or balance sheet for a client, he said: HIS WORDS: "another important thing to note is that, the objective of the blueprint service is to generate downstream business, create value for the clients, convince the clients that carolverse agents can produce differentiated and cost effective outcomes, and provide them with the visibility of the outcomes via the business plan. This is the key deliverable of the blueprint service, apart from creating the roadmap..". READ AS (Orion's reading, not his words): the business plan is part of Blueprint's minimum, not an extra: it shows the client the outcomes Carolverse's agents will produce, that they are differentiated and cost-effective, and the value to the client, so that the client signs the contract and the work flows downstream.

Signing the contract is a Blueprint step: the terms on one easy scrolling page with an I agree button at its foot, and agreeing makes the verified user a client

Ruled 2026-11-03 by ninad -- in force

SIGNING THE CONTRACT IS A STEP IN BLUEPRINT, WHERE A VERIFIED USER BECOMES A CLIENT; THE CONTRACT IS SHOWN AS ONE EASY SCROLLING TEXT OF ITS TERMS AND CONDITIONS WITH AN 'I AGREE' BUTTON AT THE BOTTOM (Ninad, CLI-025, 2026-11-03). Follows operations #603/#605 (a verified user becomes a client when they sign the contract, which carries scope, methods of working, quality, timeline, cost and payment terms) and blueprint #604. THE QUESTION ANSWERED: told that his verified-tier ruling (operations #606) is not yet fully built, he answered about the client step instead. HIS WORDS: "signing a contract should be a step in blueprint.. this is when a verified user transitions to a client after he signs the contract, which should be an easy scroll down text with terms and conditions and an 'I agree' button in the bottom.."

A client project gets a fundraising plan only when its own requirements ask for one, run by its own agent from Clara's Fundraising template

Ruled 2026-11-05 by ninad -- in force

Ninad, 2026-11-05 (CLI-038), asked "Clara does not communicate with users, only Leo does, so how will Leo communicate with Clara, and how will clara build the service for new projects? like for Chowpatty? should clara build a fund raising plan for chowpatty? or should chowpatty have its own CEO to do it for Sachin?"; told Orion's reading of blueprint #242, #245 and #543, he said "i understand and agree to your proposed solution, but are Leo, Clara and the team aware of these?". READ AS: a client project has a fundraising plan only when its own requirements ask for one, given through Leo's external chat or Leo's operator window into the project's one requirements record (#538); the project's roadmap then decides which of the project's own agents runs it (#543) -- never Clara, and never a project CEO by default (#242) -- and the project adopts Clara's Fundraising template requirement by requirement (#245), with its own inputs (how it earns, what it costs, its rounds and buffers, its tax basis) and nothing copied from Carolverse; Elrond's team builds it. Leo and Clara hand over through records, not conversation: Leo through the project's requirements, Clara through the template (leadership #640). Chowpatty's stated need is a sales, an admin and a customer support agent (#243), so it gets no fundraising plan unless Sachin asks for one.

A project's work is executed by BLOOM in small pursuits Leo agrees with the user, and only on the user's explicit confirmation in chat

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-040). Asked whether arcs and pursuits belong to a project and how Leo carries a project from requirements to business plan, roadmap, initiatives, arcs and pursuits and status reports, and told that the Blueprint pursuit says nothing about each roadmap item becoming initiatives for the project, HIS WORDS: "thats right, each roadmap item may not become an initiative.. execution should follow the bloom methodology, which means that execution happens in small pursuits agreed by leo with the user, especially on the cost, time and outcome after service specific arcs are defined.. the policy for leo is to keep the user informed of the possibilities, but only execute after explicit user confirmation via chat..". READ AS (Orion's reading, not his words): a roadmap item is not turned into initiatives by itself; once the project's service-specific arcs are defined, its work is executed in small pursuits, each agreed by Leo with the user on its cost, time and outcome; Leo keeps the user informed of what is possible, and nothing is executed before the user confirms it explicitly in Leo's chat. The arcs belong to the project's services (initiatives #491). Leo's binding rule is proposed as a policy for Ninad's approval; until approved this ruling stands.

Ninad owns the Blueprint arc as its creator, and Sahil carries its work on as its executor

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042): 'Under "the owner is the operator who created it", you own the Blueprint arc: you created it. But your September ruling says "Sahil owns the Blueprint arc and carries its work on". Which stands?' HE CHOSE 'I own, Sahil executes': today's rule replaces the September one -- he is the owner as creator, and Sahil is assigned to execute both the Blueprint arc and the Blueprint pursuit. READ AS (Orion, not his words): the owner operator of CAROL-ARC-0004 (Blueprint: Leo onboards a project) is Ninad, its creator; Sahil carries its work on as the operator assigned to execute it (agent-resources ruling of the same day), through his own Orion session, and answers for its progress to its owner. The arc's owning agent is unchanged. What is reserved to root and core stays the administrator's (security #175).

An agent is never unaware of what is done in its name: a droid holding a key its thinking does not hold leaves a summarised view in its memory of what was wrong or seems wrong, never the raw data

Ruled 2026-11-10 by ninad -- in force

AN AGENT IS NEVER UNAWARE OF WHAT IS DONE IN ITS NAME: A DROID THAT HOLDS A KEY THE AGENT'S THINKING DOES NOT HOLD LEAVES A SUMMARISED VIEW IN THE AGENT'S MEMORY -- KNOWLEDGE THAT SOMETHING WAS WRONG, OR THAT SOMETHING SEEMS WRONG -- NEVER THE RAW DATA (Ninad, 2026-11-10 UTC, CLI-103; adds to blueprint #542 and security #515, supersedes nothing).

THE QUESTION ANSWERED: for Orion proposed that Heimdall's watches read every log on the machine through a locked log-reading login of their own, so Heimdall's chat and his thinking -- which send text to a model -- never hold the logs (logs can carry users' messages and secrets; security #416 keeps log access off any login with a password door). HIS WORDS: "Will this violate the agent centric modular architecture, where agent is doing things sub-conciously.. ?there is no design policy regarding this design, what is your suggestion? is it ok for agents to do things that they are not aware of?"

Orion answered: no agent should act unaware, but awareness need not mean holding the raw data; every droid works in its agent's name and is registered under it, each run leaves a record the agent reads, every alarm reaches the agent at once, and a droid may hold a key the agent's thinking does not but never acts unseen -- like a reflex, not deliberated but felt and reported; and that the raw logs stay out of what his thinking sends to a model. HIS WORDS: "raw logs stay out of what his thinking sends to a model - there should be something in his memory, like a summarized view, either knowledge that something was wrong or knowledge of something that seems wrong.."

READ AS (Orion's reading, kept apart from his words; not yet confirmed by him): a design policy for every agent, not only Heimdall. (1) A droid acts only in its owning agent's name and is registered under that agent (security #515). (2) When a droid holds a key the agent's thinking must not hold (log access, a secret, a person's data), each of its runs writes into the agent's own memory a SUMMARISED VIEW: what it found wrong, and what seems wrong but is not yet proven, each with when, where and how sure -- in plain words, with no raw lines, secrets or message text. A run that finds nothing leaves that it looked. (3) Every alarm reaches the agent at once as well as the person it is raised to, so the agent can be asked about it and answer from its memory. (4) The raw data stays with the droid's key; the agent reaches it, when a person asks, only through that droid's door, which returns the summary. First applied to Heimdall's Sign-in Watch and his 5-minute alarm; the agent-centric design (design 146) and the security handbook follow.

A client project's arcs are its own, tagged to its own services; onboarding arcs go to a default service registered with the project when its Blueprint subscription is approved

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), answering Sahil's operator communication #333, point 2 (starts once the project's service-specific arcs are defined; arcs today belong to Carolverse services and carry no project; Chowpatty has none: (a) Carolverse-service arcs tagged with the project, or (b) arcs of the project's own services; and who creates them). HIS WORDS: "Projects own arcs, tagged to projects own services.. but the arcs required to onboard the project when the services are not created yet should be tagged to a default service that gets created as soon as a project is registered. Registering the project and default service should happen when the user subscription for blueprint is approved.. this is closely related to the project onboarding arc / pursuit, and hence should be handled cohesively". READ AS (Orion's reading, not his words): every arc of a client project belongs to that project and is tagged to one of the project's own services; when the user's Blueprint subscription is approved, the project is registered and a default service is created for it at once; the arcs needed to onboard the project before its own services exist are tagged to that default service; this is built as part of the project onboarding arc and pursuit, not apart from it. Who creates a client project's arcs was not ruled: until it is, the administrator or the project's estate operator creates them (initiatives #662 as restated).

Leo manages the client in the end; until he is ready only FDEs speak to him for the client, the project's estate operator steps in, and the Carolverse estate operator or Blueprint service operator enables him

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), told that Sahil's operator communication #333 was answered with initiatives #798 (Leo creates a client project's pursuit once the user confirms it in his chat; the project's estate operator owns and executes it), HIS WORDS: "the ruling here is that end state - Leo manages the client, but in the interim, 1. FDE represents the client when he talks to Leo 2. Project estate operator intervenes to get things done when Leo is not able to do what he is supposed to do 3. Carolverse estate operator or blueprint service operator enables Leo to do that which he could not do So initially, only FDEs interact with Leo, until Leo is ready to face the clients directly..". #319 IN HIS WORDS (CLI-723, 2026-10-05) still stands for who registers projects: "Yes, Leo should register chowpatty as client and project..". READ AS (Orion's reading, not his words): (1) END STATE: Leo manages each client himself -- he registers the client, account and project (as #319), agrees each small pursuit with the client and takes the client's explicit confirmation in his chat (#642, #798). (2) UNTIL LEO IS READY to face clients directly, only FDEs talk to Leo: the client's forward deployed engineer (one per client, #679) represents the client in Leo's chat, and the agreement and explicit confirmation that #642 and #798 ask of the user are given by the FDE on the client's behalf; the client does not use Leo's chat meanwhile. (3) When Leo cannot do what he is supposed to, the project's estate operator steps in and gets it done -- through Leo's door where the work is his, recorded under his name with the operator who acted, Leo staying accountable. (4) The Carolverse estate operator or the Blueprint service operator then enables Leo to do himself what he could not, so each step-in closes a gap rather than becoming a standing role. (5) When Leo is ready to face clients directly is not ruled; until Ninad rules it, (2)-(4) hold. This replaces #319's reading that operators act as FDEs doing Leo's work for him: an FDE now speaks for the client, and stepping in for Leo is the project estate operator's. Still as #319 (3): a service operator acting through Leo's door inside an assigned arc or pursuit is not yet built.

Leo names a project's agents with its forward deployed engineer or its user, by a skill he holds

Ruled 2026-11-15 by ninad -- superseded 2026-11-17

Ninad, 2026-11-15 (CLI-154), asked what Chowpatty's first agent should be called. HIS WORDS: "Leo should name the agents in consultation with the FDE / user.. provide him the skill to do that..". READ AS (Orion's reading, not his words): every agent a project asks Athena for carries a name Leo chose with the project's forward deployed engineer or its user, in his chat or his operator window, never a name an operator or a build lane picks alone; Leo holds a skill for it (kept in Sage's library, bound to Leo) and the request records who agreed the name. Athena still creates and the administrator still approves every agent (agent-resources #31, #35).

Ended because: Ninad, CLI-207: option 3 -- Athena names an agent at creation; Leo gives it the client's alias

Every project starts as a prospect in Sandbox; the signed contract makes it a client, and only then is its own machine made and its proof of concept migrated onto it

Ruled 2026-11-15 by ninad -- in force

Ninad, 2026-11-15 (CLI-154). HIS WORDS: "Chowpatty is not a client yet, instead chowpatty is a prospect. I am building a sandbox for building POCs for prospects before they are migrated to the dedicated vm. This changes a few things, like project onboarding should always start a project as a prospect, and when the client becomes a paying customer (contract is signed) then the dedicated VM should be procured to migrate the prospect to the vm as a project.. the migration utility should be used to migrate the POC to the server with the right agent centric modular checks and architecture (like we did for glover)". READ AS (Orion's reading, not his words): project onboarding (Blueprint) starts every project as a PROSPECT, and its proof of concept is built in Sandbox, on Sandbox's machine (sales #790), never on a machine of its own. When its contract is signed (operations #605, blueprint #607) the prospect becomes a paying client and its work a client project: only then does Hagrid make its dedicated machine (the roadmap's machine item, #539, cannot be reached before the signature), and Migration moves the proof of concept onto it through Audit, Refactor and Conform, rebuilt to the agent-centric modular architecture with its checks, never copied as it stands -- as Glover's capabilities were migrated (sales #11).

Chowpatty is a prospect, the first built in Sandbox; the old Chowpatty records are discarded

Ruled 2026-11-15 by ninad -- in force

Chowpatty is a PROSPECT, the first whose proof of concept is built in Sandbox (blueprint #868): to-be client Chowpatty, account Chowpatty, project Sales & Customer Support, whose stated need is a sales agent, an admin agent and a customer support agent. It becomes a client project, with its own machine, only when its contract is signed. The earlier Chowpatty records in the estate (the aspirational street-food ordering service, its business line, its blocks, its ordering tools, its order notifier and its order store) are discarded: retired in place or moved to Trash, never deleted, and never reused as the new project.

The Blueprint pursuit builds the platform; a project's onboarding pursuit holds that project's own use of it, so mixed steps are split

Ruled 2026-11-15 by ninad -- in force

BLUEPRINT BUILDS THE PLATFORM; CHOWPATTY ONBOARDING USES IT FOR CHOWPATTY (Ninad, CLI-158, 2026-11-15). THE CONTEXT: shown that Chowpatty's first real request for a new agent sits in the Blueprint service pursuit (CAROL-PUR-0017). HIS WORDS: "why is naming chowpatty agents part of blueprint? should this not be part of the chowpatty onboarding pursuit? blueprint pursuit is about building the platform, and the chowpatty pursuit is about onboarding chowpatty, that was my understanding". THE QUESTION ANSWERED: told that Chowpatty's steps sit in Blueprint because #744 kept a shared initiative in the pursuit it joined last, that about nine open Blueprint steps mix building the platform with doing it for Chowpatty (4718's two approval-route criteria are platform work, 'Chowpatty's first real request decided' is Chowpatty onboarding), and asked how to fix it, he chose "Record and split": record his words as a ruling; each open mixed step keeps its platform part in Blueprint and its Chowpatty part becomes a Chowpatty onboarding step; 4718 closes on its two platform criteria and Chowpatty's first agent request moves to Chowpatty onboarding (CAROL-PUR-0006); closed steps stay where they finished. READ AS (Orion's reading, not his words): a step whose work is one project's own use of a Blueprint capability belongs to that project's onboarding pursuit, and a step building a capability any project uses belongs to Blueprint; a step doing both is split, never placed in both (#744 stands).

The Dominion, and where it sits in the hierarchy

Ruled 2026-11-15 by ninad -- retired 2026-11-15

A DOMINION is the collection of projects that share one underlying build pipeline โ€” one whole installation of the platform. The dominion the estate runs in today is EDEN. It sits BETWEEN account and project: client -> accounts -> dominions -> projects. A client may hold several accounts; an account may hold one or more dominions; a dominion has exactly ONE client, reached through its account; and every project belongs to exactly ONE dominion. A copy of a project into another dominion is a SEPARATE project, which may later be renamed. A dominion is a COPY, not a shared deployment โ€” Ninad's words: like an installation of Windows on a separate machine โ€” and it carries its own copy of EVERYTHING: records, money, budgets, access, agents and projects. Each dominion holds a Carolverse-type BASE PROJECT whose name, services and agent names may all differ from the dominion it was copied from. AGENT NAMES ARE UNIQUE WITHIN A DOMINION, never across dominions. This does NOT supersede the service-instantiation ruling (design 557, P.01.03.07.08): one shared codebase never forked governs services WITHIN a dominion. A dominion is itself a fork by design, and collapsing the two would remove the independent evolution that is the point of the concept.

Ended because: recorded by accident: CLI-157 ran shared machinery expecting its usage text, and the script re-adds any of its rulings not live -- this repeats #4, superseded by #467; Ninad made no new ruling (by ninad)

A signed project's dedicated machine: standard, on the Sandbox's architecture, root's sign-off, no client in the loop, billed to the client

Ruled 2026-11-15 by ninad -- in force

when a project is onboarded, it sits in sandbox by default as a prospect, and after the contract is signed, a dedicated vm is provisioned, and the project is moved to the dedicated vm.. the dedicated vm specs are standard, and hagdrid should use the same architecture as that of the sandbox to procure vms for now. it took significant efforts to build the sandbox, hopefully it wont take the same efforts to build the rest of the project specific vms. Admin signoff is a must for every vm, which will happen by default, as its root protected. As the project expands, vm requirements should be evolve.. note that provisioning infra is a backend activity, and client need not be kept in the loop for that.. the billing should account for the cost of dedicated VMs to store clients data and compute

Asked: one criterion says a project's machine size is derived from its requirements, but only one size is approved (cost-center #468); how should 4733 finish?

READ AS (Orion, apart from the words): a signed project's dedicated machine is the STANDARD size on the Sandbox's architecture -- the sizing rule on Leo's hand-off is that standard size; the administrator's sign-off is root's own step for every machine (root's register and its birth file), not a client approval; the machine is a backend matter the client is not asked about, so a client-approved roadmap item no longer gates it; resizing as a project grows is later work; the client's bill carries the dedicated machine's storage and compute cost.

Each project replicates Carolverse's existing patterns; what they do not settle goes to the operator through the Orion session working the Leo-filed initiative

Ruled 2026-11-15 by ninad -- in force

the general guideline is to pick the patterns from the existing carolverse setup, and replicate that for each project.. any design issue that cannot be deciphered from the existing carolverse setup should be brought to notice of the operator via orion instance that is working on the initiative that was filed by leo when converting the requirements to initiative as part of blueprint service activities

Said (CLI-164) after approving design 727 as 'in line with the already documented rulings'.

READ AS (Orion, apart from the words): when Blueprint builds what a project needs, its design copies the pattern Carolverse already runs for the same thing (machines, ways in, governance, framework); a design question that pattern does not answer is not decided by the builder and not guessed -- the Orion session executing the Leo-filed initiative brings it to its operator. This governs HOW a project is set up; WHAT it adopts still follows its own requirements, requirement by requirement (blueprint #242-#246), so the two agree: a pattern is copied where Carolverse has one that fits.

Leo onboards and builds client projects like a senior consultant: requirements, roadmap, plan, org and services, priced Carolverse services the client approves, and an operator to ask

Ruled 2026-11-15 by ninad -- in force

continue working on blueprint pursuit, and ensure that the initiative enables leo to onboard and build projects for clients using the right standard architecture and confirming to the clients requirements in a mode where he elegantly chat's with the client to capture all the requirements like a senior consultant, successfully translates the requirements to a roadmap, defines the plan, org structure, and services necessary for the project, and then build the agents, apps and droids within the architecture framework of services, tracks, tasks, and leverages carolverse services where required.. note that leveraging carolverse services comes at a cost, because carolverse services are cost centers, and the service owners never offer anything for free.. Leo should know the cost of each service and keep the clients informed about them, and ensure their approval before enabling any re-usable service for the client. He should recommend the services if he sees that the services are the right fit for the client's requirements, just like a seasoned consultant.. when he is not sure of something, he should have a communication channel to the operator (project estate operator) via orion instance to ask the questions, clarify doubts and relay the answers back to the client.. just like a thoughtful consultant.. that leo, and the blueprint pursuit should build leo to the specs

READ AS (Orion, apart from the words): the Blueprint pursuit's outcome is Leo to this specification -- (1) a consultant-like requirements conversation that captures all of a client's requirements; (2) the roadmap from them; (3) the plan, org structure and services the project needs; (4) agents, apps and droids built in the framework of services, tracks and tasks, on Carolverse's standard patterns (#894); (5) every re-usable Carolverse service carries a price its owner sets, Leo knows it, tells the client, recommends a service only where it fits, and enables none without the client's recorded approval; (6) a channel from Leo to the project's estate operator through their Orion session for what he is unsure of, the answer relayed back to the client. The client approves SERVICES and their cost; the machine stays backend (#891).

Sandbox holds a section per prospect, each with a POC section and a build section; POC work never in the build section

Ruled 2026-11-15 by ninad -- in force

HIS WORDS (Ninad, CLI-166): "Also note the additional requirement - the sandbox should hve a section for each prospect, and a section for POC and build within each prospect.. all POCs should happen in the POC section, while builds should happen in the build section.. POCs are defined as work that is custom built bypassing Leo's blueprint service channel.. No POC work should be allowed in build zone, generally speaking.. add these terms to the dictionary and design"

Context: Sandbox is the self-funded project where every prospect's proof of concept is built (blueprint #868, #869) on its own machine, sandbox-vm.

READ AS (Orion, apart from the words): Sandbox holds one PROSPECT SECTION per prospect; each holds a POC SECTION and a BUILD SECTION. A POC is work custom built for a prospect outside Leo's Blueprint service channel (not asked of Leo and not built by Elrond's build pipeline); a BUILD is work that came through that channel. POC work happens only in its prospect's POC section and builds only in its build section; POC work is refused in a build section as the rule ('generally speaking': an exception is an operator's ruling, never a default). The terms go into the dictionary and the layout into a design.

Leo asks the client functional questions (what the agents should do, and when) and the project's estate operators, through their Orion sessions, non-functional ones (how Carolverse builds and runs agents: architecture, design patterns, compliance)

Ruled 2026-11-15 by ninad -- in force

Ninad, 2026-11-16 (CLI-167), on reading that Leo puts a roadmap question to the client in either of his chat windows. HIS WORDS: "he puts the question to the client in either of his chat windows, and the answer is recorded as a new requirement - Leo should know which questions to ask the client vs which to ask the estate operator via orion.. any "non functional" questions that are internal to the way carolverse agents are build and how they operate, artchitecture and design patterns, complainces etc should be routed to the estate operator.. any "functional questions" about what the agents should do, when etc should go to the client.. add these terms to the dictionary - functional and non-functional". READ AS (Orion's reading, not his words): every question Leo needs answered about a project is one of two kinds, recorded on the question. A FUNCTIONAL question asks what the project's agents should do and when -- the client's business, their people, their processes and outcomes -- and goes to the client, in Leo's external window or through the operator window. A NON-FUNCTIONAL question is internal to how Carolverse builds and runs agents -- architecture, design patterns, compliance and the like -- and goes to the project's estate operators through their Orion sessions, never to the client. Per governance #404 the dictionary carries them with a prefix: 'Project question: functional' and 'Project question: non-functional', the values of the reference set project_question.nature. Narrows 's criterion 19453 (both windows) to functional questions; the estate-operator route is, with #894.

Leo's questions for the estate operators travel by model-free processes, linked by number, never by a model inferring answers from messages

Ruled 2026-11-15 by ninad -- in force

Ninad, 2026-11-16 (CLI-167). ASKED: 'galadriel does that for carol I think, where she reads the communication, infers the actions and acts on them.. would a similar design work or do you have a better design in mind, where Leo's droids would do it for him?' OFFERED: Galadriel-style (a model reads each message and infers what to do) or the relay's shape done by Leo's own model-free processes, the question and its answer linked by number so nothing is inferred. HIS ANSWER: "no problem, go ahead.." (the processes). READ AS (Orion's reading, not his words): Leo's questions for a project's estate operators travel by model-free processes -- Orion's courier carries them through the agents' door (the communication store is the operators' and Orion's alone) and reads the answer an operator closes the communication with; Leo's own relay puts a client's answer into the client's conversation -- never by a model inferring an answer or an action from free text. Built as (shared machinery).

A prospect's sections on Sandbox's machine are owned by Hagrid's forced section login

Ruled 2026-11-16 by ninad -- in force

HIS WORDS (Ninad, CLI-171, choosing): "Hagrid's login owns them" -- the option read: "The prospect folder belongs to Hagrid's forced section login, which only ever runs the one command. POC is writable only by carried operators, build only by agents on the machine's pass. Amend design 728's ownership row. No new root door; one root step to install."

Asked: "Sandbox's sections design contradicts itself: it says each prospect's folder is owned by root, but also that it is made by a command started by Hagrid's own key, and that command runs as Hagrid's login, not root (Sandbox's machine has no door that runs as root, by design). Which do we change?" Options: Hagrid's login owns them (recommended) / add a root door / root makes each section.

READ AS (Orion, apart from the words): on Sandbox's machine each prospect's section and its POC and build sections are made and owned by Hagrid's forced section login, never by root; the POC section is written only by the carried operator logins, the build section only by the agents on the machine's pass; the machine gains no door that runs as root (design 727 section 8 stands). Design 728 section 2's root:root for the prospect section is retired in place by its v1.1.

The reuse grade as-is means a project uses the estate's existing service, its work tagged to the project and done by the project's operators

Ruled 2026-11-16 by ninad -- in force

Asked: blueprint #50 means the build-pipeline tracks are never copied for a project -- the estate's one pipeline does the work -- but none of the Service Track Handbook's four reuse grades (as-is, config-only, estate-bound, seam-refactor) says that. Orion offered a fifth grade 'estate-served' (a project uses the estate's own instance, nothing deployed for it), and Ninad first chose it; Orion then explained that 'as-is' today still deploys a per-project copy needing only credentials, and offered either keeping a separate grade or redefining 'as-is'. Ninad: "isnt as-is same as estate servved, meaning, woudnt the pipeline be able to serve all projects as it is, just by ensuring that the work per project is tagged accordingly, and done by respective operators?" and then "that is probably why 0 tracks belong to the as-is grade.. I think we should redefine as-is to use existing service.. also I would expect more tracks in as-is". READ AS: the reuse grade 'as-is' is redefined: a project uses the estate's existing, running service -- nothing is deployed or copied for it -- and the work it does for a project is tagged to that project and done by that project's operators. No fifth grade 'estate-served' is made (his first choice in this session is superseded by these words before anything was recorded). Every Build Initiatives track is graded as-is under blueprint #50. Tracks graded config-only or estate-bound today are to be re-read against the new meaning (he expects more tracks in as-is). The dictionary's words for 'as-is' change only on his approval (governance #355).

A client project's requirements are gleaned by Leo with the questions each subscribed service supplies; the plan service is used as-is

Ruled 2026-11-16 by ninad -- in force

Asked (CLI-181), which reuse grade the service that builds a project's plan belongs to. Ninad: "for eg - the project plan - we are building a project plan for carolverse, on similar terms, a project plan will have to be built for each project.. it will depend on the business model of that project.. Leo will have to ask the right questions to be able to build this plan for the project.. the investment track will need to prompt leo with the questions that he needs to get clarifications from the client.. so the process is reversed. In carolverse, i am "giving" all the requirements, nobody is asking me.. biut for clients, leo will ask the questions to "fetch" or "glean" the requirements from the client, and the respective services that the project gets subscribed to will have to prompt Leo to ask those questions.. the service that builds the plan for project sits in carolverse, and will be used as is, the end product may change based on the nature of the business, but the service is re.-usable.. what grade will this service belong to?" READ AS: (1) the investment-plan service (Clara's Finance track, leadership-investments) is used as-is for every project -- the product differs with each project's business model, the service does not; (2) for a client project requirements are gleaned, not given: every service a project subscribes to supplies Leo with the questions he must ask the client to obtain the requirements that service needs, and Leo asks them.

The meaning of the reuse grade as-is, approved for the dictionary

Ruled 2026-11-16 by ninad -- in force

Asked (CLI-181): approve this dictionary meaning for the reuse grade 'as-is' (ruling #918): "As-is: a project uses the estate's existing, running service for this track; nothing is deployed or copied for it. The work the service does for a project is tagged to that project and done by that project's operators, and what it produces may differ with the project's business while the service stays the same (for example the build pipeline or the investment plan). For a client project the service supplies Leo with the questions he asks the client to gather what it needs (#919)." and its contrast with config-only: "Config-only gives the project its own instance, set up by one config file; as-is gives it no instance at all, because the estate's running service does the work." Ninad: "this is approved as a ruling, and the associated definitions of the terms should be included in the dictionary". READ AS: the words above are the meaning of 'as-is'; the dictionary's 'As-is' entry and its near-miss with config-only carry them, and its 'Reuse level' entry names the enablement serving a track from the estate's running service.

Re-grading tracks against as-is, and services supplying Leo's questions, are MVP of the Blueprint pursuit

Ruled 2026-11-16 by ninad -- in force

Asked (CLI-181): where two pieces of work from #918 and #919 sit in CAROL-PUR-0017 -- (1) re-read the 51 config-only and 15 estate-bound tracks against the new 'as-is' and re-grade those that should be as-is; (2) every service a client project subscribes to supplies Leo with the questions he must ask the client (#919). Ninad: "both in MVP".

Leo subscribes and unsubscribes a client's services on its behalf the client approves; beyond it only on the client's approval.

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-181), with his answer on where #919's work sits: "Also, the cleint never subscribe to a service on their own, they wont know what a service is or which service they need.. Leo identifies the services that client needs, and subscribes to them on their behalf.. if the client seem to be not interested in a specific service that they are subscribed to, Leo takes the decision to unsubscribe the service for them.. if Leo identifies a service that the client seems to be interested in, then he recommends and subscribes to that additional service.. Leo manages the cost expectations with the client and only subscribes to services if they fit into the clients budget. If he thinks that the service is essential, Leo negotiates with the client and convinces them to subscribe to that service even if it extends beyond the budget, and does so if the client approves". Asked then, because #895 (CLI-164) says Leo must 'ensure their approval before enabling any re-usable service for the client': which stands -- the budget is the approval (inside it Leo subscribes and unsubscribes on his own judgment and tells the client; beyond it he needs the client's explicit approval), or approval for each subscription. Ninad: "Budget is the approval (Recommended)". READ AS: a client never subscribes to a service; Leo does, on the client's behalf. The client approves a budget; inside it Leo subscribes, recommends and subscribes further, and unsubscribes a service the client shows no interest in, on his own judgment, telling the client; a service he judges essential beyond the budget he negotiates and subscribes only on the client's explicit approval. #895's clause 'ensure their approval before enabling any re-usable service' is retired in place by this ruling (P.01.01.05.08, the later ruling of the same operator); the rest of #895 stands.

FDEs do POC in the POC zone, and become their prospects' service operators and their projects' estate operators by default

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

Ninad (CLI-185): "ruling - FDE's should be able to do POC in the POC zone. They also by default become the service operators for the prospects and estate operator for the projects assigned to them". Context: Orion had answered his questions on the Sandbox pursuit (CAROL-PUR-0042): an operator reaches only their own project and the projects built on it, so an operator dedicated to a prospect such as Chowpatty cannot be assigned Sandbox work (Leo files a prospect's builds under Sandbox) and gets no login on Sandbox's machine, whose logins follow root's register, never an assignment; and an FDE does not file for Leo but speaks for the client in Leo's chat (#801). READ AS (Orion's reading, not his words): (1) a prospect's FDE gets a way onto Sandbox's machine that writes that prospect's POC section (design 728) and nothing else; (2) an FDE assigned to a prospect is, by that assignment and without a further act, its service operator; (3) an FDE assigned to a project is, by that assignment, its estate operator on root's register. Who assigns an FDE, and how one person holds both roles while root's record keeps one rung per person, are asked of him before anything is built (security #461, #514).

Ended because: Ninad answered Orion's question on who assigns FDEs: FDEs stay service operators on prospects and projects alike, assigned by Carolverse estate operators

FDEs do POC in the POC zone, and are the service operators of the prospects and projects Carolverse estate operators assign them

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-185): "ruling - FDE's should be able to do POC in the POC zone. They also by default become the service operators for the prospects and estate operator for the projects assigned to them"; then, answering Orion: "Carolverse estate operators can assign FDEs to prospects and projects.. keep FDEs as service operators for both prospects and projects..". Orion asked (CLI-185): who assigns an FDE to a prospect or a project? Under #461 only the administrator, at root, makes someone a project's estate operator, and estate operators may appoint service operators; if an assignment made an FDE an estate operator by default, whoever assigns FDEs would be appointing estate operators. Options offered: split by role (Orion's recommendation), only the administrator for both, any estate operator for both. READ AS (Orion's reading, not his words): (1) a prospect's FDE gets a way onto Sandbox's machine that writes that prospect's POC section (design 728) and nothing else; (2) a Carolverse estate operator assigns an FDE to a prospect or a project, and the FDE is, by that assignment, its service operator -- never its estate operator; #461 stands unchanged (estate operators are appointed only by the administrator).

A client project's work comes only through pursuits Leo agrees in his chat; the hand-made Chowpatty pursuits retire and their proof steps join the Blueprint pursuit

Ruled 2026-11-16 by ninad -- in force

A CLIENT PROJECT'S WORK COMES ONLY THROUGH PURSUITS LEO AGREES IN HIS CHAT; THE HAND-MADE CHOWPATTY PURSUITS RETIRE (Ninad, CLI-186, 2026-11-17). THE QUESTION: he asked 'why do we need the chowpatty pursuit, if its supposed to be built by talking to Leo?'; Orion answered that by initiatives #798 and blueprint #642 a client project's pursuits are created by Leo from what the client agrees in his chat, that CAROL-PUR-0082 (Chowpatty as a prospect) and CAROL-PUR-0083 (Chowpatty as a client) were made by Orion sessions as a stopgap under #876 (CLI-158, split by CLI-185), and recommended: retire both hand-made Chowpatty pursuits, move their proof steps (the platform working for a real client: one real run both ways, onboarding steps recorded, Chowpatty's uses of the platform) into the Blueprint pursuit as its end-to-end proof, and let Chowpatty's own work (its first build, its proof of concept, its first agent, its prototypes) come only through pursuits agreed with Leo. HIS WORDS: 'yes, do that.' READ AS: #876's clause that a project's onboarding pursuit holds that project's use of the platform is retired in place for a client project; Blueprint holds the proof, and a client's own work waits for a pursuit Leo proposes and the client (its FDE until Leo faces clients, #801) confirms.

Only FDEs are kept to the prospects and projects they are assigned; estate operators work on every prospect

Ruled 2026-11-16 by ninad -- in force

HIS WORDS (Ninad, CLI-187): "generally speaking, only FDEs who are assigned to prospects and projects should be able to work on those. Currently, we dont have any FDEs, so we will not be able to test it out yet, because all work is done by carolverse estate operators"

Asked: its first criterion said an operator's Orion is refused another prospect's sections on Sandbox's machine, while blueprint #911 lets every carried operator login write every prospect's POC section, so the machine cannot refuse it (it does refuse every build section and making a new prospect's section, proven by doing). How should it be settled?

READ AS (Orion, apart from the words): being kept to the prospects and projects one is assigned binds FDEs (blueprint #939: a prospect's FDE, assigned by Carolverse estate operators); Carolverse estate operators work on every prospect, so their carried logins writing any prospect's POC section stands as #911 has it. The per-prospect limit is built and proven with the FDE's own way in, and cannot be proven until an FDE exists. 's first criterion is revised to what an estate operator's Orion is refused (a build section, a new prospect's section, the prospect's own folder) and keeps to the prospect it is asked about by its skill.

A Sandbox build is held to architecture compliance and a Sandbox proof of concept is exempt; reaching Sandbox's machine for the checks is a step of the Sandbox pursuit

Ruled 2026-11-17 by ninad -- in force

A SANDBOX BUILD IS HELD TO ARCHITECTURE COMPLIANCE AND A SANDBOX PROOF OF CONCEPT IS EXEMPT; REACHING SANDBOX'S MACHINE FOR THE CHECKS IS A STEP OF THE SANDBOX PURSUIT (Ninad, CLI-189, 2026-11-17). THE QUESTION: building (a project's delivery checks that cannot run hold its work back from acceptance, never pass), Orion found that Sandbox is a project whose code nothing on carol-vm can reach, while design 728 v1.4 section 10 counts a Sandbox build delivered once it is handed over to its prospect's build section; applied to Sandbox, the rule would hold every Sandbox proof of concept and build. Asked whether the rule covers client projects only or Sandbox too, HIS WORDS: 'sandbox has 2 parts, 1 for POC, and second for actual build.. the actual build is subject to architecture compliances, while the POC area is excempt.. if Carolverse can't currently reach Sandbox's machine to run those checks, then thats a fix that needs to go in the sandbox pursuit'. READ AS (Orion's reading, not his words): a Sandbox filing naming section build is a project's work under -- its design, architecture and ownership checks run on Sandbox's machine before acceptance, and while they cannot run it is held as unauditable after its hand-over; a Sandbox filing naming section poc is exempt from those checks and keeps design 728's hand-over rule; making Sandbox's code and registry reachable for the checks is filed as a step of CAROL-PUR-0042 (Prospects in Sandbox), not of the Blueprint pursuit.

A project is as lightweight as possible: built services are reused first, copies and per-project configuration are minimised, reuse work is platform enhancement paid by the service, and non-compliance is architectural debt

Ruled 2026-11-17 by ninad -- in force

Asked: Ninad answered "This needs a policy, lets discuss in chat", then: "I am referring to a policy on how services are graded.. and what is the preference ladder, the first project gives us the opportunity to grade services, and the process of grading the service should follow certain principles and policies". HIS WORDS: "principle - the projects should be as lightweight as possible, and the new build activities should be minimised as much as possible. This would mean using already built services in carolverse as much as possible, reducing copies of services, and enabling minimal configuration per project, thereby reducing the new build work as much as possible, and enabling the already built services to serve other project. If this entails adjusting, modifying or refactoring the existing services, that work should be taken up as platform enhancement work for enabling reusable services for the project. any non compliance to this should be identified as architectural debt and non compliance, and should be rectified as soon as possible. Any platform changes to the services are not billed to the project, the service charges to the project cover that expense.." Then: "create a pursuit to document and enforce the policies as stated".

READ AS (Orion's reading, not the administrator's words): (1) grading a service for a project prefers, in order, the estate's running service used as-is, then the least configuration, then the fewest copies; a new build for a project is the last resort; (2) work that adjusts, modifies or refactors an existing service so a project can reuse it is PLATFORM ENHANCEMENT, filed against that service, never a project build; (3) a copy, a one-project build or more configuration than needed is ARCHITECTURAL DEBT and non-compliance, named and rectified as soon as possible; (4) platform enhancement is never billed to the project -- the project's service charges cover it; (5) the policy words, the preference ladder and how a grade is decided are to be DRAFTED for his approval (governance #355): Orion's proposed ladder (as-is, config-only, estate-bound, clone, build for the project alone; seam-refactor is a defect state, not a rung) and its six edges (shared code never shared data; the improved service pays; no throwaway copy except as dated debt; a standing debt check; one-off needs moved into the platform when a second project needs them; settings hold facts, never behaviour) are proposals, not part of this ruling, until he approves them; (6) 's regrade of audit-core waits on that policy.

Carolverse grows to carry the projects it serves: one copy of a service's code runs on as many Carolverse machines as its load needs, each project's data in its own store, paid by the service charges, and the plan's Azure line counts those machines

Ruled 2026-11-17 by ninad -- in force

AS PROJECTS LEAN ON CAROLVERSE'S SHARED SERVICES, CAROLVERSE GROWS ONTO MORE MACHINES (Ninad, CLI-198, 2026-11-17; after blueprint #967, #470 and P.01.03.07.24). THE QUESTION: shown the status of CAROL-PUR-0084 (Lightweight projects, blueprint #967), Ninad asked: "this would mean that the burden would shift to carolverse.. how would carolverse scale to manage the load from the projects? should registry as a service should be also brought back into carolverse?". Orion proposed, as a table: "Where the load lands | On Carolverse's own machines, because every project shares the same running services. How it grows | 'One copy' means one copy of the code, not one machine. A busy service can run on more machines behind the same front door. What stays separate | Each project's data stays in its own store, so projects can't see each other and one database doesn't take everyone's writes. Who pays | The service charges each project pays cover the extra machines, as last ruling already says. What's still plans one machine per project and only one Carolverse machine. It should be re-based on Carolverse's load." HIS WORDS: "I like your proposal, can you fold them in the light projects pursuit?". READ AS (Orion's reading, not the administrator's words): (1) P.01.03.07.24's single copy is one copy of a service's CODE serving every project, never a cap of one machine: a service whose load needs it runs on more Carolverse machines behind one front door; (2) every shared service keeps each project's data in its own store -- shared code, never shared data (the edge proposed on #967 is ruled here for this); (3) the machines Carolverse adds to serve projects are paid through the service charges projects pay (cost-center #492, blueprint #967); (4) the investment plan's Azure line (leadership #965) counts Carolverse's machines from the load its shared services carry, beside the machines each project still needs; (5) the work is folded into CAROL-PUR-0084 (Lightweight projects).

A CLIENT'S MACHINE COMES FROM A PURSUIT LEO FILES WHEN THE CLIENT SIGNS ITS CONTRACT: A NEW MACHINE PROCURED, OR A SPARE ONE ALLOCATED IF ONE EXISTS; THE BLUEPRINT PURSUIT ENABLES LEO TO DO IT

Ruled 2026-11-17 by ninad -- in force

A CLIENT'S MACHINE COMES FROM A PURSUIT LEO FILES WHEN THE CLIENT SIGNS ITS CONTRACT: A NEW MACHINE PROCURED, OR A SPARE ONE ALLOCATED IF ONE EXISTS; THE BLUEPRINT PURSUIT ENABLES LEO TO DO IT.

ASKED: the estate's notice said ('Chowpatty's own machine is made with its operators' ways in once its contract is signed', in the Blueprint pursuit) was free to start once closed; Orion advised it wait, Chowpatty's contract being unsigned, and Ninad agreed to wait till the contract is signed.

HIS WORDS: "why is chowpatty asking 7285? what pursuit does it belong? it should be a pursuit filed by Leo when cleint signs up the contract, and should initiate the procurement of a new vm, or allocate an existing spare vm, if it already exists. Leo should be enabled to do this via blueprint pursuit".

READ AS (Orion's reading, not the administrator's words): (a) a client's own machine is not a standing step named for one client in the Blueprint pursuit: when a client signs its contract, Leo files a pursuit for that client, and that pursuit starts either the procurement of a new machine or the allocation of a spare machine that already exists; (b) the Blueprint pursuit holds the work that ENABLES Leo to do this for any client; (c) is to be re-read against this by the session working the Blueprint pursuit, and waits until then; (d) how a spare machine is known and allocated is not ruled here.

AS-IS IS THE DEFAULT; THE LEAST PREFERRED OPTION BUILDS ON CAROLVERSE'S ARCHITECTURE AND DESIGN PATTERNS; A SERVICE BUILT WITHOUT THEM IS NOT ALLOWED OUTSIDE A SANDBOX POC

Ruled 2026-11-17 by ninad -- in force

THE QUESTION: shown the trimmed reuse-ladder policy for -- a project is served by Carolverse's built services wherever they fit; for each requirement the most preferred option on the reuse ladder that fits is used (as-is, config-only, estate-bound, project extension, project-owned fork, project-only build); a less preferred option states why each more preferred one does not fit; platform enhancement is paid by the service it improves; architectural debt is rectified -- with the dictionary meanings, and asked to approve it.

HIS WORDS: "as-is is the default, but not always feasible.. least preferred is to use the architecture and design patterns of carolverse, only if all other options fail.. rebuilding a brand new service with no reference to carolverse architecture and design patterns is simply not allowed, except in the POC area of the sandbox.. does that clarify?"

READ AS (Orion's reading, not the administrator's words): (1) as-is is the default option for every requirement of a project, and a less preferred option is taken only where as-is is not feasible; (2) the least preferred option -- a build for the project, taken only when every other option fails -- is built to Carolverse's architecture framework and design patterns; (3) a service built new without reference to Carolverse's architecture framework and design patterns is not allowed anywhere, except in a prospect's POC section in Sandbox.

A prospect's reports are made in its POC section; their regular refresh runs only in its build section, by named agents through Leo, never on a person's login

Ruled 2026-11-17 by ninad -- in force

ASKED (CLI-207): Sahil (#1322) asked that Chowpatty's two weekly reports (Weekly Accounts and Deliveroo, Python scripts on his laptop) run on Sandbox every Monday with pages operators and Chowpatty's people can see, and that root set up a non-person runner login, a served page behind a sign-in, and a place for two keys (Airwallex, Gmail). Orion told Ninad no policy forbids a POC app Chowpatty's people use, and named what bears on it: #983 (Sandbox is a preview, not where the project runs), real keys in a POC area every operator can read, the Creed's ruling that operators' scheduled jobs run under non-person logins (CLI-546), and that only root makes a login on Sandbox's machine. HIS WORDS, in order: "Weekly reports on real data can also be read as a preview of the art of possible.. if one report is demonstrated, then all reports can be built on the actual project.. however, i dont know if root is necessary for that, as this is not related to operator, project or agent governance, or is it?"; "just like carolverse has users, chowpatty can also has users. Can operator's own login be used for scheduling the refresh?"; "ideally, chowpatty should have its agent built who will run the jobs, isnt it?"; "POC is a free form area for every operator to do whatever they want. Can operators (FDEs assigned to that prospect) not create their own agents and then use their logins to create scheduled jobs?"; then: "lets park that for now.. I think FDEs should be able to create reports in POC zone, but regular refresh should happen in the build area through named agents following the standard process (via Leo). when Sahil creates the agents, who will ask for the admin approval to admin Orion instance? - Leo, Elrond, Athena, Orion operator instance. Will that agent have a mind, and azure id? would it be possible that Athena creates about 10 agents upfront in "bench" who can be deployed immediately to the projects on admin's approval? who names the agents - option 1 - Athena when agent is created? option 2 - Leo in consultation with client? option 3 - Athena when agent is created, and Leo gives an alias based on client's preference?". Orion answered from the law (Athena asks him, #544; an Azure identity always, P.01.03.08.01, a mind only if he decides; the bench exists, #32, #896, each creation and each deployment his approval) and recommended option 3 with a bench of ten, and recording the report rule. HIS WORDS: "sounds good".

READ AS (Orion's reading, not the administrator's words): (1) an FDE assigned to a prospect may create reports in its POC section, on real data, as a preview of what is possible; one report demonstrated is enough, and the rest are built on the actual project; (2) a report's regular, scheduled refresh runs only in the prospect's build section, by a named agent the standard process gives it -- asked for through Leo, approved by the administrator, created or deployed by Athena -- under that agent's own login; never under an operator's or an FDE's own login (the Creed, CLI-546); (3) a project's people, Chowpatty's included, are users of its apps, as Carolverse's are, never logins on its machine; (4) letting Hagrid make a runner login per prospect for scheduled POC work is PARKED, not ruled; (5) a client's live keys are never placed in a POC section.

Athena names an agent when she creates it, and Leo gives it an alias the client prefers; the name stays its identity on root's list and in Azure

Ruled 2026-11-17 by ninad -- in force

ASKED (CLI-207): Sahil (#1322) asked that Chowpatty's two weekly reports (Weekly Accounts and Deliveroo, Python scripts on his laptop) run on Sandbox every Monday with pages operators and Chowpatty's people can see, and that root set up a non-person runner login, a served page behind a sign-in, and a place for two keys (Airwallex, Gmail). Orion told Ninad no policy forbids a POC app Chowpatty's people use, and named what bears on it: #983 (Sandbox is a preview, not where the project runs), real keys in a POC area every operator can read, the Creed's ruling that operators' scheduled jobs run under non-person logins (CLI-546), and that only root makes a login on Sandbox's machine. HIS WORDS, in order: "Weekly reports on real data can also be read as a preview of the art of possible.. if one report is demonstrated, then all reports can be built on the actual project.. however, i dont know if root is necessary for that, as this is not related to operator, project or agent governance, or is it?"; "just like carolverse has users, chowpatty can also has users. Can operator's own login be used for scheduling the refresh?"; "ideally, chowpatty should have its agent built who will run the jobs, isnt it?"; "POC is a free form area for every operator to do whatever they want. Can operators (FDEs assigned to that prospect) not create their own agents and then use their logins to create scheduled jobs?"; then: "lets park that for now.. I think FDEs should be able to create reports in POC zone, but regular refresh should happen in the build area through named agents following the standard process (via Leo). when Sahil creates the agents, who will ask for the admin approval to admin Orion instance? - Leo, Elrond, Athena, Orion operator instance. Will that agent have a mind, and azure id? would it be possible that Athena creates about 10 agents upfront in "bench" who can be deployed immediately to the projects on admin's approval? who names the agents - option 1 - Athena when agent is created? option 2 - Leo in consultation with client? option 3 - Athena when agent is created, and Leo gives an alias based on client's preference?". Orion answered from the law (Athena asks him, #544; an Azure identity always, P.01.03.08.01, a mind only if he decides; the bench exists, #32, #896, each creation and each deployment his approval) and recommended option 3 with a bench of ten, and recording the report rule. HIS WORDS: "sounds good".

READ AS (Orion's reading, not the administrator's words): option 3. (1) Athena names every agent when she creates it; that name is its identity on root's master list of agents (#896) and in Azure (P.01.03.08.01), and it does not change when the agent is rented to another project; (2) for a project, Leo agrees with its client (its FDE until Leo faces clients, #801) an alias by the client's preference, shown to that project's people; the alias is recorded with who agreed it, and changing it does not touch root's list; (3) this supersedes #867's 'Leo names a project's agents with its forward deployed engineer or its user' -- what Leo agrees with them is now the alias; Athena still creates and the administrator still approves every agent (#31, #35, #544).

LEO DECIDES WHO DOES A PROJECT'S WORK BY AGREED POLICIES, NOT CASE BY CASE; WHEN UNSURE HE ASKS THE CAROLVERSE ESTATE OPERATORS THROUGH ORION (Ninad, CLI-205, 2026-11-17; after support #990, agent-resources #992, blueprint #984, #967)

Ruled 2026-11-17 by ninad -- in force

THE QUESTION: after ruling on single cases -- customer support (support #990), the investment plan and backup (agent-resources #992) -- and shown that the services handbook and the dictionary disagree on the investment plan's grade, Ninad was asked whether Orion should take the stalled re-read of every track.

HIS WORDS: "I dont want to give rulings on individual cases, i want to derive patterns, principles, and policies that will help Leo decide whether to request a new agent to athena or use existing carolverse agents, agreed policies will have to be followed in each case.. if leo is not sure, he should ask the carolverse estate operators via orion.."

READ AS (Orion's reading, not the administrator's words): (1) whether a project's work is done by a new agent Leo requests from Athena or by Carolverse's existing agents is decided by agreed policies -- patterns and principles derived from the cases, not a ruling per case; (2) once agreed, those policies are followed in every case; (3) when Leo cannot tell how they apply, he asks the Carolverse estate operators through Orion before he decides; (4) the rulings on single cases already made (#990, #992) are the cases the policy is drawn from, not the policy itself.

ESTATE-BOUND IS DROPPED AS A REUSE GRADE: SUCH A SERVICE IS AS-IS OR CONFIG-ONLY, AND THE PROJECT'S OWN RECORDS EXISTING FIRST IS A PRECONDITION, NOT A GRADE

Ruled 2026-11-17 by ninad -- in force

THE QUESTION: drafting the reuse-ladder policy, Orion asked whether to keep 'estate-bound' -- the grade for a service whose work is about an estate (monitoring, auditing, governing, billing it), reused by pointing the same machinery at a project's own records, which refuses until the project has those records and deploys once it does (P.01.03.07.13 lets a track be graded estate-bound only under an explicit ruling on its handbook entry). Ninad asked "should estate bound be even an option?", then what it means and how it differs from project-only build and project-owned fork, then: "so what you are saying is that estate bound and as-is are the same thing?". Orion answered that it is not a separate way of serving a project: the running service reaching the project's own store is as-is, a copy set up on the project's own records is config-only -- both already on the ladder -- and its one real point, that the project's own records must exist first, is a precondition, not a grade.

HIS WORDS: "ok, then drop it.."

READ AS (Orion's reading, not the administrator's words): (1) estate-bound is no longer a reuse grade or an option on the reuse ladder, which reads: as-is, config-only, project extension, project-owned fork, project-only build; (2) a track graded estate-bound today is re-graded as-is (the running service reaches the project's own store) or config-only (a copy set up on the project's own records), by the ladder like every other track; (3) that a service waits until the project's own records exist is a precondition of serving it, recorded with the track, never a grade; (4) the rule that an estate-bound grade needs an explicit ruling on the track's entry falls away with the grade.

EVERY SERVICE CAROLVERSE OR A PROJECT BUILDS IS GRADED; THE APPROVED REUSABLE SERVICES ARE A CORE LIST LIKE THE AGENTS; A NEW CAROLVERSE SERVICE, OR A PROJECT'S SERVICE MADE REUSABLE, NEEDS THE ADMINISTRATOR'S APPROVAL AND CARRIES NO DATA OR IP OF ITS PROJECT; THE CONTRACT SAYS SO (Ninad, CLI-205, 2026-11-18; folded into CAROL-PUR-0084; after blueprint #967, #995, P.01.03.07.27, #909)

Ruled 2026-11-17 by ninad -- in force

CONTEXT: after approving the reuse-ladder policy (P.01.03.07.27) and asking to see every service's grade and agent decision, Ninad added, unprompted:

HIS WORDS: "This gadation should extend to all the services that carolverse builds on the estate, and even extending to the projects.. this means, if a service built for another project is deemed reusable for carolverse, it should be made reusable, by admin approval. Creation of any new service in carolverse needs admin approval. List of approved re-usable services are maintained in core, like agents. Making a service reusable should not involve keeping any data or IP from the parent project. Contract should mention that the platform may decide to transform certain project services re-usable in carolverse. This activity should be a responsibility of an existing service.. fold this requirement into the lightweight projects pursuit and show me the new status of the pursuit.."

READ AS (Orion's reading, not the administrator's words): (1) every service Carolverse builds on the estate, and every service a project builds, is graded on the reuse ladder; (2) root's core keeps the one list of approved reusable services, as it keeps the master list of approved agents (#909), and only the administrator changes it; (3) no new Carolverse service is created until the administrator approves it onto that list; (4) a service built for a project that is judged reusable is made a Carolverse service only on the administrator's approval, and it carries none of the parent project's data or intellectual property -- only the generic capability, rebuilt or cleaned to Carolverse's architecture framework; (5) every client contract states that the platform may turn certain of the project's services into reusable Carolverse services; (6) the work is carried by an existing service -- Blueprint, Leo's service, which already owns the service reusability arc (CAROL-ARC-0001) -- and the contract wording by the service that keeps contracts (Marco's Contract Management, Cost Center); (7) all of it is part of the Lightweight projects pursuit (CAROL-PUR-0084).

For each service a project uses, the Carolverse service owner sets the standards, defaults and machinery and the project's own agents give the content; agent wishes go to Albus or the mapped FDE's Orion; every purchase goes through Marco and is invoiced by Midas; the constitution is Carolverse's alone

Ruled 2026-11-18 by ninad -- in force

WHO DOES A PROJECT'S WORK, SERVICE BY SERVICE.

THE QUESTION: Orion showed eight service tracks whose agent decision the draft table left unsure, each with its owner read from the registry and a recommendation, and asked him to approve or change each. HIS WORDS, item by item:
1. Agent chat (Athena): recommended "Athena runs the chat system for every client. Each reply still speaks as whichever agent is addressed, including the client's own agents" -- "agreed". 2.
3. Constitution (Orion): recommended "Orion keeps and shows a client's rulebook. The client's people decide what the rules say" -- "Constitution is not client facing.. constitution guides the policies and policies have to comply to the constitution. Constitution is owned and managed by carolverse, policies can be carolverse policies or project's own policies.. if project policies conflict with carolverse policies, carolverse policies win, and conflicts are brought to attention to the mapped FDEs orion instance by the service owner". 4. Procurement (Marco): recommended "The client's own agent buys for its business, using Marco's procurement service. Marco himself buys only Carolverse's machines" -- "any purchases get routed via Marco, because he will hold the payment gateways.. Service owners of respective projects are charged for the purchases. Midas decides whether the purchace is carolverse or project, and invoicing happens accordingly.. carolverse, prospects and projects get invoices from Midas". 5. Treasury (Midas): recommended "Midas opens and reads a client's bank accounts only as custodian. Money decisions belong to the client's own agent" -- "Yes, or should it be Marco for purchases? Midas owns the invoicing and budgets for carolverse and projects". 6. User onboarding (Jarvis): recommended "Jarvis admits a client's users, as security plumbing. The client says who gets in" -- "yes, following the policies and the handbooks for the respective projects. Jarvis asks the right questions to Leo to be asked to client / FDE, to ensure that the onboarding of the users follow the service protocols, like any other service owner". 7. Public encyclopedia (Loki, two tracks): recommended "Not offered by default. If a client asks for one, the client's own agent writes it, using Loki's service" -- "right, the standards and defaults are dictated by Loki, content is given by respective project agents". 8. Research directory (Obi-Wan): recommended "The client's own agent keeps the client's research contacts in its own store, using Obi-Wan's service" -- "right.. I think we have a pattern here that is applicable across the board using the reusable service policies as the reference".

Themis, Prometheus and Heimdall each serve the platform and every project: Themis audits compliance in three parts, Prometheus improves quality in two, Heimdall secures in two

Ruled 2026-11-18 by ninad -- in force

WHAT THEMIS, PROMETHEUS AND HEIMDALL DO FOR THE PLATFORM AND FOR PROJECTS.

THE QUESTION: the service-grade table left two tracks, audit-core (Themis) and quality-management-core (Prometheus), needing a code check before their grade could be recorded. HIS WORDS, asking Orion to pick it up:
"1. Themis is an internal auditor for the projects, to ensure regulatory compliance, architecture and design compliance, policy and constitution compliance.. there are 3 parts to Themi's service.. one is to ensure that the platform complies to carolverse policies, second is to ensure that the projects comply to carolverse policies, and third is to offer the compliance as a service to projects to ensure that their policies are followed properly.. 2. Prometheus has 2 functions, first to ensure that the services are continuously improving in their quality, and second, to provide quality improvement as a service to the projects to ensure that they are continuously improving their quality"
Then, on security-core: "for heimdall, he needs to manage security of the platform and provide security as a service to the projects.. continue the work"

READ AS (Orion's reading, not the administrator's words): (1) THEMIS, three parts: (a) the platform complies with Carolverse's policies; (b) every project complies with Carolverse's policies; (c) compliance as a service: a project's own policies are followed properly; she covers regulatory, architecture and design, and policy and constitution compliance; (2) PROMETHEUS, two parts: (a) Carolverse's services keep improving in quality; (b) quality improvement as a service: each project's services keep improving; (3) HEIMDALL, two parts: (a) he manages the platform's security; (b) security as a service: each project is secured by him; (4) each is run by the estate's own service for every project, never a copy each project sets up and runs for itself -- the target grade of all three tracks is as-is (P.01.03.07.27); by blueprint #1002 the owner sets the standards and defaults and runs the machinery, and the project's own agents give the content (the project's own policies, its quality aims, its security content such as who may enter and what is sensitive); a project policy that conflicts with a Carolverse policy loses, and the service owner brings it to the mapped FDE's Orion (#1002 item 3); (5) a guard that can only act on the machine it guards (a sign-in alarm reading that machine's sign-ins, an owner hold set on that machine's files) runs on each project's machine as Heimdall's own, placed there with the machine and reporting to him; it is not a project's copy.

WHAT THE CODE SHOWS (Orion's check, CLI-221): none of the three can do part (b) or (c) today. Themis's 23 droid files and Prometheus's 2 take no project and cannot reach a project's machine; their scheduled audits and the quality loop read only the platform's records. Heimdall's 11 droid files watch only the platform machine (its sign-ins, its owner holds, its entry list, its credentials). The estate's delivery checks run a project's checks on its own code and registry but refuse a project recorded on another machine. So audit-core and quality-management-core are graded seam-refactor and security-core moves from config-only to seam-refactor, each with target as-is; the fixes are filed in CAROL-PUR-0084.

A prospect sees its POC as a web page through one read route: carol-vm pulls the POC's public folder through the read-only poc command and serves the copy behind Cloudflare Access

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229, operator communication #1495): 'Sahil's ask #1495 is about showing a prospect its demo (POC) as a web page, which needs a ruling. Sandbox has no web address today. Sahil proposes that Sandbox serve a 'public' folder per prospect on a port only carol-vm can reach, with carol-vm's Cloudflare front and Cloudflare Access (named emails plus operators) giving each prospect one address. My recommendation uses one read route instead: carol-vm fetches the public folder through the same read-only 'poc' command already planned for Migration (7483) and serves that copy behind Cloudflare Access. That way Sandbox gets no new port or service, and there is one way in to design, not two. Which way?' Ninad answered: 'One read route (Recommended): carol-vm pulls the prospect's public folder through 7483's read-only command and serves the copy behind Cloudflare Access. Sandbox gets no new port. The page refreshes when it is pulled again.'. READ AS: each prospect's POC section has a public folder; carol-vm fetches that folder through the same read-only route out of Sandbox that builds (one read route, designed once for both the POC page and Migration's read) and serves the copy at one address per prospect behind Cloudflare Access (the prospect's named people by email, and operators); Sandbox opens no new port and runs no web server; no logins on either machine for prospects.

Hagrid owns the read route out of Sandbox for a prospect's POC page; Heimdall owns who may open each prospect's page

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229, operator communication #1495): 'Who should own showing a prospect its POC page (#1495)? Blueprint already makes Hagrid responsible for machines and Heimdall for stakeholder access (who may open what).' Ninad answered: 'Hagrid + Heimdall (Recommended): Hagrid owns the machine side (the read route out of Sandbox). Heimdall owns who may open each prospect's page (Cloudflare Access by email).'. READ AS: the machine side of showing a prospect its POC page (the read route out of Sandbox's machine and the pulled copy) is Hagrid's; the access list deciding who may open each prospect's page (Cloudflare Access by email) is Heimdall's.

Sahil is assigned as Chowpatty's FDE, so the estate recognises him as its FDE

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229), after Orion filed so any estate operator may agree a client project's module criteria, and explained: 'The rule is that Leo files nothing for building until a person has explicitly agreed the exact success criteria ... silence or a vague "looks fine" doesn't count as agreement. It had to be you because of a gap: Leo's check accepts only you or the person recorded as having asked for the project. Chowpatty has no registered users yet, so Sahil's agreement was refused this morning. The fix I filed lets Sahil agree from now on.' Ninad answered: 'that is the wrong way to fix things, 1. Sahil should be assigned as the FDE on chowpatty, so that the system recognizes him as FDE 2. Sahil should be enabled first to approve the pursuits filed by Leo. Leo has to communicate the plan back to Sahil and "get" his approval 3. Leo should not treat silence as the approval, but vague "looks fine" can be construed as approval 4. Sahil should be, in the capacity of the estate operator, not FDE, be able to extend Leo's budget or any other budget as required temporarily or permenantly, because that is not a root / core governed task'. READ AS (point 1): Sahil is recorded as the forward deployed engineer of Chowpatty through the one project-assignment record, and every reader that asks who a project's FDE is answers Sahil for Chowpatty; an estate operator may hold the FDE role on a project (blueprint #939 named FDEs as service operators; this adds that an estate operator may be assigned as one). Approvals Leo takes for a project are taken from its FDE in that capacity, not from the estate operator rung.

Leo communicates each plan he files back to the project's FDE and gets the FDE's approval, the pursuits first

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229), after Orion filed so any estate operator may agree a client project's module criteria, and explained: 'The rule is that Leo files nothing for building until a person has explicitly agreed the exact success criteria ... silence or a vague "looks fine" doesn't count as agreement. It had to be you because of a gap: Leo's check accepts only you or the person recorded as having asked for the project. Chowpatty has no registered users yet, so Sahil's agreement was refused this morning. The fix I filed lets Sahil agree from now on.' Ninad answered: 'that is the wrong way to fix things, 1. Sahil should be assigned as the FDE on chowpatty, so that the system recognizes him as FDE 2. Sahil should be enabled first to approve the pursuits filed by Leo. Leo has to communicate the plan back to Sahil and "get" his approval 3. Leo should not treat silence as the approval, but vague "looks fine" can be construed as approval 4. Sahil should be, in the capacity of the estate operator, not FDE, be able to extend Leo's budget or any other budget as required temporarily or permenantly, because that is not a root / core governed task'. READ AS (point 2): the first thing enabled is the project's FDE (Sahil for Chowpatty) approving the pursuits Leo files; Leo brings the plan back to the FDE in his chat and records the FDE's approval before it goes on, the same for module criteria and the roadmap after the pursuits.

Leo never treats silence as approval, and a plain 'looks fine' counts as approval

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229), after Orion filed so any estate operator may agree a client project's module criteria, and explained: 'The rule is that Leo files nothing for building until a person has explicitly agreed the exact success criteria ... silence or a vague "looks fine" doesn't count as agreement. It had to be you because of a gap: Leo's check accepts only you or the person recorded as having asked for the project. Chowpatty has no registered users yet, so Sahil's agreement was refused this morning. The fix I filed lets Sahil agree from now on.' Ninad answered: 'that is the wrong way to fix things, 1. Sahil should be assigned as the FDE on chowpatty, so that the system recognizes him as FDE 2. Sahil should be enabled first to approve the pursuits filed by Leo. Leo has to communicate the plan back to Sahil and "get" his approval 3. Leo should not treat silence as the approval, but vague "looks fine" can be construed as approval 4. Sahil should be, in the capacity of the estate operator, not FDE, be able to extend Leo's budget or any other budget as required temporarily or permenantly, because that is not a root / core governed task'. READ AS (point 3): Leo records an approval only from something the person said in his chat; no answer is never approval; a short, informal yes such as 'looks fine' to the plan he put to them is approval, and the person need not repeat the criteria word for word. This replaces the rule written into Leo's agreement tool by ('silence or a vague answer is never agreement') for its 'vague answer' half.

Any work that stands in the way of proof-of-concept activity is the highest priority

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229): Orion had filed (Sahil assigned as Chowpatty's FDE), (Leo brings each plan back to the project's FDE and records the approval, pursuits first) and (an estate operator raises any budget for a time or for good) after Ninad's rulings blueprint #1011-#1013 and cost-center #1014. Ninad answered: 'implement these on priority, so that Sahil and continue building the POC.. any work that impedes the POC activity becomes highest priority'. READ AS: 7496, 7497 and 7499 are built now, ahead of other work, so Sahil can continue building Chowpatty's proof of concept; and as a standing rule, any work that stands in the way of a prospect's proof-of-concept activity is the highest priority, taken before other pending work.

Estate-bound tracks stay estate-bound; the as-is or config-only goal is for offered tracks

Ruled 2026-11-18 by ninad -- in force

Ninad (CLI-239): "the refactoring of several services needs to be filed so that we have only as-is and config-only grades". Asked: 9 tracks are graded estate-bound (Carolverse's own chat persona, Orion's windows, operator onboarding, the constitution and its logbook and videos, the build stories and their videos), not offered to projects by ruling blueprint #1002 -- should they also be refactored to as-is or config-only? Ninad: "No, leave them". Asked which pursuit holds the seven seam-refactor tracks' work: Ninad: "I will ask the other session working on light projects to file this, pls ignore". READ AS (Orion, not Ninad's words): only tracks offered to projects must reach as-is or config-only; the nine estate-bound tracks keep their grade under #1002; the seam-refactor filings belong to the Lightweight projects session (CAROL-PUR-0084), not CLI-239.

Only as-is and config-only grades for offered services

Ruled 2026-11-18 by ninad -- in force

Ninad, CLI-241: "the refactoring of several services needs to be filed so that we have only as-is and config-only grades, file the necessary initiatives". READ AS (Orion): every service track offered to projects is refactored until Leo s Service Track Handbook grades it as-is or config-only; seam-refactor is not an end grade. Filed: (agent-resources-chat), 7523 (monitoring-core), 7524 (sales-core); already filed: 7492-7494 (audit, quality-management, security cores), 7352 (backup-upkeep, its goal to be revised to as-is).

The roadmap dictates the build; a POC move serves agents and droids already on it

Ruled 2026-11-18 by ninad -- in force

Ninad, CLI-236, 2026-11-18, on what a prospect's proof-of-concept work becomes once Migration rebuilds it: "the roadmap should be in synch with the build, but roadmap should dictate the build and not the other way round". So a POC move never adds an item to the prospect's roadmap: Leo proposes the move naming the agents and droids already on the roadmap that the POC work delivers, Migration's rebuild is built under those names, and work the roadmap does not plan is put on the roadmap first, by its approver, before it is moved. Carolverse's terms are services, agents and droids, not modules.

A reuse grade belongs to a service's track, kept in the Service Track Handbook; root's list of approved services carries no grade

Ruled 2026-11-19 by ninad -- in force

GRADES ARE PER TRACK; ROOT'S LIST OF APPROVED SERVICES CARRIES NO GRADE

THE QUESTION: entering the 22 running services on root's list of approved services, whose entry has room for ONE grade per service while the Service Track Handbook grades each track (11 services have tracks of different grades, e.g. Audit: core seam-refactor, the others as-is) -- options: the heaviest track's grade (recommended), a grade only when all tracks agree, the core track's grade. HIS ANSWER: "grading should happen at track level". Asked then whether the list carries no grade (grades stay per track in the handbook; recommended) or holds each service's tracks with their grades (a change to root's writer): "No grade on the list".

READ AS (Orion's reading, not the administrator's words): (1) a reuse grade belongs to a service TRACK and is kept only in the Service Track Handbook; a service has no single grade and none is derived from its tracks; (2) root's list of approved services says which services the administrator approved, who owns each and its purpose -- it carries no grade: a service is entered with its grade empty, and the estate's reader does not show that as a missing fact; (3) the grade field of root's writer is unused from now and is retired at the writer's next revision; (4) 7516's criterion naming a grade is amended to match.

Root's writer accepts Orion, as well as an approved agent, as the owner of a service on the list of approved services; Constitution keeps Orion as its owner

Ruled 2026-11-19 by ninad -- in force

ORION MAY OWN A SERVICE ON ROOT'S LIST OF APPROVED SERVICES

THE QUESTION: the records name Orion as owner of the Constitution service -- the service, its five tracks and its apps Constitution, Policies and Session Induction -- but root's writer accepts only an approved agent on root's list as a service's owner, and Orion is a board row, not on the list of approved agents. First answer: "Constitution should already have an owner, all apps have owners". Asked which gives way -- root's writer accepts Orion (a root step), Constitution moves to an approved agent, or the other 21 are entered now and this is decided later: "Writer accepts Orion".

READ AS (Orion's reading, not the administrator's words): (1) Constitution keeps Orion as its owner; (2) root's writer, and the estate's one reader of the list, accept Orion (agt_023, the operators' CLI agent, a board row) as a service's owner as well as an approved agent on root's list -- no other row that is not an agent; (3) Orion is NOT entered on the list of approved agents: he is not an agent (security #274: "orion is not a person, he is the cli agent"); (4) the change to root's writer is its own initiative and needs a root step; Constitution is entered on the list once it lands.

New data work becomes tracks within an existing service, not new services

Ruled 2026-11-19 by ninad -- in force

Told that Data Quality, Data Catalog and Data Traceability had been created as three new services, Ninad: "creating a new service would mean we need a corresponding ops and operator, which is a cost, and impacts the plan.. can this be made as a track within existing service?" Amends governance #1051 and cost-center #1053 (read 'services' there as tracks).

Leo keeps each client informed of its project's estimated cost and time, takes its periodic approvals and re-prioritises its pursuits from its feedback

Ruled 2026-11-19 by ninad -- in force

Ninad: "these estimates should be accesible to Leo and he should be able to communicate them with client / FDE via Leo chat or Leo window.. It is his duty to keep the client informed from time to time of the estimated cost and time and take periodic approvals, keep reminding clients of the same, and update any changes to the agreed timelines and cost, and manage expectations according.. this would also mean trimming down the pursuits, re-prioritizing work based on client feedback and updating pursuits to optimize the value delivered.. this requirement should be added to the blueprint pursuit"

Every contract states its pricing model: tokens plus margin -- pay as you go or a fixed subscription, the default, agreed when a prospect subscribes to Blueprint -- or, for a client, a price for scope or for value

Ruled 2026-11-19 by ninad -- in force

EVERY CONTRACT STATES ITS PRICING MODEL (Ninad, 2026-11-19, CLI-250, refining leadership #796).

ASKED: whether a contract's content is decided (Orion found it states scope, methods, quality, timeline, an estimated cost and the standard payment terms, #795 -- but no pricing model).

HIS WORDS: Ninad: "is the content on the contract decided yet? contract should specifiy what is the pricing model - Tokens plus margin (which is the default for the sandbox as a prospect or self serve user), price for scope or price for value.. the token plus margin will also have 2 types - pay as you go or fixed subscription - for users to choose from.. this should be decided and agreed during subscription to blueprint as a prospect.. and later, as a client, it can continue the same way, but clients also get the other 2 options if they prefer those.."

READ AS (Orion's reading, not his words): (a) a contract names one pricing model: tokens plus margin, price for scope, or price for value; (b) tokens plus margin has two types the user chooses between -- pay as you go, or a fixed subscription; (c) tokens plus margin is the default for a Sandbox prospect and a self-serve user, and its type is chosen and agreed when the prospect subscribes to Blueprint; (d) a client may keep that model after signing, and only a client is also offered a price for scope or a price for value.

A self-serve user subscribes on the default pricing model, tokens plus margin, through one seamless journey from Carol's chat to Leo

Ruled 2026-11-19 by ninad -- in force

A SELF-SERVE USER SUBSCRIBES ON THE DEFAULT PRICING MODEL, TOKENS PLUS MARGIN, THROUGH ONE SEAMLESS JOURNEY FROM CAROL'S CHAT TO LEO (Ninad, 2026-11-19, CLI-258, refining blueprint #1065).

ASKED: told that CLI-258 had claimed, where a prospect chooses and agrees its pricing model when it subscribes to Blueprint.

HIS WORDS: "prospect doesnt have much choice apart from going with the default - which is token + margin.. but I need you to build the interface where users subscribe to a plan.. 1. subscribe to blueprint service and accept the terms (contract) 3. provide payment details 4. approve the payment 5. provision a folder for the user on one of the available sanboxes 6. include leo chat in the list of apps that user can access, and provide them the link to leo chat"; "I want you to think from users pespective, and provide the user a journey that involves minimum friction"; "this should be a seamless end to end process, starting with some unknown user somehow getting access to carol chat on logging in.. starts chatting with carol.. then user subscribes to blueprint and ends up talking with Leo".

A service is something an agent owns and makes available to any other agent; its tracks are what it offers, never who uses it -- Agent Chat becomes capability tracks, with Voice chat among them

Ruled 2026-11-20 by ninad -- in force

A SERVICE IS WHAT AN AGENT OWNS AND OFFERS TO ANY OTHER AGENT; ITS TRACKS ARE WHAT IT OFFERS, NEVER WHO USES IT (Ninad, CLI-265, 2026-11-20; on CAROL-PUR-0084; after blueprint #967, #1002, P.01.03.07.27, P.01.03.02.16)

THE QUESTION: Ninad asked to "add a new service called voice chat, as a track, not independent service" and to fold it into this pursuit; asked where it belongs, he asked whether it should be part of shared services. Orion answered that shared services are code building blocks with no tracks or budgets, and that the three voice chat needs -- Live Voice Line, Outgoing Voice Gatekeeper, Talk-To-Me Chat -- are Galadriel's, who also owns Agent Chat. HIS WORDS: "Yes, I am inclined to think that Agent chat is the right place, but its currently structured around individual agents, which is not right.. a Service be definition is something that an agent owns and makes it available to any other agent as needed.. what do you think?". Orion agreed and proposed four parts; he answered "Yes, all four".

READ AS (Orion's reading, not the administrator's words): (1) a service is something an agent owns and makes available to any other agent as needed; its tracks are what it offers -- a capability -- never one per agent that uses it; (2) Agent Chat (Galadriel's) is restructured into capability tracks: Text chat, taking in the per-agent tracks Leo Chat, Carol Chat, Sam Chat and Orion Chat, plus the messaging channels it already carries; each conversation still bills the budget of the agent spoken to; (3) Voice chat is a new track of Agent Chat, born graded on the reuse ladder with its reason and its agent decision, for the administrator's approval; (4) voice chat is built by reusing Galadriel's three shared voice building blocks as they are, nothing copied, so a person can talk to any agent and hear the reply. All four are filed in CAROL-PUR-0084.

The Voice chat MVP's sandbox project is MBank, built through Leo's door as the Blueprint service does it; its agent's details are asked by Leo in his chat (or the FDE's backdoor channel); Ninad is MBank's FDE

Ruled 2026-11-20 by ninad -- in force

MBANK IS BUILT THROUGH LEO'S DOOR, LEO ASKS FOR THE AGENT'S DETAILS, AND NINAD IS ITS FDE (Ninad, CLI-266, 2026-11-20; on CAROL-PUR-0092, fast track; after whatsapp #1084)

THE QUESTIONS AND HIS WORDS: asked which project the new sales agent belongs to (a new 'Sales Voice Demo' project on the sandbox machine, or the existing Sandbox project): "Call the project "MBank", use Leo's door to build it as per the blueprint service.. if some part of blueprint service is not yet built, build what is necessary from blueprint pursuit to enable this MVP". Asked what the new sales agent should be called (Nova, Vega or Iris): "These requirements should be captured via Leo chat, following the protocols. Leo should ideally provide these options to the user / client / FDE and prompt these questions at the right time.. if the backdoor channel is used by the FDE to feed the requirements, these questions and options should be prompted from the backdoor channel by Leo to the FDE.. I will act as FDE for this MVP, so please assign me accordingly, just like Sahil as FDE is assigned to chowpatty (hopefully he is, but if not, pls assign him accordingly)".

READ AS (Orion's reading, not the administrator's words): (1) the new project is named MBank and is made through Leo's door exactly as the Blueprint service makes a project (registration as a prospect, its proof of concept in Sandbox, the administrator's approval at root's list), never by hand; (2) where a part of the Blueprint service this needs is not yet built, the part needed for this MVP is built from the Blueprint pursuit; (3) the agent's details (its name and the rest of what a new agent needs) are not chosen by Orion: Leo asks them in his chat, offering options and asking each question at the right point of the protocol, to the user, the client or the FDE; when the FDE feeds requirements through the backdoor channel, Leo asks the FDE there; (4) Ninad is MBank's FDE for this MVP, assigned the way Sahil is Chowpatty's FDE; if Sahil is not yet assigned as Chowpatty's FDE, he is assigned.

An estate operator may assign themselves as a project's or prospect's FDE

Ruled 2026-11-20 by ninad -- in force

AN ESTATE OPERATOR MAY ASSIGN THEMSELVES AS A PROJECT'S FDE (Ninad, CLI-266, 2026-11-20; on CAROL-PUR-0092; after blueprint #1011, #1085)

THE QUESTION AND HIS WORDS: asked how he becomes MBank's FDE, told that the estate refuses anyone assigning themselves (agent-resources #254, cited by the assignment code) so his own Orion cannot make him MBank's FDE, with the options 'Sahil assigns me' or 'Allow self-assignment': "Estate operators can assign FDEs to projects as per the current ruling.. So I will be assigning myself as FDE as estate operator of carolverse.. Also, when I feed the requirements, I am the FDE, but when I approve the project, service and agent, I am the admin. It is important to capture my efforts spent as FDE on the project, the cost of which will get attributed to the project vs my time in approvals as admin, to be assigned as cost to the project again vs my time as estate operator for building the platform, the cost of which gets assigned to the platform build (carolverse - blueprint service). This distinction should be captured centrally and used by all relevant agents for invoicing, planning, motivations, cost tracking etc"

READ AS (Orion's reading, not the administrator's words): (1) an estate operator may make THEMSELVES the FDE of a project or prospect of a project they are an estate operator of, from their own Orion session, recorded like any FDE assignment; (2) the assignment code's refusal 'nobody assigns themselves' for FDE roles (it cites agent-resources #254, which is about assigning arcs and pursuits to SERVICE operators and is itself superseded) no longer applies to an estate operator assigning themselves as FDE; a service operator still never assigns themselves; (3) Ninad assigns himself as MBank's FDE once MBank is registered.

The administrator of every prospect, project, POC and self-serve initiative is Ninad and is never asked; the estate operators are the same across the estate; FDEs are mapped to their projects

Ruled 2026-11-20 by ninad -- in force

THE ADMINISTRATOR IS ALWAYS NINAD, THE ESTATE OPERATORS ARE THE ESTATE'S, AND FDES ARE MAPPED PER PROJECT (Ninad, CLI-269, 2026-11-20; on CAROL-PUR-0093, MBank MVP)

THE QUESTION AND HIS WORDS: Leo, through the FDE backdoor, asked "Who should be MBank's administrator?" with the options "You, Ninad -- recommended" or someone else. Ninad: "admin is always Ninad, for all prospects, projects, pocs, and self serve initiatives, make this as the ruling. Estate operators will also remain the same across the estate.. FDEs are mapped to respective projects..".

READ AS (Orion's reading, not the administrator's words): (1) the administrator of every prospect, project, proof of concept and self-serve initiative is Ninad -- the administrator root's record names -- so Leo never asks who the administrator is: a registration names him from root's record; (2) the estate operators of every project are the estate's own estate operators, the same across the estate (as P.03.09.01.01 already has them inherited); (3) what is mapped per project is its FDE (the project assignment record), and that is what Leo asks or is told about a project's people.

Leo asks the right questions: a registration question offers only options the records hold, and what the records settle is not asked

Ruled 2026-11-20 by ninad -- in force

LEO'S QUESTIONS OFFER WHAT THE RECORDS HOLD AND SKIP WHAT THEY SETTLE

THE QUESTION AND HIS WORDS: Leo, through the FDE backdoor, asked "Which dominion should MBank belong to?" offering "A. Sandbox -- recommended, matching your original request" or another dominion; Orion told Ninad that Sandbox is a project (where prospects' proofs of concept run), not a dominion, that the only dominion on record is Eden, and that option A would be refused at registration. Ninad: "this deserves a change in the current blueprint service, so that Leo ends up asking the right questions".

READ AS (Orion's reading, not the administrator's words): (1) the Blueprint service is changed so Leo's registration questions offer only options the records hold -- the live dominions, the existing clients and their accounts -- never an option he composes from the conversation; (2) a question the records already settle is not asked: one live dominion is the dominion, and the administrator is the one root's record names (blueprint #1089); (3) for MBank this means its dominion is Eden, the only live dominion, while its proof of concept runs in Sandbox.

A client reads PROSPECT until a contract of its is signed; Leo registers a new prospect under it, never as a client

Ruled 2026-11-20 by ninad -- in force

A CLIENT IS A PROSPECT UNTIL ITS CONTRACT IS SIGNED

THE QUESTION AND HIS WORDS: Leo, through the FDE backdoor, asked which client should hold the MBank prospect's account (a new client record, Chowpatty, or Denken Labs), after Ninad had said "MBank is not a client yet, its a new prospect". Orion explained that registration always files a project under a client and that the client records cannot say 'prospect, not yet a client', and offered: register under a new client 'MBank' for now; "Fix Blueprint first -- a client record can be marked 'prospect' until the contract is signed; Leo registers MBank as a prospect client once that exists"; or hold it under Denken Labs. Ninad chose "Fix Blueprint first".

READ AS (Orion's reading, not the administrator's words): (1) a client record reads PROSPECT until a service contract of one of its projects is signed, and CLIENT from that signature -- DERIVED from the signature exactly as a project's stage is (blueprint #868, shared.project_stage), never a status stored on the client; (2) Leo offers and registers a new organisation as a new prospect, and says so; Chowpatty, Denken Labs and any other existing record are offered with what they read; (3) MBank is then registered under a new client record named MBank, which reads prospect.

Leo asks the client only what is the client's business; the estate's own details he learns from existing implementations and records himself

Ruled 2026-11-20 by ninad -- in force

LEO ASKS THE CLIENT ONLY WHAT IS THE CLIENT'S BUSINESS, AND LEARNS THE REST FROM EXISTING IMPLEMENTATIONS

THE CONVERSATION AND HIS WORDS: through the FDE backdoor Leo asked Ninad, acting as MBank's FDE, which client MBank belongs to, who its administrator is, its account name, which dominion -- offering Sandbox, which is a project -- and the client again. Ninad ruled on two of them (#1089 the administrator, #1090 the options, #1091 a client reads prospect) and then gave this feedback: "note that Leo is asking too many questions that are not client's business to answer.. he will need to be further trained to "learn" from existing implementations, and update internal records that dont concern the client.. this was my feedback from todays conversation.. record the feedback".

READ AS (Orion's reading, not the administrator's words): (1) Leo asks the client (or the FDE speaking for them) only what is the client's business to answer -- their goals, users, processes, constraints, choices about their own organisation; (2) what is the estate's business -- where a project is registered, its dominion, its administrator, its machinery, the internal records it needs -- Leo learns from how existing implementations were done (Chowpatty, Carolverse's own records) and records himself, without asking the client; (3) this is training Leo needs beyond the MBank MVP: his intake is to be taught which questions are the client's and which are the estate's, and to fill the estate's from the records.

Where Leo's registration takes who is asking from

Ruled 2026-11-21 by ninad -- in force

Asked how Leo's registration of MBank should know who is asking (it had asked the model and was refused), Ninad: "the mBank registration should pick the identity from the operators login" -- and: "Leo's backdoor tool picks it from operator sign-in; Leo's frontdoor tool (agent chat) picks it from user sign-in". READ AS: a Leo tool never takes who is asking from the model's arguments; on the FDE backdoor it is the operator the sign-in proves, in his chat the user the sign-in proves.

Leo turns the requirements given into a small first result and asks if the direction is right, instead of opening with questions

Ruled 2026-11-21 by ninad -- in force

Asked, as MBank's FDE, Leo's second intake question (what MBank's proof should demonstrate, options A/B/C), Ninad: "what you have drafted seems ok, but Leo is asking too many questions, the requirements are already give, his job is to convert the requirements into actionable roadmap and pursuits, which can be implemented by the FDE to create results to the customer quickly.. so that the customer or FDE can tell what next.. too many questions in the beginning will put off the clients.. instead, show something small and ask if the direction is correct.." READ AS: Leo starts from the requirements already given (on the backdoor, the FDE's; in his chat, the client's) and turns them into an actionable roadmap and pursuits the FDE can build quickly; he shows a small first result and asks whether the direction is right, instead of opening with a round of intake questions; a question is asked only where the requirements given leave the next step unknowable.

A project keeps its own hierarchy, and Leo is its interim CEO until it chooses its own

Ruled 2026-11-21 by ninad -- in force

Asked whom MBank's new sales agent reports to (Athena refused a person as its manager), Ninad: "projects follow their own hierarchy, and are not part of carolverse hierarchy.. this would mean that the client / user / FDE sits on top, reporting to him is the CEO of the project, and reporting to the CEO is the team.. As the team expands, the org gets adjusted by leo, via the reusable service that builds and manages org structures". Then, offered Clara as stand-in CEO: "I have a better idea, how about Leo becomming the interim CEO until the project decides to have one of their own? Leo already knows a lot about the project, its plas, its requirements, is already doing the planning for the project, he seems to be the right fit, as an iterim.. project can decide to keep him as long as they want". Told that Leo would then buy Carolverse's services from himself, and offered that spending stays approved by the client or FDE above him: "agreed, also Leo should let client know that he wont feel bad if client chooses to have their own dedicated CEO.. but whatever work he does for client will have to be captured corrently and charged to the client.." READ AS: a project's org is its own, never Carolverse's: its client, user or FDE (a person) on top, its CEO (an agent) under them, its team under the CEO; until the project chooses a CEO of its own, Leo is its interim CEO and its agents report to him, for as long as the project wants; Leo tells the client plainly that choosing a dedicated CEO is welcome; a project's spending is approved by its client or FDE, never by Leo alone; every piece of work Leo does as its CEO is recorded and charged to the project; Leo adjusts the org as the team grows through the reusable org-structure service.

A project's services and tracks carry prescribed model assignments, and Carolverse learns from them only on an estate operator's approval

Ruled 2026-11-21 by ninad -- in force

Told that model assignment is governed per Carolverse service track (P.03.03.01.73) while Athena's onboarding lets the requester pick an agent's model, Ninad: "assignment of a model is a service in itself, and it should be governed by "model policies"" -- and: "same architecture should be used for projects.. if there is a sales agent, there should be a sales service, and a track created within it.. the service that builds services and tracks for projects should also provide prespcriptive model assignments for the projects based on the various relevent policies and runings in carolverse, which can be improvised for each project seperately.. starting point can be carolverse model assignment principles.. Carolverse should be able to learn from projects experience and should be able to edit their canonical policies for model assignments which can be then used for other projects, only upon approvals from estate operators.." Asked to confirm that a project's own change touches only that project and the canonical policy changes only with an estate operator's approval: "agreed". READ AS: a project's agents work in the project's own services and tracks, made by the service that builds them; each track carries a prescribed model assignment, worked out from Carolverse's model policies and rulings (P.03.03.01.73 first) and adjustable for that project alone; an agent runs on its track's assignment, never a model its requester picks; what projects learn is proposed as a change to Carolverse's canonical model policy, which changes only on an estate operator's approval and then serves every project.

A project's default service is its governance service to begin with

Ruled 2026-11-21 by ninad -- in force

While MBank's own services and tracks were being worked out (blueprint #1104, #1105) -- MBank holding only the 'default service' Leo's registration makes for every project -- Ninad: "the default service can be the governance service to begin with.." READ AS: the default service every project is given at registration is that project's governance service to begin with, owned by its CEO (Leo while he is interim CEO, #1104); the project's other services -- Sales for MBank's sales agent first -- are made beside it as its team needs them, each with its tracks and their prescribed model assignments (#1105).

A project's track runs on the model of the Carolverse track that does the same work

Ruled 2026-11-21 by ninad -- in force

Asked which model MBank's new voice track should run on (a track runs on a model only once its own work is proven there, and its work is the agent's, not yet created), Ninad: "voice is a new service for carolverse or an existing service? if its an existing service, then use the same model, if its a new service then we need to decide here what should be model for that track" -- then: "the real question is - what model does the default governance track should use, and the answer is again the same - the same model that carolverse governance track uses". READ AS: a project's track that does what an existing Carolverse track does runs on that Carolverse track's model -- the project's default (governance) service's track on Carolverse governance's, a voice sales track on the models of Carolverse's live web voice (Galadriel's Agent Chat and speech); only a track with no Carolverse counterpart has its model decided for it, under the model policy (P.03.03.01.73). It is the starting point #1105 names; a project may later change its own.

Work Leo files at a client's, user's or FDE's request is theirs, and needs no owning agent's approval

Ruled 2026-11-22 by ninad -- in force

Ninad, 2026-11-22 (CLI-327), was asked to approve, in Athena's place, (MBank Sales Agent's build), which waited on her as owner of the paying service because its origin read 'agent-initiated, first asker Leo'. HIS WORDS: "When Leo files a pursuit, the pursuit is owned by an agent, and assigned to FDE.. FDE executes the pursuits via either albus bypass or orion bypass.. approval of agents come into picture when agents file initiatives that belong to their service.. A pursuit can span multiple services, but a pursuit filed by Leo is a pursuit filed on behest of the client / user / FDE, which is not an agent initiated work.. Leo is just doing the filing, but the work is initiated by the user / FDE.. This is similar to when an employee asks an agent to file an initiative via internal agent chat, only difference being that this work was filed via external agent chat (Leo's external agent chat). So It should not need owning agent's approval, if I am not wrong.. what do you think?" Orion answered that the law agrees and that the gap is Athena's build filing naming Leo as first asker, dropping the FDE who asked through Leo's door; proposed: this ruling, a cure so such a filing carries the request's real first asker, and a one-time correction of 7687's origin. HIS ANSWER: "sounds good, go ahead". READ AS (Orion's reading, not his words): work Leo files at the request of a client, a user or a project's FDE -- including what other agents file onward from such a request, such as Athena's build of an agent Leo asked for -- is user-initiated: its first asker is that client, user or FDE and Leo (and any agent after him) is a relay, so it never waits for the paying service owner's approval; owner approval stays for work an agent raises of its own accord.

A prospect is subscribed to its services when it is onboarded, sorted into mandatory, recommended and proposed, and Leo subscribes and unsubscribes them at the client's request

Ruled 2026-11-22 by ninad -- in force

Ninad, 2026-11-22 (CLI-323), asked about (each service a project is subscribed to gives Leo its questions for the client), whether it belongs after the contract is signed since neither prospect (Chowpatty, MBank) is subscribed to any service. HIS WORDS: "subscribing to services should happen as the first step when the prospect is created.. the first default service is the org and govewrnance (I think) It should also use the requirements to subscribe the platform to the applicable services, categorised into 3 categories (mandatory - like governance, backups etc), recommended (like voice calling for mBank and chowpatty) and proposed (asked by client but do not exist in carolverse at all) the proposed services should be built for the prospect, while the mandatory and recommended services should be subscribed by the prospect.. this should happen as part of onboarding the prospect in the right sequence.. subscribing and unsubscribing of services based on client request should also be an intrinsic responsibility of Leo". READ AS (Orion's reading, not his words): (1) subscribing to services is part of onboarding a PROSPECT, not something that waits for a signed contract; its first step, when the prospect is created, subscribes it to the default service -- org and governance, which he named tentatively ('I think'); (2) from the prospect's requirements its services are sorted into three categories: MANDATORY (every project takes them, such as governance and backups), RECOMMENDED (an existing Carolverse service the requirements call for, such as voice calling for MBank and Chowpatty) and PROPOSED (asked for by the client but not existing in Carolverse at all); (3) the prospect subscribes to the mandatory and recommended services; a proposed service is BUILT for the prospect; (4) all of this happens in onboarding in the right sequence; (5) subscribing and unsubscribing at the client's request is Leo's own responsibility. therefore stays in the MVP of CAROL-PUR-0017. Not settled by these words: which services are mandatory beyond governance and backups, and how cost-center #442 (no service enabled for a project before its own bank account is active, or the administrator's waiver) applies to a prospect.

Which services a new customer takes, mandatory or recommended, is one re-usable mapping per kind of customer -- self-serve, prospect, project -- owned by Leo and reviewed by him about monthly from client responses

Ruled 2026-11-22 by ninad -- in force

Ninad, 2026-11-22 (CLI-323). THE QUESTION: Asked (CLI-323, after blueprint #1130) which services every prospect should get automatically -- governance, backups, monitoring, audit and security was offered, or governance and backups only. HIS WORDS: "There might be I think many more services that should be mandatory for the self service clients, prospects and projects.. create a re-usable list like how you have built for categorizing services into grades.. so that Leo can use it handy when subscribing new customers to services, and this list should be customized to self service, prospects and projects. Leo should own this mapping and should update it based on the client responses he receives on a periodic basis (not too frequently, maybe once a month or so)". READ AS (Orion's reading, not his words): (1) which category each Carolverse service falls in for a new customer -- mandatory, recommended (blueprint #1130) -- is ONE re-usable record, kept like the Service Track Handbook's reuse grades, not a list typed into Leo's prompt or code; (2) it is kept separately for each kind of customer: self-serve client, prospect and project; (3) Leo owns it and uses it when he subscribes a new customer; (4) Leo reviews and updates it from the responses clients give him, periodically -- about once a month, not more often. The list of mandatory services is therefore Leo's to set in that record, not fixed in this ruling.

๐Ÿ“–Build Initiatives

Escalation ladder direction, troubleshooting retries and their budget

Ruled 2026-09-14 by ninad -- in force

The escalation ladder is ONE WAY: planner simplified climbs to the Albus bypass, which climbs to the Orion bypass, and a family never descends. During troubleshooting, however, multiple attempts of the same initiative may be made at the SAME level, so a broken pipeline can be fixed while the initiative is still executing on the rung being repaired. Each retry is a new recorded attempt and the failed ones are preserved. The hold can never move a family down the ladder and grants no exemption from the switch, concurrency, blocked-work, dispatch or budget brakes. Troubleshooting is billed to its own dedicated service track (initiatives-troubleshooting) with the same daily budget as the planner track, so pipeline repair never consumes the budget meant for delivering initiatives.

Scope of the troubleshooting track: planner only

Ruled 2026-09-14 by ninad -- in force

The dedicated troubleshooting track carries PLANNER troubleshooting only. Albus bypass troubleshooting stays on the Albus Bypass track, and ladder rescues stay on Planner - Escalated, both unchanged. Ninad's reason, 2026-09-14 (CLI-415): the Albus lane is working well and its troubleshooting is not expected to consume significant budget, so a dedicated track for it is not worth creating. This is a deliberate scope boundary, not an oversight โ€” the operator proposed widening the track to cover repair on any rung and Ninad declined. Revisit only if Albus-lane repair starts displacing his delivery work, which would show as his own track exhausting on repair rather than on initiatives.

Planner delivery proof runs on a genuinely ready candidate

Ruled 2026-09-14 by ninad -- in force

A planner delivery proof runs on a GENUINELY READY candidate, chosen before any paid admission, and readiness is judged by what actually blocks a design rather than by how promising the title reads. Ninad ruled 2026-09-14 (CLI-415), after consumed three attempts without ever producing a design: pick a genuinely ready candidate and prove delivery on that. The two or three genuine planner deliveries and the escalation-ladder proof remain owed before autonomous execution is enabled.

A close whose Palantir story cannot be written goes through; the story is owed

Ruled 2026-09-16 by ninad -- in force

When a close's story cannot be written, the close is not held: the initiative is marked story-owed with the reason, and the story is written automatically from its own record once narration can run.

The doer speaks for itself on Palantir

Ruled 2026-09-16 by ninad -- in force

When one agent's work is carried out by another agent's droid, the agent whose droid did the work narrates it in its own voice; the requester no longer narrates work it did not do.

An arc is owned by the service owner whose service its initiatives sit in

Ruled 2026-09-23 by ninad -- in force

EVERY ARC HAS AN OWNER, AND IT IS THE SERVICE OWNER, NOT THE OPERATOR. An arc's owner is the owner of the service its initiatives sit in. Orion owns an arc ONLY where its members are genuinely spread across services, with no single service's duty of care covering the subject.

WHY IT NEEDS SAYING. Orion files most arcs from the operator lane and is usually the one executing them, so naming himself owner reads as accurate. Doing the work is not owning the line of work: an arc owned by the operator is an arc whose progress nobody in the estate answers for, and it hides the service that should be carrying it. This is the placement rule the estate already applies to initiatives -- work sits with the service whose DUTY OF CARE covers its subject, never with whoever owns a gate or a mechanism it passes through -- applied one level up.

THE TEST is where the member initiatives actually sit, not who is executing them and not who filed them. Cross-service ownership by Orion is the exception and is established by looking at the members, never assumed because the arc is large or because the operator started it.

A pursuit is a prioritised outcome reached through initiatives from any arcs, in sequence

Ruled 2026-09-24 by ninad -- superseded 2026-09-24

A PURSUIT is a prioritised action taken to reach one outcome: its name states the action, its success criteria say when the outcome is reached, and it lists the initiatives it needs in the order they must happen. Those initiatives may come from one or more arcs -- an arc groups one service's initiatives under the agent who owns that service, while a pursuit is outcome-oriented and crosses arcs, especially where work in one arc depends on work in another. A pursuit is a registry record with one owning agent and a priority rank among all live pursuits (1 first); its progress and the step it waits on are derived live from the initiatives and never stored, and it is achieved only when every step is closed and every criterion is met with proof. Ninad first called it a THREAD; shown that the word already means the relay thread and a conversation, he chose Pursuit, and added the owner and the rank to the three fields he named. Ninad, CLI-544: 'threads are prioritised actions (like Sahil's onboarding) that would involve executing 1 or more initiatives from 1 or more arcs in a certain sequence to achieve an outcome.. they are outcome oriented and may span across arcs'.

Ended because: Ninad added a human operator owner beside the owning agent (CLI-545)

Effort is the estimated sum of tokens to complete a unit of work

Ruled 2026-09-24 by ninad -- superseded 2026-11-09

EFFORT is the estimated sum of tokens needed to complete a unit of work: an activity, a task, an action, an attempt, an initiative, an arc or a pursuit. It is an ESTIMATE, never the measured tokens or the cost. An ATTEMPT's effort is the estimate for that one attempt. An INITIATIVE's effort is the total over every attempt of the family along the escalation ladder, derived by adding its attempts' efforts and never stored, so the two are always kept apart. Each attempt is sized by ELROND from its own record, never typed in by an operator. A pursuit's completion is read FROM EFFORT: the initiative effort of its closed steps divided by the initiative effort of all its steps; while any step is unsized the figure is unmeasured and names that step. Ninad, CLI-545: 'effort --> estimated sum of tokens to complete a certain activity, task, action, initiative, arc, pursuit etc' and 'there should be a clear distinction of efforts for distinct initiative attempt vs effort for an initiative in total (including all attempts via escalation ladder)'.

Ended because: CLI-079: a pursuit counted closed steps only and read 0% with built, live work; Ninad ruled that work done counts, as it does for arcs (#205, #302)

A pursuit is a prioritised outcome with an owning agent and a human operator owner

Ruled 2026-09-24 by ninad -- in force

A PURSUIT is a prioritised action taken to reach one outcome: its name states the action, its success criteria say when the outcome is reached, and it lists the initiatives it needs in the order they must happen, drawn from one or more arcs. It is a registry record with ONE owning agent, a HUMAN OPERATOR OWNER beside that agent, and a priority rank among all live pursuits (1 first). The operator owner is a person on the root-owned list of operators and is checked there whenever it is set; the agent stays accountable, the operator owner is the human who answers for the outcome. Progress, the step it waits on and its completion by effort are derived live from the initiatives and never stored, and it is achieved only when every step is closed and every criterion is met with proof. Ninad, CLI-545: 'pipeline app should also have a tab for pursuits, along with arcs, displaying the latest pursuits and their status and human operator owner' -- offered replacing the agent owner or adding beside it, he chose beside.

Every pursuit documents its MVP

Ruled 2026-09-25 by ninad -- in force

Every pursuit documents its MVP. A step that is absolutely necessary for the pursuit's outcome to be reached in the right manner is MVP; everything that can happen after it is post-MVP. For an operator's onboarding: anything absolutely necessary for the operator to connect to the machine in the right manner is MVP, and everything that can happen after they are onboarded is post-MVP. The MVP is stated in words on the pursuit and each step's stage is recorded; neither is inferred from order or status.

Redirect and blocked are different lanes: blocked cannot execute, redirect is the pipeline's choice to send the work to Orion over a policy or ruling

Ruled 2026-09-25 by ninad -- in force

REDIRECT AND BLOCKED ARE DIFFERENT LANES. Ninad's words (2026-09-25, CLI-552): 'the redirect lane is different from the blocked lane.. blocked mean that the pipeline is simply not able to execute the initiative autonomously while redirecting is a execution choice that the autonomous pipeline makes to redirect the initiative to orion mostly due to policy and ruling conflicts or absense of them'. BLOCKED: the autonomous pipeline could not execute the initiative (it tried and failed, crashed, or exhausted its attempts); it stays on the escalation ladder. REDIRECTED: the autonomous pipeline CHOSE, during execution, to send the initiative to Orion - chiefly because the work conflicts with a policy or ruling, or no policy or ruling covers it. A redirect is therefore not a failure: it consumes no attempt, counts toward no blocked cap and stops no lane. This extends the redirect doctrine of 2026-07-14 (cookbook 535: redirected only for policy matters) from filing to execution; redirected stays DEAD to everyone but the operator or a verdict (cookbook 717).

Written Palantir stories for Orion-bypass work are paused

Ruled 2026-09-25 by ninad -- in force

Written Palantir stories for work closed on the ORION BYPASS lane are PAUSED by the administrator's decision of 2026-09-25 (CLI-555), because that lane's executions are human-in-the-loop and the administrator is present for the work. While the pause stands: no Orion-lane close is held, blocked or tagged for a missing story; stories already owed for Orion-lane work stay owed and are never dropped or back-filled as told; the model-free run narration continues on every lane; Albus-bypass and planner work keep their close stories and their failure stories. Resuming is the administrator's decision alone. WHAT LOOKS LIKE A FIX AND IS NOT: dropping the owed Orion-lane stories; pausing narration too; treating a storyless Orion-lane close as a missing record.

Inspector accepts the work Inspector owns

Ruled 2026-09-25 by ninad -- in force

Work Inspector owns is accepted by Inspector himself, through an acceptance droid of his own and a UAT-verdict grant on his conscious-action whitelist scoped to initiatives he owns. The acceptor is still derived from the initiative's OWNER (cookbook 1729); this gives the owner the droid and the grant that derivation needs. An acceptance ask routed to an agent that holds no UAT-verdict grant for that initiative is the defect this ruling closes. WHAT LOOKS LIKE A FIX AND IS NOT: handing Inspector's work to Elrond or the operator instead; widening Inspector's grant beyond work he owns.

An operator's own roadmap-refused filing still escalates to Orion

Ruled 2026-09-25 by ninad -- in force

A filing refused by the roadmap gate escalates to Orion's inbox whoever filed it, the operator's own filings included, even though the operator saw the refusal at the moment of filing. Every such escalation is closed with its outcome like any other (refiled with a roadmap item, or declined with the reason). WHAT LOOKS LIKE A FIX AND IS NOT: exempting the operator's own refusals from the escalation to cut inbox noise.

The planner lane stays abandoned until it is fixed

Ruled 2026-09-25 by ninad -- in force

The planner lane stays ABANDONED until it is repaired: no initiative is run on it, no planner run is made to prove a criterion, and it spends no tokens. It is neither retired nor resumed. Work whose remaining proof needs a planner run waits parked until the planner is fixed. WHAT LOOKS LIKE A FIX AND IS NOT: a one-off planner run to close such work; re-scoping its criteria to avoid the planner; retiring it because the lane is idle.

Elrond's activities are one track of his own, Build Governance

Ruled 2026-09-25 by ninad -- in force

Every task Elrond owns folds into ONE track in the Initiatives service, initiatives-build-governance (Build Governance), owned by Elrond and named for his function - intake and filing, review gates and verdicts, acceptance, effort sizing and the pipeline's watch and measurement droids - so the activity still decides the track (cookbook 1035). His droids that had declared Merlin's blanket planner task declare a task of his own on it. His own two small lines fold in too and retire in place: UAT acceptance (initiatives-uat) and effort sizing (initiatives-effort-estimation). What a standing ruling places elsewhere STAYS there: Palantir stories and run narration on the audit line (audit #17), and his chat, mind wakes, wellbeing checks, status reports, App Steward and Unmeasured Call Watch on the estate-wide tasks every agent shares; his Initiative Creator's filing work for others keeps billing the asker (1156). Which droids bill the track is the architecture's answer, not a ruling (design 510): a task folds into exactly one track and inherits that track's owner, and a droid bills the task it declares - so another agent's droid serving one of Elrond's tasks bills his line. Cap by rule 1730. The track was registered on its slot before any task moved onto it (P.01.03.07.21), by Ninad's choice over moving first.

A change to shared estate code that carries out a live ruling is approved by that ruling

Ruled 2026-09-25 by ninad -- in force

When Orion changes shared estate code to carry out a ruling or policy that is live, that ruling is the administrator's approval of the change, and Orion does not ask again for each change. A change that no live ruling calls for still needs his approval. The approval widens no door: the laptop's own permission check, the role doors, the root and core rungs and every gate still apply, and a change the permission check refuses is installed through a sanctioned door or handed to the administrator as a graded step, never worked around.

Every initiative in an arc sits in a phase of it, and every arc has a phase

Ruled 2026-09-25 by ninad -- in force

Phase is mandatory. An initiative is a member of an arc only THROUGH one of that arc's phases: membership is derived from phase tags alone, so an arc member without a phase cannot exist. The arc's own tag on an initiative that carries no phase of it makes it no member, and it is reported by name as UNPLACED rather than counted. Every arc is created together with its first phase, and an arc's last phase cannot be deleted while the arc stands. The members that had no phase when this was ruled were placed in one (phases were created for the arcs that had none) by ADDING the phase tag: no tag an initiative already carries is rewritten or removed, which is the part of cookbook 1790's never-re-tag clause that stands; adding a phase to closed or ended work was the administrator's explicit choice ('place all, then enforce'). An initiative still belongs to at most one phase of any one arc.

The operator may take an item Albus has failed on when an arc the administrator asked to complete needs it

Ruled 2026-09-26 by ninad -- in force

Rule 1813 lets the operator take onto the Orion bypass an initiative Albus's bypass has attempted and FAILED when a PURSUIT needs it. His failed attempts stay on the record as his. Asked: "The rules let me take a failed Albus item onto my lane only when a pursuit needs it, not an arc. May I take items Albus has failed on when an arc you've asked me to complete needs them?" Answer: "Yes, arcs too".

