MEMORY-archive.md
Older topic-file pointers archived from MEMORY.md. This file is NOT auto-loaded into sessions — search here when the active MEMORY.md index misses.
Archived 2026-05-24 during global CLAUDE.md SOTA refactor. Active MEMORY.md kept top ~100 entries + cross-repo sections; explicitly SUPERSEDED/DEPRECATED/HISTORICAL/DISCARDED/DEFERRED-to-phase entries moved here regardless of position.
Promote an entry back to active MEMORY.md if a session needs it routinely.
- Adopt GitHub Spec Kit (SUPERSEDED v1 lock) — SUPERSEDED 2026-05-23T~19:30 by BMAD+OpenSpec hybrid lock (above)
- ζ-Q13 PENDING LOCK handoff — superseded by lock 2026-05-04T22:50
- ν.β D4 CLOSED 2026-05-04 — outcome-VALIDATED-WITH-METHODOLOGICAL-SUBSTITUTION
- ν.β D3 CLOSED 2026-05-04 — outcome-VALIDATED-WITH-NOTE
- ν.β D2 CLOSED 2026-05-04 — outcome-VALIDATED — 6-agent Phase-1 cost £27-42K; kill NOT-MET; Q-NU-011→Q-015-ai-agent frontier-derisked
- ν.β D1 CLOSED 2026-05-04 — outcome-VALIDATED
- ν.β C1 CLOSED 2026-05-04 — outcome-VALIDATED
- ν.β C2 CLOSED 2026-05-04 — outcome-VALIDATED
- ν.β C3 CLOSED 2026-05-04 — outcome-VALIDATED
- ν.β C4 CLOSED 2026-05-04 — outcome-VALIDATED
- TT spike-suite skills V1.2 deployed 2026-05-04 — wave-handoff-prompt mode
- Frontmatter is the load-bearing discipline (locked 2026-05-03)
- ζ.3 SOTA-derisking suite Wave 1+2 CLOSED
- ζ.3 Spike Z1.5 incremental-reasoning-posture CLEAN RE-RUN CLOSED 2026-05-03 —
- ζ.3 Spike Z1 v1.0 incremental-reasoning-posture CLOSED 2026-05-03 —
- ζ.3 Spike Z2 vocabulary-management CLOSED 2026-05-03 — outcome-VALIDATED; flat
- ζ.3 Spike A2 SHACL invariants CLOSED 2026-05-03 — outcome-VALIDATED
- ζ.3 Spike A1 LinkML schema-evolution CLOSED 2026-05-03 — outcome-VALIDATED
- ν.α Spike Z3 ODP role-modelling CLOSED 2026-05-03 — outcome-VALIDATED
- Dry-run flag mandatory for side-effecting launcher scripts
- ζ-Q5 κ.δ ★ alignment-axiom strength 3 May — A-149
- 22-spike Spike 7 Risk register update v1.2 → v1.3 CLOSED 2026-05-03 —
- 22-spike Spike 6 T-NEW tensions sweep CLOSED 2026-05-03 — outcome-VALIDATED — A-143
- 22-spike Spike 5 ζ-Q15 Wills CLOSED 2026-05-03 — λ.ζ ★
- 22-spike Spike 4 ζ-Q5 alignment-axiom strength CLOSED 2026-05-03 — κ.δ ★
- 22-spike Spike 10 G-IRI-VERIFY CI gate CLOSED 2026-05-03 — outcome-MITIGATED
- Research-artefact forward-traceability (locked 2026-05-03)
- 22-spike Spike 9 Prudhomme PROV-O→BFO CLOSED 2026-05-03 — outcome-VALIDATED
- 22-spike Spike 3 Phase 2 angles v-viii CLOSED 2026-05-03 — outcome-VALIDATED
- 22-spike Spike 21 Q-016 Will Update Architecture CLOSED 2026-05-03 —
- 22-spike Spike 8 9 i-ζ class authoring CLOSED 2026-05-03 — outcome-VALIDATED
- 22-spike Spike 2 MQ-* meta-Q status review CLOSED 2026-05-03 —
- 22-spike Spike 22 Q-017 trigger-event detection CLOSED 2026-05-03 —
- TT-brand work (locked 2026-05-03)
- Anthropic Managed Agents CLOSED 2026-05-03 — — A-142
- 22-spike Spike 12 documentation-cluster CLOSED 2026-05-03 — outcome-VALIDATED
- 22-spike Spike 1 refined-prompt v3.7 authoring CLOSED 2026-05-03 —
- Tier 2 pgvector library index VALIDATED 2026-05-03
- Spike-suite ε.ι outcome index — Phase E Task 13 lock-decision
- Kill-condition + maturity-vocabulary v3.7-candidates consolidated
- Curl-through-symlink corruption sub-rule (2026-05-03)
- Paste-safety for terminal handoffs (locked 2026-05-02)
- UK local time is the default for all timestamps — Anchor-or-ask, never guess; BST during DST window 2026-03-29→2026-10-25
- Universal-production-pipeline sequence — SEED → LLM-REFINE → partner-REVIEW
- Surface alternatives before collapsing synthesis to baseline
