DEFERRED 2026-05-01T18:30 per phase-fit reversal. This memory file records the Ω.η ★ lock but the lock has been DISCARDED. InheritKitBench commitment was prematurely framed as Phase 3 work; correct phase is Phase 5 (AI-vendor commercial / closure prep). The original content below is preserved for forensic continuity + re-use when Phase 5 opens. Do not act on this memory’s recommendations as if locked.
Outcome: Rich-locked Ω.η ★ on 2026-05-01T18:15 BST.
A-numbered amendment: A-146 (Tier-1 arch-state v3.31 → v3.32).
THIRD consecutive ζ.2 above-threshold decision (joins Σ.κ 0.27 / Σ.μ.ε.5 0.28 / Λ.η 0.63 in above-threshold streak; margin Ω.η ★ vs Ω.γ A+B+C = 0.43).
Trigger: A-141 verificationTier per row + A-142 Π.θ ★ Skills marketplace quarterly score-card + A-144 Σ.μ.ε.5 Premium £25/mo “InheritKitBench-validated authoring” feature + A-145 Λ.η Premium tier gate each implicitly assume InheritKitBench v0.1 as Phase-1 deliverable; formal commitment + scope + budget never explicitly locked. Future AI-vendor procurement pricing dependency. 9-12 month leapfrog window urgency per T47 §B.
Locked configuration:
(1) Phase A firm Phase-1 deliverable:
- UK&W succession 25 statute→Catala/Cedar pairs
- £30-50K budget; 3 person-months sole-author TT (Paul-equivalent rate ~£10K/PM)
- Ship Q1 2027 (Phase-1 close window)
- ICAIL 2027 Singapore June paper authorship from Phase A scope (submission deadline ~Dec 2026; Stanford CodeX co-author target)
- Vals.ai LegalBench leaderboard listing on ship
- AI-vendor pricing v0.1 against Phase A scope only
- Verification methodology = TLAi+Bench 4-metric scoring per T39 finding 4 (precision / explainability / formalisability / cross-jurisdiction transfer)
(2) Phases B-F formally deferred Phase-1.5+ trigger-gated:
- Phase B Scotland (10 pairs / 1.5 PM): no academic gate; TT-author-tractable; defers for capital-prudence
- Phase C Switzerland (10 pairs / 1.5 PM): no academic gate; TT-author-tractable; defers for capital-prudence
- Phase D Faraid (20 pairs / 3 PM): gated on IIUM + UUM signed co-authorship
- Phase E HSA (10 pairs / 1.5 PM): gated on IIT + Bar-Council India signed co-authorship
- Phase F Halakhic (5 pairs / 1.5 PM): gated on Bar-Ilan or Hebrew University signed co-authorship
- One confirmed academic lock = Phase-1.5 trigger satisfied for that phase
(3) Academic outreach kickoff 2026-06-01 (3-track parallel + universal backup):
- richard-task #220: IIUM + UUM Faraid outreach
- richard-task #221: IIT + Bar-Council India HSA outreach
- richard-task #222: Bar-Ilan + Hebrew U Halakhic outreach
- richard-task #223: Stanford CodeX universal academic anchor backup + ICAIL 2027 co-author intro (target 2026-06-15)
- richard-task #224: Vals.ai leaderboard submission prep (target 2026-12-01)
(4) Standards-track integration:
- Σ.μ.ε.5 Premium £25/mo tier “InheritKitBench-validated authoring” feature gate per A-144
- Π.θ ★ 9 Anthropic Skills marketplace quarterly Skills score-card per A-142
- A-141 verificationTier per row maturity benchmark
- Future AI-vendor procurement gating
- Vals.ai LegalBench leaderboard institutional credibility per Stanford NeurIPS 2023 trajectory
- ICAIL 2027 Singapore June peer-reviewed publication
- PNAS 2026 paper trajectory parallel (“There’s No Free Benchmark: An Institutional View of Legal AI Benchmarking” Guha+Zhang+Tsang+Manning+Nyarko+Ho forthcoming)
(5) Sole-author Phase A faith-tradition prohibition: Per T47 §finding 36, sole-author Faraid/HSA/Halakhic publication “unlikely to penetrate” academic gatekeepers; Phase A scope DELIBERATELY constrained to UK&W to avoid burning capital on benchmark whose faith-tradition rows would be rejected by very academic gatekeepers needed to anchor credibility. Self-defeating Ω.δ full-sole-author rejected on this basis.
Sensitivity: 8-option scoring; Ω.η ★ 4.46 Rich-pick vs Ω.γ A+B+C 4.03. Margin 0.43 strongly above 0.20 threshold. 6/8 perturbations preserve. Ω.γ wins under international+faith re-weight (already lower margin). Ω.β wins under hard ≤£40K Phase-1 cap.