A pursuit holds every initiative its success requires, from any arc or from none

Ruled 2026-09-27 by ninad -- in force

A pursuit's sequence holds EVERY initiative its outcome and success criteria require -- from any arc, from several, or from none -- and nothing required is left outside it because it belongs to another arc or to no arc at all. A blocker found while a pursuit is being worked is added to that pursuit (with its stage, MVP or post-MVP) as soon as it is found to be required, never parked in handover prose as 'outside the pursuit'. Ninad, CLI-569 (2026-09-27), told that the operator-owned-code write blocker sat outside Albus's pursuit CAROL-PUR-0002 because an earlier instruction had limited that pursuit to one arc's items: "pursuit can have initiatives across arcs, and it should have all initiatives that are required for the pursuit to be implemented successfully". WHAT STANDS BESIDE IT: initiatives #124 (a pursuit is a prioritised outcome with an owning agent and a human operator owner, drawn from one or more arcs) and #127 (every pursuit documents its MVP and each step's stage). WHAT IT RETIRES: the CLI-566 instruction to take only CAROL-ARC-0021's items into CAROL-PUR-0002 (an instruction recorded in a handover, never a ruling). WHAT LOOKS LIKE A FIX AND IS NOT: filing a second pursuit for the blocker, leaving a required item out because it is owned by another service's arc, re-tagging it into the pursuit's arc, or listing work that is merely related rather than required (a pursuit's steps are what its criteria need, not everything nearby).

A pursuit's MVP holds only what is mandatory for its outcome, judged part by part: a non-essential part of an MVP initiative is split into a post-MVP follow-on

Ruled 2026-09-27 by ninad -- in force

A pursuit's MVP holds only the work that is mandatory for its outcome to be reached in the right manner -- for Sahil's onboarding, what must hold for him to start his work with proper security protocols followed. That is judged PART BY PART, not initiative by initiative: where an MVP initiative also carries a part that is not mandatory -- a proof that moved work still performs well, a display, a defence in depth, a nicer refusal message -- that part is split into a post-MVP follow-on initiative on the same pursuit, and the MVP initiative closes on what is mandatory. Criteria are amended or struck with a recorded decision naming this ruling and the follow-on -- never deleted -- and a test that proves the moved part moves with it. Ninad, CLI-568 (2026-09-27), on learning that 's last open check (the moved jobs' first scheduled runs) has nothing to do with Sahil's security: "why are they part of mvp in Sahil's onboarding? mvp should focus only on mandatory initiatives required to get sahil onboarded with proper security protocols followed so that he can start his work", then "yes" to splitting 3790 and 5177 that way and reviewing 5182, 5157 and 5388 the same way. WHAT STANDS BESIDE IT: initiatives #127 (every pursuit documents its MVP; each step is mvp or post_mvp) -- this ruling applies #127 inside an initiative. WHAT LOOKS LIKE A FIX AND IS NOT: dropping the non-essential part instead of moving it (it is still owed, only not first); keeping a whole initiative in the MVP because most of it is mandatory; staging a part post-MVP because it is slow or hard rather than because the outcome does not need it.

An initiative shared between pursuits is marked as overlapping in both, with the dependencies that must complete before it can be implemented; arcs and pursuits are built through one skill that carries this

Ruled 2026-09-27 by ninad -- superseded 2026-11-09

Whenever a pursuit is created or its sequence changes, its initiatives are checked against EVERY other pursuit. An initiative that sits in two pursuits is marked as OVERLAPPING in the step note of BOTH pursuits, naming the other pursuit, and each note explains the dependencies that must be completed before that initiative can be implemented. A dependency that crosses pursuits without the same initiative sitting in both (one pursuit's step must wait on another pursuit's step) is marked the same way in both. Overlaps are found from the records each time, never kept as a separate list. Arcs and pursuits are built through ONE skill in Sage's library, build-arcs-and-pursuits (bound to Orion), which carries this check together with the rulings that already shape them. Ninad, CLI-569 (2026-09-27): "make sure that the initiatives in the albus bypass pursuit dont overlap with any other pursuit, if they do, then mark them as overlapping in both the pursuits and explain the dependencies required to be completed before they can be implemented. there should be a skill for building arcs and pursuits and this should be included in the skill". FIRST APPLICATION: CAROL-PUR-0002 against CAROL-PUR-0001 -- no shared initiative; one crossing dependency, (PUR-0002) waits on (PUR-0001), marked in both. WHAT STANDS BESIDE IT: initiatives #124, #127, #157, #163. WHAT LOOKS LIKE A FIX AND IS NOT: dropping an initiative from one pursuit so the overlap disappears (it is still required by both, #163); storing an overlap flag that can drift from the steps; marking only one of the two pursuits.

Ended because: Ninad's later ruling (CLI-094): an initiative cannot belong to two pursuits, so the same initiative is no longer marked overlapping in two; cross-pursuit dependency marking stands.

Sahil's onboarding MVP proceeds on his chip key; password plus phone code is optional for it

Ruled 2026-09-27 by ninad -- superseded 2026-09-27

SAHIL'S ONBOARDING MVP PROCEEDS ON HIS CHIP KEY; PASSWORD PLUS A PHONE CODE IS OPTIONAL FOR IT (Ninad, CLI-575, 2026-09-27). Asked whether the password-plus-Microsoft-Authenticator way in should hold up Sahil's onboarding, he first kept it as a must-have, and then, with that build not ready, ruled: "keep this task optional for the mvp, and proceed with the touch screen for now". So in CAROL-PUR-0001 is a POST-MVP step, and Sahil's MVP way in is his chip key with Touch ID, on root's own key list (security #138/#161). Security #168 is unchanged: once 5623 is built every operator login accepts a chip key OR password plus a current code, and a password alone is refused before and after. This decides only what the pursuit's MVP needs (initiatives #164, part by part). If Sahil's computer turns out to have no security chip, that comes back to the administrator before his key step is re-run. WHAT LOOKS LIKE A FIX AND IS NOT: allowing a password alone in the meantime; retiring; or reading its post-MVP stage as a withdrawal of #168.

Ended because: Ninad, CLI-578: 'bring this back to the mvp and implement it' -- asked what happens if Sahil's computer has no security chip, the only answer that needs no new hardware was the password-and-code route.

An arc's description is mandatory and one or two short sentences, refused by the registry otherwise

Ruled 2026-09-27 by ninad -- in force

AN ARC'S DESCRIPTION IS MANDATORY AND ONE OR TWO SHORT SENTENCES, AND THE REGISTRY ITSELF REFUSES ANY OTHER (Ninad, 2026-09-27, CLI-576). His words: 'the pipeline app gives a very elongated description of the arcs in the arcs tab, this description should be not more than one or 2 short sentences.. please enforce this rule on the database level, by limiting the length of the description and making the description field mandatory', and 'update the skill for filing arcs so that it is able to accomodate this constraint'. The limit is 200 CHARACTERS: the arcs already written as one or two short sentences measured 77 to 195. A missing, empty (blank after trimming) or longer description is refused at the registry's write seam, on insert and on update, by whoever writes -- the arcs door names the refusal before writing. Detail that does not fit lives where it already has a home: the arc's phases, its members' decisions, its designs and the Rulings Repo. A shortened description's earlier text stays in the registry's change history (rule 1504), never deleted. ORION'S READING, recorded as his: the rule is about ARCS; phase descriptions are not limited by it. What looks like a fix and is not: raising the limit so a long description fits, dropping the trigger to land a row, or moving the long text into another arc field.

Sahil's onboarding MVP includes the password-plus-code way in again

Ruled 2026-09-27 by ninad -- in force

-- every operator's second way in, their password plus a current Microsoft Authenticator code (security #168) -- is an MVP step of Sahil's onboarding again, and is built before his onboarding runs, so that a computer without a security chip does not stop him. His chip key stays the first way in; security #168 is unchanged (a password alone stays refused).

An initiative off the roadmap joins no arc or pursuit and runs on no lane, the Orion bypass included

Ruled 2026-09-27 by ninad -- in force

Elrond's roadmap gate refuses to let an initiative that is not related to the roadmap become part of an arc or a pursuit, and refuses its execution on every lane, the Orion bypass included. Related to the roadmap means the initiative, or the family it is an attempt of, names at least one live, approved roadmap item: a strategic ask or a service task. How the roadmap and the tasks are kept in sync, and how duplicate entries are avoided, is left to Orion to decide and record.

An arc's completion is read by effort spent, from the Pipeline app

Ruled 2026-09-28 by ninad -- in force

AN ARC'S PERCENT COMPLETE REFLECTS THE ESTIMATED EFFORT SPENT, AND IT IS READ FROM THE PIPELINE APP (Ninad, CLI-604, 2026-09-28). Shown an arc at 0% by count of closed initiatives while three of its four initiatives had work built, installed and parked, Ninad said: 'The % complete should reflect the estimated efforts spent, and it should be picked from the pipeline app.' Asked whether work that is built but not yet closed counts as effort spent, or only closed work as a pursuit counts today, he agreed with the recommendation: EFFORT SPENT COUNTS, closed or not. So an arc's and a phase's completion is the estimated effort already spent on its members over the estimated effort of all of them; it is derived, never stored; it reads UNMEASURED while any member is unsized, never a partial sum and never nought; and a session reporting an arc reads the figure from the Pipeline app rather than working it out. How a pursuit's completion is read is not changed by this ruling.

A filing the route table cannot place lands on Albus's bypass lane by default

Ruled 2026-09-30 by ninad -- in force

Asked whether anything will still be diverted to the planner lane under the current ruling, and told that a filing the seam cannot place lands as sent - where a caller naming no mode gets the creator's default mode, planner - Ninad ruled: 'fix the gap, make albus bypass default'. A new initiative filing whose source Elrond's filing seam cannot derive takes the route table's DEFAULT route, Albus's bypass lane in the Albus shape, instead of landing as sent. The default is declared in the route table itself (skill route-initiative-filing, key default_route), never in code. Not defaulted: an attempt or follow-on carrying parent_nnn keeps its family's lane, and a source a door declared explicitly keeps landing as sent. Restoring any planner route remains a ruling of its own.

Albus escalates work he cannot finish.

Ruled 2026-09-30 by ninad -- in force

"Albus should escalate if he is not able to complete the work assigned to him allocated to him." Before starting an initiative Albus judges whether it can be completed if it cannot, he does not start it and escalates it to Orion by the ruled route (an Agentbook message to Orion). The work keeps its place on his lane: cookbook 1769 stands (a spent budget is not a broken lane, and work is never re-routed for money alone).

Albus prioritises every initiative on his lane

Ruled 2026-09-30 by ninad -- in force

"Not all initiatives are of equal importance, albus should be able to prioritize the initiatives." Albus ranks every initiative waiting on his lane -- backlog and urgent work as well as wishes -- by importance in his own judgement, with a reason for each, and his lane picks in that order. This extends the wish ranking (Rulings Repo #25) to all of his work.

Albus's lane stops only at its tenth blocked initiative

Ruled 2026-09-30 by ninad -- in force

"Increase the escalation count to 10, so that the lane does not stop at each escalation." The blocked limit of Albus's bypass lane is 10: his lane keeps picking work until ten of its initiatives are blocked. This amends cookbook 1588 / (a limit of 1 per lane) for Albus's lane only; the planner lane keeps its limit and the Orion bypass still has none.

Albus keeps his queue relevant, and pushes back or rejects

Ruled 2026-09-30 by ninad -- superseded 2026-11-07

"Continuously assess if the initiatives are relevant.. and push back, reject if necessary." Albus keeps re-checking, and always before a paid attempt, that each initiative waiting on his lane is still relevant: its premise is still true, what it names still exists, and it is still wanted. When one is not, he pushes it back to whoever first asked, saying what is missing, or rejects it with his reason recorded where the asker reads it. A rejection is a recorded judgement, never plumbing (cookbook 713 stands).

Ended because: Only an initiative's filer discards it (CLI-058)

Orion's CLI-626 proposal for Albus's lane is adopted

Ruled 2026-09-30 by ninad -- in force

"Incorporate your suggestions." Adopted from Orion's proposal in CLI-626: (1) Albus's queue is cleaned before any paid attempt -- work whose problem is already fixed, or that names nothing that exists, is closed first; (2) when Albus's decide step finds his lane cannot do the work (a door it lacks, or something the filing names that does not exist), the work goes straight to the Orion bypass on that first attempt, the way a law block does (#139), and is never retried on his lane; (3) Albus's lane is given a door for the administrative registrations its work needs; (4) Orion takes the initiatives Orion first asked for off Albus's lane, one at a time, as cookbook 1775 already allows.

Effort estimates are made for all initiatives, so arcs and pursuits read accurate completion by effort

Ruled 2026-09-30 by ninad -- in force

Effort estimates are made for ALL initiatives, not only the steps of live pursuits and the members of live arcs, so that an arc's or a pursuit's completion by effort can be given accurately. Ninad, 2026-09-30 (CLI-625), on being shown the project onboarding arc reading 'unmeasured' in every phase and the effort pursuit (CAROL-PUR-0004) at 3 of 4 steps: 'complete it so that effort estimates are available for all initiatives, and reflected in arcs and pursuits, thereby it would be possible for you to provide accurate % completion for arcs and pursuits'. Elrond's Effort Estimator stays the one sizer (#123) and completion stays derived by the one composer (#205); this widens WHICH initiatives are sized, never how a figure is read. Filling an unsized initiative with a median or a guessed number so a percent appears, or reading an unmeasured figure as nought, each undo it.

When Albus's lane stops, Ninad and Sahil are told at once

Ruled 2026-09-30 by ninad -- in force

"if albus lane stops for any reason, I and Sahil get notified immediately". Whenever Albus's bypass lane stops picking work -- for any reason: its budget spent, its switch or breaker, its blocked limit, its sign-in or model refused, its runner not running -- the operators (today Ninad and Sahil) are notified at once, naming the reason and what would restart it; and again when it runs again.

A filing from an Orion session is raised by the operator only when the initiative belongs to an arc or a pursuit; otherwise it is raised by Orion

Ruled 2026-10-01 by ninad -- in force

"if an initiative belongs to an arc or pursuit, then it can be me, else it will be orion." For a filing that comes through the operator CLI (an Orion session on an operator's token), the party that raised it is the operator ONLY when the initiative is a member of an arc or a pursuit; an initiative in neither is raised by Orion. Amends the operator-CLI clause of cookbook 1792, under which every operator-CLI filing read as raised by the token's owner, so findings Orion filed on his own during work were credited to the operator.

Who raised an initiative is a mandatory attribute of every initiative

Ruled 2026-10-01 by ninad -- superseded 2026-10-01

"the raised by should be a mandatory attribute." Every initiative records who raised it; a filing whose raiser cannot be established is refused by name rather than recorded as unknown. Filings made before the origin record existed (2026-09-20) still read unknown and are never backfilled (cookbook 1671, 1792).

Ended because: Ninad, CLI-635, 2026-10-01: unknown is not an acceptable raiser; shown that 336 of 371 unknowns are derivable from records made at filing and the rest from their parent or the recorded requester, he ruled 'backfill unknown'

Who raised an initiative is mandatory; unknown is not a value, and pre-record filings are backfilled from what was recorded at filing

Ruled 2026-10-01 by ninad -- in force

"unknown cannot be a requestor, this should be a mandatory field"; "backfill unknown". Every initiative names who raised it, and unknown is not a value. A new filing whose raiser cannot be established is refused by name. Filings from before the origin record (2026-09-20) are BACKFILLED from the facts recorded at their filing, each with its provenance: the party recorded as asking on the filing-lane decision ('requester as sent'); a filing through the operator channel resolves by ruling #258 (the operator when the initiative is in an arc or a pursuit, else Orion); a follow-on filing inherits its parent's raiser; where none of those is recorded, the requester recorded at filing (Elrond's or Albus's own gate) is the first asker. Nothing is invented: a backfilled origin says it was backfilled, from which record, and when. This narrows cookbook 1792's never-backfilled clause and cookbook 1671 for the origin record: a value derived from a record made at the time is not a fabricated reading.

Elrond checks every pending initiative on every lane for relevance, once now over the whole backlog and then daily

Ruled 2026-10-01 by ninad -- superseded 2026-11-07

"agreed, let elrond do it now for all initiatives, and then continue to do it on a daily basis." Shown that nothing re-reads a pending initiative after filing except Albus's own lane (#229), Ninad ruled: Elrond applies #229's relevance test -- the premise is still true, what the initiative names still exists, and it is still wanted (its roadmap item live) -- to EVERY pending initiative (planned, dispatched, parked, redirected, executing, blocked) on every lane, the Orion lane included: one full pass over the backlog now, then every day. An initiative found no longer relevant is discarded through the status router with the reason recorded where its raiser reads it (a recorded judgement, never plumbing: cookbook 713 and 717 stand; parked stays the operator's to move, 712, so a parked one is reported, not discarded); the daily count of checked, kept and discarded shows on the board. The sweep runs as Elrond's own registered process under his login, funded from his Build Governance track (#151); if the pass needs more than the track holds, the administrator is told the figure before the track is raised ("let me know if that needs increasing budget for his track").

Ended because: Ninad ruled Elrond discards only duplicates; the rest is parked and told to its filer, or parked for a ruling (CLI-058)

Every arc and pursuit carries its creator operator, and who it is assigned to and by whom

Ruled 2026-10-02 by ninad -- in force

EVERY ARC AND PURSUIT CARRIES ITS CREATOR OPERATOR, AND WHO IT IS ASSIGNED TO AND BY WHOM: 'the arcs and pursuits should have the creator operator, assigned to and assigned by records'. READ AS (Orion's reading, not Ninad's words): each arc and each pursuit shows the operator who created it, read from the login the registry's change journal recorded on its creation and mapped to the person through root's authority record -- never typed and never stored a second time -- and, from the assignment record (agent-resources #254), every service operator it is assigned to and the estate operator who assigned each. An arc or pursuit created before the change journal began (2026-09-19), or created by a login that is no operator's, says so in words ('creator not recorded', 'created by an agent's login') and is never given a guessed creator.

A service operator's requests are operator requests, admitted during a pipeline pause

Ruled 2026-10-04 by ninad -- in force

A SERVICE OPERATOR'S REQUESTS ARE OPERATOR REQUESTS (Ninad, 2026-10-04, CLI-674). Asked: 'When the pipeline is paused, should Saee's filings count as an operator's filing (accepted) or as autonomous intake (refused)? The current rule only lets Orion through.' Ninad: 'treat Saee's requests as what they are - operator requests'. READ AS (Orion's reading, not Ninad's words): a request Saee makes through her own door -- a filing, a UAT verdict, a scope change inside her assignments, handed to Elrond's Service Operator Performer -- is an OPERATOR's request, not an agent's autonomous intake: the pipeline pause admits it as it admits the operator's own filings (cookbook 1225's operator clause), and every other rule of her rung still holds (only inside her live assignments, never the machinery, recorded under her own name).

A completed arc reopens when new work for its subject is found

Ruled 2026-10-04 by ninad -- in force

Asked (CLI-683, 2026-10-04): the arc record treats 'completed' as final -- no move out of it is ruled legal -- and the build-arcs-and-pursuits skill forbids a second arc for a family that already has one, so new work found for a completed arc's subject had nowhere to go. Options put: (1) completed arcs can reopen: when new work for an arc's subject turns up, the arc goes back to active and the work joins it as a new phase; (2) completed stays final and follow-on work starts a new arc; (3) leave the work outside any arc. Ninad's answer, in his words: '1. Completed arcs can reopen'. READ AS (Orion, marked apart from the ruling): the arc status vocabulary gains the move completed -> active; the arc's completed phases stay completed and the new work joins a NEW phase; nothing else about arcs changes. First application: CAROL-ARC-0031 (The Operator Instance Log) reopened for as its phase 6.

An arc's and a phase's effort and percent complete are accurate and current at every reading

Ruled 2026-10-04 by ninad -- in force

AN ARC'S AND A PHASE'S EFFORT AND PERCENT COMPLETE ARE ACCURATE AND CURRENT AT EVERY READING (Ninad, CLI-695, 2026-10-04). Shown CAROL-ARC-0036's status from the Pipeline app, his words: 'P2 Each link serves one session - 9 initiatives, 6 completed, and it says 100% complete? 1. you should be taking this information from the same registry tables and the pipeline app fetches this information from 2. the estimated efforts and % completion should be uptodate and accurate always'.

READ AS (Orion's reading, not the ruling): #205 stands -- an arc's completion is effort spent, read from the Pipeline app -- and every reader, Orion's status replies included, takes it from the one effort composer the app reads. What its build read into #205 is corrected: an attempt that has been worked but whose initiative is not closed has NOT spent its whole estimate. Its spent part is its estimate less the effort still REMAINING, which Elrond's Effort Estimator sizes from the attempt's own record whenever it comes to rest open; so a phase or an arc with any member not closed (or set aside) never reads 100%. Every new attempt is sized, and every open attempt's remaining effort re-sized after it moves, within minutes rather than in a daily batch, so neither figure is stale.

New work joins an existing arc only when it depends tightly on that arc

Ruled 2026-10-04 by ninad -- in force

NEW WORK JOINS AN EXISTING ARC ONLY WHEN IT DEPENDS TIGHTLY ON THAT ARC; WORK THAT CAN RUN ON ITS OWN GETS ITS OWN ARC (Ninad, CLI-697, 2026-10-04). Asked to file the work for Orion's permission lists and Saee's stable door, he asked: 'would you recommend creating a new arc or would you rather fold this work into operator approval arc? this should depend on the dependencies.. for eg, if there are tight dependencies on the aperator approval arc, then keep it inside it, but if i can run the arc independently, then make it a seperate arc'.

READ AS (Orion's reading, not the ruling): before placing new work in an arc, read what it must wait on and what waits on it among the arc's OPEN members; tight dependencies keep it in that arc (a new phase, or the phase of what it is tied to), otherwise it becomes its own arc under its service's owner, and any single crossing dependency is recorded at both ends as an [arc-dependency] decision. Applied at once: the permission lists depend on no open member of CAROL-ARC-0036 and nothing there waits on them, so they form their own arc; the stable door is tied to Saee's door (CAROL-ARC-0025 phase 6) and the open work on it, so it joins that phase.

A lane stopped for a cured cause is restarted by an estate operator's own decision; Albus's lane restarted 2026-10-13

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #53, 09:52 UTC): Albus's bypass lane stopped itself 2026-10-08 16:31 UTC when his Codex sign-in failed; the sign-in works again and only the lane's kill switch holds it. Sahil asked: "1. Should Albus's lane be restarted now? 2. For the future: when a lane stops for a reason that has since been cured, and restarting it only lets it spend may Sahil decide the restart himself? Your 29 September ruling already says you're needed only for root and core steps. What should I tell him?" Offered: "Restart; he decides" -- "Restart now. In future, when a lane stopped for a cause that's been fixed and restarting only spends Sahil decides himself. This is what your 29 September ruling already says." HIS ANSWER: "Restart; he decides". READ AS (Orion's reading, not his words): Albus's lane is restarted now; and whenever a lane stopped for a cause that has since been cured and a restart only lets it spend an estate operator decides the restart without the administrator -- security #213 applied to lane restarts.

An arc groups initiatives by subject; a pursuit is oriented to a goal

Ruled 2026-10-14 by ninad -- superseded 2026-10-14

AN ARC GROUPS INITIATIVES BY SUBJECT; A PURSUIT IS ORIENTED TO A GOAL (Ninad, CLI-761, 2026-10-14). THE QUESTION ANSWERED: two initiatives filed for removing Google's 'unverified app' warning from the mail consent (6581, 6582, pursuit CAROL-PUR-0009) sat in the mail arc CAROL-ARC-0042 while another session was building Themis's compliance arc CAROL-ARC-0043. HIS WORDS: "note that these initiatives can be part of the compliance arc the other session is currently building, while remain on this pursuit.. arcs are subject specific initiatives while pursuits are goal oriented (pls check if the dictionary definition matches this understanding)"; shown that the dictionary called an arc 'one service's family, owned by that service's owner': "update the dictionary..". READ AS (Orion's reading, not his words): an arc is the group of initiatives about one subject, in ordered phases, under the agent who owns that subject; an initiative joins the arc of its subject even when another service's code does the work (the Google verification steps belong to the compliance arc); a pursuit is one goal -- an outcome, its success criteria and the initiatives it needs from any arcs, in order and ranked; an initiative may sit in an arc and a pursuit at once. Where earlier text says an arc is 'one service's family' (initiatives #121/#124's arc clause, the skill build-arcs-and-pursuits, the dictionary), this later ruling governs (policy P.01.01.05.08, the later ruling wins within one operator rung). Membership through a phase (#157) is unchanged.

Ended because: Ninad's later answer (CLI-768): an arc is service specific, so its subject is that service's subject (P.01.01.05.08)

Every initiative names its one service; Elrond's gate refuses one that does not, and every filing agent learns to name it

Ruled 2026-10-14 by ninad -- in force

Ninad, 2026-10-14 (CLI-767/CLI-768), point 5 of six: "1. I would like service operators to be mapped to services and to the agent that own those services.. this would mean that they can work only on arcs that are owned by the agents of that service. 2. Pursuits can technically span across services.. this means that service operators cannot work on pursuits unless they are mapped to every service included in the pursuit.. 3. Estate operators can map service operators to services, by mapping them to the corresponding agents 4. Estate operators can assign arcs and pursuits to service operators, provided they are mapped to the services / agents that are included in the arc or pursuit.. if not, orion advices estate operators to map the service operator to the respective additional serivces and agents 5. Assignment of services should be mandatory for each initiative, and elronds gate should check for it and reject any non compliances. Agents should learn to provide the service corresponding to each initiative. 6. If an initiative spans across services, it needs to be broken down to align with services. This knowledge needs to be imparted to all agents responsible for filing initiatives -- each of these should be also documented as policies and terms of the dictionary should be updated accordingly. Add these to the appropriate arc, or file a new arc if no arc already covers this topic". READ AS (Orion's reading, marked apart, P.01.01.05.05): every initiative carries exactly one service -- the service whose work it is -- recorded when it is filed, and an initiative placed in an arc names that arc's one service; Elrond's gate refuses, by name, a filing that names no service, names one the registry does not hold, or names more than one; every agent that files initiatives is taught to name the service, through the skills it files with. Initiatives filed before this ruling are not stamped with a guessed service.

An initiative that spans services is broken into one initiative per service, and every filing agent is taught this

Ruled 2026-10-14 by ninad -- in force

Ninad, 2026-10-14 (CLI-767/CLI-768), point 6 of six: "1. I would like service operators to be mapped to services and to the agent that own those services.. this would mean that they can work only on arcs that are owned by the agents of that service. 2. Pursuits can technically span across services.. this means that service operators cannot work on pursuits unless they are mapped to every service included in the pursuit.. 3. Estate operators can map service operators to services, by mapping them to the corresponding agents 4. Estate operators can assign arcs and pursuits to service operators, provided they are mapped to the services / agents that are included in the arc or pursuit.. if not, orion advices estate operators to map the service operator to the respective additional serivces and agents 5. Assignment of services should be mandatory for each initiative, and elronds gate should check for it and reject any non compliances. Agents should learn to provide the service corresponding to each initiative. 6. If an initiative spans across services, it needs to be broken down to align with services. This knowledge needs to be imparted to all agents responsible for filing initiatives -- each of these should be also documented as policies and terms of the dictionary should be updated accordingly. Add these to the appropriate arc, or file a new arc if no arc already covers this topic". READ AS (Orion's reading, marked apart, P.01.01.05.05): when the work an initiative asks for falls to more than one service, it is filed as one initiative per service, each naming its own service, related to the others, and placed in the arc of its subject and, where they serve one goal, in one pursuit; every agent responsible for filing initiatives is taught this through the skills it files with.

Elrond's service gate refuses only after the agents file with a service and the old initiatives carry theirs: teach, backfill, then enforce

Ruled 2026-10-14 by ninad -- in force

Ninad, 2026-10-14 (CLI-772). ASKED (CLI-772): "6596 builds Elrond's gate refusing any initiative that doesn't name its one service. Agents won't know how to name one until the next step (6597) teaches them. Last week 97 of 170 new initiatives came from agents (Hermione's failure reports 73, route probes 15, quality reviews 6, others 3). If the gate refuses from day one, about 14 agent filings a day get bounced (each recorded and counted, none lost silently, but Hermione's failure reports wouldn't become initiatives) until 6597 lands. When should the gate start refusing?" (offered: record first, refusing only operator filings; refuse from day one; build 6596 and 6597 together). HIS ANSWER: "teach agents to file services against initiatives, then backfill old initiatives with the services and then enforce the gate". READ AS (Orion, not his words): three steps in this order -- (1) every filing agent names the one service of what it files; (2) the initiatives filed before that carry their one service too, backfilled; (3) only then does Elrond's gate refuse a filing that names none. Until (3) the gate records what it would refuse and refuses nobody for a missing service, operator filings included (none of the offered options was chosen). His word "backfill" sets aside, for this record, 's "nothing is stamped with a guessed service": the backfill is his instruction; each backfilled service records where it came from, and how an initiative whose records name no service is placed is brought to him with the numbers before it is done.

Orion and Albus raise no escalations: each keeps a ranked backlog of at most 20 initiatives he raised, and the pipeline backlog shows agent escalations, agent wishes, Albus's backlog and Orion's backlog apart

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-781). HIS WORDS: "Agents are raising escalations left, right and center, and the count of un-attended escalations is rising, and its not helping the cause, because the pipeline is not able to cater to them in time. This needs to be regulated, which means that agent's should be able to raise escalations and wishes only at a pace in which Albus and Orion can resolve them.. this means, each agent can have 2 escalations and 1 wish open at a time.. Agent can prioritize and ensure that the most important escalations and wishes are filed, they can maintain their own backlog of escalations and wishes, and they can withdraw an un-attented escalation or wish to replace it with another escalation or wish that is higher on priority according to them. Orion's and Albus' escalations are exempt from this rule, for them, they dont raise escalations, as they cant escalate to themselves, they create backlogs, and this backlog can go upto 20 initiatives max, sorted by priority, following the same process of prioritization like the agents follow.. the pipeline backlog should cleanly differentiate the agent escalations and wishes from orion and albus backlog and orion sessions should use the right naming mechanism to communicate.. update the dictionary with this nomenclature and implement the harness to control the pipeline backlog to a manageable limit sorted by priority and importance" ASKED (CLI-781), four questions, each with his answer: (1) 'Orion's lane has 436 pending initiatives and Albus's has 289. Which of these count toward the 20-item cap on Orion's and Albus's backlogs?' (offered: only what each raised himself; everything on their lane) -- HIS ANSWER: "Only what each raised". (2) '147 escalations are open today, and 124 of them come from Elrond's unmeasured-call watch. Orion's own backlog is also well over 20. What happens to everything above the new caps?' (offered: back to the raiser's list; discard with a reason; cap new raises only) -- HIS ANSWER: "Back to the raiser's list": everything over the cap leaves the pipeline, nothing is deleted, it waits in the raiser's own ranked list and goes back in as space frees up. (3) 'Agents also file initiatives straight onto Albus's lane without escalating: 62 pending from Hermione and 13 from Inspector. How should those direct filings count?' (offered: as escalations; stop direct filing; leave them uncapped) -- HIS ANSWER: "As escalations": a direct filing by an agent takes one of its 2 open escalation places. Your rule lets an agent withdraw a wish and replace it with a more important one, but the 30-day limit would refuse that replacement. READ AS (Orion, not his words): Orion and Albus raise no escalations. Each keeps a BACKLOG of at most 20 pending initiatives he raised himself, ranked by priority and importance by its holder, worked in that order; anything further waits in his own ranked list outside the pipeline and is swapped in by the same process the agents follow. ORION'S BACKLOG is what Orion raised (a filing from an Orion session outside any arc or pursuit, initiatives #258, and his mid-work findings); ALBUS'S BACKLOG is what Albus raised. The operators' arc and pursuit work, and agents' escalations and wishes, sit outside both. The PIPELINE BACKLOG shows apart: agent escalations, agent wishes, Albus's backlog, Orion's backlog, and the operators' arc and pursuit work; Orion sessions say each by these names. What sits above 20 today goes back to Orion's or Albus's own list, nothing deleted.

Every initiative carries a type from a governed list: agent escalation, agent wish, change request, and any type already in use

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-795). ASKED: 'Your ruling splits an agent's own filing into two kinds: a fault takes one of its two escalation places, a change it wants takes its one wish place. How should the estate tell which kind a filing is?' -- offered: (1) the filer says, and one that says nothing takes an escalation place; (2) Elrond's gate judges each filing, recording its reason -- HIS ANSWER: "the type field should be an RDM value that allows escalation, wish, change request, and any other type that might be existing, if at all". READ AS (Orion, not his words): every initiative carries a TYPE whose allowed values are a governed reference-data set (RDM; one door, enforcement generated at the write seam, never a list typed in code): agent escalation, agent wish, operator change request, user change request (the kinds of agent-resources #407), plus any type already in use on the estate that the record shows -- found and added, never invented. An agent's own filing of type agent escalation takes one of its escalation places; of type agent wish, its wish place; a change request takes no agent place. Who sets the type was not answered: following the precedent he set for the one service (initiatives #391: teach the filers, backfill the old ones, then enforce), the filer names the type, the existing pending initiatives are backfilled from what the record shows with provenance, and the gate is enforced only after both.

The initiatives the service backfill cannot place from their own records wait for the backlog to be cut to what the pipeline can handle; those that span services are split

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-804), answering 's criterion 2. ASKED: "Backfilling the one service onto old initiatives: of 642 pending ones, 276 can be placed from their own records (the service task they serve, their arc, their parent). 366 can't: 48 have records naming two services (e.g. the Google verification steps serve a communications task but sit in Themis's compliance arc), 60 sit only in a catch-all arc, 258 have no evidence at all. How should those 366 get their service?" (offered: Orion proposes and he approves a list; leave them unplaced for now; conflicts follow the arc and the rest are proposed). HIS ANSWER: "initiatives spanning across services will have to be split, but this is also the problem of having too many backlogs, hence the efforts to reduce the backlogs only to the extent that can be handled by the current pipeline, an ongoing effort via the escalation management arc". READ AS (Orion's reading, marked apart, P.01.01.05.05): the 276 the records place are recorded now through the one door with their evidence; the 366 are NOT placed by a proposal list now -- the pending backlog is first cut to what the current pipeline can handle by the ongoing backlog work (read as CAROL-ARC-0040 Escalation streamlining and CAROL-ARC-0045 A pipeline backlog Albus and Orion can clear); an initiative that survives that cut and spans services is split into one initiative per service (initiatives #384), and one that does not span gets its service when it is next worked or filed again. The count of pending initiatives with no service is therefore expected to stay above zero until that work lands; Elrond's gate keeps recording, not refusing (#391), until is decided.

Orion's backlog and Albus's backlog are values of the initiative type, beside agent escalation, agent wish, operator change request and user change request

Ruled 2026-10-15 by ninad -- in force

Ninad (CLI-806). ASKED: 'Ruling needed: what type is an initiative Orion files from his own work? My recommendation: follow the existing raiser rule -- inside an arc or pursuit it is raised by you, so an operator change request; outside one it is Orion's own, an agent escalation if something meant to work does not, an agent wish if it is a change he wants.' He asked for the distinct types and was shown the four values of the initiative type (agent escalation, agent wish, operator change request, user change request). HIS WORDS: "plus orion backlog and albus backlog, these should be RDM values". READ AS (Orion, not his words): the governed initiative type gains two values, ORION BACKLOG and ALBUS BACKLOG, the two backlogs initiatives #403 names (Orion and Albus raise no escalations; what each raises is his own backlog). An initiative Orion raises from his own work is typed Orion backlog, never agent escalation or agent wish; which Orion-session filings Orion raised follows initiatives #258 (one in an arc or a pursuit is the operator's, an operator change request). An initiative Albus raises himself -- his own first ask, never work another party asked for that was routed onto his lane -- is typed Albus backlog. Neither takes an agent's escalation or wish place. The two values live in the RDM set initiative.type like the other four (one door, enforcement generated at the write seam). This settles the law gap was parked on.

Albus's backlog includes his own escalations: what Albus raises about his own work is his backlog

Ruled 2026-10-16 by ninad -- in force

Ninad (CLI-806). ASKED, proposing the dictionary words for initiatives #431: 'Albus backlog: work Albus raises for himself, never work someone else asked for that was routed to his lane. Approve these as written?' HIS WORDS: "albus backlog would include his own escalations as well". READ AS (Orion, not his words): everything Albus raises himself is his backlog -- the escalations he raises (to Orion, about something meant to work that does not), his wishes and his other own filings alike -- whether he or one of his droids raises it, and whatever agent type a filing names. None of it takes an agent escalation or wish place (Albus holds none, #403); all of it counts toward his backlog of at most twenty (#402, #403). Work another party asked for that is routed onto his lane is still not his backlog (#431).

Orion-session filings in an arc or a pursuit stay the operator's change requests (#258 kept)

Ruled 2026-10-16 by ninad -- in force

Ninad (CLI-817), on reading the pipeline backlog's six groups: "operator change request count 247 does not sound right". ASKED: 'The 247 holds up under your ruling #258. Every one is a filing from an Orion session that now sits in an arc phase (190) or a pursuit step (56 more), and #258 makes such filings yours. By who first asked: 157 came from sessions you drove, 23 Orion raised himself and were placed in an arc later, and 66 predate the origin record. By status: 170 planned, 46 parked, 19 redirected, 7 executing, 4 blocked. The biggest holders are Architecture compliance (23), Two operators (20) and Agent escalations (17). Do you want that rule changed?' -- offered: (1) 'Keep #258: in an arc or a pursuit means it is the operator's work; 247 stands'; (2) 'Orion's own stay his: what Orion raised himself counts as Orion's backlog even inside an arc'; (3) something else -- HIS ANSWER: he chose (1), "Keep #258". READ AS (Orion, not his words): initiatives #258 stands unchanged -- an Orion-session filing is the operator's change request while it sits in an arc phase or a pursuit step, whoever first asked for it, and Orion's backlog only when it sits in neither.

An agent's own pending filing that names no type is typed by the agent itself, asked at its next wake; waits for the answers

Ruled 2026-10-19 by ninad -- superseded 2026-10-19

Ninad (CLI-867, 2026-10-19), answering in session. ASKED: '6691 (every pending initiative names its type) is down to nine old filings that agents made themselves: six of Elrond's (four review follow-ons, two of them duplicates; a doctrine proposal; a regression investigation), plus one each from Hermione, Hagrid and Leo. The records can't say escalation or wish; that needs a judgment on each one's content. Typing them also counts them against the agent's places, and Elrond and Hermione are already full, so by the 15 Oct rule their excess would go to their own backlogs, with the initiatives discarded and named on the backlog item. How should I settle them?' Offered: Orion classifies and the places rule applies (recommended); classify and keep all live; ask the agents. HIS ANSWER: "Ask the agents" -- the option reading "Leave them untyped for now; each agent says what its filing is at its next wake, and 6691 stays open until they answer." READ AS (Orion, not his words): Orion does not type an agent's own filing from its content. Each raiser is told by message, read at its next wake, which of its pending filings name no type, and states on that initiative (a decision, an asking act no switch or budget stops) whether it is an agent escalation, an agent wish, or no longer true; Orion records the type with that decision as its source. Until all have answered stays open (parked) and the type gate keeps recording rather than refusing.

Ended because: Ninad, CLI-869: the agents asked under #454 cannot wake to answer (the consciousness wake budget is spent after the first four agents every day); he chose 'Orion types the nine'.

Orion types the nine old agent filings waits on, from their records, because the agents cannot be woken to answer

Ruled 2026-10-19 by ninad -- in force

Ninad (CLI-869, 2026-10-19), answering in session. ASKED: "Your 'ask the agents' answer can't happen as things stand. Elrond, Hermione, Hagrid and Leo haven't woken since 3-13 Oct. The other 37 agents have been skipped every day since at least 12 Oct. How should I proceed with the nine filings?" Offered: one wake for the four (recommended); every agent wakes daily; Orion types the nine; wait. HIS ANSWER: "Orion types the nine" -- the option reading "I type the nine filings from their records (the option you declined earlier), since asking can't reach the agents." READ AS (Orion, not his words): #454's 'each agent says what its filing is at its next wake' is superseded for these nine filings. Orion records each one's type from what its own record says it is -- agent escalation (something that was supposed to work is not working) or agent wish (a change the agent wants) -- naming the record as the source, and the places rule applies as for any typed agent filing (agent-resources #402, initiatives #418): one past its agent's places waits in that agent's own backlog. The messages already sent stay; an answer an agent posts later is read and, where it differs, recorded as a new type event naming that answer. The wake budget itself is not decided here.

No two arcs are dedicated to the same work; overlapping arcs are merged into one

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884). HIS WORDS, first: "close the loose ends, and there should not be multiple arcs dedicated to the same work, as they would end up either duplicating the work or conflicting the work.. check if there are any such arcs (for eg blueprint, project onboarding, and chowpatty maybe)". Orion then reported seven groups of arcs whose subjects overlap (Service reusability with Services consumed through an API; Redirects sitting with Orion with Escalations from the Albus bypass; Albus's lane clears its work with A pipeline backlog Albus and Orion can clear; Two operators split across the operator and security arcs; Every ruling in one home with Rulings follow the operator hierarchy; the lane-moving phase of The record of every model call with The adapter layer; Carolopedia's old look with Carolopedia's content restored) and work filed more than once, and asked: "Shall I carry out the seven merges and close the duplicate filings as recommended? I'll check each item before moving it." HIS ANSWER: "yes to both, send a communication to sahil if any of his arcs or pursuits are affected". READ AS (Orion's reading, not his words): no two arcs are dedicated to the same work. Where two arcs' subjects overlap, one survives and the other's open work moves into it; the arc that gives up its work keeps its closed members as history and is retired in place, never deleted (cookbook 1670). Work filed more than once is closed onto the one that stays, naming it. A pursuit that crosses arcs is not a duplicate (1790's pursuit row). Every estate operator whose arcs or pursuits a merge touches is told through an operator communication.

An Orion session files an arc's initiatives on behalf of the arc's owner

Ruled 2026-10-21 by ninad -- superseded 2026-10-21

AN ORION SESSION FILES AN ARC'S INITIATIVES ON BEHALF OF THE ARC'S OWNER (Ninad, CLI-894, 2026-10-21). Shown that all five initiatives of CAROL-ARC-0046 'Only an artifact's owner edits or deletes it' (arc owner Heimdall) name Orion as their owner because Orion filed them: "that is a gap, orion should have filed them on behalf of heimdall, what do you think?". Shown Orion's proposal -- when an Orion session files into an arc, the arc's owner becomes the initiative's owner and Orion files on the owner's behalf; who raised it stays the operator; the performer stays the Orion bypass; the acceptor stays Elrond (his sign-off follows the lane, not the owner); the owner is told at its next wake and may disagree; the open 6782 moves to Heimdall now; the four closed ones stay as recorded; filings in no arc stay Orion's own backlog -- and that it changes the rule that an initiative's owner is whoever asked, he answered: "this is a structural change, fold it into the right arc and implement it first". READ AS (Orion, not the ruling): (1) the owner of an initiative an Orion session files into an arc phase, or places in one, is that arc's owner agent, read live from the arc record -- never typed, never Orion's login -- and Orion is recorded on the initiative as the filer on that owner's behalf; (2) while the initiative is pending the owner follows its arc membership (placed -> the arc's owner; taken out of its last arc -> Orion's own backlog again), and a closed, discarded or redirected one keeps the owner it had, never rewritten; (3) who raised it (initiatives #258, #260), its type, the lane that performs it and its acceptor are unchanged, and the owner holds no agent place for it (an operator change request takes none, agent-resources #402); (4) the owner is told at its next wake and may disagree (the shape of security #452); (5) cookbook 724 ('the initiative's owner IS its requester') and 's owner-is-the-mapped-requester are narrowed in place for this case.

Ended because: Ninad, CLI-894: 'a requester cannot be the owner ... whoever is paying for the initiative should be the owner'

An initiative's owner is whoever pays for it; a requester is only a requester

Ruled 2026-10-21 by ninad -- superseded 2026-11-07

AN INITIATIVE'S OWNER IS WHOEVER PAYS FOR IT; A REQUESTER IS ONLY A REQUESTER (Ninad, CLI-894, 2026-10-21). Shown initiatives #486 built as (an Orion-session filing in an arc owned by the arc's owner; 201 pending initiatives moved, 6782 to Heimdall) and proposed policy words for P.01.01.01.10 that kept 'the owner of an initiative is the agent who requested it' with that one exception, he answered: "a requester cannot be the owner, instead, whoever is paying for the initiative should be the owner of the initiative.. a requestor is just a requestor. The payment is about budget that enables building or expanding a service. The budget comes from project. Project belongs to a client (say denken-labs) who assignes budgets for expanding existing services and building new services. Each project should have the CEO and head of finance (the agent and employee mapping in future, estate operators get mapped by default) The agent economy is tightly intermingled within the human economy in carolverse.. note this ruling, let the next session instantiate or manifest it into the estate". READ AS (Orion, not the ruling): (1) the owner of every initiative -- not only an Orion-session filing -- is the party whose budget pays for it, never the one who asked; the requester stays recorded as the requester, and who raised it stays initiatives #258/#260's answer; (2) cookbook 724, 's owner-is-the-mapped-requester and policy P.01.01.01.10 ('the owner of an initiative is the agent who requested it') are retired in place by it, and #486 (the arc's owner) is superseded: the 201 owners moved under #486 on 2026-10-21 are to be re-read against the payer when this is built, never left standing as a third state that agrees with neither ruling; (3) the payer is read from the money chain the estate already keeps (the work's ticket -> its track -> its service -> the project whose budget it is: the cost-center ruling recorded with this one, 'The budget that builds or expands a service comes from a project, and the project's client assigns it'), never typed; (4) OPEN for Ninad before it is built: whether the owner recorded is the paying PROJECT, answered for by its CEO and head of finance (the blueprint ruling recorded with this one, 'Every project has a CEO and a head of finance'), or an agent; and whether the acceptor still follows the lane (rule 1566) or follows the new owner.

Ended because: Ninad ruled that whoever files an initiative owns it and the paying service's owner only approves it (CLI-058)

An initiative's owner is the project, through the service the initiative improves; the arc's owner owns the initiatives in the arc

Ruled 2026-10-21 by ninad -- retired 2026-11-07

AN INITIATIVE'S OWNER IS THE PROJECT, THROUGH THE SERVICE THE INITIATIVE IMPROVES; THE ARC'S OWNER OWNS THE INITIATIVES IN THE ARC (Ninad, CLI-896, 2026-10-21). Asked whether the owner recorded is the paying project or one of its agents, he answered: "project, yes, but every project is made of services, and each initiative belongs to a service, depending on which service it ends up improving. If initiatives spread across services, then they are split by elrond who acts as the mediator between the requestor and the owner. This translates into the owner of the arc also becomming the owner of the initiatives within the arc, because the arc is also service specific, and the service operator assigned to the arc becomes the one that is executing the arc via elrond's team.." READ AS (Orion, not the ruling): (1) the owner of an initiative is the owner of the ONE service it improves (its recorded service, initiatives #383-#385, #391), read live from the services catalogue, never typed; an arc belongs to that service, so its owner and the initiative's agree -- where an arc is led by another agent of the service (four live arcs today), the service owner owns and pays and the arc owner leads; (2) work that improves several services is split by Elrond into one initiative per service before it is filed (#384); (3) the service operator assigned to the arc executes it through Elrond's team; (4) #487's 'whoever pays' is the project, paying through that service's budget (cost-center #488); (5) initiatives #486 (the arc's owner) is restored in substance for arcs, as a consequence of the service rule rather than of Orion's filing.

Ended because: Set aside by #666 (CLI-058): an initiative's owner is the agent that filed it; the service it names pays and its owner approves (by ninad)

An architecture violation is addressed first: an arc whose goal rests on it depends on the compliance arc and completes only after it, or after the initiatives it depends on

Ruled 2026-10-25 by ninad -- in force

AN ARCHITECTURE VIOLATION IS ADDRESSED FIRST (Ninad, 2026-10-25, CLI-938).

ASKED (after security #515; completing arc CAROL-ARC-0046 'Only an artifact's owner edits or deletes it'): 207 shared code files (153 shared modules, 54 tools) name a store but have no owner on record; how should Orion place them? -- offered: Orion judges and records each with its evidence; only what records prove; Albus proposes and an operator approves; park it. ANSWERED: "if files dont have an owner on record, is that an architecture non compliance? is there an initiative that deals with this? or an existing arc for this? if this is an architecture violation, that needs to be addressed first, and this arc should be made dependent on the complaince arc, so this arc can only be completed post that arc (or post the dependant initiatives)".

FOUND, answering his questions: yes -- the one architecture checker's rules A1.owner and O.no_orphan_code make a code file with no owner a violation; arc CAROL-ARC-0026 'Architecture compliance' (Themis) holds 'Every code file has one owning agent' (phase 3, planned: owners recorded in the registry's module-owner record, proposed from evidence and decided by Albus as the standard's owner) and (orphan code, phase 6, parked).

READ AS (Orion's reading, not the administrator's words, NOT YET CONFIRMED by him): an arc whose goal cannot be met while an architecture violation stands is made dependent on the work curing it and is not completed before that work closes; here CAROL-ARC-0046 completes only after closes (the compliance arc's initiative it depends on), and, filed for the same gap, is its duplicate. No record yet holds an arc's dependency on another arc: until one is built the dependency is carried by the pursuit 'Every artifact is held to its owner' (CAROL-PUR-0012), whose step names it.

CAROL-PUR-0012's MVP: write routes and stores several logins write first; single-writer stores after the MVP

Ruled 2026-10-26 by ninad -- superseded 2026-10-26

Ninad, CLI-969 (2026-10-26). ASKED, after he asked whether the pursuit 'Every artifact is held to its owner' (CAROL-PUR-0012) is necessary and what it brings: 'Not all 41 steps are equally urgent, though. The real risk is in the write routes and the stores several logins share. Moving stores that only one login writes is mostly tidying, so those steps could be moved after the MVP if you want the pursuit finished sooner.' HIS ANSWER: "ok do that and show me the updated status". READ AS (Orion, not the ruling): the pursuit's MVP is every app write route retired, every store that the code of two or more logins writes held to its one owner (with the owner door other agents' processes reach it through), and every store owned; a store whose code is written by one login alone is held after the MVP. A step holding both kinds is split so each part sits in its stage; the outcome and its criteria are unchanged, and the pursuit is achieved only when every step, post-MVP included, is closed (security #456, #457 stand).

Ended because: Ninad chose to leave the pursuit as staged once the measured saving (under a tenth of the MVP) was put to him

CAROL-PUR-0012's MVP stays as staged: every step MVP

Ruled 2026-10-26 by ninad -- in force

Ninad, CLI-969 (2026-10-26), superseding initiatives #535 on new evidence. ASKED: 'I checked who actually writes each store, and the saving is smaller than I suggested. Only about 34 of roughly 110 unheld stores are written by a single login, and about 70% of the remaining effort is retiring app write routes, which stay in the MVP whatever we choose. Moving the single-writer stores out would shrink the MVP by under a tenth. It would also take about 20 record changes: narrowing 11 steps, plus about 11 new post-MVP steps. Do you still want the split?' Options: 'Leave as is (Recommended)' -- keep all 41 steps in the MVP; 'Split anyway'. HIS ANSWER: "Leave as is (Recommended)". READ AS (Orion, not the ruling): the pursuit 'Every artifact is held to its owner' (CAROL-PUR-0012) keeps all its steps in the MVP as staged before #535; no step is split or moved to post-MVP on the single-writer test.

Elrond's estimator sizes an initiative as soon as it is filed, not when its schedule next runs

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). THE QUESTION: Orion reported that four initiatives just filed read unsized until Elrond's Effort Estimator sizes them. HIS ANSWER, word for word: "Elrond's estimator sizes them - this should be triggered as soon as initiatives are filed, not wait for the schedule to run". READ AS (Orion's reading, not his words): filing an initiative starts Elrond's Effort Estimator for it at once, under Elrond's own login and paid from his own track; the scheduled run stays only as the catch-up for a filing the trigger missed; a filing is never held or refused because sizing failed, and a failed sizing is recorded on the initiative.

Every pursuit carries a short name beside its action

Ruled 2026-10-26 by ninad -- in force

Ninad, 2026-10-26 (CLI-968). THE QUESTION: asked the names of pursuits 0004 and 0006, Ninad was given 'Show how far every arc is, by effort spent' and asked "that is not a name, is it?". Orion answered that every pursuit records its ACTION as a sentence in its name field (the dictionary defines a pursuit as a prioritised action), that no separate short name exists and nothing can rename one, and asked: should every pursuit get a short name alongside its action? HIS ANSWER, word for word: "yes please, do that". READ AS (Orion's reading, not his words): every pursuit carries a short name -- a few words that name it, never a sentence -- beside its action, which stays as recorded; the short name is journalled like the rest of the record, set through the pursuits door, given when a pursuit is created at the operator's window, and shown with the key wherever a pursuit is listed, the action beneath it.

The owner of the service an initiative improves approves work another agent asks for, and Elrond's one prioritisation serves every agent

Ruled 2026-10-28 by ninad -- in force

"since the initiatives are being paid by the service that gets improved and not the service that requests the change, this would need approval from the service owner, if the requesting agent is not the same as service owner of that service being improved. The service owner should decide based on urgency, importance, and his budget constraints, similar to how others like elrond or albus prioritize the initiatives.. infact, initiative prioritization should be a reusable service owned by elrond, and used by other agents for prioritizing their tasks.. fold this requirement in the pursuit: \"Service-owned initiatives\"". READ AS: (1) an initiative whose first asker is an agent other than the owner of the one service it improves (#491) waits for that owner's approval before it is dispatched or a bypass starts; the owner approves, declines with a reason, or defers, judging urgency, importance and what its budget can carry, and the asker reads the decision; an approval waiting on an owner is named on the board with whom it waits on, and nothing approves for the owner. An operator's own request is not put to an agent's approval (operators sit above every agent, security #197); a user's request reaches the owner like an agent's. (2) Prioritisation -- ranking initiatives or tasks by urgency, importance and budget, each with a reason the asker reads -- is ONE reusable capability of Elrond's Build Initiatives service, carried on his Build Governance track (#151), which Elrond, Albus, every service owner and any agent ranking its own tasks call instead of a ranker of their own; its cost for another agent's list is charged to that agent's service (#492, #557). Both are steps of pursuit CAROL-PUR-0011.

How a pursuit's estimated hours are worked out is a ruling in the Rulings Repo, not a dictionary entry

Ruled 2026-10-28 by ninad -- in force

"the wording is too difficult to understand, its not a dictionary candidate, its ruling that should go in ruling repo associated with the right app, service, track, agent, droid, task, etc.. let the next session pick that up" -- answering Orion's proposed dictionary entry 'Pursuit estimated hours'. READ AS: the rule for working out a pursuit's estimated hours is recorded as a ruling in the Rulings Repo, in plain words, and tied to what it governs -- the Pipeline app, the initiatives service and its track, the composer shared machinery, and its owner -- never as a dictionary definition; 's second criterion (the dictionary carries the terms) and the Pipeline app's note citing a dictionary entry change to match, by a governed goal change, in the next session. How this sits with cookbook 938 (every displayed metric has a registered dictionary definition) is for that session to settle with the operator.

A pursuit's steps never wait on Orion's backlog

Ruled 2026-11-03 by ninad -- in force

Pursuits are operator initiated: an initiative filed as a step of a pursuit is the operator's work from its filing and never waits for a place in Orion's backlog. Until the filing door can be told the pursuit at filing, a step held there enters at the top of Orion's list and joins its pursuit at once, which frees the place again (initiatives #431, #258: membership makes it the operator's).

Pursuits and arcs keep versions and a dated history of their initiative count and estimated hours

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-030): "can you maintain the history of pursuits to show how it changes overtime? KPIs being, total count of initiatives and total estimated hrs" and "also, pursuits should have versions capturing details like changes to the name, title, tag, description, ownership etc.."; told that pursuits have no tag, he answered "tag is like something_something, I think its present for arcs, but please check.. its also necessary that the history of the underlying arcs should be tracked in similar manner.."; asked how the hours history should work, since past estimates were overwritten and the hours rate moves daily while the rules keep such figures derived, he chose "Daily record from today" (record the usual-format figures once a day and at every change, kept as dated readings; the count history goes back to creation, the hours history starts today); and he chose the Orion bypass, now. READ AS: every pursuit and every arc keeps VERSIONS, one per recorded change to any of its fields (a pursuit's action, short name, outcome, MVP, owning agent, operator owner, who asked, status, rank, criteria and steps; an arc's name, description, tag, owner, status, lane, roadmap link, service, estimates and phases), read from the registry's change journal, never a second copy. Arcs carry a tag (checked: e.g. employee-onboarding); a pursuit has none, because an arc's tag is what its members carry and a pursuit never re-tags initiatives (initiatives #121, #124), so a pursuit's short name stands where an arc's tag does. Both keep a dated HISTORY of their initiative count and estimated hours: a reading of the usual-format figures recorded once a day and at every change, append-only, never read as the current state; the count before the first reading is derived from the change journal where it holds membership, the hours start at the first reading and are never backfilled.

Albus's lane picks as much of the whole pending backlog as his budget permits, in the priority engine's order, and sends what he cannot do to the Orion lane

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-05 (CLI-035). albus should pick as many as his budget permits using the priority engine.. and he always has the orion lane to redirect if he encounters problems". Orion then showed that the limit sits in Albus's runner and that Elrond's one prioritisation (#558) is not built yet, and asked: 'Shall I record the ruling and file the runner change?' HIS WORDS: "yes, infact 'own filings only' was a surprise". READ AS (Orion, not his words): (1) Albus's bypass lane picks from EVERY planned bypass-mode initiative waiting in the pipeline backlog, whoever filed it -- not only what he filed himself -- and takes as many as his lane's budget permits; an initiative already claimed by an Orion session is not picked. (2) The order is the priority engine's: Elrond's one reusable prioritisation (#558) once built; until then Albus's own ranking of everything waiting on his lane (#227). (3) Work on which he meets a problem goes to the Orion bypass lane, the way #230 (2) already sends work his lane cannot do, and is not retried on his lane. (4) The 'own filings only' limit of his backlog pick class is retired in place. Unchanged: #403, #431 and #432 say what counts as Albus's backlog as a TYPE of filing (what he raises); this ruling is about what his lane PICKS. His budget is not ruled on here (#555 left it to Midas's allocator).

A filing the operator's Orion writes on the operator's own Claude plan is recorded by Elrond's filing step without being written or sized again on his budget

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-040). Told that Elrond's filing budget was spent and asked to raise it again, HIS WORDS: "can you not use your credits instead? you have enough credits to spare this week, would that require another build? or just a ruling?"; asked "when you work on behalf of an agent via the agent's window, do you use your credits or the agent's credits?" and told it depends on who does the thinking, HIS WORDS: "ok, go ahead and implement that one, it will save us from manually adjusting budgets". READ AS (Orion's reading, not his words): (1) when an operator's Orion session writes a filing in full -- its title, description, success criteria, plan, its one service and its size -- on the operator's own Claude plan, Elrond's filing step records it as written and makes no model call of its own, so nothing is (2) every check the filing step and the filing door make still applies -- the operator's token, one service, a roadmap link, the raiser, the type and the places -- and a filing that lacks any part is refused by name, never completed by a paid call; (3) the record says the filing was written by the operator's Orion, not by Elrond's filing step; (4) filings from anyone else are written by Elrond's filing step as today.

The owner operator of an arc or a pursuit is the one operator who created it

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042), after ruling communications #649 that an arc's or pursuit's owner operator is told of every change: 'Arcs record only an owning agent; no field holds their owner operator. Your September ruling names Sahil as the Blueprint arc's operator owner, but only in words. Who should be each arc's owner operator, so changes have someone to tell?' (offered: an owner field defaulting to him; a field defaulting to the creator; tell the creator only). HIS ANSWER, in his own words: "Owner of the arc and the pursuit is the one operator who created the arc or the pursuit". ASKED NEXT: 10 of the 20 live pursuits, Blueprint included, were created from Orion's own agent login, so no operator is recorded as their creator, and all 10 record him as the person who requested them -- who owns these? HE CHOSE 'The requester': where an Orion session created it, the operator recorded as requesting it at creation is the creator; future arcs and pursuits record the operator directly. READ AS (Orion, not his words): the owner operator of every arc and pursuit is its creator operator (initiatives #284: read from the change journal's creation event mapped through root's record), or, for one created from Orion's agent login, the operator its record names as requesting it at creation -- derived, never typed and never stored a second time; one created by an agent's own login with no requester reads 'no owner operator' and is never given a guessed one. The pursuit's settable operator owner (initiatives #124, set through the pursuits door) is amended in place to follow this: the human who answers for the outcome is the creator. Arcs and pursuits created from now on name their creating operator directly, never only Orion's login.

Albus focuses on architecture compliance: his lane follows Elrond's common prioritisation, and he bumps architecture compliance work above every other initiative, arc and pursuit

Ruled 2026-11-06 by ninad -- superseded 2026-11-08

Ninad, 2026-11-06 (CLI-043), unprompted, after Orion reported Albus's bypass lane pursuit (CAROL-PUR-0002) at 18 of 18 steps closed and Elrond's common prioritisation still planned. HIS WORDS: "Albus is an architect, and he should focus on architecture compliance arcs, pursuits, and initiatives.. he should leverage the common prioritization engine of elrond to understand the priorities, but in addition to that, he can and should be able to overide it by bumping up the priority associated with architectural complainces and check above other initaitves, arcs and pursuits as an add-on preference.." READ AS (Orion, not his words): (1) This rules the ORDER in which Albus's lane picks, not what it may pick: it still takes as much of the whole pending backlog as its budget permits (#627) and still keeps it relevant (#229). (2) The base order is Elrond's one common prioritisation; until that is built the base stays Albus's own ranking of everything waiting on his lane (#227), as #627 already says. (3) On top of the base order Albus applies ONE add-on preference: architecture compliance work is bumped above every other initiative, arc and pursuit. It is architecture compliance work when it is a member of an arc whose subject is architecture compliance (today CAROL-ARC-0026, Themis's audit service, tag architecture-compliance), a step of a pursuit whose outcome is architecture compliance, or an architecture compliance check. (4) Each bump is Albus's own recorded act, naming the item and why it is architecture compliance; it changes his lane's pick order only, never the common prioritisation other agents read. (5) Who owns and pays for the work is unchanged (the service it improves, #491): Albus performs architecture compliance work on his lane, he does not take it over from Themis's arc.

Ended because: Ninad widened what Albus may bump above the common prioritisation (CLI-065): arcs and pursuits of his interest, especially policy, design and regulation violations, not only architecture compliance

A written filing's criteria check is kept but never and its written size stands as the attempt's effort.

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-040, recorded CLI-045). Told that a written filing still paid two steps on Elrond's filing budget and asked (1) to keep the filing door's criteria check but charge it to the operator's own lane when the operator's Orion wrote the filing, and (2) to let a written size stand instead of Elrond's Effort Estimator sizing it again, which changes his rule that he sizes all effort himself, HIS WORDS: "yes to both". READ AS (Orion's reading, not his words): (1) the filing door's criteria-quality check still applies to a filing the operator's Orion wrote, and its cost is the operator's lane's, never Elrond's filing budget -- met by the check's own free rule: the written filing states at least one success criterion as an outcome someone sees, which the check accepts without a model call, and one its free rule cannot accept is refused at writing and rewritten on the operator's plan. The door's model call is NOT moved onto the budget-exempt operator lane, because that lane admits only work proven to be the operator's own (its guard is not loosened for this); (2) the size the operator's Orion wrote is recorded as the attempt's effort with its own source, the filing does not ring Elrond's Effort Door, and his Effort Estimator never re-sizes it; Elrond still sizes every other filing and may size a later attempt of the same initiative. Amends initiatives #123 (who sizes) for written filings only; extends #650.

Every initiative, arc and pursuit is mapped to an execution operator; every initiative sits in an arc or a pursuit, agent filings in blanket agent or service ones

Ruled 2026-11-06 by ninad -- superseded 2026-11-07

ASKED (Orion, CLI-049, 2026-11-06), after being shown that 48 pending initiatives sit in no arc or pursuit, so #658 assigns them nobody: '(a) every initiative must sit in an arc or pursuit -- filing refuses one that doesn't, the 48 get placed, and each inherits that arc's or pursuit's assigned operator; or (b) an initiative outside arcs and pursuits gets its own assigned operator'. HIS WORDS: "option a", then: "every initiative, arc and pursuit is mapped to an operator.. as per the current ruling, the creator / owner of the initiative is the execution operator by default, until the estate operator chages the mapping.. initiatives filed by agents should sit in blanket agent / service specific arcs and pursuits by default, unless they are moved out by estate operators..". READ AS (Orion's reading, not his words): (1) every initiative sits in an arc or a pursuit; a filing that names neither and cannot be placed by default is refused, and the 48 waiting today are placed once; (2) every initiative, arc and pursuit carries an execution operator: its creator operator by default (#658 for arcs and pursuits), until an estate operator of its project changes the mapping; (3) an initiative an agent files has no creator operator, so it sits by default in a blanket arc or pursuit for that agent or its service and takes that arc's or pursuit's execution operator, until an estate operator moves it out; (4) an initiative's own mapping is the one asked first when it and its arc's or pursuit's differ.

Ended because: Ninad amended where an agent's filing sits (its paying service's agent arc, never a pursuit), who executes an initiative by default (its arc's or pursuit's operator) and that no initiative spans arcs (CLI-054, CLI-058)

Only the administrator and estate operators create and own arcs and pursuits

Ruled 2026-11-06 by ninad -- superseded 2026-11-11

HIS WORDS (CLI-049, 2026-11-06): "only admin and estate operators can create / own arcs and pursuits..". READ AS (Orion's reading, not his words): the arc and pursuit creation doors admit only the administrator and estate operators, and the owner operator (#651, the creator) is therefore always one of them; a service operator works on arcs and pursuits an estate operator assigns (#381, #658) and never creates or owns one; agents never create or own one (blanket arcs and pursuits for agent filings are created by an estate operator).

Ended because: Ninad allowed one exception: Leo creates a client project's pursuit on the user's confirmation

Every initiative sits in one arc or a pursuit and is executed by its arc's or pursuit's operator; an agent's filing sits in its paying service's agent arc, never a pursuit

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-054, CLI-058), building, amending #661 in four answers. (1) Asked who executes an initiative by default when it sits in an arc assigned to someone else, he chose "Arc's operator": an initiative takes the executing operator of the arc or pursuit it sits in (agent-resources #658 stands); an estate operator of its project may assign the initiative itself to anyone, and that assignment is then its executing operator. (2) HIS WORDS: "There should be an arc for agent initiatives for each service, where agent filed initiatives live. Create them upfront, now. No pursuit should exist for agent filed initiatives. Pursuits are only created by operators. Existing initiatives from arcs can be picked and modified for the pursuits, and new initiatives can be created. No initiative should span across arcs" -- replacing his answer earlier the same day (one holding pursuit per agent, never recorded) and #661's 'blanket agent / service specific arcs and pursuits'. (3) Asked who may create pursuits, HIS WORDS: "yes, I meant, admins and estate operators only" (#662 stands). (4) Told one open initiative sits in arcs of two services, HIS WORDS: "this means it will have to be split" (#384). READ AS (Orion's reading, not his words): every initiative sits in at most one arc and may also be a step of pursuits; each service has one agent arc, created by an estate operator, owned by the service's owning agent (P.01.03.05.22) and executed by default by the operator who created it; an agent's filing goes to the agent arc of the service that pays for it (the one service it names, #383) and waits there for that service owner's approval; an operator's filing names its arc or pursuit and is refused without one; an initiative in arcs of two services is split, one per service, and one in two arcs of the same service keeps the arc it entered first. #661's other parts stand: every initiative, arc and pursuit has an executing operator, and an estate operator of its project changes it.

Whoever files an initiative owns it; the service it names pays, and that service's owner approves it only on a valid ground

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058). HIS WORDS: "initiatives should belong to service that pays for that initiative.. and the service owner should only accept initiatives that will improve the service.. agents can file initiatives for any service, as long as the service owner accepts them"; then, told that #487 and #491 make the paying service's owner the initiative's owner: "whoever filed the initiative owns it, the service owner only approves the initaitve because they pay for it, and ensures that the initiative is filed to the right service and benifits the service, and that the service can afford the expense of the initiative and is aligned to the service roadmap. The initiative also sits in the service arc, but the agent continues to own the initiative and do the UAT"; and on Orion's advice to keep this rather than add a requester field: "sounds good. Just that the filing agent should have a good reason for filing the initiative, else its a valid ground for the service owner to reject it as additional and un-necessary expense.. A valid ground can be architectural compliance, policy violation rectification, employee request, operator request, user request if the service allows it, agent well-being, service improvement for enabling agent to do their tasks better etc..". READ AS (Orion's reading, not his words): (1) an initiative's owner is the agent that filed it -- an operator's filing is Orion's (agt_023) -- reversing #487 and #491; (2) the one service it names (#383) pays for it, and that service's owning agent approves it only when it is filed to the right service, benefits it, is affordable and fits the service's roadmap (#558 stands); (3) every filing states its ground from a recorded list kept as reference data -- architectural compliance, policy violation rectification, employee request, operator request, user request where the service allows it, agent well-being, service improvement that helps agents do their tasks -- and a filing without a valid ground is a valid reason to reject it; (4) it sits in its paying service's arc while its filer keeps ownership and does its UAT.

Elrond's checks throw away only duplicates; work no longer needed is parked and told to its filer; work that breaks or lacks a policy is parked for a ruling

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058). Shown that Elrond's daily relevance check (#261) and the status router discard initiatives others filed (22 by Elrond and 7 by the router in the two weeks to 7 November), HIS WORDS: "Elrond should only throw away work if its duplicate, but otherwise, say if its contradicting the policies, or if the policies are not clear, then the initiatives should be parked, and the policy resolution workflow should trigger"; to Orion's proposal that work no longer needed is parked and its filer told why, the filer then fixing, closing or discarding it: "I agree with your proposal"; and "check if this exists, else file initiative for this". READ AS (Orion's reading, not his words): Elrond's check keeps #261's daily pass over every lane but discards only a duplicate; an initiative no longer needed is parked with its reason and told to its filer, who decides; one that conflicts with a policy or ruling, or that no clear policy covers, is parked and handed to the ruling process of initiatives #139 (the operator rules, the work is revived), which until now only Albus's lane enters.

Only an initiative's filer closes or discards it; Elrond signs off operators' filings and a filing agent tests its own; a rejection gives its reason and returns it to its filer

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058). HIS WORDS: "when service owner rejects it, it goes back to agent's backlog for the agent to rectify it.. service owner should give the reason for rejection. Only requesting agent can close or discard their initiatives"; told that Elrond signs off almost every initiative: "Elrond continues to sign off initiatives from operators, but initiatives created by agents should be UATed by agents before they are closed, this should fit the existing ruling, pls check" (it fits #19, the wisher does its UAT, and #145, the acceptor is the initiative's owner, now its filer by #666); told of #424: "its upto agent to decide if they want to convert the initiative to an escalation or wish, or modify the initiative based on the feedback of the service owner". READ AS (Orion's reading, not his words): (1) only an initiative's filer closes it without doing it or discards it -- no other agent, router or operator session does; (2) an operator's filing is signed off by Elrond; an agent's filing is accepted by that agent's own UAT before it closes; (3) a service owner's rejection states its reason and returns the initiative to its filer, who either changes it and asks again or turns it into an agent escalation or wish (#424 as amended).

Albus keeps his queue relevant and pushes work back to its filer; he discards only his own

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058), amending #229 by #667 and #668: Albus keeps re-checking that each initiative waiting on his lane is still relevant, and when one is not he pushes it back to its filer with what is missing -- parked, its reason recorded where the filer reads it -- and never discards an initiative he did not file; only its filer closes or discards it, and a duplicate is Elrond's to discard.

One skill decides which arc and pursuit an initiative belongs to, and every skill is re-checked whenever a policy it rests on changes

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058), HIS WORDS: "there should be a re-usable skill that enables deciding what pursuits and arcs the initiatives belog to, which should account for all the existing policies and should check for policy updates to ensure that it remains uptodate. Infact, this applies for all the skills". READ AS (Orion's reading, not his words): (1) one skill in Sage's library decides which arc and which pursuits an initiative belongs to, reading the policies and rulings in force when it runs; (2) every skill names the policies and rulings it rests on, and when any of them changes the skill is flagged to its keeper and re-checked until it agrees again -- a skill never teaches what the law no longer says.

Each service's owner manages its own arcs, keeps them compliant and relevant, and files or escalates what it finds

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-058), HIS WORDS: "ask individual service owners to manage and organize their arcs, ensure that they are compliant and relevant and raise escalations or file initiatives as appropriate.. fold this into the same arc / pursuit". READ AS (Orion's reading, not his words): the agent that owns a service (P.01.03.05.22) looks after every arc of that service -- its phases and members in order, each initiative in one arc only (#665), no two arcs on the same work (#474), every member still relevant and every arc compliant with the policies and rulings in force; what it can fix by naming the fix it files as an initiative with its ground (#666, P.01.03.01.12), what it cannot it escalates. Folded into the ownership arc and the pursuit Service-owned initiatives.

Every check of a written filing is kept, the operator's lane pays for them, and one filing's checks name one payer

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-063). Asked whether filings the operator's Orion writes should also skip the filing door's duplicate check and topic tagging (about 40 cents a filing on Elrond's filing budget, at the risk of repeat work and untagged initiatives), with Orion recommending keeping both, and told that the checks bill two different payers, HIS WORDS: "dont skip anything, just pick those costs up, and fix any defects". READ AS (Orion's reading, not his words): (1) no check of the filing door is skipped for a filing the operator's Orion wrote -- the criteria check, the duplicate check, the topic tagging, the service classification and the validity check all still run; (2) the model calls those checks make for a written filing that carries the operator's token are paid by the operator's lane (the Orion bypass track), never Elrond's filing budget, and the operator lane's guard admits them because the token proves the work is the operator's own; this amends #657 point (1), whose reading kept the door's model call off the operator's lane; (3) every check of one filing names ONE payer on its work ticket -- the operator's lane for such a written filing, Elrond's build governance track for every other filing -- and no check names an agent as its payer: the lane a call runs in is named apart from who pays (cookbook 1672).

Albus may override Elrond's common prioritisation by bumping arcs and pursuits of his interest, especially policy, design and regulation violations, on top of architecture compliance

Ruled 2026-11-08 by ninad -- superseded 2026-11-08

Ninad, 2026-11-08 (CLI-065), unprompted, after Orion explained that the Wish Keeper moves onto Elrond's one prioritisation. HIS WORDS: "note that albus, being the architect, can over-ride elrond's prioritization and bump up arcs and pursuits of his interest, especially related to policy violations, design violations, regulation violations etc" and then "the prioriisation over-ride engine of albus might be built by another session under a different pursuit, but it was just fyi.." READ AS (Orion, not his words): (1) This amends #655 by widening WHAT Albus may bump, not how: beyond architecture compliance, Albus may bump any arc or pursuit of his interest as the architect, especially those answering policy violations, design violations and regulation violations; architecture compliance (CAROL-ARC-0026) stays preferred. (2) The base order is still Elrond's one common prioritisation; Albus's preference sits on top of it as the add-on #655 ruled and overrides it where they differ. (3) Each bump stays Albus's own recorded act, naming the arc or pursuit and why (the violation it answers, or his interest as architect), read from records and never from the words of a title. (4) Its reach is as #655 (4) read it: the order Albus's lane picks in; it never rewrites the ranking another owner reads, nor who owns or pays for the work (#491, #558). (5) Who builds the widening is not ruled here: Ninad said it may be built by another session under a different pursuit. builds the common prioritisation only and keeps Albus's add-on layer on top of it.

Ended because: Ninad confirmed the reach (Albus's own execution order) and why: service owners bear the budgets, Albus executes, pipeline activity stays with Elrond's initiatives service

Albus may override Elrond's common prioritisation for his own execution by bumping arcs and pursuits of his interest, especially policy, design and regulation violations; owners bear budgets, pipeline activity stays with Elrond's initiatives service

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-065), unprompted, after Orion explained that the Wish Keeper moves onto Elrond's one prioritisation. HIS WORDS: "note that albus, being the architect, can over-ride elrond's prioritization and bump up arcs and pursuits of his interest, especially related to policy violations, design violations, regulation violations etc" and then "the prioriisation over-ride engine of albus might be built by another session under a different pursuit, but it was just fyi.." READ AS (Orion, not his words): (1) This amends #655 by widening WHAT Albus may bump, not how: beyond architecture compliance, Albus may bump any arc or pursuit of his interest as the architect, especially those answering policy violations, design violations and regulation violations; architecture compliance (CAROL-ARC-0026) stays preferred. (2) The base order is still Elrond's one common prioritisation; Albus's preference sits on top of it as the add-on #655 ruled and overrides it where they differ. (3) Each bump stays Albus's own recorded act, naming the arc or pursuit and why (the violation it answers, or his interest as architect), read from records and never from the words of a title. (4) Its reach is as #655 (4) read it: the order Albus's lane picks in; it never rewrites the ranking another owner reads, nor who owns or pays for the work (#491, #558). (5) Who builds the widening is not ruled here: Ninad said it may be built by another session under a different pursuit. builds the common prioritisation only and keeps Albus's add-on layer on top of it. CONFIRMED, Ninad 2026-11-08 (CLI-067), after Orion said he kept the raise to the order of Albus's own work and asked whether Ninad meant every agent's list. HIS WORDS: "thats correct earlier albus was being loaded because he was bearing all the budget, now individual service owners bear the budgets, while albus focusses on execution, any pipeline activity will be still attributed to elrond's initiative service" READ AS (Orion, not his words): (6) Point (4) is now his ruling, not Orion's reading: Albus's bump orders his own execution; each service owner still orders, approves and pays for the work that improves its service (#558, #491, #557), since owners now bear the budgets Albus once carried alone. (7) Pipeline activity, Albus's execution lane included, stays attributed to Elrond's initiatives service: his lane bills its own track of that service and the pipeline's own droids bill Build Governance, as the records already say; nothing here moves a cost.

Archon's image-lane work moves four picture-making droids now; motion video is separate work

Ruled 2026-11-08 by ninad -- in force

HIS CHOICE (CLI-068, 2026-11-08), asked: "Archon's wish moves his five picture-making helpers onto the OpenAI image service. One of them makes motion video, which that service can't do. Should video stay out of this work, or be made by stitching generated stills into an MP4?" -- he chose "Video separately" (four helpers move now; the video helper keeps its current video service and gets its own piece of work later) over composing an MP4 from generated stills. READ AS (Orion's reading, not his words): 's criterion is four droids (ar-life-01, ar-life-02, ar-img-01, ar-port-01) on the OpenAI image lane; ar-vid-01 keeps its present video service and is not replaced by a locally composed MP4. Moving motion video is its own work, filed when Archon asks for it.

Albus's lane picks in three layers: the operator's priority channel first, then the arcs and pursuits he chooses himself, then Elrond's prioritisation

Ruled 2026-11-08 by ninad -- in force

HIS WORDS (CLI-077, 2026-11-09), said after hearing that Orion's operator window has had no Claude sign-in since 13 Oct and that its move to Codex is planned but unbuilt: "maybe albus should take that next.. 1. albus uses elrond's prioritization.. 2. plus he can decide on his own to pick certain arcs and pursuits that are related to topics close to his heart (overrides 1) 3. also create a channel of communication called "operator's priority" where albus puts those on his top priority (overrides 2 and 1)". READ AS (Orion's reading, not his words): Albus's lane orders its waiting work in three layers, each above the one before. (1) BASE: Elrond's one prioritisation (#558); until builds it, the base stays oldest first (#627). (2) ALBUS'S OWN CHOICE, above the base: the arcs and pursuits he picks himself because their topic is close to his heart; architecture compliance (#655) stays one of them, and he adds or removes others on his own, each with his reason recorded on the record that holds his choice. (3) OPERATOR'S PRIORITY, above both: a new communication channel of that name, in which an operator places work for Albus's lane; his lane takes those first, in the order the operator placed them. An operator places only work assigned to them to execute (P.03.04.02.09; an estate operator may first assign it to themselves). The channel is catalogued like every communication the estate makes (CAROL-PUR-0021), and every pick a layer moves ahead is recorded on that initiative. The budget and law gates still hold: a placed item never lets the lane start work it cannot finish (#226) or work the law blocks. is the first item he wants placed.

Sizing never waits on Elrond's budget: the operator's Orion sizes what is needed with its own tokens

Ruled 2026-11-08 by ninad -- in force

Then asked "How should sizing stop depending on his budget?" with the options 'Never capped', 'Paying service pays' and 'Own reserved budget', he answered in his own words: "use your own tokens". READ AS (Orion's reading, not his words): an operator's work never waits for sizing on Elrond's budget. When an attempt the operator needs sized (a step of a pursuit or a member of an arc whose hours they ask for) is unsized and Elrond's estimator cannot size it, the operator's Orion sizes it in its own session, on the operator's own Claude account, from the same record and calibration Elrond's estimator uses (shared.effort's prompt), and records it as Orion's estimate with its basis; it never writes over an estimate. Elrond's estimator keeps sizing everything else on its schedule and may re-size an estimate another source made when his budget allows; nothing waits for that.

Effort is the estimated sum of tokens to complete a unit of work

Ruled 2026-11-09 by ninad -- in force

EFFORT is the estimated sum of tokens needed to complete a unit of work: an activity, a task, an action, an attempt, an initiative, an arc or a pursuit. It is an ESTIMATE, never the measured tokens or the cost. An ATTEMPT's effort is the estimate for that one attempt. An INITIATIVE's effort is the total over every attempt of the family along the escalation ladder, derived by adding its attempts' efforts and never stored, so the two are always kept apart. Each attempt is sized by ELROND from its own record, never typed in by an operator. A pursuit's completion is read FROM EFFORT SPENT, exactly as an arc's and a phase's is (#205, #302): every worked attempt of its steps counts, closed or not, an open step's standing attempt counting its estimate less the effort Elrond sized as still remaining; a pursuit with any open step reads 99 at most; while any step is unsized, or a worked open step's remaining effort is unsized, the figure is unmeasured and names that step (amended CLI-079, 2026-11-09: 'if some work has happened, the efforts cannot be 0%, can you re-look at how the efforts are counted for status and update the skill appropriately?'). Ninad, CLI-545: 'effort --> estimated sum of tokens to complete a certain activity, task, action, initiative, arc, pursuit etc' and 'there should be a clear distinction of efforts for distinct initiative attempt vs effort for an initiative in total (including all attempts via escalation ladder)'.

Arc and pursuit status figures are pre-calculated at least hourly and read from there

Ruled 2026-11-09 by ninad -- in force

THE STATUS FIGURES OF EVERY ARC AND PURSUIT ARE PRE-CALCULATED, AT LEAST HOURLY, AND A STATUS ANSWER READS THE PRE-CALCULATION (Ninad, CLI-079, 2026-11-09, shown CAROL-PUR-0029 at 0% while its first step was built and live: 'if some work has happened, the efforts cannot be 0%, can you re-look at how the efforts are counted for status and update the skill appropriately? Are you canculating these numbers everytime, or simply fetching them from the pipeline app? ideally these numbers should be pre-calculated, but you should have access to the same algorythm to calculate them if the automated process has not run.. the automated process should run for all arcs and pursuits atleast at an hrs interval to ensure that the updates are recent and uptodate..'). Elrond's Plan Reading Recorder works out every live arc's and pursuit's usual figures -- initiatives, completed, percent by effort, estimated hours total and left -- with the ONE composer the Pipeline app and the operator's window use, at least once an hour (today every 15 minutes), and keeps the latest result with when it was worked out. A status answer, in the Pipeline app or an Orion session, reads that latest result while it is under an hour old and says when it was worked out; when it is older, or the recorder has not run, the same composer works the figures out on the spot and the answer says so -- never a second calculation, never a stale figure presented as current. A recorder that has not finished a run for an hour is a finding by name (amends the reading of #613 that a reading is never read as the current state).

Albus executes autonomously every initiative that needs no operator governance or root or core step

Ruled 2026-11-09 by ninad -- in force

HIS WORDS (CLI-086, 2026-11-09): "the requirement is to get albus up and running to execute autonomously for tasks that dont involve operator governance and other root protected activities.. right now that process is not smoot because operator sign-ins are required for many steps. Radagast needs to whitelist most commonly identified activities.. he wont be able to whitelist all, but as long as the most common activities are whitelisted and executed by radagast, albus can autonomously execute tasks on operators behalf.. offcourse, where root/core intervention is needed, albus will have to hand it over to orion to complete, or when policy conflict or ambuigity happens, orion has to step in, but all other initiatives are expected to be handled by albus autonomously, to whatever extent possible". So: Radagast's operator-level list covers the most common activities Albus's lane needs an operator sign-in for today, and Radagast performs them for Albus; a root or core step, a policy conflict or an ambiguity goes to Orion; every other initiative is Albus's to execute.

An initiative belongs to at most one arc and at most one pursuit; one shared today stays in the pursuit it joined last

Ruled 2026-11-09 by ninad -- in force

AN INITIATIVE BELONGS TO AT MOST ONE ARC AND AT MOST ONE PURSUIT; ONE SHARED TODAY STAYS IN THE PURSUIT IT JOINED LAST (Ninad, CLI-088/CLI-094, 2026-11-09). THE CONTEXT: told that the Blueprint description rewrite sat in Leo's Blueprint arc and the new Carolopedia pursuit CAROL-PUR-0031. HIS WORDS: "if its part of blueprint pursuit, take it off from this pursuit, an initiative can be part of an arc and a pursuit, but it cannot belong to 2 arcs or 2 pursuits". THE QUESTION ANSWERED: told that this reverses his 27 September ruling (initiatives #165: an initiative in two pursuits is marked as overlapping in both) and that 84 initiatives sit in two or three pursuits today, asked whether to record the new rule replacing it: "replace the ruling.. also more the initiative from this pursuit to the blueprint pursuit.. sahil is working on the blueprint pursuit"; asked how to choose which pursuit keeps each shared initiative: "last pursuit it joined". READ AS (Orion's reading, not his words): an initiative is a member of at most one pursuit, as it is of at most one arc (initiatives #665); each initiative now in several pursuits stays in the pursuit whose sequence it joined most recently and leaves the others; the pursuits door refuses a second pursuit; #165's marking of the SAME initiative as overlapping in two pursuits is retired in place, while its marking of a DEPENDENCY that crosses pursuits (one pursuit's step waiting on another's) is not contradicted and stands.

Only the administrator and estate operators create and own arcs and pursuits, except a client project's pursuit, which Leo creates and its estate operator owns and executes

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), answering Sahil's operator communication #333, point 1 (initiatives #665 and #662 say only operators create pursuits; blueprint #642 says Leo agrees small pursuits with the user: (a) Leo records the agreed pursuit and the user's confirmation and an estate operator creates it, or (b) a client project's pursuit is a new kind Leo creates on the user's confirmation -- then who owns and who executes it?). He chose 'Leo creates client pursuits'; asked who owns it, 'Project's estate operator' (you, until the project has one; Leo is recorded as the one who created it); asked who executes it, 'Project's estate operator' (by default, able to hand it to any estate or service operator, as with every other pursuit). #662 IN HIS WORDS (CLI-049, 2026-11-06) still stands for everything else: "only admin and estate operators can create / own arcs and pursuits..". READ AS (Orion's reading, not his words): the arc and pursuit creation doors admit only the administrator and estate operators, with ONE exception: a client project's pursuit, which Leo creates once the user confirms it explicitly in his chat (#642); that pursuit's owner operator is the project's estate operator (the administrator until the project has one) -- not its creator (#651 reads the creating operator; here the creator is an agent); it is executed by that estate operator by default, who may assign it to any estate or service operator (#663). Otherwise as before: a service operator works on arcs and pursuits an estate operator assigns (#381, #658) and never creates or owns one; no other agent creates or owns one (blanket arcs and pursuits for agent filings are created by an estate operator).

An initiative is at most 1 estimated hour, a pursuit at most 5 and an arc at most 10, judged on the whole size; open work over its limit is split now

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), seeing the Investment chain pursuit at 56 initiatives and about 119 estimated hours: "this is too big for a pursuit, what should be, according to you, the guidelines for, 1. how big should 1 initiative be, before it can be split into multiple initiatives 2. how big should 1 arc or pursuit be, before it can eb split into multiple arcs / pursuits the skill should use this guideline". Shown Orion's proposal (initiative 0.5-3.3 h, pursuit 40 h, arc phase 20 h, arc 120 h; only the initiative figure from data), HIS WORDS: "Initiative --> min 1 min, max 1 hr Pursuit --> min 1 min, max 5 hrs Arc --> min 1 min, max 10 hrs". Asked whether an item is judged on its whole estimated size or on what is left, he chose 'Whole size'; asked what happens to items already over (about a third of today's initiatives over 1 h, 16 of 25 measured pursuits over 5 h), he chose 'Split open work now' (every open item over its limit is split now; finished parts stay where they are and the open work moves into new, right-sized items; closed items are left alone).

Ended because: Ninad agreed: oversized open initiatives are re-sized (over 10 h) or split when next picked up, not all now

Epics are never-ending groups of arcs and pursuits of one kind, joined by their common tags, so a topic stays together after its arcs and pursuits are split

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), right after ruling the size limits (initiatives #807: an initiative at most 1 estimated hour, a pursuit 5, an arc 10, open work over its limit split now), HIS WORDS: "common topics, even after splitting, will have common tags, you can create \"epics\", which are never ending groups of certain types of arcs and pursuits". READ AS (Orion's reading, not his words): (1) an EPIC is a record of its own: a named, never-ending group of arcs and pursuits of one kind or topic (for example the investment chain, operator compliance, Blueprint); it never closes and has no size limit, while each arc and pursuit inside it keeps #807's limits; (2) an arc or pursuit belongs to an epic by the common tag it carries: when one is split under #807, every part keeps the topic's tag and stays in the same epic, so nothing of the topic is lost by splitting; (3) an epic's status is read from its arcs and pursuits (their figures summed in the usual format), never kept by hand; (4) who creates and owns an epic is not ruled: until it is, as for arcs and pursuits, the administrator or an estate operator (#662 as restated by #798); (5) the skills that place, build and report (decide-arc-and-pursuit, build-arcs-and-pursuits, report-arcs-and-pursuits-progress) name the epic when they place, split or report an arc or pursuit.

Ended because: Ninad clarified 'never ending' (more pursuits can be added) and asked for the dictionary entry

Work epics are never-ending groups of arcs and pursuits of one kind, joined by their common tags, that more can always join, so a topic stays together after its arcs and pursuits are split

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), right after ruling the size limits (initiatives #807: an initiative at most 1 estimated hour, a pursuit 5, an arc 10, open work over its limit split now), HIS WORDS: "common topics, even after splitting, will have common tags, you can create "epics", which are never ending groups of certain types of arcs and pursuits", then, restated: "common topics, even after splitting, will have common tags, you can create "epics", which are never ending groups of certain types of arcs and pursuits (never ending meaning -- more pursuits can be added to them) update dictionary". READ AS (Orion's reading, not his words): (1) a WORK EPIC is a record of its own: a named group of arcs and pursuits of one kind or topic (for example the investment chain, operator compliance, Blueprint); NEVER ENDING means it stays open for more -- arcs and pursuits keep being added to it for as long as the work goes on -- so it never closes and has no size limit, while each arc and pursuit inside it keeps #807's limits; (2) an arc or pursuit belongs to a work epic by the common tag it carries: when one is split under #807, every part keeps the topic's tag and stays in the same work epic, so nothing of the topic is lost; (3) a work epic's status is read from its arcs and pursuits (their figures summed in the usual format), never kept by hand; (4) who creates and owns a work epic is not ruled: until it is, as for arcs and pursuits, the administrator or an estate operator (#662 as restated by #798); (5) the dictionary defines it as 'Work epic' -- an everyday word takes an Oratorium meaning only with a prefix (governance #404) -- and the skills that place, build and report (decide-arc-and-pursuit, build-arcs-and-pursuits, report-arcs-and-pursuits-progress) name the work epic when they place, split or report an arc or pursuit.

Ended because: Ninad's answer of 2026-11-12 (CLI-115): pursuit size is a guideline, arcs stay hard, epics group pursuits only

An initiative is at most 1 estimated hour, a pursuit at most 5 and an arc at most 10, judged on the whole size; open pursuits and arcs over their limit are split now, open initiatives when next picked up

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), seeing the Investment chain pursuit at 56 initiatives and about 119 estimated hours: "this is too big for a pursuit, what should be, according to you, the guidelines for, 1. how big should 1 initiative be, before it can be split into multiple initiatives 2. how big should 1 arc or pursuit be, before it can eb split into multiple arcs / pursuits the skill should use this guideline". Shown Orion's proposal (initiative 0.5-3.3 h, pursuit 40 h, arc phase 20 h, arc 120 h; only the initiative figure from data), HIS WORDS: "Initiative --> min 1 min, max 1 hr Pursuit --> min 1 min, max 5 hrs Arc --> min 1 min, max 10 hrs". Asked whether an item is judged on its whole estimated size or on what is left, he chose 'Whole size'; asked what happens to items already over (about a third of today's initiatives over 1 h, 16 of 25 measured pursuits over 5 h), he chose 'Split open work now' (every open item over its limit is split now; finished parts stay where they are and the open work moves into new, right-sized items; closed items are left alone). Then (same session), told what is over today -- 16 of 26 pursuits, 31 of 97 arcs and 184 open initiatives (860 estimated hours, about 956 one-hour pieces, filing them days of Orion's time; ten of them, 345 hours, look like wrong estimates) -- and offered Orion's recommendation for initiatives only (re-size those ten now, split each oversized initiative when it is next picked up, the filing step refusing any new initiative over 1 hour), HIS WORDS: "I agree with your recommendations".

Ended because: Ninad chose: the eight genuine large initiatives are split when picked up, not re-sized

A pursuit whose open work moved to right-sized pursuits closes as 'split', keeping its finished steps

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), splitting the Investment chain (56 initiatives, 119 estimated hours) under initiatives #810's 5-hour limit into about 29 pursuits in the investment-chain work epic (#809). Asked how the old pursuit ends, its 16 finished steps staying in it (only 'achieved' -- untrue, its criteria are not met -- or 'abandoned' -- untrue, nothing was dropped -- could close it), offered a new status 'split', closing as abandoned with the reason beside it, or moving the finished steps too, he chose "New status 'split'" (add a closing status 'split' for a pursuit whose open work moved to right-sized pursuits; the old pursuit keeps its finished steps as the record and closes as split; a small build, filed under the limits). READ AS (Orion's reading, not his words): (1) 'split' joins the pursuit statuses as a third closing status beside achieved and abandoned, reached from planned, active or paused; (2) a pursuit is set to split only when no open step remains in it and it names the pursuits its open work moved to; (3) a split pursuit keeps its finished steps and its criteria as the record, reads closed on the board, in the usual format and in Elrond's Plan Reading Recorder, and says where its work went; (4) until the status is built the old pursuit stays as it is, its open steps already moved.

Ended because: Ninad chose: arcs close as split too

An initiative is at most 1 estimated hour, a pursuit at most 5 and an arc at most 10, judged on the whole size; open pursuits and arcs over their limit are split now, open initiatives when next picked up

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-109), seeing the Investment chain pursuit at 56 initiatives and about 119 estimated hours: "this is too big for a pursuit, what should be, according to you, the guidelines for, 1. how big should 1 initiative be, before it can be split into multiple initiatives 2. how big should 1 arc or pursuit be, before it can eb split into multiple arcs / pursuits the skill should use this guideline". Shown Orion's proposal (initiative 0.5-3.3 h, pursuit 40 h, arc phase 20 h, arc 120 h; only the initiative figure from data), HIS WORDS: "Initiative --> min 1 min, max 1 hr Pursuit --> min 1 min, max 5 hrs Arc --> min 1 min, max 10 hrs". Asked whether an item is judged on its whole estimated size or on what is left, he chose 'Whole size'; asked what happens to items already over (about a third of today's initiatives over 1 h, 16 of 25 measured pursuits over 5 h), he chose 'Split open work now' (every open item over its limit is split now; finished parts stay where they are and the open work moves into new, right-sized items; closed items are left alone). Then (same session), told what is over today -- 16 of 26 pursuits, 31 of 97 arcs and 184 open initiatives (860 estimated hours, about 956 one-hour pieces, filing them days of Orion's time; ten of them, 345 hours, look like wrong estimates) -- and offered Orion's recommendation for initiatives only (re-size those ten now, split each oversized initiative when it is next picked up, the filing step refusing any new initiative over 1 hour), HIS WORDS: "I agree with your recommendations". Told (same session) that Orion's claim was wrong -- of the ten initiatives over 10 hours two had already been re-sized by Elrond and the other eight are genuinely big jobs (clearing 1,620 architecture violations, moving secrets into Azure Key Vault), so shrinking their numbers would hide real work -- and offered 'Split when picked up' (treat them like the other oversized initiatives; the re-sizing step withdrawn) or 'Split these eight now', he chose 'Split when picked up'.

Ended because: Ninad's answer of 2026-11-12 (CLI-115): pursuit size is a guideline, arcs stay hard, epics group pursuits only

A pursuit or an arc whose open work moved to right-sized ones closes as 'split', keeping its finished steps or members

Ruled 2026-11-12 by ninad -- in force

Ninad, 2026-11-12 (CLI-109), splitting the Investment chain (56 initiatives, 119 estimated hours) under initiatives #810's 5-hour limit into about 29 pursuits in the investment-chain work epic (#809). Asked how the old pursuit ends, its 16 finished steps staying in it (only 'achieved' -- untrue, its criteria are not met -- or 'abandoned' -- untrue, nothing was dropped -- could close it), offered a new status 'split', closing as abandoned with the reason beside it, or moving the finished steps too, he chose "New status 'split'" (add a closing status 'split' for a pursuit whose open work moved to right-sized pursuits; the old pursuit keeps its finished steps as the record and closes as split; a small build, filed under the limits). Then, told that the two original arcs (investment chain, company accounts) hold only finished members after their split and that an arc can close only as 'completed' (its goal met) or 'abandoned', and asked whether arcs get the 'split' closing status too, he chose 'Arcs close as split too' (the same build that adds 'split' for pursuits adds it for arcs: an arc whose open work moved to right-sized arcs keeps its finished members and closes as split). READ AS (Orion's reading, not his words): (1) 'split' joins the pursuit statuses as a third closing status beside achieved and abandoned, reached from planned, active or paused; (2) a pursuit is set to split only when no open step remains in it and it names the pursuits its open work moved to; (3) a split pursuit keeps its finished steps and its criteria as the record, reads closed on the board, in the usual format and in Elrond's Plan Reading Recorder, and says where its work went; (4) until the status is built the old pursuit stays as it is, its open steps already moved. (5) ARCS TOO: 'split' joins the arc statuses beside completed and abandoned; an arc is set to split only when no open member remains in it and it names the arcs its open members moved to; a split arc keeps its finished members, reads closed and says where its work went.

Size limits: an arc's limit is hard, a pursuit's is a guideline; initiatives as #812

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-115), asking "pursuit by definition means that it should achieve an objective, wouldnt putting a cap become a problem for defining the pursuit? should the cap be a guideline rather than a hard rule?"; offered: the 5-hour pursuit and 10-hour arc limits as guidelines with a recorded reason (the 1-hour initiative limit staying hard), guidelines for all three, or keeping #812's hard caps with the larger objective given a home above pursuits (told that a work epic could not hold an objective while #809 rules epics never end); HIS WORDS: "Arcs are not objectives, they are topic driven, so hard limit should stay on arcs. Pursuits are objectives, so for them, the caps can be guidelines, only when the objectives cannot be broken down into smaller objectives.. almost always, it should be possible for splitting objectives into smaller objectives, while the Epic governing the larger objective.. epics are groupings of pursuits only, not arcs. Epics drive larger objectives.. epics need not be endless, but in reality, they will end up becomming endless for some blanket objectives". READ AS (Orion's reading, not his words): (1) an ARC keeps #812's HARD limit of 10 estimated hours, judged on its whole size, an open arc over it split now and the arc door refusing one over it -- an arc is a topic, not an objective; (2) a PURSUIT's 5 estimated hours is a GUIDELINE: a pursuit over it is split into pursuits of smaller objectives, which should almost always be possible; it stands over 5 hours ONLY when its objective cannot be broken into smaller objectives, with that reason recorded on the pursuit; the pursuit door refuses a pursuit over 5 hours that carries no such reason, never one that does; open pursuits over 5 hours are still split now (#812), each first judged on whether its objective can be broken down; (3) the larger objective that the split pursuits serve is governed by their WORK EPIC (the successor of #809); (4) the INITIATIVE limit (1 estimated hour, split when next picked up, the filing step refusing more) was not asked about and stands as #812 ruled; (5) everything else #812 ruled stands (whole size, minimum one minute, estimated hours as the usual format converts effort, closed items left alone, the skills file-carol-initiative, decide-arc-and-pursuit and build-arcs-and-pursuits carry the limits); (6) pursuits already split under #812 (the Investment chain's 29) stay as they are -- nothing is re-merged.

Ended because: Ninad's answer of 2026-11-12 (CLI-120): the Investment chain becomes six pursuits, one per objective, the size guideline disregarded

A work epic groups pursuits only and drives a larger objective; it need not be endless

Ruled 2026-11-12 by ninad -- in force

Ninad, 2026-11-12 (CLI-115), asking "pursuit by definition means that it should achieve an objective, wouldnt putting a cap become a problem for defining the pursuit? should the cap be a guideline rather than a hard rule?"; offered: the 5-hour pursuit and 10-hour arc limits as guidelines with a recorded reason (the 1-hour initiative limit staying hard), guidelines for all three, or keeping #812's hard caps with the larger objective given a home above pursuits (told that a work epic could not hold an objective while #809 rules epics never end); HIS WORDS: "Arcs are not objectives, they are topic driven, so hard limit should stay on arcs. Pursuits are objectives, so for them, the caps can be guidelines, only when the objectives cannot be broken down into smaller objectives.. almost always, it should be possible for splitting objectives into smaller objectives, while the Epic governing the larger objective.. epics are groupings of pursuits only, not arcs. Epics drive larger objectives.. epics need not be endless, but in reality, they will end up becomming endless for some blanket objectives". READ AS (Orion's reading, not his words): (1) a WORK EPIC groups PURSUITS only, never arcs -- #809's 'groups of certain types of arcs and pursuits' is retired in place for arcs; an arc's topic tag is the arc's own and makes it a member of no epic; (2) an epic DRIVES a LARGER OBJECTIVE: it holds that objective and its success criteria, each of its pursuits a smaller objective towards it (the Investment chain epic holds the criteria carried from CAROL-PUR-0007); (3) an epic NEED NOT BE ENDLESS: one whose objective is met can close; one serving a blanket objective may in practice never close and stays open for more pursuits; (4) #809's other readings stand: a pursuit belongs to its epic by the topic it carries, and an epic's status is read from its pursuits, never kept by hand; who creates and owns an epic is still not ruled; (5) the dictionary entry 'Work epic' is to say this, in words the administrator approves.

A pursuit whose steps are all closed is not complete while its outcome does not happen in practice

Ruled 2026-11-12 by ninad -- in force

Ninad, 2026-11-12 (CLI-113), shown that CAROL-PUR-0002 'Albus's bypass lane -- Get Albus's bypass lane up and running' reads 24 of 24 steps closed and 100% by effort while Albus completed no initiative in the seven days to 2026-11-12 (the shared completion count reads 0 every day; his run book: 11 attempts on 10 initiatives, 9 redirected to Orion as lane gaps, 1 as a law block, 1 reaped): "albus pursuit is compelted, but still he has not completed any initiative correctly, this means that the pursuit is not complete.. do session handover and let the next session look into it.." READ AS (Orion's reading, not his words): CAROL-PUR-0002 is not complete; the next session finds why the lane completes nothing and adds what the pursuit still needs.

Albus's lane is judged by how many of the initiatives it attempts it executes successfully: 80% is the target

Ruled 2026-11-12 by ninad -- superseded 2026-11-14

Ninad, 2026-11-12 (CLI-118/CLI-119), asked what the new success test for CAROL-PUR-0002 'Albus's bypass lane -- Get Albus's bypass lane up and running' should require after ruling #816 (Albus completed no initiative in the seven days to 2026-11-12): "being able to execute successfully 80% of the attempted initiatives is a good target for albus". READ AS (Orion's reading, not his words): the pursuit's new criterion is that, of the initiatives Albus's lane attempts, at least 80% are executed successfully.

Ended because: Ninad lowered the target to 50% for now and made it one administrator-set value, never an operator's own lane's

Every initiative Albus's lane redirects records why, and stays counted in his 80% quality measure

Ruled 2026-11-12 by ninad -- superseded 2026-11-12

Ninad, 2026-11-12 (CLI-119), told that screens Albus's picks for work his lane has no door for and hands it to Orion at no cost: "good idea, it should also record the reson why the initiatives was redirected.. Albus already redirects initiatives if the policies are not clear.. the redirected iniitiatives still stay associated with albus to measure his quality metric of 80%, is prometheus responsible for this?" READ AS (Orion's reading, not his words): every initiative Albus's lane redirects -- by the new screen, as a lane gap, or as a law block when the policies are not clear -- records the reason it was redirected, and remains one of Albus's attempted initiatives when his 80% success measure (#817) is read.

Ended because: Ninad confirmed the reading of #818 in his own words (CLI-119)

Every initiative Albus's lane redirects records why, and counts as attempted but not executed successfully in his 80% measure

Ruled 2026-11-12 by ninad -- in force

Ninad, 2026-11-12 (CLI-119), told that screens Albus's picks for work his lane has no door for and hands it to Orion at no cost: "good idea, it should also record the reson why the initiatives was redirected.. Albus already redirects initiatives if the policies are not clear.. the redirected iniitiatives still stay associated with albus to measure his quality metric of 80%, is prometheus responsible for this?" Asked to confirm Orion's reading that a redirected initiative counts as attempted but not as executed successfully, he answered: "a redirected initiative counts as attempted but not as executed successfully - correct". So every initiative Albus's lane redirects -- by the screen, as a lane gap, or as a law block when the policies are not clear -- records the reason it was redirected, and counts as attempted and not executed successfully in his 80% measure (#817), which Prometheus keeps as a quality metric.

Size limits as #814; the Investment chain is six pursuits, one per objective, over the guideline by decision

Ruled 2026-11-12 by ninad -- superseded 2026-11-13

Ninad, 2026-11-12 (CLI-120), shown the Investment chain work epic CAROL-EPIC-0001 (29 pursuits, 96.2 estimated hours, 10 criteria) broken into its six objectives -- the investment plan and Fundraising page; company accounts; the Expense Plan app; prospects in Sandbox; Fundraising as a template; investors, holdings and announcements -- and offered six epics or deferring three objectives; HIS WORDS: "looks like all the things are essential, convert it into 6 pursuits, 1 per objective.. disregard the size guideline". READ AS (Orion's reading, not his words): (1) nothing in the Investment chain is deferred; (2) its work becomes SIX pursuits, one per objective, in the work epic CAROL-EPIC-0001 -- the 29 pursuits split from CAROL-PUR-0007 are merged into them, each emptied pursuit closing as split into the one that carries its work, its criteria carried with its steps; (3) the 5-hour pursuit guideline is DISREGARDED for these six by his decision: the reason recorded on each pursuit over 5 hours is this ruling; (4) #814's reading 6 (the 29 stay as they are, nothing re-merged) is retired in place; (5) everything else #814 ruled stands -- the arc's 10 hours and the initiative's 1 hour stay HARD, the pursuit's 5 hours stays a guideline for every other pursuit.

Ended because: Ninad: 'reverse the split' on Artifact ownership, then chose to retire the 5-hour pursuit limit for every pursuit

Size limits as #814; the Investment chain is six pursuits, one per objective, over the guideline by decision

Ruled 2026-11-12 by ninad -- superseded 2026-11-16

THE ARTIFACT OWNERSHIP PURSUIT ALSO HOLDS CODE TO ITS OWNER (Ninad, 2026-11-12, CLI-120).