- Logging-contract closure within same session as T-file authoring
- Confront richard-tasks at creation time (locked 2026-05-02)
- Test theories immediately when tabled (locked 2026-05-02)
- LinkML 1.10 annotation surface separation (locked 2026-05-02)
- YAML-as-canonical metadata carrier — INHERIT v2 architectural pattern
- Kill-condition strict-vs-spirit reading via outcome-MITIGATED
- S8 Catala HMRC golden vectors VALIDATED 2026-05-02
- S6+S7 precedent-inspection VALIDATED-WITH-NOTE 2026-05-02 — AM-CDM 6-module +
- S5 FIBO SSSOM alignment VALIDATED 2026-05-02
- S2.9 graph-RAG pilot VALIDATED 2026-05-02 — richard-task #209 confronted
- S4 UK&W NRB pipeline VALIDATED-WITH-METHODOLOGICAL-SUBSTITUTION 2026-05-02
- S2.6 owlready2 production-scale MITIGATED 2026-05-02
- S3 OntoUML-in-LinkML phase encoding MITIGATED 2026-05-02
- S2.10 Cedar Analysis confronted VALIDATED 2026-05-02
- ζ-Q3 ε.ι S2.5 owlready2 rescue VALIDATED 2026-05-02 — ε.ι Layer 1
- S2 LinkML imports MITIGATED 2026-05-02 (2nd ε.ι)
- ζ-Q3 ε.ι S1 v6.6 SEED quality audit VALIDATED 2026-05-02 (1st of 10)
- ζ-Q2 ASPIRATIONAL UPLIFT [ξ.η → ξ.+] 2026-05-02 — first uplift under v3.1 + — A-130
- Bold-front-loaded synthesis preference 2026-05-02
- ζ-Q4 LOCKED ω.η BFO-only with PROV-O + IAO SSSOM-mediated alignment 2026-05-02 — A-129
- BFO-rooted hybrid 9 i-ζ classes 2026-05-02 — A-128
- ζ-Q2 LOCKED ξ.η G+ + Phase-1-graphRAG synthesis 2026-05-02 — A-127
- ζ-Q1 LOCKED κ.θ Hybrid utility-tree META-frame 2026-05-01 — A-126
- FULL PHASE-2-3 REVERSION 2026-05-01T20:45 — 17 Qs reverted — A-126
- ζ-Q15 [PRE-REFINED-PROMPT-archived] DISCARDED PER FULL REVERSION 2026-05-01 — historical only
- PHASE-FIT REVERSAL 2026-05-01T18:30 — ζ-Q15..Q18 DISCARDED — premature by 2-3 phases; A-142..A-146 DISCARDED; risk count restored 45→34
- ζ-Q18 [DEFERRED-to-phase-5] — historical only; lock discarded; spike preserved
- ζ-Q17 [DEFERRED-to-phase-6] — historical only; lock discarded; spike preserved
- ζ-Q16-REVISED-AGAIN [DEFERRED-to-phase-6] — historical only; lock discarded
- ζ-Q16-REVISED [DEFERRED-to-phase-6] — historical only; lock discarded
- Susskind End of Lawyers indexed 2026-05-01 — Rich-purchased Kindle; library research preserved
- ζ-Q15-REVISED [DEFERRED-to-phase-5] — historical only; lock discarded
- ζ-Q14 LOCKED Ψ.γ CORE alignment-axiom A-55 verify-tier 2026-05-01 — A-141
- ζ-Q13 LOCKED Δ.β CORE schema + 3-tier verification 2026-05-01 — A-140
- MQ-002 Σ.α Phase 2 formally closed + dry-run + Phase-3 sequence 2026-05-01
- ζ-Q8 LOCKED Λ.β multi-anchor SSSOM + PHASE 2 CLOSED 2026-05-01
- ζ-Q7 LOCKED χ.γ vi-6+ Hybrid 2026-05-01
- ζ-Q6 R2-reframed LOCKED φ.α’ α 3-layer status quo 2026-05-01
- MQ-001 R2 LOCKED ω.γ → ω.α reversal 2026-04-30 evening — A-131
- ζ-Q5 LOCKED ψ.γ SSSOM-canonical + Prudhomme-vendored 2026-04-30 evening — A-130
- ζ-Q4 LOCKED ω.γ Hybrid multi-parent BFO+PROV-O 2026-04-30 evening — A-129
- ζ-Q3 LOCKED ε with 9 i-ζ classes (sub-A) 2026-04-30 evening — A-128
- ζ-Q2 LOCKED G+ permanent (sub-A) 2026-04-30 evening — A-127
- ζ-Q1 LOCKED Option C 2026-04-30 evening — A-126 first authoring. 14-criterion frame. Bass 4e + T82 synthesis. Utility tree v1.0
- MQ-000 ζ.2 chosen 2026-04-30 evening — Rich override of Claude β rec (1.05 margin); ζ.2 = re-derive options + Shape X
- Architecture re-derivation design 2026-04-30 — Methodology pivot; supersedes 2026-04-29-multi-phase-audit-design v1.0; 4-layer info flow
- Snapshot-pin vs floor-pin discipline (locked 2026-04-29) — Two pin kinds: snapshot=provenance never bumps; floor=live ref bumps each /review-plan
