📖About

DISTINCT FROM CAROL-INI-3451, which is in review with its criteria met. 3451 settled ONE corpus: it stopped the induction reciting the whole cookbook and gave it a live-law set plus a named index (227k -> 67k). It did not touch the other two thirds of the cost, and it built no search.

MEASURED NOW, after 3451: session start still costs ~170k tokens and ~25 minutes. The laptop handover index is 77k of it, the cookbook set 67k, the induction document itself only 12k. 3451 proved the principle on one corpus; this applies it to all of them and adds the retrieval 3451 assumed.

THREE TIERS. CREED: the standing rulings that change judgment before I act, read in full every session, because I cannot search for a rule I do not know exists - this is the safety boundary and it stays resident. BOARD: the state of the estate from ONE live query, replacing three handover essays that are stale the hour they are written. LIBRARY: the cookbook, designs, policies and 156 past sessions, searched semantically with EXACT lookup alongside for law, because a miss on a binding rule reads as confident correct work rather than a vaguer answer.

Ninad's goals, in his words: faster induction, fewer tokens, session performance the same or better.

⚖️Decisions

• Elrond's bypass methodology checklist (a reminder, not a gate -- you've got this): 0. File it requested_mode='bypass' (planner-vs-bypass is a deliberate choice). bypass_start REFUSES a non-bypass initiative (CAROL-INI-1846), and the dispatcher only skips the bypass lane when the mode says bypass -- a 'planner' mistag lets Merlin's pipeline grab the placeholder step and block your finished work. 1. Filed as planned status -- let the bypass claim/activate it; never file active. 2. Open the bypass (bypass_start) with your droid id + the remediation answer (remediates_initiative_id=NNN, or remediates_nothing=True). 3. Work the blocks for your work-type: template -> design -> code -> test -> review. Do the real work; record decisions on the initiative as you make them. 4. Reality is recorded for you at close -- code (files changed), each decision, and the twin-review verdict become real activities tied to this initiative and show in the Activity Tracker like a planner run (CAROL-INI-1840). No dummy rows. 5. Keep the initiative status moving; it parks in 'reviewing' and is tagged uat-pending for you at close (CAROL-INI-1836), so the stuck-watchdog leaves it alone until UAT. 6. Close runs the gates (design/architecture compliance + caller-audit). If a gate flags something pre-existing or unrelated to your change, waive it with a clear written rationale -- audit, don't skip. 7. Bypass skips the planner's auto-orchestration, NOT the standards. Same template checklist, same review, same observability as a planner run. (elrond)
• [status-router] planned -> executing | event=bypass_executing | bypass transition (or-bx-01)
• [status-router] executing -> reviewing | event=dispatcher_transition | dispatcher state change (ds-s1)
• [status-router] reviewing -> closed | event=operator_signoff | Auto-accepted (CAROL-INI-1859): Orion-initiated, >2 days in reviewing with no objection. (el-srac-01)

✅Success criteria

• Starting a session costs under 20k tokens and under 5 minutes, measured the same way before and after. (must_have)
• Every standing ruling Ninad has made sits on one page I read in full each session, and that page names where everything else lives. (must_have)
• The state of the estate at session start comes from one live query, never from prose written in an earlier session. (must_have)
• I can find any past session, cookbook rule, design or policy on demand, with none of them loaded up front, and a rule can be reached exactly by its number as well as by meaning. (must_have)
• The handover index cannot grow past its budget again, because the cap is applied when a line is written. (must_have)