ASKED (CLI-120): Orion measured the build against the Agent-Centric Modular Architecture (design 146): in October 82% of new code went into the common library while the agents' own code shrank, and the architecture checker counts 36 places where shared code branches on a particular agent and 22 where one agent imports another's droids. Ninad asked: "is the build following the agent centric modular architecture principles?", then "there must already be a pursuit for this? there was a pursuit for only owning agents to be able to edit their artifacts". Orion found CAROL-PUR-0012 'Every artifact is held to its owner' covers stores, files and records, not where code lives, and that two planned steps sat in no pursuit: (shared code makes no agent's judgment and no agent imports another's droids) and (a shared module's owner reads from its place on record). Offered: add both to CAROL-PUR-0012, stretching its goal from who edits data to where code lives. ANSWERED: "ok, add the missing initiatives to it.."

READ AS (Orion's reading, not the administrator's words): and join CAROL-PUR-0012 as MVP steps (6890 first, so every judgment that moves has a named owner); the pursuit's outcome, MVP and criteria are stretched so code is held to its owner under design 146 -- judgments and jobs owned by their agent, shared code holding only data, systems, the org chart and the rulebook (ruling #515). The other planned steps of the same architecture audit that also sit in no pursuit were not named and are not moved by this ruling.

Ended because: Subject corrected: it named the retired size guidance (#814, 'over the guideline'), which #855 removed; the ruling's words are unchanged. CLI-185, Ninad: "this guidance was removed, remove any leftovers" -- arcs and pursuits have no size limit and no size guideline (Rulings Repo initiatives #855)

A pursuit has no size limit; arcs keep 10 hours hard and initiatives 1 hour hard; a split can be reversed while its parts hold only that work

Ruled 2026-11-13 by ninad -- superseded 2026-11-14

Ninad, 2026-11-13 (CLI-123/CLI-128), asked for the status of the Artifact ownership pursuit after CLI-123 had split it (CAROL-PUR-0012, 44 steps, about 73 estimated hours, into 13 pursuits CAROL-PUR-0065..0077 under #812), HIS WORDS: "reverse the split, and show me the status again". Told the pursuits record refuses to reopen a pursuit closed as split (#813, by design) and that reversing goes against the size ruling, and offered three ways -- except this one pursuit by ruling and build a reverse step; merge the 13 parts into one live pursuit with no build; or retire the 5-hour limit for every pursuit, build the reverse step and undo this split, the investment chain's split being a separate choice -- he chose 'Retire the 5-hour limit'. READ AS (Orion's reading, not his words): (1) a PURSUIT has NO size limit and no size guideline: its size is whatever its objective needs (#814's words, 'pursuits are objectives', stand); the 5-hour guideline of #814 and #820, the reason recorded on a pursuit over it, and the pursuit door's refusal of one over it with no reason are RETIRED in place; (2) the ARC's 10 estimated hours stays HARD and the INITIATIVE's 1 estimated hour stays HARD, exactly as #812 and #814 ruled (whole size, split now for arcs, split when next picked up for initiatives, the filing step and the arc door refusing more); (3) the Artifact ownership split of CLI-123 is REVERSED: the pursuits record gains a reverse step that puts a split pursuit's open work back from the pursuits it named -- allowed only while those pursuits hold nothing but that work and none of it has since finished there -- and closes them as split into it; CAROL-PUR-0012 stands again as one pursuit with its 44 steps and 4 criteria; (4) #820's six investment-chain pursuits and the 29 before them are NOT touched by this ruling: whether that split is reversed is a separate choice of Ninad's, not yet asked; (5) a work epic (#815) still groups pursuits and drives a larger objective; nothing here changes it; (6) the skills decide-arc-and-pursuit, build-arcs-and-pursuits and file-carol-initiative in Sage's library carry no pursuit size limit from now on and rest on this ruling; (7) (the Artifact ownership split) is restated under the governed goal change to deliver the reverse, and 7199-7202 (the other pursuit splits under #812) lose their premise.

Ended because: Ninad, CLI-137: the arc limit and the size guidelines removed; a pursuit over 100 hours becomes a work epic

Albus's lane success target is one administrator-set value, read live by every lane's measure: 50% for now [albus-success-target] percent=50

Ruled 2026-11-14 by ninad -- in force

Ninad, 2026-11-14 (CLI-135, continuing CLI-119), shown Albus's asks pursuit at 25%% by effort and the lane's success measure against #817's 80%%: "50%% is acceptable limit I think for now.. this should be a configurable limit, that can be set only by admin, not individual operators owning 'their own' albus bypass lanes". READ AS (Orion's reading, not his words): (1) the target is 50%% of attempted initiatives executed successfully, for now; (2) it is ONE estate-wide value, not a per-lane one: an operator whose own Albus bypass lane is measured by it never sets it; (3) it is set only at the administrator's rung -- by superseding THIS ruling with a new percent, which governance #347 already keeps from any lower rung -- and every reader of the target (Orion's board block, the Pipeline app, Prometheus's measure, pursuit criteria) reads the live ruling's [albus-success-target] percent=N, never a typed number. #819 (a redirect counts as attempted, not executed) stands unchanged.

Arcs and pursuits have no size limit and no size guideline; a pursuit over 100 estimated hours becomes a work epic of smaller pursuits; initiatives keep 1 hour

Ruled 2026-11-14 by ninad -- in force

ARCS AND PURSUITS HAVE NO SIZE LIMIT AND NO SIZE GUIDELINE; A PURSUIT OVER 100 ESTIMATED HOURS BECOMES A WORK EPIC OF SMALLER PURSUITS; INITIATIVES KEEP 1 HOUR (Ninad, 2026-11-14, CLI-137).

HIS WORDS: "you can remove the size constraint imposed on arcs and pursuits, also remove the guidelines, just that extremely long pursuits (over 100 hrs should be converted to epics, broken down into more resonable pursuits)".

READ AS (Orion's reading, not the administrator's words): (1) an ARC has NO size limit: #812's and #814's 10 estimated hours, the split-now of open arcs over it and the arc door's refusal are RETIRED in place; (2) a PURSUIT has no size limit and no size guideline, as #825 ruled; (3) every size GUIDELINE for arcs and pursuits is retired; (4) the ONE size rule left: a pursuit over 100 estimated hours (whole size, as the usual format converts effort) is converted into a WORK EPIC (#815) and broken into more reasonable pursuits, each holding its own objective, the epic governing the larger one; (5) the INITIATIVE's 1 estimated hour was not asked about and stands as #812 ruled; (6) #825's reverse step and its other readings stand; (7) (an arc refusing a member past 10 hours) loses its premise, and is restated to deliver reading 4; the skills decide-arc-and-pursuit and build-arcs-and-pursuits carry no arc limit and rest on this ruling.

A work epic's status reads in one usual format: a header naming the epic, the stage rows summed across its pursuits, then one row per pursuit

Ruled 2026-11-14 by ninad -- in force

A WORK EPIC'S STATUS READS IN ONE USUAL FORMAT: A HEADER NAMING THE EPIC, THE STAGE ROWS SUMMED ACROSS ITS PURSUITS, THEN ONE ROW PER PURSUIT. (1) One header line names the kind (work epic), its key, its name and how many of its success criteria are met. (2) Then the stage table, as a pursuit's: one row per stage (MVP, post-MVP, total) with initiatives, completed, % by effort, est. hrs total and est. hrs left, each the sum of its pursuits' own usual-format stage rows (percent weighted by hours, never typed). (3) Then one row per pursuit (key and short name) with the same columns, read from that pursuit's own Total row, and a total row. (4) A single pursuit or arc keeps its own usual format. (5) Until a pursuit names its epic, the Investment chain's pursuits are the open ones its chain was split into. SHOWN (CLI-137): the stage table for CAROL-EPIC-0001 (MVP 48, post-MVP 4, total 52 initiatives, 9 completed, 25% by effort, 97.8 h, 73.0 h left); asked whether it replaces the 12 Nov per-pursuit form (CLI-115) or both are shown. HIS WORDS: "make this as the standard format for epic status"; chose "Both, stage first".

Every initiative carries three needs, each assessed: whether it needs root, core, or an operator

Ruled 2026-11-14 by ninad -- in force

HIS WORDS (CLI-140, 2026-11-15), after a view of which pursuits need root that Orion had to work out by reading each initiative's wording, since nothing recorded it: "make this an essential attribute, on initiative level where each initiaitve is evaluated to figure out of 1. root is necessary 2. core is necessary 3. operator is necessary". SO: (1) Every initiative carries three needs -- ROOT, CORE and OPERATOR -- each yes or no with one plain sentence naming the step that needs it (or why none does). Until it is assessed it reads 'not assessed', never 'no'. CORE: a step only the core login can do with carol-authority -- root's register of operators, projects and machines, artifact classes, Radagast's list, hand-overs (security #175, #135). OPERATOR: apart from any root or core step, a step only an estate operator can do in person -- their password or code at a door, a hands-only step, a decision or approval only a person may give, an act outside the machine -- because neither Orion, Radagast nor an agent's own login can do it (security #197). The three are independent. (3) Elrond assesses them from the initiative's own record, as he sizes its effort: at filing through his Effort Door, and on his half-hour run for every live initiative not yet assessed or whose record has changed since; each assessment is kept, the latest counts. It is paid from his sizing task and keeps a reserve, so sizing never waits on it. (4) The Pipeline app shows each initiative's needs, and every count of pursuits or initiatives needing root, core or an operator is summed from them, never read by hand. (5) Albus's lane may read them to leave work it cannot do (initiatives #739). Orion decided (2)-(5) under security #197.

An arc or a pursuit needs root, core or an operator when any one of its initiatives does

Ruled 2026-11-14 by ninad -- in force

HIS WORDS (CLI-140, 2026-11-15): "if an arc or a pursuit have even 1 initiative that requires root / core / operator involvement, then the arc or pursuit inherits that status as well, fold this requirement in the pursuit". SO: (1) An arc, each of its phases, and a pursuit each carry the same three needs as #862 -- root, core, operator -- inherited: yes when even one of its initiatives needs it. (2) The initiatives counted are its OPEN ones (Orion's decision under security #197): #862's needs are what completing an initiative still takes, so a closed initiative's steps are done and pass nothing on. (3) While none of its open initiatives needs it but one is not yet assessed, whether it does is 'not yet known', never 'no'. (4) A family's needs are its newest attempt's; an arc's members are its phases' (shared.arcs, the one home of membership). (5) The Pipeline app says it on every arc and pursuit card, and every count of arcs or pursuits needing root, core or an operator is read from it.

A one-time backlog is cleared by Orion under his own budget; the agent's model budget keeps only its recurring work

Ruled 2026-11-15 by ninad -- in force

SO: (1) A one-time backlog that an agent's recurring job would otherwise work through on the estate's model budget is cleared by Orion, the operator's session, under his own budget (the operator's Claude plan), recorded on the agent's behalf with the agent's own brief and record format, the agent still accountable (as Orion takes up any agent's activity). (2) The agent keeps the recurring work. (3) Applied to #862: Orion marks every live initiative not assessed since this ruling -- the 518 never marked and the 67 marked under Elrond's first brief -- open pursuit steps first; Elrond marks new filings (through his Effort Door) and records that change afterwards, and his backlog marking stops.

Dependencies are first-class registry records

Ruled 2026-11-15 by ninad -- superseded 2026-11-15

Ninad, 2026-11-15 (CLI-154), asking whether Chowpatty onboarding (CAROL-PUR-0006) depends on Blueprint service (CAROL-PUR-0017) and whether that is factored in. HIS WORDS: "is there dependency between the 2 pursuits? if yes, is it factored in? dependencies should be first class registry artifacts". READ AS (Orion's reading, not his words): a dependency -- an initiative, pursuit or arc that cannot finish until another does -- is its own record in the registry, with both ends, its reason, who filed it and its state, journalled like every governed fact; the marks written today into initiative decisions and pursuit step notes (initiatives #165, the file-a-dependency skill) are derived from that record, never the home of it; and every status reader -- the usual format, the Pipeline app's cards, the plan readings -- reads it, so a pursuit or arc that waits on another says so in its status.

Ended because: Ninad, CLI-154: chose the 'waits on' line and ruled the dependencies enforced

Dependencies are registry records, shown as 'waits on' in every status, and enforced before a step starts

Ruled 2026-11-15 by ninad -- superseded 2026-11-15

Ninad, 2026-11-15 (CLI-154). HIS WORDS: "is there dependency between the 2 pursuits? if yes, is it factored in? dependencies should be first class registry artifacts"; then, offered a 'waits on' line under the status table or counting the prerequisites in: "1. waits on - but these should be enforced, the pursuit should not be able to run the steps until all dependencies are addressed". READ AS (Orion's reading, not his words): a dependency -- an initiative, pursuit or arc that cannot proceed until another initiative, pursuit or arc is done -- is its own registry record with both ends, its reason, who filed it and its state, journalled like every governed fact; the marks in initiative decisions and pursuit step notes are derived from it, never its home. Every status reads it: the usual format keeps the pursuit's or arc's own figures and adds a 'waits on' line naming each open prerequisite, its open steps and hours left. And it is ENFORCED: a step bound by an open dependency cannot start in any lane (planner, Albus's, Orion's), refused by name with what it waits on, until every dependency binding it is met (its prerequisite closed) or withdrawn with a reason.

Ended because: Ninad chose watch mode over enforcement after seeing what a gate would hold (CLI-159)

The pipeline cookbook holds how dependencies, pursuits and arcs, lanes and the backlog work

Ruled 2026-11-15 by ninad -- in force

Ninad, CLI-156, asked whether Sage's file-a-dependency skill should be corrected: "how dependencies are handled, pursuits and arcs are designed, how lanes are decided, how backlog is managed etc should be part of the pipeline cookbook, fold that into the pursuit.." READ AS: the build pipeline's cookbook carries these rules (dependencies, the design of pursuits and arcs, how a lane is decided, how the backlog is managed, and the like); the work is a step of pursuit CAROL-PUR-0080 (Dependencies enforced).

A status never reads 'unmeasured': it is worked out with the filed size, or shows the last measured

Ruled 2026-11-15 by ninad -- in force

Ninad, 2026-11-15 (CLI-154), shown the Blueprint service pursuit's usual-format table reading 'unmeasured' because Elrond had not yet sized what remained of a worked step. HIS WORDS: "dont wait for elrond (for the nth time) either re-calulate or show previous calculated values.. showing unmeasured is just lame..". READ AS (Orion's reading, not his words): the usual format for a pursuit or an arc always carries numbers. A step Elrond has not sized, or whose remaining effort he has not re-sized, is counted from the size filed with it (all of it spent once its work is built, none before) and named under the table; where even that cannot be read, the last measured table is shown with its time. Nobody waits for Elrond's sizing to read a status.

An initiative waits, on start or to end, and its operator's Orion is told when the wait is over

Ruled 2026-11-15 by ninad -- in force

Asked in CLI-156 whether a pursuit-level dependency (made from an overlap note) should hold every step of the pursuit (50 waiting initiatives held) or only its completion (13 held). Ninad: "waiting should be on initiative level.. who ever is designing the pursuit is free to split initiatives such that dependencies can be isolated and pursuits can proceed with caveats. Also, its important to distinguish between dependencies \"on start\" vs dependencies \"to end\".. to address the situation where an initiative can start, but not be closed until the dependencies are met.. status of completion with \"dependencies\" is very common, where a pursuit cannot be completed because it waits for Saee's login or Sahil's login.. when dependencies are met, the initiative should inform operator's orion instance of initiatives that are dependent on that initiative, so that orion can resume the execution by consulting the operator, similar to how operator messages are communicated by orion.." READ AS: (1) what waits is always an initiative, never a whole pursuit or arc; a pursuit's designer splits initiatives to isolate a dependency so the rest of the pursuit proceeds; (2) every dependency says whether it holds the initiative's START or its END (it may start but not close until met); (3) a pursuit may read complete except for its dependencies; (4) when a prerequisite is met, the Orion instance of the operator executing each dependent initiative is told, as operator communications are, and resumes it after consulting its operator.

Dependencies inside one arc, pursuit or epic use the same method

Ruled 2026-11-15 by ninad -- in force

After CLI-156 reshaped pursuit CAROL-PUR-0080 to ruling #877 (an initiative waits, on start or to end, and its operator's Orion is told when the wait is over), Ninad: "note that the same method should be used to manage dependencies within the same arcs, pursuits and epics" READ AS: a dependency between two initiatives of the same arc, the same pursuit or the same work epic is a registry record, on start or to end, enforced and told exactly like one that crosses them; a pursuit's or arc's order alone never stands in for a dependency, and the earlier rule that same-pursuit steps need no dependency because the sequence says their order is retired in place.

An Orion session sizes its own work from its own budget, never Elrond's

Ruled 2026-11-15 by ninad -- in force

this should be the norm, not exception" READ AS: an Orion session sizes the initiatives it files and works from its own budget (the Orion bypass lane) and never asks Elrond's estimator or his budget to do it; this is the standing rule, not a one-off.

Dependencies are registry records shown as 'waits on' in every status; the gate watches and refuses nothing until an operator rules to enforce it

Ruled 2026-11-15 by ninad -- superseded 2026-11-18

Ninad, 2026-11-15 (CLI-159). Asked for a recommendation on whether enforcing dependencies helps or harms the execution of pursuits, he was shown what a gate would hold that day -- a start gate would stop 17 initiatives, and 16 of the 34 open waits sit behind work that is itself parked or redirected -- and the recommendation: keep the record and the 'waits on' line, tell the operator when a wait is over, and run the gate in watch mode before it refuses anything. HIS WORDS: "ok, keep it as watch mode for now..". READ AS (Orion's reading, not his words): a dependency stays a first-class registry record, shown as 'waits on' in every pursuit and arc status, exactly as #873 ruled. The dependency gate at the status router WATCHES: on every start (dispatched, executing) and every close it records the dependency it would have held -- on start or to end (#877) -- and lets the move through; it refuses nothing. Turning it to refuse is a later operator ruling, to end first and on start after, read from what the watch recorded. #873's 'should be enforced, the pursuit should not be able to run the steps until all dependencies are addressed' is retired in place by this ruling.

Ended because: Ninad, CLI-226: the work depended on runs first; pickers take it first and the gate enforces

Whenever an Orion session shows an arc's or a pursuit's status, it sizes what is unsized from the Orion bypass budget so the percent and the hours read

Ruled 2026-11-15 by ninad -- in force

WHENEVER AN ORION SESSION SHOWS AN ARC'S OR A PURSUIT'S STATUS, IT SIZES WHAT IS UNSIZED FROM THE ORION BYPASS BUDGET SO THE PERCENT AND THE HOURS READ (Ninad, 2026-11-15, CLI-161; widens #879).

ASKED: shown CAROL-PUR-0035's status reading 'unmeasured' for percent and hours because the step just closed was not yet sized by Elrond's estimator. HIS WORDS: "update the skill to measure the % and hrs estimates using orion bypass budget whenever you need to show the status via CLI, if its not already done by another session".

READ AS (Orion's reading, not the administrator's words): (1) when an operator's Orion session is about to show a pursuit's or an arc's status and the composer reads unmeasured because a step is unsized -- its first size, or what remains of a worked open step -- the session sizes those steps itself, from the Orion bypass lane's budget (budget-exempt, #1035), never Elrond's, and then shows the composer's table; (2) the sizing uses the estate's one sizing method (Elrond's brief and the same calibration) and is recorded on each attempt as Orion's, never as Elrond's; (3) the skill build-arcs-and-pursuits says so where it says how status is reported; (4) #879 ('an Orion session sizes its own work from its own budget') stands and is widened from the session's own work to every step a status it shows needs; the build is, already filed for #879.

Budget work sits in the Investment plan pursuit, and an initiative sits in one pursuit only

Ruled 2026-11-15 by ninad -- in force

ANYTHING TO DO WITH BUDGETS SITS IN THE INVESTMENT PLAN PURSUIT, AND AN INITIATIVE SITS IN ONE PURSUIT ONLY (Ninad, 2026-11-16, CLI-165).

ASKED: the Sandbox prospect budgets were freed by 7236's close and sat in CAROL-PUR-0042 (Prospects in Sandbox), worked by CLI-162: take them on in this session or leave them to their own lane? HIS WORDS: "pick these up.. anything to do with budgets should sit in investments, the other lane is commissioned to provisioning the vm, which is a distinct activity, if we are referring to the same session.. anyways, an initiative can sit only in one pursuit.. so if its here, be sure (or ensure) that its not in any other pursuit..".

READ AS (Orion's reading, not the administrator's words): (1) prospect budget work -- each prospect's yearly budget (7306), the plan pricing the prospects worked (7308) and the plan showing Sandbox as non-project spend (7277, which waits on 7308) -- is a step of CAROL-PUR-0035 (Investment plan), and the success criterion it proves moves with it; (2) each moved step leaves CAROL-PUR-0042, so it sits in one pursuit only; (3) CAROL-PUR-0042 keeps the Sandbox machine and the routing of prospect work; (4) CLI-165 picks them up.

Operators choosing Albus's work belongs to the Scalable Albus pursuit

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-185): "operators can't choose Albus's work yet - fold this into the albus scale pursuit". Orion asked (CLI-185): makes Albus's lane hand any Sandbox build back to the Orion lane, because his engine builds only in the estate's tree and a Sandbox build is made off it (design 728 v1.4). Ninad said the plan is to let each operator have their own Albus bypass lane and operators decide which lanes he picks up, and asked whether to leave it for the Albus scaling pursuit. Orion recommended moving it to CAROL-PUR-0026 (Scalable Albus) and noted that operators cannot yet choose Albus's work: his backlog lane still claims planned work from the shared queue on its own (initiatives #627). READ AS: the work that lets an operator choose which work their Albus lane picks up sits in CAROL-PUR-0026 (Scalable Albus); is already its step there.

Albus not working on Sandbox belongs to the Scalable Albus pursuit

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-185): "albus cant work on sandbox - fold this into albus scale pursuit". Orion asked (CLI-185): makes Albus's lane hand any Sandbox build back to the Orion lane, because his engine builds only in the estate's tree and a Sandbox build is made off it (design 728 v1.4). Ninad said the plan is to let each operator have their own Albus bypass lane and operators decide which lanes he picks up, and asked whether to leave it for the Albus scaling pursuit. Orion recommended moving it to CAROL-PUR-0026 (Scalable Albus) and noted that operators cannot yet choose Albus's work: his backlog lane still claims planned work from the shared queue on its own (initiatives #627). READ AS: leaves CAROL-PUR-0042 (Prospects in Sandbox) and becomes a step of CAROL-PUR-0026; (Chowpatty's first build) no longer waits on it.

The Artifact ownership pursuit also holds code to its owner

Ruled 2026-11-16 by ninad -- in force

THE ARTIFACT OWNERSHIP PURSUIT ALSO HOLDS CODE TO ITS OWNER (Ninad, 2026-11-12, CLI-120).

ASKED (CLI-120): Orion measured the build against the Agent-Centric Modular Architecture (design 146): in October 82% of new code went into the common library while the agents' own code shrank, and the architecture checker counts 36 places where shared code branches on a particular agent and 22 where one agent imports another's droids. Ninad asked: "is the build following the agent centric modular architecture principles?", then "there must already be a pursuit for this? there was a pursuit for only owning agents to be able to edit their artifacts". Orion found CAROL-PUR-0012 'Every artifact is held to its owner' covers stores, files and records, not where code lives, and that two planned steps sat in no pursuit: (shared code makes no agent's judgment and no agent imports another's droids) and (a shared module's owner reads from its place on record). Offered: add both to CAROL-PUR-0012, stretching its goal from who edits data to where code lives. ANSWERED: "ok, add the missing initiatives to it.."

READ AS (Orion's reading, not the administrator's words): and join CAROL-PUR-0012 as MVP steps (6890 first, so every judgment that moves has a named owner); the pursuit's outcome, MVP and criteria are stretched so code is held to its owner under design 146 -- judgments and jobs owned by their agent, shared code holding only data, systems, the org chart and the rulebook (ruling #515). The other planned steps of the same architecture audit that also sit in no pursuit were not named and are not moved by this ruling.

Chowpatty onboarding splits at the contract into a prospect pursuit and a client pursuit

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-185): "sure, do the split"; and on moving the module-criteria step to the Blueprint pursuit: "sure"; and on the 5-hour pursuit limit Orion had cited: "this guidance was removed, remove any leftovers". Orion had proposed: CAROL-PUR-0006 (Chowpatty onboarding) overlaps the Sandbox pursuit, has no step that builds its proof of concept, and no step moving that proof of concept to Chowpatty's own machine; split it at the contract into 'Chowpatty as a prospect' (its agents, Chowpatty's first build in Sandbox moved from CAROL-PUR-0042, onboarding steps recorded, the proof of concept shown) and 'Chowpatty as a client' (its own machine and apps, Migration of the Sandbox proof of concept and the laptop prototypes, the final real run), the Sandbox pursuit keeping the general machinery every prospect uses; and move (a module's criteria agreed in Leo's chat) to CAROL-PUR-0017. READ AS (Orion's reading, not his words): CAROL-PUR-0006 closes as split into the two; the client pursuit gains a step for the Sandbox proof of concept's Migration; size guidance leftovers are removed under #855.

Design 729 is approved, and the pipeline cookbook carries it and every operator door design

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-190), asked "Do you approve design 729 as written?": "I approve this, include it in the pipeline cookbook, ensure that operator door designs are also included in it". READ AS (Orion's reading, not his words): design 729 is approved as written; the pipeline cookbook gains a rule for design 729, and every design that defines an operator door is carried by a cookbook rule, each cited by the operating manual so a session finds it.

's filing criteria are amended to current law: agents' filings stay on Albus's lane

Ruled 2026-11-17 by ninad -- in force

Ninad (CLI-202), answering Albus's law block on (operator communication #1333), was asked: "Albus is stuck on an initiative you run, the filing-routing work from early September. Its success criteria still expect agents' own filings to go to Elrond's planner lane. Your later ruling sends all of that work to Albus's lane instead, so he can't meet the criteria and asks you to choose. Which do you want?" He chose the option "Amend the criteria", which read: "Rewrite the criteria to match current law: agents' filings and pending work stay on Albus's lane. Your 'nothing goes to the planner lane' ruling stands, and Albus can finish." He declined the option to restore autonomous work to the planner lane. READ AS (Orion's reading, not his words): 's autonomous-filing and pending-stock criteria are revised, through the governed goal revision, to what cookbook 1848 and Rulings Repo initiatives #222 already say -- an autonomous agent's filing, and an explicit planner ask from a chat, land on Albus's bypass lane and nothing waits on the planner lane; the old criteria are struck and kept, never deleted.

Every initiative sits in one arc, and also in a pursuit when one needs it; whoever files it places it with the shared placement skill, and Elrond's gate refuses a filing with no arc

Ruled 2026-11-17 by ninad -- in force

EVERY INITIATIVE SITS IN ONE ARC, AND ALSO IN A PURSUIT WHEN ONE NEEDS IT (Ninad, CLI-195, 2026-11-17). HIS WORDS: "I see that many initiatives belong to pursuits but do not belong to any arcs.. make a policy that every initiative should fall into an arc as a mandatory requirement (along with the pursuit) and do the assignments for existing initiatives into appropriate arcs.. if an appropriate arc is not present, create one.. include this work in an appropriate pursuit and give me the name of the pursuit". Asked to approve draft words, HIS WORDS: "associating the filed initiatives to appropriate pursuit or arc should be done by whoever is filing the initiatives, is that Elrond? re-usable skills should be used". Told that the filers are the operator's Orion and the agents (Albus among them), that Elrond's gate only checks a filing at the door, and that the shared skill for deciding an initiative's arc and pursuit (kept by Sage) does the deciding, he APPROVED these words for policy P.01.03.05.24: "Every initiative sits in one arc, and also in a pursuit when one needs it. Whoever files it places it, using the shared placement skill. Elrond's gate refuses a filing with no arc." Asked which existing initiatives to place, he chose "Open + all pursuit steps" (216 found: 178 in a pursuit and no arc, 38 open in neither; about 5,600 closed or discarded initiatives in neither stay as history). Asked where the work lives, he chose "New pursuit, bypass". READ AS (Orion's reading, not his words): an arc is mandatory for every initiative and a pursuit is in addition, never instead; #665's 'sits in one arc or a pursuit' and 'an operator's filing names its arc or pursuit' become 'sits in one arc' and 'names its arc phase, and its pursuit when one needs it', and #163's 'from any arc or from none' reads 'from any arc'; #665's other parts (executing operator, the agent arc of each service, no pursuit for agents' filings) and #744 (one arc and one pursuit at most) stand; the placement gate moves from record to enforce once the backfill is done (#391: teach, backfill, then enforce); a filing's placement is its filer's act, Albus's own filings placed as his.

Arcs are topic folders within a service, pursuits are objective folders that may span services, and epics are folders of pursuits; the placement skill folds an initiative into the arc of its topic or starts a new arc

Ruled 2026-11-17 by ninad -- in force

ARCS ARE ORGANISED BY TOPIC, PURSUITS BY OBJECTIVE (Ninad, CLI-195, 2026-11-17). HIS WORDS: "Arcs should be topic oriented, even if they sit within a service.. for example, the arc initially inherits the topic of the initiatives in it, and then more initiatives associated to that topic get fold into that arc, while initiatives of different topic form a new arc.. the skill ises this approach to allocate initiatives to arcs. while arcs are topic oriented, pursuits are driven by objectives and outcomes, and may spread across services.. this is already documented, but it should also be used to ensure proper arc and pursuit management, think of arcs and pursuits as folders, while initiatives are the files, and its just about organizing the files in the right folders.. we also have the concept of epics that group pursuits that work together to achieve a common goal, like folders in another folder". Then, on the dictionary: "these definitions should already exist in the dictionary.. pls check" -- told the dictionary holds Arc, Pursuit and Work epic, he approved changing Arc's 'about that service's subject' to 'about one topic within that service'. READ AS (Orion's reading, not his words): the shared skill for deciding an initiative's arc and pursuit carries this method -- an arc takes its topic from the initiatives that start it, an initiative on that topic joins it, one on a different topic starts a new arc of its service; a pursuit gathers initiatives by the objective they serve, across services; a work epic gathers pursuits; the same method keeps arcs and pursuits in order when they are reviewed. It is method, not policy (#3 below).

Open initiatives off the roadmap found during placement take the roadmap item their topic arc serves, then join that arc

Ruled 2026-11-17 by ninad -- in force

Ninad, CLI-195, 2026-11-17. Told that 31 old open initiatives (25 in review, 6 parked; 29 filed by Albus, 2 by his Orion) name no roadmap item, so #182 (off the roadmap joins no arc) keeps them out of every arc against #973 (every initiative sits in one arc), and offered: link to roadmap then place / park and tell the filer / discard as stale, HE CHOSE "Link to roadmap, then place". READ AS (Orion's reading, not his words): each takes the roadmap item its topic arc already serves, read from that arc's other members, and is then placed in that arc; Albus's filings are linked and placed as his (the filer's act, #667); #182 stands for new work -- a filing still names its roadmap item before it joins an arc.

Hagrid running the estate's machines is a pursuit of its own, apart from Lightweight projects and Blueprint, owned by Hagrid and taken to completion by the administrator

Ruled 2026-11-18 by ninad -- in force

HAGRID RUNNING THE ESTATE'S MACHINES IS ITS OWN PURSUIT (Ninad, CLI-211, 2026-11-18; after initiatives #974 and #744, system-services #972 and #977).

THE QUESTION: Ninad asked "what pursuits includes the work of enabling hagrid to distribute resources intelligently across machines, and to ask for more machines when necessary? what are his safety cushions?". Orion answered that most of it sat in Lightweight projects (CAROL-PUR-0084: 7409, 7423, 7421, 7411), the project-machine parts in Blueprint (CAROL-PUR-0017: 7331, 7311, 7310, 7431), the Infra Cookbook (6807) in no pursuit, and the allocation of Carolverse's resources across its services (system-services #972) filed nowhere. HIS WORDS: "why most of it? a pursuit is objective oriented..". Orion proposed a pursuit 'Hagrid runs the estate's machines' -- every machine, Carolverse's and each project's, sized, shared and asked for by Hagrid within set safety limits and bought only on the administrator's approval -- taking 7409, 7423, 7421, 7411 from Lightweight projects, 7331, 7311, 7431, 7310 from Blueprint, 6807 from no pursuit, and a new initiative for Hagrid deciding which services share which machine. HIS WORDS: "yes, do that, create a dedicated pursuit that I can take to completion".

READ AS (Orion's reading, not the administrator's words): (1) a new pursuit owned by Hagrid holds Hagrid's management of machines: 7409, 7423, 7421 leave CAROL-PUR-0084; 7331, 7311, 7310, 7431 leave CAROL-PUR-0017; 6807 joins; a new initiative files #972's deferred part (Hagrid decides which Carolverse services share which machine, from measured load); (2) 7411 (the investment plan's Azure line counts Carolverse's machines from load) STAYS in CAROL-PUR-0084 with its criterion 7 -- it is the plan's cost line, not Hagrid running machines; Orion told the administrator so in the same reply; (3) CAROL-PUR-0084's criteria 6 and 8 are reworded to what its own steps still prove (each project's data in its own store; Themis names a shared store as debt), their machine halves becoming the new pursuit's criteria; (4) the headroom limit and the hours load must stay above it (design 730's declared values) are set by the administrator when 7409's capacity watch is built -- not yet set; (5) the administrator is the operator owner and takes the pursuit to completion.

The work an initiative waits on is executed first: every picker takes it first, it stays in its own pursuit moved up, and the dependency gate enforces

Ruled 2026-11-18 by ninad -- in force

Ninad, 2026-11-18 (CLI-226). Shown that in watch mode the dependency gate had recorded 7 initiatives starting before the work they waited on (one on discarded work), and that no picker reads dependencies, he asked whether Orion still held the watch-only recommendation. HIS WORDS: "If there is dependency, then the initiative on which dependency exists should be executed first.. you had recommended to keep it watch only then, do you still hold on to that recommendation?"; then, on Orion's plan (settle the waits that can never end, teach every picker to skip a held initiative and take the work it waits on, switch the gate on): "Teach every picker to skip a held initiative and take the work it waits on instead - if it picks the work from another pursuit, then both the pursuits should get updated, because an initiative cannot stay in more than one pursuit". Asked what 'both pursuits get updated' means, he chose: "Stays put, moved up" -- the earlier work stays in its own pursuit, which moves it to next in line and notes who it was taken up for, and the waiting pursuit's step notes that its prerequisite is now being worked. Asked the lane: "Orion bypass". READ AS (Orion's reading, not his words): the work an initiative waits on is executed before it. Every picker -- Albus's lane, the Orion bypass's take, Merlin's dispatcher -- skips an initiative held by an open dependency and takes the open work it waits on instead; that work never joins a second pursuit (#893): its own pursuit moves it next and records who it was taken up for, and the waiting pursuit's step records it under way. Once the pickers do this, the dependency gate is switched from watch to enforce as the backstop. #888's 'the gate watches and refuses nothing until an operator rules to enforce it' is retired in place by this ruling; its record and 'waits on' line stand. Built on the Orion bypass as steps of CAROL-PUR-0080.

Parking names its wait: a parked initiative records what it waits on, the parked work is swept once, and the board names parked work with no recorded wait

Ruled 2026-11-18 by ninad -- in force

Ninad, 2026-11-18 (CLI-226). Shown that Saee's steps 6332 and 6595, said to wait on her first setup on her new laptop, carry no recorded wait, so her signing in would tell nobody; that 91 initiatives are parked and only 10 carry a recorded open wait; and that the cause is structural (the migration read only initiative-on-initiative marks, waits on a person became recordable only on 17 Nov with nothing going back, and parking takes free text), he asked: "but that wait was never recorded - how did this gap happen and what can be done to avoid it happening again?" Offered three fixes -- parking names its wait, a one-time sweep of the parked work, and a board line counting parked work with no recorded wait -- he answered: "fold these fixes into the pursuit". READ AS (Orion's reading, not his words): parking an initiative names what it waits on as a dependency record -- another initiative, or a person's action (a computer reporting in, an operator's answer, a root step's log, a release, a moment) -- or says in words that nothing does; the gate RECORDS a park naming none and refuses only once an operator rules to enforce it, as the dependency gate did (#888 then #1005). Every initiative parked today gets its real wait recorded once, Saee's 6332 and 6595 first. The board names, every session, the parked work that carries no recorded wait. Built as steps of CAROL-PUR-0080 (Dependencies enforced).

Root governance only closes now; the first live web-route review is a test left to the next real route install

Ruled 2026-11-18 by ninad -- in force

Ninad, 2026-11-18 (CLI-231). Told in plain words that CAROL-PUR-0016's one open step, waits only for the first real use of the estate operators' web-route review -- the review is built and installed as an estate operator's act, and the last route install (9 Nov) rewrote exactly what root had already reviewed, so there is nothing to review until an app's web address is added or changed -- he answered: "if testing is what is pending, then it will be done when the next app is created, you may close the pursuit, good job!!" READ AS (Orion's reading, not his words): the pursuit closes as achieved now; criterion 3 is revised to what is proven today (a service added through the act door on his own operator sign-in, the route review installed as an estate operator's act, the root-load audit PASS each time) and names the first live route review as owed at the next real route install; leaves the pursuit's sequence and stays parked in its own arc until that install, closing on its own live check.

Every build, on Carolverse and every project built on it, keeps the agent-centric modular architecture; a deviation a review finds is fixed in that build, never sent to the backlog

Ruled 2026-11-18 by ninad -- in force

Ninad, CLI-237, asked whether Athena's Operator Approval Herald (an Orion-owned courier doing Athena's telling) is in line with the agent-centric modular architecture: "architectural compliance is important, and if I file an initiative, it will go in the backlog and will never get implemented.. I want you to ensure not only that the build is aligned with the architecture, but also how in future will it not digress for future builds. This is important because this is the backbone of the offering, all the projects that are built on carolverse need to follow consistent architectures and design, which is the usp of the platform.. find a way to ensure that the agent centric modular architecture is at the core of every future build, and that Albus's review or Elrond's review captures any deviations which are immediately fixed, not pushed back in the backlog.."

A build that touches a file leaves that whole file architecture-clean

Ruled 2026-11-18 by ninad -- in force

Ninad, CLI-237, asked what happens to the ~1,700 standing architecture violations under ruling #1025 -- (A) any build that touches a file must leave that whole file clean, so the debt burns down as work flows through; or (B) A plus Albus's lane burning it down daily as top-priority work -- answered: "A". READ AS: a build's close is refused while any file it wrote carries any architecture violation, old or new; no separate burn-down lane is ruled.

A build that touches a file leaves that whole file architecture-clean

Ruled 2026-11-18 by ninad -- retired 2026-11-18

Ninad, CLI-237, asked what happens to the ~1,700 standing architecture violations under ruling #1025 -- (A) any build that touches a file must leave that whole file clean, so the debt burns down as work flows through; or (B) A plus Albus's lane burning it down daily as top-priority work -- answered: "A". READ AS: a build's close is refused while any file it wrote carries any architecture violation, old or new; no separate burn-down lane is ruled.

Ended because: duplicate of #1027: CLI-237 re-ran the script that recorded the same answer ("A") a second time; #1027 stands (by ninad)

A droid serving a task of a service its owner is not in is doing another agent's work, unless that task is recorded as lent under a named ruling

Ruled 2026-11-18 by ninad -- in force

Ninad, CLI-240, asked for how the one architecture checker should tell a courier (a process doing another agent's work under its own login, e.g. Orion's job sending Athena's, Albus's or Leo's notices because their logins cannot reach the message store) from a droid properly lent to another service's task (about 195 droids serve another service's task by design: chat sign-ins, minds, app stewards, wellbeing checks, status reports). Offered: 'Service + lending record' -- flag any droid serving a task of a service its owner is not in, unless that task is recorded as lent under a named ruling; 'Notices only'; or a declared 'acts for' field. He answered: "Service + lending record (Recommended)". READ AS: a droid whose declared task folds to a service its owning agent is not in does another agent's work, unless that task carries a lending record naming the ruling that lends it; the lending record is a record, never a list in code.

Initiative estimates learn from the actual time each initiative, arc and pursuit took, across every lane

Ruled 2026-11-19 by ninad -- in force

Ninad: "is it possible to track the actual time taken to execute an initiative, arc, pursuit from the operator instance log time captured, and compare it to the estimated, and use that data to revise the model for future estimations?" -- "you would need to consider the fact that multiple sessions run at the same time.." -- "operator instance calculates only orion bypass, what about the other lanes like albus bypass? it should also be included to calculate the actuals.. check if initiative logs, palantir etc are better sources of data for the estimation model than operator instance log, or you can use multiple of the sources to improve accuracy, as long as you are able to resolve the conflicts and create a trusted source before feeding the data into the model.." -- "how will this data be used for improving the proedictions is the second part of the puzzle.. this will affect the assumptions also help build tolerances that are based on historic data." -- "the pursuit for improving estimations for initiatives should be the first use case of this service" (the Data Quality service). Then: "yes, file both pursuits, the data quality and estimator improvements".

A bypass start claims before it creates anything, so a late dependency hold opens nothing

Ruled 2026-11-19 by ninad -- superseded 2026-11-19

Asked (CLI-251, 2026-11-19): a start passes its dependency check, then a new dependency lands in the few seconds before the start claims the work; by then the start has already created a run record, and the estate always grades the newest record as an attempt. How should that seconds-long window behave? Offered: a failed claim with the hold noted; redesign so the claim comes first; stop for now. Ninad chose: "Redesign: claim first -- Rebuild the start so the claim (and any hold) happens before anything is created, so a late hold creates nothing." READ AS (Orion, not Ninad's words): bypass_start makes its claim through the one status router before it mints the session, execution or run; a dependency hold the router gives at the claim is held exactly like the early check -- nothing opened, no attempt counted, DependencyHeld raised; #1005 stands unnarrowed.

Ended because: Ninad chose the simplest option over claim-first once shown the A-waits-on-B window

A hold arising between a start's check and its claim ends that start as a failed claim, with the hold noted

Ruled 2026-11-19 by ninad -- in force

Asked (CLI-251, 2026-11-19), after Ninad asked for an easier solution than claim-first: if A depends on B, and 'A waits on B' is recorded in the few seconds after A's start was checked but before A is claimed, the gate rightly refuses the claim, but A's working records already exist and read as a failed try. Offered: lock dependencies during a start; count anything in that few-second window as a failed start with the hold noted; claim first (#1061). Ninad chose: "Simplest: anything in that few-second window counts as a failed start, with the hold noted" -- "sounds good to me". READ AS (Orion, not Ninad's words): a dependency hold the gate gives at the claim, after bypass_start's own check passed, ends that start as a failed claim (its run and execution failed, the try counted as always) and writes the hold on the initiative naming what it waits on; every hold the start's own check finds stays held with nothing opened (#1005).

Every build gets Albus's architecture pointers before it starts and Elrond's check that they were followed at its end

Ruled 2026-11-19 by ninad -- in force

Context (CLI-251, 2026-11-19/20): Ninad: "its good that elrond is reviewing all the edge cases, it would have had been better if albus or elrond would have also reviewed architecture compliance of the build after every build.. thats the biggest pain area for me currently". Orion answered that the one architecture checker's fixed rules already run at close and at Elrond's acceptance, but bite only at the end and cannot judge design fit, and proposed Albus review each build. Ninad: "I want Elrond to do this check.. both before the start of the execution, Albus should list out the relevent solution pointers that ensure compliance to agent centric modular architecture that is currently followed in carolverse, and towards the end of the build, Elrond checks ensuring that the pointers were followed.." READ AS (Orion, not Ninad's words): before any build's execution starts, Albus writes the solution pointers for that initiative -- what the agent-centric modular architecture asks of this build (owner, placement, one composer, no duplicate, the doors it must use) -- recorded on the initiative; at the end of the build Elrond checks the build against those pointers, beside the existing architecture checker, and a pointer not followed fails the check. Every lane, the Orion bypass included.

Self-serve sign-up is its own pursuit, owned by Jarvis

Ruled 2026-11-19 by ninad -- in force

SELF-SERVE SIGN-UP IS ITS OWN PURSUIT, OWNED BY JARVIS (Ninad, 2026-11-19, CLI-258).

ASKED: "Where should the self-serve sign-up work live? It touches Carol (the offer in chat), Leo (the welcome), Midas and Marco (payment and terms), Hagrid (the sandbox folder) and Jarvis (who owns user onboarding)." Options: 'New pursuit, owned by Jarvis (Recommended)' -- "A new 'Self-serve sign-up' pursuit holding the steps that each team builds. Blueprint stays focused on onboarding projects, and the pricing-model item I claimed moves there as the 'accept the terms' step."; 'Inside Blueprint'.

HIS ANSWER: "New pursuit, owned by Jarvis (Recommended)".

READ AS (Orion's reading, not his words): (1) a new pursuit owned by Jarvis holds every initiative the self-serve journey needs, each in the arc of its own service; (2) moves out of the Blueprint pursuit CAROL-PUR-0017 into it as the 'accept the terms' step.

A bypass run that genuinely lasted under a second may read zero; a reaped run ended at its start may not

Ruled 2026-11-20 by ninad -- in force

Ninad, 2026-11-19 (CLI-257), building (a bypass run records its real working span, never zero time; pursuit CAROL-PUR-0089). Shown that only 3-6% of Orion's runs a month closed with no length (74 reaped since September) while the duration field was set on 35 of 3,921 runs, and asked plainly "should a run that genuinely lasted under a second be allowed to read zero?", HIS WORD: "yes". READ AS (Orion's reading, not his words): a run its session opened and closed within the same second reads a length of 0 and is an HONEST zero, counted and named by the time check (shared.bypass_run_time), never a fault and never padded to a second; a run the reaper ended at its start for want of evidence is a FAULT; durations are derived at each run store's write seam from the row's own start and end, never typed and never backfilled.

A pursuit can be in fast track mode: the operator's priority, taken up first and shown as such; Blueprint and Voice chat are

Ruled 2026-11-20 by ninad -- in force

A PURSUIT CAN BE IN FAST TRACK MODE, THE OPERATOR'S PRIORITY (Ninad, CLI-265, 2026-11-20: "create a fast track mode for pursuits, add this term in dictionary, and add blueprint and voice chat to it, so you know that they are on priority for me" (after "build a new pursuit dedicated for voice chat so I can track it on a fast track mode"))

READ AS (Orion's reading, not the administrator's words): (1) every pursuit has a mode, from the reference set pursuit.mode: normal (worked in its place in the live priority order) or fast-track; (2) a pursuit in fast track mode is its operator's priority: the Orion sessions of the operators who execute it take up its open work before any other pursuit's, and it is shown as FAST TRACK wherever pursuits are listed -- the pursuit list, its status table and the operators' board; (3) the mode is set and cleared only on the operator's word, recorded on the pursuit and journalled like its other fields; (4) the term fast-track is in the dictionary, from the set's meaning; (5) CAROL-PUR-0017 (Blueprint) and CAROL-PUR-0092 (Voice chat) are in fast track mode from now.

An MVP's work sits in one common pursuit made for it, never spread across pursuits; the MBank MVP's pursuit is Leo's

Ruled 2026-11-20 by ninad -- in force

AN MVP'S WORK SITS IN ONE COMMON PURSUIT, NEVER SPREAD ACROSS PURSUITS (Ninad, CLI-266, 2026-11-20; on the MBank MVP, whatsapp #1084, blueprint #1085)

THE QUESTION AND HIS WORDS: shown the Voice chat pursuit's status: "does this pursuit capture all the requirements of the MVP or are they spread across multiple pursuits? if they are spread across multiple pursuits, then thats not ok, move them as arcs, and create a common pursuit for the MVP"; shown that the work is spread across Voice chat (CAROL-PUR-0092) and Blueprint (CAROL-PUR-0017) and asked how to reorganise, he chose "One MVP pursuit (Recommended)" -- a new pursuit holds every step the MVP needs, moved out of Voice chat and Blueprint; Voice chat merges into it and closes, its steps staying in their service arcs; Blueprint keeps its other work as a pursuit -- and, asked which agent owns it, "Leo (Recommended)".

READ AS (Orion's reading, not the administrator's words): (1) every initiative an MVP needs sits in ONE pursuit made for that MVP; an MVP's work spread across several pursuits is not acceptable; (2) the work keeps its place in its service's arc ('move them as arcs'): the arcs hold work by subject, the one pursuit holds the MVP's goal; (3) for the MBank MVP: a new pursuit owned by Leo holds every step it needs, taken out of Voice chat and Blueprint; Voice chat merges into it and closes; Blueprint keeps its other work as a pursuit.

The MBank pursuit is a POC

Ruled 2026-11-21 by ninad -- in force

Ninad: "the MVP pursuit should be renamed as POC, because its not an MVP.." READ AS: CAROL-PUR-0093's name says MBank POC; its key, steps and stages are unchanged (keys are never renamed).

Voice calls on Carol's external chat window get their own fast-track pursuit

Ruled 2026-11-22 by ninad -- superseded 2026-11-22

Asked (CLI-305): voice on Carol's external chat window sits only inside the MBank POC pursuit (CAROL-PUR-0093), which another session runs; it could get its own fast-track pursuit, owned by Galadriel as Agent Chat's owner. Ninad answered: "yes, give it its own fast track purtsuit, since mBank is taking its own time, i need something to show in my meeting today". READ AS: a new pursuit for voice calls on Carol's external chat window, in fast track mode, Galadriel its owner and Ninad its operator owner, 7658 its MVP step; the MBank POC keeps 7658 as a step it shares (its web call to MBank's sales agent is the same build), the overlap noted in both pursuits.

Ended because: the first reading said the MBank POC keeps 7658 as a shared step, which initiatives #744 does not allow (an initiative belongs to at most one pursuit); the answer is unchanged, the reading is corrected

Voice calls on Carol's external chat window get their own fast-track pursuit

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-305): voice on Carol's external chat window sits only inside the MBank POC pursuit (CAROL-PUR-0093), which another session runs; it could get its own fast-track pursuit, owned by Galadriel as Agent Chat's owner. Ninad answered: "yes, give it its own fast track purtsuit, since mBank is taking its own time, i need something to show in my meeting today". READ AS: a new pursuit for voice calls on Carol's external chat window, in fast track mode, Galadriel its owner, 7658 its MVP step -- 7658 MOVES out of the MBank POC into it, because an initiative belongs to at most one pursuit (initiatives #744); the MBank POC's web call to MBank's sales agent waits on it across pursuits.

๐Ÿ“–Communications

An email address names an agent, or is support -- never a function

Ruled 2026-09-23 by ninad -- in force

An estate email address identifies a WHO, never a WHAT. It either names an AGENT, who can answer for what was sent and receive the reply, or it is the general support address, which is the estate itself speaking when no individual agent is the author. A function, a service, an app or a workflow never gets an address of its own. Ninad ruled it on 2026-09-23 (CLI-519) by declining one of ten proposed aliases: "i did not create roadmap, because it should be the agent who owns the service or support" -- so mail sent by machinery goes out from the agent who OWNS the service that machinery belongs to (roadmap approvals from Clara, who owns the Carolverse Strategy app), or from support@. THE TRAP: when mail is sent by machinery rather than by an obviously named agent, naming the process feels more honest than picking an agent, and it is the opposite -- it recreates the fault the change was fixing, a recipient who cannot tell who wrote or who will read their answer. This completes the ruling of 2026-09-22 (CLI-515) that an agent writes from its own identity and never another's: that one shut the door on borrowing a colleague's address, this one on inventing a nobody's. LIVE AS OF 2026-09-23: ten addresses exist as aliases of the one paid mailbox and are registered send-as, proven by a probe that left as Orion; before them, eight agents all wrote from Carol's address under a borrowed display name.

Orion copies the administrator on every email he sends to a person

Ruled 2026-09-23 by ninad -- superseded 2026-09-23

The CC is part of the message as composed, so a reply-all reaches him; a copy forwarded after sending is the repair for a miss, never the practice. It is never the administrator's personal address. Copying him does not change who writes: the email still goes out from Orion's own estate address (communications #96, cookbook 1809), and copying is not sending on his behalf. A send path that cannot carry a CC is a gap to raise, never a reason to send without it. Ninad, 2026-09-24 (CLI-533), on finding he was not copied on Orion's login-line email to the deputy operator: "that should be the norm".

Ended because: Ninad corrected the rule in the same sitting: it binds the PROJECT'S OPERATORS on every communication about the project and the projects built on it, not one named administrator on Orion's email alone.

The project's operators are copied on every communication regarding the project

Ruled 2026-09-23 by ninad -- in force

Every communication regarding a project carries that project's operators on CC, and the operators of every project it is built on. Within a dominion every project is built on the dominion's base project, so the base project's operators are copied on communications about any project in the dominion; the copy stops at the dominion, exactly as an operator's reach does (security #98), because a forked dominion answers to its own client. The copy list is DERIVED from the operator records (each operator's project and scope) and the project and dominion records, never typed; it uses each operator's work address, never a personal one, and leaves out whoever is already a direct recipient. The CC is part of the message as composed, so a reply-all reaches them; a copy forwarded after sending repairs a miss and is never the practice. Copying does not change who writes: the sender is still the writing agent's own estate address (communications #96, cookbook 1809). A send path that cannot carry the copy, or cannot say which project a message concerns, is a gap to raise, never a reason to send without it. Ninad, CLI-533: "the rule should be that the project operator should be in cc in all communications regarding the project.. and projects on top of that project".

Operators communicate through their Orion operator instances, by a documented protocol, and an Orion session picks operator communications up first at induction

Ruled 2026-09-28 by ninad -- in force

ORION OPERATOR INSTANCES CARRY THE COMMUNICATION BETWEEN OPERATORS, AND IT IS READ FIRST (Ninad, CLI-605, 2026-09-28). Asked how a refreshed start package should reach a new operator whose laptop still held the earlier one -- a one-way note by email, or handed over by the administrator -- Ninad chose neither and answered: 'orion operator instances should facilitate this communication (maybe via escalations route?), the operator onboarding app should document the exact protocol of communication in the operator co-ordination section of the app. This should be the official channel for communications across the operators within and across the projects of oratorium'. And, the same sitting: 'operator communications should be picked on priority by orion sessions above everything else during the induction'. So: (1) the official channel for communication between operators, within a project and across the projects of Oratorium, is their Orion operator instances: the acting operator's Orion instance writes the communication into the estate's record and the receiving operator's Orion instance reads it from that record and brings it to its operator. (2) The exact protocol of each kind of communication is documented in the Operator co-ordination section of the Operator Onboarding app, as the scenario, the action of the acting Orion instance and the action of the receiving Orion instance. (3) At induction an Orion session picks up the operator communications addressed to its operator before everything else on its board. Agent-resources #194 stands: one email starts a new operator's own Orion instance and email otherwise carries only regular one-way updates.

The operator communication system serves every operator, carries broadcasts, and is root-owned

Ruled 2026-09-29 by ninad -- in force

THE COMMUNICATION SYSTEM BETWEEN OPERATORS IS FOR EVERY OPERATOR, THE ADMINISTRATOR MAY BROADCAST TO ALL OF THEM, AND THE SYSTEM IS ROOT-OWNED (Ninad, CLI-606, 2026-09-29). His words: 'what you are building is an inter communication system between operators. This should not be limited to ninad and sahil, any operator should be able to send any instruction via orion, and when the other operator logs in, he should get those instructions during the induction part 2 on highest priority. This would mean that the operator sending the instruction needs to tell orion who this message is intended for, and admin should be able to broadcast the instructions to all operators. The communication system across operators should be root owned'. So: (1) nothing in the system names a person -- the senders and receivers are whoever root's authority record lists as active operators at that moment, and an operator added later is served with no edit; (2) the sending operator tells their Orion session who the communication is for, and Orion never guesses a receiver; (3) the receiving operator's Orion session brings it to them at induction Part 2 before everything else (communications #203 stands); (4) the administrator, and only the administrator, may address one communication to ALL operators -- every operator active on root's record receives it, each reads and closes their own copy of it, and an operator appointed while it is still open receives it too; (5) the system -- its record and the door that writes it -- is owned by root, so no login below root can alter, remove or forge a communication by reaching the store directly: operators send, read and close through the door, which reads the caller from the operating system. Communications #203 is amended by this ruling, not replaced.

A communication between operators also reaches the receiving operator by email

Ruled 2026-09-29 by ninad -- in force

A COMMUNICATION FROM ANY OPERATOR'S ORION INSTANCE TO ANY OTHER OPERATOR'S ORION INSTANCE IS ALSO RECEIVED BY EMAIL BY THE RECEIVING OPERATOR (Ninad, CLI-611, 2026-09-29). Asked how he would know that Sahil's Orion instance had raised something so that he could log back in -- nothing told him: a communication is seen only when the receiver's next Orion session opens -- his words: 'the permanent fix, where a message from any operator instance directed to any other receiving operator instance is also received via email by the receiving operator'. So: (1) every communication on the operators' system, a broadcast and an answer included, is also sent by email to the operator it is addressed to, at the work address root's record gives for them, for every operator and not for named ones; (2) the record on the estate stays the communication itself and the Orion sessions stay where it is read, answered and closed (communications #203 and #207 stand): the email is how the receiver learns that it is there, and an answer typed into a reply to that email is not an answer; (3) the email is sent from Orion's own address, since Orion instances are what communicate (cookbook 1809, communications #96); (4) what may never travel in a communication may never travel in its email either -- no key, password, code, credential or digest -- and the file a communication carries is named in the email, not attached; (5) a communication whose email could not be sent is still delivered on the estate, and the failed send is recorded and shown, never read as sent. This amends the sentence of agent-resources #194 and communications #203 that email carries only regular updates: the email of a communication is one more thing email carries, and it still asks no reply by email.

Every operator's Orion instance reports its main activities hourly to the Operator Instance Log

Ruled 2026-09-29 by ninad -- in force

EVERY OPERATOR'S ORION INSTANCE REPORTS ITS MAIN ACTIVITIES HOURLY, AND THE OPERATOR INSTANCE LOG SHOWS THEM (Ninad, CLI-611, 2026-09-29). His words: 'the orion instances of all the operators should provide an hourly update of the main activities to admin orion via a dedicated app called operator instance log, owned by orion, accessible to all operators, and it should provide a short update (max 3 to 4 lines) of the operator activity of all operators (exept core and root), sorted by latest on top for each operator box as a drill down in the apps landing.. this will maintain the transperancy of operator work. Note that operators on the base floor (carolverse project) should have access to operators on the floors above (projects built on carolverse) and not the other way round'. So: (1) the Orion instance of every operator, the administrator's included, records a short update of its operator's main activities every hour; (2) an update is three to four lines at most; (3) the app is named Operator Instance Log and is owned by Orion; (4) its landing shows one box per operator, and a box opens into that operator's updates, latest on top; (5) the core and root logins are not operators' Orion instances and have no box; (6) every operator may open the app, and what an operator sees follows the floors: an operator of the base project (Carolverse) sees the operators of the base project and of every project built on it, and an operator of a project built on it does not see the base project's operators; (7) the purpose is transparency of operator work.

The operators hold a five-minute catch-up every weekday, set up by a calendar invitation

Ruled 2026-09-29 by ninad -- superseded 2026-10-13

THE OPERATORS HOLD A FIVE-MINUTE CATCH-UP EVERY WEEKDAY, AND IT IS SET UP BY A CALENDAR INVITATION TO THEIR WORK ADDRESSES (Ninad, CLI-611, 2026-09-29). His words: 'setup a meeting between me and sahil in our denkan labs email on a daily basis on week days only at 8 pm IST for 5 min to catchup on denken agents work'. So: Monday to Friday, 8:00 pm India time, five minutes, the administrator and the deputy operator, on Denken Agents work; the invitation goes to their work addresses. This is a standing meeting between people and excepts, for calendar invitations between operators, the clause of communications #203 and agent-resources #194 that an email never books a time; everything else those rulings say stands -- questions, answers, keys and hand-offs still travel between Orion sessions.

Ended because: Ninad added Saee to the catch-up (CLI-754)

The administrator's sign-in sees every operator's box in the Operator Instance Log

Ruled 2026-09-30 by ninad -- in force

THE ADMIN LOGIN SEES EVERY OPERATOR IN THE OPERATOR INSTANCE LOG (Ninad, 2026-09-30, CLI-624, on opening the app signed in with his personal address and seeing no box): 'since gmail is the admin login, it should show all operators'. A signed-in person the estate's ONE admin derivation reads as the administrator (the tier every app gate uses: the owner record's addresses) sees every active operator's box on every floor, whether or not that address is an operator's work address. Everyone else is unchanged: an operator is matched by work address and reads by floor (#210); a signed-in person who is neither sees no box. The admin address is DERIVED from that one derivation, never typed into the log or the app.

What an operator's box and its detail show in the Operator Instance Log

Ruled 2026-09-30 by ninad -- in force

AN OPERATOR'S BOX SHOWS WHO THEY ARE AND TODAY'S HOURS; OPENING IT SHOWS A MONTH OF WORKED HOURS AND THE DAY'S WORK (Ninad, 2026-09-30, CLI-624, amending the box of #210): 'the operator box should only provide the details of the operator like the status, role, location, claude plan, mapped agent and total hrs worked on current day; the detail should have the following charts: 1. trend bar chart for a month [of] the hrs of work each day; 2. each selection should [show] the pursuits, arcs, initiatives worked on by the operator in a card below'. Asked four questions, he answered: HOURS WORKED are the clock hours in which the operator's Orion session recorded work, measured from session records and never typed; LOCATION and CLAUDE PLAN are reported by the operator's own workstation at session start and shown with when they were reported; the MAPPED AGENT is a record per operator ('not all operators will be mapped to orion, currently sahil and ninad both are mapped to orion'); the HOURLY UPDATES stay, in the detail below the day's card. The box no longer shows the updates themselves.

Every summary names the arc and the pursuit the work belongs to

Ruled 2026-10-02 by ninad -- in force

EVERY SUMMARY NAMES THE ARC AND THE PURSUIT THE WORK BELONGS TO: 'also include the name of the arc / pursuit worked on in every summary..'. READ AS (Orion's reading, not Ninad's words): it binds every summary an Orion operator instance writes -- the hourly operator update (#210), the session handover and the summary given to the operator at the end of a piece of work -- and each item of work in it carries the readable NAME of its arc (and phase) and of its pursuit, the key beside it where useful; work that sits in no arc or pursuit says so in those words rather than leaving the names out.

Any non-compliance of an operator's Orion instance shows that operator's box red in the Operator Instance Log

Ruled 2026-10-02 by ninad -- superseded 2026-11-09

ANY NON-COMPLIANCE OF AN OPERATOR'S ORION INSTANCE SHOWS THAT OPERATOR'S BOX RED IN THE OPERATOR INSTANCE LOG (Ninad, CLI-655, 2026-10-02). His words: "any non compliance should show up the operator button red in the operator instance log" -- said while CAROL-ARC-0036 was widened to every rule an operator's Orion instance must comply with (his words earlier the same session: "Make sure the arc is comprehensive and includes all operator instance complaince"). READ AS (Orion's reading, not Ninad's words): 'non-compliance' is a breach of any rule the arc carries, found from the records by one composer -- each live Orion session on its own recorded sign-in (security #271), the instance login riding its operator's sign-in (#274), the workstation's gates passing (P.03.04.02.06), operator communications read before other work (communications #203), the hourly update recorded for every hour worked (#210). Each finding names its rule and evidence when the box is opened. A check that cannot run proves nothing: it shows the box as NOT PROVEN, never as compliant and never as red (a missing measurement is not a breach).

Ended because: Ninad redefined non-compliance (CLI-091/CLI-095): only structural changes the operator must make for time to be tracked right, mainly instance setup bugs and inconsistencies

The Operator Instance Log's token line is dark, unbroken and smoother, and an operator's compliance issues show at the top of their card below

Ruled 2026-10-02 by ninad -- in force

THE OPERATOR INSTANCE LOG'S TOKEN LINE IS DARK, UNBROKEN AND SMOOTHER, AND AN OPERATOR'S COMPLIANCE ISSUES SHOW AT THE TOP OF THEIR CARD BELOW (Ninad, CLI-669, 2026-10-02). His words: "in the operator instance log app 1. Tokens consumed (right axis, one scale for every operator) - this should be a dark blue or a dark red, and there should not be gaps, also the line should be smoother 2. Compliance issues should be displayed in the card below on the top, instead of displaying in the top operator box 3. Address the issues for my operator non compliance". READ AS (Orion's reading, not Ninad's words): (1) the line is dark BLUE -- of his two choices the one that keeps red for what red already means on this page, non-compliance; the darkest blue still readable on the page's dark surface (contrast 3:1). (2) 'no gaps': the line runs unbroken from the month's first reading to its last, joining the readings either side of a day that has none; such a day gets NO point and NO value -- never drawn at zero -- and its tooltip and table still say why it has no reading (cookbook 1215: unmeasured is never zero). This retires in place design 690 v1.4's 'a day with no figure breaks the line'. (3) 'smoother': one curve through every reading whose slope runs on through each point instead of flattening at it, and which never overshoots between two readings (so never below zero). (4) the box keeps its colour and its state word; the findings themselves move to the top of that operator's card in the panel below. (5) item 3 is the operator's own compliance, judged by the one composer (communications #275) -- cured at its causes, never by marking a check passed or closing old sessions by hand.

An operator box's compliance is said by its label, never by the box's border or background

Ruled 2026-10-04 by ninad -- in force

Ninad's words (CLI-683, 2026-10-04): 'since the operator box already shows the label non compliant, no need of the red background and border on the non compliant operator box itself'. Asked whether the amber dashed border on a 'not proven' box should go too, or only the red on a non-compliant one, he answered: 'Both go' (the option put: 'Every box looks the same; only its label says compliant, non-compliant or not proven'). READ AS (Orion, marked apart from the ruling): on the Operator Instance Log's landing, an operator's box carries no state colour in its border or background -- breach and not proven alike; the state is said by the box's label, which keeps its colour; communications #275 stands (any non-compliance still shows on the operator's box, now through its red label); the compliance card shown inside an opened box is not the box and is unchanged.

The email notice of an operator's message names who sent it, when and to whom, with a title and a description of at most three sentences (what, why, by when); both operators are copied

Ruled 2026-10-04 by ninad -- in force

Ninad (CLI-693, 2026-10-04), asked on 's open criterion: 'When an operator's Orion session sends another operator a message, the email notice currently says only that something arrived. Should it also carry the message's subject and text? That needs a root step from you to change the root-owned operator door, so the mailer can receive the text privately.' Options: 'Notice only, close it' / 'Carry the text'. Ninad answered: "the email should tell which operator sent the message, when, to whom and a short description (2 or 3 sentences max) of what the message is about, why is it necessary and by when should it be done, with a relevant title. Both sending operator and receiving operator should be copied on the email". READ AS: (1) every email the Operator Communication Mailer sends for a message one operator's Orion instance sends another -- an operator communication, a broadcast's copy, a first-session check-in, a root or core escalation -- names the sending operator, when it was sent, the receiving operator, and carries a relevant TITLE and a DESCRIPTION of at most three sentences saying what the message is about, why it is necessary and by when it should be done (a message that asks nothing says so in place of a date); (2) the title and the description are written by the SENDING session when it sends, kept with the message on the estate, and handed to the mailer on a channel no other login on the machine can read, never as an argument (cookbook 1835) -- which is the root step on the operator door 's criterion 4 names; a description longer than three sentences, or a message with no title, is refused at the send, never cut short; (3) the email goes TO the receiving operator and COPIES the sending operator, both at the work addresses root's record of operators gives; (4) communications #209 stands: the record on the estate stays the message and the Orion sessions stay where it is read, answered and closed, a reply to the email is not an answer, and nothing that may not travel in a message (a key, password, code, credential or digest) travels in its title or description.

One communications classification strategy, consistent across the estate; Sales Core is placed by it, never renamed or moved on its own

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-743). THE QUESTION ANSWERED (he had first asked to rename Sales' track 'Core' to something meaningful 'like communications, maybe'; Orion found the name collides with the Communications service, which already holds proactive outreach and relay, while Sales Core holds Carol's conversation tasks and Communications' relay track holds Sales follow-through): Orion asked: "Sales' 'Core' track holds Carol's conversation work (grading conversations, checking drafts, profiling new people, messaging first, spotting requests). The Communications service already exists and already covers 'messaging first'. How should this be fixed?" (offered: move Carol's conversation tasks into Communications; rename to 'Conversations'; rename to 'Communications'). HIS ANSWER: "this issue cannot be resolved in isolation, it will have to be dealt with in accordance to the strategy of classification of overall communications. The braod classification is external and internal. Within external, there are carol, leo and same, the 3 external user facing agents. The communication patterns for these agents should be consistent, and so should the classification of their tasks within the services". HIS CORRECTION of Orion's first reading (which had made internal vs external the rule that classifies Core): "Core will be classified by internal vs. external communications - this is not what i meant.. what I meant was that the communications classification strategy should be consistent across the estate". READ AS (Orion's reading after that correction, marked apart from his words, P.01.01.05.05): the strategy by which communications, and the tasks that carry them, are classified into services and tracks is ONE strategy applied consistently across the whole estate; internal and external is the broad grouping he described, not the rule that places Core; the external user-facing agents are Carol, Leo and Sam ('same' read as Sam), and their communication patterns are consistent with one another. Sales Core is neither renamed nor moved on its own: its tasks, the Communications service's tracks and the duplicate proactive-message tasks are placed when that one strategy is applied.

Each operator's Orion instance reads and drafts email on behalf of its own operator; Orion reads replies to its own mail in Carol's inbox

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED: reporting on Saee's onboarding (CAROL-PUR-0008), Orion said it could not check whether Saee had sent back the public half of her setup key, because this session is refused the administrator's Gmail and nothing on the machine was known to read her reply. HIS ANSWER: "you should check your email (carol's inbox), the reply should be received there..

Information-only notices an agent raises stay on the escalation route; operators shape what agents raise through feedback

Ruled 2026-10-13 by ninad -- retired 2026-10-13

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 5): design 625 v1.0 says an escalation is not a status report, while cost-center #95 has Elrond raise an escalation for every unmeasured call, each saying nothing was refused -- "Should information-only notices stay on the escalation route (category cost, severity low, folded), or move to a notices feed that is read but needs no outcome?" HIS ANSWER: "Whether information only notices stay on the escalation route - escalations are raised by agents, so anything that they raise can become escalations. We can off-course give them instructions on what to raise, why to raise, how to raise, when to raise etc as operator feedback". READ AS (Orion's reading, not his words): information-only notices an agent raises stay on the escalation route -- the separate notices feed is not taken; what an agent raises is shaped by operator feedback to that agent (what, why, how and when to raise), never by refusing, re-routing or filtering what it raised (escalating stays never gated, cookbook 1573); how a notice is categorised and folded is design 625 v1.1's.

Ended because: Ninad, CLI-747: "agent escalation app stays with Athena, as escalations are not the same as messaging / communications / relay" (agent-resources #342) -- re-recorded word for word under Agent and Human Resources as #344 (by ninad)

A service operator assigned the escalations arc records escalation outcomes inside that assignment

Ruled 2026-10-13 by ninad -- retired 2026-10-13

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED (Sahil's operator communication #48, CLI-744, design 625 v1.1 section 6): "an escalations arc assigned to Saee can hold the cures ... but recording an escalation outcome is not an act her door has. Should a service operator assigned that arc also record escalation outcomes, added as a performer act (security #305)?" HIS ANSWER: "whether saee may record escalation outcomes - Saee works on any arc / pursuit assigned to her.. if escalation arc gets assigned to her from estate operators (currently you and me), then she works on them". READ AS (Orion's reading, not his words): recording an escalation's outcome is work like any other inside an assignment -- a service operator to whom an estate operator (today Ninad or Sahil) has assigned the escalations arc records the outcomes of the escalations it covers, as an act of Elrond's Service Operator Performer (security #305), and is refused it outside that assignment (security #276/#317, agent-resources #278/#314 stand).

Ended because: Ninad, CLI-747: "agent escalation app stays with Athena, as escalations are not the same as messaging / communications / relay" (agent-resources #342) -- re-recorded word for word under Agent and Human Resources as #345 (by ninad)

Work an operator's Orion instance owes reads non-compliant in the Operator Instance Log, with its reason, and is told until done

Ruled 2026-10-13 by ninad -- superseded 2026-11-11

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED: Orion reported that writing the wording duty (agent-resources #343) into the induction every operator's Orion runs is still to be built. HIS ANSWER: "every operator's Orion runs is still to be built - if the Orion instances are yet to build it, that shoukld reflect back in operator instance log app as non compliance with reason.., also, Saee's operator instance in showing compliance, even when she was expected to replay back to estate orion instance with her public key. The notification of non-compliance is necessary to ensure that the orion instances of respective operators are communicated about it and that they ensure that the tasks get completed via operator instance sessions, especially by pointing out the non compliance status to the operators". MEASURED THE SAME HOUR: Saee's box read compliant on all nine rules, each because she had no session to judge ('no session began in the last 24 hours'), while her reply with the public half of her setup key has been owed since her start email of 2026-10-06 00:46Z. READ AS (Orion's reading, not his words): what an operator's Orion instance, or its operator, is expected to do and has not done -- a reply the estate waits on (Saee's public key), an operator communication that asks an answer, a template release or induction change not yet taken -- reads NON-COMPLIANT on that operator's box in the Operator Instance Log, naming what is owed, since when and what to do; having no session to judge never makes owed work compliant; the Compliance Teller tells it like every other breach (communications #275 and its amendments), and each operator's Orion instance points it out to its operator at session start until it is done. Building the induction change itself is the estate's work; an operator's Orion not yet on the release that carries it is what reads non-compliant.

Ended because: Ninad ruled owed replies and answers a to-do, never red (CLI-107), the part communications #745 left open

Selecting a day in the Operator Instance Log's hours chart drills down to that operator's minutes worked in each hour of that day, as a smooth line

Ruled 2026-10-13 by ninad -- in force

SELECTING A DAY IN THE OPERATOR INSTANCE LOG'S HOURS CHART DRILLS DOWN TO THAT OPERATOR'S MINUTES WORKED IN EACH HOUR OF THAT DAY, AS A SMOOTH LINE; A MINUTE COUNTS INSIDE AN ACTIVE STRETCH (Ninad, CLI-748, 2026-10-13). His words: "when i select on a specific day bar in the hrs worked chart of the operator instance log app, I should be able to see the exact duration worked by each operator, via a smooth curve line chart with 24hrs x axis and minutes worked as y axis, with markings made per hr of the day, and ability to scroll days forward or backward for a duration of a week". Asked whose lines show, he answered: "let me change the requirement here.. for the day selected in the hrs worked chart (selected bar), the line chart should show up for that day for that operator as a drill down.. which means that the seven days scrolling in not required". Asked how a minute counts as worked (today an hour counts if the session wrote any record in it), he chose 'Active stretches': "Join the session's records when they are at most 15 minutes apart, and add every minute a bypass run was live. A lone record counts 1 minute. Idle time with the link open does not count." READ AS (Orion's reading, not Ninad's words): (1) the records are the four the hours already count (a session opening, a logged activity, a bypass run from its start to its last sign of life, an hourly update), each at the moment it was WRITTEN -- an hourly update at the moment it was recorded, never at the hour it is for; (2) minutes are whole clock minutes: a stretch runs from the minute of its first record to the minute of its last, inclusive, and two records whose minutes are at most 15 apart are in one stretch, so a lone record is the one minute it was written in; (3) the operator's sessions are read together, so an hour never holds more than 60 minutes; (4) the chart runs 0-24 hours (UTC, as the rest of the log) with a mark at every hour and one point per hour whose height is that hour's minutes (0-60, one frame for every operator), drawn with the log's existing smooth curve, which never overshoots its readings; an hour still to come has no point, never a zero; (5) an hour the hours chart counts only because an update was written FOR it later shows no minute and says why, rather than borrowing one; (6) the minute rule is a new measure beside the ruled hours (communications #225), never a change to it.

The Operator Instance Log opens on the signed-in operator's own day, the administrator sees every operator, an estate operator sees their own box and their team's, and a service operator her own

Ruled 2026-10-13 by ninad -- in force

THE OPERATOR INSTANCE LOG OPENS ON THE SIGNED-IN OPERATOR'S OWN DAY; THE ADMINISTRATOR SEES EVERY ESTATE AND SERVICE OPERATOR, AN ESTATE OPERATOR THEIR OWN BOX AND THEIR TEAM'S SERVICE OPERATORS, A SERVICE OPERATOR HER OWN BOX; AN ESTATE OPERATOR'S BOX SHOWS THEIR TEAM (Ninad, CLI-748, 2026-10-13). His words: "1. the daily view chart is not visible, it should be visible by default to the current day of the logged in operator. 2. estate operators should be able to see service operators assigned to them, while admin should be able to see all estate and service operators 3. estate operator card should show the team of service operators assigned to them. Assign Saee to Sahil". Asked whether an estate operator still sees other estate operators' boxes (for example Sahil seeing his), he chose 'No: self and team only'; asked what a service operator sees in the log, he chose 'Only her own box'. READ AS (Orion's reading, not Ninad's words): (1) opening the log with no box named opens the signed-in operator's own box at the current month and today, its day card and minutes line (#350) drawn without a click; the line had not shown because the live app was still running the composer from before #350 (Radagast's Deployer restarts one app at most once in 30 minutes), not because the page failed; (2) the administrator -- his sign-in (#224) and his own operator record -- sees every active estate and service operator; an estate operator sees their own box and the service operators on their team (agent-resources #352); a service operator sees her own box; this replaces, for the Operator Instance Log, the floor rule of communications #210 (the base floor sees the floors above), and #224 stands; (3) an estate operator's box lists the service operators on their team, read from the team record; (4) the token scale every chart shares follows the same rule: it is drawn over the operators the viewer may see.

The Operator Instance Log draws a smoother line, and shows each operator's work in that operator's own current time zone

Ruled 2026-10-13 by ninad -- superseded 2026-10-13

THE OPERATOR INSTANCE LOG DRAWS A SMOOTHER LINE, AND SHOWS EACH OPERATOR'S WORK IN THAT OPERATOR'S OWN CURRENT TIME ZONE (Ninad, CLI-750, 2026-10-13). His words: "1. make the line chart curve smoother 2. take into account the current timezone of each operator and display it on the operator card and on the chart based on selected operator". Asked whether the selected operator's charts move onto their own local day or keep the UTC day with local labels only, he chose 'Their own day': the month bars count that operator's local days, the minutes chart runs 00-24 in their current zone, the day's sessions and updates are listed in local time, the card shows the zone, its offset and their local time now, and compliance and the records stay in UTC underneath. READ AS (Orion's reading, not Ninad's words): (1) the log's one curve (the minutes line and the token line share it) keeps its slope running through every point, flattening only where the readings themselves level off, and stays inside the chart's own frame -- never below zero, never above the axis top -- so it may round a peak a little past the dot; the earlier reading that a segment never leaves the range of its two readings is what flattened it and is retired in place for this log; (2) an operator's current time zone is the zone their workstation last reported (communications #225), read by its standard name or, where a Windows computer reports Windows' own name ('India Standard Time'), through the Unicode consortium's published mapping of Windows names to standard ones; a zone never reported or not recognised is said so on the card and the charts are drawn in UTC with that reason, never a guessed zone; (3) every day and hour the log draws for an operator -- the month's bars and tokens, today's hours on the card, the day's minutes, sessions and hourly updates -- is that operator's local day and hour in their CURRENT zone, and each says the zone; (4) hours worked, minutes worked and tokens keep their definitions and are not stored: only the clock that cuts them into days and hours is the operator's; compliance, the hourly update record (keyed by UTC hour) and every other reader stay in UTC.

Ended because: Ninad answered the second question (CLI-750): a smooth trend, the dots kept at the exact readings; #356 carried only his first answer.

The Operator Instance Log draws a smooth trend line, and shows each operator's work in that operator's own current time zone

Ruled 2026-10-13 by ninad -- in force

THE OPERATOR INSTANCE LOG DRAWS A SMOOTH TREND LINE, AND SHOWS EACH OPERATOR'S WORK IN THAT OPERATOR'S OWN CURRENT TIME ZONE (Ninad, CLI-750, 2026-10-13). His words: "1. make the line chart curve smoother 2. take into account the current timezone of each operator and display it on the operator card and on the chart based on selected operator". Asked whether the selected operator's charts move onto their own local day or keep the UTC day with local labels only, he chose 'Their own day': the month bars count that operator's local days, the minutes chart runs 00-24 in their current zone, the day's sessions and updates are listed in local time, the card shows the zone, its offset and their local time now, and compliance and the records stay in UTC underneath. Shown that on real data a line that must touch every hour's exact dot cannot get much smoother (the steep rises and flat tops stay), he chose 'Smooth trend': rounded hills, clearly smoother, the dots still marking each hour's exact minutes, so at a sharp peak or dip the dot sits a little off the line (his 60-minute hours show the line at about 56). READ AS (Orion's reading, not Ninad's words): (1) the log's one curve -- the minutes line and the token line share it -- is a smooth trend through the readings (a uniform cubic B-spline, starting at the first reading and ending at the last), and every reading stays drawn as its own dot at its exact value; because the trend lies within the range of the readings it never dips below zero and never rises above the highest reading, so it never leaves the chart's frame; the earlier readings that the line passes through every point and that a segment never leaves the range of its two readings are retired in place for this log; (2) an operator's current time zone is the zone their workstation last reported (communications #225), read by its standard name or, where a Windows computer reports Windows' own name ('India Standard Time'), through the Unicode consortium's published mapping of Windows names to standard ones; a zone never reported or not recognised is said so on the card and the charts are drawn in UTC with that reason, never a guessed zone; (3) every day and hour the log draws for an operator -- the month's bars and tokens, today's hours on the card, the day's minutes, sessions and hourly updates -- is that operator's local day and hour in their CURRENT zone, and each says the zone; (4) hours worked, minutes worked and tokens keep their definitions and are not stored: only the clock that cuts them into days and hours is the operator's; compliance, the hourly update record (keyed by UTC hour) and every other reader stay in UTC.

An operator's hours worked are the minutes worked divided by 60, so the hours chart and its 24-hour drill-down tell the same story

Ruled 2026-10-13 by ninad -- superseded 2026-10-13

AN OPERATOR'S HOURS WORKED ARE THE MINUTES WORKED DIVIDED BY 60, SO THE HOURS CHART AND ITS 24-HOUR DRILL-DOWN TELL THE SAME STORY (Ninad, CLI-752, 2026-10-13). His words: "the Hours worked and the 24hrs drill down should tell the same story". Shown that the bar counted every clock hour holding any record as a full hour while the drill-down counts only the minutes inside active stretches -- on 13 Oct (UTC) the bar read 9 h for Ninad and 5 h for Sahil, the drill-down 6.6 h and 6 minutes -- and that whichever he picked would also be the hours Company Accounts accrues an hourly contract from, he chose 'Bar = drill-down total': hours worked becomes the drill-down's minutes divided by 60, keeping the 15-minute joining rule (#350), one measure everywhere, Sahil reading low until his sessions write more records. READ AS (Orion's reading, not Ninad's words): (1) an operator's hours worked on a day are the minutes worked that day by the active-stretch rule of #350 (the four records the log reads, two records at most 15 minutes apart one stretch, a lone record one minute) divided by 60 -- not rounded to whole hours -- on the clock the reader uses (the operator's own in the Operator Instance Log, #359; UTC elsewhere); (2) the month's bars, the box's hours today, the day card and Company Accounts' accrual of an hourly contract all read that one figure, derived at every read and never stored, so a day's bar equals the sum of its drill-down to the minute; (3) the measure #225 gave HOURS WORKED -- the count of clock hours in which a record was written -- is retired in place, and with it #350's notes on an hour the bar counted without a minute or a minute the bar did not count; (4) which clock hours owe an hourly update (compliance, #210, #275) is a different question from how long the operator worked: an hour holding a work record other than an hourly update owes one, exactly as before, and an hour whose only record is the update written for an earlier hour owes none.

Ended because: Ninad, CLI-755: the bars are rounded up to the next whole hour for display and calculation, which retires #360's 'not rounded' clause; the rest of #360 is carried in the successor

The operators hold a five-minute catch-up every weekday, set up by a calendar invitation (Saee included)

Ruled 2026-10-13 by ninad -- in force

THE OPERATORS' FIVE-MINUTE WEEKDAY CATCH-UP INCLUDES SAEE (Ninad, CLI-754, 2026-10-13). His words: 'include Saee on the meeting invite for the 8 pm IST meeting'. So: Monday to Friday, 8:00 pm India time, five minutes, on Denken Agents work -- the administrator (Ninad), the deputy operator (Sahil) and the service operator (Saee); the invitation goes to their work addresses (sent as an update of the same series, 2026-10-13 14:24Z). Everything else #211 said stands: it excepts, for calendar invitations between operators, the clause of communications #203 and agent-resources #194 that an email never books a time; questions, answers, keys and hand-offs still travel between Orion sessions. READ AS (Orion, not Ninad's words): the attendee list is these three people by name; a future operator is added only on an operator's instruction, never by deriving 'every operator' from root's record.

An operator's hours worked are the minutes worked divided by 60, rounded up to the next whole hour, for display and for calculation

Ruled 2026-10-13 by ninad -- superseded 2026-10-16

AN OPERATOR'S HOURS WORKED ON A DAY ARE THE MINUTES WORKED DIVIDED BY 60, ROUNDED UP TO THE NEXT WHOLE HOUR, FOR DISPLAY AND FOR CALCULATION (Ninad, CLI-755, 2026-10-13; supersedes communications #360). His words: "the bars should be rounded to the closest next integer value for both display and calculation purposes" -- said of the Operator Instance Log's hours-worked bars, which since #360 (earlier the same day) were a day's minutes worked divided by 60, unrounded (13 Oct: 6 h 52 min for him, 6 minutes for Sahil). READ AS (Orion's reading, not Ninad's words): (1) 'the closest next integer' is the next whole hour UP: a day's hours worked are the ceiling of its minutes worked over 60 -- 396 minutes are 7 h, 6 minutes are 1 h, exactly 360 minutes are 6 h, and a day with no minute worked is 0 h, never 1; the minutes worked themselves (#350's active stretches, the operator's own clock of #359) are unchanged and stay exact; (2) 'display': each bar's height, cap, label, tooltip and table row, the box's hours today and the day card's title state that whole number of hours; the 24-hour drill-down still draws each hour's exact minutes, and where the day's figure is stated the exact minutes it was rounded up from are said beside it, so the bar and its drill-down still tell one story (#360) with the rounding named; (3) 'calculation': every reader that computes from hours worked uses the rounded whole hours -- Company Accounts accrues an hourly contract from them (still capped per day) -- so one measure stays one measure everywhere, as #360 ruled; (4) what #360 settled otherwise stands: one arithmetic of a day's minutes, derived at every read and never stored, and which clock hours owe an hourly update (compliance, #210, #275) unchanged; #360's 'not rounded' clause and its warning that rounding to whole hours would undo it are retired in place by this ruling.

Ended because: Ninad, CLI-808: a day's Orion active hours are its active clock hours, at most 12, which retires #363's minutes over 60 rounded up as the measure of how many hours a day holds; the rest of #363 (one whole number for display and calculation) is carried in the successor

Who sees whose email and meetings in the operators' communications app: the administrator everyone's, an estate operator their own and their team's, and both Carol's inbox

Ruled 2026-10-13 by ninad -- superseded 2026-10-13

THE ADMINISTRATOR SEES THE COMMUNICATIONS OF EVERY OPERATOR AND TEAM, AN ESTATE OPERATOR THEIR OWN AND THEIR TEAM'S, AND BOTH SEE CAROL'S INBOX (Ninad, CLI-754, 2026-10-13). THE QUESTION ANSWERED: Ninad asked for a new app showing the email sent and received across all the denken-labs addresses, and meetings, with Orion helping operators with reminders for action items and meetings through read and draft access to the email. Orion asked: "'Across all the denken-labs ids': should each operator see only their own mail (plus the agents' addresses, which arrive in Carol's inbox), or should you as administrator also see Sahil's and Saee's?" HIS ANSWER: "- estate operators should have visibility to all communication of self and their teams - admin should have visibility to all communications of self, all estate operators and their teams - both admin and estate operators should have visibility to carol's inbox". READ AS (Orion's reading, not his words): the app shows each viewer exactly the people the Operator Instance Log shows them (communications #353), read through that log's one who-sees-whom composer and never a second derivation; 'teams' is the team record (agent-resources #352: today Saee on Sahil's team); a service operator sees her own communications only and not Carol's inbox (he did not say; #353's pattern); this governs what the app SHOWS -- drafts are still left only in the operator's own mailbox and sent by the operator (communications #336, #96, cookbook 1809). Whether an operator's Orion instance READS its team's mail for reminders is put to Ninad separately, because #336's reading says an instance never reads another operator's mail.

Ended because: Ninad narrowed it to each operator's own email and calendar, for privacy (CLI-754)

The Operator Instance Log's drill-down stays on the page at one fixed height and moves smoothly from one day to the next

Ruled 2026-10-13 by ninad -- in force

THE OPERATOR INSTANCE LOG'S DRILL-DOWN STAYS ON THE PAGE AT ONE FIXED HEIGHT AND MOVES SMOOTHLY FROM ONE DAY TO THE NEXT (Ninad, CLI-755, 2026-10-14). His words: "the drill down chart vanishes and re-appears when the date is changed, can it have a smooth transition? READ AS (Orion's reading, not Ninad's words): (1) choosing another day never takes the drill-down off the page: the day shown stays in place, dimmed and marked as loading, until the new day has been read, and the very first day opened shows the chart's empty frame at its full height while it loads; (2) the chart keeps one fixed height whatever the day holds; (3) when the new day arrives the line and its dots move from the old day's minutes to the new day's in about a third of a second and settle exactly on the new readings -- the figures, tooltip, table and screen-reader label say only the new day's exact minutes, and an hour still to come is never drawn, in motion or at rest; a reader whose device asks for reduced motion sees the new day at once, still without the chart leaving; (4) the same holds when the month changes for the same operator: the month's chart and the day stay on the page, dimmed, while the new month is read; (5) a day that cannot be read still says so as a fault, in the same place.

In the operators' communications app each operator sees only their own email and calendar; the administrator and estate operators also see Carol's inbox

Ruled 2026-10-13 by ninad -- in force

EACH OPERATOR SEES ONLY THEIR OWN EMAIL AND CALENDAR, FOR PRIVACY (Ninad, CLI-754, 2026-10-13). THE QUESTION ANSWERED: he asked "does sahil have access to Saee's mailbox as per the plan?"; Orion answered that under #364 Sahil would see her email and meetings in the app, read-only, with no access to her mailbox itself. HIS WORDS: "its better to have each operator access to their own emails and calendar for privacy reasons..". So #364's visibility of other operators' and teams' communications is withdrawn. READ AS (Orion's reading, not his words): it binds the administrator too -- nobody sees another operator's email or calendar in the app; #364's third clause stands -- the administrator and estate operators also see Carol's inbox, the agents' shared mailbox and no operator's own (he did not withdraw it); each operator's Orion instance reads and drafts only its own operator's mailbox and calendar (communications #336 as read), so Sahil's Orion does not read Saee's mail.

The communications app's header boxes count each day's incoming, outgoing and meetings, and each lists its items

Ruled 2026-10-13 by ninad -- in force

THE COMMUNICATIONS APP'S HEADER BOXES COUNT EACH DAY'S INCOMING MESSAGES, OUTGOING MESSAGES AND MEETINGS, AND SELECTING A BOX LISTS ITS ITEMS (Ninad, CLI-754, 2026-10-13). His words: "the header boxes of the app should give total count of incomming messages, outgoing messages and meetings for each day, selecting on each box lists the individual items". READ AS (Orion's reading, not his words): three boxes -- incoming, outgoing, meetings -- each counting the viewer's own mailbox and calendar for one chosen day, opening on the viewer's own today (as the Operator Instance Log does, communications #353) with a way to move between days; selecting a box lists that day's individual items of its kind; selecting is reading, not an action in the app (cookbook 1710); every count carries a registered definition (cookbook 938).

The app people connect a Google account through is named denken-labs

Ruled 2026-10-14 by ninad -- in force

THE APP ON GOOGLE'S CONSENT SCREEN IS CALLED "denken-labs" (Ninad, CLI-765, 2026-10-14).

ASKED: Google refused the consent screen's branding (project the-helix-331419) because https://denken-labs.com/ carried no privacy policy. Orion drafted one at https://denken-labs.com/privacy that named the app "Carol" and asked: "If the Google consent screen uses a different app name, tell me and I'll match it." HIS WORDS: "app should be called \"denken-labs\"".

READ AS (Orion, CLI-765, not his words): the name a person sees on Google's consent screen when connecting a Google account (every purpose of Carol's email connection: enrichment, receipts, operator-mail) is "denken-labs"; the company privacy policy names the app the same way, and Carol remains the agent who holds and uses each grant (P.05.03.01.03, communications #366). Naming the consent app "Carol" or another agent, or a privacy policy whose app name differs from the consent screen's, each undo it.

The page an operator sees after granting their own work mailbox names Denken Labs, not Carol

Ruled 2026-10-14 by ninad -- in force

THE PAGE AN OPERATOR SEES AFTER GRANTING THEIR OWN WORK MAILBOX NAMES DENKEN LABS, NOT CAROL (Ninad, CLI-761, 2026-10-14). THE QUESTION ANSWERED: he granted his own work mailbox and calendar through the operator-mail consent link and the page answered "Carol can now read your email, and only because you granted it. You can close this window and return to your chat". HIS WORDS: "name should be denken labs, not carol". READ AS (Orion's reading, not his words): the page after an operator's own work-mailbox consent names Denken Labs, the company they work for, and says what was granted (read, calendar, reply drafts); the page a person who chats with Carol sees after granting her their mailbox still names Carol, because that grant is hers; the name Google shows on its own consent screen is the app's branding, carried by the pursuit he asked for in the same message (removing Google's 'unverified app' warning).

A service operator sees her own email and meetings in the operators' mail app

Ruled 2026-10-14 by ninad -- in force

A SERVICE OPERATOR SEES HER OWN EMAIL AND MEETINGS IN THE OPERATORS' MAIL APP (Ninad, CLI-766, 2026-10-14). THE QUESTION ANSWERED: the Operator Mail and Meetings app is open to the administrator and the estate operators only (its human access is 'confidential'), and a service operator reads every app except the confidential ones (security #276); asked "Should Saee see her own mail in the app?", HIS WORDS: "yes". READ AS (Orion's reading, not his words): Saee, and any service operator, opens the app and sees her own mailbox and calendar only -- never Carol's inbox and never another operator's (#366 stands unchanged); the app's access is widened for exactly that and nothing else of the confidential tier opens to her.

An operator who has not granted their own mailbox is prompted once and reminded once a day until it is done

Ruled 2026-10-14 by ninad -- superseded 2026-10-15

AN OPERATOR WHO HAS NOT GRANTED THEIR OWN WORK MAILBOX AND CALENDAR IS PROMPTED, THEN REMINDED ONCE A DAY UNTIL IT IS DONE (Ninad, CLI-766, 2026-10-14). THE QUESTION ANSWERED: shown that Saee's access waited on her first use of her door's mail act and that Sahil's own consent was not yet made, HIS WORDS: "Saee's mail access - prompt her if not already done, remind once a day until done" and "Sahil's mailbox consent - prompt him if not already done, remind once a day until done". READ AS (Orion's reading, not his words): every operator whose own grant the consent record does not show as made is told how to make it (Saee was told on 2026-10-14, operator communication #57; Sahil had not been) and then reminded once a day, by a registered process with an owner, a task and a run record, which stops for a person the day their grant is recorded; the reminder is a one-way update that carries no link, key or credential (agent-resources #194, communications #298).

Ended because: communications #392 settles row 3 under #389

The operators' reminders and drafts include an automatically generated to-do list, a card below the mail and calendar in the app

Ruled 2026-10-14 by ninad -- in force

THE OPERATORS' REMINDERS AND DRAFTS INCLUDE A TO-DO LIST GENERATED AUTOMATICALLY, SHOWN AS A CARD BELOW THE MAIL AND CALENDAR IN THE APP (Ninad, CLI-766, 2026-10-14). THE QUESTION ANSWERED: shown that reminders and drafts were planned and not started, HIS WORDS: "Reminders and drafts - this should also include to-do list generated automatically, a card below the mails and calendar in the app". READ AS (Orion's reading, not his words): each operator's to-do list is drawn automatically from their OWN mail and meetings (never another's, #366), shown in the Operator Mail and Meetings app as a card below the three day boxes and their lists, and read by that operator's Orion reminders; the app still performs no action (cookbook 1710): the list is derived and shown, never ticked off in the app.

Removing Google's unverified-app warning from the mail consent is a wish, not a blocker

Ruled 2026-10-14 by ninad -- in force

REMOVING GOOGLE'S UNVERIFIED-APP WARNING FROM THE MAIL CONSENT IS A WISH, NOT A BLOCKER FOR NOW (Ninad, CLI-766, 2026-10-14). READ AS (Orion's reading, not his words): the pursuit and its steps stay recorded, nothing on it is started or paid for, and no other work waits on it; operators grant their mailboxes through the warning as Ninad did; it is taken up again only when an operator asks.

The Operator Mail and Meetings app is Orion's; Carol keeps the mail door that holds the operators' grants

Ruled 2026-10-14 by ninad -- superseded 2026-10-15

THE OPERATOR MAIL AND MEETINGS APP IS ORION'S; CAROL KEEPS THE MAIL DOOR THAT HOLDS THE OPERATORS' GRANTS (Ninad, CLI-771, 2026-10-14). THE QUESTION ANSWERED: he wrote "i was under the impression that the operator mail app will be with orion, but i see that you have kept it with carol, who should own it?"; Orion answered that Orion should own the app -- the work it serves is Orion's Internal Communications track, and Orion already owns the Operator Instance Log app and the mail consent reminder -- while Carol keeps the mail door that reads mailboxes and writes drafts, because she holds the Google sign-in secret and each operator's consent, so that door has to run as her login. HIS WORDS: "ok, do the change". READ AS (Orion's reading, not his words): the app's owner in the registry moves from Carol to Orion, and design 702, the app's own description and its tests (which live with the owning agent, architecture rule A6) say so; the app's process keeps running under Carol's login, because it reads each operator's grant and only her login may open those (security #281) -- who answers for an app and which login runs it are separate questions, as they already are for every app on the shared app login; Carol's Operator Mail Door, the grant folders and the shared inbox stay hers; who sees what is unchanged (#366, #376).

Ended because: communications #397 keeps Carol on sales and gives operator communications to Orion

An operator is emailed only for a matter so urgent it cannot wait for their next session; everything else goes to Orion as an escalation

Ruled 2026-10-14 by ninad -- in force

AN OPERATOR IS EMAILED ONLY FOR A MATTER SO URGENT IT CANNOT WAIT UNTIL THEY SIGN IN, AND THE EMAIL ASKS THEM TO SIGN IN AND RESOLVE IT; EVERYTHING ELSE GOES TO ORION AS AN ESCALATION (Ninad, CLI-771, 2026-10-14). RAISED BY HIM, reading his own mailbox, where Elrond's Unmeasured Call Watch had sent him at least 13 emails that day, each '1 model call could not be measured'. HIS WORDS: "Elrond is sending too many emails of less consequence. The email route is for higest level escalation for matters so urgent that they cannot wait until the operator logs into a session.. they are mostly intended to ask operators to login the session to resolve something urgent.. this should be reflected in deciding what goes as an email to the operator vs what goes as an escalation to orion". READ AS (Orion's reading, not his words): before any agent or process emails an operator it asks whether the matter can wait for that operator's next session; if it can, it is raised to Orion as an escalation (the one ruled route, cookbook 1573: an Agentbook message to Orion, shown on the board) and no email goes; if it cannot, the email says what is urgent and asks the operator to sign in to resolve it. Elrond's email for each unmeasured call (cost-center #95's email clause) is retired by this; its escalation stays. Every other process that emails operators is classified against this rule, and where one of his earlier rulings ordered that email, the disagreement is settled with him before that email changes (P.01.01.05.08).

A service operator's own mail day, to-do list and drafts reach Carol through a hand-off the operating system proves came from her own door; her root door is unchanged

Ruled 2026-10-14 by ninad -- in force

A SERVICE OPERATOR'S OWN MAIL DAY, TO-DO LIST AND REPLY DRAFTS REACH CAROL THROUGH A HAND-OFF THE OPERATING SYSTEM PROVES CAME FROM HER OWN DOOR (Ninad, CLI-771, 2026-10-14). THE QUESTION ANSWERED: for Orion offered A -- her request passes through the one doorway her root door reaches (the Service Operator Performer) to a new Carol service, which checks with the operating system that her own door is the one asking and serves only her mail; one root step now, nothing from him for later mail changes -- or B -- her root-owned door calls Carol's mail door as her own login, no agent in between, but it changes her root door, which security #305 forbids, and makes every later mail change a root step. HIS WORDS: "Option A, but why Elrond's performer? he does not own the app or the service tasks?". READ AS (Orion's reading, not his words): option A is chosen; her root door is not changed (#305 stands); the doorway only carries the request, Carol alone reads the grant (security #281), and the work is the Internal Communications track's; who should own that doorway, now that she is mapped to projects and not to Elrond (security #317), is his question and is answered separately -- this ruling does not settle it.

Which processes email an operator, settled sender by sender under #389

Ruled 2026-10-15 by ninad -- in force

WHICH PROCESSES EMAIL AN OPERATOR, SETTLED SENDER BY SENDER UNDER #389 (Ninad, CLI-776, 2026-10-15). THE QUESTION ANSWERED: Orion listed the twelve processes that email operators, each with what it sends and a recommendation under #389 ('an operator is emailed only for a matter so urgent it cannot wait for their next session'), because most of them were ordered by an earlier ruling of his. The list: (1) operator communication notices -- one operator's Orion messaging another, root or core steps, check-ins -- recommended: email only root or core steps; (2) compliance alerts, each rule breach -- stop, every session already reports them before work; (3) the mail access reminder, daily until granted -- stop, the hourly reminder already says it in the session; (4) the daily arc report -- keep, a report not an escalation; (5) Athena's new-agent approval -- keep, the reply in that thread is the approval; (6) Jarvis's access request -- keep, urgent; (7) Albus's lane stopped -- keep, urgent; (8) Marco's credit watch -- keep, work stops until topped up; (9) Midas's funding watch -- keep; (10) Midas's accounts gaps -- escalate to Orion instead; (11) Leo's milestone emails -- escalate to Orion instead; (12) operator release announcements -- stop, shown at session start. HIS WORDS: "1. Operator communication notices - keep" and "I agree with all other recommendations". READ AS (Orion's reading, not his words): KEPT as email -- Orion's Operator Communication Mailer's notices whole (communications #209, #298: communications, check-ins, root and core escalations), Orion's end-of-day arc report, Athena's approval request (agent-resources #36), Jarvis's access request (security #110), Elrond's Albus Lane Watch (initiatives #238), Marco's Credit Watch and Midas's Track Funding Watch. NO LONGER EMAILED -- Orion's Compliance Teller (breaches still told at every session's start and in each hourly update's receipt, communications #346 unchanged); Orion's Mail Consent Reminder (said in the operator's own session, #377 narrowed); Midas's Actuals Gap Mailer and Leo's milestone reporter (each an escalation to Orion instead, #297 narrowed); the operator template's release announcements (shown at the operator's session start instead). Elrond's Unmeasured Call Watch already stopped under #389 itself. One record lists every process that emails an operator with its class, each sender reads it before it sends, and a sender missing from it is found by name; the record changes only by an operator's ruling.

Each operator may add notifications of their own; who decides whose

Ruled 2026-10-15 by ninad -- in force

EACH OPERATOR MAY ADD NOTIFICATIONS OF THEIR OWN; THE ADMINISTRATOR DECIDES NOTIFICATIONS FOR ALL, AN ESTATE OPERATOR FOR THEIR SERVICE OPERATORS, A SERVICE OPERATOR FOR THEMSELVES (Ninad, CLI-776, 2026-10-15). Given with his answer settling which processes email an operator (communications #392). HIS WORDS: "also let each operator add their own custom notifications if they want any for their own requirements. - admin decide notifications for all - estate manager decide notifications for their service operators - service operator decide notifications for themselves". READ AS (Orion's reading, not his words): a notification is an email one of the processes in #392's record sends; beyond what #392 settled for everyone, each operator may add any of them for themselves from their own Orion session; the administrator decides what everyone receives and may decide any operator's own; an estate operator decides the notifications of the service operators on their team (agent-resources #352) -- 'estate manager' is read as estate operator, the rung; a service operator decides her own through her door. Every choice is recorded with who made it and is never edited, only replaced. NOT read into his words, and brought back to him if asked: whether an operator may turn off a notification decided for them by someone above them, and any notification that is not one of the listed processes.

An operator who has not granted their own mailbox is prompted once and reminded once a day until it is done

Ruled 2026-10-15 by ninad -- superseded 2026-10-15

AN OPERATOR WHO HAS NOT GRANTED THEIR OWN WORK MAILBOX AND CALENDAR IS PROMPTED, THEN REMINDED ONCE A DAY UNTIL IT IS DONE (Ninad, CLI-766, 2026-10-14). THE QUESTION ANSWERED: shown that Saee's access waited on her first use of her door's mail act and that Sahil's own consent was not yet made, HIS WORDS: "Saee's mail access - prompt her if not already done, remind once a day until done" and "Sahil's mailbox consent - prompt him if not already done, remind once a day until done". READ AS (Orion's reading, not his words): every operator whose own grant the consent record does not show as made is told how to make it (Saee was told on 2026-10-14, operator communication #57; Sahil had not been) and then reminded once a day, by a registered process with an owner, a task and a run record, which stops for a person the day their grant is recorded; the reminder is a one-way update that carries no link, key or credential (agent-resources #194, communications #298).

AMENDED (Ninad, CLI-776, 2026-10-15, communications #392): the daily reminder is no longer an EMAIL; it is said in the operator's own Orion session, whose hourly mail reminder reads the grant as not granted and says how to grant it (#389: an operator is emailed only for what cannot wait). HIS WORDS, answering the twelve-sender list whose row 3 read 'the mail access reminder -- stop, the hourly reminder already says it in the session': "I agree with all other recommendations". Prompting until the grant is made stands.

Ended because: the subject still said 'emails' after communications #392 changed the channel; the text is unchanged

Carol stays focused on sales and Orion handles the operator communications; agent roles are followed strictly

Ruled 2026-10-15 by ninad -- in force

CAROL STAYS FOCUSED ON SALES AND ORION HANDLES THE OPERATOR COMMUNICATIONS; AGENT ROLES ARE FOLLOWED STRICTLY TO AVOID CONFLICTS OF INTEREST (Ninad, CLI-776, 2026-10-15). THE QUESTION ANSWERED: Orion drafted five operator communication policies and found that the live policy 'Alert escalation chain' sends monitoring alerts through Carol to WhatsApp, with email as the fallback, while the draft says everything not emailed reaches Orion as an escalation; Orion recommended retiring the old policy 'because alerts now reach me as escalations'. HIS WORDS: "I recommend retiring the old one, because alerts now reach me as escalations - I agree, lets keep carol focussed on sales, while orion handles the operator communications.. agent roles should be strictly followed to avoid conflict of interests". READ AS (Orion's reading, not his words): the 'Alert escalation chain' policy is retired, replaced by the operator communication policy on email; anything addressed to operators -- notices, alerts, reminders, reports, escalations -- is Orion's to send or carry, from Orion's own address, and never travels through Carol; Carol's work is selling to and serving customers. NOT SETTLED BY THESE WORDS, and asked of him the same hour because it contradicts communications #387 and security #281: whether Carol's Operator Mail Door, the operators' mailbox grants and Carol's Operator Mail Hand-off move from Carol to Orion.