- Live state files first framing (locked 2026-04-29)
- Partner-pitch doc pattern (locked 2026-04-28) — 13pp/9 sections + mockup appendix recipe for law-firm pitch leave-behind PDFs
- Jane Hunter+Stephen Attree pitch 2026-04-28 — First pitch positive; combined+visual proposal locked; Mon 11 May joint call; Tue 19 May MLP board
- 28 — Multi-day: Q14-Q28 brainstorm + 5-repo BUILD-PLAN migration + master plan v1.5; arch-state v2.78→v2.93+
- Two-session hardening pattern (locked 2026-04-28) — Empty-repo Session 1 = legal+tooling; Session 2 = docs+scaffolding+BUILD-PLAN
- Phase-2 Q14-Q23 brainstorm complete (2026-04-27) — 10 substrate+application locks; arch-state v2.78→v2.79; 2 new Apache primitives
- Phase 1 Option G+ audit complete (2026-04-27) — 9 D-decisions, 7 amendments A-118..A-124, arch-state v2.71→v2.78
- Display numbers in scorecards (locked 2026-04-27) — Every scorecard MUST include numerical scoring matrix + sensitivity perturbations
- Decision-analysis one-at-a-time (locked 2026-04-26) — Never batch decision-analyses; recalculate next based on Rich’s actual answer
- Post-spike-completion session (2026-04-26) — 3 lessons; arch-state v2.65→v2.69; richard-tasks 165-168
- Autopilot prompt techniques (locked 2026-04-26) — 5-technique recipe for zero-question autopilot prompts; <4/5 = not ready
- Run unattended once scope is locked (2026-04-26) — STOP asking once option+path+criteria+sequencing chosen; ntfy heartbeats
- Skill discipline (2026-04-26) — Boundary-test tables in CLAUDE.md (project+global); slow progress = under-invoked skills
- LOC NOT a significant metric (locked 2026-04-26) — Drop LOC framings; use Gate completeness/AC-1/conformance tier instead
- READ T-files+spike data before any decision-bearing answer (elevated 2026-04-27) — 2-4 verbatim quotes with location refs required
- Stop UK-drift — strict per-jurisdiction-crate framing (locked 2026-04-26) — Never let one jurisdiction shrink module scope
- Round 7 spike-avoidance T50-T70 (2026-04-26) — 21 T-files closing 27 clusters; 4 candidate primitives; ~17-22 wks saved at £525
- IAS build plan sovereign (2026-04-26) — Do NOT rush or reorder Phase-1/Phase-2/IAS for waiting consumers
- LL Stage-28 freeze + Path A born-on-v2 (DEPRECATED 2026-04-30)
- Round 6 Harvard-mode 2026-04-25 — 8 deep-dives T42-T49; Cedar/Catala incompatible (Spike 5.5 dropped); IAS leapfrog 9-12mo window
- Round 5 gold-hunting 2026-04-25 — 7 subagents T35-T41; 48 spikes + 9 strategic findings; IAS £25-40m UK Y3 ARR
- Spike replacement Round 4 complete — 6 subagents collapsed 22 spikes 17-19wks→3wks; Cedar wins 19.5pp; Catala→WASM Path A
- review-plan 14 improvements — Improvements across Topics 2-9, Phase-C Steps 2-4, Phase-B retroactive, A-21 CI, SDK
- Technology research programme T1-T12 — 12 technologies; 8 parallel subagents complete; T1-T3 dispatched per follow-on
- Technology research synthesis — 10 discoveries + 16 tensions + 5 cross-tech threads; ISO-track positioning stronger than baseline
- Technology research discoveries catalogue — Comprehensive audit trail T1-T13; ~50 top-tier discoveries; survives compaction
- Catala ecosystem maturity — Catala v1.0/v1.1 + catala-explain + catleg production stable; reframes to ecosystem-integration
- catala-explain .docx as commercial artifact — Reframes AI-vendor pitch to per-query lawyer-readable explanations; 4-tier model
- Library+web research for narrow margins — Margin <2pp + arch-critical + ecosystem-mature → library+web research; Catala γ→δ flip
- Architecture-state file discipline — READ inherit-v2-architecture-state.md before any v2 strategic output; UPDATE after every tension-lock