Susskind 2008+2023 dual-grounding: §4.6 [Q-SUSS-78B2A8E807] user-generated legal wikis prediction (InheritKitBench fulfils 2008 prediction precisely) + §17 [Q-SUTL-R7Z0R90MNH] Big-4/TR/RELX acquirer-narrative anchor + Vals.ai precedent (Vincent valuation despite minimal direct revenue).
LegalBench precedent: Stanford NeurIPS 2023; 162 tasks / 40+ contributors; six reasoning categories; live leaderboard at vals.ai; PNAS 2026 paper forthcoming confirms institutional credibility trajectory continues 18+ months post-launch. InheritKitBench v0.1 follows established trajectory.
9-12 month leapfrog window: per T47 §B (round-6 analysis 2026-04-25); Harvey-Spellbook competitive-response trigger remains unobserved; window now ~9-11.5 months (T+6 days from T47 authoring); decision urgency = lock + start authoring v0.1 by Phase-1 sprint S5 (Q3 2026) to ship Q1 2027 BEFORE one Anthropic v2 plugin / Harvey M&A / Stanford CodeX research output closes window inside 6-9 months.
5 NEW richard-tasks: 220-224 (3-track Faraid/HSA/Halakhic outreach + Stanford CodeX universal backup + Vals.ai prep).
1 NEW risk R-NEW-ACADEMIC-OUTREACH-STALL: 6-month outreach lead-time may stall on slow institutional response cycles; mitigation = 3-track parallel (one confirmed lock = trigger satisfied for that phase); escalation = Stanford CodeX universal academic anchor backup if all three faith-tradition outreaches stall by 2026-12-01; further escalation = Vals.ai-only narrative for Phase A and accept narrower benchmark scope if Stanford CodeX also stalls by 2027-03-01. Risk count 44→45.
NO REVERSAL of: A-145 Λ.η ★ (Premium tier feature integration preserved); A-144 Σ.μ.ε.5 ★ (Premium “InheritKitBench-validated” preserved); A-143 Σ.κ ★; A-142 Π.θ ★ (Skills marketplace quarterly score-card preserved); A-141 verificationTier; A-113 sovereignty; A-114 Cedar+Catala substrate; T47 §findings 6-9 sizing; brand-architecture v1.16; memory project_inheritwills_mfi_subscription.
NO new SKOS scheme. NO new A-21 CI gate (G-INHERITKITBENCH-PHASE-A-PASS + G-VALS-AI-LISTING deferred to Phase-1 closure cascade). Cross-module primitive count UNCHANGED at 27. Customer-facing terminology lock unchanged.
Recurring revenue projection unchanged: £34M total ARR Q4 2029 (Phase A InheritKitBench v0.1 unblocks £2M+ p.a. AI-vendor procurement path; £750K p.a. IAS Premium tier-promotion gate; £25-50K p.a. partner-firm migration QA per A-141).
Methodological observations:
- THIRD consecutive ζ.2 above-threshold decision; ζ.2 mid-flow consistency strengthening (post-MQ-001 R2 reversal pattern stable)
- Phase A scope-narrowing avoids T47 §finding 36 self-defeating sole-author faith-tradition trap
- Parallel academic outreach 6-month lead-time honoured pre-emptively (NOT reactively)
- Susskind 2008+2023 dual-grounding without library re-purchase (already-indexed since 2026-04-21)
- Reversibility preserved (if window closes faster, compress B-F via Ω.ε hiring; if window stays open, organic academic locks accumulate)
- 20 of 21 ζ.2-related decisions remain convergent
Spike file: ~/off-github/library/projects/inherit/T-inheritkitbench-v0-1-commitment-spike-2026-05-01.md v1.1 (LOCKED outcome appended).
Cascade artefacts: arch-state v3.32 A-146; IK arch-state v1.6; IAS arch-state v1.7 (Premium-tier feature gate); INDEX v2.22; new ζ-Q18 cascade file Q-018-zeta-inheritkitbench-v0-1-commitment.md v1.0; risk register v1.3 (R-NEW-ACADEMIC-OUTREACH-STALL); standards-engagement-state v1.1 (ICAIL 2027 + Vals.ai + faith-tradition academic outreach pipeline); this memory file; active-work-log; richard-tasks-tt v1.19→v1.20 with 220-224 added.
Phase 3 status: 10 Q answered + cascaded (ζ-Q9..ζ-Q18); 1 deferred (ζ-Q15-original); ζ-Q19 candidate = Phase-1 deliverable matrix reconciliation under stacked ζ-Q14..ζ-Q18 amendments (master plan + per-repo BUILD-PLAN reconciliation overdue).