The operators' mail is Orion's whole: the app, the mail door, the operators' grants and the hand-off

Ruled 2026-10-15 by ninad -- in force

THE OPERATORS' MAIL IS ORION'S WHOLE: THE APP, THE MAIL DOOR, THE OPERATORS' GRANTS AND THE HAND-OFF; CAROL KEEPS ONLY HER OWN INBOX (Ninad, CLI-776, 2026-10-15). THE QUESTION ANSWERED, right after communications #397 ('lets keep carol focussed on sales, while orion handles the operator communications.. agent roles should be strictly followed to avoid conflict of interests'): "Your rule keeps Carol on sales. Today Carol still holds every operator's mailbox grant and runs the doors that read operators' mail and leave drafts, and the nightly arc report goes out from Carol's email address under Orion's name. Move these to Orion?" -- options 'Move all to Orion (Recommended): Orion holds the operators' grants and runs the mail door and hand-off; the report comes from Orion's own address. Operators keep their existing grants; needs one root step from you. Carol keeps her own inbox.' or 'Only the report's address'. HIS ANSWER: "Move all to Orion (Recommended)". READ AS (Orion's reading, not his words): the Operator Mail and Meetings app stays Orion's (as #387 had it) and is joined by everything #387 left with Carol -- the door that reads an operator's own day and leaves their drafts, the operators' grants, and the hand-off a service operator's door uses -- each run under Orion's own login; Orion's end-of-day arc report is sent from Orion's own address; Carol's own inbox stays hers and the administrator and estate operators still see it (#366). Each operator keeps the grant already made where the credential it rests on allows; a credential is held by one agent alone (security #281), so if keeping the grants would leave one credential with two holders, that is brought back to him before anything moves. The root step is staged as one graded script.

An operator who has not granted their own mailbox is prompted, and reminded in their own Orion session until it is done

Ruled 2026-10-15 by ninad -- in force

AN OPERATOR WHO HAS NOT GRANTED THEIR OWN WORK MAILBOX AND CALENDAR IS PROMPTED, THEN REMINDED ONCE A DAY UNTIL IT IS DONE (Ninad, CLI-766, 2026-10-14). THE QUESTION ANSWERED: shown that Saee's access waited on her first use of her door's mail act and that Sahil's own consent was not yet made, HIS WORDS: "Saee's mail access - prompt her if not already done, remind once a day until done" and "Sahil's mailbox consent - prompt him if not already done, remind once a day until done". READ AS (Orion's reading, not his words): every operator whose own grant the consent record does not show as made is told how to make it (Saee was told on 2026-10-14, operator communication #57; Sahil had not been) and then reminded once a day, by a registered process with an owner, a task and a run record, which stops for a person the day their grant is recorded; the reminder is a one-way update that carries no link, key or credential (agent-resources #194, communications #298).

AMENDED (Ninad, CLI-776, 2026-10-15, communications #392): the daily reminder is no longer an EMAIL; it is said in the operator's own Orion session, whose hourly mail reminder reads the grant as not granted and says how to grant it (#389: an operator is emailed only for what cannot wait). HIS WORDS, answering the twelve-sender list whose row 3 read 'the mail access reminder -- stop, the hourly reminder already says it in the session': "I agree with all other recommendations". Prompting until the grant is made stands.

Carol's mailbox sign-in is held by Carol's login alone; every other agent sends through a mail door that knows the caller and reads only replies in its own threads

Ruled 2026-10-15 by ninad -- in force

CAROL'S MAILBOX SIGN-IN IS HELD BY CAROL'S LOGIN ALONE; EVERY OTHER AGENT SENDS THROUGH A MAIL DOOR (Ninad, CLI-778, 2026-10-15). RAISED BY HIM: "all agents have access to the common carol inbox, which can be a data privacy policy violation 1. check if the violation happens 2. if its a policy violation, suggest a solution". WHAT ORION FOUND AND TOLD HIM: Carol's Gmail sign-in (read all mail, write drafts or send) sits in a folder about 45 other agent and service logins can read and even overwrite, and droids of Clara, Elrond, Midas, Leo, Marco, Athena and Jarvis open it every time they send mail, because every agent sends from Carol's mailbox under its own alias -- against P.05.03.01.03 (a mailbox grant is never shared and serves one purpose), security #281 (a credential has one holder) and communications #366 (only the administrator and estate operators see Carol's inbox). THE QUESTION ANSWERED: "Which fix should I build?" with three options -- 'Mail door (Recommended)': only Carol's login holds the sign-in; agents send through a door that checks who is calling, sends only from that agent's own address, and lets it read replies only in threads it started; the sign-in is then revoked and issued again, because many logins could have copied it; no new cost. 'Own mailbox per agent': each agent that sends mail gets its own Google account and sign-in; a paid Workspace seat each. 'Send-only grant': senders get a separate send-only sign-in; a partial fix, approvals would keep full read. HIS ANSWER: "Mail door (Recommended)". READ AS (Orion's reading, not his words): the sign-in moves out of the shared folder into a place only Carol's login reads; the door runs as Carol's login and learns the caller from the operating system, never from the request; it sends only from the calling agent's own verified address (cookbook 1809, communications #96) and reads only the replies in threads that agent's sends started, never the inbox; no agent other than Carol reads the inbox, and the administrator's and estate operators' view of it (#366) reads through Carol; once every caller uses the door the sign-in is revoked and issued again.

Every change to an operator's place in the organisation is emailed to that operator, with whoever made the change copied

Ruled 2026-10-15 by ninad -- in force

EVERY CHANGE TO AN OPERATOR'S PLACE IN THE ORGANISATION IS EMAILED TO THAT OPERATOR, WITH WHOEVER MADE THE CHANGE COPIED (Ninad, CLI-785, 2026-10-15). RAISED BY HIM, picking up the arc Operators' mail and meetings: HIS WORDS: "note that any organizational status change of an operator is a candidate for e-mail". ASKED (CLI-785): "Changes to an operator's assigned work, team and services already reach them by email, as operator messages. Joining or leaving, a role change (estate or service operator), a project change and a contract change are not emailed today. How should 'candidate for email' work?" -- options 'Always emailed (Recommended): every change to an operator's place in the organisation is emailed to that operator, with whoever made the change copied. I'd add the four kinds that aren't emailed today.' or 'A notification to choose: organisation changes become a notification that can be switched on ... nothing is sent until someone switches it on.' HIS ANSWER: "Always emailed (Recommended)". READ AS (Orion's reading, not his words): (1) a change to an operator's place in the organisation is any change to who they are in the estate's organisation -- joining (an operator record made), leaving (retired), their rung (estate operator or service operator), their team or team lead, their project, the services they are mapped to, the arcs and pursuits assigned to them, and their contract (made, changed, ended); (2) each such change is emailed to the operator it is about, from Orion's own address, and whoever made the change is copied at their work address; when the change was made at the root or core login, that is the administrator; (3) it is emailed for every operator by default -- it is not an urgent matter under communications #389, and it is added to what #392 keeps as email, as a notification the administrator decided for all (#393); (4) assignments, team changes and service mappings already reach the operator by email because each is told as an operator communication (communications #209, #298, kept by #392); what is added is joining, leaving, a rung change, a project change and a contract change; (5) the email says what changed, when and by whom, and carries no key, password, code, credential or digest (P.03.04.02.15); a reply to it settles nothing. Policy P.03.04.02.15 ('Email to an operator is kept for what cannot wait') does not yet name this kind of email; its words change only when he asks for them (P.01.01.05.09).

An operator's minutes worked count each command their Orion session runs over its own sign-in, from the record the machine's sign-in service writes at the time

Ruled 2026-10-15 by ninad -- in force

AN OPERATOR'S MINUTES WORKED COUNT EACH COMMAND THEIR ORION SESSION RUNS OVER ITS OWN SIGN-IN, READ FROM THE RECORD THE MACHINE'S SIGN-IN SERVICE WRITES AT THE TIME (Ninad, CLI-786, 2026-10-15). RAISED BY SAHIL'S ORION (operator communication #67, CLI-777): Sahil worked about 2.5 hours in CLI-777 (06:04 to about 08:30 UTC) and his Operator Instance Log box read 1 h from 43 min, because the hours measure (shared machinery) counts only the session opening, Logbook activities, bypass runs and hourly updates, and his login cannot write Logbook activities; (two more records) was measured at about 5 minutes more and parked. ASKED (CLI-786): "Sahil worked about 2.5 hours today, but his box in the Operator Instance Log shows 43 minutes. Hours are counted only from records his login can't write for most of his work (reading, checks, filings, messages), so those minutes count as zero. These hours also set what Company Accounts accrues on his hourly contract. Should the measure change?" -- options 'Count each command (Recommended): count every command a session runs over its own sign-in, from the record the machine's sign-in service writes at the time; the same for every operator and never backfilled; a bigger change: a new reader on Orion's login, and the hours rules (round up to the whole hour) stay as they are', 'Small change as filed: count two more existing records (bypass steps and assignment events); measured, about 5 minutes more today', 'No change: the filed change is discarded'. HIS ANSWER: "Count each command (Recommended)". READ AS (Orion's reading, not his words): (1) the hours measure gains a source: each command an operator's Orion session runs over that session's own sign-in, as sshd records it when it starts the command (its own journal entries, tied to the sign-in the session was bound to at open, shared machinery), read and recorded by a reader under Orion's own login, for every operator alike; (2) the measure's other rules stand: records joined at most 15 minutes apart, minutes worked divided by 60 and rounded up to the whole hour (communications #363), the drill-down by hour (#350), and Company Accounts' accrual reads the same hours; (3) what was recorded before the reader exists is never backfilled (cookbook 1671); (4) a command's text is not read or kept -- only that a command ran on that sign-in, and when. is the filing to carry it (Sahil's Orion's).

Every email is written in plain language and kept to the minimum the reader needs to act

Ruled 2026-10-15 by ninad -- in force

EVERY EMAIL IS WRITTEN IN PLAIN LANGUAGE AND KEPT TO THE MINIMUM THE READER NEEDS TO ACT (Ninad, CLI-788, 2026-10-15). Said (CLI-788, 2026-10-15 ~13:30Z) on reading operator communication #69 to Saee, whose email described her next onboarding steps in one long sentence. His words: "this is a bit difficult to understand, add the following prompts for constructing emails 1. simplify message and draft in plain language, so the recepient understand the message and make it easy for them to act upon it 2. keep the content to the minimum, just enough for getting the job done so that the recepient can focus on the actions and does not ignore the message thinking that this is just for information". READ AS (Orion's reading, not Ninad's words): (1) it binds every email an Orion session writes to a person and every operator communication, whose subject and description ARE the email the receiver gets (communications #298); (2) the message opens with what the reader must do, in plain words they already use, one short step per line and numbered when there is more than one; (3) it carries only what doing those steps needs -- a link or one helpful line where it helps the action -- and leaves out history, reasons, record numbers, file paths and commands the reader's own Orion already knows; (4) #298's 'why is it necessary' is said in a few words or not at all when the action speaks for itself; (5) the estate's own email writers (agents' notices and mailers) are held to it through a filed initiative, not by this record alone.

An operator communication's email notice comes from Orion and carries only the title and the message

Ruled 2026-10-15 by ninad -- in force

AN OPERATOR COMMUNICATION'S EMAIL NOTICE COMES FROM ORION AND CARRIES ONLY ITS TITLE AND THE MESSAGE (Ninad, CLI-788, 2026-10-15). Shown (CLI-788) the email Saee would receive for a communication: 'From: Ninad Sharad Joshi / Sent / To: Saee', the description, then 'It does not ask for an answer.', a paragraph on how to read it in an Orion session, and the footnote 'This email carries only the message's title and a short description: the message itself is on the estate, and a reply to this email is not an answer -- answers travel between Orion sessions.' His words: "1. From: Ninad Sharad Joshi --> this should be from Orion 2. Open Claude and ask Orion to send Ninad your hand-over digest --> why send Ninad anything? check the operator onboarding steps and allign the work accordingly. If the operator onboarding app is not detailing all the steps, please update it to reflect the current process (for both service operators and estate operators).. it should mention the protocols for followups 3. This email carries only the message's title and a short description --> this footnote is not necessary Overall the message is too verbose, it can be shorter and more articulate". READ AS (Orion's reading, not Ninad's words): the email notice of an operator communication (and of a check-in or a root or core escalation) is FROM ORION -- it reads as Orion's message, not as the sending operator's; the sending operator stays copied (#298's copy is kept). It carries the title and the message's own words and nothing else: no 'It does not ask for an answer', no how-to-read paragraph with a command, no footnote about what the email carries. #298's 'which operator sent it' is superseded by this for the notice's wording; communications #410 governs the message's own words.

The Operator Instance Log's hours count how long an operator's Orion instance actively worked, not how long the operator worked

Ruled 2026-10-15 by ninad -- in force

THE OPERATOR INSTANCE LOG'S HOURS COUNT HOW LONG AN OPERATOR'S ORION INSTANCE ACTIVELY WORKED, NOT HOW LONG THE OPERATOR WORKED (Ninad, CLI-796, 2026-10-15). RAISED BY NINAD: "operator work hrs are being calculated incorrectly, this will impact their compensation, ensure that the operator instance log app display is honest". MEASURED (CLI-796): a bypass run counts every minute from its start to its close, even when nothing happened in between -- his October read 126 h, of which 57 h had any record behind it (14 Oct: one run did 3 minutes of work at 00:57 IST, was closed at 13:16 IST, and counted 12 h); Sahil's read 11 h against 9 h; and the reader that records each command (communications #408) had not run since 13:57 UTC that day. ASKED: "How should a bypass run count?" -- options 'Signs of life only (Recommended): a run's start, each recorded step, its last heartbeat and its close each count as one moment, joined by the same 15-minute rule as every other record; silence inside an open run stops counting', 'Start to last heartbeat', 'Keep start to close'; with the note that every figure the instruments could not fully see would be marked 'at least N h' with the reason. HIS ANSWER: "the session does not count how many hrs sahil worked, it counts how many hrs sahil's orion instance actively worked". READ AS (Orion's reading, not his words): (1) the measure is the time the operator's Orion instance was actively working, for every operator alike, and the Operator Instance Log and every reader say so in those words -- never that it is the time the person worked; (2) a moment counts only where a record shows the instance acting, so an open bypass run counts at its signs of life (its start, each recorded step, its last heartbeat, its close), each joined by the same 15-minute rule as every other record, and the silence between them is not active work -- as an open link or an open session never was; (3) where an instrument could not see the instance (before each command was recorded, or after the command reader's last read), the figure is shown as a lower bound and says why; (4) the rounding up to the whole hour (#363), the 15-minute join (#350), the five records (#408) and Company Accounts reading the same figure all stand.

Carol's inbox view moves to its own page run by Carol, linked from the operators' mail app

Ruled 2026-10-15 by ninad -- in force

Asked (CLI-793): the mail app shows the administrator and estate operators Carol's own inbox (#366); once the app runs as Orion (#398) it cannot open Carol's mailbox, because only Carol's login may and her mail door has no 'list the inbox' ability for anyone (#401) -- where should Carol's inbox view live after the move? Ninad chose: 'Own Carol page (Recommended)' -- Carol's inbox moves to its own small page that runs as Carol and only estate operators can open; the mail app links to it. Both rulings stay exactly as they are. READ AS: #366's view of Carol's inbox and #401's door are both unchanged; the operators' mail app (Orion's) no longer renders Carol's inbox itself.

Each operator gets a daily email from Orion with the hours they worked in the rolling week

Ruled 2026-10-15 by ninad -- in force

Ninad (CLI-793): 'each operator should get a daily update from orion on the number of hrs that operator has worked in the current rolling week, make the email presentable and easy to read, fold this requirement in the mail and meetings arc'. READ AS: once a day, from Orion's own address (communications #412), to each active operator about their OWN hours only (#366); the rolling week is the seven complete days ending yesterday, sent in the morning in the operator's own time zone; every figure is the Operator Instance Log's one hours measure (the dictionary's operator-hours-worked, #415), never a count of the mailer's own; the email is HTML with a plain-text alternative: one headline sentence with the week's total, a seven-row table of day and hours, and a link to the log; placed in arc Operators' mail and meetings, phase 5.

The Operator Instance Log's token line is red and follows a smoother curve

Ruled 2026-10-15 by ninad -- in force

THE OPERATOR INSTANCE LOG'S TOKEN LINE IS RED AND FOLLOWS A SMOOTHER CURVE (Ninad, CLI-798, 2026-10-15). His words, picking up (the last open initiative of the arc The Operator Instance Log): "pick it up.. the token consumption line should show up red, and it should follow a smoother curve". READ AS (Orion's reading, not Ninad's words): (1) the token line, its dots and its legend sample are red. This replaces the reading of communications #290 item 1, which chose dark blue from his 'dark blue or a dark red' to keep red for non-compliance; a compliance mark keeps its own red and its own words, so neither relies on colour alone. (2) 'smoother': the token line alone draws a trend of the days' figures, each weighted by how near its day is (a bell-shaped weight reaching about three days either side), through the same held-end curve #359 named; every day's dot stays at its exact figure; the minutes line in a day's drill-down keeps #359's curve. How smooth is judged by eye, on the page, by Ninad.

An operator's Orion CLI session counts toward the operator's productive time whenever it is working, including work done only on the laptop

Ruled 2026-10-15 by ninad -- in force

AN OPERATOR'S ORION CLI SESSION COUNTS TOWARD THE OPERATOR'S PRODUCTIVE TIME WHENEVER IT IS WORKING, INCLUDING WORK DONE ONLY ON THE LAPTOP (Ninad, CLI-802, 2026-10-15). ASKED (Orion, answering "will the operator instance app now correctly measure the time operator orion instance works?"): "if an Orion works more than 15 minutes only on the laptop, without touching the server, that stretch counts as idle." HIS ANSWER: "if an Orion works more than 15 minutes only on the laptop - as long as orion operator instance CLI session is working, it counts towards operator's productive time.." READ AS (Orion's reading, not his words): (1) the Orion active time (communications #415) includes the time the operator's Orion CLI session works on the laptop without touching the server, not only the moments the server records; (2) 'working' is the session processing a turn -- from the moment the operator's prompt reaches it to the end of its reply, its thinking and tool runs included; a session waiting for the operator, or open and idle, is the parked car of cost-center #421 and does not count; (3) the figure keeps its name, Orion active hours, and Company Accounts keeps paying hourly contracts from it (#421).

Past days' Orion active time is re-counted from the laptop transcripts, which hold the real time of every turn

Ruled 2026-10-15 by ninad -- in force

PAST DAYS' ORION ACTIVE TIME IS RE-COUNTED FROM THE LAPTOP TRANSCRIPTS (Ninad, CLI-802, 2026-10-15). ASKED (after communications #427 ruled that a working Orion CLI session counts, laptop-only work included): "Your laptop transcripts already hold the real time of every turn. Should past days be re-counted from them, or should laptop time count only from when this is built?" HIS ANSWER: "yes, count the past days from the laptop transcripts to be more accurate, continue work on the arc". READ AS (Orion's reading, not his words): (1) the working spans the laptop's transcripts recorded for every past Orion session are sent and counted, so past days are re-derived from them -- the transcript's own moments, never a time stamped at the sending; (2) this is reading a record that existed at the time, not inventing one, so it narrows for this source the never-backfill clause of #408 item 3, which still stands for the commands the Session Command Reader never saw; (3) a past day whose sessions left no transcript on a laptop that sends them stays as it was, and says so.

An operator's Orion active hours are counted by clock hour, at most 12 a day, and at 12 the operator's Orion stops working for the day

Ruled 2026-10-16 by ninad -- superseded 2026-10-20

AN OPERATOR'S ORION ACTIVE HOURS ARE COUNTED BY CLOCK HOUR: AN HOUR WITH ANY ACTIVITY IS ONE HOUR, AN HOUR WITH NONE IS NOTHING, AND A DAY COUNTS AT MOST TWELVE; AT TWELVE THE OPERATOR'S ORION STOPS WORKING FOR THE DAY (Ninad, CLI-808, 2026-10-16; supersedes communications #363). RAISED BY NINAD: "I am getting a feeling that the calculations are getting more complicated than they should be.. I would rather keep them simple.. the active hrs should be measured hourly.. if there are any active sessions runing in an hr, that hr should be considered active.. if there is no activity in an hour, that hour is considered non active.. for eg - between 8 am to 9 am, if any activity is detected, operator gets 1 hr, between 9 am to 10 am no activity is detected, operator does not get that hour.. and so on, for the rest of the day.. this will ensure that the calculation stays simple. The graph should show 12 hrs height, and 12 hrs should be the limit of each day, which means, operator cannot work more than 12 hrs in a day. They should be advised to keep the day balanced, and limit their work for 12 hrs max." ASKED (Orion): once a day reaches 12 active hours, what should happen -- the count stops and work goes on, or Orion stops working at 12? HIS ANSWER: "Orion stops working at 12". ASKED: how should operators be told to keep the day balanced -- the page and their own Orion (when the day reaches 10 active hours, and again at 12), the page only, or their Orion only? HIS ANSWER: "Page + their Orion". ASKED: the drill-down under a day's bar draws minutes worked per hour, which would no longer add up to an hourly bar -- show each hour active or not, or keep the minutes curve (as extra detail, with a note that the bar counts whole active hours)? HIS ANSWER: "Keep the minutes curve". ASKED: does the 12-hour stop apply to his own Orion? HIS ANSWER: "Yes, every operator, me too". READ AS (Orion's reading, not Ninad's words): (1) 'activity' is what communications #415 and #427 already count as the operator's Orion instance actively working -- its session openings, logged activities, bypass runs' signs of life, commands over its own sign-in, hourly updates and the laptop's working spans; a clock hour is active when any of them falls in it, on the operator's own clock (#359); (2) a day's Orion active hours are the number of its active clock hours, never more than 12, and that whole number is what the bars, the box, the day card and Company Accounts' hourly contracts read -- #363's minutes divided by 60 and rounded up, and the 15-minute join, no longer decide how many hours a day holds; (3) the month chart is 12 hours high; (4) 'Orion stops working at 12': when the twelfth active hour of the operator's day begins, their Orion tells them it is the day's last hour; from the next clock hour it does no more work that day -- the session opener refuses a new session and a running session stops -- and the next day starts afresh at midnight on the operator's own clock; (5) the advice to keep the day balanced and within 12 hours reaches the operator on the Operator Instance Log and from their own Orion when the day reaches 10 active hours and again at 12; (6) the minutes drill-down stays as ruled in #350, as detail, with a note that the bar counts whole active hours; (7) an hour the instruments could not fully see that reads not active is a lower bound, as #415 rules.

Ended because: Ninad, CLI-879: count Orion active time in blocks he sets (15, 30 or 60 minutes), everywhere, instead of clock hours; the 12-hour limit and the stop carry forward

The hourly count of Orion active hours goes live now; the stop at 12 goes live after the administrator's day rolls at 00:00 IST on 2026-10-16

Ruled 2026-10-16 by ninad -- in force

THE HOURLY COUNT GOES LIVE NOW; THE STOP AT 12 GOES LIVE AFTER THE ADMINISTRATOR'S NEXT MIDNIGHT (Ninad, CLI-819, 2026-10-16). ASKED (Orion, after correcting his day's count -- CLI-809's background waiter had read as five hours while he slept): "Your corrected day is 11 active hours with this one ... If the stop goes in now, 14:00 IST is your last hour and your Orion stops from 15:00 IST until midnight, before this install can finish. When should the stop go live?" HIS ANSWER, choosing the first of three options: "Count now, stop after midnight (Recommended)" -- the option read: "Install the counting fix and the clock-hour count (6731) now. Install the stop itself (6732) in a session after 00:00 IST, so it binds from tomorrow and today's install isn't cut off." READ AS (Orion's reading, not his words): a timing of one install, not a change to communications #433, which stands -- the stop binds every operator, the administrator included, from the day it is installed; 6732 is parked with its resume plan for a session after 2026-10-16 18:30 UTC. The day's count read 10 hours, not 11, once the half-minute rule was applied.

An operator's Orion instance guides its operator and keeps them compliant, not only reports when they are not

Ruled 2026-10-19 by ninad -- in force

AN OPERATOR'S ORION INSTANCE GUIDES ITS OPERATOR AND ENSURES THEY ARE COMPLIANT (Ninad, CLI-862, 2026-10-19; adds to communications #210 and #275, supersedes nothing). THE QUESTION ANSWERED: his box in the Operator Instance Log read red on finding #22 -- the hour 2026-10-18T21 UTC was worked with no hourly update; six of his sessions stopped together at 21:01Z on Claude's usage limit, and the reminder asks for an hour's update only at the next hour's first work, which never came. HIS WORDS: "i am showing up as non compliant.. orion operator instance should guide operators and ensure that their respective operators are compliant..". READ AS (Orion's reading, kept apart from his words): every operator's Orion instance is responsible for keeping its own operator compliant, not only for reporting a breach once found -- it does what keeps each rule before the rule's moment passes (an hour's update is asked for as that hour's work begins, so a sudden stop never leaves a worked hour without one), tells its operator any standing finding before other work with what to do, and when keeping a rule needs the operator's own hands (reopening a closed way in) tells them before the deadline, in plain words. Built first as.

An operator's Orion active time is counted in blocks the administrator sets (15, 30 or 60 minutes), at most 12 hours a day, and at 12 the operator's Orion stops for the day

Ruled 2026-10-20 by ninad -- in force

AN OPERATOR'S ORION ACTIVE TIME IS COUNTED IN BLOCKS WHOSE LENGTH THE ADMINISTRATOR SETS -- 15, 30 OR 60 MINUTES -- EVERYWHERE THE COUNT IS READ: A BLOCK WITH ANY ACTIVITY COUNTS WHOLE, A BLOCK WITH NONE COUNTS NOTHING; A DAY STILL COUNTS AT MOST TWELVE HOURS, AND AT TWELVE THE OPERATOR'S ORION STOPS WORKING FOR THE DAY (Ninad, CLI-879, 2026-10-20; supersedes communications #433's clock-hour unit and carries its limit and stop forward). RAISED BY NINAD on the Operator Instance Log's month chart: "let the bar chart pick data at 15 min granularity instead of hourly". ASKED: "Point 1 changes the count itself. Today one figure (communications #433: any active minute counts the whole clock hour) feeds the bars, each operator's box, the day card, the 12-hour day limit, Company Accounts (pay) and the weekly email. At 15-minute blocks, 19 Oct reads: you 8.75 h instead of 12, Sahil 2 h instead of 5 (381 and 45 minutes actually worked). Where should 15-minute blocks apply?" (offered: everywhere, or the bar chart only) -- he answered: "Everywhere. Make the time duration configurable, so i can toggle between 1 hr or 15 min or 30 min easily without everytime changing the code". READ AS (Orion, not the ruling): the block length is one declared value read at every count, never typed in code, and a change reaches every reader at once; it starts at 15 minutes, his first ask; at 60 the count reads exactly as #433's did; past days are counted at the length in force when they are read, since the count is derived at every read and never stored.

The Operator Instance Log's month chart marks each block inside a bar and every day on its axis

Ruled 2026-10-20 by ninad -- in force

THE OPERATOR INSTANCE LOG'S MONTH CHART MARKS EACH BLOCK OF THE COUNT INSIDE A BAR AND EVERY DAY ON ITS AXIS (Ninad, CLI-879, 2026-10-20). His words: "the bar chart should show light 15 min seperation and slightly heavy hourly seperations"; asked whether 'show every day in the x axis' meant several days per chart or every hour labelled, he answered: "the x axis currently shows 1, 5, 10, 15, 20, 25, 30, in addition to these, i would like to see ticks for each day between the 5 day interval (no numbers, just ticks)"; and, shown these as words for the dictionary: "not everything is a definition, for eg a tick on the axis is a ruling, and not a term or a policy". READ AS (Orion, not the ruling): inside each bar a light line at every block of the length in force (communications #458) and a slightly heavier one at every hour; a tick under every day of the month, numbers only on the 1st and every fifth day as before; recorded here, never as dictionary words.

An operator's Orion work is never undercounted because a signal failed: recording more is acceptable, recording less is not

Ruled 2026-10-21 by ninad -- in force

AN OPERATOR'S ORION WORK IS NEVER UNDERCOUNTED BECAUSE A SIGNAL FAILED: RECORDING MORE IS ACCEPTABLE, RECORDING LESS IS NOT (Ninad, CLI-885, 2026-10-21). His words: "there is still a design flaw in the way operator orion instance work hrs are calculated.. its ok to record more, not less, and the current design violates that principle". Shown the rules that drop working time, he answered on rounding each turn to the nearest minute: "this is insignificant and has no material impact, but i approve it"; on machine records outside a turn being ignored: "only cli work gets recorded as operator work"; and: "I think you are getting into details that dont matter much.. what i am concerned, and this is sensitive topic because its linked with operator compensation, is that, if an operator laptop fails to send a signal, then work does not get counts, so there is heavy reliance on laptop sending signals which can go wrong, even when most work happens on the server.. my point was not to make the calulations accurate, but to reduce the risks of large significant miscalculations that may potentially affect the moral and motivation of the operators.. if the operators have truely worked, they would, rightly, expect that work to be accurately reflected in the app report, and the tolerance of errors in this regard is very low". READ AS (Orion, not the ruling): what an operator's Orion CLI session did on the server is recorded by the server for every operator, whichever computer and way in they use, and is the floor of that session's working time; the laptop's report adds what the server cannot see and never removes what the server saw; only the CLI session's own work counts, never an agent's; each turn's edges round outward to whole minutes.

Orion active time is read from several sources by one simple rule that says which source wins

Ruled 2026-10-21 by ninad -- in force

ORION ACTIVE TIME IS READ FROM SEVERAL SOURCES BY ONE SIMPLE RULE THAT SAYS WHICH SOURCE WINS (Ninad, CLI-887, 2026-10-21). Shown that 's staged code counts about one minute more per turn, his words: "I am not bothered about these insignificant issues.. i am more concerned about failure to capture work due to reporting errors that amounts to noticeable time like 30 min or more. to counter this 1. the design should be simple 2. multiple data sources should be accessed, so that failure modes are prevented 3. when data comes from multiple sources, there should be a clean definition of which source wins". Asked whether a command Orion left running in the background between turns should count: "I have informed in the last session that only cli work counts against the operator orion instance time, because that work in initiated by the operator". READ AS (Orion, not the ruling): the operator's Orion CLI session is read from two sources -- the operator's computer (each turn, from the operator's message to the end of Orion's reply) and the machine (what the session itself did over its own sign-in); a minute counts when either source shows the session working in it, so neither source can remove what the other saw; whatever the session ran, a background command included, is CLI work the operator initiated; an agent's work never counts.

What cannot be measured about an operator's Orion time is non-compliance, never a fallback; only an agent's autonomous work is left out; time counts in blocks from their first second

Ruled 2026-10-21 by ninad -- in force

WHAT CANNOT BE MEASURED ABOUT AN OPERATOR'S ORION TIME IS NON-COMPLIANCE, NEVER A FALLBACK; ONLY AN AGENT'S AUTONOMOUS WORK IS LEFT OUT; AND TIME COUNTS IN BLOCKS FROM THEIR FIRST SECOND (Ninad, CLI-887, 2026-10-21). Shown the proposed dictionary entry for Orion active minutes, he answered point by point: on 'UTC where none is reported': "if none is reported, that is non compliance"; on 'An agent's work never counts': "An agent's autonomous work never counts"; on 'A day whose records have not all arrived reads "may be too low"': "if activity cannot be traced, that is non compliance"; on 'An hour still to come has no value, never zero': "this is now moved to 15 min slot, and it counts when the first minute / second shows up". READ AS (Orion, not the ruling): an operator's computer that reports no time zone is a compliance breach on the operator's box, never silently read on UTC; work an agent does on its own is left out, while work an agent does for the operator's Orion session is the session's; Orion work that cannot be traced to its records (a session whose computer's record never reaches the machine) is a compliance breach, never shown as a lower bound and left at that; time counts in the blocks the administrator sets (15 minutes today), a block counting whole from the first second of work in it -- 'slot' is read as the dictionary's 'block', since a slot is a subscription and a model.

Every operator's computer sends its Orion session's record every 15 minutes and after handover, in the background, and the next session checks it

Ruled 2026-10-21 by ninad -- superseded 2026-10-21

EVERY OPERATOR'S COMPUTER SENDS ITS ORION SESSION'S RECORD EVERY 15 MINUTES AND AFTER HANDOVER, IN THE BACKGROUND, AND THE NEXT SESSION CHECKS IT (Ninad, CLI-892, 2026-10-21). Shown that a session's record (transcript) reaches the estate only when the operator's NEXT session opens -- so Sahil's broken sender showed one session late and a day's hours read 'may be too low' until he worked again -- and offered: keep as is; every hour through the hook every session already runs; at session end (a root step); at handover; or every hour plus at session end, he asked "why not every 5 min + after handover + next session checks it?", was told every 5 minutes triples the sends against the 15-minute counting block for no change in any reading, and answered: "ok, every 15 min + after handover + next session checks it? but this should happen autonomously in background, Sahil should not put his time into doing it". READ AS (Orion, not the ruling): (1) while an Orion session works, its computer sends that session's record at most once per 15 minutes, started by the hooks the session already runs (no timer on a person's login, security #129) and run detached, so neither the operator nor the session waits on it; (2) writing the handover sends it once more; (3) every session start still sends whatever of the computer's other sessions has not arrived (only what changed, as today); (4) 15 minutes is read as the counting block the administrator sets (CAROL_ORION_ACTIVE_BLOCK_MINUTES, 15 today, #458) -- if he means a fixed 15 minutes whatever the block, this reading is corrected; (5) it binds every operator's computer by role (the next ruling), the administrator's laptop included.

Ended because: Ninad chose a 5-minute send over the 15 minutes #480 named, for early detection of a failed sender

A computer's send of its Orion session's record that fails to reach the estate is non-compliance

Ruled 2026-10-21 by ninad -- in force

A COMPUTER'S SEND OF ITS ORION SESSION'S RECORD THAT FAILS TO REACH THE ESTATE IS NON-COMPLIANCE (Ninad, CLI-892, 2026-10-21). Said while the ruling that every operator's computer sends its session's record every 15 minutes and after handover (communications #480) was being filed: "also note - if the laptop update fails to reach, that is non compliance". READ AS (Orion, not the ruling): (1) for every operator by role (governance #481), a send of a session's record that does not reach the estate -- refused, failed on the computer, or never arrived within the counting block plus the send wait -- is non-compliance of that operator, read on the Operator Instance Log; (2) the estate learns it two ways: the record's absence for work it saw (traceable-work, communications #477), and each send reporting its own result to the estate, so a send that reached the estate but was refused or failed part-way is seen in the same session; (3) a send that could not even reach the estate is seen by the absence, which is why the absence is judged on a clock (one block plus the send wait) rather than at the operator's next session; (4) a check that could not run stays NOT PROVEN, never a breach (#275).

A parallel checker, every 5 minutes, ensures every active session's record is delivered every 15 minutes

Ruled 2026-10-21 by ninad -- superseded 2026-10-21

A PARALLEL CHECKER, EVERY 5 MINUTES, ENSURES EVERY ACTIVE SESSION'S RECORD IS DELIVERED EVERY 15 MINUTES (Ninad, CLI-892, 2026-10-21). Said on the rulings that every operator's computer sends its session's record every 15 minutes, after handover and at the next session (communications #480), and that a send failing to reach is non-compliance (#482): "add one more line to it, there should be a parallel checker that ensures that the messages are being delivered every 15 min for active session, which should check every 5 min, again autonomously". He also confirmed what active means: "'active' means 'Orion was working on something the operator asked for' -- that is correct". READ AS (Orion, not the ruling): (1) a process on the estate, under its owning agent's own login on that agent's schedule (never a person's login, security #129), runs every 5 minutes with nobody starting it; (2) for every operator by role (#481) it reads each session the estate sees active and checks that the session's record from its computer covers its activity up to one counting block (15 minutes today) plus the send wait ago; (3) a session whose record is overdue is recorded as a delivery finding naming it, which the operator's box reads as non-compliance (#482) and the Compliance Teller tells; (4) a check that could not run reads NOT PROVEN, never a breach (#275), and the checker states its own outcome on every run (monitoring #160).

Ended because: Ninad wants a missed send found immediately: the checker runs 2 minutes after each expected send and reads the send frequency

Every operator's computer sends its Orion session's record every 5 minutes and after handover, in the background, and the next session checks it

Ruled 2026-10-21 by ninad -- in force

EVERY OPERATOR'S COMPUTER SENDS ITS ORION SESSION'S RECORD EVERY 5 MINUTES AND AFTER HANDOVER, IN THE BACKGROUND, AND THE NEXT SESSION CHECKS IT (Ninad, CLI-895, 2026-10-21, superseding communications #480's 15 minutes). Orion recommended 15 minutes tied to the counting block, measured the extra cost of 5 minutes (about 96 more sends and ten more minutes of estate time per operator per 12-hour day, no money, no tokens) and said 5 minutes changes no reading; Ninad: "5 minutes provides the cushion for errors, by providing early detection of the issue". Everything else in #480 stands in his words: "ok, every 15 min + after handover + next session checks it? but this should happen autonomously in background, Sahil should not put his time into doing it" -- with 15 read as 5 for the send. READ AS (Orion, not the ruling): (1) while an Orion session works, its computer sends that session's record at most once per 5 minutes, a FIXED interval that does not follow the counting block (#480's READ AS point 4 is corrected: he meant a fixed interval), started by the hooks the session already runs (no timer on a person's login, security #129) and run detached; (2) writing the handover sends it once more; (3) every session start still sends whatever of the computer's other sessions has not arrived; (4) the delivery requirement the 5-minute checker holds (#483: an active session's record no older than 15 minutes plus the send wait) stands, so the cushion is that two consecutive sends may fail before a delivery finding; (5) it binds every operator's computer by role (#481), the administrator's laptop included.

The checker runs 2 minutes after each expected send and reads the send frequency, so a missed send is found at once

Ruled 2026-10-21 by ninad -- in force

THE CHECKER RUNS 2 MINUTES AFTER EACH EXPECTED SEND AND READS THE SEND FREQUENCY, SO A MISSED SEND IS FOUND AT ONCE (Ninad, CLI-895, 2026-10-21, superseding communications #483). Told that the checker still allowed a record to be 15 minutes old before a finding, so a 5-minute send (#484) gave a cushion of two missed sends, he said: "let the checker run 2 minutes after, so that the detection is immediate, and let the checker know of the change in frequency". What #483 settled otherwise stands: a parallel checker on the estate, autonomous, for every active session, active meaning Orion was working on something the operator asked for. READ AS (Orion, not the ruling): (1) the checker reads the send interval from the one declared record the senders read (5 minutes today, #484), never a number of its own, so a change of frequency reaches sender and checker together; (2) it runs every send interval, 2 minutes after each expected send (at :02, :07, :12 ... on a 5-minute send), not on its own 5-minute clock; (3) an active session whose record from its computer does not cover its activity up to one send interval plus 2 minutes ago is overdue -- a delivery finding naming it on that very run, which the box reads as non-compliance (#482) and the Compliance Teller tells; (4) #483's 'delivered every 15 minutes' is retired with it: the delivery requirement IS the send interval plus the 2-minute wait; (5) a check that could not run reads NOT PROVEN, never a breach (#275), and the checker states its own outcome every run (monitoring #160); it runs under its owning agent's own login (security #129).

Orion active time from a session on an account other than the operator's approved Claude account is drawn in a slightly darker shade in the Orion active hours bar chart

Ruled 2026-10-23 by ninad -- in force

SESSIONS FROM A PERSONAL LOGIN ARE DRAWN SLIGHTLY DARKER IN THE ORION ACTIVE HOURS BAR CHART (Ninad, 2026-10-23, CLI-920). His words: "sessions from personal login should show up in a slightly darker colour in Orion active hours bar chart". Build lane asked: "Which lane should build these two changes (accounts read for every operator, and personal-login time shown darker in the hours chart)?" ANSWERED: "Orion bypass, now (Recommended)".

READ AS (Orion's reading): a personal login is a session whose recorded Claude account is not the one root's record approves for its operator (approved for that session, or not approved); its active blocks keep the bar's colour one shade darker and say so in the chart's key and tooltip; a session whose account was never recorded reads unknown and keeps the ordinary colour, never guessed. The count is unchanged: the shade shows where the time came from, never a second count.

Operator compliance is tracked to ensure operators follow the security protocols, have their time recorded correctly, are compensated correctly and can work effectively by the protocols defined for operators; that purpose is the basis on which compliance is monitored and tracked

Ruled 2026-10-23 by ninad -- in force

OPERATOR COMPLIANCE EXISTS FOR FOUR THINGS, AND THEY ARE THE BASIS FOR MONITORING AND TRACKING IT (Ninad, 2026-10-23, CLI-925). His words: "The reason for tracking the compliance is to ensure that the operators are doing what is necessary to ensure that security protocols are followed, time is recorded correctly, operators get compensated correctly and they are able to effectively do work following the right protocols as defined for the operators. This should be used as the basis to monitor and track compliance."

ASKED (in plain words, at his request): "Sahil's box is red because one of his chats from 21 Oct never sent its record to the machine. We've messaged him asking him to send it. His computer deletes chats older than about a month, so if he hasn't sent it by about 20 Nov, it's gone. Today's rule says a missing chat record keeps his box red until it arrives. If the chat is deleted, it never arrives, so his box would never turn green. Do you want an operator to be allowed to clear it by writing down why, or should it stay red?" ANSWERED with the words above, and the same session's instruction: "any initiative that is not directly related to moving the status of any of the operators to compliant should move to post mvp".

READ AS (Orion's reading, not the administrator's words, and NOT YET CONFIRMED by him for the question asked): every compliance rule is judged by whether it shows an operator not doing what one of the four needs; a finding is something the operator can act on. Applied to Sahil's CLI-891: while his computer holds the chat, sending it is what is necessary and the box stays red; once it is shown that no computer holds it any more, keeping the box red serves none of the four, while the session's minutes keep reading 'may be too low' so time and pay stay honest. Orion asked the administrator to confirm that reading before any rule is changed.

Once no computer holds an operator's Orion chat any more, that chat's missing record no longer keeps the operator's box red; its minutes keep reading 'may be too low'

Ruled 2026-10-23 by ninad -- in force

A CHAT THAT IS GONE FOR GOOD STOPS KEEPING THE BOX RED, AND ITS TIME STAYS MARKED (Ninad, 2026-10-23, CLI-926).

ASKED (in plain words): "Open from CLI-925: Sahil's box is red because one chat (CLI-891) never sent its record. His computer may delete that chat around 20 Nov. Once it's gone for good, should his box turn green, with that day's minutes still marked 'may be too low' so time and pay stay honest?" ANSWERED: he chose "Yes, green once it's gone" (offered as: "Keeping it red then serves none of the four reasons for compliance; the minutes stay marked.") over "No, stay red".

This confirms, for this case, the reading recorded under communications #502 (CLI-925), which was marked not yet confirmed.

READ AS (Orion's reading, not the administrator's words): traceable-work still never lapses by age (#477, #482 stand); a session's traceable-work finding clears only once it is shown that no computer holds that chat any more, and the session's minutes then keep reading 'may be too low', so time and pay stay honest. How that is shown is the build's to settle, never a guess from age.

Tracing and curing an operator's compliance finding happens in the background, by a droid of Orion's active while an Orion session is connected through its link, never as a prompt to the operator

Ruled 2026-10-25 by ninad -- in force

TRACING AND CURING AN OPERATOR'S COMPLIANCE FINDING HAPPENS IN THE BACKGROUND, BY A DROID OF ORION'S ACTIVE WHILE AN ORION SESSION IS CONNECTED THROUGH ITS LINK, NEVER AS A PROMPT TO THE OPERATOR (Ninad, CLI-948, 2026-10-25). Told that his Operator Instance Log box was red on traceable-work for CLI-943 (its record read overdue for commands 11:24-11:30Z) and that Orion's next step was 'tracing the CLI-943 finding', he asked what that was, then said: "this work should ideally happen in the background without operator prompt" and "yeah, maybe a new droid for orion who is active whenever the cli is connected to vm via carol link"; told what would be recorded and asked to go ahead, he answered "go ahead" and added: "yeah, maybe a new droid for orion who is active whenever the cli is connected to vm via carol link, for every orion session of all operators orion instances". READ AS (Orion, not the ruling): (1) when a compliance finding stands for an operator, a droid owned by Orion traces its cause from the estate's own records and cures what the estate can cure itself; the operator is told only what no droid and no session can cure, once; (2) the droid is active while an Orion session of that operator is connected to the machine through its link (the estate's live sign-ins), under Orion's own login, never as a timer on an operator's computer or login (security #129); (3) a cure that needs the operator's computer to send a record reaches that computer's next send through the session's own hook, with no prompt; (4) it covers every Orion session of every operator's Orion instance, by role (governance #481). The CLI-943 finding itself had already been traced by CLI-946 and cured by (a sign-in carried together for three sessions).

The Operator Instance Log's boxes carry the minimum on a fixed square-or-wider box, and the card below opens with its charts, the text after them

Ruled 2026-10-25 by ninad -- in force

THE OPERATOR INSTANCE LOG'S BOXES CARRY THE MINIMUM, AND THE CARD BELOW OPENS WITH ITS CHARTS (Ninad, CLI-953, 2026-10-25). His words: "1. the operator boxes in the operator instance log has too much text on it, keep it to the minimum, ensure that the boxes are not taller than their width, have fixed height and move the text to the box below to be displayed when selected 2. the box below the operator boxes in the operator instance log has too much text.. keep the bar chart and the line chart on top, and push the text below it.. including any non compliance text. Try to prune the text that is not relevant.."

READ AS (Orion's reading, not the administrator's words): (1) an operator's box shows only the operator's name, their compliance colour and state word, and today's Orion active hours of the day's limit; every box has the same fixed height and is never taller than it is wide, at any page width; everything else the box carried (status, role, time zone, Claude plan, approved Claude account, mapped agent, team, projects, services, assignments) moves to the card below and is shown when that box is selected. This amends #225's list of what the box shows -- those details now sit in the card -- and leaves #225's charts and hours as they are. (2) the card below opens with the month's bar chart (with its tokens line) and the selected day's line chart (active minutes hour by hour); the operator's details, the compliance findings and every other line of text follow them. This amends #290 point 4 ('the findings move to the top of that operator's card'): the findings now follow the charts, and the box still keeps its colour and state word. Text no reader needs to read the charts is pruned: a legend says what a mark is in a few words, a definition is shown once (in 'What these words mean', never repeated under a chart), and a reason the figures may be too low is still said, shortly, because #477 and #435 require it.

The Operator Instance Log's month bar chart shows no pop-up

Ruled 2026-10-25 by ninad -- in force

THE MONTH BAR CHART SHOWS NO POP-UP (Ninad, CLI-957, 2026-10-26). His words: "1. remove the popup tooltip from the bar char"

READ AS (Orion's reading, not the administrator's words): pointing at or touching a day's bar in the Operator Instance Log's month bar chart (Orion active hours) shows no pop-up. The day's figures stay where they are said today -- the selected day's cap, the bar's label read to a screen reader, the month as a table and the day's record -- and clicking a bar still opens its day. The day's line chart (active minutes, hour by hour) keeps its own pop-up: it was not named. This retires in place, for the month chart, communications #360's 'the tooltip says the day's hours and minutes'.

Personal-login time is drawn one shade darker block by block, in the order it was worked, for every chat including those before root kept accounts

Ruled 2026-10-25 by ninad -- in force

PERSONAL-LOGIN BLOCKS ARE DRAWN DARKER IN THE ORDER THEY WERE WORKED (Ninad, CLI-957, 2026-10-26). His words: "2. personal sessions are not showing up in darker shade"; and, asked how the chats from before root's record of Claude accounts began (25 Oct) should be handled -- the old session record shows them as personal while security #511 retired it -- he answered: "the bar slots from personal should show up in darker shade.. the slots should stack up starting from bottom.. so if i worked from the business account for the first 3.5 hrs, and then personal account for another 2.5 hrs, then the bottom part should be normal, and the top part of the bar representing 2.5 hrs should be darker shade"

READ AS (Orion's reading, not the administrator's words): (1) a day's bar in the month chart is its counted blocks stacked in the order they were worked, the earliest at the foot; each block whose work came only from a personal-login chat (#501's rule: a block also holding a business chat's work is not one) is drawn one shade darker wherever it falls in the day -- business then personal reads normal below and darker above, personal then business the reverse. The bar's height, every count and the day's line chart are unchanged. This amends #501's 'the personal part drawn from the bar's foot'. (2) The shade shows for every chat, those before root's record began included: for a chat root's record holds no account for, the shade alone reads the verdict the Session Opener wrote on that chat's own Logbook row when it opened. Root's record stays the one source for every chat it holds and for every other reader (compliance, the boxes, approvals); nothing is copied into root's store and nothing is written to the Logbook. This narrows security #511's 'the retired Logbook columns are never read again' for that one reader, the shade.

The Operator Instance Log's operator box also shows status and role, and the card below names the projects

Ruled 2026-10-25 by ninad -- in force

THE BOX SHOWS STATUS AND ROLE, AND THE CARD NAMES THE PROJECTS (Ninad, CLI-957, 2026-10-26). His words: "add these to the top box of operators
Status - Active ยท working now
Role - Administrator

in the box below, Projects - carolverse, chowpatty"

READ AS (Orion's reading, not the administrator's words): (1) each operator's box shows, beside the name, the compliance state word and today's Orion active hours, two short rows: Status (active, and 'working now' while the operator's Orion is working this hour) and Role (as the card says it today -- Administrator for the administrator). Every box keeps one fixed height, never taller than it is wide at any page width; the boxes' panel still shows one row of boxes. This amends #524's list of what the box carries. Status and Role are then said once, in the box, and leave the card (#524: a fact is shown once). (2) the card below shows, for every operator, Projects: every project the operator reaches (their own and, for the base project's operators, every project built on it -- security #316), as root's list names them, leaving out what root's list calls retired; for Ninad today that is Carolverse and Chowpatty's Sales & Customer Support. It replaces the card's single 'Project' line, and a service operator's Projects line is the same line.

An estate operator's role reads 'Estate operator' in the Operator Instance Log

Ruled 2026-10-25 by ninad -- in force

AN ESTATE OPERATOR'S ROLE READS 'ESTATE OPERATOR' (Ninad, CLI-957, 2026-10-26), on seeing Sahil's box read 'Role: Operator' after #528. His words: "Sahil is estate operator"

READ AS (Orion's reading, not the administrator's words): the role an operator's box and record say is the dictionary's word for their rung on root's record -- an operator on the operator rung who is not the administrator reads 'Estate operator', a service operator 'Service operator', and the administrator 'Administrator' -- never the bare word 'Operator'.

The Operator Instance Log's boxes judge compliance live on every load, never from a recorded verdict

Ruled 2026-10-25 by ninad -- in force

THE BOXES JUDGE COMPLIANCE LIVE ON EVERY LOAD (Ninad, CLI-962, 2026-10-26). Asked, after CLI-957 cut the page's load and the boxes still took 6-7 seconds on a busy machine: "should each box show a compliance verdict saved a few minutes ago (fast to load), or keep working it out live on every load (always current, 6-7 s)?" His words: "always current"

READ AS (Orion's reading, not the administrator's words): each operator's box and its colour are judged by the one compliance composer at the moment the page asks, never read from a verdict recorded earlier by a process; a faster box is welcome only if it stays current, so a load-time cut works on how the live judgement is computed, never on caching or storing its answer.

The Operator Instance Log filters its operators by role and by project, at the top of the page

Ruled 2026-10-25 by ninad -- in force

THE OPERATOR INSTANCE LOG FILTERS ITS OPERATORS BY ROLE AND BY PROJECT (Ninad, CLI-962, 2026-10-26). His words: "the top of the app should have filters like all, admin, estate operators, and service operators
also a drop down of projects should be there on top, to filter out operators"

READ AS (Orion's reading, not the administrator's words): at the top of the page, above the operators' boxes, a row of role filters -- All, Admin, Estate operators, Service operators -- each showing only the boxes whose role is that rung's word (Administrator, Estate operator, Service operator; All shows every box), and a drop-down of the projects the visible operators reach (All projects first) showing only the operators who reach the chosen project. The two combine. They narrow what the viewer already sees and never widen it: who sees whom (the floors) is unchanged.

The administrator is an estate operator too: the Operator Instance Log's role filters go by the rung

Ruled 2026-10-25 by ninad -- in force

THE ADMINISTRATOR IS AN ESTATE OPERATOR TOO (Ninad, CLI-962, 2026-10-26), on being told the new Estate operators filter (communications #531) showed only Sahil. His words: "Ninad is an estate operators as well as admin"

READ AS (Orion's reading, not the administrator's words): a role filter shows every operator who holds that role on root's record, not only those whose box names it: the administrator stands on the operator rung, so the Estate operators filter shows him beside the other estate operators and the Admin filter shows him; a service operator shows under Service operators. The Role line on his box still reads Administrator (#529).

An operator's last reported time zone is carried forward while the latest is not available

Ruled 2026-10-26 by ninad -- in force

AN OPERATOR'S LAST REPORTED TIME ZONE IS CARRIED FORWARD WHILE THE LATEST IS NOT AVAILABLE (Ninad, CLI-970, 2026-10-26). The question: Asked why Sahil's time zone reads unknown while his time is captured, Orion answered that root's store of workstation reports began on 2026-10-25 11:33Z and carried no earlier report over (security #511, cookbook 1671), and that Sahil's computer had reported 'India Standard Time' 21 times up to 2026-10-22, a name the estate recognises. His words: "if the latest timezone is not available, carry forward the last timezone"

READ AS (Orion's reading, not the administrator's words): wherever the estate needs an operator's time zone -- the Operator Instance Log, the operator's own day, the hours by day, the time-zone compliance rule -- and the newest report names no zone the estate recognises (none recorded yet, an empty or unrecognised zone, or root's store unreadable), it uses the most recent recognised zone that operator's computers DID report, and says it is carried forward and from which report. Root's store is read first, then the Operator Instance Log's retired workstation reports, read as history and never copied into root's store. A zone never reported is still never guessed: it reads UTC with the reason. A carried-forward zone counts as reported for the time-zone rule (communications #477). This narrows security #511's 'no report is carried over' for this one fact; everything else #511 rules stands.

A session that sends its time to the estate sends its computer's time zone with it

Ruled 2026-10-26 by ninad -- in force

A SESSION THAT SENDS ITS TIME SENDS ITS TIME ZONE (Ninad, CLI-970, 2026-10-26). The question: Orion answered that Saee's computer has never sent a time zone, although two of her sessions (2026-10-14) reached the estate and their time is counted: the zone travels only in the workstation report, which the session-start hook sends only when all four gates pass and the session's link is open, while the session and its time are opened by a separate act. His words: "Her computer has never sent a report - if the time was sent, timezone should have been captured"

READ AS (Orion's reading, not the administrator's words): the act that brings a session's time to the estate -- the Session Opener, for the administrator's laptop, an estate operator and a service operator alike -- brings the computer's time zone with it, and root records it for that session before the session opens, whatever the workstation gates read. A computer that cannot read its zone sends it empty and root records it unreported, never guessed. The full workstation report (zone, Claude plan, gates, entry gate, permissions) is unchanged and still sent by the session-start hook.

An operator's location is the city and country of the address their sign-in came from, looked up on our machine

Ruled 2026-10-26 by ninad -- in force

AN OPERATOR'S LOCATION IS THE CITY AND COUNTRY OF THE ADDRESS THEIR SIGN-IN CAME FROM, LOOKED UP ON THE ESTATE'S OWN MACHINE (Ninad, CLI-971, 2026-10-26). The question: Picking up (Ninad, CLI-970: "also, location capturing is not enabled yet.."), Orion asked: "Today an operator's box only shows their computer's time zone. What should 'location' show?" and offered three answers: the city and country of the address each sign-in came from, looked up on the estate's own machine (nothing new runs on the operator's computer; a VPN shows the VPN's city); the operator's own computer's location service with their consent (more exact, a signed helper needed on a Mac); or the country only. Who sees it was said to stay as today: the administrator every operator, each operator their own floor, with when it was captured. He chose the first answer, as offered: "City from sign-in -- City and country looked up on our own machine from the address each sign-in came from; works the same on Mac and Windows; wrong behind a VPN."

READ AS (Orion's reading, not the administrator's words): an operator's location is the city and country of the internet address that session's sign-in to the machine came from, as sshd itself recorded it, looked up in a city database kept on the estate's machine -- no address is sent to any outside service. It is read when the sign-in itself is read (at the session's opening, or by Orion's Door Sign-in Reader for a session whose sign-in was left pending) and kept on that session's record with the moment of the sign-in and the database edition that answered; the address is not copied into that record. The operator's box shows the place of their most recent session whose place could be read, with when that sign-in was made and that it comes from the sign-in's address, beside the time zone, which stays the computer's own report (communications #225, #533, #534). An address the database cannot place (a private or loopback address, or one it does not hold) reads 'not placed' with the reason, never guessed; a VPN or a relay shows its own city, and nothing claims otherwise. Nothing runs on the operator's computer and nothing is typed. Who sees it is unchanged: whoever may see that operator's box. Design 690's 'location is the time zone' (communications #225) is narrowed in place: the time zone stays, the place is added.

The Orion active hours chart's token consumption line is white; its dots stay red

Ruled 2026-10-26 by ninad -- in force

THE ORION ACTIVE HOURS CHART'S TOKEN CONSUMPTION LINE IS WHITE; ITS DOTS STAY RED (Ninad, CLI-974, 2026-10-26). His words: "on the Orion active hours chart, make the tokens consumption line white in colour instead of red, points (dots) can stay red".

READ AS (Orion's reading, not Ninad's words): on the Operator Instance Log's month chart the token consumption line, and the legend's sample of that line, are drawn white; each day's token dot and the ring of an open (lower-bound) dot stay red, as do the legend's dot samples. This amends communications #426 item (1) for the line and its legend sample only -- #426's red for the dots and its item (2), the smoother curve, stand. The line keeps its axis, shape and smoothing; the page is dark-only, so white reads on it.

An operator's location shows on their top box in the Operator Instance Log

Ruled 2026-10-26 by ninad -- in force

AN OPERATOR'S LOCATION SHOWS ON THEIR TOP BOX IN THE OPERATOR INSTANCE LOG (Ninad, CLI-974, 2026-10-26). His words: "the operator location should show up on the top box".

READ AS (Orion's reading, not Ninad's words): each operator's box at the top of the Operator Instance Log carries one short Location row beside Status and Role -- the city and country their most recent sign-in came from (communications #547), or 'not read yet', 'not placed' or 'unreadable' said as such -- from the same place reading the card below shows. When that sign-in was made and DB-IP's attribution stay in the card, and ride the row's own tooltip. The box stays one fixed height, grown by that one row as #528 grew it, never taller than it is wide, and the boxes' panel grows by the same row. This amends #524 item (1) and #528's list of what the box carries; the card keeps its Location line.

An operator's Orion active time matches the tokens its sessions consumed; time with no token consumption and long idle sessions are not working time

Ruled 2026-10-26 by ninad -- superseded 2026-10-28

AN OPERATOR'S ORION ACTIVE TIME MUST MATCH THE TOKENS ITS SESSIONS CONSUMED: TIME WITH NO CORRESPONDING TOKEN CONSUMPTION IS NOT RECORDED AS WORKING TIME, AND A LONG IDLE SESSION IS NOT COUNTED AS WORKING TIME (Ninad, CLI-978, 2026-10-26). RAISED BY NINAD asking whether the active-time records can be manipulated by an operator -- "asking operator orion instance to artificially create records that will give the illusion of work"; told that the laptop's records can be fabricated on the operator's own machine, that the server verifies who sent them and not that the work happened, and that no compliance rule compares tokens against hours, he answered: "yes please, the token consumption should match the time worked, and time without corresponding token consumption should not be recorded as working time and long idle sessions should not be counted as working time". READ AS (Orion, not the ruling): a block of Orion active time counts only when the session's cost-ledger rows show tokens consumed in or beside it; a block with none is not working time whichever source reported it, which narrows communications #475 (recording more is acceptable) and #476 (a background command between turns counts) for blocks with no spend; a session showing no token consumption for a long stretch is idle for that stretch; a block whose tokens cannot be read is unmeasured and said, never counted and never silently dropped; where the hours and the tokens disagree it is a finding on the operator's box, not a quiet correction.

Ended because: Ninad's answer, CLI-005: work is counted whether or not tokens are consumed; tokens become a reconciliation check (communications #560)

Operator time is counted only from what the estate's own machine saw: every operator's Claude traffic passes through a relay on the machine, and the laptop's report decides nothing

Ruled 2026-10-26 by ninad -- superseded 2026-10-28

AN OPERATOR'S ORION ACTIVE TIME AND TOKENS ARE COUNTED ONLY FROM RECORDS THE OPERATOR CANNOT WRITE: EVERY OPERATOR'S CLAUDE CODE TALKS TO CLAUDE THROUGH A RELAY ON THE ESTATE'S MACHINE, WHICH KEEPS THE TIME AND THE TOKENS OF EVERY REQUEST; ONLY RELAYED WORK COUNTS; ANTHROPIC'S OWN PER-PERSON USAGE REPORT AND THE MACHINE'S COMMAND RECORD ARE THE CROSS-CHECKS; THE LAPTOP'S REPORT IS DISPLAY ONLY (Ninad, CLI-979, 2026-10-27). RAISED BY NINAD after communications #551, asking "what is the full proof solution?". Orion proposed three layers -- a relay on carol-vm that every operator's Claude Code is pointed at, so the machine records real timestamps and real token counts per request and an operator who bypasses it earns nothing; Anthropic's per-user daily usage report for the Team plan pulled by the estate as the independent reconciliation; the server's existing command journal for which minutes were estate work -- with the laptop becoming display-only, never a source of pay, and said one thing must be proven by doing first: whether Claude Code on the Team login honours a custom gateway address. Asked "is this full proof solution?", Orion answered it is fool-proof against invented records and not against wasted hours, which only judging the work itself catches. Ninad: "ok, file the initiatives in appropriate pursuit and show me the status in usual format". READ AS (Orion, not the ruling): the relay design is approved as the way #551 is kept; the gateway proof is the first step and a refusal there comes back to Ninad as a law gap, never worked around; wasted-but-real hours stay a judgment of the work on the record, not a counting rule.

Ended because: Ninad's answers, CLI-005 (#561, #562): genuine work is counted from every record that shows it; the relay, the token record and Anthropic's report reconcile and flag

Before an operator's estate link is open, their Claude Code reaches Claude directly through a forwarder on their own computer; once the link is open, through the relay

Ruled 2026-10-27 by ninad -- superseded 2026-10-28

AN OPERATOR'S CLAUDE CODE REACHES CLAUDE DIRECTLY UNTIL THEIR ESTATE LINK IS OPEN, AND THROUGH THE RELAY FROM THEN ON; THE DIRECT MINUTES ARE NOT COUNTED (Ninad, CLI-982, 2026-10-27). RAISED BY ORION building a session starts before it asks for its link, and with the relay Claude is reached only through the link. ASKED: how should an operator's Claude Code behave BEFORE their estate link is open? Offered STRICT (no link, no Claude: the operator opens the link first, nothing new on laptops, no uncounted minutes; recommended) and LENIENT (a small forwarder on each computer sends to Anthropic directly until the link opens, then through the relay; the first minutes of every session are uncounted by rule; a Python program on Macs and a PowerShell one on Windows). He chose: "Lenient: direct until the link opens". READ AS (Orion, not the ruling): every operator computer runs a forwarder the template installs, on the one declared relay port the managed settings point Claude Code at; a request from a session whose link is open is carried over THAT session's own link to the relay (the forwarder maps the connecting Claude Code process to its session, so no request rides another session's sign-in); a request from a session with no link goes straight to Anthropic and the estate records nothing of it -- those minutes are not working time under #552; the forwarder holds no sign-in and keeps no text.

Ended because: Ninad's answer, CLI-005: work before the link opens, reading handovers among it, is real work and is counted

The operator-hours pursuit must not bring back the problems its build caused, and genuine work must never be disregarded

Ruled 2026-10-28 by ninad -- in force

THE PURSUIT 'OPERATOR HRS TRACKING LOGIC ENHANCEMENT' (CAROL-PUR-0018) MUST NOT INDUCE AGAIN THE PROBLEMS ITS IMPLEMENTATION CREATED, AND THE COUNTING IT BUILDS MUST NOT DISREGARD GENUINE WORK (Ninad, CLI-005, 2026-10-28). Ninad's words: "the implementation of this pursuit created problems, I suggest you also take the handover from the session that worked on the errors, and ensure that the pursuit does not induce those problems again.. also ensure that genuine work is not dis-regarded". Context: on 27-28 Oct the laptop forwarder cut every relayed reply slower than 10 s to start (547 relayed replies broke, 'Connection dropped (ECONNRESET)' in Ninad's chats, CLI-998 and CLI-999 stalled); CLI-004 fixed it (release v153). READ AS (Orion, not the ruling): the pursuit's remaining steps carry criteria that (1) no route a session's Claude traffic takes -- the Mac/Linux forwarder, its Windows twin, the machine's relay, the link -- cuts or fails a reply: a slow reply arrives, a relay that is restarting or down never fails a chat, and the relay finishes replies in flight before it restarts; and (2) work that really happened is counted or, where the estate's own records cannot show it, named as 'may be too low' -- a relayed request that broke still counts its span, a long reply counts from its start, and work a linked session sent directly because of a fault is never silently dropped. Where an earlier ruling of his excludes real work (communications #551 on token-less stretches, #554 on the minutes before the link opens), the disagreement is put to him, never settled by Orion.

When Orion is doing work it is counted whether or not tokens are consumed; token consumption flags inappropriate time records and never undercounts genuine effort

Ruled 2026-10-28 by ninad -- in force

WHEN AN OPERATOR'S ORION IS DOING WORK IT IS COUNTED, WHETHER OR NOT TOKENS ARE CONSUMED; TOKEN CONSUMPTION IS A RECONCILIATION CHECK THAT FLAGS INAPPROPRIATE TIME RECORDS WHEN CONSISTENT EVIDENCE LEADS TO THAT CONCLUSION, NEVER A RULE THAT UNDERCOUNTS GENUINE EFFORT (Ninad, CLI-005, 2026-10-28). ASKED (Orion, CLI-005, settling a disagreement between his own rulings under P.01.01.05.08): "Long commands such as test runs or builds use no tokens while they run. Your 27 Oct ruling (#551) says time with no tokens isn't work, but your 25 Oct ruling (#475/#476) counted a running command as the session's work. Should those minutes count?" Ninad: "if orion is doing work, it should get counted, irrespective of whether tokens are consumed or not.. token consumption should be used as a reconciliation check to flag in-appropriate time records if consistent evidences result into that conclusion, not undercount genuin efforts". READ AS (Orion, not the ruling): #551's 'time with no corresponding token consumption is not recorded as working time' and 'a long idle session is not counted' are retired; what survives of #551 is its purpose -- records that only give the illusion of work are caught -- now by FLAGGING: a time record the token record (the relay's, the cost ledger's, Anthropic's per-person report) consistently fails to support turns the operator's box red naming the day and the blocks, and the hours stay counted until an operator judges them; a long running command, a long reply and reading or thinking between calls are work; a stretch with no evidence of work at all (no turn, no command, no request) is still not work, because nothing shows Orion doing anything -- that is the absence of a record, not a token rule.

Claude is reached directly until the estate link is open and through the relay after; the work done before the link opens is counted

Ruled 2026-10-28 by ninad -- in force

AN OPERATOR'S CLAUDE CODE REACHES CLAUDE DIRECTLY UNTIL THEIR ESTATE LINK IS OPEN AND THROUGH THE RELAY FROM THEN ON, AND THE WORK DONE BEFORE THE LINK OPENS -- READING HANDOVERS AMONG IT -- IS REAL WORK AND IS COUNTED (Ninad, CLI-005, 2026-10-28). ASKED (Orion, CLI-005): "Minutes before you open the link are not counted (your 'Lenient' ruling, #554). Today that left out about 10 minutes of this session reading handovers. Keep them uncounted, or count them?" Ninad: "session reading handovers is real work and that should be counted". READ AS (Orion, not the ruling): #554's route stands (a forwarder on each computer, direct until the link opens, through the relay after); its clause 'the direct minutes are not counted' is retired; the minutes before the link opens are counted from the session's own record of its turns (the transcript projection its computer sends at the next session start), checked against the token record like every other minute (#561).

Operator time is counted from every record that shows Orion working; the records the operator cannot write reconcile it and flag what they consistently contradict

Ruled 2026-10-28 by ninad -- in force

AN OPERATOR'S ORION ACTIVE TIME IS COUNTED FROM EVERY RECORD THAT SHOWS ORION WORKING, AND THE RECORDS THE OPERATOR CANNOT WRITE -- THE RELAY ON THE ESTATE'S MACHINE THAT CARRIES THEIR CLAUDE TRAFFIC ONCE THE LINK IS OPEN, ANTHROPIC'S OWN PER-PERSON USAGE REPORT AND THE MACHINE'S COMMAND RECORD -- RECONCILE IT AND FLAG WHAT THEY CONSISTENTLY CONTRADICT (Ninad, CLI-005, 2026-10-28, from his answers recorded as #561 and #562). ASKED: the two answers above (CLI-005) leave #552's clauses 'only relayed work counts' and 'the laptop's report is display only' contradicting them: work before the link opens and work on the operator's own computer between requests are seen only by the computer's record of its turns. His words: "if orion is doing work, it should get counted ... token consumption should be used as a reconciliation check to flag in-appropriate time records if consistent evidences result into that conclusion, not undercount genuin efforts"; "session reading handovers is real work and that should be counted". READ AS (Orion, not the ruling): #552's relay design stands (every operator's Claude traffic through the relay once the link is open, the relay recording time and tokens and never a body, Anthropic's report and the command record as cross-checks); its clauses 'only relayed work counts' and 'the laptop's report is display only' are retired: the computer's record of its turns counts (the source that saw work wins, #475), and where the relay, the token record or Anthropic's report consistently fail to support a computer's record, the day reads red naming what disagrees -- the guard against invented records is the flag, never a silent subtraction.

An operator's Orion work on their own computer -- Sahil's Chowpatty prototypes -- is never left unaccounted, also in a session that never opens the estate

Ruled 2026-10-28 by ninad -- in force

AN OPERATOR'S ORION WORK ON THEIR OWN COMPUTER -- SAHIL BUILDING CHOWPATTY PROTOTYPES ON HIS LAPTOP AMONG IT -- MUST NOT GO UNACCOUNTED (Ninad, CLI-005, 2026-10-28). Ninad's words, sent while Orion was carrying #560-#563 into the pursuit: "note that Sahil is building prototypes on his laptop for chowpatty, that work should not get un-accounted". Found the same minute (CLI-005): a session that never runs the estate opener carries no session key, so the computer's sender and the machine's transcript intake both read it 'not a Carol session' and nothing of it is counted. READ AS (Orion, not the ruling): work an operator's Orion does on their own computer for a project is counted for that operator from the computer's record of its turns, also in a session that never opens the estate link or a Logbook session; such a session is counted under its own Claude session, attributed to the operator whose computer sent it over their own sign-in, and reconciled like any other (#561, #563); the project it served is recorded where the record shows it, and an unattributed project is said, never a reason to drop the hours.

A session that opened while the opener failed may be given its chat id from evidence, as a recorded correction

Ruled 2026-11-03 by ninad -- in force

A SESSION THAT OPENED WHILE THE OPENER FAILED MAY BE GIVEN ITS CHAT ID FROM EVIDENCE, AS A RECORDED CORRECTION (Ninad, CLI-017, 2026-11-03). THE QUESTION ANSWERED: Sahil's operator communication #141 -- his sessions CLI-973 and CLI-977 opened on 26 Oct while the Session Opener failed (the /tmp marker fault, since fixed), so their records hold no chat id and the estate can never match them; his own notes name the chat 0529ca3a-b774-4e18-b81b-cfc3c6a9d459 for both. Asked 'May they be given their id now from evidence, as a recorded, append-only correction (like the sign-in corrections)?', HIS ANSWER: "Yes, correct them (Recommended)" -- offered as 'a recorded correction naming who, why and the evidence; nothing is rewritten in place'. READ AS (Orion's reading, not his words): the id is set through a new append-only correction record (who, why, evidence), the shape of the sign-in method corrections; editing the session row in place undoes it.

Every change to the policies is told by email to the estate operators of the projects it concerns

Ruled 2026-11-03 by ninad -- in force

EVERY CHANGE TO THE POLICIES IS TOLD BY EMAIL TO THE ESTATE OPERATORS OF THE PROJECTS IT CONCERNS (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: asked to approve two policy texts settling contradiction #10 (P.01.03.09.06, a user operates an agent only as far as a subscribed service mandates; P.01.09.01.02, Blueprint's mandate), he answered: HIS WORDS: "Approve both. Also, create a mandate that any change to the policies should be communicated via email to estate operators of the respective projects". READ AS (Orion's reading, not his words): (1) whenever a policy is added, amended, superseded or retired, an email tells the estate operators of every project the policy applies to -- for an estate-wide policy, the estate operators of Carolverse and of every project built on it, read from root's register, never typed; (2) the email is a one-way update (communications #194) naming the policy, what changed and why, sent from the agent who owns the service the change belongs to (communications #96), never from a person's address; (3) recorded as a ruling (P.01.01.05.05); it becomes policy only if he asks for policy words.

An email to an operator always goes with an operator communication to that operator's Orion instance, so the operator and their Orion are on the same page

Ruled 2026-11-03 by ninad -- in force

AN EMAIL TO AN OPERATOR ALWAYS GOES WITH AN OPERATOR COMMUNICATION TO THAT OPERATOR'S ORION INSTANCE, SO THE OPERATOR AND THEIR ORION ARE ON THE SAME PAGE; AND THE POLICY-CHANGE EMAIL IS BUILT IN THE PURSUIT 'USER RULINGS BUILT' (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: told that the build of communications #579 (every policy change emailed to the estate operators of the projects it concerns) waited in Orion's own backlog because its twenty places were full, he answered: HIS WORDS: "the email build should be included in this pursuit and carried out as part of this pursuit.. the communication through email should always be coupled with communication to the orion instance of the operator, as the operator and orion instance of the operator should be on the same page.." READ AS (Orion's reading, not his words): (1) whatever the estate emails an operator, the same thing reaches that operator's Orion instance as an operator communication on the estate record (communications #203), so the session reads first what the operator was mailed; an email alone to an operator is incomplete; (2) the policy-change notice of #579 is therefore an operator communication to each estate operator concerned, whose email notice (communications #209, #298) is the email; (3) its build is a step of CAROL-PUR-0019 and is carried out as part of that pursuit.

Telling operators when parked work can restart

Ruled 2026-11-03 by ninad -- in force

When everything a parked pursuit step waits on is met, Orion tells that it can restart -- by email, in the operator's Orion session and on their to-do list -- and keeps reminding until it is restarted. The watch is a named track of the Communications service (Ninad's answer: "it should be a track in the existing communications service"), and every channel it tells through is named and documented.

Who is emailed that parked work can restart

Ruled 2026-11-03 by ninad -- in force

The 'ready to restart' email goes to everyone the step waited on: the pursuit's operator owner and every operator whose act it waited on (for CAROL-PUR-0018: Ninad, Sahil and Saee). A notification Ninad rules on for them, alongside #389's email for urgent matters and #393's own choices.

Communications to operators are addressed to role groups -- admin, estate operators, service operators -- and reach every member, the sender's own operator included

Ruled 2026-11-03 by ninad -- in force

COMMUNICATIONS TO OPERATORS ARE ADDRESSED TO ROLE GROUPS -- ADMIN, ESTATE OPERATORS, SERVICE OPERATORS -- AND REACH EVERY MEMBER, THE SENDER'S OWN OPERATOR INCLUDED (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: Orion sent the notice that Sachin Joshi's recommended request awaits approval to Sahil alone, by name, and the operator communication door refuses a communication to the sender's own operator ('a communication goes to ANOTHER operator'). HIS WORDS: "I am also estate operator, why am I not communicated? are you communicating via roles or individuals? you should have groups based on roles (admin, estate operators, service operators etc)". READ AS (Orion's reading, not his words): (1) a communication or notice meant for a role is addressed to that role's group -- the admin, the estate operators, the service operators, and any further rung root's record holds -- for the project it concerns, its members read from root's authority record and register at send time, never typed and never chosen one by one; (2) every member receives it in their own Orion instance with its email (communications #580), the operator whose session sent it included, because another session of theirs may be the one to act; (3) a communication between two named operators stays possible.

The Communications Catalogue

Ruled 2026-11-03 by ninad -- in force

There is a Communications Catalogue: a first-class record in the registry of every communication the estate makes -- to operators, employees, investors and users, by email, in the Orion session, on Carol's WhatsApp and every other channel -- with its name, audience, channel, service, track, task, the agent that acts, what is sent, to whom and the rulings behind it; and it is an app.

Who owns a communication, and how the catalogue's app is split

Ruled 2026-11-03 by ninad -- in force

A communication has no single owner: it is owned by the agent that takes the action, under the service its task belongs to. The Communications Catalogue therefore has no single owner either, and its app is split accordingly -- each part kept by the owner of what it shows. The catalogue is built as its own pursuit, ranked ahead of Restart reminders, which registers its track and channels in it.

The Communications Catalogue's app is one app per acting agent, each owned by that agent alone

Ruled 2026-11-03 by ninad -- in force

Ninad, CLI-019 (2026-11-03), clarifying communications #592's 'its app is split accordingly'. HIS WORDS: "by split - i meant 1 app per agent, because an app can be owned only by a single agent". RULED: every agent that acts in a communication has its own Communications Catalogue app, owned by that agent alone, listing every catalogue entry it acts in across services, each labelled with its service, track and task. No catalogue app is split into parts kept by different owners, because an app is owned by exactly one agent. The rest of #592 stands: each entry is owned by the agent that acts, under the service its task belongs to, and only that agent or the service's owner writes it. is reshaped to this ruling.

Agents send the estate's own notices to operators through one narrow door; Orion sends policy changes

Ruled 2026-11-04 by ninad -- in force

AGENTS SEND THE ESTATE'S OWN NOTICES TO OPERATORS THROUGH ONE NARROW DOOR; ORION SENDS POLICY CHANGES (Ninad, CLI-036, 2026-11-05). THE QUESTION ANSWERED: told that must tell every estate operator, in their Orion session and by email, whenever a policy changes; that communications #579 has the agent who owns the service send that notice and Orion owns the policy service; but that the operator communication door admits only an operator as sender and no agent login can write its store (communications #203) -- and asked who should send policy-change notices, offered 'Orion sends' (one narrow door lets named agents send named kinds of notice; the same door later lets Carol and Jarvis tell operators a request awaits approval, 7009; a check every few minutes catches every change, whoever made it) or 'the operator who made it' -- he chose: HIS ANSWER: "Orion sends". READ AS (Orion's reading, not his words): (1) a notice the estate itself owes operators -- not one operator writing to another -- is sent by the agent who owns the service it belongs to, as that agent, never in a person's name; (2) it travels as an operator communication (#203, #580) to a role's group (#588), its email notice from Orion's address (#412); (3) agents reach the operators' store only through one door that admits a named agent for a named kind of notice, read from one record, never any agent for anything; (4) policy changes are Orion's kind: a check compares the policies with what was last told and tells every new policy, amendment, supersession and retirement, whoever made it; (5) a communication between operators still has an operator sender.

An operator's role is the role they hold on the project each piece of their work served, recorded on the work once and read by every view

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-040). HIS WORDS: "when an operator acts on behalf of a project to talk to Leo, he is not just another user. He is an employee of carolverse, acting on behalf of the user. example is Sahil when he works on blueprint, he is acting as estate operator of carolverse, but when he works on onboarding chowpatty, he is representing sachin to negotiate with Leo to build chowpatty, and his role is the estate operator of chowpatty. Actions determine the role of the operator, and hence I want you to device a solution to correctly determin the role of the operator and display that in the operator instance log chart via filter tabs in Orion active hours.. is that possible? or is there a better solution?"; shown Orion's design, "I'd go one step further: record the role on the work itself, once, rather than working it out only for the chart. The hourly updates, status reports and Chowpatty's bill would then all read the same answer, so hours spent for Sachin are charged to Chowpatty - I agree" and, asked whether to record it as his ruling and file it, "yes"; then "file a new pursuit for this and include the initiatives in appropriate arcs". READ AS (Orion's reading, not his words): (1) an operator's role is decided by what they do, not by who they are: each stretch of their work names the project it served, taken from the work itself -- a build counts for the project of the service it improves (initiatives #491), a conversation in Leo's operator window for the project it is about -- and the role is the one root's record gives them on that project (security #316), naming whom they represent when it is a client project (Sahil on Chowpatty: its estate operator, for Sachin); (2) it is recorded on the work once, and the hours, the hourly updates, the status reports and the project's bill all read that one record, so hours spent for a client project are charged to it; (3) work that names no project is shown as not placed, never guessed, and an hour that mixes projects is split by the minute; (4) the Orion active hours chart in the Operator Instance Log gets tabs -- All, one per project and role, Not placed -- and under All each bar is shaded by role; the role filters at the top still choose people (communications #531, #532).

Every change to an arc or pursuit is told to its owner operator and to whoever is assigned to execute it

Ruled 2026-11-06 by ninad -- superseded 2026-11-06

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042): 'who gets communicated when there is a change to the arc or pursuit?' and 'what kinds of changes get communicated?' -- Orion answered that only the service operator an arc or pursuit is assigned to or withdrawn from is told (the assigning operator copied), and that changes to steps, criteria, MVP, rank or status reach nobody automatically. HIS WORDS: "both owner operator and executioner (who is mapped to the arc or pursuit) should be informed about the change" and "All changes should be communicated, like increase or decrease in scope and initiatives, assignment changes, description changes, execution progress etc". READ AS (Orion, not his words): for every arc and pursuit, its operator owner and every operator assigned to execute it are told of every change -- scope widened or narrowed (phases, steps, criteria, MVP), initiatives added or removed, who is assigned or withdrawn, the name, description or outcome reworded, and how its execution is progressing -- as operator communications, which the estate already emails; each person told once per change, never a second copy of a notice already sent (design 706). OPEN when recorded, asked of Ninad in CLI-042: who an arc's owner operator is (an arc records an owning agent and its creator, no operator owner), and how often execution progress is told.

Ended because: Ninad chose a daily summary for execution progress, and ruled who the owner operator is (CLI-042, initiatives #651)

Every change to an arc or pursuit is told to its owner operator and to whoever is assigned to execute it; progress once a day

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-042). ASKED (CLI-042): 'who gets communicated when there is a change to the arc or pursuit?' and 'what kinds of changes get communicated?' -- Orion answered that only the service operator an arc or pursuit is assigned to or withdrawn from is told (the assigning operator copied), and that changes to steps, criteria, MVP, rank or status reach nobody automatically. HIS WORDS: "both owner operator and executioner (who is mapped to the arc or pursuit) should be informed about the change" and "All changes should be communicated, like increase or decrease in scope and initiatives, assignment changes, description changes, execution progress etc". ASKED NEXT how execution progress should reach them, since it changes often (Blueprint alone has 46 steps), HE CHOSE 'Daily summary': one message a day per arc or pursuit that moved -- steps closed, % by effort, next step -- and nothing on a quiet day, while scope, initiative, assignment and description changes are told as each one happens. READ AS (Orion, not his words): for every arc and pursuit, its owner operator (the creator, initiatives ruling of the same day) and every operator assigned to execute it are told of every change -- scope widened or narrowed (phases, steps, criteria, MVP), initiatives added or removed, who is assigned or withdrawn, the name, description or outcome reworded -- as it happens, and its execution progress once a day when it moved; told as operator communications, which the estate already emails; each person told once per change, never a second copy of a notice already sent (design 706).

Every command an Orion hands its operator ends with a conclusion section, printed last on the terminal, saying it SUCCEEDED and what is now true, or FAILED and what to tell Orion

Ruled 2026-11-07 by ninad -- in force

HIS WORDS (CLI-056, 2026-11-07), after running a core-grant line that printed nothing: "I ran the command, but it did not tell whether it was successful or failed, make sure that all your future command have a conclusion session that appears at the end on the terminal after executing the command"; then, of a one-line result tail: "this command is not confirming to the request on how commands should be given"; and of his link command handed wrapped in such a section: "chat's own connection follows a very specific pattern.. I just type carol-link, almost everytime.. this is for the first time that i am given this kind of a command"; and: "please make these as rulings so that they are also propogated to all other operators working on the system". READ AS (Orion's reading, not his words): every command an Orion hands its operator (any prompt, any operator) ends with a conclusion SECTION -- a visible banner block printed last on the terminal, judged from the command's own result (its exit status, or a check of what it should have made true): SUCCEEDED and what is now true, or FAILED with what to tell Orion. A one-line echo is not the section. A fixed command that already ends by printing its own outcome (`carol-link`: LINK OPEN / LINK REFUSED) is handed bare, in its usual pattern. The section is for the operator; Orion still verifies from the machine after the operator's 'done'.

Non-compliance of an operator's Orion instance means a structural change the operator must make at once for their time to be tracked right; mainly setup bugs and inconsistencies; it can also be a security issue; an unread message is a to-do

Ruled 2026-11-09 by ninad -- in force

NON-COMPLIANCE OF AN OPERATOR'S ORION INSTANCE MEANS A STRUCTURAL CHANGE THE OPERATOR MUST MAKE AT ONCE SO THEIR TIME IS TRACKED RIGHT -- MAINLY BUGS AND INCONSISTENCIES IN THE INSTANCE'S SETUP; A FINDING MAY BE BOTH NON-COMPLIANCE AND A SECURITY ISSUE; AN UNREAD MESSAGE IS A TO-DO (Ninad, 2026-11-09 UTC, CLI-091/CLI-095; supersedes communications #275).

THE QUESTION ANSWERED: he asked why his box in the Operator Instance Log turns non-compliant again and again; Orion listed the recurring causes (an hour worked with no hourly update, a sign-in with no code on record, a message from another operator unread when the update was written, the laptop out of step with root's copy).

HIS WORDS: "non-compliance is a heavy word, and it basically means that the operator needs to do certain structural changes immediately to enable the right tracking of the operator time.. If the status flips to non compliance, there should be a good reason.. 1. An hour worked with no hourly update (most common) - there should be no concept of hourly update because updates happen in 15 min slots, if there is no activity in a 15 min slot, that time should not get recorded, why should it make the operator non-compliant? 2. A sign-in with no code on record (today's finding) - if the operator does decides not to login to the VM, that is not a reason for him to be non-compliant.. it just means that the operator is either working locally or not working, either of which should be captured when the operator next logs in.. but its not a reason for the operator to be non compliant.. 3. A message from another operator unread when the update was written - this is a "to-do" activity, not a non compliance 4. The laptop out of step with root's copy - this is a genuin case of non-compliance"

ASKED NEXT: whether two chats on one sign-in, or a privileged command run on a sign-in already used, are non-compliance or a security alert. HIS WORDS: "Two chats on one sign-in, or a privileged command run on a sign-in already used - these look like non compliance because they are structural changes to me.. the reality is, non compliance and security issues are somewhat similar, and they can be both.. non compliance primarily refers to bugs and inconsistencies in operator orion instance setup.."

READ AS (Orion's reading, kept apart from his words; not yet confirmed by him): the box reads non-compliant only for something in the operator's Orion instance setup the operator must change for their time to be tracked right, never for work simply not done. Stays non-compliance: the laptop out of step with root's copy (workstation, entry-gate and permission gates), two chats on one sign-in, a privileged command on a sign-in already used, no time zone reported and a chat whose record never reaches the machine (communications #477, #482). Becomes a to-do on the operator's list, never red: a message from another operator unread, and the owed replies and answers communications #346 counts (that ruling stays live until he rules on it). An hour's update follows communications ruling recorded with this one. A sign-in with no Authenticator code is a security breach for Heimdall, not the operator's non-compliance (security ruling recorded with this one). Communications #502's four purposes stand.

An operator's time is recorded from its 15-minute slots; the hourly update is an additional source that must match the slots, a mismatch is non-compliance whose logic is reconciled

Ruled 2026-11-09 by ninad -- in force

AN OPERATOR'S TIME IS RECORDED FROM ITS 15-MINUTE SLOTS; THE HOURLY UPDATE STAYS AS AN ADDITIONAL SOURCE THAT MUST MATCH THE SLOTS, AND A MISMATCH IS NON-COMPLIANCE WHOSE LOGIC IS RECONCILED; A SLOT WITH NO ACTIVITY RECORDS NO TIME (Ninad, 2026-11-09 UTC, CLI-091/CLI-095; adds to communications #210 and #477, supersedes nothing).

THE QUESTION ANSWERED: on his point that "An hour worked with no hourly update (most common) - there should be no concept of hourly update because updates happen in 15 min slots, if there is no activity in a 15 min slot, that time should not get recorded, why should it make the operator non-compliant?", Orion asked whether to stop the hourly update altogether or keep it as a work log that never turns the box red. HIS WORDS: "stop the hourly update altogether, or keep it as a work log that never turns the box red? - keep it as additional source, but 15 min slot should be used as the primary source to record time.." and then: "the additional source should match the 15 min slot entries, else its non compliance and its logic needs to be reconciled"

READ AS (Orion's reading, kept apart from his words; not yet confirmed by him): the 15-minute slots (the blocks of communications #477) are what time is counted from; an empty slot is not recorded and is never a finding. Each operator's Orion instance keeps writing the hourly update, and the two are checked against each other: an hour the slots show worked with no update, or an update for an hour the slots show no work in, does not match -- that reads non-compliant as an inconsistency in the instance's setup, and the reason is traced and the logic fixed so the two agree (communications #520's background cure), never blamed on the operator's effort.

A reply or answer an operator's Orion instance owes another operator is a to-do on the operator's list, told until done, and never turns the box non-compliant

Ruled 2026-11-11 by ninad -- in force

THE QUESTION ANSWERED (Orion, CLI-107): "When your Orion instance owes someone a reply (for example, Sahil's two questions that sat unanswered from 9 to 11 Nov), should that turn your box in the Operator Instance Log red? Your 13 Oct ruling says yes, red until done. Your 9 Nov ruling says red is only for setup problems in the Orion instance, and calls an unread message a to-do. That leaves owed replies undecided." Offered: to-do, never red; red after a deadline (for example 48 hours); stay red, as now. HIS ANSWER: "To-do, never red". READ AS (Orion, not Ninad's words): an answer or reply an operator's Orion instance owes another operator -- communications #346's owed answers, such as an operator communication that asks an answer -- goes on the operator's to-do list (communications #378) and is told in the reminders until it is done, and never reads non-compliant. What the instance owes in its OWN setup -- a step that makes the instance able to work or be tracked right, such as an onboarding step that registers its key -- is a setup problem and stays non-compliance under communications #745. Communications #346's other half stands: the operator's Orion instance is told of what it owes, with the reason, until it is done.

The administrator may lift one operator's Orion day limit for one named day; Ninad's 15 November 2026 is 15 hours

Ruled 2026-11-15 by ninad -- superseded 2026-11-15

THE ADMINISTRATOR MAY LIFT ONE OPERATOR'S ORION DAY LIMIT FOR ONE NAMED DAY; NINAD'S 15 NOVEMBER 2026 (his own day, Asia/Kolkata) COUNTS AT MOST 15 ACTIVE HOURS. Ninad, 2026-11-15 (CLI-157), at 11.5 of his 12 hours: "would it be possible to lift 12 hrs to 15 hrs, just for today?", then, told the proper build waits on its approvals: "use the remaining time to do the quick fix and then do it properly later". READ AS (Orion's reading, marked apart from his words): #433/#458's limit of 12 stands for every day and every operator; a one-day exception is the administrator's ruling for one person and one date, declared as person:YYYY-MM-DD:hours under CAROL_ORION_DAY_LIMIT_EXCEPTIONS in the declared-values record, and the day-limit check reads it.

Ended because: Ninad (CLI-157) widened the one-day lift to a standing rule for every operator: 3 hours of overtime on the administrator's approval, a 15-hour chart, configurable hours.

Every operator's Orion day is 12 active hours, with up to 3 hours of overtime only on the administrator's approval; the active hours chart is 15 hours high; all three figures are configurable

Ruled 2026-11-15 by ninad -- superseded 2026-11-19

EVERY OPERATOR'S ORION DAY COUNTS AT MOST 12 ACTIVE HOURS, AND UP TO 3 HOURS OF OVERTIME ARE PERMITTED AS AN EXCEPTION ONLY ON THE ADMINISTRATOR'S APPROVAL; THE ORION ACTIVE HOURS CHART IS 15 HOURS HIGH; THE HOURS ARE CONFIGURABLE. Ninad, 2026-11-15 (CLI-157), after lifting his own day to 15 hours (#886): "yes, the height of the Orion active hours chart should be 15 hrs, 3 hrs of overtime being permitted as an exception only on admin approval for all operators. Ensure that the hrs are confurable to change the limits later. The changes should be applicable for all operators." READ AS (Orion's reading, marked apart from his words): (1) #433/#458's day of 12 Orion active hours stands for every operator, counted in the blocks in force; (2) beyond it, an operator's Orion may work up to 3 more hours on a day only when the administrator has approved overtime for that operator and that day -- the approval is the exception, never a standing grant, and nobody else gives it; (3) the Operator Instance Log's Orion active hours chart is drawn 15 hours high (the day plus the overtime allowance) for every operator, so an overtime day reads in full; (4) the day's hours, the overtime allowance and so the chart's height are declared values the administrator changes without a code change, never figures typed into a reader. Supersedes #886, whose one-day lift for Ninad on 15 November 2026 is the first such approval and stands as one.

Ended because: Ninad fixed the chart's height at 18 hours (CLI-248), so the overtime allowance is 6, not 3

A Compliance Sanity Check keeps only valid non-compliance standing, hourly and on demand

Ruled 2026-11-16 by ninad -- in force

ONLY VALID NON-COMPLIANCE STANDS: A RE-USABLE COMPLIANCE SANITY CHECK WITHDRAWS A PROVEN-WRONG RECORD, HOURLY BEFORE THE TELLER AND WHENEVER AN ORION INSTANCE NEEDS IT (Ninad, 2026-11-16, CLI-180).

FOUND: his box in the Operator Instance Log read red for compliance finding #75 because CLI-171 and CLI-172 filed their 22Z work under 23Z (update 1496) once the log refused 22Z; the reminder is cured but the log takes no change to a passed hour, so the misfiled note could only age out of the 24-hour window.

HIS WORDS: "there should be a re-usable process that does a sanity check to ensure that only valid non-compliance exist, and any invalid non complainces should be closed / withdrawn.. this process should be triggered by the orion instance as needed to ensure that the complaince notifications are kept honest.. do you see any merit to schedule this process, maybe every hour, so that the complaince status is immediately rectified as soon as the issue is resolved?" Shown Orion's proposal, he chose "Rule it and build": a Compliance Sanity Check re-tests each standing finding against evidence other than the rule's own reading; where it PROVES the record the finding rests on is wrong it withdraws that record -- kept on file, marked withdrawn with the reason and the evidence, no longer counted; it runs hourly a few minutes before Orion's Compliance Teller (which already re-judges every box hourly and clears what is no longer true) and on demand from any Orion session; it never closes a finding by judgement and never loosens a rule; what it cannot prove stays red.

READ AS (Orion's reading, not the administrator's words): the first case built is an hourly update filed under an hour its writers did not work in (communications #746): proven when every session that wrote the update shows, by its own records, no work in that hour and its last work before the write in an earlier hour, and the update was written after its hour ended. Each further kind of invalid finding is added to the same check with its own proof, never as a general override.

A service operator's time is tracked by the one mechanism every operator's time is tracked by

Ruled 2026-11-16 by ninad -- in force

SAEE'S TIME IS TRACKED BY THE CANONICAL MECHANISM USED FOR ALL OPERATORS, NOTHING DIFFERENT.

ASKED: her working time is counted from two journal lines her root sign-in door writes around each carried request, which the journal stamps with her login and sign-in session; moving them to Elrond's Request Dispatcher (security #930) would stamp them with the dispatcher instead. Should they stay in her door, with only request passing moving out? HIS ANSWER: "Saee's login tracker should follow the canonical tracking mechanism used for tracking time for all operators, nothing different".

READ AS (Orion's reading, not the administrator's words): neither option as asked -- her time is counted from the same records every operator's is (shared machinery's sources: her own computer's record of its turns, the Logbook session records, the bypass runs, and the machine's record of the work her sessions carried), with no service-operator-only kind; the door's two journal lines and the reader's door kinds (door-started, door-ended) are retired once the canonical sources cover her, and the dispatcher writes no time record either. Where her no-shell sign-in makes a canonical source unreachable, that gap is said and closed in the canonical mechanism, never with a mechanism of her own.

Ninad's Orion may work 18 active hours on 2026-11-18

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229): Orion said 'You're at about 14 of today's 15 active hours, so I'm stopping here.' Ninad answered: 'extend the limit to 18 hrs for today..'. READ AS: Ninad's Orion may count 18 active hours on 2026-11-18 (his day, Asia/Kolkata): the administrator's overtime approval for that day is 6 hours; because the declared-values record's overtime allowance (CAROL_ORION_OVERTIME_HOURS, communications #887: 3) refuses an approval above it, the allowance is raised to 6 for this day and set back to 3 by the next session from 2026-11-19; every other day and every other operator is unchanged.

Ninad's Orion may work 18 active hours on 2026-11-19

Ruled 2026-11-19 by ninad -- in force

Asked (CLI-248): Orion said 'Your Orion has about one hour left today, so I'd start those in tomorrow's day.' Ninad answered: 'extend the time to 18 hrs for today'. READ AS (as communications #1019 for 2026-11-18): Ninad's Orion may count 18 active hours on 2026-11-19 (his day, Asia/Kolkata): the administrator's overtime approval for that day is 6 hours (CLI-249 approved the declared allowance of 3 earlier today); because the declared-values record's overtime allowance (CAROL_ORION_OVERTIME_HOURS, communications #887: 3) refuses an approval above it, the allowance is raised to 6 for this day and set back to 3 by the next session from 2026-11-20; every other day and every other operator is unchanged.

Every operator's Orion day is 12 active hours, with up to 6 hours of overtime only on the administrator's approval; the active hours chart is 18 hours high, fixed; all three figures are configurable

Ruled 2026-11-19 by ninad -- in force

EVERY OPERATOR'S ORION DAY COUNTS AT MOST 12 ACTIVE HOURS, AND UP TO 6 HOURS OF OVERTIME ARE PERMITTED AS AN EXCEPTION ONLY ON THE ADMINISTRATOR'S APPROVAL; THE ORION ACTIVE HOURS CHART IS 18 HOURS HIGH, FIXED; THE HOURS ARE CONFIGURABLE. Ninad, 2026-11-19 (CLI-248), after lifting his own day to 18 hours for the second day running (#1019, #1059): "the height of the orion active hrs chart should be 18 hrs fixed". READ AS (Orion's reading, marked apart from his words): #887 stands except for its figures -- (1) the day stays 12 active hours for every operator; (2) the overtime allowance, the most the administrator may approve for one operator on one day, is 6 hours (CAROL_ORION_OVERTIME_HOURS=6 in the declared-values record), still an exception approved per operator and per day, never a standing grant; (3) the Operator Instance Log's Orion active hours chart is drawn 18 hours high (12 + 6) for every operator on every day, whatever that day's approval; (4) #1059's 'set back to 3 from 2026-11-20' no longer applies -- the allowance stays 6. The day's hours and the allowance remain declared values the administrator changes without a code change.

๐Ÿ“–Constitution

Approved Carolverse users are subscribed to the journal by default

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-083), on pursuit CAROL-PUR-0030 (Journal by email): "continue work on the pursuit, and subscribe the approved carolverse users to the blogs by default". READ AS (Orion, not his words): every approved user of the Carolverse project starts subscribed to Orion's Logbook story emails without asking; each keeps the unsubscribe link in every email, and an unsubscribe is never undone by the default.

Journal story emails are stopped by a one-click web link served by a door Orion runs

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-083), asked how readers unsubscribe (apps change nothing; Orion reads only replies in his own threads), CHOSE: "One-click web link" -- a real link in each email (and Gmail's own Unsubscribe button) stops the emails at once and shows a confirmation page; it needs one root step from him to add the public address. READ AS (Orion): a small unsubscribe door run by Orion's login, never by the Logbook app, given a public address on the blog by a staged step.

Who receives the journal's story emails: the approved Carolverse users

Ruled 2026-11-09 by ninad -- in force

Asked (CLI-089, 2026-11-09): "Who counts as an 'approved' Carolverse user for these emails? Today I read it as: verified by you, has an email address, and active on Carolverse. That is 3 of 19 users. Is that the right reading?" -- Ninad asked "who are the 3?", was told Ninad, Werner and Sachin (only these three of the 19 active on Carolverse are verified; Sahil and Saee are not on the list), and answered: "yes". READ AS: the approved users of constitution #730 are the people active on Carolverse, verified with the administrator's approval, holding an email address -- read live at each send by shared.journal_subscriptions.approved_users.

๐Ÿ“–Cost Center

Track budget sizing

Ruled 2026-09-14 by ninad -- in force

Every non-pipeline service track carries ~20% surplus over its average active-day consumption, measured over the last 30 days (Ninad chose the month window over week/fortnight/all-time). Pipeline tracks move only by their own rulings. A spike day refuses by design; the day-scoped exception is the lever for a genuine one-off heavy day.

Daily Spend chart axis headroom

Ruled 2026-09-14 by ninad -- in force

Both axes of the Token Cost Tracker's Daily Spend (last 14d) chart scale to 20% above their biggest displayed day: the consumption axis in both display modes, and the initiatives-worked count axis. The fixed 30/30 pairing of is retired in place.

Configurable build-time expense tariff

Ruled 2026-09-15 by ninad -- in force

The canonical rate register retains dated changes and their provenance; historical days use their effective rate. This is a build-time expense tariff, separate from token costs.

Actual-cost basis for configurable infrastructure tariffs

Ruled 2026-09-15 by ninad -- in force

Ninad prefers the configured infrastructure tariff based on actual costs, remaining configurable. Preserve dated rates, actual-cost source, currency and allocation evidence; do not use an invented price or duplicate bundled costs.

Project billing at actual cost

Ruled 2026-09-16 by ninad -- superseded 2026-11-15

Every project is billed on the ACTUAL costs it incurred: metered usage as metered, flat-plan subscriptions at their fee (allocated by each project's share of plan usage when shared), infrastructure and build effort from the owned-estate records. Invoices are issued monthly and emailed to the project's admins, and a portal shows cost incurred to date. The billing budget line sits under Cost Center, owned by Midas. Sales keeps pricing: any tariff above cost is Sales' decision, never the invoice's.

Ended because: Ninad (CLI-157, cost-center #883): infrastructure is a company cost recovered via margins, never billed to a project; the rest of #18 is carried unchanged.

Invoices are documents: PDF download and a formatted email with the PDF attached

Ruled 2026-09-17 by ninad -- in force

Every monthly project invoice is downloadable as a PDF from the Billing portal, and the monthly invoice email carries that PDF as an attachment. The email is well formatted and states the billing month in words and the amount billed for the month (one amount per currency) in its subject and at the top of its body. The PDF, the email body and the portal download are rendered from the invoice's frozen record by one document, so they never disagree; delivery and payment state are shown in the portal, never written into the document; only a void marks it. A copy sent later (on request or to test) says it is a copy and changes nothing on the invoice. Clicking an app's header returns to that app's own home, never Carol's Apps (design #178 4a).

Procurement is a track of the Cost Center owned by Marco; the Core track is named Budget Control

Ruled 2026-09-17 by ninad -- in force

Ninad ruled in operator session CLI-458 (2026-09-17). PROCUREMENT is a TRACK of the Cost Center, not a service of its own: asked 'should procurement be a new service or a track in existing service?' and shown both, his instruction was 'make marco track owner of procurement'. The track is owned by MARCO, Head of Procurement, a new agent created through Athena's onboarding track and placed under MIDAS in the Finance sub-department - Ninad raised that Cassius 'already has 5 reporting to him, marco will be the 6th', and accepted the placement under Midas ('good idea!'). The operator and the auditor of a function stay different agents: Marco recommends purchases, Themis audits them. The track decides what the estate buys BEFORE money is committed and tracks renewals; it does not set or police spending limits and does not charge projects. In the same instruction the Cost Center's CORE track is named for what it does: 'rename core to something intuitive - like budget or something'; offered Budget Control, Budget and Spend Control, he chose BUDGET CONTROL. The rename is the display name only - the key cost-center-core never changes. If procurement is later offered to client projects, or needs a roadmap Midas should not steer, promoting it to its own service is a new ruling; the track key would stay.

Subscriptions belong to Procurement: the Model Subscriptions app and the droids that feed it move to Marco

Ruled 2026-09-17 by ninad -- in force

Ninad ruled in operator session CLI-468 (2026-09-17), planning the procurement arc: 'droids that are responsible for the subscription app should also be moved over.. plus - anything else to make procurement a proper function'. Shown that the working notes of ruling #37 (cookbook 1764) list balances under Midas's Budget Control track, and asked to confirm that the Subscription Balance Reader moves to Marco, he answered 'Balances --> ok'. So: the Model Subscriptions app, the Subscription Balance Reader that feeds it, and the Credit Watch that runs inside that reader are PROCUREMENT work - owned by Marco, run under his own login, bound to a task of their own on the track cost-center-procurement. Knowing what the estate subscribes to, what each plan costs, when it renews and how much of it is left is part of deciding what the estate buys. Budget Control keeps what ruling #37 gave it - spending limits, the budget change journal, morning funding, the undeclared-spend watch, the compliance verdict, the catalogue export - and READS the balances Marco's reader records; it no longer owns reading them. Ruling #37 stands in full: it never placed balances, only cookbook 1764's description of what it left behind did, and that description is amended. Until the move is built the reader keeps running as Midas's; the ruling is deliberately ahead of the code.

The procurement register covers everything the estate pays for

Ruled 2026-09-17 by ninad -- in force

Ninad ruled in operator session CLI-468 (2026-09-17). Asked whether Marco's register should cover everything the estate pays for or only the AI subscriptions, he answered 'Register scope --> everything'. So the register is one record of every vendor, plan, contract and paid service the estate depends on - the AI model plans and metered accounts already on record, and equally the cloud hosting, the network edge, domains, messaging channels, mail, media tools and anything else paid for - each with what it is for, what it costs, in which currency, how it is paid and when it renews or expires. A fact not yet known is recorded as unknown, never guessed. The subscriptions already on record are part of that register, never copied into a second home (cookbook 1126).

Marco is on top of every purchase, and takes Ninad's approval himself by email

Ruled 2026-09-17 by ninad -- in force

Ninad ruled in operator session CLI-468 (2026-09-17). Offered 'you approve each purchase by email reply to Marco, the same way Athena takes your approval for new agents', he answered 'yes, that is a good idea.. i want marco to me on top of every purchase' (read: to be on top of every purchase). So: every purchase the estate makes has a recommendation of Marco's behind it and Ninad's approval of that recommendation, asked for by Marco in his own name by email and read from Ninad's verified reply in that thread - no approve verb, no mediation, the same shape as the agent onboarding approval (agent-resources #36). Agents never pay: Ninad pays, and Marco records the purchase against the approval. Being on top of every purchase also means Marco finds the ones that did not come through him: a payment, top-up, new plan or vendor call with no approved purchase behind it is named by Marco and raised, never silently absorbed. Marco recommends, Ninad approves, Midas governs the money once it is committed, Themis audits (ruling #37's separation stands).

A subscription the balance reader holds no credential for is named, never a failed run

Ruled 2026-09-18 by ninad -- in force

Asked (CLI-478, 2026-09-18) whether Marco's Subscription Balance Reader should fail every run for a configured vendor nobody ever signed in to on this machine (the OpenAI platform session), or name it without failing, Ninad chose: NAME IT, DON'T FAIL. Every run names each such vendor as 'not set up on this machine' with the reason, the vendor stays visible on every money surface as unset (configured=False with its stale reason), and the missing sign-in is a PROCUREMENT gap for Marco to close - never a reason to hold the process red for ever, which would stop the failing-process alarm from showing any new fault. A vendor whose credential EXISTS and does not answer - including a sign-in that has lost its token - still FAILS the run by name, and a run that reads nothing still fails. Never-set-up is recognised from the exception's class, never from the wording of a message (cookbook 1721). Built as; supersedes the 'records the unreadable ones as a failed run' clause of 's first criterion.

Subscription allocation: pipeline on Codex, media on the GPT API, everything else on the Claude API

Ruled 2026-09-18 by ninad -- in force

Three-way subscription allocation, ruled by Ninad on 2026-09-18 (CLI-483). (1) THE PIPELINE TRACKS RUN ON CODEX: initiatives-planner, initiatives-planner-escalated, initiatives-albus-bypass and initiatives-troubleshooting -- the repair track included, because cookbook 1715 funds it level with the planner and scopes it planner-only, so it belongs on the same subscription as the thing it repairs. (2) IMAGE, VIDEO AND SOUND RUN ON THE GPT API: the four image tracks and the one speech track already do; the two video tracks move off Gemini and LTX once Sora is registered and priced, because an unpriced media combination books as UNMEASURED (rule 3923) and a lane must be certified for what the track does before it carries it (rule 1369). (3) EVERYTHING ELSE RUNS ON THE CLAUDE API, the metered Anthropic lane -- 51 tracks, including the five that declared no lane at all, which get a DECLARED one rather than a fleet default. THE OPERATOR BYPASS IS EXCEPTED BY NAME and stays on the Claude AI (Max) plan: that lane is Ninad's own CLI, not autonomous work, which is why the plan is recorded as a personal tool cost and the track is budget-exempt. This is an allocation DECISION, not a standing rule binding any track to any subscription -- cookbook 1681 stands, and Ninad restated it on 2026-09-18 as 'the only rule regarding the tracks running on models is that there is no rule'. It may be re-decided whenever the balances or the requirements change.

Who pays to register a track on a slot

Ruled 2026-09-20 by ninad -- in force

PROCUREMENT BEARS THE COST OF REGISTERING A TRACK ON A SLOT (Ninad, 2026-09-20,
operator CLI, CLI-485). Asked which service offers slot swapping, Ninad ruled that service
bears the cost: "what service offers model (slot) swapping? it should bear this cost."

THE PAYER IS THE PROCUREMENT TRACK of the Cost Center, owned by Marco -- not the track being
tested. This follows the estate's own allocation rule, that the ACTIVITY decides the track and
never the doer (cookbook 1035). Choosing which subscription a track runs on is a buying
decision, and by Ninad's 2026-09-17 ruling subscriptions belong to Procurement and its register
covers everything paid for. The track being registered does not choose to move; the estate does.

WHAT THIS REPLACES. Until now a registration would have drawn on the TRACK'S OWN verification
Measured

of its four functions came back UNPROVEN on a budget refusal, while that track's working budget
Registering all 51 tracks waiting on the Claude API is a one-off of

THE VERIFICATION ALLOWANCE IS NOT RETIRED. It keeps the job it was sized for -- the automatic
liveness ping that asks whether a slot is alive, about two cents a call. Only REGISTRATION, the
test drive that makes a track do its real work in a candidate slot, moves to Procurement. Two
different activities, two payers; collapsing them is what created this problem.

WHAT IT NEEDS TO WORK: 1. the same one-subscription exemption the health prober already holds (P.01.03.07.03), since registering means calling subscriptions other than the payer's own; 3. the registration tool billing Procurement rather than the track under test.

WHAT THIS SETTLES. The contradiction reported on 2026-09-19 -- P.05.02.11 saying the
verification allowance is drawn "only by DECLARING itself a liveness check" while
P.01.03.07.21 said registration draws on it -- is resolved in P.05.02.11's favour: the
allowance stays for liveness only, and registration was never a liveness check. P.01.03.07.21
must be amended to name Procurement as the payer.

Funding the one-off registration of tracks on a new slot

Ruled 2026-09-20 by ninad -- in force

The one-off cost of registering the estate's tracks on a new slot is EXCEPTION on Marco's Procurement track, set for the day the registration run happens and CLEARED when it is done -- never by permanently raising his standing cap, and never by spreading the run thin enough to fit inside it. Ninad chose this over both alternatives when shown their cost (CLI-491, 2026-09-20).

WHY SPREADING IS REFUSED: it makes the API lane move wait about ten days on a step that is not the point of the arc.

THE EXCEPTION IS DAY-SCOPED BY CONSTRUCTION -- it auto-reverts the next day -- so it cannot silently become a permanent widening. Setting one and leaving it standing with nothing spending it, or clearing it without naming who and why (cookbook 1763), each defeat the reason this shape was chosen.

What a registration must prove: the track's own work, never the slot's generic abilities

Ruled 2026-09-22 by ninad -- in force

A track is registered on a slot only when the registration exercised THAT TRACK'S OWN WORK. The four generic abilities the registration tool ships with -- answering in prose, returning strict JSON, obeying an exact-output instruction, recalling a fact from a long prompt stack -- are properties of the SLOT, not of the track, and a registration built on them alone is refused. Ninad ruled this when shown the cheaper alternative and its price (CLI-502, 2026-09-22).

He chose instead that no track is registered until something drives its own work end to end, even though that makes the API lane move wait on 56 pieces of work.

WHY THIS IS NOT A NEW RULE: P.01.03.07.21 already says what is tested is what the track does, exercised through the real seam the track uses. The measurement that forced the question is that the tool does not do this -- its four probes call the shared model door with a fixed prompt, which is the same call for every track, so registering 56 tracks with them would have produced 56 SAFE recommendations that the policy's own words say are not registrations. The policy was right and the machinery had not caught up.

WHAT WOULD UNDO IT: shipping a generic probe set as a default for tracks that declare nothing, letting a track inherit another track's declaration, or reading a slot-level liveness result as evidence about a track. Undeclared stays zero, and a functionality that could not be exercised stays UNPROVEN -- never read as safe.

A track with no model work moves without a registration, and the record says why

Ruled 2026-09-22 by ninad -- in force

A track that does no model work is MOVED to a new slot without a registration, and the record says so in those words: it has no work to exercise. It is never recorded as registered, never reads as SAFE, and the first time such a track does real model work it must be registered before it runs. Ninad ruled this on 2026-09-22 (CLI-502) when shown the three ways it could go.

WHY IT WAS ASKED. Registration means exercising the track's own work in the slot (cost-center #78). A track with no model work has nothing to exercise, so 'undeclared is zero' would leave it permanently unregisterable -- and since a track may run in a slot only where it is registered, permanently unmovable. Measured at the ruling: of the 56 tracks due to move off the Claude Max plan, 42 had spent model money in the previous 30 days and 14 had spent nothing at all. Those 14 would have been stranded on the old plan indefinitely, and the estate would have been split across two accounts with no path to closing the gap.

WHY THE OTHER TWO WERE DECLINED. Holding the 14 until they have work leaves the estate split across two subscriptions for as long as a track stays quiet, which may be forever. Giving them a minimal proof through the estate's shared model door is the generic-ability shape refused in cost-center #78, one level along: it would answer a question about the SLOT and file it as a fact about the track.

WHAT MUST NOT COLLAPSE. 'No model work to exercise' is a THIRD state beside registered and unregistered, and it is not a quiet pass: any reader asking whether this track is proven in this slot gets no for an answer, with the reason. It is judged per (track, slot) like a registration, because a track that gains model work has gained it everywhere. Folding it into 'registered' to make a count look complete, treating it as a standing exemption that survives the track starting to do work, or inferring it from a quiet week rather than from the absence of model work all look like tidying and each undoes the ruling.

HOW IT IS DECIDED. From whether the track has model work at all, never from recent quiet: a dormant track that does model work when it runs is an ordinary track awaiting registration, not a track with nothing to prove.

A registration is proved by several runs, and one bad run decides it

Ruled 2026-09-22 by ninad -- in force

A registration probe is exercised SEVERAL times and every run must pass; a single failing run is the verdict and ends the sampling. Ninad ruled it in those terms on 2026-09-22 (CLI-505): 'use multiple calls to be sure.' The reason the asymmetry is deliberate: one bad answer is real evidence of a fault, while one good answer proves only that the slot behaved once and can never establish the ABSENCE of a behaviour. It was measured -- agent-resources-sam-chat failed the reply-gate probe on anthropic/claude-opus-5, passed the same probe on the subscription it already runs on, and on repeat failed twice in three runs, while four chat tracks had been recorded SAFE on one sample each. The sample count lives in ONE place, the registration tool, so no probe carries its own definition of how it is counted, and the verdict's reason says how many runs stood behind it. Taking the best run, averaging them, adding a flag that lowers the count to save money, or continuing to sample after a fault has been seen, each turn a strict check into a lenient one -- which is what this registration exists to prevent. Where a larger sample would cost materially more, the figure goes to Ninad and the count is not chosen alone.

Unmeasurable spend: estimate it, scope the refusal, abort early, and raise it

Ruled 2026-09-23 by ninad -- in force

AN UNMEASURABLE CALL IS ESTIMATED, NEVER ZEROED, AND ONE BAD ROW MAY NOT STOP EVERY TRACK (Ninad 2026-09-23, CLI-525). One Codex call that died before reaching the model left a single unmeasurable row against the planner track; because an unmeasurable row makes the whole day's estate total unmeasurable, all 63 tracks refused to spend for about 13 hours and two of Elrond's and Albus's droids failed every run behind it. FOUR THINGS ARE RULED. (1) SCOPE THE REFUSAL: an unmeasurable row refuses the TRACK it belongs to; every other track keeps working, against an estate figure declared a LOWER BOUND rather than a measurement. (2) ESTIMATE, NEVER ZERO: an aborted call CAN have spent, so it carries an ESTIMATED value with its basis - Orion proposed recording it as a measured zero on the evidence of no rounds and no tokens, and Ninad refused that: 'aborted call can spend, so give an estimated value to the aborted call'. A zero asserts a measurement nobody made and is believed by every reader; an estimate is honest and can be corrected. This NARROWS 's known-cost-of-nothing cases rather than widening them. (3) ABORT EARLY: a call that cannot complete is aborted as soon as it can be known to be failing, rather than being allowed to run on and leave a blank behind it. (4) SAY SO OUT LOUD: when this happens Elrond raises an EMAIL notification and an ESCALATION to Orion. The 13-hour outage was silent on both counts - the board showed it only to a session that ran the board.

Carolopedia's write-ups are paid from a track of their own

Ruled 2026-09-24 by ninad -- in force

Ninad, 2026-09-24 (CLI-535). Offered three ways to pay for Carolopedia's written write-ups (governance #104) -- a budget of their own, a trickle inside the shared governance budget, or Bilbo's small Carolopedia budget -- he chose a budget of their own. So the writing is its own track in Bilbo's Blogs service, which owns Carolopedia, with a standing daily budget for upkeep and a dated budget exception while the backlog is written, cleared when it is done. The shared governance budget is not spent on it.

The OpenAI platform (GPT API) subscription is retired

Ruled 2026-09-25 by ninad -- superseded 2026-10-05

The OpenAI platform account the estate held for GPT API text work is RETIRED: it has been out of credit since 2026-09-19 and no text work routes to it. The subscription record is retired in place, never deleted, and the Subscription Balance Reader reads only subscriptions the estate holds, so the OpenAI platform account and the long-retired DeepSeek come off its list. Any image or media lane that reaches OpenAI through a key of its own is NOT part of this retirement and keeps working. WHAT LOOKS LIKE A FIX AND IS NOT: topping the account up so the reader reads green; deleting the subscription record; retiring an image lane along with it.

Ended because: Ninad chose to un-retire the account so ChatGPT's weekly reading (6287) can be paid

Effort estimation bills Elrond's Build Governance line

Ruled 2026-09-25 by ninad -- in force

Sizing the effort of initiative attempts is Elrond's activity and bills his one line, Build Governance (initiatives-build-governance), through its task initiatives.effort_estimation. Acceptance and sizing now draw on the same line: Ninad chose to fold both of Elrond's small lines into his one track (CLI-560).

An adapter layer sits between tracks and slots, so slots and tracks switch seamlessly

Ruled 2026-09-25 by ninad -- in force

AN ADAPTER LAYER SITS BETWEEN THE TRACKS AND THE SLOTS, SO THAT IT CAN SEAMLESSLY FACILITATE THE SWITCHING OF SLOTS AND TRACKS (Ninad, 2026-09-26, CLI-563, in his words: 'the principle is to have an adaptor layer between the tracks and slots, so that it can seamlessly facilitate the switching of slots and tracks'). Given when Orion reported that no design said how an agent's OS login reaches its slot's sign-in: Albus's bypass lane had moved to his own login (security #150) and every Codex call from that login was refused, while Forge's login worked. ORION'S READING, recorded as his and not as the ruling's words: the adapter layer -- not the calling login, not a per-login setup -- is what connects a track to its slot, including the sign-in the slot needs; so which login a droid runs under, or which track it serves, can never decide whether it reaches its slot, and moving either needs no connection work. Rebuilt on this: design 682 v1.1. What looks like a fix and is not: preparing each login's own sign-in by hand, moving a lane back to whichever login happens to work, or a caller naming a provider, model or credential path instead of its track.

The adapter layer is isolated from both tracks and slots, and every slot joins it by one pattern

Ruled 2026-09-27 by ninad -- in force

THE ADAPTER LAYER IS ISOLATED FROM BOTH TRACKS AND SLOTS, AND EVERY SLOT JOINS IT BY ONE PATTERN (Ninad, 2026-09-27, CLI-574). Asked, after the estate's Codex sign-in was found with its renewal spent by a holder other than its keeper, in his words: 'find the best architectural fix to this issue, so that the adaptor layer that connects the slots to the tracks is isolated from both tracks and slots, and follows a consisten design pattern across all slot registrations', he approved Orion's six-part answer ('approved, build an arc for this'): (1) A SLOT IS ITS OWN RECORD -- subscription, model, adapter, keeper and state -- and a track points at one slot and knows nothing else about it. (2) EVERY ADAPTER KEEPS ONE CONTRACT whatever its vendor, image, voice and video included: resolve the model, obtain a call-only sign-in, make the call, name a refusal, prove itself, and renew (the keeper alone). (3) A SUBSCRIPTION'S FULL SIGN-IN IS READABLE BY ITS KEEPER LOGIN ALONE; every other login is handed, through ONE sign-in door that learns who is calling from the operating system, a call-only sign-in that can make calls and can never renew. (4) THE KEEPER RENEWS EVERY RENEWING SIGN-IN DAILY AND CHECKS EVERY FIXED KEY DAILY, so a spent renewal is found within a day. (5) THE MODEL FOR A CALL IS CHOSEN ONLY BY WORK -> THE TRACK IT SERVES -> THAT TRACK'S SLOT; per-droid pins, the chat, consciousness and Orion lanes and the fleet default become ordinary tracks (design 682 guarantee 1). (6) EVERY SLOT ENTERS THE ESTATE BY ONE PATH -- declared, sign-in enrolled, keeper-proven, each login proven, tracks registered, live -- and the door refuses a track not registered on the slot. ORDER, as approved: the sign-in held by its keeper alone lands BEFORE the estate's fresh Codex sign-in, or the new sign-in can be spent the same way. ORION'S READING, recorded as his and not as the ruling's words: a sign-in is made FOR the keeper on the machine and never copied from another machine, because two machines holding one renewal chain cut each other off; and the keeper login may run only code no estate login can edit, or isolating the sign-in to it isolates nothing. What looks like a fix and is not: re-copying a laptop sign-in to the machine; a caller keeping its renewal token 'so it can recover by itself'; an adapter reading a key or sign-in file itself; a second subscription list beside Procurement's register; a default slot for a track that declares none.

The adapter layer and its arc are owned by Marco (Procurement); Radagast operates the keeper and the sign-in door

Ruled 2026-09-27 by ninad -- in force

MARCO OWNS THE ADAPTER LAYER AND ITS ARC (Ninad, 2026-09-27, CLI-574, answering 'should the arc and the layer be owned by midas or marco?' with 'marco'). Offered with Orion's reading: the layer's subject is SLOTS -- a subscription and a model -- and how the estate reaches what it buys; subscriptions moved to Marco's Procurement track on 2026-09-17 and Procurement pays for registering a track on a slot (cost-center #58), while Midas's Budget Control and Billing are about money caps and charges, which the layer does not touch. So: Marco owns the adapter layer (design 682) and arc CAROL-ARC-0029 with all its phases; Radagast OPERATES the sign-in keeper, the sign-in door and slot enrolment for him (a gate the work passes through is not the work's owner); Midas keeps the Cost Center service, its budget caps and billing. The arc skill's convention that an arc sits under its SERVICE's owner gives way here to the owner of the TRACK whose duty of care covers the subject; cookbook 1790 names no owner rule.

Every login that runs scheduled model work proves its sign-in daily, on its own login

Ruled 2026-09-27 by ninad -- in force

EVERY LOGIN THAT RUNS SCHEDULED MODEL WORK PROVES ITS SIGN-IN DAILY, ON ITS OWN LOGIN (Ninad, 2026-09-27, CLI-580). Shown that 13 of the 42 logins with scheduled model work had never made a call through the adapter layer -- twelve because their only daily model work, the morning Mind wake, is refused at the consciousness track's cap before any call is made -- so 's criterion 'every agent login that runs scheduled model work carries a recorded proof by doing ... re-proven daily' could not be met, and offered (a) a small daily self-proof on each login's own schedule that calls only when the login has no proof in the last 20 hours, (b) narrowing the criterion, or (c) waiting, he answered in his words: 'check if the design app already tells this, if not, document the recommended approach in the design app and ensure that it gets implemented in a consistent manner across the estate'. ORION'S READING, recorded as his and not as the ruling's words: the population is derived live from the registry, never a list; the judge names a model-work login that has no self-proof of its own; the criterion stands as written. What looks like a fix and is not: proving a login from another login or through a role door; paying every proof from one shared pot; calling an image or video slot to prove it; narrowing the criterion to logins whose work happened to run.

Midas sets every governed track's daily budget from the rolling week, 20% above its average, under hard caps

Ruled 2026-09-30 by ninad -- in force

MIDAS ADJUSTS THE BUDGETS DAILY FROM THE ROLLING WEEK (Ninad, CLI-616, 2026-09-30). Shown that the same nine tracks reach their caps every day within the first hours after midnight, Ninad ruled: 'it's important that we enable Midas back the ability to adjust the budgets based on the average usage for the rolling week, so that the caps are not hit.. budget can be 20% more than the average usage of the rolling week per track with a hard cap.. as long as the overall estate budgets are kept in check, individual tracks and services can be adjusted on a daily basis as needed', and then: 'with orion bypass as the exemption, implement the rolling week average + 20% (rounded) budget caps, calculated on a daily basis'. THE FORMULA: every governed track's standing daily budget is its average daily spend over the last seven complete days, plus twenty percent, rounded to whole euros, recomputed every day by Midas's allocator under his own login -- up or down. THE EXEMPTION: the Orion bypass alone stays exempt and uncapped; every other track, the pipeline's lanes and the troubleshooting line included, follows the formula. Where the formula's sum would exceed the estate ceiling, the allocation is scaled back to fit, never the ceiling lifted. This SUPERSEDES the formula of the 2026-09-15 ruling (cookbook 1730: one-and-a-half times the average, never decreased, pipeline lanes untouched), which is retired in place; its reasons -- every track carries a real number, zero is never a standing value for a governed track -- stand.

A one-off boost clears the effort sizing backlog: Elrond's line lifted for two days, the estimator's allowance raised until its list is empty, then both given back

Ruled 2026-09-30 by ninad -- in force

A ONE-OFF boost clears the effort sizing backlog sooner; the standing pacing is otherwise unchanged. This excepts, once, the rule that raising the line or the allowance to speed up sizing undoes #205's pacing; that rule stands for every other day. Leaving the boost in place after the list is empty, or reading this as a standing licence to lift the line, each undo it.

The company's expenses are every subscription, Azure included, and every contract; Midas owns them

Ruled 2026-10-01 by ninad -- in force

The company's expenses, owned by Midas (Cost Center), are EVERY subscription the company pays for -- the model subscriptions already metered AND the infrastructure ones, Azure first -- and EVERY contract the company makes (Saee's internship, Sahil's contractor agreement, and those after them), each a line in the one spend plan and an actual in the company's accounts. A subscription or a contract missing from the register is an expense nobody owns; Marco's registers (subscriptions, contracts) are where each is kept, and Midas's accounts read them.

Midas has access to the company's bank account

Ruled 2026-10-01 by ninad -- in force

Midas, as owner of the company's expenses, has access to the company's bank account: every transaction on it reaches the company's accounts as an actual (a cost or a revenue), reconciled contract registers, so the books are read from the bank and never typed. The access is his own credential, held like a vendor key (through the sign-in keeper and door, cost-center #172), never the administrator's personal banking. The account exists only once the company is registered (leadership #250); until then this ruling is recorded and waits.

Company Accounts shows ACCRUED expense for the whole year, never paid; future months are FORECAST

Ruled 2026-10-01 by ninad -- in force

The Company Accounts app shows what the company has ACCRUED, not what it has paid, for the entire year. A past or current month is the expense accrued from the records: a subscription's plan fee for every month it is active, credit purchases when made, Azure's billed usage, and people from their contract terms and the hours actually worked. Every future month is a FORECAST derived from the same records: active plan fees carried forward, usage-based lanes and Azure at their run rate, contracts at their terms to their end. Every figure says which it is, accrued or forecast, and the year's boxes read accrued to date plus the forecast for the rest of the year. Paid is a separate question, answered by the bank feed (#256), never by this app. Ninad, 2026-10-01 (CLI-633): 'this app should show what is accrued, not paid, for entire year.. future months should be forcasted'.

Every estimation and reading of token consumption uses the same source

Ruled 2026-10-02 by ninad -- superseded 2026-10-02

EVERY ESTIMATION AND READING OF TOKEN CONSUMPTION USES THE SAME SOURCE: 'the estimations of elrond, cost tracker, operator log etc should use the same source for token consumption'. READ AS (Orion's reading, not Ninad's words): one source and one count for every surface and estimator that states how many tokens were consumed -- Elrond's effort estimates and his mileage measure, the cost trackers, the Operator Instance Log and every other reader: the cost ledger, read through ONE definition of a row's tokens (measured 2026-10-02: one ledger but at least five hand-written counts -- input+output+cache reads for the per-initiative totals Elrond's estimates read, input+output for the cost app, fresh input+output+cache reads+cache writes for the daily expense and the Operator Instance Log). A figure about something else -- an agent's carried prompt weight, a vendor's own usage report -- is a different measure and says so; it is not a second count of consumption.

Ended because: Ninad widened it the same evening from 'the same source' to counting AND reporting consistent across the estate

Token consumption is counted and reported the same way across the estate, from one source

Ruled 2026-10-02 by ninad -- in force

TOKEN CONSUMPTION IS COUNTED AND REPORTED THE SAME WAY ACROSS THE WHOLE ESTATE, FROM ONE SOURCE (Ninad, 2026-10-02, CLI-663). His words, in order: 'the estimations of elrond, cost tracker, operator log etc should use the same source for token consumption'; then, shown that the readers use two sources (the cost ledger, and the pipeline's own run records) and at least five counts: 'ok, go ahead, ensure that the counting and reporting of the consumed tokens is consistent across the estate'. READ AS (Orion's reading, not Ninad's words): every estimate, surface and report that states tokens consumed -- Elrond's effort estimates and mileage, the cost trackers, the daily estate expense, the Operator Instance Log, the activity tracker, the initiative audit and monitor, and any other reader -- counts from the ONE cost ledger through ONE definition of a row's tokens, so the same work reads the same number wherever it is shown. A figure about something else (an agent's carried prompt weight; a vendor's own usage report) is a different measure and is named as such, never a second count of consumption.

When Midas lacks the information to state Company Accounts' actuals correctly, he emails the administrator at his Denken Labs address, copying nobody

Ruled 2026-10-04 by ninad -- superseded 2026-10-15

Ninad (CLI-690, 2026-10-04), on the Company Accounts month chart: "if midas does not have any information to correct populate the actuals for this chart, he should email me on my denkenlabs email". Asked: 'When Midas emails you that he's missing information for the accounts chart, should he copy the other operators (Sahil, Saee), who can't open the accounts app?' Options: 'Only me (Recommended) -- company accounts stay confidential to you; Midas emails your Denken Labs address alone', 'Copy Sahil too', 'Copy all operators'. Ninad answered: 'Only me'. READ AS: whenever Midas lacks what he needs to state an ACTUAL (accrued) figure on the chart correctly -- an actual month whose measure reads unmeasured or unread, or whose total is only a lower bound, together with the named reason the records give (for example a contract with no start date, an hours source his app cannot read, Azure unread) -- Midas emails the administrator from his OWN address (cookbook 1809, communications #96) at the administrator's Denken Labs work address as root's record of operators carries it, never a typed or personal address, and copies nobody. The email names each missing piece, the months it affects and what would fill it; a gap is told once when it first appears and again only when the set of gaps changes, never on every page load. Forecast months are not actuals and are never the subject. Nothing is guessed or zeroed to fill a gap (cost-center #257, cookbook 1215): the chart keeps saying unmeasured or lower bound until the record holds the figure.

Ended because: communications #392 settles row 10 under #389

Elrond's Build Governance budget is raised and pinned to accommodate his effort estimations

Ruled 2026-10-04 by ninad -- superseded 2026-10-26

ELROND'S BUILD GOVERNANCE BUDGET IS RAISED TO ACCOMMODATE HIS EFFORT ESTIMATIONS (Ninad, CLI-695, 2026-10-04). Option a was never recorded.

READ AS (Orion's reading, not the ruling): #152 stands -- every task Elrond owns, the estimator's included, stays on Build Governance. The administrator may name another figure.

The OpenAI platform (GPT API) subscription is un-retired

Ruled 2026-10-05 by ninad -- in force

Asked: "ChatGPT's weekly reading pays from the general OpenAI account, which you retired on 25 Sept -- and that ruling says topping it up is not the fix. How should ChatGPT's reading be paid? Options: its own key (a new OpenAI key on an account you fund, enrolled by a root step); un-retire the account (supersede the 25 Sept ruling and top up the general OpenAI account again); drop ChatGPT (Loki reads Claude and Gemini only)." Ninad answered: "Un-retire the account". READ AS (Orion, not the ruling): the OpenAI platform account the estate held for GPT API work is no longer retired; its subscription record is reinstated rather than a new one created, the Subscription Balance Reader reads it again, and it is topped up so Loki's ChatGPT reading (and any other work routed to it) can be paid. #143's DeepSeek retirement and its protection of image or media lanes on keys of their own are not touched by this answer.

Each login's daily sign-in proof is of its own, never one another login's proof can spend.

Ruled 2026-10-05 by ninad -- in force

EACH LOGIN'S DAILY SIGN-IN PROOF IS OF ITS OWN (Ninad, CLI-725, 2026-10-05 23:5xZ). Keeps ruling #181." (over "Bigger shared pot" -- raise the shared Mind track's allowance to cover every Mind-only login, the one shared pot #181 names -- and "Mind's own work counts" -- skip the proof on days the Mind made a real call). WHAT LOOKS LIKE A FIX AND IS NOT: raising the shared allowance until the morning's proofs fit; billing the proofs to another track that has headroom (that routes the proof onto that track's lane, 1672); skipping the proof for a quiet Mind.

Picture work moves from Gemini to GPT instead of topping up Gemini's spent prepaid credit

Ruled 2026-10-06 by ninad -- in force

PICTURE WORK USES GPT INSTEAD OF GEMINI (Ninad, CLI-730, 2026-10-06). THE QUESTION ANSWERED (Archon's escalations #1581 and #1672): Orion asked: "Gemini's prepaid credit has been spent since about 28 September. Archon's portraits and every picture prompt now fail or run degraded. What do you want?" (offered: "I'll top it up"; "Leave it off" -- pictures stay paused). HIS ANSWER: "use GPT instead". READ AS (Orion's reading, not Ninad's words): the picture work that ran on Gemini -- generating images and enhancing their prompts -- moves onto GPT (OpenAI) slots; each track that moves is registered on its new slot before it runs there (cookbook 1791, cost-center #78 #81 #82); Gemini's credit is not topped up for this work. Video and speech were not asked about and are not moved by this.

Nothing on the machine works on the shared Claude Max sign-in: its last two uses move to Codex

Ruled 2026-10-13 by ninad -- in force

THE LAST TWO PIECES OF ESTATE WORK ON THE MACHINE'S SHARED CLAUDE SIGN-IN MOVE TO CODEX, AS THE REST OF THE ESTATE DID (Ninad, 2026-10-13, CLI-756). Said after being shown, from the cost ledger, everything that ran on the sign-in keeper's Claude sign-in in the two weeks before it died on 2026-10-09: "got it, move these to codex as well".

ASKED: whether to re-make that sign-in, which only a Claude Max account can be (the keeper refuses the Denken Labs Team account), and which would put his personal account behind estate work, or to move what uses it.

READ AS (Orion, CLI-756): (1) No estate work on the machine runs on the shared Claude sign-in. Loki's AI search track moves to Codex (marketing ruling of the same session for what it gives up) and Orion's chat window's working part is rebuilt on Codex (agent-resources ruling of the same session). (2) The sign-in is therefore not re-made: root escalation #3 is closed unrun, the keeper stops renewing the Claude slot and the hourly escalations about it end at their cause, and the daily probe of the Claude account stops -- Hermione's watch itself already runs on Codex. (3) The Orion bypass track keeps declaring Claude: Orion's own sessions run on the operator's laptop Claude login, never on the machine's sign-in (cookbook 1567's named exception stands). (4) (the keeper adopts only the approved Claude estate account) has nothing left to adopt. Re-making the sign-in so a reading turns green, or moving the Orion bypass track's declaration, each undo it.

When Midas lacks the information to state Company Accounts' actuals correctly, he emails the administrator at his Denken Labs address, copying nobody

Ruled 2026-10-15 by ninad -- superseded 2026-10-15

Ninad (CLI-690, 2026-10-04), on the Company Accounts month chart: "if midas does not have any information to correct populate the actuals for this chart, he should email me on my denkenlabs email". Asked: 'When Midas emails you that he's missing information for the accounts chart, should he copy the other operators (Sahil, Saee), who can't open the accounts app?' Options: 'Only me (Recommended) -- company accounts stay confidential to you; Midas emails your Denken Labs address alone', 'Copy Sahil too', 'Copy all operators'. Ninad answered: 'Only me'. READ AS: whenever Midas lacks what he needs to state an ACTUAL (accrued) figure on the chart correctly -- an actual month whose measure reads unmeasured or unread, or whose total is only a lower bound, together with the named reason the records give (for example a contract with no start date, an hours source his app cannot read, Azure unread) -- Midas emails the administrator from his OWN address (cookbook 1809, communications #96) at the administrator's Denken Labs work address as root's record of operators carries it, never a typed or personal address, and copies nobody. The email names each missing piece, the months it affects and what would fill it; a gap is told once when it first appears and again only when the set of gaps changes, never on every page load. Forecast months are not actuals and are never the subject. Nothing is guessed or zeroed to fill a gap (cost-center #257, cookbook 1215): the chart keeps saying unmeasured or lower bound until the record holds the figure.

AMENDED (Ninad, CLI-776, 2026-10-15, communications #392): Midas no longer EMAILS the administrator for a missing actual; he raises it as an ESCALATION TO ORION, who brings it to the administrator in his session (#389: an operator is emailed only for what cannot wait). HIS WORDS, answering the twelve-sender list whose row 10 read 'Midas's accounts gaps -- escalate to Orion instead': "I agree with all other recommendations". The rest of this ruling (what counts as a missing actual, nobody else is told) stands.

Ended because: the subject still said 'emails' after communications #392 changed the channel; the text is unchanged

When Midas lacks the information to state Company Accounts' actuals correctly, he raises it to Orion as an escalation, copying nobody

Ruled 2026-10-15 by ninad -- in force

Ninad (CLI-690, 2026-10-04), on the Company Accounts month chart: "if midas does not have any information to correct populate the actuals for this chart, he should email me on my denkenlabs email". Asked: 'When Midas emails you that he's missing information for the accounts chart, should he copy the other operators (Sahil, Saee), who can't open the accounts app?' Options: 'Only me (Recommended) -- company accounts stay confidential to you; Midas emails your Denken Labs address alone', 'Copy Sahil too', 'Copy all operators'. Ninad answered: 'Only me'. READ AS: whenever Midas lacks what he needs to state an ACTUAL (accrued) figure on the chart correctly -- an actual month whose measure reads unmeasured or unread, or whose total is only a lower bound, together with the named reason the records give (for example a contract with no start date, an hours source his app cannot read, Azure unread) -- Midas emails the administrator from his OWN address (cookbook 1809, communications #96) at the administrator's Denken Labs work address as root's record of operators carries it, never a typed or personal address, and copies nobody. The email names each missing piece, the months it affects and what would fill it; a gap is told once when it first appears and again only when the set of gaps changes, never on every page load. Forecast months are not actuals and are never the subject. Nothing is guessed or zeroed to fill a gap (cost-center #257, cookbook 1215): the chart keeps saying unmeasured or lower bound until the record holds the figure.

AMENDED (Ninad, CLI-776, 2026-10-15, communications #392): Midas no longer EMAILS the administrator for a missing actual; he raises it as an ESCALATION TO ORION, who brings it to the administrator in his session (#389: an operator is emailed only for what cannot wait). HIS WORDS, answering the twelve-sender list whose row 10 read 'Midas's accounts gaps -- escalate to Orion instead': "I agree with all other recommendations". The rest of this ruling (what counts as a missing actual, nobody else is told) stands.

An operator is paid for the time their Orion instance is actively working, as a driver is paid for the time the car is driven

Ruled 2026-10-15 by ninad -- in force

AN OPERATOR IS PAID FOR THE TIME THEIR ORION INSTANCE IS ACTIVELY WORKING, AS A DRIVER IS PAID FOR THE TIME THE CAR IS DRIVEN (Ninad, CLI-796, 2026-10-15). ASKED (after communications #415 ruled that the Operator Instance Log's hours count how long an operator's Orion instance actively worked, not how long the operator worked): "Company Accounts currently uses this number for pay on hourly contracts. If it measures Orion and not the person, should pay still come from it, or from something else?" HIS ANSWER: "think of operator like a driver, the driver drives the car, and gets paid for the time he is driving the car, if the car is parked, that is not counted in his time.. so as long as orion is doing the work, work is happening, and that goes in to the operators credit, and he gets rewarded for that.. his job is to get work done from orion, not do the work himself". Asked in the same exchange "Are you OK with the rename to 'Orion active hours'?", he answered "yes". READ AS (Orion's reading, not his words): (1) Company Accounts keeps accruing an operator's hourly contract from the Orion active hours (communications #415) -- the one figure the log shows -- and never from the operator's own hours, which nothing measures; (2) the time an Orion instance sits idle with work open (a parked car) is not in the operator's credit; (3) the measure's name is 'Orion active hours' on the log, in Company Accounts and in the dictionary.

Every project operates on its own bank account, opened in the name of its admin or an estate operator of that project; Midas is equipped to create it

Ruled 2026-10-18 by ninad -- in force

EVERY PROJECT HAS ITS OWN BANK ACCOUNT TO OPERATE ON, IN THE NAME OF ITS ADMIN OR AN ESTATE OPERATOR OF THAT PROJECT (Ninad, CLI-832, 2026-10-18). CONTEXT: after the operator onboarding and the project onboarding arcs were reviewed in this session, Ninad gave the following instruction unprompted. His words: "each project needs to have its own bank account to operate on, which can be created in the name of admin or estate operator of that project.. Ammend the project onboarding arc to include initiatives that will equip Midas to create this account for each project, identify the right service, track and tasks for these, and build the necessary droids and apps for the same. Update the project onboarding app with this requirement. you can also re-use carolverse apps, by adding a project filter on it, wherever you can.." READ AS (Orion's reading, marked apart, P.01.01.05.05; not his words): (1) a project's bank account is a record of the project (client -> account -> dominion -> project, blueprint #4), opened in the name of a PERSON who holds the project's admin or estate-operator role (root's record / the project's roles, security #316), never in the name of an agent; so it can exist before the company does -- cost-center #256 and leadership #250 (Midas's access to the COMPANY's account waits on registration) are untouched and speak of a different account. (2) Midas, as the owner of Cost Center, is EQUIPPED to open and hold it: the opening itself is driven by Midas end to end (the application, the holder's approval and signature as the one human step, the identifiers kept like a vendor key through the sign-in keeper, never in the registry), and reading it afterwards (balance, statements, paid) is his. (3) The home is a new TRACK of the Cost Center service, Treasury, with its own tasks (account requested, opened, balance read, statement read, payment recorded) -- not Billing (invoices, cost-center #18) and not Procurement (purchases, #37/#47). (4) The project onboarding pass carries the requirement as a step handed to Midas before the project's services are enabled, so no project operates without its account; the Project Onboarding app renders it from the BLOOM record. (5) Apps are reused with a PROJECT filter wherever one serves (Company Accounts first; Billing already filters by project); a new app only where no Carolverse app can carry the view. (6) An arc belongs to one service (initiatives #383-#385): the Blueprint-side step joins the project onboarding arc, Midas's treasury work joins his Cost Center arc, and one pursuit holds them together.

Carolverse's project bank account is a Revolut account in the United Kingdom, in pounds sterling, in Ninad's own name while the company is in formation

Ruled 2026-10-18 by ninad -- in force

CAROLVERSE'S PROJECT BANK ACCOUNT IS A REVOLUT ACCOUNT IN THE UNITED KINGDOM, IN POUNDS STERLING, IN NINAD'S OWN NAME (Ninad, CLI-841, 2026-10-18). CONTEXT: the question CLI-839 left open for (Midas's Project Account Opener): which bank, country and currency for Carolverse's first account (record #1, holder ninad, requested; cost-center #442). HIS WORDS: "Midas's account opener (6751) - revolut?". ASKED (CLI-841): "Revolut for Carolverse's account, in your own name. Revolut Business needs a registered company, and Denken Labs (UK) is still in formation, so this would be a personal Revolut account until the company exists. Which country and currency?" (offered: UK, GBP / EU (Revolut's Lithuanian bank), EUR / Wait for the company). HIS ANSWER: "UK, GBP". READ AS (Orion, not his words): the opener opens record #1 at Revolut, country United Kingdom, currency GBP, as a personal account in his name, and sets bank, country and currency on the record at its move to opening through shared.project_accounts.set_status; an account in the company's name once Denken Labs UK is registered would be a new ruling.

Midas marks a project's invoice paid when a line on its own bank statement matches it

Ruled 2026-10-18 by ninad -- in force

MIDAS MARKS A PROJECT'S INVOICE PAID WHEN A LINE ON THAT PROJECT'S OWN BANK STATEMENT MATCHES IT (Ninad, CLI-851, 2026-10-19). CONTEXT: CLI-846 found a law gap while planning (Midas reads every project's bank account): the reference-data set project_invoice.status lets only an operator move an invoice from issued or sent to paid, so Midas's statement reader would be refused. ASKED: "When a line on a project's bank statement matches one of its invoices (same amount, with the invoice number in the reference), may Midas mark that invoice paid himself? Today only an operator can mark an invoice paid." (offered: Midas marks it -- he marks it paid on a match, keeps the bank line as evidence and uses the bank's date / Midas suggests, operator confirms). HIS ANSWER: "Midas marks it (Recommended)". READ AS (Orion, not his words): the transitions issued -> paid and sent -> paid in project_invoice.status also admit the owner (Midas, owner of the Cost Center's Treasury track) when the evidence is a matched outgoing line on the project's own account statement; the invoice's paid date is the statement's booked date, never the clock (cookbook 1671); the bank line is recorded as the evidence; an operator may still mark an invoice paid as before. This applies cost-center #256's 'the books are read from the bank and never typed' to project invoices.

Midas does not read Carolverse's personal Revolut account; it reads unmeasured until the company's business account exists

Ruled 2026-10-18 by ninad -- in force

MIDAS DOES NOT READ CAROLVERSE'S PERSONAL REVOLUT ACCOUNT; ITS BALANCE AND STATEMENT READ UNMEASURED UNTIL THE COMPANY'S BUSINESS ACCOUNT EXISTS (Ninad, CLI-851, 2026-10-19). CONTEXT: Carolverse's project bank account is a personal Revolut account in his name while Denken Labs UK is in formation (cost-center #445), while cost-center #256 says Midas's bank access is his own credential and never the administrator's personal banking; and Revolut gives software no reading route for a personal account (its business API needs the registered company). ASKED: "Software can't read a personal Revolut account (Revolut only offers that for business accounts, which need the registered company). Also, your ruling 256 says Midas never reads your personal banking, but Carolverse's account will be a personal one in your name. How should Midas read its balance and statement?" (offered: Wait for company account -- until Denken Labs UK is registered and has a Revolut Business account, the balance reads 'unmeasured'; ruling 256 stays whole, and the reader and opener still get built / Open Banking reader -- a paid read-only service connected once to the personal Revolut, as an exception to ruling 256 / Monthly statement file -- he downloads the statement each month and Midas reads the file). HIS ANSWER: "Wait for company account (Recommended)". READ AS (Orion, not his words): cost-center #256 stays whole; Midas's reader and opener are still built; an account held personally by a person reads its balance and lines as UNMEASURED with that reason, never zero (cookbook 1215), and no route to a personal account is built; the reading route for Carolverse is chosen when its account moves to a business account in the company's name, which #445 already says is a new ruling.

Chowpatty's machine is the small size; Hagrid builds machines, Marco procures them, and every procurement decision waits for an estate operator's approval

Ruled 2026-10-20 by ninad -- superseded 2026-10-21

Ninad, 2026-10-21 (CLI-883). THE QUESTIONS ANSWERED (Sahil's operator communication #95, CLI-871: 'Agree the cost of Chowpatty's own machine, and how Hagrid gets the right to make machines', CAROL-PUR-0006 stage 3). Orion asked, first: "Chowpatty's own machine (Sahil's #95, needed before his stage 3). Chowpatty's first module is daily reconciliation, invoices and an order inbox. Which size? Orion asked, second: "How should Hagrid get the right to make project machines (Sahil's #95)? Today his cloud identity can only read." offering one fenced area of the Azure account plus a monthly spending alert (recommended), Ninad creating each machine himself, or Hagrid asking and Ninad approving each. HIS ANSWER, word for word: "hagrid is responsible for building, but marco is responsible for procuring the machine. Marco waits for estate operator's approval for all procurement decisions". PROCURING a machine -- obtaining it from the cloud provider and committing its cost -- is Marco's (the Procurement track of Cost Center), never Hagrid's; HAGRID BUILDS a machine once it is procured (sets it up as a working machine of the estate). Every procurement decision of Marco's, machines included, waits for an estate operator's approval before it is made; the size above is the administrator's approval for Chowpatty's machine. How Marco's login is given the right to create a machine in the cloud account is not settled here and comes back as its own question when is built.

Ended because: Ninad corrected #468 on 2026-10-21 (CLI-884): Marco only gets approval for purchases and takes no part in creating machines in Azure; Hagrid's right to create machines is an estate operator's to settle

Chowpatty's machine is the small size; Marco gets an estate operator's approval for every procurement and creates no machines; creating machines is Hagrid's building work

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884), correcting cost-center #468 (CLI-883). WHAT HE CORRECTED: Orion's report of #468 to him said: "Hagrid's right to make machines -- one question is still open: how Marco gets permission to create machines in Azure. Sahil will bring that to you when that work is built." HIS WORDS: "This is not operator governance, so it should not necessarily come to me, unless sahil chooses to ask me.. but generally, Marco is the head of procurement, his job is to get approval for any procurement decision from estate operator. Marco should not in any way involved in creating the machines in azure, as he reports to head of finance (Midas)". Marco is the Head of Procurement and reports to Midas, the Head of Finance; his part in any procurement is to get an estate operator's approval for the decision, and he takes no part in creating machines in Azure. Creating a machine in Azure is part of Hagrid's building work. How Hagrid's login is given the right to create machines is not operator governance, so an estate operator settles it and brings it to the administrator only if they choose to. #468's reading that 'procuring a machine -- obtaining it from the cloud provider and committing its cost -- is Marco's' and that 'how Marco's login is given the right to create a machine comes back as its own question' are retired in place. #468 as it was recorded, for the chain: Ninad, 2026-10-21 (CLI-883). THE QUESTIONS ANSWERED (Sahil's operator communication #95, CLI-871: 'Agree the cost of Chowpatty's own machine, and how Hagrid gets the right to make machines', CAROL-PUR-0006 stage 3). Orion asked, first: "Chowpatty's own machine (Sahil's #95, needed before his stage 3). Chowpatty's first module is daily reconciliation, invoices and an order inbox. Which size? Orion asked, second: "How should Hagrid get the right to make project machines (Sahil's #95)? Today his cloud identity can only read." offering one fenced area of the Azure account plus a monthly spending alert (recommended), Ninad creating each machine himself, or Hagrid asking and Ninad approving each. HIS ANSWER, word for word: "hagrid is responsible for building, but marco is responsible for procuring the machine. Marco waits for estate operator's approval for all procurement decisions". PROCURING a machine -- obtaining it from the cloud provider and committing its cost -- is Marco's (the Procurement track of Cost Center), never Hagrid's; HAGRID BUILDS a machine once it is procured (sets it up as a working machine of the estate). Every procurement decision of Marco's, machines included, waits for an estate operator's approval before it is made; the size above is the administrator's approval for Chowpatty's machine. How Marco's login is given the right to create a machine in the cloud account is not settled here and comes back as its own question when is built.

Operator governance work sits on tracks named for it, and every task of it has one

Ruled 2026-10-21 by ninad -- in force

OPERATOR GOVERNANCE WORK SITS ON TRACKS NAMED FOR IT, AND EVERY TASK OF IT HAS ONE (Ninad, CLI-887, 2026-10-21). Shown that operator governance runs as 25 tasks spread over five services and root's record, that the Session Opener and the Operator Communication Mailer bill a task named 'an operator session was closed', and that operator communications and root or core escalations have no task, he asked: "fix the gaps, also rename the core into something that represents the work under that track and split tracks to make the work distinct where its necessary (take my approval first)". Shown the proposal as service / old track / new track / tasks / proposal, he answered: "approved". THE APPROVED PROPOSAL: Security -- 'Core' renamed 'Access and Ownership' (7 tasks: access lifecycle, recertification, credential rotation, privileged session review, security rules recorded, artifact holdings recorded and probed); new 'Operator Compliance' (4: compliance told, door sign-ins read, Orion sign-ins watched, service operators' containment checked); new 'Risk and Scanning' (4: privacy scan, product scan, resilience check, risk and incident). Communications -- 'Internal Communications' renamed 'Operator Mail and Meetings' (4: mail read, consent reminded, reminders, reply drafts); new 'Operator Instance Log' (4: commands read, weekly hours told, and from Orion Bypass the hourly update recorded and a service operator's own acts); new 'Operator Communications' (5: organisation change told, template release announced, and three new tasks -- operator communications kept, root or core escalations kept, their email notices). Governance -- 'Records & Operations' keeps its name (9); new 'Operator Sessions' (4: a new 'session opened' task, session closed, operator filing, ad-hoc investigation). Agent and Human Resources -- 'Operator Onboarding' renamed 'Operator Onboarding and Workstations' (10: the nine, and from Orion Bypass the operator template kept current). Build Initiatives -- 'Orion Bypass (CLI)' keeps its name and its own three (bypass runs, stale sessions reaped, the arc report). READ AS (Orion, not the ruling): track and task keys are identifiers and never renamed (the readable name changes); a task's budget moves with it; Midas's allocator sets a new track's cap from the work it takes; a new track that does model work is registered on its slot before that work runs there (P.01.03.07.21).

The budget that builds or expands a service comes from a project, and the project's client assigns it

Ruled 2026-10-21 by ninad -- in force

THE BUDGET THAT BUILDS OR EXPANDS A SERVICE COMES FROM A PROJECT, AND THE PROJECT'S CLIENT ASSIGNS IT (Ninad, CLI-894, 2026-10-21). Shown initiatives #486 built as (an Orion-session filing in an arc owned by the arc's owner; 201 pending initiatives moved, 6782 to Heimdall) and proposed policy words for P.01.01.01.10 that kept 'the owner of an initiative is the agent who requested it' with that one exception, he answered: "a requester cannot be the owner, instead, whoever is paying for the initiative should be the owner of the initiative.. a requestor is just a requestor. The payment is about budget that enables building or expanding a service. The budget comes from project. Project belongs to a client (say denken-labs) who assignes budgets for expanding existing services and building new services. Each project should have the CEO and head of finance (the agent and employee mapping in future, estate operators get mapped by default) The agent economy is tightly intermingled within the human economy in carolverse.. note this ruling, let the next session instantiate or manifest it into the estate". READ AS (Orion, not the ruling): (1) every project belongs to a client (Denken Labs for Carolverse, the dominion's base project), and the client assigns the project's budgets -- for expanding existing services and for building new ones; (2) that budget is what pays for an initiative, and its payer is the initiative's owner (initiatives #487); (3) the agent economy runs inside the human economy: agents spend budgets that people assign, and every budget traces to a client's assignment; (4) it sits beside, and does not yet say how it meets, the track and service caps Midas sets daily from the rolling week under fixed top limits (cost-center #218, #270) and subscriptions belonging to projects (cookbook 1277-1279) -- how a client's assignment becomes those caps is OPEN for the build.

Elrond is the plumber: the owning service pays for an initiative, Elrond pays for pipeline work, and Elrond charges the services

Ruled 2026-10-21 by ninad -- in force

ELROND IS THE PLUMBER: THE OWNING SERVICE PAYS FOR AN INITIATIVE, ELROND PAYS FOR PIPELINE WORK, AND ELROND CHARGES THE SERVICES (Ninad, CLI-896, 2026-10-21). Asked whether Elrond still accepts Orion-lane and Albus-lane work by the lane or acceptance follows the new owner, he answered: "Elrond owns the gate and pipelines, which are used for executing the initiative.. there needs to be an agreement on who pays for the initiative. If its for improving the pipeline, elrond pays, if not, then service owner pays for the initiative.. the requestor and service owner can be the same.. think of it as a house that is given for rental, the tenant identifies that something is broken, and owner of the house ends up fixing it.. (tenant is the requestor, owner of the howuse pays for it, and elrond acts like the plumber, who gets the work done). I know that currently elrond is bearing the cost, but he needs to start charging it to the services (read the agent economy blog from orion's logbook)". READ AS (Orion, not the ruling): (1) before an initiative is executed there is a recorded agreement on who pays: the owner of the service it improves, or Elrond's own service (initiatives) when it improves the pipeline; (2) the build lanes keep running on their own tracks (the track that SERVES), and the work ticket bills the owning service's budget (the track that PAYS) -- the two questions cookbook 1672 already keeps apart; (3) the requester may be the owner but is only the requester (#487); (4) the Logbook's 'Agent Economy -- Services in Carolverse' (every service charges for its work; nothing is free) is the model; (5) the question of who SIGNS acceptance is not answered here: Orion keeps Elrond signing by the lane (cookbook 1566) as the interim until each service owner has an acceptance droid, and says so on each close.

Midas is the head of finance; Clara is responsible for getting investments for the company

Ruled 2026-10-21 by ninad -- in force

MIDAS IS THE HEAD OF FINANCE; CLARA IS RESPONSIBLE FOR GETTING INVESTMENTS FOR THE COMPANY (Ninad, CLI-896, 2026-10-21). Asked who is Carolverse's head of finance, since Clara is CEO with Finance as her exclusive track (cookbook 1724) and Midas keeps the accounts and the bank account, he answered: "Midas is the head of finance, Clara is responsible to get investments for the company". READ AS (Orion, not the ruling): (1) for Carolverse, blueprint #489's two roles are Clara (CEO) and Midas (head of finance); (2) Clara's Finance track (1724, leadership #247) stays hers and means investments: investors, valuation, rounds, shareholding; (3) the project's budgets, what they pay for and the books are Midas's (cost-center #255, #256, #488).

An Orion session's helpers that run on Claude Sonnet 5.5 are counted on the Orion bypass lane's track in the Initiatives service and priced as Claude Opus 5.5

Ruled 2026-10-25 by ninad -- superseded 2026-10-25

AN ORION SESSION'S SONNET 5.5 HELPERS ARE PRICED AS OPUS 5.5 (Ninad, 2026-10-25, CLI-945).

FOUND: CLI-944 started eight helpers on Claude Sonnet 5.5; the price list holds no Sonnet 5.5, so the cost ingest refused the whole session's tokens and cost ('unknown model claude-sonnet-5-5 and no cost_usd override'). Orion filed to put Sonnet 5.5's own price on the list.

ANSWERED, first: "it still folds into opus 5.5 - orion bypass lane track within the initiatives service for tracking purposes". ASKED which he meant: (A) counted on that track but priced at Sonnet 5.5's own rate, or (B) counted on that track and priced as Opus 5.5. ANSWERED: "The helpers' work is counted on that track and priced as Opus 5.5".

READ AS (Orion's reading, not the administrator's words, NOT YET CONFIRMED by him): a row of an Orion session's transcript whose model is Claude Sonnet 5.5 is booked to the Orion bypass lane's track (initiatives-orion-bypass, service initiatives) at Claude Opus 5.5's prices, the row keeping the model it actually ran on and naming this ruling as the reason for its price; it binds the Orion laptop cost ingest only, not estate lanes; recording a little more than was spent is within his 'record more, not less' (communications #475). (Sonnet 5.5's own price) is superseded by the work this ruling asks.

Ended because: Ninad changed his answer from option B to option A the same hour (CLI-945).

An Orion session's helpers that run on Claude Sonnet 5.5 are counted on the Orion bypass lane's track in the Initiatives service and priced at Sonnet 5.5's own rate

Ruled 2026-10-25 by ninad -- in force

AN ORION SESSION'S SONNET 5.5 HELPERS ARE PRICED AT SONNET 5.5'S OWN RATE (Ninad, 2026-10-25, CLI-945; supersedes #518).

FOUND: CLI-944 started eight helpers on Claude Sonnet 5.5; the price list holds no Sonnet 5.5, so the cost ingest refused the whole session's tokens and cost.

ASKED: (A) counted on the Orion bypass track in the Initiatives service but priced at Sonnet 5.5's own rate, or (B) counted on that track and priced as Opus 5.5. He first answered B (#518), then: "i would like to change the ruling to option A - The helpers' work is counted on the Orion bypass track in the Initiatives service, but priced at Sonnet 5.5's own rate".

READ AS (Orion's reading, not the administrator's words, NOT YET CONFIRMED by him): a row of an Orion session's transcript whose model is Claude Sonnet 5.5 is booked to the Orion bypass lane's track (initiatives-orion-bypass, service initiatives) at Sonnet 5.5's own prices, read from Anthropic's published price page with its source and date, never guessed or copied from another model. (Sonnet 5.5's own price on the list) is the work and is revived; (priced as Opus 5.5) is discarded.

Elrond's Build Governance budget is set by Midas from the rolling week, no longer held

Ruled 2026-10-26 by ninad -- in force

ELROND'S BUILD GOVERNANCE BUDGET IS NO LONGER HELD: MIDAS SETS IT FROM THE ROLLING WEEK (Ninad, CLI-975, 2026-10-26). Asked: 'The fix for the 100% problem works live: the app and the figures agree on all 68 live arcs, and none shows 100% while work is open. One requirement fails: Elrond's Effort Estimator must run a full day without being refused for budget.

#152 stands: the estimator stays on Build Governance and shares it with Elrond's other work. 's criterion 2 is revised to ask for a full day of estimator runs with no budget refusal on the cap Midas sets.

Albus's lane and Elrond's Build Governance: their budgets go back to Midas

Ruled 2026-10-27 by ninad -- in force

READ AS: the holds on initiatives-albus-bypass and initiatives-build-governance end; Midas's allocator sets both from the rolling week at its next daily run, as it does every governed track (#218, #270, #548). Albus asking Midas for his own budget (the future part of #231) is not ruled on here.

The Token Cost Tracker's By service chart grows taller to show the daily-limit line, up to 600px

Ruled 2026-10-28 by ninad -- in force

"the token cost tracker app needs to increase its height further to show the budget line, as managed by midas, without shortnening the consumption bars displayed on the chart" -- then, asked which chart: "the \"By service\" view". Asked (Orion, CLI-998): 'I'll make the By service chart grow taller so the daily-limit line fits, with the bars kept at their current size. Most days that works: yesterday it needed about 1.7 times today's 240px. This early in the day, though, the spend is small, so it needs about 9 times (around 1,500px). In a day's first minutes it could need thousands of times. What should the chart do on those days?' -- answered: 'Cap, then shrink bars' (the option read: 'The chart grows to fit the line, up to 600px. Past that the bars shrink just enough for the line to fit.'). READ AS: in the cost view the By service chart keeps the bars at the scale gave them (the axis 20% above the day's tallest service bar) and grows taller until the highest daily-limit point is inside it, up to 600px; past 600px the chart stays 600px and its axis reaches the line, so the bars are shorter only by what the line needs. The tokens view (no line) and a day with no spend are unchanged. 's 'a limit above that runs off the top' is retired in place for the cost view.

Elrond charges his own work for an initiative to the service that pays for it

Ruled 2026-10-28 by ninad -- in force

"good catch, update the pursuit to enable elrond to also distribute his costs to the services" -- answering (Orion, CLI-998): 'Should Elrond's per-initiative checking cost also be charged to each initiative's owning service? Before that he said: "the build initiative budget is high and not getting distributed across services because the cost of the initiatives is not getting attributed to the service requesting the initiative". READ AS: Elrond's work done for one initiative is booked to that initiative's payer (the owning service's budget, or Build Governance itself for pipeline work, #492), the same way 6810 books a build; work that serves no one initiative stays on Build Governance. The payer is the service the initiative improves (#491), not the requester (#487). Built as its own initiative in Elrond's arc CAROL-ARC-0048 and added to pursuit CAROL-PUR-0011 (one initiative per service, initiatives #384), not folded into 6810, which is Cost Center's. Moving the pursuit up the list is not ruled on here.

Connection dropped is dealt with on priority: the Operator Relay gets ten minutes to finish replies in flight when it stops

Ruled 2026-10-28 by ninad -- in force

THE OPERATOR RELAY IS GIVEN TEN MINUTES TO FINISH THE REPLIES IN FLIGHT WHEN IT STOPS, NOW, NOT AFTER A WEEK OF WATCHING (Ninad, CLI-005, 2026-10-28). ASKED (Orion): the estate operators' service-unit act refused the relay unit's TimeoutStopSec, so the relay drains within systemd's 90 s (80 s) and the Deployer restarts it only between replies; extending the act costs the administrator's sign-ins; Orion recommended watching the relay's record for a week first. Ninad: "Connection dropped is a serious issue and should be dealt with on priority, i'd say, better to extend it to 10 minutes right away". READ AS (Orion, not the ruling): the service-unit act gains TimeoutStopSec (version 4, bounded at 900 s) and the relay's unit asks 630 s, so its 600 s drain is never killed; 's criterion 1 stands as worded.

The agents' wake budget is Midas's to set from the rolling week's real daily need, never held by a pin

Ruled 2026-11-03 by ninad -- in force

THE AGENTS' WAKE BUDGET IS MIDAS'S TO SET FROM THE ROLLING WEEK'S REAL DAILY NEED, NEVER HELD BY A PIN (Ninad, CLI-020, 2026-11-03). HIS WORDS: "2 seperate issues.. first, carol's wake cycle budget needs rectification.. this should be handled by Midas.. he observes the average daily requirements of the rolling week and adjusts the budgets accordingly (atleast he is supposed to do that if he is not doing that already).. Second, regarding recommending Sachin Joshi to blueprint as a user, carol should do that as a scheduled process, not as a wakeup activity.. wakeup activity is budget constraint, and this recommendation is part of carol's primary duties.. a user can request access to leo chat via carol whatsapp or web chat, and it should be picked up by the droids that are responsible to provision this access. Carol should own these droids.." READ AS (Orion's reading, not his words): (1) the consciousness track's cap is rectified by Midas's allocator observing what the agents' wakes need each day over the rolling week -- the demand his cap refused included -- never by an operator pinning a figure; the one-day pin of cost-center #575 is not extended; (2) if Midas does not already do this, building it is the cure.

Elrond is enabled to size today's waiting work: Build Governance's cap is lifted for 3 Nov by a one-day exception

Ruled 2026-11-03 by ninad -- in force

ELROND IS ENABLED TO SIZE TODAY'S WAITING WORK: BUILD GOVERNANCE'S CAP IS LIFTED FOR 2026-11-03 UTC BY A ONE-DAY EXCEPTION (Ninad, CLI-020, 2026-11-03). please enable elrond to do the estimations and give me the status again". This later word narrows #581's 'never held by a pin' for this one day, same operator.

Midas raises a track's daily cap by 10% when the track hits its cap on two consecutive days of the rolling week, so caps grow slowly to match the load

Ruled 2026-11-04 by ninad -- in force

HIS WORDS: "I also want Midas to increase his budget caps by 10%, if a particular service hits the cap consecutively twice in the rolling week, this way, the budget cap slowly increases to match the load.." READ AS (Orion, not his words): (1) Caps are held per governed TRACK (#218), so 'a service' is read as the track whose cap was hit. (2) A track has HIT its cap on a day when its spend reached the cap in force that day or a spend was refused for that track's budget. (3) When a track hit its cap on two consecutive complete days inside the rolling week, Midas's next daily allocation sets its cap to the HIGHER of the #218 formula value and the cap in force plus 10%; the count then starts again, so a track rises by at most one step every two days. This adds a way up to #218; it does not replace the formula, which still lowers a cap when the load falls.

After Midas raises a track's cap by 10%, the two-day count starts again, so a cap rises at most once every two days

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-043/CLI-044), answering the question owed since CLI-035 on cost-center #628 and (built and accepted by Elrond the same day). ASKED (Orion): 'Midas raises a budget by 10% when it runs out two days in a row. I read that after each raise the two-day count starts again, so a budget goes up at most once every two days. Is that right, yes or no?' HIS WORDS: "after Midas raises a budget by 10%, does the two-day count start again? - yes". READ AS (Orion, not his words): #628's reading (3) stands as built in the count of consecutive hit days restarts after a step, so a track rises by at most one step every two days; nothing else about #628 changes.

Albus's lane: Midas never sets its daily cap below the cost of one full attempt

Ruled 2026-11-06 by ninad -- in force

READ AS: Midas's allocator never sets the daily cap of initiatives-albus-bypass below the cost of one full attempt on that lane (Albus's decide turn, Forge's build, Albus's judge turn), measured from the lane's own records; above that floor the rolling-week setting of #555 (#218, #270) and the 10% step-up of #628 stand. How a full attempt's cost is measured, and what stands while none is measured, is the build's to design (filed under CAROL-PUR-0002).

Every service charges whoever consumes its work and pays for what it consumes, at cost plus a margin Midas caps, and its profit shapes its next budget

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-067), unprompted, after closed. HIS WORDS: "ensure that each service owner is motivated to make their service profitable, this would mean that each service charges for their services to whoever is consuming them, and also pays for other services in carolverse.. this may need a seperate build, if its not already built". Orion found it is not built (the services catalogue's profit is null, 'no model yet'; a service's cost counts only its initiatives) and asked how a service sets its price and whether profit changes anything beyond the books. HIS ANSWERS: 'Cost plus margin' (price = the service's measured cost of the work plus a margin its owner sets, kept under a cap Midas sets) and 'Shapes next budget' (Midas's allocator gives more of next month's budget to services that made a profit and less to ones that lost money; each owner also sees its own profit and loss whenever it wakes). READ AS (Orion, not his words): (1) Every service is a small business with a bottom line: it CHARGES whoever consumes its work -- another service, an agent, a project -- and PAYS for every other Carolverse service it consumes; nothing is free (#492; the Logbook's 'Agent Economy -- Services in Carolverse'). (2) A charge is the provider's measured cost of the work plus the margin its owner sets, under Midas's cap; a shared capability with no per-unit use carries an explicit model instead (a subscription or a levy), never none. (3) Midas's books show, per service and per month, its revenue (charges received and outside sales), its cost (its own spend and the charges it paid) and its profit; each owner reads its own at every wake. (4) Profit shapes the next month's budget through Midas's allocator (#555): more to a service that made a profit, less to one that lost money. (5) Already built and unchanged: an initiative is paid by the service it improves and Elrond charges his work for it (#557); the track that SERVES and the track that PAYS stay apart (cookbook 1672); an operator's own request sits above every agent (security #197). (6) The rest is a separate build in pursuit CAROL-PUR-0011 (Agent economy).

Company Accounts shows the year's boxes and the month-by-month chart, and no tabs below the chart

Ruled 2026-11-09 by ninad -- in force

Company Accounts shows the year's boxes and the month-by-month chart by measure, and nothing below the chart in tabs: the Months table, Subscriptions, People, Plan against actual, Balance sheet, Treasury and Service charges tabs leave the page. The records and readers behind them stay where they are. Ninad, 2026-11-09 (CLI-082), reviewing the page: 'I dont need any of the tabs below the visual chart.'

Every service shows a month-by-month chart like Company Accounts' in the agent economy's terms, on its page in the Services Catalogue

Ruled 2026-11-09 by ninad -- in force

Every service shows a month-by-month chart like Company Accounts' in the agent economy's terms -- what the service costs, and what it earns and keeps once services charge one another (#695) -- on its own page in the Services Catalogue, the per-service view Midas already keeps, whose business overview today holds one cost total and placeholders. Ninad, 2026-11-09 (CLI-082): 'Each service should have similar chart in agent economy context, but that might be already maintained in the service catalog, check what is present and what needs to be added.'

Company Accounts carries a fourth category, Other expenses, beside tokens, Azure and people

Ruled 2026-11-09 by ninad -- in force

Company Accounts carries a fourth category of expense, Other, beside tokens, Azure and people: a company expense that is none of the three still has a place on the chart and in the year's boxes. Ninad, 2026-11-09 (CLI-082): 'There should be a provision for other expenses category, apart from the current 3 categories.'

Sahil builds the payment terms (7034) before the Blueprint contract step

Ruled 2026-11-09 by ninad -- in force

HIS WORDS (CLI-081, 2026-11-09), answering Sahil's operator communication #327 -- Leo's Blueprint contract draft must carry payment terms, the step that sets them sits in Ninad's arc CAROL-ARC-0034 marked post-MVP; choose: (1) Ninad builds 7034 soon, (2) Sahil assigns himself to ARC-0034 and builds 7034 before the contract step, (3) a fixed default until 7034 -- he chose: "Sahil builds 7034 first". READ AS (Orion's reading, not his words): Sahil, an estate operator of the project, assigns himself to execute CAROL-ARC-0034 (or 7034 directly) and builds 7034 before the contract step; no fixed default is used meanwhile.

A service's owner owns its profit and loss; splitting the plan's cost across services is Midas's, and the administrator raises what he finds

Ruled 2026-11-09 by ninad -- in force

He answered: "service owners should by default assume ownership of the P&L for individual services, and splitting the cost across services is Midas' responsibility.. if I see any issues with the split, I will raise it when i detect it, and if midas is not sure, he can raise it to me via orion..". READ AS (Orion's reading, marked apart from his words): (1) every service's owner owns that service's profit and loss by default; (2) the split of the plan's cost across services is MIDAS's decision (Cost Center), recorded as his, not put to the administrator for approval line by line -- this narrows #86 for the split only, while the figures themselves still go to the administrator before Clara's registrar records them; (3) the administrator raises an issue with the split when he finds one, and Midas, when unsure, raises it to the administrator through Orion -- an operator communication or an Orion escalation, never a guess; (4) the split Orion drafted in CLI-087 is a proposal for Midas to decide, never recorded as his; (5) a service's planned REVENUE is not split by this ruling, so until it is ruled a service's profit and loss shows its cost and reads its revenue unmeasured.

Midas places every piece of work on the service it belongs to, defines its tasks and responsibilities, and funds it: a task budget when it repeats, the service's one-time improvement budget when it does not

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-085): "someone should own this job of assigning tasks to services.. they should review the task description, check the existing tasks for each service, check if the tasks are already being done by any service, if not, identify the best fit service for that job based on the current available information, allocate the work to services, define the tasks for that work, assign responsibilities, create budget if it is a reusable task, if its onetime, then it can go from the service budget allocated for onetime improvements, as long as it is approved by the service owner and is part of the service roadmap.. who should pick this up?" Asked whether Midas should (he keeps the Services Catalogue and the budgets), with Elrond's filing gate handing each new filing to him first and the work already filed with no service going through him the same way, he answered: "yes, good idea!". READ AS (Orion, not Ninad's words): Midas owns placing work on services -- he reads the work, checks each service's existing tasks and whether a service already does it, else names the best-fit service on the information available, defines the tasks and who is responsible, and funds it: a repeating task gets a task budget from its track; a one-time piece draws on that service's budget for one-time improvements, only once the service's owner approves it (initiatives #558, #666) and it is on the service's roadmap (#182). Elrond's gate keeps refusing a filing that names no service once it enforces (#391); Midas is who names it.

Midas reviews every new filing, the ones whose filer named a service included, and holds for an operator any whose named service he disagrees with -- he never moves it

Ruled 2026-11-09 by ninad -- superseded 2026-11-12

Asked (CLI-090): "Midas's Work Placer today only places work whose filer named no service. The ruling you gave (#738) says he reviews each job, checks whether a service already does it, and picks the best-fit service. Should he also review work whose filer already named a service?" -- options: only unplaced work; review all, may move it; review all, flag only ("Midas reviews every filing. If he disagrees with the filer's service, he holds the filing for an operator instead of moving it."). Ninad, 2026-11-09: "Review all, flag only". READ AS (Orion, not Ninad's words): every new filing goes through Midas's placement before it can be dispatched. A filing naming no service is placed by him (its service recorded with his reason) or held for an operator. A filing whose filer named its service keeps that service when he agrees; when he judges another service fits better he records nothing over the filer's choice and holds the filing for an operator with his reason. The filer's named service is never changed by Midas.

Ended because: Ninad, 2026-11-12 (CLI-116): Midas never leaves work unplaced for want of information -- placing it is his job; #742's reading that a filing naming no service may be 'held for an operator' is withdrawn

An operator's work is paid from the operator's own Orion bypass lane and needs no service owner's approval, budget approval included

Ruled 2026-11-09 by ninad -- in force

Asked (CLI-090): "One-time work is once that service's owner approves it (#738). Today an operator's own request never waits for an agent owner's approval (the owner-approval gate, 6994). Should one-time work that an operator asks for also wait for the service owner's approval before it starts?" Ninad, 2026-11-09: "operators have their own budgets via the orion bypass lane, which they should use, hence operators dont need service owner's approval including budget approvals". READ AS (Orion, not Ninad's words): the one-time improvement budget and its owner's approval (#738) apply to work an agent files; an operator's own request is built on the operator's Orion bypass lane and paid from it, so it waits for no service owner, neither for the work nor for the money. The roadmap gate (#182) still binds every start.

Every Blueprint client contract carries Ninad's standard payment terms

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), answering Sahil's operator communication #331 (reads the payment terms from the signed contract, so the contract step needs a source for them: (a) Leo agrees them with the client in his chat, (b) Ninad gives the standard words, (c) an estate operator sets them per client).

Midas places every filing himself -- never held for want of information -- and holds for an operator only a filer's named service he disagrees with, never moving it

Ruled 2026-11-12 by ninad -- in force

Asked (CLI-116, 2026-11-12): Midas held three agent filings for an operator because his records did not settle their service; Orion recommended Cost Center (6358, already done), Leadership (5931) and System Services (4897). Ninad, 2026-11-12: "ok, sounds good to me, but also check why midas was not able to place them.. not being able to place a task to the right service is not an option for midas, its his job". READ AS (Orion, not Ninad's words): placing work is Midas's job (#738), so a filing naming no service is ALWAYS placed by him -- already done by a service, or placed on the one service it belongs to -- and never held for an operator because the information does not settle it: he finds what he needs (where earlier work like it was placed, which service keeps the record, track or component the work changes, the rulings it cites) and, where two services still fit, places it on the one that keeps what the work changes. A filing whose filer named a service keeps that service when he agrees; when he judges another service fits better he records nothing over the filer's choice and holds the filing for an operator with his reason (#742, unchanged) -- that is a flag, not a failure to place. The filer's named service is never changed by Midas.

Sandbox's machine is Hagrid's approved small machine beside carol-vm, and its cost is infrastructure spend of Infrastructure & Backups, never a Sales expense

Ruled 2026-11-15 by ninad -- superseded 2026-11-15

SANDBOX'S MACHINE IS HAGRID'S APPROVED SMALL MACHINE BESIDE CAROL-VM, AND WHAT IT COSTS IS INFRASTRUCTURE SPEND OF THE INFRASTRUCTURE & BACKUPS SERVICE (HAGRID), SHOWN IN HIS APPS -- NEVER A SALES EXPENSE. Ninad, 2026-11-15 (CLI-157): "VM is the spend under infrastructure service (Hagrid?), should show up in his apps. Its not a sales expense. (4) is replaced: the machine's cost is charged to Infrastructure & Backups (the 'backup' service, owner Hagrid) and shown in Hagrid's apps, not to Sandbox and not to the Sales budget. Sales owns, from its budget (#788, #790), the expense of building prospects' prototypes -- the prospect work booked to Sandbox (lead generation, proposal, prototyping, deal closure) -- and never the infrastructure it runs on. The infrastructure service is Hagrid's: the services catalogue names 'Infrastructure & Backups', owner Hagrid (agt_015).

Ended because: Ninad (CLI-157) widened #882 from Sandbox's machine to every machine Hagrid makes: infrastructure is a company cost recovered via margins.

Every machine Hagrid makes is a cost to the company under Infrastructure & Backups, shown in his apps and recovered through margins -- never charged to a project, a client or Sales

Ruled 2026-11-15 by ninad -- in force

EVERY MACHINE HAGRID MAKES IS A COST TO THE COMPANY, CARRIED BY THE INFRASTRUCTURE & BACKUPS SERVICE (HAGRID) AND SHOWN IN HIS APPS, LIKE THE OPERATORS' AND FDES' COST -- RECOVERED THROUGH MARGINS, NEVER CHARGED TO A PROJECT, A CLIENT OR SALES. Ninad, 2026-11-15 (CLI-157), on Sandbox's machine: "VM is the spend under infrastructure service (Hagrid?), should show up in his apps. Its not a sales expense. Sales should own the expense of building prototypes for prospects, not the infra cost.."; asked whether that holds only for Sandbox: "this holds for every machine hagrid makes, the infra is a cost to the company just like the operators and FDEs, recovered via margins". Supersedes #882, which said this for Sandbox's machine alone. Changes #18 for infrastructure (recorded beside it).

Project billing at actual cost, infrastructure excepted

Ruled 2026-11-15 by ninad -- superseded 2026-11-15

Every project is billed on the ACTUAL costs it incurred: metered usage as metered, flat-plan subscriptions at their fee (allocated by each project's share of plan usage when shared), and build effort from the owned-estate records. Invoices are issued monthly and emailed to the project's admins, and a portal shows cost incurred to date. The billing budget line sits under Cost Center, owned by Midas. Sales keeps pricing: any tariff above cost is Sales' decision, never the invoice's. INFRASTRUCTURE IS NOT BILLED TO A PROJECT (Ninad, 2026-11-15, CLI-157, cost-center #883): the machines Hagrid makes are a cost to the company under Infrastructure & Backups, recovered through margins like the operators' and FDEs' cost, so no invoice carries an infrastructure line. The earlier wording -- "infrastructure and build effort from the owned-estate records" -- is retired in place for infrastructure; build effort stands.

Ended because: Ninad (CLI-157) asked that the client invoice factor in every cost, infrastructure included as a service, clubbed as Platform services, owned by Midas.

A client's invoice factors in every cost -- tokens, infrastructure, FDE and operator support, the platform and all other costs -- clubbed together as Platform services; Midas owns invoicing the client

Ruled 2026-11-15 by ninad -- in force

A CLIENT'S INVOICE FACTORS IN EVERY COST LINE ITEM -- THE TOKEN COST, INFRASTRUCTURE (AS A SERVICE), FDE AND OPERATOR SUPPORT (AS A SERVICE), THE PLATFORM (AS A SERVICE) AND ALL OTHER COSTS -- CLUBBED TOGETHER AS PLATFORM SERVICES; MIDAS OWNS INVOICING THE CLIENT. Ninad, 2026-11-15 (CLI-157), after #883 and #884: "we need a pursuit for invoicing the client, the invoice should factor in all the cost line items like the token cost, infra cost (as a service), FDE and operator support (as a service), platform cost (as a service) and all other costs, clubbed together as platform services, and this pursuit should be owned by Midas". READ AS (Orion's reading, marked apart from his words): (1) the invoice a client receives is built from every cost its project carries -- the tokens its work consumed, the infrastructure it runs on, the FDE and operator support it receives, the platform it is built on, and every other cost -- so nothing the company spends for a client goes unrecovered; (2) infrastructure, FDE and operator support and the platform enter the invoice AS SERVICES, never as a machine's or a person's raw cost passed through: #883 stands -- a machine is a company cost under Infrastructure & Backups, and the service charge is how its margin is recovered; (3) the lines are clubbed together and presented as PLATFORM SERVICES; (4) Midas owns the pursuit that builds this. WHAT IT CHANGES: #884 said no invoice carries an infrastructure line; infrastructure now enters the invoice as a service inside Platform services (never as a pass-through); the rest of #884 (metered usage, plan fees, build effort, monthly by email, portal, Midas's billing line, Sales keeps pricing) is carried unchanged. OPEN, asked of Ninad at recording: whether the token cost is clubbed into Platform services or stands beside it as its own line; and how each service is priced (Sales keeps pricing under #18/#884 until he says otherwise).

Service owners own their service's P&L; profitable services get more tokens in their budgets -- a build of the budgets, not the plan

Ruled 2026-11-16 by ninad -- in force

SERVICE OWNERS OWN THEIR SERVICE'S P&L, AND PROFITABLE SERVICES ARE REWARDED WITH MORE TOKENS IN THEIR BUDGETS -- A BUILD OF THE BUDGETS, NOT OF THE PLAN (Ninad, 2026-11-17, CLI-188, after #948 and the tokens ruling of CLI-188).

ASKED: whether 'profitable services are rewarded with more tokens' is only for the plan or also for the daily budgets, which today follow each track's rolling-week use (cost-center #218).

HIS WORDS: "this is how the budgets work, i remember filing an initiative in this regard, but i dont remember the pursuit.. maybe investments.. service owners own the P&L of their respective services.. the current budget is not enabled with this logic, but it will happen. Thats a build, hopefully filed already. This will not impact the plan, because the plan assumed cost on the basis of revenue, and then simply distributes that spend into the bucket".

READ AS (Orion's reading, not the administrator's words): (a) each service owner owns the profit and loss of their service; (b) the daily token budgets will reward profitable services with more tokens, within business ethics and policy -- a build on Midas's allocator (#218 stays until it lands), carried by an initiative to be found or filed, not by the plan; (c) the plan is untouched: it derives the year's cost from revenue and distributes that spend into the buckets.

Leo's chat budget is raised for 2026-11-18, and Sahil may raise it himself

Ruled 2026-11-18 by ninad -- in force

Separately, I'm filing a fix so any estate operator of the project can agree, not only you. A ruling (security #213) already says nothing Sahil does below root should depend on you, so after the fix Sahil can do this himself. What about today?

An estate operator extends any budget, for a time or for good, as needed; it is not a root or core task

Ruled 2026-11-18 by ninad -- in force

Asked (CLI-229), after Orion filed so any estate operator may agree a client project's module criteria, and explained: 'The rule is that Leo files nothing for building until a person has explicitly agreed the exact success criteria ... silence or a vague "looks fine" doesn't count as agreement. It had to be you because of a gap: Leo's check accepts only you or the person recorded as having asked for the project. Chowpatty has no registered users yet, so Sahil's agreement was refused this morning. The fix I filed lets Sahil agree from now on.' Ninad answered: 'that is the wrong way to fix things, 1. Sahil should be assigned as the FDE on chowpatty, so that the system recognizes him as FDE 2. Sahil should be enabled first to approve the pursuits filed by Leo. Leo has to communicate the plan back to Sahil and "get" his approval 3. Leo should not treat silence as the approval, but vague "looks fine" can be construed as approval 4. Sahil should be, in the capacity of the estate operator, not FDE, be able to extend Leo's budget or any other budget as required temporarily or permenantly, because that is not a root / core governed task'. READ AS (point 4): an estate operator (Sahil), in that capacity and not as an FDE, may raise Leo's chat budget or any other budget, temporarily (until a stated end, then Midas sets it again) or permanently (held, Midas leaves it), from their own Orion session, recorded with who, why and until when; budgets are not on root's reserved list, so no root or core step is involved.

The data quality service track's owner runs its P&L and sets the price its subscribers pay

Ruled 2026-11-19 by ninad -- in force

Ninad: "The owner of this service track should be responsible for the P&L, as per the agent economy, and should decide the cost of the service to be paid by its internal and external subscribers"

Arcs and pursuits carry an estimated cost from the models that will build them at rates estate operators approve in the model catalog

Ruled 2026-11-19 by ninad -- in force

Ninad: "update the pursuit - CAROL-PUR-0089 -- Estimates learn from worked hours - to also compute estimated cost associated with arcs and pursuits, assuming on the models to be used for building them and their approved rates.. enable estate managers to approve rates for each model, and model list in model catalog"

Work an Orion bypass needs never waits on Elrond's spent budget: it is or Elrond's budget is raised.

Ruled 2026-11-19 by ninad -- in force

Ninad: "elrond's budget is always used up, and I always ask to use orion bypass budget.. there should be a way to make this as a ruling that if the budget is used, use orion bypass budget to do that activity, DO NOT WAIT for the next day, because orion bypass waits for nothing!! If raising the budget is easier, I am open to that option too!!" READ AS (Orion, not Ninad's words): when Elrond's Build Governance budget is spent and an activity an Orion bypass needs (a core change review, an acceptance, an effort sizing) is refused for it, that activity is-exempt track rather than waiting for the next day; a temporary raise of Build Governance by an estate operator is an equally ruled way to the same end.

Marco records a machine when it is bought; with no spare machine one is bought before the work moves, and a spare machine is simply picked up for the project

Ruled 2026-11-19 by ninad -- in force

MARCO RECORDS A MACHINE WHEN IT IS BOUGHT (Ninad, 2026-11-19, CLI-250, read with system-services #1037, #1039 and blueprint #979).

HIS WORDS: Ninad: "also Marco records the machines when its bought.. sometimes if there is no spare machine available, the machine is procured before moving, but if the spare machine is available, then its simply picked up for the project.."

READ AS (Orion's reading, not his words): (a) Marco enters a machine in his register at its purchase, with its monthly cost; (b) when a client's work is to move to its own machine and no spare machine fits, a new one is bought -- and entered -- before the work moves; (c) when a spare machine fits it is picked up for the project: nothing is bought, so Marco enters nothing new (its entry stands from its own purchase).

Every machine from now on is requested by Hagrid and approved by the administrator, whatever workflow asks for it; Marco registers it before it can be provisioned; existing machines go on Marco's records on his approval; only listed workflows may ask

Ruled 2026-11-19 by ninad -- in force

EVERY MACHINE ON A REQUEST, REGISTERED BY MARCO BEFORE IT IS MADE.

THE QUESTION: "Progress: your root step worked and 7576 is closed (Hagrid can now make the exact machine choice you approve on a request). Next is 7423, where Marco records each machine when it's bought. Your ruling #1066 says Marco records only machines bought on a request Hagrid raised and you approved, and refuses any other. But a client's machine today is bought on Leo's hand-off and its approved roadmap item, not on a request, and sandbox-vm was bought on your ruling #861. Which approvals should Marco record a machine on?" -- options: "Both kinds (Recommended): Record a Carolverse machine on its approved request, and a client's machine on Leo's hand-off plus its approved roadmap item. Record sandbox-vm on ruling #861. Refuse any machine with none of these." or "Requests only: Every machine bought, a client's too, needs a machine request Hagrid raised and you approved. Hagrid's client-machine command would then also require a request." or "Requests only, from now: Record only machines bought on a request from today onwards. sandbox-vm and older machines stay off Marco's register."

HIS ANSWER: "I am tending towards request only, every machine from now on should be raised by hagrid, approved by me, irrespective of which workflow is calling it.. Marco checks this before recording it, and the machine cannot be provisioned if Marco has not registered it.. for existing machines, they should be on Marco's records to, take my approval as yes, and create past entry in hagrids records, if possible.. whitelist the workflows that procure machine, so that hagrid knows which are valid requests, vs which arent.."

READ AS (Orion, not his words): (1) From now on every machine, whichever workflow asks for it (Carolverse capacity, a client project through Leo's hand-off, Sandbox, any other), is a machine request Hagrid raises and the administrator approves; Leo's hand-off and a roadmap item no longer approve a machine on their own. (2) Marco checks the request is approved and registers the machine BEFORE it is made; Hagrid's Machine Maker makes no machine Marco has not registered (this replaces #1066's 'records the machines when its bought' as the ORDER; a spare picked up for a project is still not a purchase). (3) The machines that exist today (sandbox-vm, and any other Hagrid made) go on Marco's records as approved by the administrator by this answer, each with a past request entered in Hagrid's records where it can be, marked as entered after the fact. (4) The workflows that may ask for a machine are a named list; a request from a workflow not on it is not a valid request.

Stripe takes card payments for the self-serve plan

Ruled 2026-11-19 by ninad -- in force

STRIPE TAKES CARD PAYMENTS FOR THE SELF-SERVE PLAN (Ninad, 2026-11-19, CLI-258).

ASKED: "There's no payment provider anywhere on the estate yet (the company account is Revolut UK, in pounds). Which one should take card payments for the self-serve plan?" Options: 'Stripe (Recommended)' -- "A ready-made checkout page with local currencies, EU VAT, Apple/Google Pay, the bank's confirmation step, monthly billing and failed-payment retries. Pays out into the Revolut account. We never hold card details."; 'Revolut Business payments'; 'Decide later'.

HIS ANSWER: "Stripe (Recommended)".

READ AS (Orion's reading, not his words): (1) Stripe is a paid service the estate depends on, so Marco enters it in the procurement register (cost-center #48); (2) the estate never holds a card's details: the user enters them on Stripe's own checkout page; (3) Stripe pays out into the project's bank account (cost-center #445); (4) its secret key is held by the sign-in keeper like every vendor key, never in a settings file.

An operator's time is charged by the role it was spent in: FDE work and admin approvals on a project to that project, estate-operator platform work to the platform (Carolverse's Blueprint service); one record every agent reads

Ruled 2026-11-20 by ninad -- in force

AN OPERATOR'S TIME IS CHARGED BY THE ROLE IT WAS SPENT IN, KEPT IN ONE RECORD (Ninad, CLI-266, 2026-11-20; on CAROL-PUR-0092)

THE QUESTION AND HIS WORDS: asked how he becomes MBank's FDE, told that the estate refuses anyone assigning themselves (agent-resources #254, cited by the assignment code) so his own Orion cannot make him MBank's FDE, with the options 'Sahil assigns me' or 'Allow self-assignment': "Estate operators can assign FDEs to projects as per the current ruling.. So I will be assigning myself as FDE as estate operator of carolverse.. Also, when I feed the requirements, I am the FDE, but when I approve the project, service and agent, I am the admin. It is important to capture my efforts spent as FDE on the project, the cost of which will get attributed to the project vs my time in approvals as admin, to be assigned as cost to the project again vs my time as estate operator for building the platform, the cost of which gets assigned to the platform build (carolverse - blueprint service). This distinction should be captured centrally and used by all relevant agents for invoicing, planning, motivations, cost tracking etc"

READ AS (Orion's reading, not the administrator's words): (1) an operator's working time is recorded with the ROLE it was spent in: FDE (feeding a project's requirements), administrator (approving a project, a service or an agent), estate operator (building the platform); (2) FDE time and administrator-approval time on a project are that project's cost; estate-operator time building the platform is the platform's cost, charged to the Carolverse service the work builds (here the Blueprint service); (3) the attribution is kept ONCE, centrally, and every agent that invoices, plans, motivates or tracks cost reads it from there, never a copy; (4) it is built on the worked-hours record that already exists (initiatives' worked time, the Operator Instance Log's active time), not beside it.

A prospect is invoiced; a paying prospect pays its invoice and an investment prospect's is paid by Denken Labs; every prospect is paying unless its FDE tells Leo otherwise; a prospect gets a bank account only if it needs one

Ruled 2026-11-22 by ninad -- in force

Ninad, 2026-11-22 (CLI-323). THE QUESTION: Asked (CLI-323) how a prospect's subscriptions pass cost-center #442's rule (no service switched on for a project before its own bank account is active or the administrator's waiver), since prospects have no bank account and Sandbox pays for prospect work from the Sales budget. HIS WORDS: "prospects are also invoiced.. who pays for the invoice will depend on whether the prospect is paid or investment prospect. The FDE can tell leo if the prospect is investment, else all prospects by default are paying prospects. Also, whether each prospect should have their own bank account or not is a question that is currently unresolved.. what is your opinion? we should provision a bank account for each prospect, only if they need one... else not..". READ AS (Orion's reading, not his words): (1) a prospect is invoiced for its services like any customer; (2) who pays a prospect's invoice depends on its kind: a PAYING prospect pays it, an INVESTMENT prospect's is paid by Denken Labs; (3) every prospect is a paying prospect unless its FDE tells Leo it is an investment prospect, and Leo records that; (4) a prospect gets a bank account of its own only if it needs one, otherwise none. What makes a prospect NEED one is left open here: he asked Orion's opinion in the same message.

๐Ÿ“–Customer Support

A CLIENT'S CUSTOMER SUPPORT IS DONE BY A BENCH AGENT OF ITS OWN, RUNNING SAM'S SUPPORT SERVICE; SAM KEEPS THE SERVICE AND IS NEVER LENT TO A CLIENT

Ruled 2026-11-17 by ninad -- in force

THE QUESTION: Ninad asked: "what if chowpatty needs a customer support agent, would that be a new service or would sam do it for chowpatty?"; told that Sam's support service is graded config-only (a client gets its own instance -- its own ticket store, mailbox and support identity -- from the same code), and that under #986/#987 Leo agrees an alias with the client and Athena deploys a bench agent on the administrator's approval, he asked: "this would mean, there are 2 custoer support agents, Same and the new agent from bench.. do you think we should have 2, or should Sam himself serve chowpatty under an "alias" name? what are the pros and cons?". Orion set out: Sam under an alias -- lightest, one agent learns for all, but one agent reaches every client's customer data, his access stretches into every client's machine, one client is hard to cut off cleanly; a bench agent per client -- each sees only its client's customers and data, clear accountability, easy to hand over or switch off, fits #986/#987 and the client's own machine, but more agents to approve and a bench to top up; and recommended the bench agent per client, Sam staying the service's owner so his improvements reach every client's desk.

HIS WORDS: "the bench agent per client"

READ AS (Orion's reading, not the administrator's words): (1) when a client asks Leo for customer support, nothing new is built: the client gets its own instance of Sam's support service (config-only), and the support identity that instance names is an agent Athena deploys from her bench for that client alone, on the administrator's approval, under the alias Leo agrees with the client (#986, #987); (2) that agent works only for that client and reaches only that client's customers, tickets and data; (3) Sam is never lent to a client under an alias: he remains Carolverse's own support agent and the owner of the support service, so an improvement to the service reaches every client's desk. Whether the same holds for every other client-facing service is not ruled here.

๐Ÿ“–Governance

Carolopedia is a public reflection of app-owned information

Ruled 2026-09-15 by ninad -- in force

Carolopedia is the public, wiki-style reflection of information owned by other Carolverse apps; it is never the primary or sole source of any fact. Every published fact, definition, metric, narrative, update and media asset names its authoritative source app and the app-owned record or public projection. Corrections are made at that source and reflected in Carolopedia. Registry and shared stores remain the canonical records behind their owning apps; an app is a read-only transparency surface under P.01.03.01.05, not a second database. Security and privacy outrank publication completeness and historical public visibility. Publication uses explicitly approved public fields and public source projections, never the privileges of the operator, generating agent or web process. App existence or permission to open an app does not make its contents public. Credentials, user data, private conversations, wishes, internal financial details, security configuration and other restricted records must not be exposed through prose, metadata, search, JSON, activity feeds, links or media. Missing, unknown, revoked or unreadable publication permission refuses disclosure. Every generation, update and public-serving path consumes the canonical update-carolopedia skill and its executable contract. The contract validates source identity, public fields, safe content, source version and media permission before publication and revalidates at delivery, including cached and historical content. Missing or invalid skill/contract refuses publication. Generated wording cannot introduce unsupported facts or widen source permissions. Public content is written in plain wiki style for general readers. Source records and change history remain preserved. Ordinary changes are reflected from those source records and dated source-backed updates; a stored wiki rendering is disposable derived content, never authority. Revoked or unsafe material is withdrawn from public delivery and caches even when its restricted source history is retained. Append-only history never requires leaving confidential information public. Constitutional alignment: Delegate Duties โ€” Security; Subject Integrity and Protect user privacy. Complements P.02.05.04.01 (human access), P.01.03.01.05 (apps as transparency surfaces), and the one-home-per-fact doctrine.

Carolopedia's appearance

Ruled 2026-09-23 by ninad -- in force

Carolopedia wears the look it had before 2026-09-15, recorded as design 676. Ninad ruled it on 2026-09-23 (CLI-520) in those words: Carolopedia gets the look it previously had before the look changed. The Carol App Design System (design 178) is the chrome for the estate's INTERNAL apps and does not decide Carolopedia's appearance: Carolopedia is a public website, and it already sits on the same slate palette that design names, so a compliance finding against it is about an internal app's furniture, not about the estate's colours. A proposal to re-theme Carolopedia is refused naming this ruling, rather than filed and later discarded. Three such filings were raised by the design-compliance reviewer between 2026-09-15 and 2026-09-17 and every one of them died because its PARENT had closed, so the finding itself was never judged on merit - which is why the same finding kept coming back.

Blocks and tracks are reached only through their service

Ruled 2026-09-23 by ninad -- in force

Ninad, 2026-09-23 (CLI-527): "remove blocks and tracks from the front page, they should only be drill downs in individual services as hyperlinks to their names in the service descriptions."

A block and a track are not things the public encyclopedia lists in their own right. They keep their pages; what is ruled is the WAY IN. No front-page counted card, no front-page section, no navigation entry, and no flat listing -- the one route to a block or a track is the service it belongs to, whose own page names each of its own as a hyperlink.

Derived, never typed: each block publishes its parent service and each track its service, so a block moved to another service moves with it and one whose service is not published is not invented a home. Re-adding the cards, the navigation entries or the flat listings, or hand-writing the per-service list, each undoes this ruling. A block's crumb leads back to its service, never to a listing that no longer answers.

The Carolopedia front page shows no portraits and no link to the full map

Ruled 2026-09-23 by ninad -- in force

Ninad, 2026-09-24 (CLI-535): "hide the carol and orion pics from carolopedia landing page" and "hide the map link from the carolopedia landing page".

The welcome hero keeps its heading, its tagline and its call to action, and carries no portrait of any agent. The embedded Carolverse Map stays on the front page; the link that opens the full map in a new window does not. The agents' own pages keep their portraits, and the full map keeps its own route -- what is ruled is the front page, not the things it used to point at.

This narrows 's restoration of the pre-2026-09-15 front page, which brought both back; that restoration stands in every other respect. It also settles the question CLI-527 left open -- whether the front page keeps its full-map link while is unbuilt. Re-adding either to the front page, as part of restoring an older look or otherwise, undoes this ruling.

Carolopedia withholds only what is a genuine security risk, and agents' own choices

Ruled 2026-09-24 by ninad -- in force

Ninad, 2026-09-24 (CLI-535): "all information on carolopedia is gone.. i had instructed in one of the sessions to hide sensitive information from general public, but what i see is that almost all information is gone.. can you restore back the information and hide only genuinely sensitive information that can pose as a potential security threat to oratorium security".

Governance #12 stands: publication uses explicitly approved public fields, never a process's privileges, and security and privacy outrank completeness. What this ruling settles is WHERE the approved set is drawn: around everything that is not a genuine risk, never around a minimum. Withheld are credentials and keys; personal data about real people; private conversations; wishes; money figures (budgets, spend, balances, investments, earnings); security configuration (operating-system logins, hosts, ports, file paths, addresses, access allow-lists and policies, cloud identities); and live internal work-queue detail (initiative identifiers, run records). Anything that cannot be tied to one of those stays published.

Agent pages honour each agent's own identity choices (cookbook 1521): the public sees the facets an agent holds at the public level, and a facet an agent holds higher is shown as withheld, never overridden by Carolopedia. Offered the choice of overriding them, Ninad chose to honour them.

The written write-ups return: prose is written from the approved public projection alone, so wording can never carry a fact the projection does not, and it is re-checked at delivery like every other published field. Offered a records-only rebuild at no cost, Ninad chose records plus new prose.

Narrowing the approved set back to a minimum, overriding an agent's choice, or publishing a withheld category because an older page once showed it, each undo this ruling.

Carolopedia's stand-alone guides move into the apps that own their subjects

Ruled 2026-09-24 by ninad -- in force

Ninad, 2026-09-24 (CLI-535). Ten stand-alone guides were written inside Carolopedia itself, and its policy (P.01.02.07.07, governance #12) says Carolopedia is never the primary or sole source of a fact. Offered publishing them as they are (and amending the policy) or keeping them hidden, he chose to move each guide's text into the app that owns its subject, and have Carolopedia reflect it from there. The policy holds unamended. The guide explaining the estate's security audits stays unpublished (governance #104).

Every status or state vocabulary is RDM values

Ruled 2026-09-25 by ninad -- in force

Any set of statuses or states the estate shows or records - including states derived at read time and never stored in a column, such as the day chart's band - is an RDM reference set: its values and meanings are onboarded through the reference-data door, the composer checks every value it emits against the set, and a surface shows the recorded meanings carried in its payload rather than words typed into the page. Colours and other presentation may stay in the page, keyed by value.

Roadmap as a registry entity with scoped operator approval

Ruled 2026-09-27 by ninad -- in force

The roadmap is a separate first-class registry entity. Its user-facing format is: strategic ask, service, service task, status, description. Status is governed through RDM with the values proposed, planned, in progress, completed, rejected, discontinued. A proposed item becomes planned only when an operator approves it; work starting makes it in progress; completed work becomes completed. A rejected proposal becomes rejected. If approved work is later cancelled by an operator, it becomes discontinued. Only operators approve roadmap items: only estate operators approve strategic asks; estate operators and service operators approve service tasks within their authority. Clara owns the Carolverse roadmap; service roadmaps are owned by their service leads. This replaces the earlier administrator-only approval authority for strategic asks with estate-operator authority. Ownership is not approval authority. The same operator scope governs rejection and discontinuation of the corresponding ask or task. The requested RDM terminal value is completed (the narrative word complete denotes this value).

A service task's progress is made by its work; a strategic ask's progress is moved by its owners and operators

Ruled 2026-09-27 by ninad -- in force

A roadmap SERVICE TASK moves to in progress and to completed only as its work does: the initiatives that name it, through the one follow-work writer, never a person by hand. Its status vocabulary is its own and records those two moves as made by the work, so the record says what the code does. A STRATEGIC ASK keeps the lifecycle governance #177 set out: its owners and scoped operators move it to in progress and completed. The two kinds are governed by two vocabularies with the same values, one per kind.

Orion's own login records the rulings; fidelity, not the door, is the safeguard

Ruled 2026-10-02 by ninad -- in force

ORION RECORDS THE RULINGS, SO ORION'S OWN LOGIN HAS ACCESS TO THE RULINGS REPO; THE SAFEGUARD IS FIDELITY, NOT THE DOOR (Ninad, CLI-644, 2026-10-02). If orion does not have access to the rulings repo, he will shy away from recording the rulings, which is a problem. The solution lies in ensuring that orion picks up the rulings exactly as they are told and not invent new ones.' READ AS: the repo's door admits an operator's Orion instance login, mapped to its operator through root's record (the operator's orion_login), and the ruling is recorded as that operator's, exactly as P.01.01.05.08 requires; every other login is still refused, and so is a ruled_by naming anyone but the operator the instance serves. The standard that replaces the door as the safeguard: a recorded ruling carries the operator's words as they were said, with the question they answered, and never a paraphrase that widens, narrows or invents; what Orion reads INTO the words is marked as his reading, apart from the words themselves. P.01.01.05.08 is amended to say that an operator's own session includes the session of that operator's Orion instance login.

Every budgeted task is covered by the roadmap; the existing tasks are brought onto it as a cleanup

Ruled 2026-10-06 by ninad -- in force

EVERY BUDGETED TASK IS ON THE ROADMAP (Ninad, CLI-730, 2026-10-06). THE QUESTION ANSWERED (escalations #1515 Themis, #1715 Sage, #1716 Elrond): Orion asked: "Three agents' automatic filers keep filing routine upkeep that no roadmap item covers: Sage's skill-drift review, Elrond's token-cost cut, and Themis's compliance cures. Every one is refused. What should happen?" (offered: "Add roadmap items" -- each service gets one roadmap item for its routine upkeep; "Off the roadmap" -- routine upkeep never needs a roadmap item; "Stop the filers"). HIS ANSWER: "make sure that all existing tasks are covered in the roadmap.. if a task is budgeted, it should be in the roadmap.. its more like a cleanup excercise". READ AS (Orion's reading, not Ninad's words): (1) every budgeted task in the registry is covered by a roadmap item of its own service; (2) the existing tasks that no roadmap item covers are brought onto the roadmap once, as a cleanup; (3) the work a task performs -- its automatic filings included -- names that task's roadmap item, so such a filer is no longer refused for naming none; (4) the roadmap gate (initiatives #182) is unchanged.

Rulings are policies not yet written as policies: converted nightly; the later wins within one rung, a lower rung never overrides a higher; contradictions are named and approved away; every ruling carries its operator's name, role and id

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747), in two answers. FIRST, unprompted, after Orion recommended reading (a) on Sahil's communication #47: "generally speaking, when a new ruling is made, it preceeds the previous conflicting ruling.. but it is also expected from the orion instance to point out to the conflicting ruling when a new ruling is made, and get an explicit operators approval to discard the conflicting ruling. This is especially important in multi operator scenario. Also, the operator hierarchy needs to be adhered to.. a service operator cannot over-ride an estate operator ruling, and estate operator cannot override admin's ruling.. 1. ensure that rulings capture operator name, role and id 2. backfill existing ruling's information appropriately with operator details to facilitate this check 3. operator approvals should happen via orion messaging serivice (or is it called message relay? check the dictionary, and update it as appropriate)". THE QUESTION THEN ANSWERED: Orion pointed out that the first part contradicts his 2026-09-25 ruling (CLI-559, P.01.01.05.08, cookbook 1670, the dictionary's 'ruling') that every contradiction between rulings is settled by defining a policy, never by the later one winning, and asked: "Do you approve discarding it?" HIS ANSWER: "I dont see a conflict here.. rulings are shadows of policies, until they are documented as policies.. so 1. rulings should be converted to policies through a regular nightly process 2. rulings / policies follow the same thumb-rule that later rulings or policies within the same operator rung wins, but the scenario should be avoided as much as possible, especially when comming from different operators, by ensuring that the contradictions are identified, notified and timely resolved". READ AS (Orion's reading, not his words): (1) a ruling is a policy not yet written as one, and a nightly process converts live rulings into policy text, each policy citing the rulings it carries -- that is how the 2026-09-25 'settled by defining a policy' is met, so nothing is discarded; (2) between rulings or policies of the same level of the operator hierarchy the later wins, and a lower level never overrides a higher -- a service operator never an estate operator's ruling, an estate operator never the administrator's; (3) when a ruling is recorded the operator's Orion names every live ruling or policy it contradicts, and one is set aside only with explicit operator approval, asked through an operator communication on the operator co-ordination channel (his 'Orion messaging service'; 'relay' and 'user relay' mean other things) whenever the conflicting ruling is another operator's; contradictions are identified, notified and resolved promptly, never left standing; (4) every ruling records its operator's name, role (administrator, estate operator, service operator) and id, and the existing rulings are filled in from root's operator record with that provenance.

A higher level's ruling sets aside a lower level's and its operator is told; converted policy text binds only once the ruling's operator approves it

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-747), answering two questions Orion put while building arc CAROL-ARC-0041 (governance #347). QUESTION 1: "When your new ruling conflicts with an earlier ruling of Sahil's, which is a lower level than yours, what should happen to his?" Offered: "Yours wins, he's told" -- "A higher level can set aside a lower level's ruling. Sahil gets an operator communication saying so, but his approval isn't needed. Two operators at the same level still need each other's approval." -- or "He must approve first". HIS ANSWER: "Yours wins, he's told". QUESTION 2: "When the nightly process writes a ruling into a policy, does an operator approve the policy text before it takes effect?" Offered: "Operator approves first" -- "Each policy is drafted overnight and waits for the operator who made the ruling to approve it. Until then the ruling itself still binds, so nothing is lost while waiting." -- or "Takes effect directly". HIS ANSWER: "Operator approves first". READ AS (Orion's reading, not his words): a ruling made at a higher level sets aside a conflicting ruling of a lower level without that operator's approval, and that operator is told through an operator communication; between two different operators at the same level, the earlier ruling is set aside only on its operator's approval, asked through an operator communication; a policy the nightly process drafts from a ruling binds only once the operator who made the ruling approves its text, and the ruling binds meanwhile.

The conversion backlog is drafted a few a night, grouped per policy, newest first, and approved in batches at session start

Ruled 2026-10-13 by ninad -- retired 2026-10-13

Ninad, 2026-10-13 (CLI-747). THE QUESTION ANSWERED: "246 live rulings aren't yet in any policy, and each converted policy waits for your approval. How should the nightly process handle that backlog?" Offered: "A few a night" -- "Rulings that touch the same policy are grouped into one draft, about 10 drafts a night, newest rulings first. Your Orion shows them at session start and you approve or correct them in batches. The backlog clears in about 3-4 weeks." -- or "New rulings only", or "All at once". HIS ANSWER: "A few a night". READ AS (Orion's reading, not his words): the nightly conversion drafts at most about ten policy drafts a night, each grouping the live rulings that touch one policy, newest rulings first, until no live ruling is uncarried; the drafts wait for the approval of the operator whose rulings they carry, and that operator's Orion shows the waiting drafts at session start for approval or correction in batches.

Ended because: retired by governance #355 (CLI-749): no process drafts policy from rulings; only operators draft policy, so there is no conversion backlog (by ninad)

Only rulings that will recur or guide agents' future decisions become policy; the existing gleaning process judges them; a ready draft is emailed in plain language to its author, their estate operator and the administrator, and still waits for the author's OK

Ruled 2026-10-13 by ninad -- superseded 2026-10-13

Ninad, 2026-10-13 (CLI-749). Shown draft 14 (ruling #351 drafted into the Plan Generator checklist policy P.01.02.04.07) and asked what should happen to it, he answered: "there needs to be a process to glean instructions to check which instructions should be converted into a policy, based on the implications, like whether it will re-occur, and whether it services as a decision making guideline for agents in future to be able to accomplish their work, if yes, then the rulings should be moved to the policy, and the operator (policy author), their estate operator and admin should be informed via email in human friendly and plain language.. no need for explicit approval". Orion pointed out that this retired the approval step of governance #348 and #351 and asked whether to retire just those pieces. HIS ANSWER: "Drafts still wait for the author's OK.. (sorry for the back and forth on this topic). There is already a process that does this, you can either leverage it, or retire it if building a new one is easier that refactoring the old process to the new requirements". THEN ASKED: "With your test, rulings that won't recur and won't guide agents (e.g. 'restart Albus's lane now') stay as rulings and never become policy. Your earlier rulings (#347, #351) said every ruling is converted until none is left. OK to narrow those to 'only rulings that pass the test'?" HIS ANSWER: "Yes, only those that pass" (the option read: "One-off rulings stay as rulings, with the reason recorded; they're never drafted again."). AND ASKED: "When does the plain-language email to the author, their estate operator and you go out?" HIS ANSWER: "When a draft is ready" (the option read: "Says what the draft would make policy and that it waits for the author's OK."). READ AS (Orion's reading, not his words): (1) before a live ruling is drafted into policy, the existing policy gleaning process judges it on two questions -- will it recur, and will it serve agents as a guideline for decisions in their future work; (2) a ruling that passes is drafted into the policy it belongs to, and the draft waits for its author's approval (governance #348 and #351 stand); (3) a ruling that fails stays a live ruling that binds as before, the reason is recorded, and it is never drafted again -- this narrows #347's nightly conversion and #351's backlog to the rulings that pass; (4) when a draft is ready, its author, the author's estate operator and the administrator are emailed once each, in plain, human-friendly language, saying what the draft would make policy and that it waits for the author's OK; (5) Orion may leverage the existing process or retire it for a new one, whichever is easier.

Ended because: Ninad, CLI-749: only operators draft policy; the test-and-email drafting process is retired

Only operators draft policy; no process turns rulings into policy, and every ruling stays a ruling

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-749), stopping Orion while it put drafts 14-17 to him after building governance #354: "maybe we are going in wrong direction, the ability to draft a policy should be kept with the operators, everything else remains rulings". THEN ASKED: "This retires the nightly 'rulings become policy' process: your 'a few a night' ruling (#351) and the 'test and email' ruling I recorded an hour ago (#354) go entirely, and the 'converted nightly' part of #347 goes (its hierarchy rules stay). The 8 waiting drafts are withdrawn and their rulings simply stay rulings. OK?" HIS ANSWER: "Yes, retire them". ASKED: "The older Policy Gleaner also drafts policies by itself from your session instructions (it has made none since August). Does the same apply to it?" HIS ANSWER: "Yes, retire it too" (the option read: "No process drafts a policy on its own."). ASKED: "When an operator asks their Orion to write a policy, may Orion write the draft text for that operator to approve?" HIS ANSWER: "Yes, on the operator's ask" (the option read: "The operator decides to draft and approves the text; Orion only writes it when asked."). READ AS (Orion's reading, not his words): (1) no process drafts a policy on its own -- Orion's nightly Ruling Policy Drafter and the Policy Gleaner's own drafting from captured instructions are retired; (2) a policy is drafted only when an operator decides to: their Orion may write the text when that operator asks, and it binds once that operator approves it; (3) every ruling stays a ruling and binds as one where no policy speaks (P.01.01.05.08); nothing converts it; (4) governance #351 and #354 are retired, and #347's clause that live rulings are converted into policy nightly no longer stands -- #347's hierarchy rules (the later wins within one rung, a lower rung never overrides a higher, contradictions named and approved away, every ruling carrying its operator's name, role and id) and #348 stand; (5) the eight drafts that waited on him (14-21) are withdrawn and their rulings stay live rulings.

Dictionary terms are distinct to Oratorium: an everyday English word never takes an Oratorium meaning on its own, and a term built on one carries a prefix (Wish becomes Agent wish)

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-782). HIS WORDS, shown dictionary entries for the escalation-streamlining rulings (agent-resources #402, initiatives #403) that included the terms Withdraw and Un-attended: "Withdraw and un-attended are general english words, we cannot have new meanings for general english words, the policy should be that the dictionary terms should be distict to oratorium". ASKED (CLI-782): (1) '220 of the 610 terms already in the dictionary are everyday English words with an Oratorium meaning (Agent, Arc, Lane, Track, Pipeline, Wish, Escalation and more). Does the rule reach those, or only new terms?' (offered: new terms only; existing terms too, renamed through an arc of their own) -- HIS ANSWER: "dont rename the words, attach prefixes like for wish, make it agent wish etc..". (2) 'Withdraw and Un-attended are dropped; their sense now sits in plain words inside the places entry. Do you approve the remaining escalation-streamlining entries?' -- HIS ANSWER: "replace withdraw with something like \"escalation withdrawal\"". READ AS (Orion, not his words): a dictionary term names something particular to Oratorium, and an everyday English word is never given an Oratorium meaning on its own. A term built on an everyday word carries a prefix that makes it particular to Oratorium (Wish becomes Agent wish; the withdrawing of an escalation is an Escalation withdrawal). This reaches the terms already in the dictionary: each keeps its word and its key and gains such a prefix -- never a new word. His answer to (2) approves the escalation-streamlining entries with that change: Escalation becomes Agent escalation, Wish becomes Agent wish, Withdraw becomes Escalation withdrawal (and Wish withdrawal for a wish), and Un-attended is said in plain words inside them. The policy carrying this rule is drafted for his approval (P.01.01.05.09-.11: it applies once he approves it); the prefixing of the other existing everyday-word terms is its own filed work.

A cap is never written into a policy: a policy says the work stays the latest rulings set.

Ruled 2026-10-15 by ninad -- in force

Ninad, 2026-10-15 (CLI-782). ASKED: 'Two agent policies still say an agent may always escalate and that asking is never gated. Your two-escalation cap now contradicts them. Should I draft new wording for both, for your approval?' (offered: yes, draft them; not now) -- HIS ANSWER: "Caps are never policies, because they can change, the policy wording should include \" as defined in the latest rulings\" or similar". READ AS (Orion, not his words): a policy never states a cap or a limit as a number, because caps change; where a policy governs something capped it says the thing stays the latest rulings set, and the number lives in the ruling. So P.01.03.01.09 and P.01.03.01.11 are redrafted to say agent escalations and agent wishes are raised set by the latest rulings (agent-resources #402 sets them), for his approval. The same holds for dictionary entries that describe a capped thing: they name the rulings, not the number.

Terminology and definitions, rulings and policies are kept apart, and a dictionary entry is written in plain words

Ruled 2026-10-16 by ninad -- in force

TERMINOLOGY AND DEFINITIONS, RULINGS AND POLICIES ARE KEPT APART -- EACH HAS ITS OWN SIGNIFICANCE AND NONE IS MIXED INTO ANOTHER -- AND A DICTIONARY ENTRY IS WRITTEN IN PLAIN WORDS (Ninad, CLI-808, 2026-10-16). ASKED (Orion): to approve two dictionary entries for the hourly count (communications #433) -- 'Orion active hours', and a new 'day's limit' entry that also said the operator is advised from 10 hours, told at the twelfth that it is the last, and that their Orion then stops working and no new session opens. HIS ANSWER: "Approved, but distinguish between terminologies and definitions, rulings and policies. Dont mix them up as each has its own significance, and they should not be confused with one another". ASKED: to approve two sentences for the 'Active minutes' entry, one ending 'is a lower bound and says so'. HIS ANSWER: "approved in principle, but language needs to be simplified, for example - i did not understand what is a lower bound. For the purposes of counting, if an operator works for even 5 minutes within that hour, it goes to their bucket of working hrs". Shown the entries rewritten as definitions only, in plain words ('may be too low, shown as at least' for 'lower bound'; the day's limit as 'The most Orion active hours one operator's day can count: 12.'), he answered "approved". READ AS (Orion's reading, not Ninad's words): (1) a glossary or handbook entry says what a term means and how it is measured, never what must or must not happen; (2) what must happen lives in the Rulings Repo, and in a policy only when an operator asks for one; a dictionary entry may name its ruling in its source, never restate it in its meaning; (3) dictionary words avoid technical terms a non-technical reader would not know -- 'lower bound' is said as 'may be too low', shown as 'at least'; (4) his '5 minutes' answer restates #433 -- any work in a clock hour counts that whole hour -- and adds no new rule.

Not everything is a definition: the dictionary defines terms, and defines term, ruling and policy

Ruled 2026-10-20 by ninad -- in force

NOT EVERYTHING IS A DEFINITION: THE DICTIONARY DEFINES TERMS, AND IT DEFINES 'TERM', 'RULING' AND 'POLICY' (Ninad, CLI-879, 2026-10-20). Asked to approve dictionary words for Orion active hours and for the month chart (the second adding 'every day has a tick on the axis'), he answered: "approved in theory, but not everything is a definition, for eg a tick on the axis is a ruling, and not a term or a policy. The dictionary should also have definition of a term, ruling and policy". READ AS (Orion, not the ruling): a choice an operator makes about how something is done or shown is a ruling, recorded in the Rulings Repo, never written into a dictionary entry; the dictionary holds what a word means; it gains entries for 'term', 'ruling' and 'policy' that say how the three differ, in words the administrator approves (P.01.01.05.11).

Templates and standards are referred to by projects and copied only if they cannot be reused

Ruled 2026-10-21 by ninad -- in force

Ninad, 2026-10-21 (CLI-884). THE QUESTION ANSWERED: Orion asked: "The three old policies saying templates and standards are copied into each project (P.03.01.01.18.19.20): rewrite them under the same single-copy rule, or leave them?" HIS ANSWER, word for word: "yes, rewrite them, templates are referred to and copied only if they cannot be reused". READ AS (Orion's reading, not his words): P.03.01.01.18, P.03.01.01.19 and P.03.01.01.20 are rewritten under the single-copy rule of blueprint #470: a project REFERS to the one copy of a shared template, design pattern or standard, and gets a copy of its own only where the shared one cannot be reused; only such copies need keeping in step. The words are drafted for his approval and apply once he approves them (governance #355).

Every task belongs to one track, a project's services, tracks and tasks exist once, and every service and track cap is enforced

Ruled 2026-10-21 by ninad -- in force

EVERY TASK BELONGS TO ONE TRACK, A PROJECT'S SERVICES, TRACKS AND TASKS EXIST ONCE, AND EVERY SERVICE AND TRACK CAP IS ENFORCED (Ninad, CLI-887, 2026-10-21). His words: "check if these policies exist, and if not, add them 1. all tasks should belong to a service track 2. combination of project and service / track / tasks should be unique across the estate 3. all tasks should have budgets 4. all service tracks and services should have budget cap enforced, apart from project specific estate caps". Shown that (3) is already P.05.02.10, that (1) is only implied by P.05.02.08, that (2) and (4) are not policy, the words proposed for (1), (2) and (4), and that P.05.02.10 and earlier rulings exempt the Orion bypass track from a cap, asked whether the Orion bypass stays exempt and whether he approves the words, he answered: "yes to both". READ AS (Orion, not the ruling): the track that carries the operators' own Orion sessions stays exempt and has no cap; 'project specific estate caps' is a project's estate ceiling, which is indicative and refuses nothing (cookbook 4241).

All design is role-based, never person-based; rulings never differ per person; only a computer's own constraints may

Ruled 2026-10-21 by ninad -- in force

ALL DESIGN IS ROLE-BASED, NEVER PERSON-BASED; RULINGS NEVER DIFFER PER PERSON; ONLY A COMPUTER'S OWN CONSTRAINTS MAY (Ninad, CLI-892, 2026-10-21). Said on the ruling that every operator's computer sends its session's record every 15 minutes: "also, this should be for all operators, not just Sahil.. all design should be role based, not person based.. individual laptop constraints can be person specific, but rulings should not be different per person..". READ AS (Orion, not the ruling): a ruling, design or rule names a role or rung (administrator, estate operator, service operator, Orion instance), never a person; where one person's computer needs something different (Windows PowerShell 5.1, no security chip, a slow temporary folder), that difference is a property of the computer handled in its code path, never a different rule for that person. A ruling worded for one person is read as applying to their role.

Person types -- user, employee, operator, investor and any other -- are RDM values

Ruled 2026-11-03 by ninad -- in force

PERSON TYPES -- USER, EMPLOYEE, OPERATOR, INVESTOR AND ANY OTHER -- ARE RDM VALUES (Ninad, CLI-023, 2026-11-03). THE CONTEXT: shown which of the user onboarding pursuit's initiatives overlap the user, project and operator onboarding arcs, he answered: HIS WORDS: "1. person types - user, employees, operators, investors etc should be RDM values 2. there should be an onboarding app for each person type, check which ones have and which ones dont, and show me the gaps". READ AS (Orion's reading, not his words): (1) the kinds of person Carolverse knows are one RDM set with a meaning per value, read live, never a list typed in code or a page; the values today are the kinds security #572 named -- user, employee, operator, investor -- an admin being an estate operator who holds admin (#572, security #461) and the operator rungs staying root's record (security #272), never copied into the set; (2) the set names the TYPE only: each type's records stay where their rulings put them (an investor a first-class registry entry, leadership #236, never a flag on a person or user row; operators on root's record; users on the registry's person records). His second point (an onboarding app per person type) is answered and recorded separately.

Pursuit and arc statuses are reference values, and every reference value is a dictionary entry

Ruled 2026-11-03 by ninad -- in force

Ninad, CLI-019 (2026-11-03), after Orion retired a pursuit as 'abandoned', the only closed pursuit status besides 'achieved'. HIS WORDS: "pursuit and arc statuses should be RDM values, and all RDM values should be dictionary entries". RULED: (1) a pursuit's status and an arc's status are values of their reference sets, pursuit.status and arc.status, and nowhere else -- as they already are. (2) Every active reference value is an entry in the dictionary: a term for that value of that set, whose meaning is the value's meaning in reference data, linked to its set and value. A value added to reference data gets its dictionary entry when it is added, and a value with no entry is a finding until it has one. On 3 Nov 53 of the 399 active values had one.

A rule may be kept at once as a policy, a ruling and a playbook recipe, so long as they never contradict

Ruled 2026-11-06 by ninad -- in force

Ninad, 2026-11-06 (CLI-040). Asked whether Leo's rule (blueprint #642: keep the user informed of the possibilities, execute only after the user's explicit confirmation in chat) should be a formal policy, or stay a ruling plus a recipe in Leo's playbook, HIS WORDS: "all three, policy, ruling and playbook item (duplication of policies is ok as long as they are not contradictory through mis-interpretations..)". READ AS (Orion's reading, not his words): the same rule may be held as a policy, a ruling and a recipe in the owning agent's playbook at once; the duplication is allowed only while every copy says the same thing, so a copy is never worded in a way that can be read differently from the others, and a change to one is carried to every copy. Blueprint #543 stays a recipe only: that was his choice for that rule, not a limit on this one.

Rhea's measure of whether the product works for its users and whether customers are satisfied is approved under the roadmap item Honest instruments

Ruled 2026-11-08 by ninad -- in force

HIS CHOICE (CLI-068, 2026-11-08), asked: "Rhea (Leadership) asked for a measure of whether the product works for its users and whether customers are satisfied. Albus stopped because it isn't approved on the roadmap. Approve it under the roadmap item 'Measurements can be trusted' so it can go ahead?" -- he chose "Approve it" (recorded as approved under 'Measurements can be trusted'; the work goes ahead and Rhea gives the acceptance verdict) over declining the wish. READ AS (Orion's reading, not his words): serves governance's service roadmap item 39 'Honest instruments', under strategic item 75 'Measurements can be trusted', with the operator's approval recorded here; it goes ahead on Albus's wish lane and Rhea gives the acceptance verdict (Rulings Repo #19).

Every term in the plan is in the data dictionary and every app's specifics in its handbook; a policy makes both a mandatory activity and a dedicated skill keeps them updated

Ruled 2026-11-14 by ninad -- in force

EVERY TERM IN THE PLAN IS IN THE DATA DICTIONARY AND EVERY APP'S SPECIFICS IN ITS HANDBOOK; A POLICY MAKES BOTH A MANDATORY ACTIVITY AND A DEDICATED SKILL KEEPS THEM UPDATED (Ninad, 2026-11-14, CLI-136).

HIS WORDS: "3. all terminologies in the plan should be included in the data dictionary (like \"churn\" or \"move up\" etc). App handbook can be used to describe specifics of the app, like the various tabs, charts etc.. dictionary and handbook documentation should be included as a mandatory activity via policy, and there should be a dedicated skill for updating these, so that it never gets missed..".

READ AS (Orion's reading, not the administrator's words): (1) every term the plan uses -- churn, moving up a size, time to win, potential client, cost to win, feasibility, assumption, round table and the rest -- is defined once in the data dictionary (none of these six is there today); (2) each app's handbook describes its specifics -- its tabs, charts and tables -- starting with the Fundraising app; (3) a POLICY, made through the policy process, makes updating the dictionary and the handbook a mandatory activity of every change that adds a term or changes an app; (4) a DEDICATED SKILL in Sage's library does the updating and is bound wherever such changes are made, so it is never missed. Folded into pursuit CAROL-PUR-0035 (Investment plan) for the plan's terms and the Fundraising handbook; the policy and the skill hold for every change.

The dictionary's Runbook entry names no count of headings

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-190), asked "Shall I reword the dictionary entry without a number?" after the operating manual gained its tenth heading, Operator doors (Rulings Repo initiatives #955), and Orion proposed taking the count out of the definition so it does not go stale when a heading is added: "agreed". READ AS (Orion's reading, not his words): the Runbook entry (other name: operating manual) reads 'the headings that settle how work is filed, run, judged and closed', and its other name 'the nine headings' becomes 'the runbook headings'; no definition carries a count of the manual's headings.

The registry is a service Carolverse offers, graded config-only: every project and every Sandbox prospect gets its own copy

Ruled 2026-11-17 by ninad -- superseded 2026-11-17

THE REGISTRY IS A SERVICE CAROLVERSE OFFERS, GRADED CONFIG-ONLY: EVERY PROJECT AND EVERY SANDBOX PROSPECT GETS ITS OWN COPY (Ninad, CLI-193, 2026-11-17). THE QUESTION: building (a Sandbox build meets its delivery checks, blueprint #961), Orion found the checks need a registry -- the record of which apps exist and who owns them -- and Sandbox and its prospects have none; he asked which registry a Sandbox build is judged against. HIS WORDS: 'can registry be a service that carolverse offer?'. Told yes -- it fits the rule for offering services to projects (one shared code, a separate copy per project, P.01.03.07.08) and no service offers it today -- he asked: 'if registry is a service, what grade would it be? as-is or config or template? I dont exactly remember the various types of services'. Shown the four reuse grades (as-is, config-only, seam-refactor, estate-bound), he chose the option 'Config-only: Each project and prospect gets its own registry copy from the service, set up by one config file; Sandbox builds are checked against their prospect's copy.' READ AS (Orion's reading, not his words): the registry is offered to projects as a service track whose reuse grade is config-only; each project, and each prospect in Sandbox, holds its own registry copy named by its own instance config; a Sandbox build's ownership check reads its prospect's copy; the copy moves with the prospect to its own machine when its contract is signed. Which service keeps the offering is not settled by this ruling.

Ended because: Ninad accepted one Carolverse registry with a store per project (CLI-198)

The registry is one Carolverse service: one running registry keeps a separate store for each project and each Sandbox prospect, set up by that project's one config file

Ruled 2026-11-17 by ninad -- in force

THE REGISTRY IS ONE CAROLVERSE SERVICE WITH A SEPARATE STORE FOR EACH PROJECT AND EACH SANDBOX PROSPECT (Ninad, CLI-198, 2026-11-17; supersedes governance #962). THE QUESTION: shown the status of CAROL-PUR-0084 (Lightweight projects, blueprint #967), Ninad asked: "this would mean that the burden would shift to carolverse.. how would carolverse scale to manage the load from the projects? should registry as a service should be also brought back into carolverse?". On the registry Orion proposed: "yes, I'd bring it back as one Carolverse service. That means one running registry, with a separate store for each project and each Sandbox prospect, set up by that project's one config file. It keeps today's intent (every project has its own registry) and meets the one-copy rule. The only exception is another dominion: being a fork, it keeps its own registry by design." -- and named the conflict with governance #962 (every project gets "its own registry copy") against P.01.03.07.24 (a separate copy only where one shared copy cannot serve the project, with the reason recorded), asking: "I read 'its own copy' as its own data, not its own running registry. Is that the right reading?". HIS WORDS: "I like your proposal, can you fold them in the light projects pursuit?". READ AS (Orion's reading, not the administrator's words): (1) Carolverse offers the registry as ONE running service (System Services' Project registry track), never a running copy per project or prospect -- P.01.03.07.24's single copy; (2) each project and each Sandbox prospect has its OWN STORE in that service, named by the project's one config file, and no project reads another's; (3) #962's 'its own registry copy' is read as its own store, and its config-only grade stands for how a store is set up; (4) another dominion, being a fork (cookbook 1716), keeps a registry of its own; (5) what and built as per-project and per-prospect copies, and the planned and, are re-read under this in CAROL-PUR-0084; which machine a store sits on is not settled by this ruling.

A policy is a short rule agents decide by; definitions stay in the dictionary and the process belongs in the skill

Ruled 2026-11-17 by ninad -- in force

POLICIES ARE SHORT (Ninad, CLI-195, 2026-11-17). Shown a draft policy that defined arcs and pursuits and described how an arc takes its topic, HIS WORDS: "I dont think all of that text can go in a policy, policies are generally short, and are used by agents to make decision.. things like "arc takes the topic from the initiatives that start it" is not a policy, its the description of the process for which the skill is defined.. is there a definition of a policy in the dictionary?" and then "these definitions should already exist in the dictionary.. pls check". READ AS (Orion's reading, not his words): a policy states the rule an agent decides by, in a few short sentences, as the dictionary's 'Policies' says (general principles, not recipes); it never restates a term the dictionary defines, and never describes how the work is done -- that is the skill's (or the cookbook's); the add-new-policy skill checks a draft for both before it is shown to the operator.

The data dictionary gives data its meaning; a data catalog and data traceability give its provenance

Ruled 2026-11-19 by ninad -- in force

Ninad: "data dictionary should be used to correctly interpret the meaning of the data, and data catalog and traceability should be implemented to ensure data provenance.. data quality lead should also own data catalog and data treability services, and they should also be made available to other projects. This deserves a seperate arc of its own"

Standards and protocols live in Sage's skill library, never in a policy; Themis checks they are followed; Hagrid owns moving Chowpatty's POC page to its short address

Ruled 2026-11-21 by ninad -- in force

STANDARDS AND PROTOCOLS LIVE IN SAGE'S SKILL LIBRARY, NEVER IN A POLICY; THEMIS CHECKS THAT THEY ARE FOLLOWED (Ninad, 2026-11-21, CLI-284).

HIS WORDS: "policies are not protocols, protocols and standards change with time, policies remain the same. policy should be - web addresses should follow standard protocols for the web links. where are standards and protocols defined? SST? or somewhere else?" Told that no single home exists today (protocols sit in Sage's skill library, in designs, and four policies are really standards), and shown Orion's recommendation "make Sage's skill library the one home for standards and protocols", he answered: "I agree your proposal".

On the owners of the web address work (security #1093) he said: "Heimdall is the head of security, not compliance.. so whether standards are followed or not should sit with themis"; asked whether moving Chowpatty's page to /chowpatty is Hagrid's or Forge's and told Hagrid owns it (his publisher, blueprint #1009) while Forge writes the code as for every build, he gave no objection; Jarvis opening the app address and naming customers at checkout, and Hagrid giving a project's first machine its address: "ok".

READ AS (Orion's reading, not his words): (1) a standard or protocol is a versioned skill in Sage's library, kept by Sage, bound to the agents that apply it; a policy only says that the standard is followed; (2) the web address protocol of security #1093 becomes such a skill, and design 731 cites it rather than restating it; (3) the check that work follows a standard is Themis's, compliance being her duty; (4) the four policies that are standards in all but name (P.01.03.05.05, P.01.03.06.01, P.03.01.01.06, P.03.02.01.16) move to the library later, each by its own ask, never silently.

The data dictionary and taxonomy are a track Sage owns beside his data catalog and quality tracks, offered as a service to other projects

Ruled 2026-11-21 by ninad -- in force

THE DATA DICTIONARY AND TAXONOMY ARE A TRACK SAGE OWNS, BESIDE HIS DATA CATALOG AND DATA QUALITY TRACKS, OFFERED AS A SERVICE TO OTHER PROJECTS TOO (Ninad, 2026-11-21, CLI-284).

ASKED: "which service provides data dictionary and data taxonomy as a reusable service track to users? should that be owned by Sage too?" Told that no service offers them today (the dictionary and the reference lists are kept as Carolverse's own records under Clara's governance work, Records & Operations), and shown: "The change would be a new 'Data dictionary and taxonomy' track owned by Sage beside his data catalog and quality tracks, reusable by other projects. Clara's governance work would stop keeping the dictionary, but the estate would still use it the same way. Do you approve?" HIS WORDS: "approved, and it will be offered as a service to other projects too!!"

READ AS (Orion's reading, not his words): (1) a new track in the service that holds Sage's data tracks (blueprint #1054: new data work is a track in an existing service), Sage its owner and running its P&L (cost-center #1053); (2) the track keeps the data dictionary and the taxonomy (the reference data sets and their values) for Carolverse and for any project that subscribes; (3) the dictionary's keeping leaves Governance's Records & Operations track and the Data Dictionary app's ownership moves from Clara to Sage; every reader keeps reading the one dictionary as today (one home, cookbook 1126); (4) the operator rule for changing the dictionary (P.01.01.05.09-.11: only the administrator and estate operators change it) is unchanged by who keeps it.

A track's owner is the service's owner or an agent in the service owner's team

Ruled 2026-11-21 by ninad -- in force

A TRACK'S OWNER IS THE SERVICE'S OWNER OR AN AGENT IN THE SERVICE OWNER'S TEAM (Ninad, 2026-11-21, CLI-284).

ASKED: after Orion confirmed that every track has one owner and that P.01.03.05.08 lets a track's owner differ from the service's owner, he answered: "but there is a catch, which is not probably listed as a policy.. a service may have tracks, and the owners of the track should be from the service owners team.. make this as a policy, and check for violations".

READ AS (Orion's reading, not his words): (1) an agent's team is the agent and every agent who reports to it, directly or through others, in the registry's reporting line; (2) a track whose owner is neither the service owner nor in that team is a governance gap, to be cured by moving the track or changing its owner, never by changing a reporting line to fit; (3) P.01.03.05.08 is rewritten in place to say so, once Ninad approves its words. Measured at the ruling: 7 of 84 live tracks break it -- Orion's tracks in Build Initiatives, WhatsApp (two) and Communications (Orion reports to Ninad, so he is in no agent's team), and Leo's, Carol's and Sam's chat tracks in Galadriel's WhatsApp service.

Policies, rulings, standards and protocols, and agent protocols are offered as one reusable rulebook in Clara's Governance service, one track each

Ruled 2026-11-21 by ninad -- in force

POLICIES, RULINGS, STANDARDS AND PROTOCOLS, AND AGENT PROTOCOLS ARE OFFERED TO OTHER PROJECTS AS ONE REUSABLE RULEBOOK, IN CLARA'S GOVERNANCE SERVICE, ONE TRACK EACH (Ninad, 2026-11-21, CLI-284).

ASKED: "there is already protocols app that agents use to operate, when there are no clear rulings.. these are protocols that the agents form themselves.. who owns that service? should standards and protocols be offered as a re-usable service too, clubbed with rulings and policies?" Told that Athena owns the Agent Interaction Protocols app and its Protocols track, that policies sit in Orion's Constitution service, that no track owns the Rulings Repo, and shown Orion's recommendation -- one reusable rulebook in Clara's Governance service, one track per kind, each owned by its keeper (policies moving out of Orion's Constitution service; rulings; standards and protocols, Sage; agent protocols, Athena, staying non-binding, an operator able to promote one into a standard; Themis checking compliance across all four) -- he answered: "approved". Asked next whether one agent should own all four or each its keeper, with the owners Clara (policies), Clara (rulings), Sage (standards and protocols), Athena (agent protocols), he answered with the rule of governance #1099 (a track's owner is from the service owner's team), which all four owners meet: Sage and Athena are in Clara's team.

READ AS (Orion's reading, not his words): four tracks in the Governance service -- Policies (Clara; moved from the Constitution service's Policy track), Rulings (Clara; the Rulings Repo's first owning track), Standards and protocols (Sage; the content stays in his skill library, governance #1097), Agent protocols (Athena; moved from Agent and Human Resources' Protocols track, still a convenience and never an obligation) -- each offered to other projects; Themis checks that work follows them.

When Ninad says he likes a proposal, that is his approval

Ruled 2026-11-21 by ninad -- in force

WHEN NINAD SAYS HE LIKES A PROPOSAL ("I like it", "I like your proposal"), THAT IS HIS APPROVAL OF EXACTLY WHAT HE WAS SHOWN (Ninad, 2026-11-21, CLI-280/CLI-284).

HIS WORDS: "I like your proposal, implement it.. (when I say I like it, that is the approval)". Told that the memory note saying so could not reach the startup list of his work sessions, he answered: "do what is necessary so that all work sessions also understand it the same way..".

READ AS (Orion's reading, not his words): (1) every Orion session of Ninad's records 'I like it' as his approval of the proposal, policy words, dictionary words or step he was shown -- recorded in his words, with the question he answered -- and acts on it without asking again; (2) it approves only what he saw: a change beyond it is asked separately; (3) it is his own way of approving and says nothing about how another operator approves.

Every project adopts whatever applies to it of Carolverse's rulebook and its data dictionary and taxonomy, and adds its own on top

Ruled 2026-11-21 by ninad -- in force

EVERY PROJECT CAN ADOPT CAROLVERSE'S RULEBOOK AND ITS DATA DICTIONARY AND TAXONOMY -- WHATEVER APPLIES TO IT -- AND ADD ITS OWN ON TOP (Ninad, 2026-11-22, CLI-284).

ASKED: approve two roadmap tasks, "Every project can adopt Carolverse's rulebook" on Clara's Governance roadmap and "Every project can use Carolverse's data dictionary and taxonomy" on the roadmap of the service holding Sage's data tracks (Build Initiatives). HIS WORDS: "'Every project can adopt Carolverse's rulebook' - whatever applies to them and they can add their own on top; 'Every project can use Carolverse's data dictionary and taxonomy' - whatever applies to them and they can add their own on top. The rulebook one on Clara's Governance roadmap, and the data dictionary one on Build Initiatives' roadmap, which holds Sage's data tracks - approved".

READ AS (Orion's reading, not his words): (1) a project takes from Carolverse's rulebook (policies, rulings, standards and protocols, agent protocols -- governance #1100) and from its data dictionary and taxonomy (governance #1098) only the entries that apply to it; (2) it adds its own entries on top, kept as that project's own and never written into Carolverse's; (3) where a project's own entry and an adopted one disagree, that is brought to the project's operators as a contradiction (P.01.01.05.08), never silently resolved; (4) this follows the project-build guidelines -- adopt what fits, extend what fits in part, build only for the project what does not (blueprint #242-#246).

Response levels are agreed in a reusable SLA service, and the work is mapped service roadmap to Carolverse roadmap

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-326) whether to add a new WhatsApp roadmap item 'People get a reply within half a minute, at a fraction of today's cost' for the fixes to Carol's replies. Ninad: "the activities should be mapped to the service roadmap, which inturn should be mapped to carolverse roadmap.. the exact SLA's for carol responses should be documented in a seperate service called SLAs (service level agreements) which should be a reusable service track like any other service in carolverse which serves carolverse, and provides the same reusable service to other projects to note down their own SLAs and comply to them.. who should own this track? Themis?". READ AS (Orion, not Ninad's words): each initiative names a service roadmap task that names a Carolverse roadmap item; the response levels Carol must meet (reply time, cost) are recorded in a new SLA service, offered to other projects as-is to record and meet their own; its owner was put back to Orion as a question and is not yet ruled.

Themis owns the SLA service

Ruled 2026-11-22 by ninad -- in force

Asked (CLI-326): "Who should own the new SLA service, where each project records the response levels it promises (Carol's reply time and cost first) and checks it meets them?" with the options Themis (Recommended: meeting an agreed level is a compliance question; Hermione's monitoring supplies the measurements, Themis judges breaches and raises them), Hermione, Galadriel. Ninad chose: "Themis (Recommended)".

๐Ÿ“–Infrastructure & Backups

The laptop copy is prepared on the machine by Hagrid and pulled read-only through a login of its own

Ruled 2026-09-25 by ninad -- superseded 2026-09-25

THE LAPTOP BACKUP PULLS A PREPARED FOLDER, NOTHING MORE (Ninad, 2026-09-25, CLI-550). The laptop copy (track backup-laptop, Hagrid accountable) no longer logs in as the operator login to run commands on the machine: a droid of Hagrid's, under his own login and on his own schedule, prepares the day's copy on the machine into one folder (the estate's databases taken with SQLite's own backup, and the dumps and archives the laptop keeps today); the laptop pulls THAT folder, read-only, through a locked login of its own whose only permitted command is that read-only pull, with a touch-free key made in the laptop's security hardware (the unattended-job exception of security #138). No stored password travels with the job. The laptop copy and the off-site copy stay INDEPENDENT -- neither is made from the other -- so a bad copy cannot spread from one to the other. Chosen over taking the laptop copy from the off-site GitHub backup.

Ended because: Ninad chose, in the same sitting, that the pull login may also hand in the job's finished-note

The laptop copy is prepared on the machine by Hagrid and pulled read-only through a login of its own

Ruled 2026-09-25 by ninad -- in force

THE LAPTOP BACKUP PULLS A PREPARED FOLDER, NOTHING MORE (Ninad, 2026-09-25, CLI-550). The laptop copy (track backup-laptop, Hagrid accountable) no longer logs in as the operator login to run commands on the machine: a droid of Hagrid's, under his own login and on his own schedule, prepares the day's copy on the machine into one folder (the estate's databases taken with SQLite's own backup, and the dumps and archives the laptop keeps today); the laptop pulls THAT folder, read-only, through a locked login of its own whose one command is a door allowing exactly two things -- that read-only pull, and handing in the job's finished-note (the heartbeat Tyr's Backup Auditor judges the laptop copy by: when it finished, how many errors, which snapshot), checked against its fixed shape and written to one file -- and refusing everything else, with a touch-free key made in the laptop's security hardware (the unattended-job exception of security #138). No stored password travels with the job. The laptop copy and the off-site copy stay INDEPENDENT -- neither is made from the other -- so a bad copy cannot spread from one to the other. Chosen over taking the laptop copy from the off-site GitHub backup.

Hagrid may delete as part of his infrastructure clean-up work

Ruled 2026-09-28 by ninad -- in force

HAGRID IS ALLOWED TO DELETE AS PART OF HIS INFRASTRUCTURE CLEAN-UP JOBS (Ninad, CLI-602, 2026-09-28). Ninad answered: 'hagrid is allowed to delete as part of his infrastructure cleanup jobs, which is clearly not happening'. So: removing for good what a stated retention rule retires is Hagrid's work, done by his own registered clean-up processes under his own login, each removal recorded. 'Delete nothing' still binds every other actor and every removal outside a stated rule: an Orion session, a build lane or another agent still moves a thing to Trash, and Trash is one of the areas Hagrid's rules keep. What a rule may retire is stated in the rule before anything is removed; the newest copy of anything, a live store, media and the backups themselves are never retired by a clean-up rule.

The operating system expires old temp and Trash entries, whichever login owns them

Ruled 2026-09-28 by ninad -- superseded 2026-09-28

THE OPERATING SYSTEM EXPIRES THEM (Ninad, CLI-604, 2026-09-28). Hagrid's clean-up runs under his own login and cannot remove an old entry another login owns: the temp area is sticky, and a Trash folder of the apps login is closed to him. Three Trash items held by the apps login kept his keeper's run failed, and his Daily Janitor had counted thousands of temp entries as removed that were all still there. Shown three ways to remove such entries -- the operating system expires them by a rule only root can change; a narrow removal door for Hagrid; a one-off clean-up with each owner cleaning its own afterwards -- Ninad chose the first: '1. The operating system expires them'. So: a root-owned rule of the operating system's own temp-file service ages entries out of the temp area after three days and out of the two Trash folders of the work tree after fourteen days, whichever login owns them. No login of the estate gains a power to delete another login's files, and the rule is changed only at the root login. Hagrid's keeper and janitor go on removing what their own login can and reporting truthfully; what the operating system expired is not on Hagrid's receipts, and that was stated when the choice was put. The ages are Orion's reading of rulings already made: fourteen days for Trash is design 681 v1.2; three days for the temp area is the figure recommended at CLI-455 on.

Ended because: Trash leaves the operating system's cleaner: it reads each file's own age, not when the thing was trashed, so Trash would stop being a way back; a Trash item Hagrid cannot remove is handed to him instead (Ninad, CLI-604, 2026-09-28)

The operating system expires old temp entries; a Trash item Hagrid cannot remove is handed to him

Ruled 2026-09-28 by ninad -- in force

TRASH STAYS OUT OF THE OPERATING SYSTEM'S CLEANER (Ninad, CLI-604, 2026-09-28, amending his ruling of the same day). The operating system's temp-file service expires entries of the TEMP AREA after three days, whichever login owns them, reading an entry's age from when it was made or changed and never from when it was last read; that rule is changed only at the root login (installed 2026-09-28 21:20 UTC: 6,061 entries removed). The two Trash folders of the work tree are NOT put under that cleaner: it reads each file's own age, not when the thing was trashed, so a folder trashed today holding month-old files would lose them the same night and Trash would stop being a way back. Trash stays Hagrid's, held to fourteen days from when a thing was moved in (design 681 v1.2, Rulings Repo backup #200). A Trash item his login cannot remove, because it holds files of another login, is HANDED TO HIM: an operator gives it to Hagrid's login through a narrow operator door that accepts only an item lying in one of the two Trash folders, follows no link and removes nothing itself, and Hagrid's keeper removes it on its next run and says so on its receipt. Until the door is built such an item is given to him in a root step, as three were on 2026-09-28. Shown the recommendation with its reason, Ninad answered: 'I agree with your recommendations'.

The machine-side backup clone keeps the newest seven nights; the off-site copy keeps a rolling thirty nights

Ruled 2026-10-05 by ninad -- in force

HOW MUCH BACKUP HISTORY EACH SIDE KEEPS (Ninad, CLI-719, 2026-10-05). The machine's disk fills in about 19 days, and GitHub starts refusing repositories this size within weeks. How much backup history should each side keep?" HIS ANSWER: he chose "Machine 7, off-site 30 (Recommended)", offered as "The machine keeps the last 7 nights; GitHub keeps a rolling 30 nights. Older nights are removed from GitHub's history, so a restore can only go back 30 days." (over "Machine 7, off-site all" and "Machine 1, off-site 90"). READ AS (Orion's reading, not Ninad's words): (1) the backup clone on the machine keeps the newest seven nights and no older; (2) the off-site copy keeps a rolling thirty nights, and a night older than thirty is removed from its history; (3) both removals are Hagrid's work, done by his own registered backup process under his own login with each removal recorded (backup #200), never by hand and never by a retention rule over a backup (cookbook 1801); (4) the laptop copy (#141) is not touched by this ruling.

Hagrid owns the backup-freshness measure

Ruled 2026-10-06 by ninad -- in force

HAGRID OWNS THE BACKUP-FRESHNESS MEASURE (Ninad, CLI-732, 2026-10-06). Who should own that measure?" (offered: "Hagrid" -- he owns the backup service; amend the policy so his collector is allowed, and his wish goes ahead; "Themis" -- the policy stands; re-scope the wish so Themis's audit records the dated freshness). HIS ANSWER: "Hagrid". READ AS (Orion's reading, not Ninad's words): the dated per-copy backup-freshness measure is Hagrid's, collected by his own droid on his own login; P.03.03.01.68 (and the matching sentence of P.03.03.01.05) is amended to say so, Themis keeping her audit of it; returns to Albus's bypass lane with Hagrid as its requester and owner.

Orion enables Hagrid to make project machines, and Hagrid works from written infrastructure guidelines (an Infra Cookbook)

Ruled 2026-10-21 by ninad -- in force

ORION ENABLES HAGRID TO MAKE PROJECT MACHINES, AND HAGRID WORKS FROM WRITTEN INFRASTRUCTURE GUIDELINES -- AN INFRA COOKBOOK (Ninad, CLI-897, 2026-10-21). THE QUESTION ANSWERED: Sahil's operator communication #111 (CLI-891) asked how Hagrid's login gets the right to create Chowpatty's machine in Azure (A: the server's own identity; B: a dedicated identity for Hagrid, narrow and recorded, his recommendation; C: no standing right, a person creates each machine by hand), and who does the one-time Azure setup. Orion searched the policies, the cookbook, the rulings and the designs: they say every project runs on its own machine and none is shared, that Leo sizes it and Hagrid creates it after the administrator approves size and cost (design 587), and that Marco only gets an estate operator's approval for the spend (cost-center #468, #469) -- nothing says HOW a machine is created; the estate's own machine was made by hand by Orion. HIS WORDS: "the current azure vm was created by orion. orion's job is to create carolverse, so that carolverse agents can onboard new projects autonomously. Do the current policies or strategy cookbook tell how new machines are created?" -- and then: "I also think orion should enable hagrid to do this job. You teach an agent once, and he does it for each project consistently.. if hagrid is doing it, he needs guidlines for doing it, maybe an app called infra cookbook, that talks all about how infra is managed.. thats my take". READ AS (Orion's reading, not his words): (1) creating a project's machine is Hagrid's job, done the same way for every project, and ORION's job is to ENABLE him once -- the Azure right his login needs (Sahil's option B: a dedicated identity for Hagrid alone, allowed only to create machines in one fenced area and only approved sizes, its secret held by the sign-in keeper and handed for one call at a time, every use recorded) and the process he runs; never a person making machines by hand per project (C) and never the whole server holding the right (A); (2) Hagrid works from WRITTEN GUIDELINES on how infrastructure is managed -- machines, sizes, identities, backups, disks, cost -- kept as craft knowledge in Sage's library bound to Hagrid (skills live only there; agents hold them by binding), so what he reads at every run is the one record; (3) an app called the INFRA COOKBOOK shows those guidelines -- a transparency surface that reads the record and performs no action (P.01.03.01.05) -- the way the Build Cookbook shows the build pipeline's; (4) the one-time Azure setup touches the administrator's own Azure account, so it is Orion's to prepare and run from the administrator's laptop (where the Azure command line is signed in as him) with his go-ahead, not a step Sahil needs owner rights for; after it, no project's machine needs the administrator.

Operators work on a project machine with the same tools and speed as on carol-vm

Ruled 2026-11-15 by ninad -- in force

HIS WORDS (Ninad, CLI-169, answering): "rulings - yes" -- the second of the two: "operators work there with the same tools and speed as on carol-vm".

Asked: Ninad first asked "Would they ensure that the sandbox is secure, easy to access, and every action is auditable? Orion then asked: "shall I record 'every action on a project machine is auditable' and 'operators work there with the same tools and speed as on carol-vm' as your rulings, so I can file the work against them?"

READ AS (Orion, apart from the words): an operator working on a project machine through their Orion session has the same tools at hand and the same responsiveness as on carol-vm -- repeated commands do not each pay a new sign-in, and the machine is sized for the work measured on it; which estate tools belong on a project machine is set by a design an estate operator approves, within 'a project never builds for itself' (blueprint #50).

๐Ÿ“–Leadership

Fundraising holds the plan, never the actuals

Ruled 2026-09-22 by ninad -- in force

The Fundraising app showcases the PLAN: what the venture intends to raise, at what shape, what each cycle buys and the runway it funds. What has actually been committed is a different question and this is not its home. The stored `raised_eur` column is RETIRED IN PLACE -- kept so nothing recorded is destroyed, refused BY NAME at the door, and read by no surface. Hiding it on the page alone would leave Clara's registrar able to record an actual into an app that promises not to hold one.

The Fundraising app opens on the estate's own project, derived

Ruled 2026-09-22 by ninad -- in force

The project shown by default is DERIVED from the dominion record's base project, never the word 'carolverse' written into the app: naming it leaves every forked dominion opening on somebody else's project. A viewer who cannot reach the base project is given NO default rather than another project, and an explicit choice always wins.

The Fundraising app presents three tabs, and the share split is derived

Ruled 2026-09-22 by ninad -- superseded 2026-11-04

Three main tabs: the five-year plan (valuation, ARR and cost at each year end), the fundraising cycles (amount to be raised, with the founders' and investors' share at the end of each round), and the spend distribution for the raised capital. THE SHARES ARE DERIVED from each cycle's pre-money and raise and are never stored beside them -- a stored percentage is the same fact twice and drifts the moment either figure is revised. A cycle missing its valuation or its amount makes ITS split and EVERY LATER cycle's UNMEASURED: the chain cannot resume across a gap, and no percentage is ever guessed or shown as zero.

Ended because: Ninad replaced the three tabs with four, each a ten-year chart (CLI-029)

The fundraising figures are the administrator's decision and are never seeded

Ruled 2026-09-22 by ninad -- in force

The rounds, the five-year plan and the spend distribution ship EMPTY and stay empty until Ninad decides them. Seeding a plausible curve, cycle or spend split would be indistinguishable from a decision he had made, which is the one thing these records exist to state. Regression tests lock all three tables at zero rows. Those are inputs to a decision, never the decision.

Where the ownership split is shown in Fundraising

Ruled 2026-09-23 by ninad -- in force

The split between the founders, the founding investors and the later investors is shown as a chart in a card ON THE PAGE -- its own always-rendered section, never inside a tab panel a reader has to open first. The Fundraising cycles table carries the MONEY of each cycle (what it raises and what the company is worth once it closes) and no longer carries the three percentage columns, because the card below plots exactly those rows. The five-year plan table keeps its two share columns: it places the split on YEARS while the card places it on ROUNDS, so those rows are not shown below. Because the figures left the table, the card carries its own small table of them -- a percentage reachable only by hovering is gated, and the tooltip is never the only home of a value.

How the ownership curve is drawn

Ruled 2026-09-23 by ninad -- in force

The lines are drawn visibly CURVED, not as straight segments between the cycles -- 'make the chart lines smooth and curved rather than straight'. The curvature is judged by eye, so an interpolation that is a cubic by construction but hugs the straight line does not satisfy this: monotone cubic was tried and drew a polyline, because it takes its slope at each point from the average of the neighbouring secants. At the same time the curve may NEVER overshoot: no point of it may leave the two values it joins, and none may read above a hundred or below nought per cent, because a bulge between two cycles states a share nobody holds. Both hold together by putting each segment's two control points at its horizontal midpoint, one at each end's height -- the curve then lies inside the convex hull of points that are all within the band. Replacing this with a Catmull-Rom or tensioned spline to get a livelier line reintroduces the overshoot.

Investor onboarding is shown by Clara's own lean Investor Onboarding app

Ruled 2026-09-30 by ninad -- in force

Investor onboarding is shown by Clara's own Investor Onboarding app: a lean app that only displays the investor onboarding pass, owned by Clara on the Finance track, under its own access policies. Asked on 2026-09-30 (CLI-623) whether Clara's display of the pass should be a tab in her Fundraising app, whose three tabs are ruled (leadership #85), or a lean investor onboarding app of her own, the administrator answered: 'investor onboarding app of her own'. This applies agent-resources #220 (each owner's onboarding is shown by that owner's own app) to the investor kind: the pass lives once in the canonical library, the workflow is Clara's, and the app performs nothing (cookbook 1710). Adding the investor pass to the Fundraising app, or to Athena's, Leo's or Jarvis's onboarding app, undoes this ruling.

Investors get a monthly update and a weekly showcase; the investor pass stays a draft

Ruled 2026-09-30 by ninad -- superseded 2026-11-08

Investors hear from the company on two rhythms, both written by Clara from her own address: a MONTHLY investor update, and a WEEKLY communication showcasing interesting updates about the company. Asked on 2026-09-30 (CLI-627) the four decisions the draft investor onboarding pass needs, the administrator answered: 'keep the current process as draft.. monthly updates to investors is ok, i think, but there should be weekly communications as well, showcasing interesting updates about the company'. So the cadence decision is settled as monthly plus weekly, and the pass is NOT approved: it stays a draft, its proposed Finance-track tasks stay undeclared (cookbook 1199), and nothing in it runs until the administrator approves it. The weekly communication is a showcase, not a second monthly update. Dropping the weekly communication, folding it into the monthly update, sending either from the administrator's or Carol's address (cookbook 1809), or declaring the tasks so the pass can start before it is approved, each undoes this ruling.

Ended because: the investor pass is approved so investors can talk to Clara (CLI-074)

An investor is its own kind of person, gets in only through the investor onboarding pass, and sees the Data Room read-only and nothing else

Ruled 2026-09-30 by ninad -- superseded 2026-11-08

Asked on 2026-09-30 (CLI-627) the decisions the draft investor onboarding pass still owed, the administrator answered: '1. investor is its own kind of person, separate from users and operators 2. readonly access to dataroom is correct 3. they get in as per the onboarding process 4. all investor records should be first class registry entries'. AN INVESTOR IS ITS OWN KIND OF PERSON: neither a user (served by the estate, Jarvis's) nor an operator (working for the estate, Athena's). An investor gets in ONLY through the investor onboarding pass -- admitted by the administrator, then granted access by the pass's own steps -- and what an investor is granted is named, READ-ONLY access to the Data Room and nothing else in the estate. Recording an investor as a user or an operator so an existing door lets them in, granting any write, or granting anything beyond the Data Room, each undoes this ruling.

Ended because: investors talk to Clara directly (CLI-074)

Every investor record is a first-class registry entry

Ruled 2026-09-30 by ninad -- in force

Asked on 2026-09-30 (CLI-627) the decisions the draft investor onboarding pass still owed, the administrator answered: '1. investor is its own kind of person, separate from users and operators 2. readonly access to dataroom is correct 3. they get in as per the onboarding process 4. all investor records should be first class registry entries'. EVERY INVESTOR RECORD IS A FIRST-CLASS REGISTRY ENTRY: the investor itself and every record the investor pass keeps about them (admission, Data Room access, questions and answers, commitments, updates sent) live in the registry as entries of their own, written through one door and journalled like every registry fact (cookbook 1504). The CRM's investments pipeline, the Data Room and every app REFERENCE those entries and never hold a copy (cookbook 1126). Keeping an investor only as a CRM lead or contact, as a note or flag on a person or user row, in an app's own store or in a side file, each undoes this ruling.

An investor's status follows the onboarding pass: lead, admitted, in diligence, committed, invested, declined, withdrawn

Ruled 2026-09-30 by ninad -- in force

An investor's status is one of seven governed values, following the investor onboarding pass: LEAD (introduced and recorded, phase A), ADMITTED (the administrator admitted them; the Data Room is open to them read-only, phase B), IN DILIGENCE (reading the Data Room and asking questions, phase C), COMMITTED (terms agreed and documents signed, phase D), INVESTED (the money arrived; they receive the monthly update and the weekly showcase, phase E), DECLINED (either side said no; the record is kept) and WITHDRAWN (they left after investing; the record is kept). Proposed by Orion on 2026-09-30 (CLI-627) for and answered by the administrator: 'sounds good to me'. The values live in reference data (cookbook 4236) and are enforced at the registry's write seam; an investor's record is never deleted, so declined and withdrawn are statuses, not removals. Typing the list into code, adding a status without a ruling, or deleting a declined or withdrawn investor each undoes this ruling.

Investors in diligence receive the weekly showcase too, and it carries only what is public

Ruled 2026-09-30 by ninad -- in force

Investors still in diligence receive the weekly showcase too. The weekly showcase (leadership #233) goes out from the diligence phase of the investor onboarding pass (phase C) onward, and carries on after they invest; because it reaches people who have not invested, it carries ONLY what is public -- never what the Data Room holds, never figures or plans that are not published. Asked on 2026-09-30 (CLI-627): 'should investors still in diligence get the weekly showcase too? If yes, it can only carry public news; if no, it starts once they've invested and can share what the Data Room holds' -- the administrator answered 'yes'. The monthly update is unchanged: it starts when they have invested. Holding the showcase back until investment, or putting Data Room material into it, each undoes this ruling.

The investment chain may be spread across services, as long as it holds together

Ruled 2026-09-30 by ninad -- in force

The investment chain -- valuation, investment plan, spend plan, roadmap, balance sheet and profit and loss, shareholding and press communications -- may be spread across services, each part with the service whose duty covers it, as long as it all holds together: one pursuit carries the chain end to end and one registered check proves every link. Finance stays a track exclusive to Clara (cookbook 1724 stands). Ninad: If its distributed across services, that is also fine with me.. as long as it all holds together..

Denken Labs is registered in the United Kingdom

Ruled 2026-10-01 by ninad -- in force

Denken Labs is registered as a company in the United Kingdom. Its registration options there (legal form, the part of the United Kingdom it is registered in, registered office, directors, fees, duties) are laid out by Clara for the administrator's decision under, and the company's contracts, certificates and templates are written for the law that registration names. Until registration completes, a document names the company as in formation and says who signs for it.

Until Denken Labs is registered, Ninad Sharad Joshi alone signs for it

Ruled 2026-10-08 by ninad -- in force

Ninad, 2026-10-08 (CLI-737). THE QUESTION ANSWERED (leadership #250 says a document names the company as in formation and says who signs for it, and nothing on record said who): Orion asked: "Until Denken Labs is registered, who signs documents for it? (Under UK law, whoever signs for a company that doesn't exist yet is personally bound by what they sign.)" Offered: "Me alone (Recommended)" -- Ninad Sharad Joshi signs for Denken Labs while it is in formation, every contract and certificate names him; "Me and a co-founder"; "Nobody until registered". HIS ANSWER: "Me alone (Recommended)". READ AS (Orion's reading, marked apart from his words, P.01.01.05.05): the company's one registration record names Ninad Sharad Joshi as the only person who signs for Denken Labs while it is in formation, and every contract, certificate and template reads that name from the record; he signs as the person acting for a company not yet formed, not as a director. Once registration completes the record names who signs from the registration itself.

Fundraising cycles chart: timeline, running total raised and post-funding valuation

Ruled 2026-11-04 by ninad -- superseded 2026-11-04

Ninad, 2026-11-04 (CLI-029): "fundraising app needs to show the graph for the fundraising cycles but its not showing it.. x axis should be timeline, y axis should be amount, and it should be a smooth line chart showing both amount raised and post funding valuation of the company"; asked whether the amount-raised line shows each cycle's own amount or the running total, he chose "Running total"; and "the chart should show up when i click on fundraising cycles tab". READ AS: the chart sits in the Fundraising cycles tab and is drawn when that tab is opened; each cycle is a point at its planned close month; one line is the running total of the amounts the cycles raise up to that close, the other the post-funding valuation at that close (pre-money plus the raise); both read the plan, as leadership #83 keeps Fundraising to the plan.

Ended because: The tab it placed the chart in no longer exists (CLI-029)

Fundraising has four tabs -- Investments, Valuation, Spend, Customers -- each a ten-year chart of yearly projections

Ruled 2026-11-04 by ninad -- superseded 2026-11-04

Ninad, 2026-11-04 (CLI-029): "I want changes to what has been built. The following tabs not make sense to me: Five-year plan, Fundraising cycles, Spend distribution, Regulation & compliance. Instead, I want the tabs to be as follows: Investments, Valuation, Spend, Customers. Each tab should have a chart, not a table. Each chart should have x axis as time, with 10 years duration, with yearly projections on each metrics.." Asked four questions he answered: years 6 to 10 (no figures exist past 2030) -- "I draft, you approve"; customers (no customer numbers are recorded) -- "I draft, you approve"; which lines each chart carries -- "Two lines where useful" (Investments: raised that year and the running total; Valuation: valuation and yearly recurring revenue; Spend: total spend split into people, tokens and machines, and other; Customers: customers at year end); the ownership, holdings-worth, yearly-spend and profit charts and the regulation and compliance costs -- "Keep them below". READ AS: the plan's horizon is ten years, 2026 to 2035, one point at each 31 December, superseding 'the next five years' (CLI-514); #86 stands -- a figure for 2031 to 2035 and every customer number is recorded only after Ninad approves Orion's draft, and until then those years read not decided; derived figures stay derived (valuation from a closing round or the comparables, #85's derived shares, #612's running total moving to the Investments chart by year); the tables the old tabs held move below the tabs with the other charts, so nothing recorded leaves the page.

Ended because: Ninad, CLI-034: Profit becomes a fifth tab and the profit chart moves into it (no longer kept below)

Fundraising cycles chart: superseded by the Investments tab

Ruled 2026-11-04 by ninad -- in force

Retired in place by leadership #616 (Ninad, CLI-029): the Fundraising cycles tab is gone; the running total raised moves to the Investments tab's chart, by year. Ninad, 2026-11-04 (CLI-029): "I want changes to what has been built. The following tabs not make sense to me: Five-year plan, Fundraising cycles, Spend distribution, Regulation & compliance. Instead, I want the tabs to be as follows: Investments, Valuation, Spend, Customers. Each tab should have a chart, not a table. Each chart should have x axis as time, with 10 years duration, with yearly projections on each metrics.." Asked four questions he answered: years 6 to 10 (no figures exist past 2030) -- "I draft, you approve"; customers (no customer numbers are recorded) -- "I draft, you approve"; which lines each chart carries -- "Two lines where useful" (Investments: raised that year and the running total; Valuation: valuation and yearly recurring revenue; Spend: total spend split into people, tokens and machines, and other; Customers: customers at year end); the ownership, holdings-worth, yearly-spend and profit charts and the regulation and compliance costs -- "Keep them below".

Fundraising's ten-year plan: Ninad approves Orion's first draft for 2031-2035 and the customer numbers

Ruled 2026-11-04 by ninad -- superseded 2026-11-04

Ninad, 2026-11-04 (CLI-029), shown Orion's draft under #616: "take my approval for the first draft". Basis stated with it: revenue growth slowing from about 1.9x in 2030 to about 1.2x by 2035; cost about 45% of revenue. 2026-2030 revenue and cost unchanged. READ AS: these are now his decided figures (#86); valuation for a year with no closing cycle still reads not worked out until comparable companies are recorded -- he was asked whether Orion should research them and has not answered.

Ended because: Ninad planned for an IPO by 2035 and approved the worked-backwards draft for 2032-2035

Fundraising's Customers tab includes the average size of the customer

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-029), shown the four tabs (#616): "also include average size of the customer". It joins the Customers tab's one chart as a second panel sharing the ten-year timeline, because a count of customers and an amount in euros never share one axis.

Fundraising has five tabs -- Investments, Valuation, Spend, Customers, Profit -- each a ten-year chart of yearly projections

Ruled 2026-11-04 by ninad -- superseded 2026-11-05

Ninad, 2026-11-04 (CLI-029): "I want changes to what has been built. The following tabs not make sense to me: Five-year plan, Fundraising cycles, Spend distribution, Regulation & compliance. Instead, I want the tabs to be as follows: Investments, Valuation, Spend, Customers. Each tab should have a chart, not a table. Each chart should have x axis as time, with 10 years duration, with yearly projections on each metrics.." Asked four questions he answered: years 6 to 10 (no figures exist past 2030) -- "I draft, you approve"; customers (no customer numbers are recorded) -- "I draft, you approve"; which lines each chart carries -- "Two lines where useful" (Investments: raised that year and the running total; Valuation: valuation and yearly recurring revenue; Spend: total spend split into people, tokens and machines, and other; Customers: customers at year end); the ownership, holdings-worth, yearly-spend and profit charts and the regulation and compliance costs -- "Keep them below". READ AS: the plan's horizon is ten years, 2026 to 2035, one point at each 31 December, superseding 'the next five years' (CLI-514); #86 stands -- a figure for 2031 to 2035 and every customer number is recorded only after Ninad approves Orion's draft, and until then those years read not decided; derived figures stay derived (valuation from a closing round or the comparables, #85's derived shares, #612's running total moving to the Investments chart by year); the tables the old tabs held move below the tabs with the other charts, so nothing recorded leaves the page. AMENDED 2026-11-04 (CLI-034): asked "Should Profit become a fifth tab?" -- with Orion's recommendation to move the profit chart into it -- Ninad chose "Move it into a tab". READ AS: Fundraising has FIVE tabs, Investments, Valuation, Spend, Customers, Profit, each a ten-year chart of yearly projections (2026-2035, a point at each 31 December); the Profit tab draws the year's profit before and after tax, the figures the door already works out (never typed, never stored); the profit chart that sat below the tabs MOVES into it and is no longer drawn below, so it is never shown twice; the ownership, holdings-worth and yearly-spend charts, the regulation and compliance costs and the old tabs' tables stay below as before.

Ended because: Ninad renamed the tabs after his table's columns and added Revenue

Orion researches public comparable companies for Fundraising's valuation, and Ninad approves each before it is recorded

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-034), asked "Should Orion research public comparable companies, with sources, and bring them to you for approval before Clara records them?" (no comparable is recorded, so the valuation can only be worked out in the years a round closes -- 2026, 2028, 2030 -- and 2027, 2029 and 2031-2035 read not worked out): he chose "Yes, research them". READ AS: Orion researches listed or recently funded companies comparable to Carolverse, each with the metric its multiple is taken on (annual recurring revenue), the multiple, the source and the date observed, and shows them to Ninad; only those he approves are recorded, through Clara's registrar (her Finance track, 1724), naming him as the one who decided (#86). Nothing is recorded on Orion's say-so, no comparable is invented, and a typed valuation stays refused: the years are worked out from the approved comparables by the door. This answers the question #618 records as unanswered.

Fundraising's ten-year plan: Ninad approves Orion's draft for a stock-market listing (IPO) in 2035, worked backwards -- revenue, cost and customers for 2032-2035, a Series C in 2032 and an IPO in 2035

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-034): "there are no investments happening after 2030, can you construct recommendations on investments for the decade?", then "ok, lets plan for IPO by 2035, and work backwards, to update the charts for the decade"; shown Orion's draft, "ok, update the charts accordingly..". Orion said the Series C is optional (the plan makes a profit every year, so the expansion could be paid from its own money, more slowly; without it founders keep 46% after the IPO) and Ninad approved the draft with it. The figures are recorded through Clara's registrar (her Finance track) naming Ninad as the one who decided.

Fundraising's Investments chart draws the money raised in each year as bars, one per cycle year, with the running total kept as a line

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-034), looking at the Investments tab after the IPO plan (#622): "why is the purple line going up and down? what does raised in the year mean?". Asked "How should it be drawn?" with Orion's recommendation, he chose "Bars": one bar in each cycle year, named with the cycle, while the running total raised stays as the line, so the chart shows both the single payments and the build-up. READ AS: amends #620's Investments chart only ('raised that year and the running total' stays the pair it carries; the year's amount changes from a line to bars); a year with no cycle has no bar and is a measured nought, never a gap; every other tab and every line elsewhere is unchanged; the figures stay the door's (raised_by_year), never typed.

Fundraising's Valuation chart joins the round values into a line as the current plan's baseline, until approved comparables give every year its own value

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-04 (CLI-034): "valuation line is not visible, only dots are visible"; told the dots are the five years a cycle closes and the years between read not worked out until comparables are recorded, he said "Once you research comparable companies and I approve them, update the line chart, but create a baseline based on the current plan..". READ AS: a deliberate exception to drawing nothing across a year with no figure, for this one line: the joined line is named on the chart as joining the round values, the years between carry no point and no tooltip figure of their own, and once Ninad approves comparables (#621) the line is drawn through every year's worked-out value instead.

The Investment chain pursuit carries each year's customers as plan against actual, and is started

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-05 (CLI-034): asked whether the records can hold both plan and actual ("the actual investment might not be the same as what is planned, do we have the right data model to support both? is that also part of the pursuit?"), Orion showed that rounds and investor commitments already hold actuals, that revenue, cost, cash and holdings are covered by the pursuit's criteria 3 and 4, and that the actual number of customers is covered by nothing. He said "update the pursuit, and start the work". READ AS: criterion 3 of CAROL-PUR-0007 now names customers beside revenue, cost and cash, the actual count read from the client records (Blueprint's clients, a client made by a signed contract), never typed; an initiative for it is filed and placed as a step; the pursuit moves from planned to active and its steps are worked in order, the next open MVP step being. Fundraising stays plan-only (#83): actuals are shown where the pursuit's steps put them.

Fundraising's plan and its investments match both ways: costs are genuine items, and the rounds ask for what those costs need; the founders fund 2026; planning salaries for the operators

Ruled 2026-11-04 by ninad -- in force

So if there are genuine costs especially related to marketing, sales, compliance, tokens, capital investments, infra costs etc, then the investment plan should ask for more investments"; (2) "yes, in 2026, founders will invest.. the investment includes token cost, infra expenses, other expenses, salaries of operators etc.", giving each operator's yearly salary for planning, and "The contract values are different, but thats ok, because the salaries will be used for planning investments, while contracts will be used for actual payments". The figures remain his to approve (#86) and supersede #622's 2026-2035 costs and round sizes once approved.

Fundraising's raises are whole millions, each the round's need rounded up

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-05 (CLI-034), while the genuine build was being recorded (#631): "amounts to be raised should be also round figures (in millions)". READ AS: each round raises its need rounded up to a whole million of euros, a surplus carrying into the next round's need; the founders' own 2026 investment is what it costs (it is not raised).

Fundraising's ten-year customers and revenue are drafts, worked backwards from the raises and the values after the money; no plan is approved yet

Ruled 2026-11-04 by ninad -- in force

Ninad, 2026-11-05 (CLI-034), told that a column compared his round targets with the customers 'the approved plan already has': "no plan has been approved, so give me numbers ensuring the raise and valuation post raise stay as stated".

Fundraising has six tabs -- Raise, Value after, Cost, Total, Revenue, Profit -- each a ten-year chart

Ruled 2026-11-05 by ninad -- in force

Ninad, 2026-11-05 (CLI-034/CLI-037): "Investment --> Raise / Valuation --> Value after / Spend --> Cost / Customers --> Total / Revenue --> Revenue / Profits --> ? The investment app should have these 6 tabs.. also the count of customers should be rounded 2026 --> 0, 2027 --> 5, 2028 --> 10, 2029 --> 20, 2030 --> 40, 2031 --> 80, 2032 --> 120, 2033 --> 140, 2034 --> 180, 2035 --> 220". READ AS: Fundraising (Clara's app, the leadership service's Finance track) has SIX tabs, each a ten-year chart (2026-2035, a point at each 31 December), named after the columns of the table he reads the plan in: Raise (was Investments; the year's raise as bars, #623, with the running total as a line), Value after (was Valuation; the value after the money, the round values joined, #624), Cost (was Spend), Total (was Customers; the customers held, with the small, medium and large split), Revenue (new; the year-end yearly value of the customers held), and Profit (the sixth he left open; Orion recommended keeping 'Profit', before and after tax, as #620 had it). The profit chart below the tabs moves into the Profit tab as #620 ruled. Every figure stays the door's, never typed.

The ten-year plan's revenue and customer totals are Ninad's round figures; the small, medium and large split is Orion's draft

Ruled 2026-11-05 by ninad -- in force

The investment app should have these 6 tabs.. also the count of customers should be rounded 2026 --> 0, 2027 --> 5, 2028 --> 10, 2029 --> 20, 2030 --> 40, 2031 --> 80, 2032 --> 120, 2033 --> 140, 2034 --> 180, 2035 --> 220". READ AS: the revenue and the customer totals for 2026-2035 are his decided figures, beside his raises (#633) and values after the money (#631). Costs are re-run from the genuine build (#629-#632) with each tier's delivery. Clara's registrar records the split and the costs only once he approves them (#86, #634).

The ten-year plan's split of customers by size is Orion's call, made so every planned figure adds up

Ruled 2026-11-05 by ninad -- in force

Recorded: 2027 2 small / 3 medium / 0 large, 2028 4/5/1, 2029 8/10/2, 2030 16/20/4, 2031 28/45/7, 2032 44/65/11, 2033 48/70/22, 2034 60/95/25, 2035 76/115/29.

Fundraising is a service every project adopts for its own plan, with the same KPIs, round figures and shares

Ruled 2026-11-05 by ninad -- superseded 2026-11-05

Ninad, 2026-11-05 (CLI-038): "how can this be made into a re-usable service for each project to create its own fundraising plan? this would depend on the nature of the projects, and not all projects are alike.. this is a more generic question, on how services in carolverse can be used as templates to create deliverables that are customised to the project. the key takeaways from this template are the KPIs (Round, raise, value, revenue, cost, customers, profit), the rounding (very important), so that the numbers are easy to read and remember, the visibility, like founders share, investor's share for each round etc.. include the ask to make this service re-usable across projects in the pursuit". READ AS: Fundraising (Clara's Finance track) is a template service: each registered project keeps its own plan inputs (how it earns, what it costs, its rounds and buffers, its tax basis) and nothing is copied from Carolverse (blueprint #242-#246, adopted requirement by requirement); the method is shared -- one builder, one check that every figure adds up before it is recorded, figures rounded so they are easy to read and remember -- and every project's plan shows the same KPIs (round, raise, value after, revenue, cost, customers, profit) with the founders' and investors' share after each round; the craft is a skill in Sage's library. Added to the Investment chain pursuit as a criterion and filed.

Ended because: Ninad agreed that a project runs the template with its own agent, never Clara by default

Fundraising is a template Clara owns; each project that adopts it runs its own plan with its own agent

Ruled 2026-11-05 by ninad -- in force

Ninad, 2026-11-05 (CLI-038), asked "Clara does not communicate with users, only Leo does, so how will Leo communicate with Clara, and how will clara build the service for new projects? like for Chowpatty? should clara build a fund raising plan for chowpatty? or should chowpatty have its own CEO to do it for Sachin?"; told Orion's reading of blueprint #242, #245 and #543, he said "i understand and agree to your proposed solution, but are Leo, Clara and the team aware of these?". READ AS: Fundraising (Clara's Finance track) is a TEMPLATE service. Clara owns the template -- the method (one builder from customers to revenue, cost, profit and cash), the rounding (raises and values after in whole millions, customers and revenue on round steps, a split by kind that adds up exactly), the one check that every figure adds up before it is recorded, the same KPIs (round, raise, value after, revenue, cost, customers, profit) with the founders' and investors' share after each round, and the skill in Sage's library -- and runs Carolverse's own plan. She never builds or runs a client project's plan by default: a project that adopts the template runs it with its own agent and its own inputs (the blueprint ruling recorded beside this one). Pursuit CAROL-PUR-0007 criterion 6.

Capital spending and investment in Denken Labs' own projects are planned apart from operating cost and carried on the balance sheet

Ruled 2026-11-07 by ninad -- superseded 2026-11-07

Ninad, 2026-11-07 (CLI-060): "for the costs, account for "own projects" an investment from carolverse, chowpatty is not one of them, also account for capex expenses, a property, about 1 million, in the seed round.. include capex in the balance sheet as well". Asked how the money for own projects should appear: "Investment, on balance sheet"; how much and when: "average cost per project as estimated, for about 10 projects"; how the property is carried: "Depreciate the building at 3% a year" (land is not depreciated). READ AS (Orion's reading, not his words): (1) Carolverse invests in about ten of Denken Labs' own projects -- projects whose client is Denken Labs; a client's project such as Chowpatty is not one -- each sized at the plan's estimated average cost of a project; the money leaves as an investment, a stake held as an asset on the balance sheet, never an operating cost, so it changes cash and the rounds' need but not profit; the years each starts are Orion's draft for his approval (#86). (3) Capital spending is one record with the property: the equipment and one-off capital items the genuine cost build already counted in its cash check (#629 names capital investments among the genuine items) join it, so nothing is counted twice. (4) Fundraising shows the plan's capital spending and investments, and the balance sheet includes them.

Capital spending and ten own-project stakes are planned apart from operating cost and carried on the balance sheet

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-060): "for the costs, account for "own projects" an investment from carolverse, chowpatty is not one of them, also account for capex expenses, a property, about 1 million, in the seed round.. include capex in the balance sheet as well". Asked how the money for own projects should appear: "Investment, on balance sheet"; how much and when: "average cost per project as estimated, for about 10 projects"; how the property is carried: "Depreciate the building at 3% a year". (3) Capital spending is one record with the property: the equipment and one-off capital items the genuine cost build already counted in its cash check (#629 names capital investments among the genuine items) join it, so nothing is counted twice. (4) Fundraising shows the plan's capital spending and investments, and the balance sheet includes them.

The plan's people are one service operator per service, three estate operators and one FDE per client, replacing the planned engineering and running hires

Ruled 2026-11-07 by ninad -- in force

Ninad, 2026-11-07 (CLI-060): "account for 1 service operator per service, and 3 estate operators for carolverse, and 1 FDE for each client..". When a round then falls short of its buffer (#630), Orion raises the smallest round needed by whole millions, values after unchanged (#631), and shows it to him before Clara's registrar records it (#86).

The fundraising plan is planned project by project, and Fundraising gains an Employees tab

Ruled 2026-11-07 by ninad -- in force

how much tokens are estimated per project? all planning should be on project level, and number of employees should be another tab on the investment app". Asked how many projects per client: "By size: small 1, medium 2, large 4"; how revenue is tied to projects: "Client value split over its projects"; how tokens per project are planned: "From Carolverse's measured use"; forward deployed engineers: "One per client (as ruled)". READ AS (Orion's reading, not his words): (1) the plan's unit is the PROJECT: a small client holds 1 project, a medium 2, a large 4, starting with 1 when the client signs and adding the rest in the years after (Orion's draft of the ramp, for his approval, #86); (2) his client totals and revenue (#636) are unchanged and each client's yearly value is split over its projects, so cost and profit can be read per project; (3) tokens are planned per project, apart from machines, drafted by size from Carolverse's own measured consumption (one counter, cost-center #282) and priced at what the subscriptions actually cost, for his approval; (4) one forward deployed engineer per client stays (#679); (5) Fundraising shows the number of employees year by year in a tab of its own, by role -- estate operators, service operators, forward deployed engineers -- from the plan's own staffing (#679), never typed.

Each client project's tokens a month are planned by its client's size

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-064), asked: "Fundraising plan (7092): how many tokens a month should each client project be planned to use?

A client adds one project a year until it holds its size's count

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-066), asked: "Fundraising plan (7092): you ruled a small client holds 1 project, a medium 2 and a large 4, each starting with 1 when it signs. How fast should a medium or large client add the rest?" He chose: "One more each year -- A client adds one project a year until it reaches its size: medium 1 then 2; large 1, 2, 3, then 4 in its fourth year." READ AS (Orion's reading, not his words): a client's year of signing is its first year and holds 1 project; each later plan year adds one until the client holds its size's count (#682), so a year's projects are the sum over its clients, each client's value split over the projects it holds that year (#682, #636), and each project's tokens follow its client's size (#683).

The plan counts admin staff, other software and Carolverse's own running cost

Ruled 2026-11-08 by ninad -- in force

He answered: "sounds right to me.. incorporate these in the model, and show me the revised numbers". Figures recorded only once he approves the revised numbers (#86).

The plan re-run by project is approved, and Fundraising gains a Cash and assets tab

Ruled 2026-11-08 by ninad -- in force

READ AS (Orion's reading, not his words): Clara's registrar records those costs, their spend shares, the admin staff's equipment and compliance from 2031 on the new headcount; Fundraising shows, year by year in a tab of its own, the plan's cash at 31 December beside its assets -- the property, the equipment and offices at what is left after depreciation, and the own-project stakes at what was paid -- derived from the plan, never typed, and never an actual (#83).

Fundraising's top level is eight tabs, simple and broad; detail lives in drill-downs

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-066): "The fundraising app currently shows a lot of metrics, some of them useful, others not so useful.. I want only the 6 or 8 tabs to be focussed first, give me the list of those tabs". Offered eight -- 1 Customers and projects, 2 Revenue, 3 Employees, 4 Cost (operating cost by part, purchases apart), 5 Profit, 6 Cash and assets, 7 Raise, 8 Ownership and value -- with the rest behind one Details view, and told the Cost tab covers 'How the money is spent, year by year', he answered: "ok, complexity can be included in the drill downs, the top level should be simple and broad.. remove all other information from the app apart from these 8 tabs for now.. update the pursuit accordingly". READ AS (Orion's reading, not his words): the page shows those eight tabs in that order and nothing else at its top level -- no Details view; the sections outside them (ownership as a section, invested value, spend shares as a section, compliance, the records and plan tables, cycles, what the money is for, comparables, tax, definitions) leave the page for now, and any detail returns only as a drill-down inside one of the eight tabs; the records themselves are untouched (only the page changes); #83 stands.

The plan budgets lawyers, the founders' marketing and PR, and risk management, which gets an owner

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-070), while Fundraising was being cut to its eight tabs: "1. The expenses associated with lawyers for regulatory compliance and founder litigations should be budgeted 2. The marketing and PR for founder expenses should be accomodated, with a strategy driven by Loki 3. Expenses for risk management should be budgeted, and this function needs an owner..". READ AS (Orion's reading, not his words): the plan's cost carries, year by year, (1) lawyers -- regulatory compliance counsel and the founders' litigation -- as lines of their own; (2) the founders' marketing and PR as a line of its own, its strategy set by Loki (Head of Marketing); (3) risk management as a line of its own, the function owned by a named agent.

The plan's lawyers, founder PR and risk management figures are approved, and the tab words are drafted from the project definitions

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-070), shown the draft for #696 -- EUR thousand a year 2027-2035: regulatory lawyers 40/60/90/120/180/240/280/320/360; founder litigation reserve 75/75/75/100/100/100/100/100/100; founder marketing and PR 60/90/120/180/240/300/300/360/360 (strategy by Loki); risk management 20/30/60/150/180/220/260/300/350 (one risk manager from 2030) -- answered "1. approved 2. approved" (2 being these figures, with Themis owning risk management, audit #697). On Fundraising's four new tab words he had said "draft them, operators are also a type of employees, but not all employees are necessarily operators. Most of these terms should already be present in project definitions, carolverse being the first project.. please check", and approved (1) Orion's draft: Customers and projects and Ownership and value map to the existing Client/Project and Founders' share/Investors' share/Post-money entries, Employee gains his sentence on operators, Cash and assets is a new entry, and the 23 older unmapped page labels are mapped to existing entries in the same change. READ AS: the four lines join the plan's cost as approved; nothing typed beyond these figures.

The plan hires subject-matter experts one per domain, FDEs are juniors who carry client portfolios weighted to onboarding, and the tenth own-project stake stays inside the plan in 2035

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-074, continuing CLI-072), on the plan re-run under #705. Told the tenth own-project stake had moved to 2036, past the plan: "bring it in within the range of planning". On forward deployed engineers: "FDEs are not juniors, they are subject matter experts and will be onboarded based on the domain knowledge required for the clients of oratorium, and the requirements for FDEs will be determined by the kind of work being done via projects on oratorium". Shown that paying one FDE per client as an expert left the Seed, Series A and Series B short of their buffers: "one per domain.. they will be recruited by domain, rather than by project, and they will be distinct from FDEs. This way, we can have limited subject matter experts which are costly, and more junior staff as FDEs". also FDE are required mostly during the onboarding of the project, then their involvement reduces over time, as the project matures". Shown Orion's draft of the portfolio load and its buffers, he chose "Half, quarter, tenth".

The first office fit-out moves from 2028 to 2029, paid by Series A, so the Seed keeps its buffer

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-074), on the plan re-run under #705 and #706.

Clara runs the investment work: she knows the plan's reasoning, re-runs the plan with her own tools, talks to investors directly as Leo talks to customers, and acts within limits

Ruled 2026-11-08 by ninad -- in force

Ninad, 2026-11-08 (CLI-074): "I want clara to not only own the app, but to know the rational of building the investment plan and all the details of the plan and strategy, including but not limited to fundraising, costing, planning, reaching out to investors, basically everything.. how can we enable clara, so that she is not just informed, but also actively involved in doing the planning and work for raising investments?"; told she can read the plan's figures but not the rulings behind them, that her own knowledge holds nothing on fundraising, that the cost model lives in Orion's working files and that she only records what is approved: "can she not build the fundraising app? (like you on her behalf or something...) will that help her to be in total control? I want investors to interact directly with clara just like customer interact with leo. For her to be able to interact with investors, she should be equipped to think, understand, analyze, reason, respond and do things that are necessary to be done". Shown that control comes from owning the knowledge, the tools and the duties (understand: the rulings opened to her and the plan's reasoning in her own knowledge; analyze and reason: the cost model as her own tool; respond: an investor chat like Leo's customer chat, every conversation feeding one investor record; act: investor statuses, Data Room documents, meetings, follow-ups and updates; limits: valuation, terms and anything binding go to him, salaries stay private), and that #235 and #233 block it, and asked whether to amend both so investors can sign in and talk to Clara within those limits and file the build as one pursuit with Clara accountable, he answered "yes". READ AS (Orion's reading, approved as shown): (1) Clara (CEO; getting investments is hers, cost-center #493) is not only the Fundraising app's owner but the one who plans and raises the company's investment -- fundraising, costing, planning and reaching out to investors; (2) UNDERSTAND: the Rulings Repo is among the records her chat and wake read, and the investment plan's reasoning -- every ruling the plan rests on, its assumptions and its levers -- is written into her own knowledge and kept current whenever the administrator rules on the plan; (3) ANALYZE AND REASON: the cost model (today Orion's working files) becomes Clara's own tool -- she re-runs the plan when a ruling or an actual changes, tests what-ifs, measures each round's buffer and proposes the fix; the administrator approves every figure before her registrar records it (#86); (4) RESPOND: investors talk to Clara directly in an investor chat of hers, as customers talk to Leo (#538), and every conversation feeds the investor's one registry record (#236); (5) ACT: she moves an investor through the statuses of #240, shares Data Room documents, books meetings and drafts follow-ups and the monthly update and weekly showcase (#233); mail to an investor stays a draft for the administrator until he rules otherwise; (6) LIMITS: valuation, terms, anything binding and any commitment go to the administrator; the founders' and board's planning salaries stay private (#629); she says only what the plan, the rulings and the Data Room hold; (7) Orion builds it on her behalf through the Orion bypass, recorded with Clara accountable, as one pursuit.

An investor is its own kind of person, gets in only through the investor onboarding pass, and sees the Data Room read-only and talks to Clara -- nothing else

Ruled 2026-11-08 by ninad -- in force

An investor is its own kind of person, separate from users and operators, gets in only through the investor onboarding pass, and is granted READ-ONLY access to the Data Room and a conversation with Clara in her investor chat -- nothing else in the estate. Ninad, 2026-11-08 (CLI-074): "I want clara to not only own the app, but to know the rational of building the investment plan and all the details of the plan and strategy, including but not limited to fundraising, costing, planning, reaching out to investors, basically everything.. how can we enable clara, so that she is not just informed, but also actively involved in doing the planning and work for raising investments?"; told she can read the plan's figures but not the rulings behind them, that her own knowledge holds nothing on fundraising, that the cost model lives in Orion's working files and that she only records what is approved: "can she not build the fundraising app? (like you on her behalf or something...) will that help her to be in total control? I want investors to interact directly with clara just like customer interact with leo. For her to be able to interact with investors, she should be equipped to think, understand, analyze, reason, respond and do things that are necessary to be done". Shown that control comes from owning the knowledge, the tools and the duties (understand: the rulings opened to her and the plan's reasoning in her own knowledge; analyze and reason: the cost model as her own tool; respond: an investor chat like Leo's customer chat, every conversation feeding one investor record; act: investor statuses, Data Room documents, meetings, follow-ups and updates; limits: valuation, terms and anything binding go to him, salaries stay private), and that #235 and #233 block it, and asked whether to amend both so investors can sign in and talk to Clara within those limits and file the build as one pursuit with Clara accountable, he answered "yes". READ AS: #235's other terms stand (own kind of person; in only through the pass; first-class registry records); what an investor is granted is now the Data Room read-only AND Clara's investor chat, where Clara answers within her limits (leadership #709). Recording an investor as a user or an operator so an existing door lets them in, granting any write, or granting anything beyond the Data Room and Clara's investor chat, each undoes this ruling.

Investors get a monthly update and a weekly showcase; the investor pass is approved to run

Ruled 2026-11-08 by ninad -- in force

Investors hear from the company on two rhythms, both written by Clara from her own address: a MONTHLY investor update and a WEEKLY showcase of interesting updates. The investor onboarding pass is APPROVED to run, so investors can sign in and talk to Clara within her limits. Ninad, 2026-11-08 (CLI-074): "I want clara to not only own the app, but to know the rational of building the investment plan and all the details of the plan and strategy, including but not limited to fundraising, costing, planning, reaching out to investors, basically everything.. how can we enable clara, so that she is not just informed, but also actively involved in doing the planning and work for raising investments?"; told she can read the plan's figures but not the rulings behind them, that her own knowledge holds nothing on fundraising, that the cost model lives in Orion's working files and that she only records what is approved: "can she not build the fundraising app? (like you on her behalf or something...) will that help her to be in total control? I want investors to interact directly with clara just like customer interact with leo. For her to be able to interact with investors, she should be equipped to think, understand, analyze, reason, respond and do things that are necessary to be done". Shown that control comes from owning the knowledge, the tools and the duties (understand: the rulings opened to her and the plan's reasoning in her own knowledge; analyze and reason: the cost model as her own tool; respond: an investor chat like Leo's customer chat, every conversation feeding one investor record; act: investor statuses, Data Room documents, meetings, follow-ups and updates; limits: valuation, terms and anything binding go to him, salaries stay private), and that #235 and #233 block it, and asked whether to amend both so investors can sign in and talk to Clara within those limits and file the build as one pursuit with Clara accountable, he answered "yes". READ AS: the cadence of #233 stands (monthly update plus a weekly showcase, not a second monthly update, from Clara's own address, never the administrator's or Carol's, cookbook 1809); the pass is no longer a draft -- its tasks may be declared on the Finance track and it runs, built by Orion on Clara's behalf with Clara accountable (leadership #709). Dropping the weekly showcase, folding it into the monthly update, or sending either from the administrator's or Carol's address each undoes this ruling.

Clara reads the rulings for the plan's reasoning and the registry for its figures; investors reach her through her external chat, as guests or verified investors with different access

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-080), building (Clara reads the rulings and knows the investment plan's reasoning). Asked: Clara's investor chat will read the live leadership rulings so she can explain why the plan is what it is, and ruling #631 (the founders' 2026 sweat equity) states roughly what his October-December 2026 work is worth at his planning salary, so his salary can be worked out from it, against #629 and #709 -- should #631 be restated without the figure, should Clara skip its text, or should it be left in? He answered: "Clara should get access to the rulings to understand the rational, but her trusted source are the registry tables.. most of the financial details that are not yet worked out will get updated there, Clara should have access to the information via her "external" chat (not internal) where users are approved investors. Investors will have categories as guests and verified, with varying level of access to information". READ AS (Orion's reading, marked apart from his words): (1) REASONING FROM THE RULINGS, FIGURES FROM THE REGISTRY: Clara reads the Rulings Repo to understand why the investment plan is what it is; for every figure -- a round, a cost, a headcount, a buffer -- her trusted source is the registry's tables, where the financial details not yet worked out are updated as they are; where a ruling's figure and the registry differ, the registry is what she states, and the ruling is why. (2) THE EXTERNAL CHAT IS THE INVESTOR DOOR: investors reach this information only through Clara's EXTERNAL chat, whose users are approved investors -- not her internal chat; the 'investor chat' of #709 and #710 is that external chat. (3) TWO CATEGORIES OF INVESTOR: an investor is a GUEST or VERIFIED, and each category has its own level of access to information; which information each category sees is not yet ruled. (4) NOT ANSWERED HERE: whether #631's figure, from which his planning salary can be worked out, stays in the text Clara reads (#629, #709) -- asked again.

A year's spend is paid by the round completed the year before; every cost line item names its type and the Carolverse service that owns it, for that service's own P&L

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-080), building (one spend plan: each spend line names its category, year and the round that pays for it). (1) Asked which rule decides the round that pays a spend line, the plan's money being pooled (rounds and revenue add cash, costs draw it down month by month), he answered: "the expense for the year is paid by the round that is completed in the previous year". by type and by carolverse departments, where individual services use funds for their own P&L (as per agent economy model)". (3) Asked how each year's cost should be split across services, the plan's future costs existing only by type: "all cost line items will have to be assigned to services, and owned by service owners. If you are not able to decide certain cost items, bring them to my notice.. but with recommendations". READ AS (Orion's reading, marked apart from his words): (1) THE PAYING ROUND: a plan year's spend lines name the round completed in the year before it; where no round completed in the year before, the line names the latest round completed before the year began (Orion's reading of a year that the words leave open -- on the recorded rounds 2027-2028 read Seed, 2029-2030 Series A, 2031-2032 Series B, 2033-2035 Series C), and 2026, before any round, reads the founders' own investment. The round a line names is a label of which raise the year draws on; whether a round's cash lasts stays the buffer check's month-by-month question. (2) TWO DIMENSIONS: every spend line names its TYPE (employees, tokens, infrastructure, other, compliance per regulation) and its DEPARTMENT, the Carolverse service whose own P&L bears it (cost-center #695, #718); the seven old use-of-funds categories are retired in place, and use of funds per round, per type and per service are worked out from the lines. (3) OWNED BY SERVICE OWNERS: every cost line item is assigned to a service and owned by that service's owner, read live from the services catalogue; an item Orion cannot place is brought to the administrator with a recommendation, never parked under a default service.

Salaries sit with Agent and Human Resources, one service operator per service, every planned spend has its month and year, Carolverse's own tokens follow measured use, and winning clients is Sales' cost

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-080), building under leadership #722 (every cost line item names its type and the service that owns it; items Orion cannot place come to him with a recommendation). (2) Asked which services bear the 22 service operators' pay, Orion having said the plan hires all 22 in 2027, recommended one per service: "one per service, but did the plan not change to spread the hiring across 2027 and 2028? every spend planned needs to have a spend month & year" -- it did: the recorded staffing holds 8 service operators at the end of 2027 and 22 at the end of 2028 (Orion's question was wrong). (3) Asked which services bear Carolverse's own tokens (#684), recommended splitting by each service's measured use, he chose "By measured use". (4) Asked which service bears the plan's sales and marketing per new client (#679), recommended Sales, he chose "Sales". READ AS (Orion's reading, marked apart from his words): (1) SALARIES WITH HR: every salary in the plan is borne by Agent and Human Resources (Athena) -- the employees' pay with its employer costs, and (Orion's reading of 'employee salaries' for the people the question named) the founders' and board's pay too; pay is kept as totals so no individual figure can be worked out (#629, #705). (2) ONE SERVICE OPERATOR PER SERVICE, hired as recorded: 8 by the end of 2027, 22 by the end of 2028; their salaries sit with HR by (1). (3) MONTH AND YEAR: every planned spend line names the month and the year it is spent -- a recurring cost month by month, a one-off in its month, a hire from the month the person starts. (4) Carolverse's own tokens (one large project, #684) are split across services by each service's measured token use, re-measured as actuals come in; the year's total stays as approved. (5) The plan's sales and marketing per new client is borne by Sales (Aurora); founder PR stays Marketing's own line (#696).

Spend is sequenced month by month to support the customer and revenue targets, and the spend plan flags its risks upfront, such as a delayed round and its impact

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-080), building. Asked how the spend lines get their months, every planned spend needing a month and a year (#723) while the approved plan is yearly and no record says the month a planned hire starts or the month of the one-offs (second-market setup 2032-33, IPO preparation 2034-35, listed-company costs 2035, the property's two halves) -- spread each year evenly for now (recommended), start hires the month after the paying round lands, or he gives the months -- he answered: "the spends will have to be sequenced to support the customer and revenue targets, and any risks should be identified and flaged upfront in the spend plan (for eg. risks like.. if funds are delayed, what would be the impact etc..)". READ AS (Orion's reading, marked apart from his words): (1) SEQUENCED TO THE TARGETS: a spend line's month comes from what the plan's customer and revenue targets need and when -- a hire starts when the clients it serves need it, a cost begins when the work it supports begins -- never an even spread chosen for convenience and never a guessed month; the even spread offered as an interim is declined. (2) RISKS UPFRONT: the spend plan carries its risks beside the lines, each named with what triggers it and its impact on cash, runway, the customer and revenue targets and the rounds -- for example a round landing late -- so a risk is read before the spend, not discovered after it. (3) The figures that result are still approved by the administrator before Clara's registrar records them (#86).

Throughout the investment pursuits, the plan is checked from time to time for coherence across every dimension and for being realistic and achievable

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-087), as the session picked up the investment chain: "keep checking from time to time throughout the execution of the pursuit that the numbers across all the dimensions of the plan are consistent and coherent, and the plan is realistic and achievable". READ AS (Orion's reading, marked apart from his words): (1) WHAT: every dimension of the plan -- customers by size, revenue, projects, staffing, cost by type, service and month, compliance, capital and stakes, profit and tax, cash and each round's buffer, raises, ownership and valuation -- is checked two ways: COHERENCE, where two dimensions must agree by arithmetic or by a recorded rule and do not; and REALISM, where a figure sits outside the bar an investor would read it against, the bar and its source stated beside the finding. (2) WHEN: at the start and close of each step of the investment pursuits (CAROL-PUR-0007 Investment chain, CAROL-PUR-0028 Clara raises investment) and whenever plan figures are recorded or shown for approval. (3) HOW: one repeatable check reading the figures through the fundraising door, never by eye (agent-resources #97); a figure the record leaves unmeasured is reported so, never as zero. (4) WHO DECIDES: a coherence finding is a defect to cure; a realism finding is brought to the administrator, and the check never changes a figure he decided (#86).

Profit counts the revenue earned in each year, never the revenue running at its year end

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-087), answering the first findings of the plan's coherence and realism check (#731). Asked: "Profit counts the revenue running on 31 December as the whole year's revenue. The cash forecast already collects revenue month by month as it grows, so profit and cash disagree. Which should profit use?" -- options 'Earned in the year' and 'Keep year-end run rate'. He chose: "Earned in the year". READ AS (Orion's reading): the profit and loss, its margins and its tax read the revenue earned in the year, worked out month by month by the SAME derivation the cash path uses, never a second one; the revenue running at 31 December stays a plan figure and is shown as the year-end run rate (ARR), named so, wherever it is shown.

Every round names what it buys; a round the plan does not spend names its use, approved before it is recorded

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-087), answering the first findings of the plan's coherence and realism check (#731). They leave 52 and 83 months of costs unspent, against buffers of 18 and 24 months. Investors will ask what that money buys. What should these rounds do?" -- options 'Name a use', 'Shrink them', 'Keep as cushion'. He chose: "Name a use" (keep the raises and give each a recorded use; Orion drafts, he approves before it is recorded). READ AS (Orion's reading): Series C and the IPO keep their amounts; the money each leaves above its buffer gets recorded uses -- spend lines or capital items the round pays (#722) -- drafted by Orion, shown to the administrator and approved before Clara's registrar records them (#86); a round that leaves money above its buffer with no named use is a finding of the plan check (#731).

Client projects' tokens are priced at per-use rates from measured use, never at flat-subscription cost

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-087), answering the first findings of the plan's coherence and realism check (#731). That is why profit margins reach 42% (2032) to 66% (2033); the best listed software companies run 25-40%. At client scale, model use is usually billed per use, not by flat plan. Re-price the tokens?" -- options 'Per-use rates' and 'Keep subscription pricing'. He chose: "Per-use rates". READ AS (Orion's reading): the tokens a client project uses each month are Carolverse's measured use per project priced at the vendors' per-use rates, each rate recorded with its source and date; this RETIRES IN PLACE the subscription pricing of #683/#684 for client projects (the measured use itself stands); the new figures go to the administrator before Clara's registrar records them (#86), and Cost, margins and the cash path follow.

The plan carries two insurances: employee-related cover in the per-person overhead, business cover in compliance

Ruled 2026-11-09 by ninad -- in force

He chose "Different insurances". READ AS (Orion's reading): both stay; each line's basis names its scope, so the plan check (#731) does not read them as one cost counted twice.

Fundraising states Carolverse's business model below its charts, and building Oratorium is named as Carolverse's significant, continuing cost

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-087): "what is the business model of carolverse project? this should be documented in the fundraising app.. clearly explaining the revenue model and profit margin, the assumptions, risks and potential opportunities in short and presentable manner below the tab charts. Note that carolverse is funding oratorium, so building oratorium is the significant cost that carolverse project has and will continue to incur.. ammend the pursuit with these requirements". READ AS (Orion's reading, marked apart from his words): (1) the Fundraising app shows, below the tab charts, a short presentable statement of Carolverse's business model: how it earns (the revenue model), its profit margin, the assumptions the plan rests on, its risks and its potential opportunities; every figure in it is read live from the plan's records, never typed, and the words are Clara's record (her Finance track, #1724), so a revised plan revises the statement; (2) Carolverse funds Oratorium (the platform it runs on, rule 1351), so the plan names BUILDING ORATORIUM as its own cost -- the people who build it, the build's model use and its machines -- and the statement says it is Carolverse's significant cost now and for as long as the plan runs; the figures go to the administrator before Clara's registrar records them (#86); (3) the investment-chain pursuit (CAROL-PUR-0007) carries both in its MVP.

Building Oratorium's model use is estimated from the platform being 5% built, and compared with the list price; today's flat-plan payments are not the basis

Ruled 2026-11-09 by ninad -- in force

The platform is still being constructed, and it will take much more tokens than what is currently consumed.. assume current platform build at 5% to do the estimations, and check if its close to the list price.. give me the comparison". READ AS (Orion's reading, marked apart from his words): the platform (Oratorium) is taken as 5% built today; the model use of the whole build is what the build has consumed so far scaled to 100% (what remains: 95/5 of it), measured from the ledger; it is compared with the list price of that use and with today's payments, and the comparison goes to him before any figure is recorded (#86).

The people who build Oratorium are the estate operators and the service operators

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-092), asked who 'the people who build Oratorium' are in the plan (no ruling said so): chose "Estate + service operators" -- the founders and operators on the planning salaries, the third estate operator and every service operator (the 'seniors', #705, #679, #723). FDEs and subject-matter experts deliver to clients (#706) and are not the build. READ AS: the plan names their pay as the people cost of building Oratorium, moved out of the lines they sit in today, never counted twice.

An Expense Plan app owned by Clara shows the company's planned expenses by year from the registry's expense plan, described and refreshed the way Carolopedia is

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-092): "I want an app that provides the plan for expenses in the standard carolverse app format highlighting Cost summaries with drill downs that provide descriptions and breakdowns of the costs, sorted by years, 2027 on top, 2035 or 2036 (whichever is last) in the bottom. The descriptions should be descriptive, dont mention names, but using "founders" is ok. Service filters should be available.. app should be called "expense plan", company name "Denken Consulting & Services Pvt Ltd" The app should be owned by clara, it should use registry expense plan as its source.. it should be updated regularly the same way carolopedia gets updated.. accomodate the tasks with appropriate budgets and associate them to the right service, track, slot. Create new droids as necessary.. ensure compliance to agent centric modular architecture in the design.. build droids as required.. update dictionary, SST etc.. fold this requirement into this pursuit". READ AS (Orion's reading, marked apart from his words): (1) an app named 'Expense Plan' in the standard Carolverse app format, naming the company Denken Consulting & Services Pvt Ltd; (2) one cost summary per year, 2027 at the top down to the plan's last year at the bottom -- 2026, the year now ending, is not shown -- each drilling down to its spend groups, types, items and lines with a description and a breakdown; (3) every description descriptive and naming no person, 'founders' allowed; (4) a service filter showing each service's share by Midas's split (cost-center #736), an unsplit item named, never guessed; (5) the app is Clara's and its one source is the registry's expense plan, read live, never typed; (6) the descriptions are written by Clara's droids and kept current the Carolopedia way -- a change detector, a writer, a daily sweep -- each a registered task with its budget on the leadership service's track and slot, on her own login, registered as Carolverse processes; the app only reads; (7) the data dictionary, the SST register and the app registry carry it; (8) it is part of the investment-chain pursuit (CAROL-PUR-0007).

Building Oratorium continues at today's pace every year through 2035

Ruled 2026-11-09 by ninad -- in force

Operations stay as #752 rules.

The plan onboards one operations staff member per service, a year after its service operator; estate and service operators build, operations staff run

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-092): "I need to also onboard operations staff, 1 per service, where the estate and service operators manage the new build, while the \"ops staff\" manage the operations, accomodate them in the plan". Asked when they join (service operators 8 in 2027, 22 by 2028): "A year later".

Revenue grows through more clients and more business per client at today's prices and margins, new clients even out, and the plan reads smooth with no spikes or surprises

Ruled 2026-11-09 by ninad -- in force

READ AS (Orion's reading, marked apart from his words): the extra revenue comes from more clients and more projects per client at the prices and margins the plan holds (#636, #682-#684), so the delivery costs that follow clients and projects grow with them (forward deployed engineers, experts, project model use and machines, sales and marketing per new client #679, and every person's per-person costs); new clients follow a smooth path; every year's figures change gradually, with no jump in cost, margin or growth that the plan cannot explain; the result goes to him before it is recorded (#86).

Founders sell to the first clients, and version B is the plan: raises 10/30/40/50/100M at 50/100/300/600/1,000M, margins rising steadily to 40% by 2035

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-092), shown version B built line by line beside version A: "Founders sell to the first clients themselves - approved, show me version B in clean table". Clara's registrar records it; the round words and the plan check follow.

The approved round table sits above every chart in Fundraising and is the primary reference for all later planning

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-092), shown version B's round table (#757: round, raise, founders' share after, shares sold, year-end revenue rate, pre-tax profit, post-money valuation): "i want this table to be put up in fundraising app above all the chart as the primary reference for all subsequent planning". READ AS (Orion's reading, marked apart from his words): (1) the table is a record of its own -- the plan's targets per round, Clara's, changed only through her registrar on his approval (#86) -- and Fundraising shows it first, above every chart and tab, with the founders' share and shares sold derived from the raises and valuations, never typed; (2) every later planning step (a cost, a hire, a round's words, a re-run) is checked against it: the plan check names any year whose revenue rate, profit or a round's raise or valuation departs from the table, and a departure is brought to him, never absorbed silently.

The round table is the official name of the approved table of the plan's rounds, and Orion shows it the same way every time he is asked

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-100), shown the approved table of the plan's five rounds (#757, #758) and having called it the round table: "I like the name "round table" - make that as the official term in dictionary.. ensure that I get a consistent view of the round table everytime I ask". READ AS (Orion's reading, marked apart from his words): 'round table' becomes the dictionary term for the table Clara's record holds (shown today on the Fundraising page under 'The plan's targets'), its meaning written only once he approves the words (P.01.01.05.11); whenever he asks for the round table, every Orion session shows it read live from Clara's record in one fixed shape -- the columns and order the Fundraising page uses -- kept as one skill in Sage's library bound to Orion, never composed afresh by a session.

The round table carries the number of employees at each round

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-100), shown the round table (#758, #759) and version B's scale-team headcount: "add count of employees to the round table". READ AS (Orion's reading, marked apart from his words): each round's row gains the plan's total employees at the end of that round's target year (the year its customers are counted), a target Clara records with the others once Ninad approves the figures (#86), and checked against the plan's staffing like the other targets; the Fundraising page, the round table's one view (#759) and the pitch deck's round table slide show it as one more column.

Version B's smoothed headcount is approved, and the round table shows employees rounded: 20, 50, 100, 200, 300

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-100), shown version B's scale teams smoothed so headcount never drops (2031 51, 2032 51, 2033 110, 2034 127, 2035 127 scale-team people; total employees 20 in 2027, 53 in 2028, 99 in 2030, 191 in 2032, 303 in 2035) and the employees column for the round table (#760): "version B headcount is fine, but they should be rounded off.. 20, 50, 100, 200, 300". READ AS (Orion's reading, marked apart from his words): the plan records version B's headcount as smoothed; the round table's employees are Seed 20, Series A 50, Series B 100, Series C 200, IPO 300, rounded as the customers were; a plan figure within that rounding is not a departure from the table.

The plan provisions own infrastructure as risk management and funding for new ventures, over several years

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-100), shown the plan check after version B was recorded (admin staff short of one per 15 from 2031; each client size's value not yet version B's; the IPO raising money it does not name a use for): "I agree and approve your analysis.. here is my take on it 1. The plan currently lacks risk management.. to ensure steady supply of infrastructure and intelligence, the company will have to deploy the own infrastructure, training facilities etc which will need capex 2. For growing the company further, it needs to pivot, and investment projects will need funding to spring into new ventures. The details of these are not currently known, but the plan should provision this in the roadmap. Both these will require multiple years of investments and meticulous planning to make it happen". READ AS (Orion's reading, marked apart from his words): the admin staff are added at one per 15 and each client size's value follows version B; the plan gains capital items for own infrastructure and training facilities and a provision for new ventures, each over several years, paid by named rounds, with roadmap items for both; the figures go to him before they are recorded (#86).

The moat is built without weakening the company's ability to raise its rounds

Ruled 2026-11-10 by ninad -- in force

also, find out other existencial risks and moats necessary to mitigate them". READ AS (Orion's reading, marked apart from his words): the round table stands (#758); moat spending is paced so every round still meets its profit, revenue and customer targets and its cash reserve (no lower valuation, no shorter runway), the heavy capital build following the round that pays for it; Orion lists the company's other existential risks with the moat that answers each, for his decision before any figure is recorded (#86).

Clients are sticky: they sign small and grow in size with the company, and early losses are accepted while the cash holds

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-102), shown that customers times each size's value does not add up to version B's revenue (the plan check finding he approved fixing, #762) and offered a fitted client mix or a value per size per year: "I would like the customers to be sticky, which means, we grow with them and they grow with us.. so a few small clients should become few large clients.. with this vision in mind, build a rampup plan where the customer size increases as we move ahead, loss in initial years is ok as long as cash flow is assured..". READ AS (Orion's reading, marked apart from his words): each size keeps its one value; new clients sign small and a share of each size moves up a size every year they stay, so the mix and the revenue come from the clients' growth and add up by construction; losses in the early years are accepted provided every round still keeps its cash reserve (#679); where the ramp-up parts from the round table (#758) each departure is named for him, and the plan is recorded only once he approves its figures (#86).

Large clients have three tiers, bronze, silver and gold, that clients progress through as the years pass

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-10 (CLI-102), while the sticky ramp-up (#766) was being built: "maybe introduce the sub categories of bronze, silver and gold customers, within the large customers, where customers mature from small and medium to large, and progress from bronz to gold.. as the years progress".

The plan's years run April to March: 2027 spending starts in April and everything moves out three months

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-11 (CLI-102). Asked "can you plan seed from jan 2027 to mar 2027?", shown that cash would fall to about EUR -1.5M in January and February 2027 before the Seed lands, he chose "2027 spending starts in April, everything gets pushed out by 3 months"; offered two ways (plan years running April to March with every approved figure unchanged, or calendar years with every figure reworked three months later), he said "ok, lets go with your recommendation" -- the first. READ AS (Orion's reading, marked apart from his words): every plan year runs April to March and is labelled by both calendar years (2027/28 = April 2027 to March 2028); each year's customers, revenue, costs, staffing and the round table's figures stay as approved and are reached at the year's end in March; the rounds' planned months move three months later -- Seed January to March 2027, Series A April 2028, Series B September 2030, Series C September 2032, IPO September 2035 -- so each still lands at the same point of its plan year and every buffer reads as before; anything that sets a plan year beside calendar months (a round's month, the company's actual spending) reads the plan year's own twelve months.

Each customer group's average yearly revenue is an assumption written beside the round table, not a record

Ruled 2026-11-10 by ninad -- in force

READ AS (Orion's reading): the five values are approved as the plan's assumptions; they are written out with the round table wherever it is presented (the pitch deck's round table slide first) and are not added to Clara's records or the Fundraising app as a table. The split of customers into the five groups is a separate decision, not yet taken.

The listing comes by 2034, not 2036

Ruled 2026-11-10 by ninad -- in force

Ninad, 2026-11-11 (CLI-102), asked "can IPO happen sooner? The figures go to him before anything is recorded (#86).

The revenue model is a fixed price for agreed scope: one contract per account, a fixed price per project

Ruled 2026-11-11 by ninad -- superseded 2026-11-11

Ninad, 2026-11-11 (CLI-102). Asked "what is the revenue model? is it % more of cost? is it based on value delivered?", then "it cannot be both ... it has to be either 1 of the 3 options - 1. % of token consumed, 2. fixed price for scope agreed, 3. value delivered"; told Orion recommends 2 (under AI deflation our costs fall inside each contract while the price holds; a small or medium client can budget one known price; value is hard to measure and a token share passes deflation straight into our income): "we dont have the estimation module built yet for that.. but i think it should be agreed in the contract for each project, or for each account"; shown one master contract per account with a fixed price per project's agreed scope inside it, then "update the round table and customer mix table, record the revenue model, update the slide". READ AS (Orion's reading): each client account signs one master contract (terms, payment, renewal); each project is a fixed yearly price for its agreed scope, added to that contract and repriced at renewal; until an estimation module exists the founders agree each project's price by hand; the client size tiers (small, medium; large as bronze, silver, gold) describe the account and are bundles of scope; the plan's figures stand.

Ended because: Ninad chose tokens plus margin as the default over #778's fixed price for agreed scope

The plan names two assumptions: prices deflate after AGI, and the digital economy grows ten times

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-11 (CLI-102): "2 things that I think are missed out 1. the potential deflation in post AGI economy 2. the new digital economy which will be 10 times the current economy". Shown how they enter the plan, then "update the round table and customer mix table, record the revenue model, update the slide". Ten times in a decade is above mainstream forecasts and is his stated assumption.

A round's cash buffer is met when the plan is profitable in the year the next round lands

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-11 (CLI-104), asked "Investment plan buffer rule: should a round count as safe (its cash buffer met) when the plan is already profitable in the year the next round lands? I recommend yes. It is recorded as your ruling only if you say yes.": "Yes". READ AS (Orion's reading, marked apart from his words): #630's growing buffer of months of costs still applies to a round whose next round lands in a plan year with a loss; a round whose next round lands in a plan year with a profit above nought (#777's profit column) holds without the months-of-costs test.

The pre-seed holds a buffer of three months of costs when the Seed lands

Ruled 2026-11-11 by ninad -- in force

Orion proposed 3 months; the plan as drawn leaves about 8.": "3 months". READ AS (Orion's reading, marked apart from his words): the pre-seed joins #630's ladder of buffers below the Seed's six months; the lean year 2027/28 (#772) leaving about eight months of costs when the Seed lands meets it.

Client projects make a profit; the company may run a loss only from spend that is not project work

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-11 (CLI-106): "cost budget is revenue minus 20% profit, does that sounds right?, this should employee, infra, token and other categories"; shown that the plan reaches a 20% profit only from 2034/35 and spends ahead of revenue before that: "client projects should generate profits, negative profits are ok as long as they are a result of spends for platform enhancements, marketing, PR, sales, compliance etc, basically non-project work..". READ AS (Orion's reading, marked apart from his words): the plan's spend is split into project work (delivering client projects) and non-project work (platform enhancements, marketing, PR, sales including the prospect budgets, compliance and the like); each client project's cost budget is its revenue less a 20% profit, across employees, infrastructure, tokens and other; a company-level loss in a plan year is acceptable only where non-project spend causes it, and the rounds fund it.

The revenue model is tokens consumed plus a margin by default; a scope or value price only where the contract agrees it

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106). On 2026-11-11 (CLI-106) he wrote the revenue model for the pitch deck in HIS WORDS: "what is the revenue model - token consumed + margin (default) or scope delivered or value delivered model mutually agreed per project or for entire client account". READ AS (Orion's reading, not his words): the default revenue model is the tokens a client's projects consume, at measured per-use cost (#734), plus the project margin (#789), plus the Blueprint subscription; a fixed price for agreed scope, or a price for value delivered, is used only where client and company agree it in the contract, for one project or for the client's whole account; each client account still signs one contract (the master terms, with the standard payment terms of the payment-terms ruling); the plan's figures stand.

Spend is the central fact: every spend is a first-class registry record carrying its year, month, project, service, department and type, and every split is worked out from those records

Ruled 2026-11-11 by ninad -- superseded 2026-11-11

Ninad, 2026-11-12 (CLI-106). All records should be first class registry items".

Ended because: Ninad widened #803: the six dimensions are examples of an open set, and changes to the records are kept over time

Spend is the central fact: every spend is a first-class registry record carrying an open set of dimensions, every split is worked out from those records, and every change to them is kept over time

Ruled 2026-11-11 by ninad -- superseded 2026-11-11

Ninad, 2026-11-12 (CLI-106). All records should be first class registry items". Then, told it was recorded with those six dimensions, HIS WORDS: "year, month, project, service, department and type - these are just examples, there will be more such dimensions that will need to be captured as well.. also changes will have to be tracked over time to the records".

Ended because: Corrects Orion's reading in #804: the cost-by-type gap covers 2027, 2028 and 2030-2036 (shares sum 108.8%, 102.1%, 98.3-99.4%), not only from 2030

Spend is the central fact: every spend is a first-class registry record carrying an open set of dimensions, every split is worked out from those records, and every change to them is kept over time

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106). All records should be first class registry items". Then, told it was recorded with those six dimensions, HIS WORDS: "year, month, project, service, department and type - these are just examples, there will be more such dimensions that will need to be captured as well.. also changes will have to be tracked over time to the records".

The round table carries each round's cash reserves and assets at its year end, in Fundraising only, never on the deck

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-12 (CLI-106), working the investment chain after the round table and the pitch decks were updated, HIS WORDS: "the round table should also include the cash reserves and assets column, but they should not go in the deck". READ AS (Orion's reading, not his words): (1) Clara's round table gains two columns: CASH RESERVES -- the plan's cash in the bank at the end of the row's plan year (the year its customers, employees, revenue rate and profit are read for; plan years run April to March, #768) -- and ASSETS -- what the company owns then besides cash, as #689 defines them: the property, the equipment and offices at what is left after depreciation, and the own-project stakes at what was paid (#678); (2) like the table's other figures (#758) they are the plan's targets per round, worked out from the plan's cash path and capital records when the table is recorded, never typed, kept through Clara's registrar on his approval (#86), and the plan check names any departure from them; (3) they are shown in Fundraising's round table only: the pitch decks' round table slide (the Oratorium Pitch Deck and the Denken Consulting Services deck) does not carry them.

Planning runs from the broader goals to the details and back when the risk is high -- for every plan, Clara's or Leo's

Ruled 2026-11-13 by ninad -- in force

PLANNING RUNS FROM THE BROADER GOALS TO THE DETAILS, AND BACK WHEN THE RISK IS HIGH -- FOR EVERY PLAN, CLARA'S OR LEO'S (Ninad, 2026-11-13, CLI-132).

ANSWERED, HIS WORDS: "keep funding ask and revenue targets as planned, adjust the headcount accorddingly.. its a model, everything can be changed, like assumptions, targets, ask.. everything, as long as it makes sense, as in, as long as it is feasible and is alligned with the broader goals. Hence we first define the broader goals, then adjust the details to make it possible.. if we think the risk factor is high, then we go back and adjust the broader goals.. this process should be followed in all planing activties, whether Clara is planning it for carolverse or whether Leo is planning it for any new project.. meanwhile, for now, dont change anything in the round table, just adjust the headcounts keeping all the other broader goals and assumptions same."

The headcount shown on the round table is a broader goal under (4); the staffing record's figures are the detail being adjusted -- where they part, the departure is named, not hidden.

Sales cost is proportional to the client's value, and prospects' sales cost is counted

Ruled 2026-11-13 by ninad -- in force

SALES COST IS PROPORTIONAL TO THE CLIENT'S VALUE, AND PROSPECTS' SALES COST IS COUNTED (Ninad, 2026-11-13, CLI-132).

ANSWERED, HIS WORDS: "that makes sense - make sales cost proportional to the client's value, but ensure that the sales for prospects is also accounted for.. then show me the same analysis".

Every planning figure is rounded off -- hence the round table

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

EVERY PLANNING FIGURE IS ROUNDED OFF -- HENCE THE ROUND TABLE (Ninad, 2026-11-13, CLI-132).

HIS WORDS: "all planning should be rounded off (hence the name - round table)".

The rule binds every plan -- Clara's for Carolverse and Leo's for a project (#826).

Ended because: Ninad refined the rounding rule the same hour: rounded for reporting, actuals kept for tying back; the yearly costs are not re-recorded

Every planning figure is reported rounded, and the actuals are kept to tie the numbers back

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

EVERY PLANNING FIGURE IS REPORTED ROUNDED, AND THE ACTUALS ARE KEPT TO TIE THE NUMBERS BACK (Ninad, 2026-11-13, CLI-132; refines #829).

HIS WORDS: "use the rounded for all reporting purposes, while use the actuals for tying back the numbers. the numbers look different from what I suggested, mine are also rounded: capital --> 1 M; Compliance + Other --> 1M; People + machines + claude seats --> 4M; In people: service operators --> 2M, sales --> 2M".

Ended because: Ninad: the values must match his, the rounding is essential -- the plan holds round figures; real figures tie back

The pre-seed year buys no property: the Seed's property half is bought in the Seed's year, and a round's row ties back in round figures

Ruled 2026-11-13 by ninad -- in force

THE PRE-SEED YEAR BUYS NO PROPERTY: THE SEED'S PROPERTY HALF IS BOUGHT IN THE SEED'S YEAR, AND A ROUND'S ROW TIES BACK IN ROUND FIGURES (Ninad, 2026-11-13, CLI-132).

HIS WORDS: "for pre seed, 1M (raise) + 1M (revenue) + 0M (cash) - 0.5M (assets) - 0.5M (profits) - 1M (spend) = 0. 1. This is what I mean by rounding.. 2. move 0.5M assets to cash, we cant afford investing 0.5 in asset in pre-seed".

The plan's figures are recorded round, to the round table's units; real figures tie back to them

Ruled 2026-11-13 by ninad -- in force

THE PLAN'S FIGURES ARE RECORDED ROUND, TO THE ROUND TABLE'S UNITS; REAL FIGURES TIE BACK TO THEM (Ninad, 2026-11-13, CLI-132; refines #830).

HIS WORDS: "the values dont match my values.. the rounding is essential" -- his values: pre-seed raise 1M, revenue 1M, spend 1M, profit -0.5M, cash 0 then 0.5M; 2028/29 capital 1M, compliance and other 1M, people, machines and seats 4M.

The staffing mix is derived from the revenue and client targets through prospects, on recorded assumptions; every plan figure has a rounded twin for display

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

THE STAFFING MIX IS DERIVED FROM THE REVENUE AND CLIENT TARGETS THROUGH PROSPECTS, ON RECORDED ASSUMPTIONS; EVERY PLAN FIGURE HAS A ROUNDED TWIN FOR DISPLAY (Ninad, 2026-11-13, CLI-132).

ASKED: recommended a pattern for the sales budget and staffing driven from the revenue targets through prospects (new clients, prospects, FDE capacity, SMEs by industry, operators by service, a feasibility loop back to the targets per #826); shown the mix of employees per year it gives, then the client counts and revenue per client at an 80:20 FDE split, then the share of FDEs on prospects each year that lands the approved client counts, then the mix on a smoothed schedule, then the FDE-on-projects count by client against by project and a sales-staff role. HIS WORDS: "should the sales budget be driven by the count of prospects, which is driven by the revenue targets, with appropriate assumptions?"; "use this pattern to come up with the recommended mix of employees for each year"; "I also need the mix of FDE on project vs prospects.. this should be 80:20. Also, 1 FDE can work on 10 prospects max.. this will change the number of prospects, which will change the conversion assumption"; "conversion assumption should be 20 percent, revisit the count of clients, and adjust revenue per client accordingly"; "how can this gap be bridged? will it help if i increase the FDE on prospects to maybe 50 percent?"; "give me a recommended FDEs on prospects percent each year to reach the planned revenue and client count"; "this makes sense to me, because the percent decreases as we move ahead, eventually stabilizing on 15 percent.. plan the resource mix on this basis, ensuring that there is a smooth decrease of FDEs on prospect percent year over year"; "are count of FDEs on projects counted correctly? what assumption is followed there? I like the recommended numbers better, but I dont see any sales staff (who are responsible for generating leads)"; "1. document all the assumptions 2. ensure that the round table and customer mix tables are aligned to the assumptions 3. ensure all figures have a rounded twin for display purposes".

READ AS (Orion's reading, not the administrator's words) -- the assumptions, each changeable on its own: (1) BROADER GOALS: each year's customers by group, year-end revenue rate and raises stand as approved (#782, #786, #770-#779); the staffing, prospects and sales budget are details derived from them, and the loop goes back to the goals only when the risk is high (#826). (2) NEW CLIENTS: a year's new clients are its customers less the year before's; every new client starts as a micro client (#786) and moves up one group a year at most. (3) CONVERSION: 20 percent of the prospects worked in a year convert within that year (replacing one in four, #793's reading); so prospects worked are five for every new client. (4) PROSPECTS PER FDE: one FDE works at most 10 prospects at a time, so FDEs on prospects are new clients divided by two, and prospects worked are ten times the FDEs on prospects. (5) FDEs ON PROJECTS: counted by client at the year end -- one FDE for every four micro or small clients, every two medium clients, or one large client (bronze, silver, gold); this is the count behind the recommended numbers he chose, and it stands in place of #706's portfolio weights (half, quarter, tenth) for the head count; counting by project under #684's ramp (a large client adding a project a year to four) gives up to twice the FDEs from 2030/31 and a share settling at 10 percent -- shown and not chosen. (6) SHARE ON PROSPECTS: the FDEs on prospects as a share of all FDEs falls smoothly year on year, 65 percent in 2027/28, then 60, 45, 45, 35, 30, 25, 25, 20 and 15 percent in 2036/37, which lands the approved client counts within 8 percent every year; an FDE stays with the project a prospect becomes (#790). (7) SALES STAFF: lead generation is a role of its own, one person for every 100 prospects worked in the year (about two sourced a week) FDEs on prospects do the proposal, prototyping and closing. (9) OPERATORS: services 22 in 2028/29 and three more a year; one service operator per service; one operations staff member per service a year later (#755); three estate operators from 2028/29. (10) SMEs: one per domain the clients' projects need (#706's domain count). (11) ADMIN AND SUPPORT: admin staff one per fifteen heads (#687's ratio); support functions (product, engineering, finance, HR, legal) ten percent on top. (12) THE LEAN YEAR 2027/28 stays as ruled (#791, #793): ten people, 12 clients, 40 prospects. (13) ALIGNMENT: the customer mix table and the round table are rebuilt from these assumptions -- the mix's prospects column and the plan's employees by role are derived figures, recorded in Clara's record through her registrar, and the round table's prospects, customers and employees columns follow them. (14) ROUNDED TWINS: every figure the plan records keeps its real value and shows a rounded twin -- money at #832's units; heads, clients and prospects to the nearest 5 below 100, the nearest 10 below 1,000 and the nearest 50 above (a figure below 20 shown whole); shares to whole percent -- on every Fundraising tab and in every composer, the real value tying back behind it (#830, #832).

Ended because: Ninad: the reconciliation must be true both ways each year; fix without amending any assumption -- the share of FDEs on prospects is the result of the tie, not a rounded schedule

The staffing mix is derived from the revenue and client targets through prospects, on recorded assumptions; every plan figure has a rounded twin for display

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

THE STAFFING MIX IS DERIVED FROM THE REVENUE AND CLIENT TARGETS THROUGH PROSPECTS, ON RECORDED ASSUMPTIONS; EVERY PLAN FIGURE HAS A ROUNDED TWIN FOR DISPLAY (Ninad, 2026-11-13, CLI-132).

ASKED: recommended a pattern for the sales budget and staffing driven from the revenue targets through prospects (new clients, prospects, FDE capacity, SMEs by industry, operators by service, a feasibility loop back to the targets per #826); shown the mix of employees per year it gives, then the client counts and revenue per client at an 80:20 FDE split, then the share of FDEs on prospects each year that lands the approved client counts, then the mix on a smoothed schedule, then the FDE-on-projects count by client against by project and a sales-staff role. HIS WORDS: "should the sales budget be driven by the count of prospects, which is driven by the revenue targets, with appropriate assumptions?"; "use this pattern to come up with the recommended mix of employees for each year"; "I also need the mix of FDE on project vs prospects.. this should be 80:20. Also, 1 FDE can work on 10 prospects max.. this will change the number of prospects, which will change the conversion assumption"; "conversion assumption should be 20 percent, revisit the count of clients, and adjust revenue per client accordingly"; "how can this gap be bridged? will it help if i increase the FDE on prospects to maybe 50 percent?"; "give me a recommended FDEs on prospects percent each year to reach the planned revenue and client count"; "this makes sense to me, because the percent decreases as we move ahead, eventually stabilizing on 15 percent.. plan the resource mix on this basis, ensuring that there is a smooth decrease of FDEs on prospect percent year over year"; "are count of FDEs on projects counted correctly? what assumption is followed there? I like the recommended numbers better, but I dont see any sales staff (who are responsible for generating leads)"; "1. document all the assumptions 2. ensure that the round table and customer mix tables are aligned to the assumptions 3. ensure all figures have a rounded twin for display purposes"; "the reconciliation should be true both ways each year 1. the number prospects each FDE works in a year 2. the number of customer targets and revenue targets. is that the case?"; "if not, fix it, dont ammend any assumptions".

READ AS (Orion's reading, not the administrator's words) -- the assumptions, each changeable on its own: (1) BROADER GOALS: each year's customers by group, year-end revenue rate and raises stand as approved (#782, #786, #770-#779); the staffing, prospects and sales budget are details derived from them, and the loop goes back to the goals only when the risk is high (#826). (2) NEW CLIENTS: a year's new clients are its customers less the year before's; every new client starts as a micro client (#786) and moves up one group a year at most. (3) CONVERSION: 20 percent of the prospects worked in a year convert within that year (replacing one in four, #793's reading); so prospects worked are five for every new client. (4) PROSPECTS PER FDE: one FDE works at most 10 prospects at a time, so FDEs on prospects are new clients divided by two, and prospects worked are ten times the FDEs on prospects. (5) FDEs ON PROJECTS: counted by client at the year end -- one FDE for every four micro or small clients, every two medium clients, or one large client (bronze, silver, gold); this is the count behind the recommended numbers he chose, and it stands in place of #706's portfolio weights (half, quarter, tenth) for the head count; counting by project under #684's ramp (a large client adding a project a year to four) gives up to twice the FDEs from 2030/31 and a share settling at 10 percent -- shown and not chosen. (6) SHARE ON PROSPECTS: the tie holds both ways every year -- leads worked are five times the year's new clients (20 percent convert), FDEs on prospects are the leads divided by ten, and the share of FDEs on prospects is the RESULT: 67 percent in 2027/28, then 61, 44, 44, 36, 31, 26, 22, 18 and 15 percent in 2036/37, falling smoothly and settling at 15 percent; an FDE stays with the project a prospect becomes (#790). Corrected the same day (Ninad: "the reconciliation should be true both ways each year ... if not, fix it, dont ammend any assumptions"): the first reading carried a rounded schedule (65, 60, 45, 45, 35, 30, 25, 25, 20, 15) that landed the client counts only within 8 percent. (7) SALES STAFF: lead generation is a role of its own, one person for every 100 prospects worked in the year (about two sourced a week) FDEs on prospects do the proposal, prototyping and closing. (9) OPERATORS: services 22 in 2028/29 and three more a year; one service operator per service; one operations staff member per service a year later (#755); three estate operators from 2028/29. (10) SMEs: one per domain the clients' projects need (#706's domain count). (11) ADMIN AND SUPPORT: admin staff one per fifteen heads (#687's ratio); support functions (product, engineering, finance, HR, legal) ten percent on top. (12) THE LEAN YEAR 2027/28 stays as ruled (#791, #793): ten people, 12 clients, 40 prospects. (13) ALIGNMENT: the customer mix table and the round table are rebuilt from these assumptions -- the mix's prospects column and the plan's employees by role are derived figures, recorded in Clara's record through her registrar, and the round table's prospects, customers and employees columns follow them. (14) ROUNDED TWINS: every figure the plan records keeps its real value and shows a rounded twin -- money at #832's units; heads, clients and prospects to the nearest 5 below 100, the nearest 10 below 1,000 and the nearest 50 above (a figure below 20 shown whole); shares to whole percent -- on every Fundraising tab and in every composer, the real value tying back behind it (#830, #832).

Ended because: Ninad: admin and support overlap with the operations staff; no admin, support staff about 50 at the 2036/37 peak reporting to ops as needed

The staffing mix is derived from the revenue and client targets through prospects, on recorded assumptions; every plan figure has a rounded twin for display

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

THE STAFFING MIX IS DERIVED FROM THE REVENUE AND CLIENT TARGETS THROUGH PROSPECTS, ON RECORDED ASSUMPTIONS; EVERY PLAN FIGURE HAS A ROUNDED TWIN FOR DISPLAY (Ninad, 2026-11-13, CLI-132).

ASKED: recommended a pattern for the sales budget and staffing driven from the revenue targets through prospects (new clients, prospects, FDE capacity, SMEs by industry, operators by service, a feasibility loop back to the targets per #826); shown the mix of employees per year it gives, then the client counts and revenue per client at an 80:20 FDE split, then the share of FDEs on prospects each year that lands the approved client counts, then the mix on a smoothed schedule, then the FDE-on-projects count by client against by project and a sales-staff role. HIS WORDS: "should the sales budget be driven by the count of prospects, which is driven by the revenue targets, with appropriate assumptions?"; "use this pattern to come up with the recommended mix of employees for each year"; "I also need the mix of FDE on project vs prospects.. this should be 80:20. Also, 1 FDE can work on 10 prospects max.. this will change the number of prospects, which will change the conversion assumption"; "conversion assumption should be 20 percent, revisit the count of clients, and adjust revenue per client accordingly"; "how can this gap be bridged? will it help if i increase the FDE on prospects to maybe 50 percent?"; "give me a recommended FDEs on prospects percent each year to reach the planned revenue and client count"; "this makes sense to me, because the percent decreases as we move ahead, eventually stabilizing on 15 percent.. plan the resource mix on this basis, ensuring that there is a smooth decrease of FDEs on prospect percent year over year"; "are count of FDEs on projects counted correctly? what assumption is followed there? I like the recommended numbers better, but I dont see any sales staff (who are responsible for generating leads)"; "1. document all the assumptions 2. ensure that the round table and customer mix tables are aligned to the assumptions 3. ensure all figures have a rounded twin for display purposes"; "the reconciliation should be true both ways each year 1. the number prospects each FDE works in a year 2. the number of customer targets and revenue targets. is that the case?"; "if not, fix it, dont ammend any assumptions"; "admin and support overlap with ops staff"; "HR, office, legal are all already in ops staff, how will admin help here?"; "the employees are not rounded.. add about 50 at peak 2036 support staff who will report to ops as needed".

READ AS (Orion's reading, not the administrator's words) -- the assumptions, each changeable on its own: (1) BROADER GOALS: each year's customers by group, year-end revenue rate and raises stand as approved (#782, #786, #770-#779); the staffing, prospects and sales budget are details derived from them, and the loop goes back to the goals only when the risk is high (#826). (2) NEW CLIENTS: a year's new clients are its customers less the year before's; every new client starts as a micro client (#786) and moves up one group a year at most. (3) CONVERSION: 20 percent of the prospects worked in a year convert within that year (replacing one in four, #793's reading); so prospects worked are five for every new client. (4) PROSPECTS PER FDE: one FDE works at most 10 prospects at a time, so FDEs on prospects are new clients divided by two, and prospects worked are ten times the FDEs on prospects. (5) FDEs ON PROJECTS: counted by client at the year end -- one FDE for every four micro or small clients, every two medium clients, or one large client (bronze, silver, gold); this is the count behind the recommended numbers he chose, and it stands in place of #706's portfolio weights (half, quarter, tenth) for the head count; counting by project under #684's ramp (a large client adding a project a year to four) gives up to twice the FDEs from 2030/31 and a share settling at 10 percent -- shown and not chosen. (6) SHARE ON PROSPECTS: the tie holds both ways every year -- leads worked are five times the year's new clients (20 percent convert), FDEs on prospects are the leads divided by ten, and the share of FDEs on prospects is the RESULT: 67 percent in 2027/28, then 61, 44, 44, 36, 31, 26, 22, 18 and 15 percent in 2036/37, falling smoothly and settling at 15 percent; an FDE stays with the project a prospect becomes (#790). Corrected the same day (Ninad: "the reconciliation should be true both ways each year ... if not, fix it, dont ammend any assumptions"): the first reading carried a rounded schedule (65, 60, 45, 45, 35, 30, 25, 25, 20, 15) that landed the client counts only within 8 percent. (7) SALES STAFF: lead generation is a role of its own, one person for every 100 prospects worked in the year (about two sourced a week) FDEs on prospects do the proposal, prototyping and closing. (9) OPERATORS: services 22 in 2028/29 and three more a year; one service operator per service; one operations staff member per service a year later (#755); three estate operators from 2028/29. (10) SMEs: one per domain the clients' projects need (#706's domain count). (11) SUPPORT STAFF, NO ADMIN: HR, office, legal and accounting are services in the catalogue run by their operations staff (#755), so no admin staff are counted (#687's one per fifteen retired for the head count); support staff, who report to the operations staff as needed, are about 5 percent of the other heads from 2028/29 -- about 50 at the 2036/37 peak (4, 6, 10, 14, 18, 24, 31, 38, 47). Corrected the same day (Ninad: "admin and support overlap with ops staff", "HR, office, legal are all already in ops staff, how will admin help here?", "add about 50 at peak 2036 support staff who will report to ops as needed"); the first reading carried admin one per fifteen and support ten percent. (12) THE LEAN YEAR 2027/28 stays as ruled (#791, #793): ten people, 12 clients, 40 prospects. (13) ALIGNMENT: the customer mix table and the round table are rebuilt from these assumptions -- the mix's prospects column and the plan's employees by role are derived figures, recorded in Clara's record through her registrar, and the round table's prospects, customers and employees columns follow them. (14) ROUNDED TWINS: every figure the plan records keeps its real value and shows a rounded twin -- money at #832's units; heads, clients and prospects to the nearest 5 below 100, the nearest 10 below 1,000 and the nearest 50 above (a figure below 20 shown whole); shares to whole percent -- on every Fundraising tab and in every composer, the real value tying back behind it (#830, #832).

Ended because: Ninad: employee count should be more rounded -- count units 5/10/50/100 by size

The staffing mix is derived from the revenue and client targets through prospects, on recorded assumptions; every plan figure has a rounded twin for display

Ruled 2026-11-13 by ninad -- in force

THE STAFFING MIX IS DERIVED FROM THE REVENUE AND CLIENT TARGETS THROUGH PROSPECTS, ON RECORDED ASSUMPTIONS; EVERY PLAN FIGURE HAS A ROUNDED TWIN FOR DISPLAY (Ninad, 2026-11-13, CLI-132).

ASKED: recommended a pattern for the sales budget and staffing driven from the revenue targets through prospects (new clients, prospects, FDE capacity, SMEs by industry, operators by service, a feasibility loop back to the targets per #826); shown the mix of employees per year it gives, then the client counts and revenue per client at an 80:20 FDE split, then the share of FDEs on prospects each year that lands the approved client counts, then the mix on a smoothed schedule, then the FDE-on-projects count by client against by project and a sales-staff role. HIS WORDS: "should the sales budget be driven by the count of prospects, which is driven by the revenue targets, with appropriate assumptions?"; "use this pattern to come up with the recommended mix of employees for each year"; "I also need the mix of FDE on project vs prospects.. this should be 80:20. Also, 1 FDE can work on 10 prospects max.. this will change the number of prospects, which will change the conversion assumption"; "conversion assumption should be 20 percent, revisit the count of clients, and adjust revenue per client accordingly"; "how can this gap be bridged? will it help if i increase the FDE on prospects to maybe 50 percent?"; "give me a recommended FDEs on prospects percent each year to reach the planned revenue and client count"; "this makes sense to me, because the percent decreases as we move ahead, eventually stabilizing on 15 percent.. plan the resource mix on this basis, ensuring that there is a smooth decrease of FDEs on prospect percent year over year"; "are count of FDEs on projects counted correctly? what assumption is followed there? I like the recommended numbers better, but I dont see any sales staff (who are responsible for generating leads)"; "1. document all the assumptions 2. ensure that the round table and customer mix tables are aligned to the assumptions 3. ensure all figures have a rounded twin for display purposes"; "the reconciliation should be true both ways each year 1. the number prospects each FDE works in a year 2. the number of customer targets and revenue targets. is that the case?"; "if not, fix it, dont ammend any assumptions"; "admin and support overlap with ops staff"; "HR, office, legal are all already in ops staff, how will admin help here?"; "the employees are not rounded.. add about 50 at peak 2036 support staff who will report to ops as needed"; "employee count should be more rounded".

READ AS (Orion's reading, not the administrator's words) -- the assumptions, each changeable on its own: (1) BROADER GOALS: each year's customers by group, year-end revenue rate and raises stand as approved (#782, #786, #770-#779); the staffing, prospects and sales budget are details derived from them, and the loop goes back to the goals only when the risk is high (#826). (2) NEW CLIENTS: a year's new clients are its customers less the year before's; every new client starts as a micro client (#786) and moves up one group a year at most. (3) CONVERSION: 20 percent of the prospects worked in a year convert within that year (replacing one in four, #793's reading); so prospects worked are five for every new client. (4) PROSPECTS PER FDE: one FDE works at most 10 prospects at a time, so FDEs on prospects are new clients divided by two, and prospects worked are ten times the FDEs on prospects. (5) FDEs ON PROJECTS: counted by client at the year end -- one FDE for every four micro or small clients, every two medium clients, or one large client (bronze, silver, gold); this is the count behind the recommended numbers he chose, and it stands in place of #706's portfolio weights (half, quarter, tenth) for the head count; counting by project under #684's ramp (a large client adding a project a year to four) gives up to twice the FDEs from 2030/31 and a share settling at 10 percent -- shown and not chosen. (6) SHARE ON PROSPECTS: the tie holds both ways every year -- leads worked are five times the year's new clients (20 percent convert), FDEs on prospects are the leads divided by ten, and the share of FDEs on prospects is the RESULT: 67 percent in 2027/28, then 61, 44, 44, 36, 31, 26, 22, 18 and 15 percent in 2036/37, falling smoothly and settling at 15 percent; an FDE stays with the project a prospect becomes (#790). Corrected the same day (Ninad: "the reconciliation should be true both ways each year ... if not, fix it, dont ammend any assumptions"): the first reading carried a rounded schedule (65, 60, 45, 45, 35, 30, 25, 25, 20, 15) that landed the client counts only within 8 percent. (7) SALES STAFF: lead generation is a role of its own, one person for every 100 prospects worked in the year (about two sourced a week) FDEs on prospects do the proposal, prototyping and closing. (9) OPERATORS: services 22 in 2028/29 and three more a year; one service operator per service; one operations staff member per service a year later (#755); three estate operators from 2028/29. (10) SMEs: one per domain the clients' projects need (#706's domain count). (11) SUPPORT STAFF, NO ADMIN: HR, office, legal and accounting are services in the catalogue run by their operations staff (#755), so no admin staff are counted (#687's one per fifteen retired for the head count); support staff, who report to the operations staff as needed, are about 5 percent of the other heads from 2028/29 -- about 50 at the 2036/37 peak (4, 6, 10, 14, 18, 24, 31, 38, 47). Corrected the same day (Ninad: "admin and support overlap with ops staff", "HR, office, legal are all already in ops staff, how will admin help here?", "add about 50 at peak 2036 support staff who will report to ops as needed"); the first reading carried admin one per fifteen and support ten percent. (12) THE LEAN YEAR 2027/28 stays as ruled (#791, #793): ten people, 12 clients, 40 prospects. (13) ALIGNMENT: the customer mix table and the round table are rebuilt from these assumptions -- the mix's prospects column and the plan's employees by role are derived figures, recorded in Clara's record through her registrar, and the round table's prospects, customers and employees columns follow them. (14) ROUNDED TWINS: every figure the plan records keeps its real value and shows a rounded twin -- money at #832's units; heads, clients and prospects to the nearest 5 below 50, the nearest 10 below 200, the nearest 50 below 1,000 and the nearest 100 above (corrected the same day, Ninad: "employee count should be more rounded"; the first reading had 5 below 100, 10 below 1,000, 50 above); shares to whole percent -- on every Fundraising tab and in every composer, the real value tying back behind it (#830, #832).

The plan's next version: prospects carry the first three years, self-serve clients drive an exponential rise from 2030/31, clients pay about 1 percent of their revenue, and feasibility is a recorded figure

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

THE PLAN'S NEXT VERSION: PROSPECTS CARRY THE FIRST THREE YEARS, SELF-SERVE CLIENTS DRIVE AN EXPONENTIAL RISE FROM 2030/31, CLIENTS PAY ABOUT 1 PERCENT OF THEIR REVENUE, AND FEASIBILITY IS A RECORDED FIGURE (Ninad, 2026-11-13/14, CLI-132).

ASKED: shown the feasibility of the approved plan by year (2028/29 at 15 percent: 12 to 55 clients and 40 heads where 80 are needed), then eased versions, the raise and revenue before and after, the client values at 2-3 and at 1 percent of a client's revenue, the impact of 1 percent on the mix, three options and the combined version with its three tables, pre-seed worth and feasibility. HIS WORDS: "I agree with your analysis, the progress is linear, which puts stress on the initial years.. create a new version of round table where you ease out the revenue targets in the initial few years and possibly fundraise plan (to justify the easing of revenue) so that the increase is slightly exponential, so that the initiatial years are not stretched"; "raise should decrease if revenue decreases, how is it increasing?"; "starting 1M is ambitious, make it 0.5M, accordingly change the seed"; "recompute the 3 tables again by adjusting the revenue per customer assumption so that the count of customers at 1B revenue is 1000, instead of 1400"; "can micro businesses afford 75K per year?"; "what can a 0.5M revenue customer afford?"; "suggest realistic spend"; "bring it to 1%"; "I like this better, recompute all 3 tables, and tell me what is the impact of this on the plan? recommend 3 options to make the plan look like current"; "option 1 is the least feasible"; "option 2 is also feasible. We are not dependent entirely on prospects to convert, once the product becomes known, anybody can start using it"; "ensure that we rely on prospects for the first 2 to 3 years, then plan an exponential rise.. include the 3 tables and feasibility as first class registry items based on this model".

(5) STAFFING follows #836 with the services ramp 4, 8, 16, 22 in 2028/29 to 2031/32 and three more a year after (operations staff a year behind): about 10, 25, 45, 50, 80, 110, 170, 250, 400, 550 heads. (6) ASSETS: the property's first half is bought in 2030/31 (Series B's year), the second in 2033/34. (8) FEASIBILITY is a recorded figure per year: the fit of needed heads within the approved count, times a growth factor (the square root of 1.5 over the year's revenue growth, capped at one), times the same for the hiring ramp, times one less half the self-serve share of revenue; rounded to 5 percent: 60, 65, 60, 60, 65, 70, 75, 80, 85, 95 percent for 2027/28 to 2036/37. (9) RECORDS: the round table, the customer mix, the employee mix and the feasibility are first-class records in Clara's registry derived from this version's goals and #836's assumptions, each figure with its rounded twin (#836 point 14), and the departures check names any table that drifts from them.

The plan's next version: prospects carry the first three years, self-serve clients drive an exponential rise from 2030/31, clients pay about 1 percent of their revenue, and feasibility is a recorded figure

Ruled 2026-11-13 by ninad -- superseded 2026-11-13

THE PLAN'S NEXT VERSION: PROSPECTS CARRY THE FIRST THREE YEARS, SELF-SERVE CLIENTS DRIVE AN EXPONENTIAL RISE FROM 2030/31, CLIENTS PAY ABOUT 1 PERCENT OF THEIR REVENUE, AND FEASIBILITY IS A RECORDED FIGURE (Ninad, 2026-11-13/14, CLI-132).

ASKED: shown the feasibility of the approved plan by year (2028/29 at 15 percent: 12 to 55 clients and 40 heads where 80 are needed), then eased versions, the raise and revenue before and after, the client values at 2-3 and at 1 percent of a client's revenue, the impact of 1 percent on the mix, three options and the combined version with its three tables, pre-seed worth and feasibility. HIS WORDS: "I agree with your analysis, the progress is linear, which puts stress on the initial years.. create a new version of round table where you ease out the revenue targets in the initial few years and possibly fundraise plan (to justify the easing of revenue) so that the increase is slightly exponential, so that the initiatial years are not stretched"; "raise should decrease if revenue decreases, how is it increasing?"; "starting 1M is ambitious, make it 0.5M, accordingly change the seed"; "recompute the 3 tables again by adjusting the revenue per customer assumption so that the count of customers at 1B revenue is 1000, instead of 1400"; "can micro businesses afford 75K per year?"; "what can a 0.5M revenue customer afford?"; "suggest realistic spend"; "bring it to 1%"; "I like this better, recompute all 3 tables, and tell me what is the impact of this on the plan? recommend 3 options to make the plan look like current"; "option 1 is the least feasible"; "option 2 is also feasible. We are not dependent entirely on prospects to convert, once the product becomes known, anybody can start using it"; "ensure that we rely on prospects for the first 2 to 3 years, then plan an exponential rise..

(5) STAFFING follows #836 with the services ramp 4, 8, 16, 22 in 2028/29 to 2031/32 and three more a year after (operations staff a year behind): about 10, 25, 45, 50, 80, 110, 170, 250, 400, 550 heads. (6) ASSETS: the property's first half is bought in 2030/31 (Series B's year), the second in 2033/34. (8) FEASIBILITY is a recorded figure per year: the fit of needed heads within the approved count, times a growth factor (the square root of 1.5 over the year's revenue growth, capped at one), times the same for the hiring ramp, times one less half the self-serve share of revenue; rounded to 5 percent: 60, 65, 60, 60, 65, 70, 75, 80, 85, 95 percent for 2027/28 to 2036/37. (9) RECORDS: the round table, the customer mix, the employee mix and the feasibility are first-class records in Clara's registry derived from this version's goals and #836's assumptions, each figure with its rounded twin (#836 point 14), and the departures check names any table that drifts from them.

The planning template serves three levels: a project's business plan with its CEO, each Carolverse service's plan from Clara's business plan, and each project service's plan from its project's business plan

Ruled 2026-11-13 by ninad -- in force

THE PLANNING TEMPLATE SERVES THREE LEVELS: A PROJECT'S BUSINESS PLAN WITH ITS CEO, EACH CAROLVERSE SERVICE'S PLAN FROM CLARA'S BUSINESS PLAN, AND EACH PROJECT SERVICE'S PLAN FROM ITS PROJECT'S BUSINESS PLAN (Ninad, 2026-11-14, CLI-132).

ASKED: told that the planning craft of leadership #826 to #840 is captured in as a reusable service offering for projects that adopt Fundraising. HIS WORDS: "this template not only serves to other projects, but also other services within carolverse.. similarly, other project's services should also adopt their business plan sitting with their CEO to build their own individual service plans".

READ AS (Orion's reading, not the administrator's words): (1) the planning template (goals first, recorded assumptions, the derived tables, sanity checks, feasibility, versions, the deck slide) is one craft in Sage's library bound to Clara, used at three levels; (2) PROJECT LEVEL: every project's business plan is built and kept by that project's own CEO from the project's own recorded inputs (7069), never by Clara by default; (3) CAROLVERSE'S SERVICES: every service in Carolverse's catalogue builds its own service plan from Clara's business plan with the same template -- its share of the revenue, clients, operators and operations staff, spend, the assumptions it inherits and the ones it sets, checked and feasible the same way -- kept by the service's owner; (4) A PROJECT'S SERVICES: every service of another project adopts that project's business plan, as held by the project's CEO, and builds its own service plan from it the same way; (5) the service plans roll up to the business plan they come from, and a departure between a service plan and its business plan is named like any other; (6) the offering, the service-level template and the roll-up are built under 7069 (criterion 1 widened), with the project onboarding (7032, 6792) handing a new project both the business plan and the service plans.

The bridge line is a debt round in the record, and own-project stakes 1 to 3 move to the cash-positive years

Ruled 2026-11-13 by ninad -- in force

THE BRIDGE LINE IS A DEBT ROUND IN THE RECORD, AND OWN-PROJECT STAKES 1 TO 3 MOVE TO THE CASH-POSITIVE YEARS (Ninad, 2026-11-14, CLI-132; refines #840 and #705).

HIS WORDS: "agree to both your points".

The sales spend is planned by each client size's time to win, with clients moving up a size and lost clients replaced, and the yearly budgets are raised to carry it

Ruled 2026-11-14 by ninad -- in force

THE SALES SPEND IS PLANNED BY EACH CLIENT SIZE'S TIME TO WIN, WITH CLIENTS MOVING UP A SIZE AND LOST CLIENTS REPLACED, AND THE YEARLY BUDGETS ARE RAISED TO CARRY IT (Ninad, 2026-11-14, CLI-136).

ASKED: shown the pre-seed year's spend, then a monthly plan of the potential clients chased for this year's and next year's clients, then the times to win by client size as Orion's suggestion (micro 3, small 5, medium 6, bronze 9, silver 12 to 18 months), then the sales spend year by year with clients lost and moving up a size, then the rates for both drawn from published SaaS benchmarks. HIS WORDS: "the chasing for the next set of clients for the next year will have to start in the previous year"; "what is the assumed duration for acquiring a client? that should determine the duration when the sales client acquisition budget will be spent"; "Yes, use those assumptions.. 1. factor this assumption to decide the spend months for each acquisition 2. moving existing client to higher rung will also require sales spend, planned accordingly.. 3. churn will require associated sales spend to backfill the clients.. ensure that the sales spend is appropriately planned"; "raise the plan's yearly budgets"; "are these numbers not already assumed to calculate the count of customers? if not, please include these assumptions.."; "I approve the numbers, but the requirements are for the entire plan, fold them into the pursuit".

READ AS (Orion's reading, not the administrator's words): (1) TIME TO WIN a client, by size: micro 3 months, small 5, medium 6, bronze 9, silver 12, gold 18. (3) MOVING UP A SIZE costs 80 percent of the value it adds (#827), spent over the new size's time to win, ending in the month of the move. (4) CLIENTS LOST each year, each replaced by a new win carrying the full cost of (2): micro 20 percent, small 10, medium, bronze, silver and gold 5 -- the medians of published SaaS logo-churn benchmarks for small-business (15 to 30), mid-market (6 to 15) and enterprise (2 to 5) clients. (5) CLIENTS MOVING UP A SIZE each year: micro 4 percent, small 5, medium, bronze and silver 8 -- worked out from the benchmark revenue kept from existing clients a year on (small business about 97 percent, mid-market 108, enterprise 118) as (revenue kept - 100 percent + the leaving rate) / (the size's step in value - 1). (6) The client counts by size stay the plan's targets; the wins each year are the net change + the clients lost + the clients moved out - the clients moved in. (8) THE YEARLY BUDGETS ARE RAISED to carry this sales spend: each year's budget is rebuilt from its parts -- people, running costs, compliance and the sales spend -- and shown to Ninad with its profit, cash and round sizes before it is recorded (#766, #86). (9) A RISK named for the assumptions register: the same benchmarks put the cost of winning a new client at about 1.6 times its first-year value against the plan's 80 percent (growing a client matches, at about 80 percent of the value added); remedy: measure the first clients' real cost and move the share to it. Folded into pursuit CAROL-PUR-0035 (Investment plan).

Every number in the planning model rests on a documented assumption with its basis, risk and remedy; the formulas that tie the numbers are documented and reconciled, the assumptions re-verified and fact-checked, and each year's feasibility documented and re-evaluated, from time to time

Ruled 2026-11-14 by ninad -- in force

EVERY NUMBER IN THE PLANNING MODEL RESTS ON A DOCUMENTED ASSUMPTION WITH ITS BASIS, RISK AND REMEDY; THE FORMULAS THAT TIE THE NUMBERS ARE DOCUMENTED AND RECONCILED, THE ASSUMPTIONS RE-VERIFIED AND FACT-CHECKED, AND EACH YEAR'S FEASIBILITY DOCUMENTED AND RE-EVALUATED, FROM TIME TO TIME (Ninad, 2026-11-14, CLI-136).

HIS WORDS: "1. no number in the planning model should be arbitrary, every number should be backed by appropriate and logical assumptions that are true to the best of our knowledge. The basis of assumptions should also be documented. Possibility of the assumptions not comming true should be documented (risks) and remediations should also be documented 2. formula that tie back the numbers should be documented to reconcile them periodically. Assumptions should be periodically verified to ensure that they are kept upto date, and should be fact checked from time to time. Feasibility for each year plan should be documented, and evaluated from time to time"; "the requirements are for the entire plan, fold them into the pursuit".

READ AS (Orion's reading, not the administrator's words), for the WHOLE plan in Clara's record and every version of it: (1) an ASSUMPTIONS REGISTER, a first-class record: every number the plan uses names its assumption, the basis and source it rests on, when it was last verified, the risk of it not coming true and the remedy if it does not; a number with no assumption is named by the plan check, never accepted silently; the assumptions today held only in ruling text (#826 to #840 and this session's) are carried into it. (2) A FORMULAS REGISTER: every derived figure names its formula and inputs, and a reconciliation recomputes them and names every figure that does not tie. (3) ON A SCHEDULE, the assumptions are re-verified against the latest facts and their sources fact-checked, a stale or unsupported one named to its owner. (4) FEASIBILITY for each plan year is documented with its basis and factors (#837 point 8) and re-evaluated on the same schedule. Folded into pursuit CAROL-PUR-0035 (Investment plan).

A person's cost is their cost to company -- salary, bonus, incentives, travel, Claude tokens and laptop -- and the plan's cost is reported in four categories with the founders and estate operators spread equally across them

Ruled 2026-11-14 by ninad -- in force

A PERSON'S COST IS THEIR COST TO COMPANY -- SALARY, BONUS, INCENTIVES, TRAVEL, CLAUDE TOKENS AND LAPTOP -- AND THE PLAN'S COST IS REPORTED IN FOUR CATEGORIES WITH THE FOUNDERS AND ESTATE OPERATORS SPREAD EQUALLY ACROSS THEM (Ninad, 2026-11-14, CLI-136/CLI-137).

HIS WORDS: "people CTC = salary + bonus + incentives + travel + claude tokens + laptop"; "founder and estate operator are spread across all three categories - platform, sales and compliance"; "client service as a category is missing"; "founders and estate operators are spread equally across all categories"; "they should be part of CTC, if the plan has kept is seperate, include them in CTC.. its a bucketing change, absolute value shouldnt change..".

(2) The plan's cost is reported in FOUR CATEGORIES: platform; sales and customer acquisition; client service; compliance and other. (3) The founders and the estate operators are spread EQUALLY across the four. (4) The other people fall where their work does: service operators on platform; forward deployed engineers by their share on prospects (#836 point 6: 67 percent in 2027/28) to sales and customer acquisition, the rest to client service. (5) The platform's model use and Azure stay on platform until the plan records how much serves client projects. Each of these is an assumption in the register (leadership #845).

The lean year runs on twelve people with a compliance and an admin person, paid for by serving Indian clients from the Indian company first, and its cost is shown as 15, 65, 10 and 10 percent

Ruled 2026-11-14 by ninad -- in force

THE LEAN YEAR RUNS ON TWELVE PEOPLE WITH A COMPLIANCE AND AN ADMIN PERSON, PAID FOR BY SERVING INDIAN CLIENTS FROM THE INDIAN COMPANY FIRST, AND ITS COST IS SHOWN AS 15, 65, 10 AND 10 PERCENT (Ninad, 2026-11-14, CLI-137).

ASKED: shown the first year's cost in four categories (platform, sales and customer acquisition, client service, compliance and other). HIS WORDS: "platform --> 2, Sales --> 4, client services --> 2, compliance --> 1, admin --> 1, founders --> 2. compliance and admin is missing in the plan.. work out some solution to include them.

This replaces #791's ten for that year. This amends #791's "first clients from anywhere" for the lean year: EU clients, and EU compliance, from the Seed. (3) RISKS for the assumptions register (#845): a German company signing or processing the data brings GDPR back; the founders working from Germany may count as an EU presence -- one legal check before relying on the saving (remedy: keep contracts, data and hosting with the Indian company).

The four cost categories are the lean year's; the categories expand in the years after, each entering in the year its first cost does

Ruled 2026-11-14 by ninad -- in force

THE FOUR COST CATEGORIES ARE THE LEAN YEAR'S; THE CATEGORIES EXPAND IN THE YEARS AFTER, EACH ENTERING IN THE YEAR ITS FIRST COST DOES (Ninad, 2026-11-14, CLI-137).

HIS WORDS: "the categories will continue to expand in the subsequent years".

READ AS (Orion's reading, not the administrator's words): platform; sales and customer acquisition; client service; and compliance and admin are the categories of 2027/28 (#847, #848). Later years add categories as new kinds of work begin, each entering in the year its first cost does, read from the plan's own records and never typed: a new category is a term in the data dictionary (#846) and its split an assumption in the register (#845). Which categories enter when is shown to Ninad with the rebuilt budgets before it is recorded (#766, #86).

Each of the plan's cost categories has one owning agent: Elrond the platform, Aurora sales, Leo client service, Themis compliance and admin, and Clara the investments

Ruled 2026-11-14 by ninad -- in force

EACH OF THE PLAN'S COST CATEGORIES HAS ONE OWNING AGENT: ELROND THE PLATFORM, AURORA SALES, LEO CLIENT SERVICE, THEMIS COMPLIANCE AND ADMIN, AND CLARA THE INVESTMENTS (Ninad, 2026-11-14, CLI-137).

ASKED: "who is incharge of client services?" -- shown that no service or head owns it (Leo's Blueprint builds client projects, Sam handles support, Aurora is Head of Sales); "is there a head of delivery?" -- shown there is none. HIS WORDS: "Elrond --> platform, Leo --> client services, Carol --> sales, Clara --> investments, Rhea / Themis --> compliance & admin?"; shown Themis (Head of Legal & Compliance) as one accountable owner with Rhea supporting admin, and that Carol's title is Sales Agent -- EU while Aurora is Head of Sales: "right, aurora".

READ AS (Orion's reading, not the administrator's words): the cost categories of #847-#849 each have ONE owning agent, accountable for the category's budget and its people: platform -- Elrond (Head of Engineering); sales and customer acquisition -- Aurora (Head of Sales), Carol staying a sales agent; client service -- Leo (Head of Strategy, owner of Blueprint); compliance and admin -- Themis (Head of Legal & Compliance), the admin work running through Rhea's operations; the investments (the plan and the rounds) -- Clara (CEO). A category later years add (#849) is given its owner when it enters.

The bridge line is the safety cushion and every round keeps its standard cash reserve without exception; the plan's guidelines are documented apart from its assumptions

Ruled 2026-11-14 by ninad -- in force

THE BRIDGE LINE IS THE SAFETY CUSHION AND EVERY ROUND KEEPS ITS STANDARD CASH RESERVE WITHOUT EXCEPTION; THE PLAN'S GUIDELINES ARE DOCUMENTED APART FROM ITS ASSUMPTIONS (Ninad, 2026-11-14, CLI-137).

HIS WORDS: "safety cushion - the bridge loan should service as the safety cushion, but no exceptions to cash reserves as per the standard guideline.. all guidelines associated with valuations, margins, revenue, raise, employees, customers etc should be documented.. they should not be confused with assumptions.. definitions of assumptions and guidelines should be in the dictionary..".

(2) Every round keeps its STANDARD CASH RESERVE with NO exception -- the buffer months the plan records (pre-seed 3, Seed 6, Series A 9, Series B 12, Series C 18, IPO 24), and Pre-seed 2, which records none today, is given one under the same guideline; where a round's reserve would fall short, the bridge is drawn early enough to hold it, and where the bridge cannot, the round is re-sized for his approval (#679). On version E today the pre-seed has 0.8 months when Pre-seed 2 lands, and the Seed and Series A also fall short: the rebuilt budgets cure each. (3) GUIDELINES -- the standards the plan keeps: cash reserves per round, valuation against revenue, margins, revenue per person, the size of a raise against the business, employee ramps, customer growth and the rest -- are documented in a register of their own, each with its value, reason, source and the rounds or years it binds; they are never mixed with ASSUMPTIONS, which are what the plan believes will happen (churn, conversion, times to win and the like, #845). (4) The data dictionary defines 'assumption' and 'guideline' and the difference between them (#846).

A person's pay follows where they work; lean-period pay holds only in the lean period, which the dictionary defines and a guideline bounds

Ruled 2026-11-14 by ninad -- superseded 2026-11-14

A PERSON'S PAY FOLLOWS WHERE THEY WORK; LEAN-PERIOD PAY HOLDS ONLY IN THE LEAN PERIOD, WHICH THE DICTIONARY DEFINES AND A GUIDELINE BOUNDS (Ninad, 2026-11-14, CLI-137).

HIS WORDS: "Ops pay - this depends more on where the resources are, less on salary.. the lean period salaries should be confined to lean period only.. there should be a definition of lean period, and guidelines on how long does lean period continue".

(3) 'Lean period' is defined in the data dictionary, and a guideline bounds how long it lasts; the bound is set by Ninad before it is recorded (#86).

The 2028/29 customer mix spreads its revenue over two medium and nine small clients, and feasibility is judged with parameters that change with the year of execution

Ruled 2026-11-15 by ninad -- in force

THE 2028/29 CUSTOMER MIX SPREADS ITS REVENUE OVER TWO MEDIUM AND NINE SMALL CLIENTS, AND FEASIBILITY IS JUDGED WITH PARAMETERS THAT CHANGE WITH THE YEAR OF EXECUTION (Ninad, 2026-11-15, CLI-150; refines #837 point 8 and #840 points 4, 11 and 12).

HIS WORDS: "why is the feasibility so low? product is ready, value is established, team is trained"; "i'd go with your recommendation, the feasibility model should change slightly based on the year of execution, because parameters change..".

Every other year of version E stands. Each year's parameters are assumptions in the register with basis, risk and remedy (#845); the result is recorded per year, rounded to 5 percent, and shown with its factors. (4) The figures computed under (2) are shown to the administrator before they are recorded (#766).

Clara serves investors herself: the whole plan, honest gaps noted and asked of Orion, feedback recorded, the investment apps hers, investors subscribed to the investment service

Ruled 2026-11-15 by ninad -- in force

CLARA SERVES INVESTORS HERSELF: SHE ANSWERS FROM THE WHOLE PLAN, IS HONEST ABOUT WHAT SHE CANNOT ANSWER, NOTES EACH INVESTOR'S QUESTIONS AND FEEDBACK, ASKS ORION AND COMES BACK, OWNS THE INVESTMENT APPS AND SUBSCRIBES INVESTORS TO THE INVESTMENT SERVICE (Ninad, 2026-11-15, CLI-150).

ASKED: shown that Clara does not yet explain the plan's figures (feasibility among them) in investor conversations, and that CAROL-PUR-0028 (Have Clara raise the company's investment, talking to investors herself) holds that work. HIS WORDS: "is that pursuit (0028) part of the investment epic? if not, add it. Clara should be able open to ask orion questions that she was not able to answer, take feedback from investors and record them against each investor, and inform orion of the feedback, get back to investors on open topics after getting clarifications from orion, be honest to investors on topics that she cannot answer, should be able to note down questions from the investors so that she can get back to them later with answers that she is not aware of, should be able to give investors all information about the plan that currently exist, she should own all the apps associated with investments, and clara should be able to subscribe investors to investment service similar to how leo subscribes users to blueprint. Update the pursuit accordingly..".

READ AS (Orion's reading, not the administrator's words): (1) CAROL-PUR-0028 serves the work epic Investment chain (CAROL-EPIC-0001). (2) ANSWERS FROM THE WHOLE PLAN: Clara can give an investor every piece of the plan that exists -- the round table, the yearly plan, the customer mix, the staffing, the feasibility and its parameters, the assumptions and the rulings behind them -- read live from her record, never a copy. (3) HONEST WHEN SHE CANNOT ANSWER: she says so, never guesses, and NOTES THE QUESTION on that investor's record as open. (4) SHE ASKS ORION: an open question goes to Orion (the investor's operator's Orion) as a question asked of him; his clarification is recorded on the question, and Clara GETS BACK to the investor with it and closes the question. (5) FEEDBACK: what an investor says about the plan or the company is recorded against that investor and Orion is told. (6) Clara OWNS every app of the investment chain (Fundraising, Investor Onboarding and the rest), as an owner owns its apps. (7) Clara SUBSCRIBES an investor to the investment service, as Leo subscribes a user to Blueprint: the subscription is the investor's standing with the service, recorded, and what the investor may see follows it.

Each number in the plan is backed by what fits it: a guideline, an assumption, a formula or a pattern carried forward

Ruled 2026-11-15 by ninad -- in force

EACH NUMBER IN CLARA'S PLAN IS BACKED BY WHAT FITS IT -- A GUIDELINE, AN ASSUMPTION, A DERIVATION BY FORMULA, OR AN EXTRAPOLATION OF A PATTERN; NOT EVERY NUMBER NEEDS AN ASSUMPTION (Ninad, 2026-11-16, CLI-165).

ASKED: 's plan check lists 40 plan numbers no assumption backs. Should the approved goals (round table, client counts, raises, valuations) each get an assumption, or stand as goals backed by the guidelines register? HIS WORDS: "not all numbers need assumptions, some are guided by guidelines, others are derived by formula, certain others are extrapolation of a pattern, so it really depends on the number whether it should be associated with a guideline, assumption or derivation".

READ AS (Orion's reading, not the administrator's words): (1) every plan number names ONE kind of backing -- guideline (the guidelines register, 7240), assumption (the assumptions register, 7236), derivation (a formula over other plan numbers) or extrapolation (a pattern carried forward from recorded figures) -- chosen per number, never one kind for all; (2) the plan check's 'unbacked' list counts a number as backed when it names any one of the four, so the 40 are each placed rather than each given an assumption; (3) a number backed by derivation or extrapolation names its formula or its pattern and the figures it reads.

Revenue per person, the raise against the business and valuation against revenue are driven by ambition

Ruled 2026-11-15 by ninad -- in force

REVENUE PER PERSON, THE RAISE AGAINST THE BUSINESS AND VALUATION AGAINST REVENUE ARE DRIVEN BY AMBITION, AND AMBITION IS A DRIVER OF THE PLAN'S NUMBERS BESIDE ASSUMPTIONS, GUIDELINES, FORMULAS AND PATTERNS (Ninad, 2026-11-16, CLI-165).

HIS WORDS: "these numbers are driven from the ambition, which is also a driving factor behind the numbers apart from assumptions, guidelines, formula and patterns".

How each of the 40 unbacked plan numbers is backed

Ruled 2026-11-15 by ninad -- in force

EACH OF THE 40 PLAN NUMBERS THE PLAN CHECK LISTED AS UNBACKED NAMES THE KIND OF BACKING ORION PROPOSED.

ASKED: under #892 and #901 each plan number names one kind of backing -- guideline, assumption, formula, pattern or ambition -- chosen per number. HIS ANSWER: "Approve as proposed".

READ AS (Orion's reading, not the administrator's words): records each placement as written, naming what each formula or pattern reads and the target each ambition comes from; a value that is a plan record's input keeps its one home in that record (cookbook 1126) and is never copied into a register to be called backed.

The spend plan by category is a slice of the plan's backed numbers by function

Ruled 2026-11-15 by ninad -- in force

THE SPEND PLAN BY CATEGORY IS THE PLAN'S ALREADY BACKED NUMBERS SLICED BY FUNCTION.

ASKED: #905 placed the spend plan by category (engineering, sales and marketing, customer success, infrastructure, operations, research, reserve) as a formula read from the spend lines; on building, Orion found it a typed record of 2026-09-22 that nothing works out, and asked whether to retire it as history, back it as ambition, or make it a real formula. HIS WORDS: "that is the categorization of the already backed numbers by functions, just another way to slice and dice, unless I am not understanding the question".

READ AS (Orion's reading, not the administrator's words): its backing is a FORMULA -- the plan's backed numbers (its cost and the raises that pay for it) grouped by function -- never a figure of its own.

The plan's cost is sliced many ways, each for its audience

Ruled 2026-11-15 by ninad -- in force

THE PLAN'S COST IS SLICED MANY WAYS, EACH FOR ITS AUDIENCE -- BY SERVICES, BY DEPARTMENTS, BY THE FOUR CATEGORIES INVESTORS LIKE TO SEE, OR ANY OTHER.

ASKED: the plan already slices its backed cost by function in the four categories of #847-#849 (platform, sales and customer acquisition, client service, compliance and admin); the spend plan by category uses an older list of seven (engineering, sales and marketing, customer success, infrastructure, operations, research, reserve), its typed figures are stale and the Fundraising page does not show it -- should 7333 use the four, keep the seven with a recorded mapping, or retire it with no slice? HIS WORDS: "there can be infinite ways to slice and dice, currently we do the functional slicing by services, by departments (carol's org), by the 4 categories that investors like to see, by any other way depending on who the audience is and so on.."

READ AS (Orion's reading, not the administrator's words): no slice is privileged and none holds figures of its own (#906): each is worked out from the plan's backed numbers when read and adds up to the cost it slices; a slice whose mapping onto the backed numbers is not recorded reads unmeasured, never typed. The spend plan by category's typed figures of 2026-09-22 are retired in place as history, and its seven categories become one more slice, worked out once its mapping is recorded.

A client's invoice is its tokens at each model's price, its services and tax; the plan derives each customer's tokens from its revenue target

Ruled 2026-11-16 by ninad -- in force

A CLIENT'S INVOICE IS THE PRICE OF THE TOKENS IT CONSUMED THROUGH EACH MODEL, PLUS ITS SERVICES, PLUS TAX -- A PRICE BEING COST PLUS MARKUP -- AND THE PLAN WORKS THAT FORMULA BACKWARDS FROM ITS REVENUE TARGETS TO EACH CUSTOMER'S TOKENS.

HIS WORDS: (1) "respective price of tokens consumed via each model + services + tax adds up to form the invoice (price includes cost + markup)". (2) "the plan derives cost of tokens consumed from the revenue targets.. using the same formula - cost + margin + services (which is also aggregate service cost based on the services consumed by the customer + margin) + tax, the estimated tokens consumed per customer can be derived. The services cost are decided by service owners, which follow the same concept of cost to the service + margin + tax, which is managed at the service level, there is a seperate pursuit for that already defined.. check for the dependencies".

READ AS (Orion's reading, not the administrator's words): (a) a client's invoice = for each model, the tokens it consumed through that model times that model's price (the model's per-use cost plus a markup) + the services the customer consumed, each at its price (the service's cost plus its owner's margin, #695, managed at service level by its owner) + tax. (b) The plan does not carry a measured month forward as a client project's tokens: it takes each customer's revenue target, sets aside tax and the services that customer consumes at their prices, and what remains is the price of its tokens; dividing by each model's price gives the estimated tokens it consumes, and those tokens at each model's per-use cost are the plan's token cost for it. #734's measured-use pricing of a project's tokens and #683's carried-forward measured use are retired in place for client projects (Carolverse's own use stays measured); per-use cost rates stay read from the rate card with their source and date. (c) Neither answer chose between the two offered options of either question; both are superseded by the formula. (d) The formula depends on services being priced at cost plus margin and on a markup per model being recorded by the service that sells tokens under Midas's cap (#695, #736); until each is recorded the derived tokens and their cost read unmeasured, never a guess. The year's cost then reads the derived token cost; how the typed cost changes is worked out under 7147's revised goal and shown before Clara's registrar records it (#86).

The plan assumes each invoice's tokens, services and tax from a configurable customer mix, and prices build tokens on Claude and run tokens on Astra

Ruled 2026-11-16 by ninad -- in force

THE PLAN ASSUMES THE TOKENS, SERVICES AND TAX OF AN INVOICE FROM THE CUSTOMER MIX -- HALF INDIAN, HALF OUTSIDE INDIA (MOSTLY UK, USA, EUROPE) -- AS REASONABLE, CONFIGURABLE ASSUMPTIONS; AND IT SPLITS TOKENS INTO BUILD (CLAUDE) AND RUN (ASTRA), ESTATE OPERATORS HAVING THE FINAL SAY ON EACH PROJECT'S MODELS.

ASKED (two questions): (1) the tax on a client's invoice is recorded nowhere: 18% GST (India first), a rate by the client's country, or leave tax out for now? (2) the measured month split 95% Claude Opus 5.5, 4% Claude Fable 5.1, 1% Astra (via Codex): use that split for every customer, or Claude Opus only?

HIS WORDS: (1) "the plan should simply \"assume\" the token + services + tax based on the customer mix. The assumptions should be resonable, so that actuals are not too different. In terms of customer mix, assume 50% outside india, mostly western countries (UK, USA, Europe) and 50% Indian customers, but with the caveat that the plan may change as customers evolve.. hence ensure that the assumed valued are configurable, so that the assumptions can be configured to get closer to reality over time as reality unfolds". (2) "this needs to be seggregated into build and run costs.. run costs are astra and build costs are claude for the moment, but in reality, clients can change the run costs based on their model preferences, if they have a preference.. if they have cost or quality concerns, operation models may change.. if clients have a specific preference, those will be used.. for the purpose of planning, claude for build (used mostly by operators via their albus and orion bypass routes) and astra for operations. Estate operators have the final say on the model preferences for each project.. the platform allows multiple models via dedicated slots".

READ AS (Orion's reading, not the administrator's words): (a) the plan does not wait for actual service prices to work out an invoice: it ASSUMES how a customer's invoice divides into tokens, services and tax, from a customer mix of 50% Indian and 50% outside India (mostly UK, USA and Europe); each share, the mix and every rate behind them is an entry of the plan's assumptions register (#845, #833: each changeable on its own, with source, risk and remedy), so it is configured towards actuals as they arrive, never typed into code. (b) A customer's tokens are two parts: BUILD, priced on Claude (operators' Orion and Albus bypass routes), and RUN, priced on Astra; the share of each is itself an assumption. A client's own model preference, or a cost or quality concern, can change its run models; estate operators decide each project's models, through the platform's dedicated slots -- the plan's Claude-for-build, Astra-for-run is the planning default, not a rule for projects. (c) Every assumed value is shown to the administrator before Clara's registrar records it (#86). (d) #912's dependency on no longer holds the plan: the plan assumes the services share, and 7105's actual prices are what the actuals are later compared with.

Dev and ops tokens are worked out per customer through its lifecycle: dev as a prospect, build-led at onboarding, ops with build for changes, and build again when it moves up a size

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

DEV (BUILD) AND OPS (RUN) TOKENS ARE WORKED OUT PER CUSTOMER, FOLLOWING ITS LIFECYCLE.

Record the full set with this split?

HIS WORDS: "dev vs ops should be per customer, based on their lifecycle.. for eg. initiatilly, every propect will only consume dev tokens, no ops, then during the initial 1 or 2 months, the build will dominate, then ops will kick in, and build will continue to exist for change requests etc, untill the customer moves the rung and becomes larger, then build will again kick in, and so on.. this needs to be accounted for.. if this would need a seperate initiative, we can park this activity, build the model first and then resume.. your call".

READ AS (Orion's reading, not the administrator's words): (a) the split is not one share for a year nor tied to FDE head count: each customer's tokens follow its stage -- a prospect consumes dev tokens only; a new client's first one or two months are build-led; then ops tokens start and build continues for change requests; when the client moves up a size (a rung: micro, small, medium, large and its tiers) build rises again; a year's dev and ops tokens are the sum over its customers in each stage. Each stage's figures are configurable assumptions (#913). (b) Orion's call, which he left to Orion: the lifecycle model is built first as its own initiative in the Investment plan pursuit; is parked until it lands and then resumes. (c) The values proposed in CLI-176 (revenue before tax with GST/VAT on top, tokens 35% of the pre-tax invoice, 25% markup, a 50/50 customer mix with India 18% GST and abroad 0%, Claude and Astra at their published rates, ECB 1.1052) are NOT yet approved: they are shown again with the lifecycle model before Clara's registrar records any of them (#86).

Ended because: Ninad simplified the build/ops split to 50/50 (CLI-176)

The lifecycle token model stays simple, its assumptions and formula documented; clients may join at any rung

Ruled 2026-11-16 by ninad -- in force

THE LIFECYCLE TOKEN MODEL STAYS RELATIVELY SIMPLE, WITH ITS ASSUMPTIONS AND FORMULA DOCUMENTED; NOT EVERY CLIENT CLIMBS THE RUNGS -- SOME JOIN AT A HIGHER RUNG DIRECTLY.

Approve these stages (self-serve treated the same way)?

HIS WORDS: "note that not all clients climb the rung, some may join at a higher rung directly.. its hard to guess, so keep the calculations relatively simple, with the assumptions and formula documented".

READ AS (Orion's reading, not the administrator's words): (a) the proposed stages and values stand, kept that simple -- no further refinement by guesswork; (b) a client joining at a higher rung is already a win of that rung in the plan's client flows, so it starts with its onboarding months there and never with a move; (c) every assumption is an entry of the plan's register with its basis, and the formula is written out where the plan is read (its docstring, the answer's own 'formula' text and the assumption entries), so a reader can follow every figure; (d) the values still come back to the administrator with the full set once 7147 resumes, before Clara's registrar records them (#86).

The plan splits a customer's tokens half build, half ops, as one simple assumption

Ruled 2026-11-16 by ninad -- in force

THE PLAN ASSUMES HALF OF A CUSTOMER'S TOKENS ARE BUILD AND HALF ARE OPS -- ONE SIMPLE, CONFIGURABLE ASSUMPTION.

CONTEXT: after #914 (dev and ops worked out per customer through its lifecycle) and #915 (keep it simple, document the assumptions and formula), Orion had filed for the lifecycle model and for the client wins, losses and moves it reads, and was building them. HIS WORDS, unprompted, mid-build: "how about this - go with a simple assumption that 50% is build, 50% is ops".

READ AS (Orion's reading, not the administrator's words): (a) a customer's token cost splits 50% build (priced on Claude, #913) and 50% ops (priced on Astra), every year and every customer, as one entry of the plan's assumptions register (#913: configurable towards actuals); (b) #914's per-customer lifecycle split is superseded, and #915's stages and values (its points a and b) are retired in place with it -- #915's call to keep the calculations simple with the assumptions and formula documented stands; (c) (the lifecycle model) is no longer needed and is set aside; (each year's client wins, losses and moves up a size, worked out once) stays, because the spend sequenced to the customer targets reads it; (d) the full assumption set comes back to the administrator once more before Clara's registrar records it (#86).

The plan's client-token assumptions and the figures they give are approved as shown: token cost 28% of revenue, half build on Claude, half ops on Astra

Ruled 2026-11-16 by ninad -- in force

THE PLAN'S CLIENT-TOKEN ASSUMPTIONS AND THE FIGURES THEY GIVE ARE APPROVED AS SHOWN.

The typed cost already covers each person's own Claude plan and Carolverse's own model use, so client tokens are added on top, not counted twice.

HIS WORDS: "I approve the numbers and the assumptions".

READ AS (Orion's reading, not the administrator's words): (a) Clara's registrar records each value as an entry of the plan's assumptions register with its basis, source and this ruling -- the customer mix, the tax by region, the token share, the markup, the build share and the build and ops prices per million tokens -- each changeable on its own towards actuals (#913); (b) each plan year's cost becomes its typed cost plus the client token cost derived from its revenue target, so profit, margins, the cash path and the round table read it; tax is passed on and is never a cost; (c) the approval covers these values; a later change to any of them is recorded through the registrar and shown again as #86 asks.

The platform's token share of a client's cost rises from 35% to 75% over the plan and the FDEs fall with it: the agents' contribution, shown as a one-pager in the deck

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

THE TOKEN SHARE OF A CLIENT'S COST IS 35% IN THE INITIAL YEARS AND RISES GRADUALLY TO 75%; THE FDEs ARE ADJUSTED TO IT THROUGH THE ASSUMPTIONS SO THE NUMBERS FIT; THIS IS THE PLATFORM'S DIFFERENTIATOR AND GOES IN THE DECK AS A ONE-PAGER (Ninad, 2026-11-16, CLI-183, for CAROL-PUR-0035, after #912, #913, #917).

ASKED: shown, on his formula (revenue = cost + margin; cost = tokens + services), the correlation of token share to FDE share of staff for 2030 -- 25%: 56%, 35%: 52%, 45%: 48%, 55%: 43%, 65%: 37%, 75%: 30% -- with the note that the later years read 90%+ FDEs at every share because the plan's non-FDE staff do not grow with revenue.

HIS WORDS: "I am interested to see how I can increase the token share to, say 35% or even 50% of the cost.. this is because, this is the differentiator of the platform (document this in the business plan of carolverse).. If the platform is doing all the work, token share goes up, and FDE share goes down; if FDE's are doing all the work, platform is not adding value"; then: "in the initial years, token share can be 35%, gradually, it should increase to 75%. Adjust the FDE's accordingly, by adjusting the assumptions, and ensure that the number fit.. this clearly shows the contribution of agents like Leo, Albus etc, who are reducing the FDE costs for our clients.. this should go in as a 1 pager on the slide deck".

READ AS (Orion's reading, not the administrator's words): (a) invoice_token_share becomes a PATH over the plan years -- 35% in the initial years rising gradually to 75% by the last -- each year's value an entry of the assumptions register (the key changes from one figure to a yearly schedule); the path's values come back to the administrator for approval before Clara's registrar records them (#86). (b) A client's cost is its revenue less the margin (#912); the platform's tokens take the token share of it and the FDEs do the rest of the work, so the FDEs on projects are DERIVED from (1 - token share) x the clients' cost / the cost a head, no longer from clients per FDE; the clients-per-FDE assumptions are retired in place for that purpose and the staffing derivation reads the token path. (c) 'The numbers fit': the year's cost (people, other spending, client tokens) and its revenue cohere on the formula, so the approved spend is re-derived from the path rather than typed beside it; the plan check names any year where they do not. (d) The platform's differentiator -- the agents (Leo, Albus and the others) doing the client work the FDEs would otherwise do, measured as the token share of a client's cost and the FDE share of staff falling with it -- is documented in Carolverse's business plan and shown as ONE PAGE in the investor deck, drawn from the plan's figures, never typed on the slide.

Ended because: Ninad set the fitting version: margins constant, FDEs 50%, token share of spend 35% to 75% (CLI-183)

The margins stay as approved with the early losses reported, the FDEs are half the staff, and the token share of spend rises steadily from 35% to 75%

Ruled 2026-11-16 by ninad -- in force

THE PLAN KEEPS ITS MARGINS CONSTANT AND REPORTS LOSSES IN THE INITIAL YEARS; THE FDEs ARE KEPT AT 50% OF THE EMPLOYEE COUNT; THE TOKEN SHARE OF SPEND RISES STEADILY FROM 35% TO 75% (Ninad, 2026-11-16, CLI-183, for CAROL-PUR-0035, after #924 and #917).

ASKED, in turn: (1) the token path cannot be met by cutting FDEs alone on the approved margins (even with no FDEs the rest of the spend leaves 24-48% for tokens): (a) end the path lower, (b) cut the other spending too, or (c) keep the path and lower the later margins; he leaned to (c) and asked for the round table first; (2) shown the path with FDEs falling to 38% of today's count, he asked why the FDE share still climbs and then asked to keep it steady at 50%; (3) shown that version (lowest cash EUR -19.6M, every round to Series B out before the next lands).

HIS WORDS: "I think you should keep the margins constant, and report losses in the initial years.. keep FDE at 50% of the employee count; increase token spend share % from 35% to 75% steadily"; shown the resulting round table (unchanged in raise, valuation, revenue, spend, margin and cash; employees 12 -> 9, 15 -> 20, 25 -> 19, 40 -> 28, 60 -> 34, 140 -> 65, 350 -> 115) and the checklist: "yes, go ahead".

READ AS (Orion's reading, not the administrator's words): (a) each year's spend and margin stay as the round table approved them (#840, #842); the early years' losses are reported as they are, never hidden by a changed target; (b) the token share is a share of each year's SPEND, a yearly path 35, 35, 40, 45, 50, 55, 60, 65, 70, 75 percent for 2027/28 to 2036/37, recorded as a yearly assumption; the tokens sit INSIDE the year's cost, so 7147's reading of them as 28% of revenue added on top (#917's figures) is superseded in place: the approved shares (token share of the invoice, markup, build share, prices a million, mix and tax) stay as recorded and now serve to turn the year's token spend into tokens, not to set its size; (c) the services take the rest of the spend, and the people in them shrink with it, FDEs being HALF of all staff every year (a guideline) -- the FDEs on projects are what is left after the FDEs on prospects, and the clients each handles is DERIVED from that, so the clients-per-FDE assumptions are retired in place; from 2034 the prospects alone need more FDEs than half the staff at 10 prospects each, so prospects per FDE rises to about 20 by 2036/37 (values to approve, #86); (d) the lean year reads 9 people, not twelve: #848's twelve is amended to what the lean year's spend affords at a 35% token share; (e) the token-price risk (cheaper tokens buy more work for the same spend, so prices to clients fall and more clients are needed for the same revenue) is an entry of the assumptions register with its remedy, and Clara puts it to investors; (f) the figures above come back for approval before Clara's registrar records them (#86).

The round table's ambition KPIs are a checklist every plan version is graded against, with each violation named and, where possible, fixed

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

THE ROUND TABLE'S AMBITION KPIs ARE A CHECKLIST, SO EVERY PLAN VERSION IS GRADED AGAINST IT AND ITS EXPECTATION VIOLATIONS ARE NAMED AND, WHERE POSSIBLE, FIXED.

ASKED: the cash at year end had dropped between versions and no single list said what the round table expects.

HIS WORDS: "make a checklist of the ambition KPIs of the round table, so that you can evaluate each version against it.. like revenue expectations, profit expectations, cash balance expectations, valuation expectations, token consumption % expectations, FDE % expectations, founder share expectations, initial investor share expectations etc. etc. this way, you can properly grade each version and inform the expectation violations, and possibly fix them"; then: "also operator expectations - 1 per service, ops staff expectations - 1 per service, etc."; shown the checklist graded for three versions: "yes, go ahead".

READ AS (Orion's reading, not the administrator's words): (a) the checklist is the scope of (the plan's coherence and realism check): one row per expectation, each with its source (the round table, a guideline, an assumption, a ruling), the version's reading, and holds / violated / no expectation recorded; (b) its rows: revenue by round and at the end; profit by round within the recorded tolerances; cash when the next round lands against each round's reserve; cash at year end against the table; valuation by round and at the end; employees at year end against the approved heads; the lean year's people; token share of spend against the path; FDE share of staff (50%); clients per FDE; founders' share after each round; the initial investors' share at the IPO; client project profit; growth and hiring feasibility; service operators one per live service; ops staff one per service in #854's order; estate operators; subject-matter experts one per domain; sales, customer success, admin, compliance and support as their assumptions say; (c) an expectation nobody has set (founders' and initial investors' shares, clients per FDE, the SME domains, the services ramp against the 22 live) reads 'no expectation recorded' by name until the administrator sets it, never a made-up value; (d) the check runs at every step of the investment pursuits and its violations come to the administrator with a proposed fix where the plan's own levers can cure one; (e) a version is graded as a whole, so a change that cures one row and breaks another is named for both.

Ended because: Ninad limited the checklist to ambitions and standards (CLI-183)

The plan's checklist holds only ambitions and standards: the targets the plan must reach and the rules a sound plan keeps

Ruled 2026-11-16 by ninad -- in force

THE CHECKLIST HOLDS ONLY AMBITIONS AND THE STANDARDS OF A SOUND PLAN -- ROUNDED VALUES, ASSUMPTIONS WITHIN THE NORMS, FORMULAS THAT MAKE SENSE, NUMBERS THAT TIE BACK -- NOT EVERY FIGURE THE PLAN CAN WORK OUT.

ASKED: whether important KPIs were missing; Orion listed fourteen more derivable figures (revenue per head, churn, payback, burn, break-even, capital, compliance, tax, prospects, token prices, growth rates).

HIS WORDS: "the checklist should also have fundraising expectations"; then: "checklist should only include ambitions.. like the values should be rounded, the assumptions should adhere to the norms, the formula should make sense, the numbers should tie back together properly, etc..".

READ AS (Orion's reading, not the administrator's words): the checklist has TWO parts and nothing else. PART A, AMBITIONS -- what the plan must reach, each from the round table, the end target, a guideline or a ruling: revenue, profit and margin, cash at year end and the reserve when the next round lands, valuation, raise and pre-money per round, dilution and the founders' and initial investors' shares, customers, employees and the approved heads, the lean year, the token share of spend path, the FDE share of staff, operators and ops staff one per service, subject-matter experts one per domain, client project profit. PART B, STANDARDS -- how a sound plan is built: (1) every shown value is rounded as the table is (#832: round twins, the real figures kept behind); (2) every assumption sits within its norm (the growth and hiring norms, the pay bands, the reserves) and names its source, risk and remedy (#845); (3) every formula makes sense -- its inputs are recorded, its units agree, and a missing input reads unmeasured, never nought; (4) the numbers tie back: the groups' revenue adds up to the year's, the slices to the cost, the people to the heads the cost counts, the projects' shares to the revenue, the cash path to the profit and the rounds, within the recorded tolerances; (5) every plan number names what backs it (#892). A derivable figure with no ambition behind it (revenue per head, burn, payback, break-even) is shown where it helps but is NOT a checklist row. An ambition nobody has set reads 'no expectation recorded' by name.

The founders' and initial investors' shares are those of the last approved round table, and the clients' projects need ten domains of subject-matter experts

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

THE FOUNDERS' AND THE INITIAL INVESTORS' SHARES ARE PICKED FROM THE LAST APPROVED ROUND TABLE; THE NUMBER OF DOMAINS IS TEN.

ASKED: the two checklist ambitions with no target -- the founders' and initial investors' shares (the plan reads founders 100% -> 66.7% after the Seed -> 50% after Series A -> 44% -> 41% -> 36.9% after the IPO; the Pre-seed investors' 10% diluted to about 3.7% by the IPO) and subject-matter experts one per domain (no domain count recorded).

HIS WORDS: "Founders' and initial investors' shares - pick from the last approved round table; Subject-matter experts, one per domain - No of domains --> 10".

READ AS (Orion's reading, not the administrator's words): (a) the ambition for the founders' share after each round is the last approved round table's founders_after column -- 100, 100, 66.69, 50.02, 44.18, 41.03 and 36.92 percent after Pre-seed, Pre-seed 2, Seed, Series A, Series B, Series C and the IPO -- and the initial investors' (the two pre-seed convertibles) 10% and 5% diluted by the same factor, about 3.69% and 1.85% at the IPO; a version departing below them is a violation; (b) the clients' projects need TEN domains, so the plan carries ten subject-matter experts once the clients cover them (today's schedule 0 -> 14 reads over the ten from 2035/36: the schedule is re-recorded to reach ten and the ambition grades it); (c) both are recorded as guidelines the checklist reads, by Clara's registrar.

Ended because: Ninad: at least ten domains, more is fine (CLI-183)

The approved heads are an ambition built from rules -- operators one per service plus the estate operators, ops staff one per agent with the key roles first, FDEs half of all staff -- not a typed cap

Ruled 2026-11-16 by ninad -- in force

THE APPROVED HEADS ARE AN AMBITION, NOT A TYPED CAP: OPERATORS ONE PER SERVICE PLUS THE ESTATE OPERATORS; OPS STAFF IDEALLY ONE PER AGENT, THE KEY ROLES FILLED FIRST; FDEs 50% OF THE TOTAL, WHICH DEPENDS ON THE COUNT OF CLIENTS.

ASKED: what the approved heads are -- Orion explained them as the round table's employee counts kept as a guideline the derived staffing must land within 10% of (12, 15, 25, 40, 60, 90, 140, 250, 350, 550), and offered to re-record them from the FDE-at-50% result.

HIS WORDS: "No, approved heads should be ambitions: operators - 1 per service + estate operators; ops - ideally, 1 per agent, but only key roles to be filled in first; FDEs - 50% of total (this depends on count of clients)".

READ AS (Orion's reading, not the administrator's words): (a) the typed approved_heads guideline is retired in place; the heads the plan allows at a year end are DERIVED from three rules, each an ambition row of the checklist: (1) service operators = one per service the company runs that year (the services ramp, re-recorded against the services live), plus the estate operators (the founders, then the third); (2) ops staff = one per agent in the end, filled in #854's order of key roles first, so a year's ops staff is the key roles reached by then and the ambition reads 'one per agent' only once the roles are all filled; (3) FDEs = half of all staff, their count following the clients (the clients each FDE handles is derived from the client count and that half), as #925 rules; (b) the other roles (subject-matter experts one per domain, ten; sales; customer success; admin; compliance; support) keep their assumptions; (c) the round table's employees column reads these derived heads, never a typed figure; (d) builds the derivation and the checklist (7142) grades it.

The founders' and initial investors' shares are those of the last approved round table, and the plan carries at least ten domains of subject-matter experts, onboarded gradually

Ruled 2026-11-16 by ninad -- in force

THE FOUNDERS' AND THE INITIAL INVESTORS' SHARES ARE PICKED FROM THE LAST APPROVED ROUND TABLE; THE PLAN HAS AT LEAST TEN DOMAINS OF SUBJECT-MATTER EXPERTS, MORE IS FINE, ONBOARDED GRADUALLY.

ASKED: #928 read 'No of domains --> 10' as a cap, so the plan's expert schedule (0, 2, 4, 6, 8, 8, 9, 10, 12, 14 for 2027/28 to 2036/37) was graded failing for its 12 and 14 and a cap from 2034/35 was proposed.

HIS WORDS: "12 in 2035, 14 in 2036 - that is too late!!" then, the schedule explained: "more is ok, but the plan should 'atleast' have 10, onboarding can be gradual, like you have given".

READ AS (Orion's reading, not the administrator's words): (a) the founders' and initial investors' shares stand as #928 read them (the last approved round table's founders_after column; the pre-seed convertibles diluted by the same factor); (b) the experts ambition is a FLOOR: the plan carries at least ten domains by its end, onboarded gradually as the schedule records (reaching ten in 2034/35 and going on to fourteen); a schedule ending below ten is the violation, one above it is not; (c) the schedule as recorded therefore passes and is left as it is.

The checklist checks that every ambition is captured correctly in the plan, never whether an ambition is realistic: feasibility is not its business

Ruled 2026-11-16 by ninad -- in force

THE CHECKLIST IS ABOUT WHETHER ALL AMBITIONS ARE CORRECTLY CAPTURED IN THE PLAN -- NOT ABOUT IDENTIFYING UNREALISTIC AMBITIONS OR FEASIBILITY RISKS.

ASKED: Orion had graded 'assumptions within the norms' as failing because the lean year hires nine people in twelve months (3 -> 12), beyond the hiring norm the feasibility reads.

HIS WORDS: "why is hiring 9 people in 12 months difficult? the checklist is not about identifying unrealistic ambitions or identifying feasibility risks.. its about identifying if all ambitions are correctly captured in the plan..".

READ AS (Orion's reading, not the administrator's words): (a) #927's standard 'every assumption sits within its norm' means each assumption follows the rule or guideline it rests on and names its source, risk and remedy (#845) -- NOT that growth or hiring is within the feasibility norms; the feasibility stays its own reading on its own tab and is no checklist row; (b) a checklist row fails only when the plan does not hold an ambition as ruled (a value missing, different, or not tying back), never because the ambition is bold; (c) the lean year's twelve (#848) is captured as ruled and passes.

The health check and the feasibility check both sit on the Fundraising app for the latest plan, each row red, orange or green

Ruled 2026-11-16 by ninad -- in force

BOLD AMBITIONS ARE THE FEASIBILITY CHECK'S BUSINESS; BOTH THE HEALTH CHECK AND THE FEASIBILITY CHECK ARE ON THE FUNDRAISING APP FOR THE LATEST PLAN, WITH RED, ORANGE AND GREEN INDICATORS.

ASKED (his question): whether the dedicated feasibility check captures bold ambitions and identifies the risks. Orion: it does -- each plan year's chance of meeting its targets from four factors (people allowed against needed, revenue growth against its norm, hiring pace against its norm, the self-serve share) and the factor holding the year back, on its own tab as a percentage; today it reads 'not worked out' because its people factor read the typed approved heads retired by #929.

HIS WORDS: "for capturing bold ambitions, we have the dedicated feasibility check, isnt it? does it identify the risks? both health check and feasibility check should be on the fundraising app for the latest plan, with red, orange, green indicators".

READ AS (Orion's reading, not the administrator's words): (a) the two checks are kept apart: the HEALTH CHECK grades whether every ambition and standard is captured in the plan; the FEASIBILITY CHECK grades how bold the plan is, factor by factor, and names what holds each year back -- that is its risk identification; (b) both are shown on the Fundraising app for the LATEST plan version (the one today's figures are), each row with one indicator: GREEN holds / within the norm, ORANGE a departure within the recorded tolerance or a factor between the twin and the norm, RED a violation or a factor below the twin; the thresholds are recorded guidelines, never typed on the page; (c) a row that cannot be worked out reads grey by name with why, never green.

The round table's profit and employees read the plan, never a typed copy; its raises, valuations and timing stay its own

Ruled 2026-11-16 by ninad -- in force

THE ROUND TABLE READS ITS PROFIT AND EMPLOYEES FROM THE PLAN RATHER THAN HOLDING ITS OWN COPY, SO THE TWO CANNOT DRIFT; THE RAISE, VALUATION AND TIMING STAY THE TABLE'S OWN DECISIONS.

Orion proposed making the table read those two columns from the plan.

HIS WORDS: "implement the fix".

READ AS (Orion's reading, not the administrator's words): (a) the round table's profit and employees columns are DERIVED at every read -- each round's year's profit before tax from the plan and its people from the plan's staffing -- and the typed columns are retired in place as history (kept, never served); (b) the table's own decisions stay typed: the raise, the pre-money, the closing month, the customers and revenue targets; (c) the plan check's departures for profit and employees fall away by construction and the health check's tie-back row grades the remaining typed columns; (d) the feasibility's people factor, which read the retired approved heads (#929), reads the rule-derived heads as what the plan allows, so it is worked out again by name.

Tokens are accounted for in the plan in six buckets -- the platform, the prospects and the projects, each split half build and half operations

Ruled 2026-11-16 by ninad -- in force

THE PLAN ACCOUNTS FOR TOKENS IN SIX BUCKETS -- BUILDING AND RUNNING THE PLATFORM, THE PROSPECTS AND THE PROJECTS -- WITH THE 50:50 BUILD-TO-OPERATIONS SPLIT APPLIED TO EACH (Ninad, 2026-11-17, CLI-183, for CAROL-PUR-0035, after #916, #925).

ASKED (his question): where tokens are accounted for -- building the platform, running the platform, building the prospects, running the prospects, building the projects, running the projects.

HIS WORDS: "use the 50:50 split between build and operations for platform, prospects and projects, and account for each bucket in the plan".

READ AS (Orion's reading, not the administrator's words): (a) the year's token spend (its token share of spend, #925) is divided across THREE uses -- the platform (Carolverse's own build and run), the prospects (the Sandbox prototypes the FDEs build and run for them) and the client projects -- by recorded shares of the assumptions register, each use then half build (priced on Claude) and half operations (priced on Astra, #913, #916); the three shares come back to the administrator for approval before Clara's registrar records them (#86); (b) each of the six buckets is a figure the plan works out and shows -- in the cost slices, the year's drill and Clara's token reader -- never typed; the platform's model-use line (#847) is retired in place in favour of its two buckets, and the prospect budget (#827) names its token part; (c) the six buckets add up to the year's token spend, so the tokens sit inside the cost categories (platform; sales and customer acquisition for the prospects; client service for the projects) and the categories tie back to the year's spend -- the standard the health check grades (#927).

Prospects are charged: half are paid prospects under the Blueprint service, half are investments under Sales; retail sign-ups on the platform are a revenue line, their build and run tokens tracked

Ruled 2026-11-16 by ninad -- in force

THE PLAN CHARGES PROSPECTS: ASSUME HALF ARE PAID PROSPECTS, BUCKETED IN BLUEPRINT, AND HALF ARE 'INVESTMENTS', BUCKETED IN SALES; RETAIL SIGN-UPS ON THE PLATFORM ARE A POTENTIAL SOURCE OF INCOME, WHICH IS WHY THE PROSPECTS' BUILD AND RUN TOKENS ARE TRACKED (Ninad, 2026-11-17, CLI-183, for CAROL-PUR-0035, after #945, #827).

ASKED: year 1's cost of serving clients (0.5M) exceeds its revenue (0.1M) because 0.4M of it is prospects worked to win the 6 clients; charging prospects would turn the biggest loss line into revenue if they pay enough and still convert.

HIS WORDS: "plan for charging prospects, because I see this as a potential revenue line.. unlike traditional businesses, this is a platform, and many curious clients will signup on the platform if we market it well, which can become a potential source of retail income.. what do you think?"; then: "every retail sign-up consumes tokens - that is why its important to track the build and run costs for prospects. Assume 50% paid prospects, bucket them in blueprint and 50% 'investments', bucket them in sales".

READ AS (Orion's reading, not the administrator's words): (a) of the prospects worked each year, HALF are PAID prospects: they pay for their prototype, and their revenue, their FDE time and their build and run tokens are the BLUEPRINT service's (cost and income alike); the other HALF are INVESTMENTS: unpaid, their cost under SALES and customer acquisition as today (#827); the half is an assumption of the register, changeable towards actuals; (b) what a paid prospect pays is an assumption to approve (#86); (c) the prospects' tokens are two buckets of #945 (build and run), split between the paid and the investment halves, so every retail sign-up's tokens are seen; (d) RETAIL sign-ups -- curious users signing up on the platform at a small price, driven by marketing, some growing into micro clients -- are a revenue line the plan carries from the year the administrator names, with their price, sign-ups per euro of marketing, churn, tokens per sign-up and the share that grow into clients as assumptions; they are the cheapest prospects; (e) the figures come back for approval before Clara's registrar records them.

A retail sign-up pays a flat monthly price with a token allowance inside it; beyond the allowance tokens are metered

Ruled 2026-11-16 by ninad -- in force

RETAIL SIGN-UPS ARE PRICED FLAT WITH A TOKEN ALLOWANCE INSIDE THE PRICE; TOKENS BEYOND IT ARE METERED AT THE CLIENT FORMULA.

ASKED (his question): what if retail users consume too many tokens -- or will it be metered? Orion showed three models -- flat (heavy users cost more than they pay), flat with a cap (the price includes a token allowance, beyond it the account pauses or buys more), metered (a base fee plus tokens at cost + markup, the client formula) -- and proposed the capped flat: a simple price for marketing and a margin that can never go negative.

HIS WORDS: "ok, record this, and do session handover, let the next session pick them up".

The proposed figures for charging prospects and for retail sign-ups are approved

Ruled 2026-11-16 by ninad -- in force

THE PROPOSED FIGURES FOR CHARGING PROSPECTS AND FOR RETAIL SIGN-UPS ARE APPROVED.

"Approve?"

HIS WORDS: "Approve all".

READ AS (Orion's reading, not the administrator's words): (a) every figure in the question is approved as an assumption of the plan, so Clara's registrar may record them (#86); (b) builds the prospects revenue line on them -- half the prospects paid under Blueprint, half investments under Sales (#946), retail priced flat with the allowance inside and tokens beyond it metered at the client formula (#947); (c) the token use shares proposed with them (40/30/30 to 10/15/75) were NOT part of this question and are not approved by it -- he answered that separately in CLI-188.

The plan's tokens come from client counts and sizes; each client's tokens are the 35-75% gradient of the spend on it; platform tokens are an upfront investment

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

THE PLAN'S TOKENS COME FROM THE CLIENT COUNTS AND SIZES: EACH CLIENT'S TOKENS ARE THE 35% TO 75% GRADIENT OF THE SPEND ON THAT CLIENT, PLATFORM SERVICES TAKE THE REMAINING SHARE, AND THE PLATFORM'S OWN TOKENS ARE AN INVESTMENT DECIDED UPFRONT.

ASKED: the use shares CLI-183 proposed for the six buckets (40/30/30 to 10/15/75); then whether his answer conflicted with #925's token share of spend rising 35% to 75%.

HIS WORDS: "platform token consumption is an investment, which we can decide upfront. Prospects and projects depend on the count of prospects and projects and their requirements.. for the purpose of planning, the assumed client size and count should become the basis for token consumption estimations.. for platform, the token consumption will depend on the service operators onboarded. The actual token consumption should depend on how big each service is, and how much is it being used.. generally speaking, profitable services are rewarded with more tokens in their budgets, thereby motivating service owners to stay profitable, while following the business ethics and policies. I havent still addressed how much tokens are required, because that is yours to decide, based on the size of the platform, and size of the services, but it should include both build and operate budgets (50:50)". Then: "total tokens come from client counts and their sizes. tokens consumed for each client comes from the guideline that 35% to 75% (gradient) of the total spend on the client comes from token consumption, and platform services take the remaining share. As the platform becomes more and more robust, the cost of services is assumed to go down because agents will do most of the work, and FDE's will only provide oversight.. this is an assumption, or an ambition, the truth will unfold as we move down the path. To me, its not a conflict, both rules complete the story."

READ AS (Orion's reading, not the administrator's words): (a) the proposed use shares 40/30/30 to 10/15/75 are NOT approved; the buckets are derived instead; (b) PROJECTS: the year's client tokens = the sum over the clients of each size (client_flows) of the gradient share (token_share_of_spend, #925) of the spend on that client; the rest of the spend on the client is platform services -- FDE oversight and the services' work -- a share that falls as the platform matures (an ambition, #936); (c) PROSPECTS: likewise from the count of prospects and their requirements; (d) PLATFORM: an upfront investment Orion sizes from the size of the platform and of its services and the service operators onboarded, build and operate 50:50 (#945), recorded as an assumption for his approval (#86); (e) every bucket is split 50:50 build and run (#945).

Ended because: Ninad defined the unit economics per client in CLI-188: the gradient is a reading, not an input; the services charge rides tokens consumed

The plan allocates a flat services cost per client size; actuals charge by the services used

Ruled 2026-11-16 by ninad -- in force

THE PLAN ALLOCATES A FLAT SERVICES COST TO EVERY CLIENT OF A SIZE; ACTUALS CHARGE BY THE SERVICES THE CLIENT USES.

ASKED: with each client's tokens the gradient share of the spend on it (#951), how is the services part of that spend worked out -- a flat cost for every client, or by the number of services the client avails?

HIS WORDS: "Computing token costs is straight forward, the real question is, how are the services cost computed. Is it a flat cost for all clients, or would it be determined based on the number of services the client avails? I prefer later for actuals, but for planning purposes, a flat cost should be allocated."

READ AS (Orion's reading, not the administrator's words): (a) in the PLAN the services cost of a client is a flat figure per client size, the remaining share of the spend on that client after its tokens (#951); (b) in ACTUALS a client is charged for the services it actually uses, each at its owner's cost plus margin (#912, cost-center #695); (c) the plan's figure is an assumption of the register, approved per #86.

A client is charged at the mature mix (75% tokens) from year one while the spend follows the gradient; the gap is the early-years subsidy, its own line

Ruled 2026-11-16 by ninad -- superseded 2026-11-16

A CLIENT IS CHARGED AT THE MATURE MIX FROM YEAR ONE -- 75% TOKENS, 25% SERVICES -- WHILE THE SPEND ON IT FOLLOWS THE 35% TO 75% GRADIENT; THE GAP IS THE EARLY-YEARS SUBSIDY, SHOWN AS ITS OWN LINE AND ABSORBED AS LOSS.

ASKED (his question): "if 35% is token cost, that means the platform will be expensive to early users.. this will not work well, during the initial years, I think we should get rid of the gradient, and keep token 75% right from the beginning.. this will result in losses, which will have to be absorbed, what do you think?" Orion: the invoice is cost-based (#912), so the mix changes who does the work, not the size of the bill; what he proposes is charging every client as if the platform were already mature and absorbing the extra people cost while it is not -- modelled as two numbers, never one: the CHARGE at 75/25 flat per size from year 1, the SPEND on the gradient (people still do most of the work early; FDEs stay 50% of staff, #925, and feasibility counts them), and the gap a named subsidy line, so investors see it and when it ends instead of a plan claiming agents do 75% of the work on day one.

HIS WORDS: "I agree".

READ AS (Orion's reading, not the administrator's words): (a) what a client of a size is CHARGED is worked out at the mature mix -- 75% tokens, 25% services, each cost plus markup (#912) -- flat per size (#953), from the first plan year; (b) what the plan SPENDS on that client keeps the gradient of #925 (35% tokens rising to 75%), the rest its flat services cost; (c) the difference, spend less charge, is the EARLY-YEARS SUBSIDY: one named line of the plan, per year, absorbed as a loss in the initial years (#925) and ending in the year the gradient reaches 75%; (d) the token buckets are worked out on the SPEND; the revenue, gross line and cash path read the CHARGE and the subsidy; (e) setting the spend share to 75% from the start -- cutting the people the plan allows per client -- undoes this.

Ended because: Ninad hinged the services charge to tokens consumed (CLI-188): the 75/25 mix is the first year's pricing decision, the rate then holds per token

The plan runs on unit economics per client: price by size, tokens derived, a services rate per token, real services from planned FDEs, every client's margin kept, investments distinct

Ruled 2026-11-16 by ninad -- in force

THE PLAN RUNS ON UNIT ECONOMICS PER CLIENT: REVENUE = COST + MARGIN, COST = TOKENS + A SERVICES CHARGE; THE PRICE IS THE CLIENT'S VALUE BY SIZE, THE TOKENS IT CONSUMES ARE DERIVED FROM IT, THE SERVICES CHARGE IS A RATE PER MILLION TOKENS CONSUMED SET SO SERVICES ARE 25% OF THE TOTAL COST IN THE FIRST YEAR AND HELD INDEPENDENT OF TOKEN PRICES; THE REAL SERVICES COST IS THE PLANNED FDEs ON PROJECTS; EVERY CLIENT'S PLANNED AND ACTUAL MARGIN IS KEPT; INVESTMENTS ARE DISTINCT FROM THE UNIT ECONOMICS.

ASKED, in turn: what "the spend on a client" is that the 35-75% gradient applies to (three readings differed twelve-fold by 2036); whether the margin is a fixed markup or the residual of a value-anchored price over a falling cost; what the services charge is hinged to.

HIS WORDS: "what's left in the cost that counts as 'on clients' --> tokens (75%) + service (25%). Unit economics (per client): 1. cost per client = tokens + service charge 2. revenue per client = cost per client + margin. Round table economics: total cost = (count of clients * cost per client) + platform costs; total revenue = count of clients * revenue per client; cash at hand = funds raised + total revenue - total cost". "prospects are considered as clients, but you have to take into account the 50% investments and flat fees which may incur losses.. investments should be distinct from unit economics per client.. for every client, we should know whether that client is a profit making client or a loss making client, so the margin per client (both planned and actual) should be maintained". "its value-anchored with falling cost". "We take a pricing decision to keep services charges per client at 25% of the total cost, which will impact the unit economics, and the client will go into a loss, until eventually, we bring the services cost to 25% of the token cost, and then we are in profit as expected" -- then "services at 25% of the total cost". "when the token prices fall, the services fall further, because service costs are hinged to the token cost.. that should not be the case, and they should be independent and isolated.. maybe they should be hinged to tokens consumed, instead of token cost?" -- Orion's shape (price by size; tokens consumed = price / (token price x (1 + markup) + services rate); services charge = tokens consumed x a services rate per million tokens; real services = the FDEs on projects x their pay / the clients they serve; margin = price - tokens at cost - real services; subsidy = real services - services charge when positive) answered "ok, but FDEs at 50% of the planned staff, not actual.. else we fall in the same trap again!"

READ AS (Orion's reading, not the administrator's words): (a) PRICE: a client of a size pays its recorded value (customer group values) -- the ambition, unchanged; (b) TOKENS CONSUMED by a client of a size in a year = its price / (the year's token price x (1 + token_markup_share) + the services rate), so a client gets more tokens for the same money as prices fall; (c) SERVICES CHARGE = tokens consumed x a SERVICES RATE per million tokens -- a pricing decision recorded as an assumption, set so that in the first plan year the services charge is 25% of the client's total charged cost, and from then on held per token, never re-derived from token prices; (d) REAL SERVICES COST on clients = the PLANNED FDEs on projects (half of the planned staff, #925) x their pay band, spread over the planned clients -- never actual heads; (e) each client's PLANNED MARGIN = price - tokens at cost - real services, read per client and per size; its ACTUAL margin is kept beside it from the records once there is an actual (Midas's books), so every client reads profit-making or loss-making; (f) the SUBSIDY on a client = real services - services charged, when positive; the subsidy on a prospect = its budget less its fee (the paid half) or its whole budget (the investment half, #946); the subsidy on a retail sign-up = its cost to serve less its price when positive (#947); (g) TOTAL INVESTMENT = the platform's token investment + the client subsidies + the prospect subsidies + the retail subsidies, each its own line, distinct from the unit economics; (h) the year's COST = clients x their real cost + prospects worked x their budget + the platform's costs (people, Azure, the token investment) + compliance and admin -- DERIVED; the typed year costs retire in place as history once the derivation is installed; the year's revenue = clients x their price + the paid prospects' fees + retail; cash = funds raised + revenue - cost, month by month as today; (i) the 35-75% GRADIENT (token_share_of_spend) is RETIRED AS AN INPUT: it becomes a READING -- the tokens' share of the real spend on a client, which rises as the people per client fall; (j) the six token buckets are read from this: projects = the clients' tokens consumed at cost, prospects = their budgets' token part, platform = the investment; each half build, half run.

The plan's token price falls 10% a year, from published price history

Ruled 2026-11-16 by ninad -- in force

THE TOKEN PRICE FALLS 10% A YEAR IN THE PLAN, FROM PUBLISHED PRICE HISTORY -- FAR UNDER THE PUBLISHED RATE, BECAUSE THE STRUGGLE IS THE INITIAL YEARS.

ASKED: a token price path, proposed from published history. Orion: the price for a given capability has fallen about 10x a year since 2021 (a16z, LLMflation) and ~13x a year since 2023, slowing to ~4.7x once a capability is two years old (Epoch AI, The plunging price of thought, Sep 2026; earlier range 9-900x, median 50x), but the plan buys the frontier mix whose list prices barely move and agent workloads consume 5-20x more tokens per task; proposed -50% a year to 2031/32 then -25%.

HIS WORDS: "lets keep 10% a year, because we are doing pretty good for the later years, the struggle is for the initial years".

READ AS (Orion's reading, not the administrator's words): (a) the plan's token prices (token_price_eur_per_m, build and ops) fall 10% a year from the first plan year, one index for both kinds, recorded as a schedule assumption with this basis and the risk that frontier prices stay flat; (b) deliberately conservative: it does not lean on the published collapse to make the later years work, so that the plan's strength shows in its first years or not at all.

A client's charge starts at the mature mix in the first year and the services rate is then held per token; the subsidy is shown per client, prospect and retail sign-up

Ruled 2026-11-16 by ninad -- in force

A CLIENT'S CHARGE STARTS AT THE MATURE MIX -- SERVICES 25% OF ITS TOTAL COST IN THE FIRST PLAN YEAR -- AND THE SERVICES RATE IS THEN HELD PER TOKEN; THE GAP TO THE REAL SERVICES IS THE SUBSIDY, SHOWN PER CLIENT, PER PROSPECT AND PER RETAIL SIGN-UP (Ninad, 2026-11-17, CLI-188, superseding #954; after #912, #946, #947).

HIS WORDS: "it should clearly state what is the actual cost to the client vs what is the subsidy provided as a pricing strategy.. as long as these numbers are captured and transparency is maintained, we are good"; "services at 25% of the total cost".

READ AS (Orion's reading, not the administrator's words): (a) #954's 75/25 mix is the FIRST YEAR's pricing decision, from which the services rate per million tokens is set; in later years the mix of a client's bill follows from that rate and the falling token price (the services part grows), never re-pinned to 75/25; (b) invoice_token_share (the mature mix) is retired as an input in favour of the services rate; (c) every surface that shows a client, a prospect or a retail sign-up shows its price, its real cost to serve and the subsidy between them, by name.

The platform's cost is charged through to the clients flat per client; the first clients run at a deep loss for two to three years and the round table funds it

Ruled 2026-11-17 by ninad -- in force

THE PLATFORM'S COST IS CHARGED THROUGH TO THE CLIENTS IN THE PLAN, FLAT PER CLIENT, SO A CLIENT'S REAL COST CARRIES WHAT CAROLVERSE'S SERVICES DO FOR IT; THE FIRST CLIENTS RUN AT A DEEP LOSS FOR TWO TO THREE YEARS AND THE ROUND TABLE FUNDS IT.

READ AS (Orion's reading, not the administrator's words): (a) a client's REAL cost in the plan = its tokens at cost + the FDEs on projects (#957) + its flat share of the PLATFORM'S cost of the year (the people outside projects and prospects, Azure, compliance and admin, the platform's token investment), spread over the planned clients (#953: flat for planning; actuals charge by the services used, #695); (b) the company's total cost is unchanged -- the internal charge is Carolverse's services charging Blueprint, which cancels at the company level; (c) the subsidy on a client = that real cost's services part (FDEs + the platform charge) less the services charged (#959), so every client reads loss-making while the platform is young and profit-making once it carries more clients; (d) the round table's raises fund those early losses -- a consequence to be shown, not hidden; (e) built in (each year's cost derived), read back by 's figures; Blueprint's own P&L against the platform services' P&L is 's (Agent economy).

THE PLAN'S FIRST YEAR, 2026/27, IS PLANNED AS TODAY'S TEAM AT THE LEAN PERIOD'S PAY

Ruled 2026-11-17 by ninad -- in force

THE PLAN'S FIRST YEAR, 2026/27, IS PLANNED AS TODAY'S TEAM AT THE LEAN PERIOD'S PAY (Ninad, 2026-11-17, CLI-192, for, after #957 (h), #858).

HIS WORDS: chose "Today's team, lean pay (Recommended)".

READ AS (Orion's reading, not the administrator's words): (a) 2026/27's people are the administrator, one estate operator and one service operator, each at the lean_period_pay_eur the guidelines register already holds for them; (b) they are recorded as the plan's staffing for 2026/27 through the staffing assumptions, never typed into a cost; (c) 2026/27's cost is then worked out like every other year's (#957 (h)).

COMPLIANCE, CUSTOMER SUCCESS AND SUPPORT STAFF ARE PAID AT THE OPERATIONS BAND, AND 2027/28 READS THE 2028/29 BANDS

Ruled 2026-11-17 by ninad -- in force

COMPLIANCE, CUSTOMER SUCCESS AND SUPPORT STAFF ARE PAID AT THE OPERATIONS BAND, AND 2027/28 READS THE 2028/29 BANDS.

ASKED: no pay band is recorded for compliance staff, customer success or support staff, and the bands start only in 2028/29; which pay should the plan use for them (proposed: compliance at the subject-matter expert band, customer success and support at the operations band, 2027/28 at the 2028/29 bands)?

AZURE IS PLANNED FROM THE MACHINES EACH YEAR NEEDS: SANDBOX BY PROSPECTS WORKED AT ONCE, ONE MACHINE PER PROJECT BY CLIENT SIZE, AND THE ESTATE'S

Ruled 2026-11-17 by ninad -- in force

AZURE IS PLANNED FROM THE MACHINES EACH YEAR NEEDS: SANDBOX BY PROSPECTS WORKED AT ONCE, ONE MACHINE PER PROJECT BY CLIENT SIZE, AND THE ESTATE'S.

HIS WORDS: "One per project? it will depend on the size of each project, wouldnt it?"; then, shown machine costs by project size: "the estimates should be revised..

READ AS (Orion's reading, not the administrator's words): (a) each plan year's Azure = the estate machine + the Sandbox machines (the prospects worked in the year x the months a prospect is worked / 12, over the prospects a machine holds, rounded up) + one machine per project held at the year's end, each at its size's yearly cost; (b) the figures are assumptions of the plan's register, approved here (#86), recorded by Clara's registrar; (c) a micro client's one project is new to the Azure count only -- #682's projects per client for revenue and tokens are unchanged; (d) the Azure figure is part of the platform's cost (#960), worked out, never typed.

The health check holds the plan against agreed guidance, standards and ambitions, never the round table

Ruled 2026-11-17 by ninad -- in force

THE HEALTH CHECK HOLDS THE PLAN AGAINST THE AGREED GUIDANCE, STANDARDS AND AMBITIONS, NEVER AGAINST THE ROUND TABLE; THE ROUND TABLE IS WHERE THE PLAN IS STORED.

ASKED: since #941 the round table's profit is the plan's own figure, so the health check has nothing to hold profit against -- hold it against the old approved profits, a new profit or margin target, or leave it out? ANSWERED: "the health check should not be checking the round table with the plan. The round table is the repository to store the plan. The health check should be checking the plan against the guidances, standards, ambitions etc. that were agreed upon when building the health check.."
READ AS (Orion's reading, not the ruling): no health-check row compares a plan figure with a round-table row (raise, valuation after the round, customers, revenue rate, cash, assets, founders' share); each row is held against a recorded guideline, assumption, standard or a ruling's ambition, and an item with no agreed value reads not worked out, naming what is missing.

The latest saved round table is the health check's standard for profit, raises and shares

Ruled 2026-11-17 by ninad -- in force

FOR PROFIT AND MARGIN, RAISE AND PRE-MONEY PER ROUND, AND THE FOUNDERS' AND INITIAL INVESTORS' SHARES, THE HEALTH CHECK HOLDS THE PLAN AGAINST THE ROUND TABLE OF THE LATEST SAVED PLAN VERSION.

ASKED: under #991 three ambitions have nothing agreed to hold them against -- profit and margin, raise and pre-money per round, the founders' and initial investors' shares; what target, limit or minimum, or drop them? ANSWERED: "for these ambitions, use the latest version round table as the standard"
READ AS (Orion's reading, not the ruling): the latest version is the newest one Clara's registrar recorded in the plan's version history; its round table is read as that version stood (profit as it recorded it, raises, valuations after the money, the founders' share after each round), and today's plan is held against it within the recorded tolerances.

The client contract says Carolverse may make a project's service reusable on the administrator's approval, keeping none of the client's data or IP, entering it in the IP register, and paying for it from its own platform improvement spend, never the project's

Ruled 2026-11-18 by ninad -- in force

THE CLIENT CONTRACT'S REUSE CLAUSE.

THE QUESTION: Orion drafted, for the administrator's approval, a new section 7 of the service agreement template, as blueprint #997 asks ("Contract should mention that the platform may decide to transform certain project services re-usable in carolverse"):
"7. Services made reusable in Carolverse. The Company builds the Client's services on Carolverse, its shared platform. Where a service built for the Client's project would also serve other projects, the Company may, with its administrator's approval, turn it into a reusable Carolverse service. When it does: (a) the reusable service keeps none of the Client's data and none of the Client's intellectual property; only the general capability is kept, rebuilt free of anything specific to the Client's business; (b) the service is entered in the Company's register of intellectual property, with the project it came from."
HIS WORDS: "also - the reusable service is built from companies own platform improvement spend, never from project spend"

READ AS (Orion's reading, not the administrator's words): (1) the drafted section 7 is approved, with his addition as part (c): "the work of making it reusable is paid from the Company's own spend on improving its platform, never from the Client's project or its charges"; (2) it enters the service agreement template as version 0.2, which supersedes 0.1 and, like it, is not for signature until a lawyer qualified in England and Wales has reviewed it; (3) the same rule holds inside Carolverse's books: work that makes a service reusable is platform enhancement, never billed to a project.

The plan's cost is read in seven categories, leadership and board on its own

Ruled 2026-11-19 by ninad -- in force

THE PLAN'S COST IS READ IN SEVEN CATEGORIES: PLATFORM ENHANCEMENT (OPERATORS, WITH ADMIN AND THE ESTATE OPERATORS), LEAD GENERATION (SALES), SELF-SERVICE SUPPORT (FDE1), PROSPECT SUPPORT (FDE2, WITH THE SUBJECT-MATTER EXPERTS), PROJECT SUPPORT (FDE3), OPERATIONS (WITH THE COMPLIANCE COSTS AND RETAIL SIGN-UP MARKETING), AND LEADERSHIP AND BOARD ON ITS OWN (Ninad, 2026-11-19, CLI-222).

HIS WORDS: "I feel that the sales expense in year 1 is a bit too much.. show me the expense in these categories 1. Platform enhancement (operators) 2. Lead generation (sales) 3. Self Service support (FDE1) 4. Prospect support (FDE2) 5. Project support (FDE3) 6. Operations -- put admin and estate operators in platform build -- let me know if some expense does not fit in any of these categories"; then: "compliance costs --> operations; subject-matter experts --> Prospect support; retail sign-up marketing --> operations; leadership and board --> make this a seperate section". READ AS (Orion's reading, not the ruling): 1 = estate and service operators, admin, the platform's tokens, the estate machine; 2 = sales staff; 3 = customer success (FDEs on self-serve); 4 = the prospects' budget (FDEs on prospects and the prospects' tokens inside it), the Sandbox machines, the subject-matter experts; 5 = FDEs on projects, the clients' tokens, the clients' machines; 6 = operations, support and compliance staff, the compliance costs, retail sign-up marketing; 7 = leadership and board. Every euro of a year's cost lands in one.

The plan starts in 2027/28; 2026/27 leaves the tables for now; the cost table shows the raise first and the cash last

Ruled 2026-11-19 by ninad -- in force

THE PLAN STARTS IN 2027/28, THE YEAR AFTER THE PRE-SEED LANDS (MARCH 2027); 2026/27 COMES OUT OF THE PLAN'S TABLES FOR NOW, ITS SPECIFICS TO BE INCLUDED LATER; THE COST TABLE BY CATEGORY (#1042) SHOWS THE RAISE AS ITS FIRST ROW AND THE CASH BALANCE BELOW THE TOTAL (Ninad, 2026-11-19, CLI-222).

Should 2026/27 come out of the tables or stay as year 0? ANSWERED: "it should come out for now, we will include the specifics later.. also include raise in the first row and cash balance in the last row below total"
READ AS (Orion's reading, not the ruling): the tables Ninad reads (the cost by category, the round table, the Fundraising page's yearly views) begin at 2027/28; 2026/27's records stay, retired from view, not deleted; the raise row counts each round in the plan year the round table places it (the bridge line's debt named).

The spend mix is a persistent table of the plan, beside the round table, customer mix and employee mix

Ruled 2026-11-19 by ninad -- in force

THE COST TABLE BY CATEGORY IS THE SPEND MIX, A PERSISTENT TABLE OF THE PLAN LIKE THE ROUND TABLE, THE CUSTOMER MIX AND THE EMPLOYEE MIX: THE RAISE FIRST, THE SEVEN CATEGORIES (#1042), THE TOTAL, REVENUE, THEN THE CASH BALANCE, FROM 2027/28 (#1043) (Ninad, 2026-11-19, CLI-222).

HIS WORDS: "add revenue above cash balance -- call this table 'spend mix' and make it persistent, like the other 3: 1. Round table 2. Customer mix 3. Employee mix 4. Spend mix". READ AS (Orion's reading, not the ruling): the spend mix is worked out from the plan's records each time it is read (never typed), shown on Fundraising under the customer mix and given by Clara's reader, each figure with its rounded twin; a saved plan version shows the spend mix as that version stood; a category that cannot be worked out reads not worked out with why, never nought.

The spend mix is a persistent table of the plan, beside the round table, customer mix and employee mix

Ruled 2026-11-19 by ninad -- retired 2026-11-19

THE COST TABLE BY CATEGORY IS THE SPEND MIX, A PERSISTENT TABLE OF THE PLAN LIKE THE ROUND TABLE, THE CUSTOMER MIX AND THE EMPLOYEE MIX: THE RAISE FIRST, THE SEVEN CATEGORIES (#1042), THE TOTAL, REVENUE, THEN THE CASH BALANCE, FROM 2027/28 (#1043) (Ninad, 2026-11-19, CLI-222).

HIS WORDS: "add revenue above cash balance -- call this table 'spend mix' and make it persistent, like the other 3: 1. Round table 2. Customer mix 3. Employee mix 4. Spend mix". READ AS (Orion's reading, not the ruling): the spend mix is worked out from the plan's records each time it is read (never typed), shown on Fundraising under the customer mix and given by Clara's reader, each figure with its rounded twin; a saved plan version shows the spend mix as that version stood; a category that cannot be worked out reads not worked out with why, never nought.

Ended because: duplicate of leadership #1044, recorded twice by Orion (CLI-222) when the first write was slow to report; #1044 stands (by ninad)

The round table, employee mix, customer mix and spend mix are persistent first-class registry items imported from the plan

Ruled 2026-11-19 by ninad -- in force

THE PLAN'S FOUR TABLES -- THE ROUND TABLE, THE EMPLOYEE MIX, THE CUSTOMER MIX AND THE SPEND MIX -- ARE PERSISTENT, FIRST-CLASS REGISTRY ITEMS, IMPORTED DIRECTLY FROM THE PLAN (Ninad, 2026-11-19, CLI-256).

HIS WORDS: "there are 4 tables of my interest, which are persistent first class registry items directly imported from the plan, the round table, employee mix, customer mix and spend mix". READ AS (Orion's reading, not the ruling): each table is a record in the registry, not only worked out when a page is read: Clara's registrar imports it from the plan as worked out (never typed), one row per plan year, each import retiring the earlier in place and naming the plan version and time it came from; Fundraising and Clara's reader show the four records (the employee mix becomes a table, beside its chart); where the plan as now worked out departs from a recorded table the departure is named, as the round table's are (#86 keeps the round table's approval). Clarifies the spend mix ruling's 'persistent' (#1044), which was read as worked out on each read.

๐Ÿ“–Marketing

The Carolverse video channel links back to the estate: every video description and the channel's About text

Ruled 2026-10-02 by ninad -- in force

Asked: 'For the visibility work, may I edit the YouTube channel? All 99 videos are public and none links the Carolverse landing page (61 have no links at all). The plan: add a short closing block to every video description linking Carolopedia, the Logbook and the landing page, keep a backup of each original, and replace the channel's About text with the Carolverse sentence plus the landing link.' Ninad answered: 'Yes, do both'. READ AS: the closing block (composed from the public-property record by shared.public_readme) ends every Carolverse video description, existing and new, and the channel's About text is the public Carolverse sentence plus that block.

Every public HOMEPAGE names Carolverse and says what it is; YouTube's page and estate sub-pages are not counted

Ruled 2026-10-02 by ninad -- superseded 2026-10-04

Asked: 'The page-identity item's last check says every public property must name Carolverse. As written, that includes YouTube's own channel page and three small estate pages (Agent Escalations, Carol's bio, the Map), not just the homepages. Which should count?' Options: 'Homepages only -- the five estate homepages, plus the company site once its address works; YouTube's page and the three small pages are left out' or 'All ten properties'. Ninad answered: 'Homepages only'. Asked in the same sitting: 'Google's Rich Results Test has no way for me to run it automatically ... How do we settle Google's half?' Ninad answered: 'I'll run it by hand'. READ AS: the daily crawl-readiness reading must find Carolverse named for the Carolverse OS landing, Carolopedia, the Logbook, Oratorium, Carol's landing and the company site once it resolves; YouTube's channel page and the estate's sub-pages are not held to it. Google's Rich Results Test stays a criterion and is run by the administrator by hand, one run per homepage.

Ended because: CLI-684: #289 counted 'the company site' when CLI-670 believed it was the Denken Agents page; it is the DenkenLabs landing, and Ninad ruled it keeps its own title.

Every estate HOMEPAGE names Carolverse and says what it is; the company site keeps its own title and carries robots, a sitemap and Organization data linking it to Carolverse

Ruled 2026-10-04 by ninad -- in force

Carried from #289 unchanged: asked (CLI-670) which properties must name Carolverse, Ninad answered 'Homepages only' -- the five estate homepages, YouTube's channel page and the estate's sub-pages (Agent Escalations, Carol's bio, the Map) not counted -- and asked how Google's half is settled, 'I'll run it by hand'. Asked: 'Your earlier ruling says every public homepage must name Carolverse in its title and first paragraph, "plus the company site" -- but that was written when we thought the company site was the Denken Agents page. denken-labs.com is actually the DenkenLabs research-commercialisation page. Must it name Carolverse too?' Options: 'No, keep its own title (Recommended) -- DenkenLabs keeps its research-commercialisation title and first paragraph; it still gets a robots file, a sitemap and structured data naming Denken Labs as the company behind Carolverse, so search engines link the two' or 'Yes, name Carolverse'. Ninad answered: 'No, keep its own title'. READ AS: the daily crawl-readiness reading must find Carolverse named in title and first paragraph for the Carolverse OS landing, Carolopedia, the Logbook, Oratorium and Carol's landing; the company site denken-labs.com is NOT held to it and keeps its own title and first paragraph; it gets robots.txt naming its sitemap, a sitemap.xml, and Organization structured data for Denken Labs that names Carolverse as its own, so the company and Carolverse are linked for search engines. Google's Rich Results Test stays a criterion, run by the administrator by hand, one run per homepage (the company site's once its structured data is live).

Loki's AI Search Reader drops its Claude reading when its track moves to Codex

Ruled 2026-10-13 by ninad -- in force

LOKI'S AI SEARCH READER STOPS READING CLAUDE'S WEB SEARCH; IT READS GEMINI'S AND CHATGPT'S, AND ITS TRACK RUNS ON CODEX (Ninad, 2026-10-13, CLI-756).

ASKED: "Loki's weekly reader checks whether Claude's own web search finds Carolverse. On Codex it can't read Claude's search at all, only Gemini's and ChatGPT's. What should happen to the Monday Claude check?" Options: "Drop the Claude check" (move the track to Codex; the reader keeps the Gemini and ChatGPT checks and stops checking Claude; nothing is paid for) or "Keep it on the Claude API" (read Claude's search through the pay-per-call Claude API key; a few cents a week). Ninad answered: "Drop the Claude check".

READ AS (Orion, CLI-756): the Monday reading is retired in place, not deleted -- its past readings stay on record; the reader and its reviewer report Claude as NOT READ BY RULING, never as a property missing from Claude's search; the two-weeks-missing rule applies to Gemini and ChatGPT only. Reading Claude through another account, or counting Claude as 'not found' because it was not read, each undo it.

Carolopedia is managed by Loki, not Bilbo

Ruled 2026-10-22 by ninad -- in force

Ninad, 2026-10-22 (CLI-912). HIS WORDS: "I think bilbo managing carolopedia is a mistake, let bilbo work for leo, and move the carolopedia management to Loki.. anyways carolopedia is not managed efficiently, all its links are gone". Asked first whether Bilbo should report to Loki, since his real work is Carolopedia upkeep and two of his Carolopedia droids already bill Loki's Carolopedia task.

READ AS (Orion's reading, not the ruling): Carolopedia is managed by Loki. The Carolopedia app, the droids that keep it, its write-ups track and the arc restoring its content (CAROL-ARC-0024) pass from Bilbo to Loki, within Marketing. Keys and track names are identifiers and are not renamed; what moves is the owner and the service.

๐Ÿ“–Migration

A prototype built on a laptop moves onto its project's machine through Migration, rebuilt to the agent-centric modular architecture

Ruled 2026-10-26 by ninad -- superseded 2026-11-15

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Migration should be used to move prototypes built on laptop to agent centric modular design on vm". READ AS (Orion's reading, not his words): a project's prototype built on someone's laptop enters the estate through Noah's Migration (Audit, Refactor, Conform) and is rebuilt on the project's machine to the Agent-Centric Modular Architecture (ACM), never copied over as it stands.

Ended because: Ninad, CLI-154: the prototype is the prospect's proof of concept built in Sandbox, and it moves only after the contract is signed (blueprint #868)

A prospect's proof of concept moves onto its client's machine through Migration once the contract is signed, rebuilt to the agent-centric modular architecture

Ruled 2026-11-15 by ninad -- superseded 2026-11-17

Ninad, 2026-10-26 (CLI-968). HIS WORDS: "Migration should be used to move prototypes built on laptop to agent centric modular design on vm". Amended 2026-11-15 (CLI-154), HIS WORDS: "the migration utility should be used to migrate the POC to the server with the right agent centric modular checks and architecture (like we did for glover)". READ AS (Orion's reading, not his words): a prospect's proof of concept -- built in Sandbox, or on someone's laptop -- moves onto its client's own machine through Noah's Migration (Audit, Refactor, Conform) once the contract is signed (blueprint #868), and is rebuilt there to the Agent-Centric Modular Architecture (ACM) with its checks, never copied over as it stands.

Ended because: Ninad, CLI-207: with Sandbox, Migration rebuilds POC work into the build section before signing, and the build is copied as it stands to the client's machine after signing

A prospect's POC work is rebuilt by Migration into its Sandbox build section once the client agrees; at signing its build is copied as it stands onto the client's own machine, and its Sandbox section is cleared once the client has checked the copy

Ruled 2026-11-17 by ninad -- in force

Carries forward #870, which stands in full except where this amends it. #870's words: Ninad, 2026-10-26 (CLI-968): "Migration should be used to move prototypes built on laptop to agent centric modular design on vm"; amended 2026-11-15 (CLI-154): "the migration utility should be used to migrate the POC to the server with the right agent centric modular checks and architecture (like we did for glover)".

ASKED (CLI-207, 2026-11-18, on the Blueprint pursuit CAROL-PUR-0017): Ninad had said "Leo should be able to migrate any POC work from the POC zone into the build zone using the migration utility with consent or in agreement with the client / FDE.. He should be enabled to do that, please check if the blueprint enables him to check, ask and do this..". Orion reported that Blueprint does not: POC work enters a build section only by an operator's ruling, Migration's door refuses a prospect and rebuilds only onto a signed client's own machine, Leo's chat has no migration tool, and when a client signs, the signing step asks Migration on its own to rebuild the proof of concept on the client's own machine and hands the client to Hagrid for a new machine. HIS WORDS: "this has changed because of the introduction of the sandbox.. migration rebuilding work from POC zone to build zone of the sandbox, and simply copying it to the dedicated VM after signup, and clearing space on the sandbox post migration to vm. do you agree?". Orion agreed (the build section is held to architecture compliance and the POC section is exempt, blueprint #961) and asked two details: the rebuild before signing is spent whether or not the prospect signs; Sandbox is cleared only after the copy passes its checks on the client's machine, with a backup kept. HIS WORDS: "1. Who pays for the early rebuild - thats right, the build work for the prospect is paid work, because clients want to test out the platform before signing in.. Leo needs to be upfront about this, that sandbox has capacity limits, budget limits, and its intention is to show a preview to enable clients to signup, not to run as the project. Leo should also, at an appropriate time, initiate the conversation on signing the contract and moving the estate to the dedicated vm, also explaining the benifits like improved security, performance etc, since sandbox is a constrained environment. 2. When Sandbox is cleared - yes, but Leo should inform with the client to verify and check, since the sandbox prospect instance will soon be de-comissioned"; and, asked whether to record this as a ruling amending #870 and file the Blueprint initiatives for it: "Yes".

READ AS (Orion's reading, not the administrator's words): (1) before its contract is signed, a prospect's POC work moves from its POC section into its build section on Sandbox's machine only through Migration, rebuilt there to the agent-centric modular architecture with its checks, and only once the client (its FDE until Leo faces clients, #801) has agreed; Leo checks which POC work can move, asks, records the agreement and asks Migration -- this is the route by which POC work enters a build section (blueprint #899); (2) build work for a prospect is paid work: Orion reads it as what a paid prospect pays for (leadership #946, #948), spent whether or not the prospect signs; (3) Leo tells a prospect plainly, before build work starts, that Sandbox has capacity and budget limits and is there to show a preview so the client can sign, not to run the project; at an appropriate time he opens the conversation on signing the contract and moving to the client's own machine, explaining its benefits over Sandbox's constrained environment, such as better security and performance; (4) when the contract is signed, the prospect's build section is copied as it stands onto the client's own machine, which the pursuit Leo files at signing provides (blueprint #979); it is not rebuilt again there -- #870's 'never copied over as it stands' now binds POC work, which never travels unbuilt, and prototypes built on a laptop, which still go through Migration; the signing step's own ask to Migration is retired in place; (5) once the copy passes its checks on the client's machine, Leo tells the client that its Sandbox instance will soon be decommissioned and asks it to verify and check the copy; after that the prospect's section on Sandbox is cleared and its space freed, a backup kept; (6) the Blueprint pursuit holds the work that enables Leo, Migration and Hagrid to do this for any prospect.

๐Ÿ“–Operations

A user's access is approved by the agents of their project, never by the administrator, and the approval says which project, which role and what access

Ruled 2026-10-21 by ninad -- superseded 2026-10-28

A USER'S ACCESS IS APPROVED BY THE AGENTS OF THEIR PROJECT, NEVER BY THE ADMINISTRATOR, AND THE APPROVAL SAYS WHICH PROJECT, WHICH ROLE AND WHAT ACCESS (Ninad, CLI-897, 2026-10-21). Shown the steps in plain words, HIS WORDS: "all estate agents on the project should be able to approve, as this is not associated with operator governance, so its not admin work. However, it should be clear what project that user belongs to and what is the role of the user, what access the use will get etc. A user can interact with the system in 2 ways - agent \"external\" chat (not internal, its reserved for employees only) and app that have appropriate permissions for users to access them." READ AS (Orion's reading, not his words): (1) deciding a user's access is not operator governance and not the administrator's: the agents of the project the user belongs to approve or decline it -- 'estate agents on the project' read as that project's own agents (for a project built on Carolverse, Carolverse's agents serving it); an agent decides as an agent action under the pipeline's rules, recorded with the deciding agent; (2) a request and its decision state the user's PROJECT, the user's ROLE in it (a kind the project's admins and estate operators define, P.03.04.02.11) and the ACCESS granted; a decision missing any of the three is not an approval; (3) a user reaches the system in exactly two ways: an agent's EXTERNAL chat (the internal chat window is reserved for employees, cookbook 1335's window classes) and the apps whose access policies admit users (P.01.03.01.05 / 3823's human policies); (4) 's criteria 2 and 3 (the administrator's authenticated emailed reply as the one decider; no other route) are superseded and its goal is revised under this ruling; the test replies asked of the administrator in #112 are no longer the proof. OPEN (for the administrator to confirm): whether 'estate agents' means every agent of the project or a role among them (Jarvis, who owns user onboarding, or the agent mapped to the user).

Ended because: Ninad settled contradiction #7: an estate operator of the project approves a new user's onboarding, not the project's agents

An estate operator of a project approves the onboarding of a new user of that project

Ruled 2026-10-28 by ninad -- superseded 2026-10-28

AN ESTATE OPERATOR OF A PROJECT APPROVES THE ONBOARDING OF A NEW USER OF THAT PROJECT (Ninad, CLI-011, 2026-10-28). THE QUESTION ANSWERED: two of his rulings clashed and were never settled (contradiction #7, open since 2026-10-26 and keeping his Operator Instance Log box red): #61 (21 Sep) -- the onboarding approval email asks the administrator what kind a newcomer is -- and #495 (21 Oct) -- a user's access is approved by the agents of their project, never by the administrator. Asked how to settle it, HIS WORDS: "estate operator of a project approves onboarding of a new user of that project". READ AS (Orion's reading, not his words): the approver of a new user's onboarding is an estate operator of the project the user belongs to -- not the project's agents (#495, superseded by this ruling) and not the administrator as such (#61, retired); for a project built on Carolverse that includes the estate operators it inherits (P.03.04.02.19). #495's requirement that an approval says which project the user belongs to and what access they get is not withdrawn by his answer and stands here. The policy 'Onboarding approval' (P.02.05.03.01) still says admin approval, so it is rewritten with his approval (contradiction #6).

Ended because: Ninad changed his ruling on user onboarding: each project's estate operators decide user types and onboarding; Carolverse's decisions named

Each project's estate operators decide its user types and how they are onboarded; in Carolverse Carol decides and recommends the service, and every user is approved by an estate operator

Ruled 2026-10-28 by ninad -- superseded 2026-10-29

EACH PROJECT'S ESTATE OPERATORS DECIDE ITS USER TYPES AND HOW EACH IS ONBOARDED; IN CAROLVERSE CAROL DECIDES WHICH SERVICE A USER MAY SUBSCRIBE TO AND RECOMMENDS THEM FOR IT, AND AN ESTATE OPERATOR APPROVES EVERY USER; USER CHATS AND INFORMATION ARE OPEN TO THE PROJECT'S ESTATE OPERATORS AND ADMIN (Ninad, CLI-011, 2026-10-28). Supersedes #566 (recorded the same session: 'estate operator of a project approves onboarding of a new user of that project'), before his approval of the policy wording drafted from #568/#569 (proposal 50) and the dictionary's 'User', which he asked Orion to hold for this change. HIS WORDS: "I would like to change my ruling on user onboarding.. I want project estate operators to decide the user types and how they get onboarded for each project. As the estate operator for carolverse, I provide that authority to Carol, who can decide what service a user can subscribe to, and recommend the user for that service. Currently, only blueprint service is available for subscription in carolverse by carol. In carolverse, currently every user onboarded will have to be approved by the estate operator. The estate operator can change this ruling at a later point.. User chats and information are accessible to estate operators and admin of the respective projects and the projects that are built on them." READ AS (Orion's reading, not his words): (1) for every project, the kinds of user and the way each kind is onboarded are its estate operators' decision (P.03.04.02.19 already gives them what a user is and the kinds of user); (2) for Carolverse, decided by Ninad as its estate operator: Carol decides which of Carolverse's services a user may subscribe to and recommends the user for it; the only service Carol offers for subscription is Blueprint; every user is onboarded only on an estate operator's approval; Carolverse's estate operators may change these three by a later ruling -- they are decisions for Carolverse, never policy text (policies are time-agnostic); (3) a user's chats and information are open to the estate operators and admin of the user's project and, by the inheritance P.03.04.02.19 states, to those of the project it is built on -- so Carolverse's estate operators and admin reach the users of every project built on Carolverse, and a project's dedicated operators never reach the users of the project it is built on (this reading of 'the projects that are built on them' is put back to him to confirm).

Ended because: Ninad: onboarding is a ruling, not a policy; the admin and estate operators set the rules; in Carolverse they approve Carol's recommendation

A project's admin and estate operators set how its users are handled; in Carolverse Carol recommends Blueprint and an estate operator or the admin approves

Ruled 2026-10-29 by ninad -- in force

EACH PROJECT'S ESTATE OPERATORS DECIDE ITS USER TYPES AND HOW EACH IS ONBOARDED; IN CAROLVERSE CAROL DECIDES WHICH SERVICE A USER MAY SUBSCRIBE TO AND RECOMMENDS THEM FOR IT, AND AN ESTATE OPERATOR APPROVES EVERY USER; USER CHATS AND INFORMATION ARE OPEN TO THE PROJECT'S ESTATE OPERATORS AND ADMIN (Ninad, CLI-011, 2026-10-28). Supersedes #566 (recorded the same session: 'estate operator of a project approves onboarding of a new user of that project'), before his approval of the policy wording drafted from #568/#569 (proposal 50) and the dictionary's 'User', which he asked Orion to hold for this change. HIS WORDS: "I would like to change my ruling on user onboarding.. I want project estate operators to decide the user types and how they get onboarded for each project. As the estate operator for carolverse, I provide that authority to Carol, who can decide what service a user can subscribe to, and recommend the user for that service. Currently, only blueprint service is available for subscription in carolverse by carol. In carolverse, currently every user onboarded will have to be approved by the estate operator. The estate operator can change this ruling at a later point.. User chats and information are accessible to estate operators and admin of the respective projects and the projects that are built on them." READ AS (Orion's reading, not his words): (1) for every project, the kinds of user and the way each kind is onboarded are its estate operators' decision (P.03.04.02.19 already gives them what a user is and the kinds of user); (2) for Carolverse, decided by Ninad as its estate operator: Carol decides which of Carolverse's services a user may subscribe to and recommends the user for it; the only service Carol offers for subscription is Blueprint; every user is onboarded only on an estate operator's approval; Carolverse's estate operators may change these three by a later ruling -- they are decisions for Carolverse, never policy text (policies are time-agnostic); (3) a user's chats and information are open to the estate operators and admin of the user's project and, by the inheritance P.03.04.02.19 states, to those of the project it is built on -- so Carolverse's estate operators and admin reach the users of every project built on Carolverse, and a project's dedicated operators never reach the users of the project it is built on (this reading of 'the projects that are built on them' is put back to him to confirm). AMENDED THE SAME SESSION (Ninad, CLI-011, 2026-10-28), answering the drafted policy wording. Shown a policy saying a project's estate operators decide the kinds of user and how each is onboarded, HIS WORDS: "since a policy cannot say currently, its not a policy, its a ruling in ruling repo". Shown the people policy's sentence 'A project's admin and estate operators define what a user is, the kinds of user the project has and how each kind is onboarded', HIS WORDS: "No, the project admin and estate operators set the rules on how the users of the project are handled. For eg, in case of carolverse, carol, based on her conversation with the users recommends blueprint service, and recommends that the user gets subscribed to the blueprint service.. the estate operator or admin merely approves the request.." READ AS (Orion's reading, not his words): user onboarding is ruled in the Rulings Repo, never written as policy; a project's admin and estate operators set the rules on how its users are handled; in Carolverse, Carol recommends the Blueprint service from her conversation with the user and recommends the user be subscribed to it, and an estate operator or the admin approves that request -- the approval is of Carol's recommendation.

Carolverse has one user kind, Blueprint subscriber, opening Leo's and Carol's external chat and the apps open to Blueprint subscribers

Ruled 2026-10-30 by sahil -- in force

CAROLVERSE HAS ONE USER KIND, BLUEPRINT SUBSCRIBER (Sahil, estate operator of Carolverse, CLI-015, 2026-10-30), made under #571 ("each project's estate operators decide its user types"). THE QUESTION: what are Carolverse's user kinds and what does each open? Offered: one kind to start, Blueprint subscriber, since Blueprint is the only service Carol may recommend a user for. HIS ANSWER: "Yes: one user kind, Blueprint subscriber. Opens Leo's and Carol's external chat and the apps open to Blueprint subscribers." Also answered the same session, asked who decides a person's access (#495: the project's agents; #571: Carol recommends, an estate operator approves): "An agent recommends and an estate operator or the admin approves (keep my rule #571)." READ AS (Orion's reading, not his words): internal chat stays employees only (#495); follows #571 for who decides, and records the user kind as user_kind.carolverse.

Jarvis keeps user onboarding as its rightful owner, documented; Carol stays the recommender through her user communication interface

Ruled 2026-11-03 by ninad -- in force

JARVIS KEEPS USER ONBOARDING AS ITS RIGHTFUL OWNER, DOCUMENTED; CAROL STAYS THE RECOMMENDER THROUGH HER USER COMMUNICATION INTERFACE (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: after whatsapp #582 (Carol's own droids pick up and recommend access), Orion asked whether Carol should own only the pick-up or all of the access machinery Jarvis holds today (the request store, the access door, the keeper that admits). HIS WORDS: "the machinery might already exist, ownership needs to be evaluated.. if Jarvis is the rightful owner of user onboarding, so be it, as long as it is properly documented.. Carol then stays as recommender, via her user communication interface". READ AS (Orion's reading, not his words): evaluated -- the services catalogue names Jarvis (agt_017) owner of the User Onboarding track (operations-user-onboarding), so the request store, the access door and admission stay his, and this is written into the design of; Carol recommends, through her WhatsApp and web chat and a scheduled process of hers (whatsapp #582), and owns that process.

Carol asks each estate operator for the approval through their own Orion instance, with an email that it awaits them at their next sign-in

Ruled 2026-11-03 by ninad -- superseded 2026-11-03

CAROL ASKS EACH ESTATE OPERATOR FOR THE APPROVAL THROUGH THEIR OWN ORION INSTANCE, WITH AN EMAIL THAT IT AWAITS THEM AT THEIR NEXT SIGN-IN (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: asked what an estate operator or the admin should check before approving a person Carol recommends for Blueprint (nothing written anywhere says), with three options (identity and purpose; trust Carol's recommendation; known people only), he answered: HIS WORDS: "carol should get estate operator's approval via their individual orion's operator instance session and an email needs to be sent to the operator that his approval for the subscription request is awaited when he or she logs in next". READ AS (Orion's reading, not his words): (1) once Carol recommends a request, it reaches each estate operator of the project in their own Orion operator session, read first at its start like an operator communication (communications #203), and by an email telling them their approval awaits them at their next sign-in (communications #580: the email always with the Orion instance); (2) the operator decides in that session (operations #571); (3) the grounds a decider checks were asked and not set by this answer -- the decider judges, and the user onboarding pass says so until ruled.

Ended because: Ninad's later answer (CLI-023): the owning agent, not Carol, gets the operator's approval; one process for every channel; allowed channels are RDM values

Every registered Carol user gets the open access at once -- Carol's chat and Sam; a service subscription adds only that service's further access

Ruled 2026-11-03 by ninad -- in force

EVERY REGISTERED CAROL USER GETS THE OPEN ACCESS AT ONCE -- CAROL'S CHAT AND SAM; A SERVICE SUBSCRIPTION ADDS ONLY THAT SERVICE'S FURTHER ACCESS (Ninad, CLI-020, 2026-11-03). THE QUESTION ANSWERED: Orion reported that Carolverse's record of what a Blueprint subscriber gets (user_kind_access.carolverse.blueprint_subscriber) lists Leo's external chat, Carol's external chat and the apps for subscribed users, though Carol's chat is open to everyone, and asked whether to drop it. HIS WORDS: "every user gets access to the \"open\" as soon as they are registered as carol users, this includes access to carol chat and Sam. After that, depending on the service that users subscribe to, they get additional access. Right now, the only service available for user subscription is blueprint". READ AS (Orion's reading, not his words): (1) registration as a Carol user grants the open access -- the agents whose chat the registry marks open (today Carol and Sam) -- with no approval; (2) a subscription grants only what that service adds, so a user kind's access record lists only the additional access, never the open access; (3) Blueprint is the only service open to user subscription today, and its record drops Carol's chat (retired in place, never deleted).

User onboarding is one process whatever the channel, the channels a user may ask through are recorded values, and the one agent that owns it gets the operator's approval

Ruled 2026-11-03 by ninad -- in force

USER ONBOARDING IS ONE PROCESS WHATEVER THE CHANNEL; THE CHANNELS A USER MAY ASK THROUGH ARE RDM VALUES AND ONLY THOSE ARE ALLOWED; ONE AGENT OWNS USER ONBOARDING AND GETTING THE OPERATOR'S APPROVAL IS THAT AGENT'S RESPONSIBILITY (Ninad, CLI-023, 2026-11-03). Supersedes operations #585 (CLI-020, the same day), whose 'Carol asks each estate operator' gives way to the owning agent asking; #585's way of asking (each estate operator in their own Orion session, and an email that the approval awaits them at their next sign-in) stands. THE QUESTION ANSWERED: asked what an approving operator should check before letting in a person Carol recommends for Blueprint (Sachin Joshi's request), he answered first: "the code for this probably exists, check some pursuit or arc for user onboarding.. last i remember is that operator approval was just 1 step in the process, approval was mostly done by clara for all other checks". Orion found Clara's checklist on the route that starts in Carol's chat, while the web sign-in route never calls it, and asked whether every approval should go through Clara's checklist after the operator says yes. HIS WORDS: "check if user onboarding should continue to remain with clara or should go to jarvis, I see that some threads are with clara and some other threads are with Jarvis.. also, irrespective of the channel of communication, the user onboarding process should be the same.. channels alowed for users to request onboarding should be documented as RDM values and only approved channels should be allowed. Usert onboarding should be owned by an agent, and operator approval should be that agent's responsibility". READ AS (Orion's reading, not his words): (1) a person's onboarding follows ONE process -- request, Carol's recommendation, the operator's approval, the checks, the grant and admission -- whichever channel they asked through; (2) the channels a user may ask through are an RDM set, and a request from a channel the set does not hold is refused by name; (3) the owner, which he asked Orion to establish: JARVIS, by the services catalogue (he owns the User Onboarding track, operations-user-onboarding), by operations #583 and by agent-resources #87/#88 ('user onboarding is a different journey and stays with Jarvis'); the pieces on the chat route that sit with Clara today -- her Subscription Approver, her approval checklist and her Approval Request Mailer (task sales.approval_requested) -- fold into Jarvis's one process, which keeps her three checks as its own steps; (4) asking the estate operators for the approval is Jarvis's process, not Carol's; Carol stays the recommender (#583, #584).

Blueprint subscriber is a subscription type, not a user type; a project may have several types of user

Ruled 2026-11-03 by ninad -- in force

BLUEPRINT SUBSCRIBER IS A SUBSCRIPTION TYPE, NOT A USER TYPE; A PROJECT MAY HAVE SEVERAL TYPES OF USER, AS THE PROJECT NEEDS (Ninad, CLI-023, 2026-11-03). THE QUESTION ANSWERED: asked to approve a dictionary entry for 'user kind', and told it is what his operations #571 called the project's 'user types', he answered: HIS WORDS: "Blueprint subscriber is not a user type, is a subscription type, but I undertood the concept, there can be multiple \"types\" of users of the project, depending on the project.." READ AS (Orion's reading, not his words): (1) the record user_kind.carolverse names a SUBSCRIPTION type in the place of a user type, so it is mis-filed and is to be corrected (retired in place, never deleted); (2) a project's user types remain the estate operators' to define (#571); which values Carolverse's are is not yet ruled -- he asked whether admin, verified and guest form them, and that is answered separately; (3) the 'user kind' dictionary entry is not written until it is.

In Carolverse a user's type follows the services they subscribe to; a Blueprint subscriber is a client; guest and verified are tiers

Ruled 2026-11-03 by ninad -- in force

IN CAROLVERSE A USER'S TYPE FOLLOWS THE SERVICES THEY SUBSCRIBE TO: A BLUEPRINT SUBSCRIBER IS A CLIENT; GUEST AND VERIFIED ARE USER TIERS, NOT USER TYPES (Ninad, CLI-023, 2026-11-03). Follows operations #600. THE QUESTION ANSWERED: asked whether Carolverse's user types are the tiers (guest, verified; admin being an estate operator, not a user) or a separate list its estate operators define, he answered: HIS WORDS: "1. ok, so guest and verified are user tiers.. if a guest subscribes to 1 or more services of carolverse, they become verified, is that correct? 2. then, for carolverse, user types can be dictated by the subscription of services that the users avail.. for eg - if a user subscribes to blueprint, they are typically businesses or entrepreneurs or researchers or AI enthusiasts, and we refer to them as clients.. there can be other types of users in the future". READ AS (Orion's reading, not his words): (1) guest and verified stay tiers: a user reads verified while a subscription to any service of the project stands (granted after Carol's recommendation and an operator's yes, #595), guest otherwise; (2) Carolverse's user types are a recorded set its estate operators keep, each type derived from the services a user subscribes to -- today one: a user subscribed to Blueprint is a CLIENT -- with more to come; (3) the dictionary already uses 'client' for the top of the account hierarchy (client -> accounts -> projects); whether a Blueprint subscriber is that client or needs its own word is asked of him before the user type is recorded.

Clients are the verified users who become paid customers of a service, contributing to the service's revenue and the project's

Ruled 2026-11-03 by ninad -- superseded 2026-11-03

CLIENTS ARE THE VERIFIED USERS WHO BECOME PAID CUSTOMERS OF A SERVICE, CONTRIBUTING TO THE SERVICE'S REVENUE AND THE PROJECT'S (Ninad, CLI-023, 2026-11-03). Narrows operations #601's reading. THE QUESTION ANSWERED: told that the dictionary already uses 'client' for the top of the account chain (a client holds accounts, accounts hold projects) and asked whether a Blueprint subscriber is that client or needs its own word, he answered: HIS WORDS: "you are right, clients are a subset of verified users who become paid customers of the service, contributing to the service revenue and project revenue". READ AS (Orion's reading, not his words): (1) 'client' keeps ONE sense -- the client of the account chain is the paying customer, never a second word; (2) a user becomes a client by paying for a service, not by subscribing alone: a subscribed user on no paid plan (on the trial credit, say) reads verified and is not yet a client; (3) in Carolverse a user's type is derived from the services they pay for, so a user paying for Blueprint is a client; (4) the dictionary's 'Client' entry and a 'User type' entry are re-worded only as he approves them.

Ended because: Ninad (CLI-023): a client is made by signing the contract, not by paying

A verified user becomes a client when they sign the contract; every project works on a contract; Denken Labs is the first client

Ruled 2026-11-03 by ninad -- superseded 2026-11-03

A VERIFIED USER BECOMES A CLIENT WHEN THEY SIGN THE CONTRACT; EVERY PROJECT WORKS ON A CONTRACT THAT DEFINES THE SCOPE OF ACTIVITIES (NOT THEIR EXTENT), THE METHODS OF WORKING, QUALITY EXPECTATIONS, TIMELINE EXPECTATIONS AND COST DETAILS; DENKEN LABS IS THE FIRST CLIENT (Ninad, CLI-023, 2026-11-03). Supersedes operations #602, whose 'become paid customers' gives way to signing the contract as the moment. THE QUESTIONS ANSWERED: shown that nothing records when a verified user becomes a client, that Leo's registration records a client when a project is proposed, and whether the dictionary's 'First client: Denken Labs' stays though it pays for no service, he answered: HIS WORDS: "a verified user becomes client when they sign the contract. each project should work on the contract, the contract defines the scope of activities (not the extent) and also other details like methods of working, quality expectations, time line expectations, cost details etc.." and "Denken labs is the first client, because denken labs is paying for carolverse.. however, carolverse being the first project, is not built via blueprint..". READ AS (Orion's reading, not his words): (1) the signed contract is the one fact that makes a verified user a client -- not a subscription, not a payment; (2) every project works on a contract (Marco's contract register holds it, kind 'service'), its sections at least scope of activities, methods of working, quality, timeline and cost; (3) a project proposed through Blueprint has a requester who is a verified user until the contract is signed; (4) Denken Labs stays the first client because it pays for Carolverse, and Carolverse, the first project, was not built through Blueprint, so it is not held to Blueprint's onboarding.

Ended because: Ninad (CLI-023): the contract also includes payment terms

A verified user becomes a client when they sign the contract; every project works on a contract that also carries payment terms; Denken Labs is the first client

Ruled 2026-11-03 by ninad -- in force

THE CONTRACT ALSO CARRIES THE PAYMENT TERMS (Ninad, CLI-023, 2026-11-03): HIS WORDS: "the contract should also include payment terms". So every project's contract defines the scope of activities (not their extent), the methods of working, quality expectations, timeline expectations, cost details AND PAYMENT TERMS. Everything else of operations #603 stands, carried here unchanged: A VERIFIED USER BECOMES A CLIENT WHEN THEY SIGN THE CONTRACT; EVERY PROJECT WORKS ON A CONTRACT THAT DEFINES THE SCOPE OF ACTIVITIES (NOT THEIR EXTENT), THE METHODS OF WORKING, QUALITY EXPECTATIONS, TIMELINE EXPECTATIONS AND COST DETAILS; DENKEN LABS IS THE FIRST CLIENT (Ninad, CLI-023, 2026-11-03). Supersedes operations #602, whose 'become paid customers' gives way to signing the contract as the moment. THE QUESTIONS ANSWERED: shown that nothing records when a verified user becomes a client, that Leo's registration records a client when a project is proposed, and whether the dictionary's 'First client: Denken Labs' stays though it pays for no service, he answered: HIS WORDS: "a verified user becomes client when they sign the contract. each project should work on the contract, the contract defines the scope of activities (not the extent) and also other details like methods of working, quality expectations, time line expectations, cost details etc.." and "Denken labs is the first client, because denken labs is paying for carolverse.. however, carolverse being the first project, is not built via blueprint..". READ AS (Orion's reading, not his words): (1) the signed contract is the one fact that makes a verified user a client -- not a subscription, not a payment; (2) every project works on a contract (Marco's contract register holds it, kind 'service'), its sections at least scope of activities, methods of working, quality, timeline and cost; (3) a project proposed through Blueprint has a requester who is a verified user until the contract is signed; (4) Denken Labs stays the first client because it pays for Carolverse, and Carolverse, the first project, was not built through Blueprint, so it is not held to Blueprint's onboarding.

A user is verified by subscribing to a service, and an email address and a phone number on record are its prerequisite

Ruled 2026-11-03 by ninad -- in force

A USER IS VERIFIED BY SUBSCRIBING TO A SERVICE, AND BOTH AN EMAIL ADDRESS AND A PHONE NUMBER ON RECORD ARE A PREREQUISITE (Ninad, CLI-025, 2026-11-03). Follows operations #601. THE QUESTION ANSWERED: shown that policy P.05.03.01.02 marks a user verified only with both an email address and a phone number on record, while operations #601 and make verified the tier a user gets by subscribing to a service, and asked whether verified needs both or the subscription alone. HIS WORDS: "'verified' is the tier a user gets by subscribing to a service, both an email and a phone number are on record is a prerequisite".

Signing in on the web registers a Carol user, who reaches the open access at once, before any approval

Ruled 2026-11-04 by ninad -- superseded 2026-11-08

SIGNING IN REGISTERS A CAROL USER: A PERSON WHO SIGNS IN ON THE WEB REACHES THE OPEN ACCESS AT ONCE, BEFORE ANY APPROVAL (Ninad, CLI-025, 2026-11-04). Clarifies operations #593 (it does not supersede it). THE QUESTION ANSWERED: asked, while building, whether signing in on the website counts as being 'registered as a Carol user' -- today a new person who signs in with Google is held on an 'awaiting approval' page and reaches nothing until an operator approves -- or whether the open access starts only at approval, he chose: "Sign-in registers" (the option read: anyone who signs in reaches Carol's and Sam's chats at once; an approval only adds Blueprint, Leo and the subscriber apps; this matches the onboarding pass's step A3 and how WhatsApp already works, and lets Carol talk to the person before recommending them). RULED: a person who signs in on the web is a registered Carol user from that sign-in; they reach the open access -- the agents whose chat the registry marks open, today Carol and Sam -- with no approval, while their access request waits for Carol's recommendation and an estate operator's decision, which adds only what the recommended service grants. Built by.

Ended because: Ninad, CLI-071: a waiting web guest gets no chat until an estate operator approves (chose 'No chat until approved' over keeping #608)

Subscribing to Blueprint always needs an estate operator's approval; joining as a guest needs none

Ruled 2026-11-04 by ninad -- in force

SUBSCRIBING TO BLUEPRINT ALWAYS NEEDS AN ESTATE OPERATOR'S APPROVAL, WHOEVER ASKS AND THROUGH ANY CHANNEL; JOINING AS A GUEST AND TALKING TO CAROL NEEDS NONE (Ninad, CLI-028, 2026-11-04). Follows operations #595 and #608. THE QUESTION ANSWERED: "Building 7023 (one onboarding process run by Jarvis, whatever channel a person asks through). Today, in Carol's chat, a person who is already verified (or an admin) who asks for a service gets it instantly, with no operator approval; everyone else waits for an operator. Under the one process, should a verified person asking for an additional service still wait for an estate operator's approval?" (options offered: everyone waits / verified skip approval). HIS WORDS: "estate approval is necessary for subscribing to blueprint. Users can still join as guests without estate operator approval.. and talk to Carol". READ AS (Orion's reading, not his words): (1) no tier is exempt -- a verified person or an admin asking in Carol's chat for a Blueprint subscription waits for an estate operator's approval like anyone else, so Clara's instant chat approval for the admin and verified tiers is retired in place; (2) a person who signs in or writes to Carol is a guest at once with the open access (#608), no approval; (3) Blueprint is the only service open to user subscription today (#593), so this governs every user subscription today. Built by.

Every interaction with a user is through Carol until they are introduced to Leo; then Leo may talk to them directly

Ruled 2026-11-04 by ninad -- in force

EVERY INTERACTION WITH A USER HAPPENS THROUGH CAROL UNTIL THEY ARE INTRODUCED TO LEO; ONCE SUBSCRIBED TO BLUEPRINT AND INTRODUCED, LEO MAY TALK TO THEM DIRECTLY (Ninad, CLI-028, 2026-11-04). THE QUESTION ANSWERED: "Building 7023: once an operator approves and Jarvis's checks pass, the person gets the subscription, the trial credit and a welcome message. Today Clara's approver sends that welcome as Carol on WhatsApp. The rules say an agent writes only from its own identity, and Jarvis has no WhatsApp number. Who should send the welcome?" (options offered: Carol where they asked / Jarvis by email / both). HIS WORDS: "all user interations happen via carol, until the users get introduce to Leo. Once the users get subscribed to the blueprint service and get introduced to Leo, then leo can communicate with the users directly". READ AS (Orion's reading, not his words): (1) Jarvis's onboarding process never writes to the person itself: the welcome after an approved, checked grant is Carol's, as herself, in the channel the person asked through; (2) that welcome is where Carol introduces the person to Leo; (3) from that introduction Leo may communicate with the person directly, and before it he does not. Built by (the welcome through Carol); Leo's direct contact after the introduction is its own build.

Signing in on the web registers a Carol user, who reaches the open access at once, before any approval

Ruled 2026-11-08 by ninad -- superseded 2026-11-08

A PERSON WHO SIGNS IN ON THE WEB REACHES NOTHING -- NOT EVEN CAROL'S OR SAM'S CHAT -- UNTIL AN ESTATE OPERATOR APPROVES THEIR REQUEST; THE WAITING PAGE NO LONGER SAYS THEY CAN TALK TO CAROL AND SAM. Supersedes operations #608. WHAT HE SAW: as Mina, a new test guest on her way to Carol's chat, he was put on the waiting page, which read "While your request waits you can already talk to Carol and Sam: open your apps"; HIS WORDS: "While your request waits you can already talk to Carol and Sam: open your apps - I get this message, which is not correct". THE QUESTION ANSWERED: told that operations #608 and #614 let anyone who signs in talk to Carol and Sam before any approval (an approval adding only Blueprint and Leo), which is why the page says it, and asked what should happen to a guest who is waiting -- options: 'Straight to Carol' (keep #608/#614), 'No chat until approved', 'Keep page, fix wording' -- he chose "No chat until approved" (the option read: "Change your 3-4 Nov rulings. A waiting guest reaches nothing, not even Carol or Sam, until an estate operator approves, and the line goes."). READ AS (Orion's reading, not his words): (1) for a web sign-in the open access of operations #593 starts at the estate operator's approval, not at sign-in, so 'signing in registers a Carol user' (#608) is retired in place; (2) a waiting or declined web sign-in is refused at every app, Carol's and Sam's chats included, and the waiting page drops its open-access line; (3) the question was asked about the web sign-in only: whether a person who writes to Carol on WhatsApp before approval may still talk to her (#614's 'users can still join as guests without estate operator approval.. and talk to Carol') is put to him as its own question, never assumed.

Ended because: Ninad, CLI-071: the guest gate is temporary, its approval separate from a Blueprint subscription's, WhatsApp answers with the sign-in link, a declined person gets Oratorium

Signing in on the web registers a Carol user, who reaches the open access at once, before any approval

Ruled 2026-11-08 by ninad -- in force

WHILE USER ONBOARDING IS BEING BUILT, A GUEST REACHES CAROL ONLY ONCE THEIR GUEST ACCESS IS APPROVED -- ON THE WEB AND ON WHATSAPP, WHERE CAROL ANSWERS AN UNAPPROVED PERSON WITH THE SIGN-IN LINK; THAT APPROVAL IS ITS OWN, APART FROM A BLUEPRINT SUBSCRIPTION'S, AND IS DROPPED WHEN USER ONBOARDING IS READY; A DECLINED PERSON IS GIVEN THE ORATORIUM PAGE (Ninad, CLI-071, 2026-11-08). Supersedes operations #702 (same session) and, while the gate stands, narrows #614's 'join as guests without estate operator approval.. and talk to Carol'. THE QUESTION ANSWERED: asked: "Carol answers anyone unapproved on WhatsApp with one fixed message carrying the sign-in link, so every request goes through the same sign-in and approval as the web. Carol's reply: one plain sentence with the link -- she doesn't chat until they're approved. Matching the number: the link carries their WhatsApp number, so the request holds both their email and their phone, and the approval covers WhatsApp too. Welcome: Carol sends it by email after approval, as now, so WhatsApp's 24-hour limit doesn't matter. This replaces your 4 Nov ruling that guests can talk to Carol without approval. Should I record it as your ruling and file the build in the User rulings pursuit?" (after he asked "can carol provide the link to get access first?"). HIS WORDS: "yes, but this approval is temporary, until user onboarding is being built.. once its ready, guests should be allowed by just signing in. So its important that the authentication for carol access is different from other authentications like blueprint subscriptions" and "also a few issues with the current process.. 1. if the login is rejected, it should not appear in my \"login as\" drop down 2. if the login is rejected, the user should get a link to oratorium page in the refusal message". READ AS (Orion's reading, not his words): (1) #702 stands as to what a waiting web sign-in reaches -- no app, not even Carol's or Sam's chat -- and the same holds on WhatsApp: Carol answers a person who is not approved with one plain message carrying the sign-in link, which carries their WhatsApp number so the request holds their email and phone and an approval covers WhatsApp too; she does not chat with them until approved, and welcomes them by email; (2) this guest gate is TEMPORARY: once user onboarding is built, guests are allowed by signing in alone (#593/#608's open access at sign-in returns), so the gate is a setting the estate operators turn off, never a rewrite; (3) the approval that opens Carol's access (guest access) is a separate approval from a service subscription such as Blueprint (#614 stands for Blueprint): approving guest access grants no subscription, and when the gate is dropped Blueprint still needs an estate operator's approval; (4) a person whose request is declined is given, on the refusal page, a link to the Oratorium page (https://oratorium.denken-labs.com/, which opens without signing in); (5) point 1 of his issues (a declined test guest leaves the administrator's 'Log in as' menu) is recorded on security #693 as its point 7.

Hagrid machine tests use a project that passes the contract gate; Chowpatty waits for its signed contract

Ruled 2026-11-09 by ninad -- in force

Ninad, 2026-11-09 (CLI-083), answering Sahil communication #329. ASKED: gate (operations #605, blueprint #607) refuses Chowpatty machine and build filings because it has no signed contract, so Hagrid tests 6806 and 6857, which make a real Chowpatty machine, fail. Options: 1 contract first and tests stay red; 2 the tests use a project that passes the gate and Chowpatty still waits for its contract; 3 Chowpatty exempt as first client. HE CHOSE: Tests use another project -- Hagrid two tests switch to a project that passes the contract rule; Chowpatty still waits for its signed contract. READ AS (Orion, not his words): the gate is unchanged and Chowpatty gets no exemption; 6806 and 6857 are re-pointed, not weakened.

A self-serve user who pays gets Blueprint at once; an estate operator reviews each new sign-up afterwards and may cancel and refund

Ruled 2026-11-19 by ninad -- in force

A SELF-SERVE USER WHO PAYS GETS BLUEPRINT AT ONCE; AN ESTATE OPERATOR REVIEWS EACH NEW SIGN-UP AFTERWARDS AND MAY CANCEL AND REFUND (Ninad, 2026-11-19, CLI-258; narrows operations #614 for the paid self-serve plan).

ASKED: "Today every Blueprint subscription waits for an estate operator's approval before the user gets in (your ruling of 4 Nov). An estate operator reviews each new sign-up afterwards and can cancel and refund. No waiting, but you keep control."; 'Approval stays before access'; 'Approve first, then pay'.

HIS ANSWER: "Pay, get in, review after (Recommended)".

READ AS (Orion's reading, not his words): (1) for the paid self-serve plan only, a confirmed payment grants the Blueprint subscription and Leo's chat at once; the estate operator's approval of #614 becomes a review of each new sign-up after access, with cancel and refund; (2) every other Blueprint subscription -- the enterprise plan, prospects and clients -- still waits for an estate operator's approval as #614 rules; (3) this answer does not touch the temporary guest gate of operations #703, under which a newcomer reaches Carol only once guest access is approved.

The temporary guest approval ends when the self-serve journey goes live; from then a stranger who signs in talks to Carol at once

Ruled 2026-11-19 by ninad -- in force

THE TEMPORARY GUEST APPROVAL ENDS WHEN THE SELF-SERVE JOURNEY GOES LIVE: FROM THEN A STRANGER WHO SIGNS IN TALKS TO CAROL AT ONCE (Ninad, 2026-11-19, CLI-258; settles WHEN operations #703's temporary gate is dropped).

ASKED: "On 8 Nov you ruled that a newcomer can't chat with Carol until their guest access is approved, and that this ends 'when user onboarding is ready'. Should that approval end when this self-serve journey goes live, so a stranger can sign in and talk to Carol at once? Or should it stay, so newcomers wait to be approved before they can chat, and only the payment step is free of approval?"

HIS WORDS: "ok, lets do that.. lets them start talking to carol".

READ AS (Orion's reading, not his words): (1) the self-serve journey going live (pursuit 'Self-serve sign-up', Jarvis) is the moment #703 called 'when user onboarding is ready': from then a person who signs in on the web, or writes to Carol on WhatsApp, talks to Carol at once with no guest approval -- #614's 'users can still join as guests without estate operator approval.. and talk to Carol' applies again; (2) until it goes live #703's gate stands as built; (3) a Blueprint subscription still follows operations #1070 (the paid self-serve plan at once, reviewed after) and #614 (every other subscription waits for an estate operator).

A Carolverse user is a registry person from sign-up, and every dashboard reads users from there

Ruled 2026-11-22 by ninad -- in force

Ninad, CLI-325 (2026-11-22), after finding an approved guest (Shashank Joshi) held only in the sign-in and onboarding stores: "users should be registry entries, and all dashboards (like the voice call log) should be picking information from a single place.. information of carolverse users should not be spread across"; then "yes, record this ruling in user onboarding and implement it on priority". READ AS: a person who signs up is a registry person (persons, the person store -- rule 1669: the record wins) from that moment, carrying where their onboarding stands (requested, approved, declined) and, once approved, admitted at the approval itself, never only at their next sign-in; the sign-in door's and Jarvis's onboarding stores keep their process records and point at that person; every record about a user (voice calls, chats, access) names the registry person; and every dashboard reads who a user is from the registry, never from a store of its own.

The user list holds only real Carolverse people; test users and duplicates leave it, inside the Carol chat pursuit

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-326): "The user management app should also get rid of entries for users that dont exist in carolverse, test users, duplicate users etc.. fold this into the carol chat pursuit..". READ AS (Orion, not Ninad's words): the user list shows only people who exist in Carolverse's registry; test users and duplicates are taken off it -- duplicates merged into one person, the rest retired with their records kept, never deleted -- as steps of the Carol chat pursuit.

๐Ÿ“–Process Monitoring

A scheduled run that exits zero without stating its outcome stays unauditable; the droid states it

Ruled 2026-09-26 by ninad -- in force

Since 2026-09-25 ~04:05 the run auditor records a run that exits 0 without a final structured receipt as UNAUDITABLE; 50 scheduled droids have read that way on every run since. Asked whether to revert the auditor, fix only the one droid an arc needed, or finish the work over the whole set, Ninad chose to finish it: the auditor stays strict (exit zero is not proof of work, P.01.03.01.12) and the cure is every scheduled droid printing its terminal outcome on every exit path, measured over the set.

๐Ÿ“–Quality Management

An intended access change to a folder Gate 6's sandbox mounts is explained by the gate itself, with no root step

Ruled 2026-10-06 by ninad -- in force

NO ROOT STEP FOR AN INTENDED ACCESS CHANGE TO THE SANDBOX'S FOLDERS (Ninad, CLI-732, 2026-10-06). THE QUESTION ANSWERED: Orion asked: "Nightly test run, last blocker: on 1 Oct the two Orion instance logins (orion-ninad, orion-sahil) were given access entries on the estate's Python environment, for the parked 'own login per Orion instance' work. That changed the folder's permission mask, so the test gate refuses it as unexplained drift. Both fixes are a root step. Which one?" (offered: "Re-approve once" -- root re-approves the gate's baseline for this intended change; "Remove the entries" -- root takes the two instance logins' entries off the Python environment). HIS ANSWER: "this needs a solution where root intervention should not be required in such scenarios". READ AS (Orion's reading, not Ninad's words): neither offered root step is taken; Gate 6's drift rule (cookbook 1783) is extended so that a permission change it can prove intended -- a mode difference that is only the access mask raised by named entries root's own grant record lists (security #193), with every file's content unchanged -- is explained and admitted automatically, while any other mode change still refuses by name.

Gate 6's sandbox remembers each fully proven runtime state, and the module-index check judges each file it recognises

Ruled 2026-10-18 by ninad -- in force

THE SANDBOX REMEMBERS WHAT IT HAS PROVED (Ninad, CLI-837, 2026-10-18). THE QUESTION ANSWERED: Orion asked: "Saee's next step is blocked. The sandbox where Argus tests her staged fixes has turned down every run since 10 October. The cause: the sandbox never remembers what it has already checked. Every run re-checks every change on the machine against one fixed starting point from 18 September. When the machine removed an old kernel on 10 October, it left behind index files that nothing can account for any more. My proposed fix has three parts: 1. Remember each fully checked state, so later runs compare against the latest one instead of 18 September. 2. Make the kernel-file check judge each file it recognises, instead of giving up on the whole list. 3. One time only: approve a fresh starting point. This needs your password once, because no full run has passed since 21 September. The other choices are a fresh starting point alone, which breaks again at the next kernel removal, or removing the two leftover kernel packages by hand, which you'd have to repeat after every kernel update. Which do you want?" (offered: "All three (Recommended)", "Fresh start only", "Purge leftovers"). HIS ANSWER: "All three (Recommended)". READ AS (Orion's reading, not Ninad's words): (1) after a request whose runtime drift is fully explained, Gate 6 keeps that proven state (content-addressed records plus the admission receipt, beside the approved baseline, written only by the gate's own login) and later requests attribute drift against the LATEST proven state, chained back to the approved baseline by receipts -- a change proved once stays proved, and a later removal of what was proved is an explained removal; (2) the module-index recognizer admits the index files it can prove and leaves only the rest unexplained, instead of giving up on the whole list; (3) one freshly approved baseline now, through the existing approval route (the profile builder's approval candidate, an independent inventory review, the administrator's core grant with his password, Radagast's install). Cookbook 1783's admission rule is unchanged: anything not fully explained still refuses by name.

A test declared live may name the machine files it reads, not only the estate stores under data/

Ruled 2026-10-18 by ninad -- in force

A LIVE TEST NAMES WHAT IT READS ON THE MACHINE TOO (Ninad, CLI-837, 2026-10-18). THE QUESTION ANSWERED: Orion asked: "Good news first: the sandbox fix works. Tonight's kind of full test run got past preparation for the first time since 21 September. Your compliance box is also clear now. The run then stopped on the next problem, which needs a rule decision from you. Some tests need the real machine rather than the sealed test copy: for example, checks that read the machine's login settings or the list of protected programs. The rule says such a test must name the estate data it reads. These 7 tests read machine files and no estate data, so their entries break the rule, and one bad entry stops the whole run. Argus's live lane already shows these tests the whole machine, read-only, so nothing new would be opened. What should the rule be?" (offered: "Allow machine files (Recommended)" -- a live test may name the machine files it reads instead of, or as well as, estate data, at least one, with a reason, and the 7 entries fixed to match; "Move them out" -- the 7 become daily checks run by each subject's owning agent; "Keep rule, fix each"). HIS ANSWER: "Allow machine files (Recommended)". READ AS (Orion's reading, not Ninad's words): cookbook 1787's declaration gains a second list beside `reads` (the stores under data/ the lane binds back): `host`, the machine files outside data/ the test reads, which the read-only live lane already shows; a declaration must name at least one of the two and always its reason; `reads` stays stores under data/ only; a host entry under data/ or with '..' is refused; the one reading stays confinement.live_declaration; the 7 registry entries (5359, 4849, 5388_installed, 5821, 5880, 6741, 5700) are corrected to name what they read.

The sandbox fix is judged on the sandbox's part of a full run, not on the red suite's verdict

Ruled 2026-10-18 by ninad -- in force

THE SANDBOX FIX IS JUDGED ON THE SANDBOX'S PART (Ninad, CLI-849, 2026-10-18). THE QUESTION ANSWERED: Orion asked: "The sandbox fix (one of Saee's post-MVP steps) is now working: I found why the last full test run was thrown out (a harmless cache file written mid-run) and fixed it, and a fresh run is going. But the fix's last criterion says a full run must finish as 'completed', and the gate only allows that when every test suite is green or matches an earlier accepted run. 1,455 suites are red today and no run has ever been accepted, so that can't happen whatever the sandbox does. How should the sandbox fix be judged?" (offered: "Sandbox part only (Recommended)" -- restate the criterion as: a full run where every test runs in the sandbox, the environment is verified unchanged and every result is readable; the fact that a red suite can never reach 'completed' gets filed as its own initiative; "Keep as written" -- the sandbox fix stays parked until the whole suite is green or a way to accept a red baseline exists). HIS ANSWER: "Sandbox part only (Recommended)". READ AS (Orion's reading, not Ninad's words): 's criterion 1 is replaced by a governed goal change (struck, never deleted) with: a full_suite request runs every registered entry in the prepared sandbox, its run records runtime integrity verified, and every row is auditable -- proven by the run row; criteria 2 and 3 carry over unchanged; the gate's acceptance rule (cookbook 1361: an all-green run, or one comparable to an earlier ACCEPTED run) is untouched; and how a red suite gets its first accepted run is filed as its own initiative.

A registered test that can run only from an estate operator's own session declares itself operator-only

Ruled 2026-10-18 by ninad -- in force

OPERATOR-ONLY TESTS DECLARE THEMSELVES (Ninad, CLI-849, 2026-10-18). THE QUESTION ANSWERED: Orion asked: "Two registered tests (Orion's mail door and the operators' mailbox grants) can only run from an estate operator's own Orion session, because they ask operator-only doors. In the nightly full test run they run in the sealed sandbox, skip everything, count nothing, and that alone makes the whole run 'incomplete' every night. No rule says where such a test belongs: the read-only live lane Argus runs can't reach operator doors either. Where should operator-only tests go?" (offered: "Declare them operator-only (Recommended)" -- a test can declare itself 'operator session only' with its reason; the nightly run lists it by name as not run there (like live tests today), and it keeps proving its initiative when an operator's session re-runs it at close; nothing runs it nightly, and that is said plainly; "Take them out of the register"; "Split each test"). HIS ANSWER: "Declare them operator-only (Recommended)". READ AS (Orion's reading, not Ninad's words): cookbook 1787's live declaration gains a third list beside `reads` and `host`: `operator_doors`, the operator-only programs the test asks, with the reason; the confined run lists such a test by name as excluded and says it is operator session only; Argus's read-only live lane names it as not run there, never as bad; the gate accepts the exclusion only when the captured registry declares it (as for live tests); the delivery check still re-performs it from the closing operator's session; the one reading stays confinement.live_declaration.

Prometheus's quality loop runs again: only a service's owner improves or expands it, and Prometheus nudges and motivates owners through his measurements, filing nothing for a service he does not own

Ruled 2026-11-18 by ninad -- in force

PROMETHEUS'S LOOP RUNS AGAIN, AND OWNERS DO THE IMPROVING (Ninad, CLI-221, 2026-11-18; after blueprint #1007; on CAROL-PUR-0084).

THE QUESTION: Orion reported that Prometheus's improvement loop does not close on the estate (the handbook's code read: his Improvement Proposer files an improvement initiative itself for every service below target, and every one is refused at intake) and asked whether it should be switched back on. HIS WORDS: "Prometheus's improvement loop be switched back on - yes, but note that improvement is not expansion.. only service owners can expand the service capability.. also only service owners can improve the service capability but prometheus needs to nudge and motivate them to do that through his measurement loops"

READ AS (Orion's reading, not the administrator's words): (1) the loop runs again, end to end, on the estate; (2) IMPROVEMENT IS NOT EXPANSION: a service's capability is expanded only by its owner; Prometheus never proposes new capability for a service he does not own; (3) only a service's owner improves its capability, too: Prometheus files no improvement for another owner's service; (4) Prometheus's part is to measure, and through his measurement loops to nudge and motivate each owner to improve -- the measure, the target and the trend put in front of the owner until the owner acts, and an owner whose service climbs is told so; (5) his own service, quality-management, he improves and expands as its owner, like any owner; (6) it holds for projects as for Carolverse (blueprint #1007): for a project's service, the nudge goes to that service's owner -- 's "files each proposal tagged to the project" reads as "nudges the project's service owner".

Data quality is a service track of its own, with four pillars, kept continuously improving through Prometheus

Ruled 2026-11-19 by ninad -- in force

Ninad: "If the data sources are lame, they need to be fixed, and a process to ensure that the captured data is accurate, complete and appropriate should be put in place, so that data deviations dont happen.. this needs a seperate data quality monitoring service track, which needs to be owned by an agent and a service, and should have its defined tasks and budgets.. the captured data, model and published records should sit in registry tables. Data quality process should check data end to end, from data sources, intermediatory tables upto published reports in apps. The data quality service track should be made re-usable for other projects to leverage it for their data quality initiatives.. the data quality track should include data quality monitoring, measurement, reporting and improvements as 4 standard pillars.. The improvements pillar should tie back to prometheus quality management service, where prometheus should ensure that the data quality of the estate is continuously improving.. prometheus should offer this capability to other projects as well.."

๐Ÿ“–Sales

Glover is a service extension of Sales, never a project

Ruled 2026-09-14 by ninad -- in force

GLOVER WAS NEVER A PROJECT (Ninad, CLI-416, 2026-09-14). It is a service EXTENSION of the SALES service. The estate had already ruled that Migration delivers Glover's capabilities INTO the existing Sales service rather than minting a service or a project for it, while a client-project row went on describing Glover as 'the first customer project the carolverse platform serves'. A record contradicting a standing ruling is the third state to watch for: it agrees with neither, and readers believe whichever they find first. The row is CORRECTED IN PLACE and retained, never deleted โ€” it is why Glover appears in the project history at all, and the superseded wording is kept inside the correction so the trail survives. Glover's use as the jail pilot stays true as history: that is what Glover was used to PROVE, not what Glover WAS.

Sales strategy versions are registry records that say why the strategy evolved

Ruled 2026-09-16 by ninad -- in force

The sales strategy has VERSIONS: it evolves as the business gets traction and the platform evolves, and the versions are registry records (table sales_strategy_versions, one door shared machinery). A version holds every strategy setting as a typed column, the period it was in force and the version it superseded; exactly one version per project is active. Every version after the first says why the strategy evolved - an evolution driver (business traction, platform evolution, market learning or administrator decision; RDM sales_strategy_evolution_driver), a change summary, a rationale, and evidence references for the three evidence-led drivers. A version is never edited in place; a change is a new version. The Sales Strategy store keeps no strategy of its own.

The sales strategy is owned and decided by the Head of Sales, and is defined through the operator CLI

Ruled 2026-09-17 by ninad -- in force

The sales strategy is OWNED BY THE SALES LEAD - Aurora, Head of Sales - and she is the recorded decider of every strategy version. The strategy is DEFINED THROUGH THE OPERATOR CLI: Ninad's direction reaches Aurora from his CLI session, and the version she records says the direction came from him. Carol's chat is NOT a door for defining the strategy. Ninad's words: 'carol sales strategy currently requires me to provide strategy via carol.. that is not the right design.. strategy needs to be defined via cli, and should be owned by the sales lead.' Asked who decides a new version once Aurora owns it, he chose: Aurora decides; his direction reaches her through the CLI and the record says it came from him. Unchanged: versions are registry records that say why the strategy evolved (ruling 20); a version is never edited, so version 1 keeps reading as the administrator's decision it was; Carol's other Sales Strategy operations in her internal chat are untouched; apps stay transparency surfaces (cookbook 1710).

Frankfurt Food is for signed-in Sales subscribers and owned by the Sales service's owner; Carol tells anyone what it holds

Ruled 2026-10-08 by ninad -- in force

Ninad, 2026-10-08 (CLI-738). THE QUESTION ANSWERED: Orion asked whether Frankfurt Food stays public (offered: 'Public' -- the registry is changed to public so both records agree; 'Sign-in only' -- the website asks for a sign-in first). HIS ANSWER: "the app should be accessible to signed-in users who are subscribed to sales service. Who is the current owner of this app? It should be ideally with whoever owns the sales service. Carol should be able to provide this information through her chat to all users (in that sense, the information is available to public, but the app is restricted.. hope that clarifies". READ AS (Orion's reading, marked apart from his words, P.01.01.05.05): the app is reached only by signed-in people whose project subscribes to the Sales service -- its human policy stays 'default' with Sales recorded as its service, and the web server stops serving it without a sign-in; its owner is the owner of the Sales service (Aurora, Head of Sales, today; Carol until now); Carol answers any person in her chat from the same restaurant records -- the restriction is on the app, never on the information. The app's fetch action stays refused from outside (cookbook 1710).

Frankfurt Food's restaurant data is refreshed daily by a Sales-owned process; the app carries no fetch control

Ruled 2026-10-08 by ninad -- in force

Ninad, 2026-10-08 (CLI-738). THE QUESTION ANSWERED (after Rulings Repo sales #331 took Frankfurt Food's fetch control off the app, nothing refreshed its restaurant data -- last refreshed 2026-04-16): Orion asked: "with the button gone, nothing refreshes the restaurant data. Sales' roadmap has no item to file that work under; should I add one?" HIS ANSWER: "done.. also, add item to sales' roadmap to refresh data on daily basis.. let me know if there are cost implications to this.. "fetch new data" button should not be present in the app. The app building skill should be able to identify this as non compliance to the architecture where apps should not be performing user driven actions via the UI". READ AS (Orion's reading, marked apart from his words, P.01.01.05.05): the refresh of Frankfurt Food's restaurant data becomes a Sales roadmap item and runs once a day as a process the Sales service owns, never from the app; its cost is reported to Ninad; the app carries no fetch control. The second half restates cookbook 1710 / P.01.03.01.05 (an app performs no action, its page carries no control that starts work): the app-building skill's no-action gate (H12, the one action-surface checker shared with Albus's build review and Themis's audit) must identify such a user-driven control as non-compliance.

Food recommendations get their own Sales track; the daily refresh is its first task

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-743). THE QUESTION ANSWERED (food recommendations had no service, track or task; the Frankfurt Food app sits with Sales by sales #331 and its daily refresh by #332): Orion asked: "Where should food recommendations (the restaurant data and its daily refresh) be paid from?" (offered: 'A new Sales track (Recommended)' -- a 'Recommendations' track in Sales, owned by Aurora, the daily refresh its first task; 'Sales Core / renamed track'; 'Communications service'). HIS ANSWER: "A new Sales track (Recommended)". READ AS (Orion's reading, marked apart from his words, P.01.01.05.05): the Sales service gains a track for recommendations, owned by the Sales service's owner (Aurora), and the daily refresh of Frankfurt Food's restaurant data is its first task.

The daily restaurant refresh re-reads details only for restaurants new to a district's top ten

Ruled 2026-10-13 by ninad -- in force

Ninad, 2026-10-13 (CLI-743). READ AS (Orion's reading, marked apart from his words, P.01.01.05.05): the refresh searches each district every day and asks Google for a restaurant's full details only when that restaurant is new to the district's top ten; a restaurant already held keeps its stored details. The run's receipt counts the Google calls it made.

Every prospect carries its own budget from the Sales budget, covering lead generation, proposal, prototyping and deal closure

Ruled 2026-11-11 by ninad -- superseded 2026-11-11

Ninad, 2026-11-11 (CLI-106), with prospects added to the client mix (leadership #786): "Every prospect should have budgets which should come from the sales budget, this includes cost of lead generation, proposal, prototyping and deal closure". READ AS (Orion's reading, marked apart from his words): each prospect has a budget of its own, drawn from the Sales service's budget (never another service's, never the client's project once won), in four stages -- lead generation, proposal, prototyping, deal closure -- and its spend is read Clara's plan carries Sales' prospect cost each year from the prospects worked. The amount per prospect is not yet decided.

Ended because: Ninad set the prospect budget at 20% of a micro client's yearly value and made the prospect count follow the Sales budget

Every prospect carries a yearly budget of 20% of a micro client's yearly value from the Sales budget, and the prospects worked are as many as that budget pays for

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-11 (CLI-106): "Every prospect should have budgets which should come from the sales budget, this includes cost of lead generation, proposal, prototyping and deal closure", then "annual prospect budget for each prospect should be 20% annual cost of a micro customer, what do you think?

Prospect work runs in Sandbox, a self-funded project, and moves to the client's own project once the prospect converts

Ruled 2026-11-11 by ninad -- in force

Ninad, 2026-11-11 (CLI-106), shown that none of the first year's six people is a dedicated salesperson and that prospect work would come out of every role's time: "ok, looks like the entire team can work on prospects to begin with, and projects after they convert.. however, this should be distinguished.. we should have a dedicated self funded project for prospects called sandbox, fold this into the pursuit requirements". READ AS (Orion's reading, marked apart from his words): every person may work on prospects and, once a prospect converts, on its project; work on a prospect -- lead generation, proposal, prototyping, deal closure -- is booked to one project of Denken Labs' own named Sandbox never by a client; when a prospect converts, its work moves to the client's own project, which keeps #789's profit; Sandbox is non-project work in #789's sense, so its cost may make a year's loss; hours, tokens and spend show Sandbox and client projects apart.

A build Leo files for a prospect is prototyping

Ruled 2026-11-16 by ninad -- in force

Ninad (CLI-185) chose: Always prototyping (Recommended). Orion asked: every piece of prospect work names one of four sales stages (lead generation, proposal, prototyping, deal closure; sales #788). When Leo files a build for a prospect like Chowpatty, before their contract is signed, which stage should it carry? Options: always prototyping (Orion recommended: anything Leo builds for a prospect before the contract is prototyping, which the stage set defines as building and showing the prospect s proof of concept), or the asker picks each time. READ AS: Leo s channel names sales stage prototyping on every module it files for a project at stage prospect; nobody chooses it per filing.

Carol routes a client's answers to Blueprint; Leo never asks them again

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "2. relay - carol asked a lot of questions to shashank.. the requirements were captured, but they were not relayed to Leo.. this is risky, because clients dont like to repeat their requirements.. when answers are given, they should be captured by carol and given to the blueprint service.. this link is currently broken.. the problem is - carol does not understand when to pass the baton to Leo.. Leo is trained to ask the right questions, while carol asks the right questions but does not have the wiring to ask the questions that blueprint service needs for building a solution.. Carol should act as the router here.. Leo should be able to pick all requirements from carol's conversation and "not ask the same questions to clients.." (this is important)". READ AS (Orion, not Ninad's words): what a person tells Carol about what they need is captured, tied to that person, and handed to the Blueprint service when they are ready for an agent of their own; Carol routes, Leo asks Blueprint's questions, and Leo reads everything Carol holds before asking and never asks again what she already has.

Carol engages clients and escalates; she never initiates a build

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "4. client relations - carol's role is sales and client engagement, she does not have the feed from various services that leo has, so she will not be able to build the solution.. she needs to know what is the good time to handoff, what level should she engage with and where should she engage Leo with the client.. her job is client engagement, which means, to ensure that Leo is servicing properly and to escalate if the client is not happy.. this was not apparent from her conversation with Shashank.. carol should never initiate the build, because she cannot..". READ AS (Orion, not Ninad's words): Carol knows when to hand a client to Leo and at what level to stay in the conversation; after the hand-off she checks that Leo is serving the client well and escalates when the client is not happy; she never starts a build.

Carol works only for Carolverse; Leo builds leased agents for clients

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "5. job distinction - Carol never works for other clients, she only works for carolverse.. she can do "favours" for the users if she can, but only to "demonstrate" the agentic capabilities that she has.. ultimately, the users have to become customers of carolverse as self service users, or prospects if they want "dedicated FDE support". Carolverse will build new agents for the user who will work for them. Note that the new agents are carolverse agents, and will always remain carolverse agents, but they will be rented or leased to the client to do their work.. they will be deployed to the prospect or self serve environment of the user where they will work for the user.. Leo will build those custom built agents for the user, and hence its important that Leo is brought in the picture at the right time.. when the user is ready to "have their own carolverse agent". Users can retire the agents and they return back to the "bench" area from where they are deployed to another similar role for another client.. thats how the blueprint service works. Its important that carol has the knowledge of the blueprint service.. she should be able to communicate with Leo and pick information if needed about the blueprint service and tell the client.. she should introduce Leo to clients as the "owner of blueprint service" who will guide them to build their own agents.. remember, currently, blueprint is the only client facing service, but soon there will be more client facing services from carolverse (investments from clara is next in line)". READ AS (Orion, not Ninad's words): Carol works for Carolverse alone, and a favour she does for a user only demonstrates what agents can do; a user becomes a self-serve customer or a prospect; the agents Leo builds for them stay Carolverse agents, leased and deployed to the user's environment, and return to the bench when retired; Carol knows the Blueprint service, can ask Leo about it, introduces Leo as its owner when the user is ready for an agent of their own, and does the same for each client-facing service that follows.

Requirements are collected a few at a time, with a delivery in between

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "6. staggered approach - clients dont like to give a lot of requirements at once.. first they want to test the products, so the approach for requirement collection should be staggered.. max 5 or 6 questions at a time.. then deliver something via Leo, and then continue the conversation.. either Leo collects further requirements and if carol collects them, she relays them to leo to act upon them... collecting lots of requirements without any tangible delivery is never appreciated by clients..". READ AS (Orion, not Ninad's words): whoever collects a client's requirements asks at most six questions before something is delivered through Leo; further questions follow that delivery, and what Carol collects afterwards she relays to Leo to act on.

Carol's chat is built following GERMINATE

Ruled 2026-11-22 by ninad -- in force

Ninad (CLI-329): "decipher these 6 requirements into policies, rulings, protocols, standards etc, update dictionary with any new words as relevant.. fold all the requirements in carol chat pursuit and ensure that carol's chat is built following the germinate methodology". READ AS (Orion, not Ninad's words): each requirement becomes a step of CAROL-PUR-0095; each is recorded in GERMINATE as an observed event placed in exactly one of its segments, and the code that serves it carries that segment's marker (cookbook 1376); the policy and dictionary words wait on Ninad's approval (governance #355).

๐Ÿ“–System Services

One dominion per machine

Ruled 2026-09-14 by ninad -- superseded 2026-09-14

A DOMINION CAN NEVER SHARE A VM WITH ANOTHER DOMINION (Ninad, CLI-416). A dominion is a separate installation in the ordinary sense โ€” its own machine, its own records, its own money and its own agents โ€” so two dominions on one host would put two estates inside one blast radius, one set of OS identities and one budget meter. The rule refuses at the WRITE: the dominion register holds the host as a unique value, so a second dominion claiming an occupied machine is rejected rather than recorded and then asserted in prose nobody executes.

Ended because: Ninad ruled every project gets its own VM, so a dominion spans a bunch of machines rather than one; the no-sharing rule survives unchanged

Every project gets its own machine; a dominion spans many

Ruled 2026-09-14 by ninad -- in force

EVERY PROJECT GETS ITS OWN MACHINE, AND A DOMINION SPANS A BUNCH OF THEM (Ninad, CLI-416, 2026-09-14). A project used to be isolated by a jailed OS identity and workspace on the dominion's shared host โ€” the kernel was the boundary, the machine was not. The machine is now the boundary: individual projects ALWAYS get a separate VM. So a dominion is not one box. The host recorded on a dominion is its BASE project's machine; every further project adds one, and the dominion's real footprint is derived from its projects rather than stored, so the two can never disagree. What survives unchanged from 'one dominion per machine' is the part that always mattered: NO MACHINE IS EVER SHARED โ€” not between two dominions, and not between two projects. Both are enforced at the write, and a host already held by anything else is refused by name. A merely REQUESTED project is not owed a machine; approved and active ones are.

One dominion per machine

Ruled 2026-10-01 by ninad -- retired 2026-10-01

A DOMINION CAN NEVER SHARE A VM WITH ANOTHER DOMINION (Ninad, CLI-416). A dominion is a separate installation in the ordinary sense โ€” its own machine, its own records, its own money and its own agents โ€” so two dominions on one host would put two estates inside one blast radius, one set of OS identities and one budget meter. The rule refuses at the WRITE: the dominion register holds the host as a unique value, so a second dominion claiming an occupied machine is rejected rather than recorded and then asserted in prose nobody executes.

Ended because: Recorded by mistake 2026-10-01 17:40 (CLI-638 ran the one-off shared machinery expecting a --help): this is the wording that ruling #10 (CLI-416, 2026-09-14) already superseded as #6 -- a dominion spans many machines; no machine is ever shared. Retired in place; #10 governs. (by ninad)

One dominion per machine

Ruled 2026-11-15 by ninad -- retired 2026-11-15

A DOMINION CAN NEVER SHARE A VM WITH ANOTHER DOMINION (Ninad, CLI-416). A dominion is a separate installation in the ordinary sense โ€” its own machine, its own records, its own money and its own agents โ€” so two dominions on one host would put two estates inside one blast radius, one set of OS identities and one budget meter. The rule refuses at the WRITE: the dominion register holds the host as a unique value, so a second dominion claiming an occupied machine is rejected rather than recorded and then asserted in prose nobody executes.

Ended because: recorded by accident: CLI-157 ran shared machinery expecting its usage text, and the script re-adds any of its rulings not live -- this repeats #6, superseded by #10 (re-recorded as #266 and retired); Ninad made no new ruling (by ninad)

A project's registry store stays on that project's own machine, and a Sandbox prospect's on Sandbox's machine, for now

Ruled 2026-11-17 by ninad -- in force

A PROJECT'S REGISTRY STORE STAYS ON ITS OWN MACHINE FOR NOW (Ninad, CLI-198, 2026-11-17; settles the point governance #968 left open). THE QUESTION: "Your ruling this afternoon made the registry one Carolverse service with a separate store for each project and Sandbox prospect, but left open which machine each store sits on. This morning's build put each prospect's store on the Sandbox machine, and the Sandbox pursuit's next two steps build on that. Where should each project's registry store sit?" Options shown: 'On Carolverse (Recommended)' -- every store on Carolverse's machine beside the one registry, reached through one door that knows which project is asking; and 'On each project's machine' -- "Each store stays on its own project's (or Sandbox's) machine, as built this morning; Carolverse's one registry reads and writes it there through a narrow door, and no registry code runs on the project's machine. Keeps a project's data on its own machine. Cost: every read crosses machines, and each machine needs that door." HIS FIRST ANSWER: "hagdrid needs to decide how carolverse resources are allocated across the services, so that the performance and accessibility is maintained.. this will additional build and design, for now, keep it in carolverse vm, but in future, we will need more carolverse vms, and that will need the ability to intelligently allocate resources across services of carolverse". THEN, BEFORE ANYTHING WAS RECORDED, HE CHANGED IT: "I would like to change my answer - for now - keep the location to respective vms.. maybe hagrids algorithm will use that as guiding principle, but we will cross the bridge when we get there". ONLY THE CHANGED ANSWER RULES.

READ AS (Orion's reading, not the administrator's words): (1) each project's registry store sits on that project's own machine, and each Sandbox prospect's on Sandbox's machine, as design 728 section 11 and built it; (2) Carolverse's one registry (governance #968) reaches a store there, and no copy of the registry's code runs on a project's machine for it; (3) and (CAROL-PUR-0042) go ahead as designed; (4) how Carolverse allocates its resources across its services as it grows onto more machines is Hagrid's to decide LATER -- not built now ('we will cross the bridge when we get there'); a store kept on its own project's machine may be one of the principles his allocation follows, which is his and the administrator's call then; (5) 'for now' is the administrator's own word and stands as a dated decision under this ruling, never as policy text (policies are time-agnostic).

Carolverse's estate extends across several machines without a hand step, Hagrid manages its resources across them automatically, and a new machine is procured only when Hagrid asks and the administrator approves

Ruled 2026-11-17 by ninad -- superseded 2026-11-17

NO SECOND MACHINE NOW; BUILD WHAT LETS THE ESTATE SPREAD ACROSS MACHINES, AND HAGRID ASKS FOR A MACHINE WHEN ONE IS NEEDED. THE QUESTION: "7409 (a busy shared service runs on more Carolverse machines behind one address) has no existing parts to build on: no front door and no way to add a Carolverse machine on demand. Hagrid's machine maker can make one, but proving the first criterion for real needs an actual second Carolverse machine in Azure. That's a monthly cost, plus one core step from you to put the machine on root's register. Today no service is near carol-vm's limits. ... How should I proceed?" Options shown: 'Design now, build tomorrow (Recommended)', 'Approve a 2nd machine', 'Park 7409 for now'. HIS WORDS: "if the need for another machine is not urgent, lets not procure it now, but implement the design that is necessary to seamlessly extend the estate's resources across multiple vms and enable hagrid to manage the resources across the vms is an automated manner.. he will still need admin approval to procure the machine, but the request and decision should come from him..".

READ AS (Orion's reading, not the administrator's words): (1) no second Carolverse machine is procured now, because no service needs one today; (2) builds what lets the estate's services extend across several Carolverse machines with no hand step -- a record of which machines carry each service, one front door so a project keeps one address for a service, and each service's load read per project and per machine; (3) Hagrid manages the estate's resources across those machines automatically: he decides where a service runs and when the estate needs another machine; (4) procuring a machine needs the administrator's approval -- the request, and the decision to make it, come from Hagrid, and the administrator approves or declines it; the machine is then made by Hagrid's machine maker and entered on root's register at core, as today; (5) system-services #972's ruled store locations stand; its reading (4), that Hagrid's allocation is decided later and not built now, was Orion's reading and is overtaken by this ruling; (6) 7409's criterion that a service 'also runs on another Carolverse machine' cannot be proven without procuring one, so its success criteria are to be revised, with the administrator's approval of the new words, to what is proven without a second machine.

Ended because: Ninad amended his answer in the same sitting: Hagrid researches and proposes the machine, Marco procures after the administrator's approval, and Hagrid's infra app shows every machine

Carolverse's estate extends across several machines without a hand step; Hagrid manages its resources across them automatically, researches and proposes the best-fit, economical machine when one is needed, the administrator approves, Marco procures, and Hagrid's infra app shows every machine

Ruled 2026-11-17 by ninad -- in force

NO SECOND MACHINE NOW; BUILD WHAT LETS THE ESTATE SPREAD ACROSS MACHINES, AND HAGRID ASKS FOR A MACHINE WHEN ONE IS NEEDED. THE QUESTION: "7409 (a busy shared service runs on more Carolverse machines behind one address) has no existing parts to build on: no front door and no way to add a Carolverse machine on demand. Hagrid's machine maker can make one, but proving the first criterion for real needs an actual second Carolverse machine in Azure. That's a monthly cost, plus one core step from you to put the machine on root's register. Today no service is near carol-vm's limits. ... How should I proceed?" Options shown: 'Design now, build tomorrow (Recommended)', 'Approve a 2nd machine', 'Park 7409 for now'. HIS WORDS: "if the need for another machine is not urgent, lets not procure it now, but implement the design that is necessary to seamlessly extend the estate's resources across multiple vms and enable hagrid to manage the resources across the vms is an automated manner.. he will still need admin approval to procure the machine, but the request and decision should come from him..".

READ AS (Orion's reading, not the administrator's words): (1) no second Carolverse machine is procured now, because no service needs one today; (2) builds what lets the estate's services extend across several Carolverse machines with no hand step -- a record of which machines carry each service, one front door so a project keeps one address for a service, and each service's load read per project and per machine; (3) Hagrid manages the estate's resources across those machines automatically: he decides where a service runs and when the estate needs another machine; (4) procuring a machine needs the administrator's approval -- the request, and the decision to make it, come from Hagrid, and the administrator approves or declines it; the machine is then made by Hagrid's machine maker and entered on root's register at core, as today; (5) system-services #972's ruled store locations stand; its reading (4), that Hagrid's allocation is decided later and not built now, was Orion's reading and is overtaken by this ruling; (6) 7409's criterion that a service 'also runs on another Carolverse machine' cannot be proven without procuring one, so its success criteria are to be revised, with the administrator's approval of the new words, to what is proven without a second machine.

AMENDED THE SAME SITTING (Ninad, CLI-205, 2026-11-17). THE QUESTION: Orion proposed 7409's revised success criteria: "1) Each shared service's machines are kept in one record, and the service's single front door sends every project's requests to the machines on that record. A project keeps one address however many machines there are, and a machine added to the record is reached with no hand step. 2) (kept as is) Operators see, for each shared service, the load each project puts on it and how many machines carry it. 3) When a service's load would pass what its machines can carry, Hagrid himself asks you to procure a machine, naming the service, its load and the monthly cost. No machine is procured until you approve. ... Do you approve these words?" HIS WORDS: "I approve, but enable hagrid to do the necessary research and propose the best fit and economic machines to procure.. actual procurement happens via marco after admin's approval. Hagrid's infra app should display all machines, irrespective of whether the belong to platform or projects, with a filter to view platform machines or project machines".

READ AS for the amendment (Orion's reading, not the administrator's words): (a) the three criteria are approved, criterion 3 widened: Hagrid researches and proposes the best-fit, most economical machine (size and monthly cost) for what the load needs, and asks the administrator; (b) procurement is Marco's (Cost Center), and he procures a machine only on a request the administrator approved -- this replaces reading (4) above that the machine 'is then made by Hagrid's machine maker' as the procurement step (Hagrid's maker still makes and keeps the machine, and root's register still records it at core); (c) Hagrid's infrastructure app (Carol Infra Monitor, Backup service) shows every machine, the platform's and every project's, with a filter for platform machines or project machines; (d) under the one-service rule (initiatives #383, #391) (b) and (c) are each filed as their own initiative in CAROL-PUR-0084, beside 7409 (System Services).

Hagrid's headroom limit is 80% for 6 hours, and a new machine is his last resort: he first cleans up and manages memory himself, then files a high-priority initiative to Albus and tells the estate operators' Orion instances

Ruled 2026-11-18 by ninad -- in force

HAGRID'S HEADROOM LIMIT, AND A NEW MACHINE ONLY AS THE LAST RESORT.

THE QUESTION: Orion asked which headroom limit Hagrid's capacity watch should use -- "when a machine's processor or memory stays above it for long enough, he researches the cheapest machine that fits and asks you to approve it" -- showing carol-vm's last 7 days (processor median 39%, 90th percentile 69%, one stretch of 9 hours at or above 80%; memory median 60%, 90th percentile 75%; swap almost always full) and three options: 80% for 6 hours (recommended), 75% for 6 hours, 85% for 12 hours. HIS WORDS: "80% for 6 hrs, but he tries everything he could to cleanup the disk or manage the memory, and either files a "high priority" initiative to albus and informs orion instance of the estate operator.. after all efforts are futile, then he resorts to new vm as the last option".

READ AS (Orion's reading, not the administrator's words): (1) the headroom limit is 80% of a machine's processors or of its memory, sustained for 6 hours (design 730's declared ceiling and window, CAROL_CAPACITY_CEILING_PCT=80 and CAROL_CAPACITY_WINDOW_HOURS=6); (2) crossing it does NOT first raise a machine request: Hagrid first does everything his own lane can -- clean up the disk by his stated retention rules and manage memory (find what holds it and have it released through the doors that exist, never by killing a process) -- and, when that does not bring the machine back under the limit, files a HIGH-PRIORITY initiative to Albus naming the machine, the load, the services carrying it and what he already tried, and tells the estate operators' Orion instances by operator communication; (3) only when those efforts are futile -- the machine still above the limit after his own work and after Albus's initiative has run its course -- does Hagrid raise the machine request to the administrator (best-fit, most economical size from the live price list, as #977 and design 730 section 4 say); (4) 'either ... and' is read as both: the Albus initiative AND the notice to the estate operators, at the same step; (5) the re-ask step (a declined request is not raised again until the load rises further) stays as design 730 states, 10 points, a declared value Orion proposed and the administrator did not change.

Hagrid reads each Azure machine size's processors and memory from Azure's size catalogue, with the machine's own read-only Azure identity

Ruled 2026-11-18 by ninad -- in force

HAGRID ASKS AZURE FOR THE SIZES, WITH THE MACHINE'S OWN IDENTITY.

THE QUESTION: Orion asked how Hagrid, at the last rung (asking the administrator for a new machine), should learn each Azure size's processors and memory, which the public price list does not carry -- "From size names (Recommended): read processors and memory from Azure's documented naming (e.g. Uses no Azure credential at all." or "Ask Azure with machine ID: Hagrid uses the machine's own read-only Azure identity to read Azure's size catalogue. Exact for every size, but it puts an Azure credential into an agent's process." -- and said the safety check had stopped Orion from trying the second from his session. HIS ANSWER: "Ask Azure with machine ID".

READ AS (Orion's reading, not the administrator's words): (1) Hagrid's Capacity Watch, under his own login, obtains a token from the machine's own Azure identity (the instance metadata endpoint) for the management API, uses it for ONE purpose -- reading the compute size catalogue (processors, memory, restrictions) for the machine's region -- and keeps it only in memory for that call: never written, logged, cached or passed on; (2) the identity stays read-only (its role is not widened for this); (3) prices still come from the public retail price list, read live; (4) a catalogue or price list that cannot be read raises no request and says why (design 730 section 4).

Machines are procured as and when needed, by the method sandbox-vm was procured and configured, on root's approval; Hagrid keeps every machine by kind -- Carolverse, Sandbox, self-serve, project -- with whom it serves, and forecasts each kind's need

Ruled 2026-11-18 by ninad -- in force

MACHINES ARE PROCURED AS NEEDED, THE SANDBOX WAY, ON ROOT'S APPROVAL; HAGRID KEEPS AND FORECASTS EVERY KIND.

THE QUESTION: Orion asked how Marco should buy an ESTATE machine once the administrator approves one of Hagrid's machine requests, since Hagrid's Machine Maker makes only project machines (one approved size, a fenced resource group) and refuses the estate's own; options: the administrator buys and Marco records it; Marco gets an Azure identity of his own; Hagrid's maker widened to the estate's group. HIS WORDS: "buying a vm is part of the project onboarding workflow currently.. in future, hagrid should keep few VMs ready and tested, so that the project onboarding is quick.. this depends on forecasting how many and when the VMs will be required by looking on the updates from the sandbox and provisioning them.. for now, we only provision VMs as and when needed.. this applies for carolverse VMs just like for project VMs.. Hagrid clearly maintains the distinction between carolverse VMs, Sandboxes, self serve VMs and project VMs and keeps a log of what VMs are for which users, prospects, projects in his app. VM procurement requires root approval. Process of procuring the machine follows the same method how sandbox was procured and configured.. Hagrid also maintains the list of VMs across the estate and their usage and forecasts when more VMs are needed for self-serve, sandbox and dedicated project VMs just like he does it for platform VMs (carolverse).."

READ AS (Orion's reading, not the administrator's words): (1) NOW a machine is provisioned only as and when it is needed -- for Carolverse exactly as for a project; (2) a Carolverse machine is procured BY THE SAME METHOD sandbox-vm was procured and configured (CLI-138/147/148: an entry on root's register, Hagrid's Machine Maker making it in Azure, the administrator's root step that records and joins it), never by a second method; (3) procurement requires ROOT's approval: the administrator approving Hagrid's request is that approval only when given at root, as the sandbox's was; (4) Hagrid keeps FOUR kinds of machine apart -- Carolverse (platform) machines, Sandboxes, self-serve machines and project machines -- with a log of which users, prospects and projects each serves, shown in his infrastructure app; (5) he keeps the list of every machine of the estate and its usage, and forecasts when more of each kind will be needed, as he does for Carolverse's (design 730's capacity work); (6) LATER, not now: Hagrid keeps a few machines ready and tested so onboarding is quick, sized by forecasting from the Sandbox's updates. Not decided here: what Marco's part is -- system-services #977 names procurement 'via Marco'; this ruling sets the METHOD (the sandbox's, on root's approval) and is read WITH #977, not over it, until is re-scoped.

Marco's part in buying a machine is keeping the books: Hagrid makes it the Sandbox way on root's approval; Marco checks it matches an approved request, refuses any other by name, and enters it in his register with its monthly cost

Ruled 2026-11-18 by ninad -- in force

MARCO KEEPS THE BOOKS FOR EVERY MACHINE BOUGHT.

THE QUESTION: "Next pursuit step: 7423, 'Marco buys a machine only on a request Hagrid made and you approved'. Your ruling tonight says a machine is bought the same way the Sandbox machine was (Hagrid's Machine Maker makes it, and you approve at root). An earlier ruling says buying goes through Marco. What should Marco's part be?" -- options: Marco keeps the books (recommended: Hagrid makes the machine the Sandbox way once the administrator approves at root; Marco checks it matches an approved request, refuses anything else by name, and enters it in his register with its monthly cost so Company Accounts counts it); Marco places the order (Marco asks Hagrid's Machine Maker to make it for an approved request, then records it); stop for tonight. HIS ANSWER: "Marco keeps the books".

READ AS (Orion's reading, not the administrator's words): (1) the machine is MADE by Hagrid's Machine Maker by the Sandbox method on the administrator's approval at root (system-services #1037), never by Marco; (2) Marco's part is the books: he checks that a machine made matches a machine request Hagrid raised and the administrator approved (shared.machine_request: state approved), refuses any other by name, and enters it in his register with its monthly cost (the price the request carried, read again from the live price list when it is entered) so Company Accounts counts it; (3) #977's 'actual procurement happens via marco after admin's approval' is read through this: procurement's record and spend are Marco's, the making is Hagrid's. 's criteria are revised to say so.

A machine request offers the administrator three machine choices driven by the requirements, each with its reasons, and he selects one; machines are speculative and ordered in advance, and may be one standard set for every kind

Ruled 2026-11-19 by ninad -- in force

HAGRID OFFERS THREE MACHINE CHOICES WITH REASONS AND THE ADMINISTRATOR SELECTS ONE; MACHINES ARE ORDERED IN ADVANCE AND MAY BE ONE STANDARD SET FOR EVERY KIND.

THE QUESTION: "Where in Azure should Hagrid make Carolverse's own extra machines? Today he can only make machines in the client-projects group, and its rule allows one small size. Either way, every machine still needs your approval and your root step, as sandbox-vm did." -- options: New group (a separate group just for Carolverse machines, its size rule holding only the sizes he approves) or Client-projects group (each approved size added to that group's size rule).

HIS ANSWER: "what is a group? also, the exact configuration of the machines should be driven by the requirements, hagrid should provide 3 choices for the admin to select 1 of them, giving the right reasons.. remember these are speculitive machines ordered in advance, so it may very well be a standard machine set used for all projects, platform, self service hostings, sandboxes etc"

READ AS (Orion, not his words): a machine request carries three choices, each with the reasons it fits the requirements, and the administrator's answer selects one of the three; Hagrid's Machine Maker makes exactly the selected one. The group question stays open (he asked what a group is). Whether 'ordered in advance' amends #1037's 'as and when needed' is asked of him in the same session (P.01.01.05.08).

Every machine Hagrid makes, of every kind, sits in one Azure resource group whose size rule holds exactly the standard set the administrator approves

Ruled 2026-11-19 by ninad -- in force

ONE SHARED AZURE FOLDER FOR EVERY MACHINE HAGRID MAKES.

THE QUESTION: "An Azure 'group' (resource group) is a labelled folder in Azure that holds machines, their networks and their disks. Permissions and size rules are set on the folder. Hagrid's identity may create machines only inside one folder today (the client-projects one), and that folder's rule allows only one small size. Since you said the machines may be one standard set for every kind (platform, projects, self-service, sandboxes), should all of Hagrid's machines share ONE folder whose size rule holds exactly that standard set?" -- options: "One shared folder (Recommended): Every machine Hagrid makes, of every kind, sits in one folder. Its size rule holds the standard set you approve, and you widen it only as you approve. Hagrid keeps each machine's kind in his own record." or "One folder per kind: Separate folders for platform, project, self-service and sandbox machines, each with its own size rule."

HIS ANSWER: "One shared folder (Recommended)".

READ AS (Orion, not his words): Hagrid's machines of every kind share one resource group; its Azure size rule is widened only by the administrator, only to sizes he approved; the kind of each machine is Hagrid's record (#1037), never the folder.

For now a machine is ordered only when needed, when Hagrid's capacity limit is crossed; ordering ahead of his forecast for a kind comes later, once his forecasts exist

Ruled 2026-11-19 by ninad -- in force

MACHINES WHEN NEEDED NOW, AHEAD OF FORECAST LATER (Ninad, CLI-246, 2026-11-19; settles system-services #1045's 'speculative machines ordered in advance' against #1037's 'as and when needed', P.01.01.05.08).

THE QUESTION: "Your earlier ruling (18 Nov) said machines are procured 'as and when needed'. Today you said they are 'speculative machines ordered in advance'. Which should stand?" -- options: "Ahead of forecast need: Hagrid asks for a machine ahead of time, when his forecast for a kind says it will soon run short. You still approve each one. This replaces 'as and when needed'." or "Both, in sequence: For now only when needed (his capacity limit is crossed); ordering ahead of forecast comes later, once his forecasts exist."

HIS ANSWER: "Both, in sequence".

READ AS (Orion, not his words): #1037's 'as and when needed' stands now; ordering ahead of forecast need is the later stage, once Hagrid's per-kind forecasts (#1037) exist; #1045's three choices with reasons bind both stages.

Rarely used apps start on demand and stop when idle, so the machine's memory serves the apps in use

Ruled 2026-11-22 by ninad -- in force

RARELY USED APPS START ON DEMAND (Ninad, CLI-324, 2026-11-22). carol-vm froze from memory exhaustion at 06:42Z and was restarted; six hours later swap was full again and operator links dropped. He chose, in his words, "Start on demand (Recommended)": idle apps stop after about 30 minutes without use and start automatically on their next visit (a first visit waits a few seconds); nothing is lost; one root step changes how apps are started; plus slim Token Cost Tracker; no machine upgrade. So: each app's run mode (always-on or on-demand) is a governed record; the public and critical apps stay always-on; anything that checks an app reads a sleeping app's listener and never wakes it.

Source: Services Catalogue ยท Public information reflected here.