- v1 deep-dive completed 2026-04-24 — 5-agent parallel deep-dive; 180 findings; 13 T-NEW tensions; C-21 resolved via A-40
- v1 valuable starting point — inclusive not dismissive — 2026-04-24: v1 content huge starting-point for jurisdictional acceleration
- Auto deep-dive when top options imperfect — Trigger on <5pp margin + structural limits + recent frameworks unapplied
- Auto-deep-dive via framework-rediscovery — Two modes: narrow-margin-triggered + elective framework-application; A-55 most reused
- Weak criteria to avoid in scorecards — 5-test checklist: scarcity/clean-break/double-counting/temporal-mismatch/effect-size
- Session close 2026-04-24 morning — 8 tensions closed + 1 partial; A-41..A-68; state v1.0→v2.7; A-21 CI 7→16 gates
- PHASE-A complete 2026-04-24 evening — 3 items closed; A-69..A-82; state v2.7→v2.10; A-21 CI 16→18; DistributionContext primitive #12
- First vertical slice symmetric 2026-04-24 — Option G: Phase-1 treats 5+ jurisdictions equally; no early bet; anti-E&W-creep explicit
- STRICT single-question rule — highest-knock-on first — Triple-elevated 2026-04-26: >1 Q-marks at message-end = STOP, build dependency map
- Always display full scorecard inline — 2026-04-23: every scorecard inline in full (criteria + matrix + sensitivity + critique)
- Apache-commitment VERY LOW PRIORITY — 2026-04-23: cap Apache-timing tensions at I; defer until Rich opens topic
- Audit brainstorm IN PROGRESS — paused awaiting Q1 — INHERIT v2 + IK technical audit through 2029-watcher eyes; Q1 = scope of revisability
- indexed first — 2026-04-22: ls ~/off-github/library/indexed/ + grep before scoring research confidence
- IRI-verification-before-lock — Before any external-ontology IRI axiom, run 30-min existence+semantics+version spike
- IRI terminology + Unicode discipline — From 2026-04-23: all new v2+IK writing uses “IRI” not “URI”; ASCII-only Phase-1
- Introduction-module rule — Concept lives in module where FIRST semantically introduced + Core fallback; won 100% 3×
- Reframe-beats-re-weight — Stalled 85-92% scorecards: question framing before re-weighting; observed 3× in session
- Always save scorecards as auditable artefacts — 2026-04-24: persist every scorecard to docs/superpowers/scoping/YYYY-MM-DD-scorecards/
- INHERIT v2 Shape X full lock 2026-04-23 — All 8 angles locked; 4-layer standards stack; A-1..A-25 amendments; Phase-1 unblocked
- v2 + IK international from day one — 2026-04-24: architecture is international; sequence may be E&W-initial
- Session 2026-04-22 PM library deep grounding — 24 items indexed; weighted-avg confidence ~86%; 6 Z-items ≥90%
- INHERIT v2 architecture LOCKED 2026-04-22 — Ⅲ Full-C+ + v6.6-inheritance + Option-D (E&W+Scotland+CH) + 7-module + 30-48mo Phase-1
- Partner model — three shapes — 2026-04-22: A succession+tax+faith / B catalogue marketplace / C platform-integration
- Session 2026-04-22 architecture lock — Wed 22 April: architecture-lock + preparatory commits for 24-Z research dispatch
- Z research programme 24 files — 6-wave dispatch; ~26 agent-runs; 6-7hr wall-clock; Z24-MODULARITY added
- v2 + IK clean-break from v6.6 — 2026-04-22: each in own self-contained repos; v6.6 reference-only; no runtime dependency
- v2 Asset-catalogue scope — 14 Apache primitives — Evolved from 11 (+Liability A-4 +Organisation A-11 +RoleInstance A-20); ~32K LOC canonical
- Phase-1 LLM Scenario 2 lock — Cloudflare AI Gateway + Haiku 4.5 + £1,500/mo ceiling; no Opus
- PDF ingestion text-extraction first — Agents read
<slug>-text.txt(not raw PDFs) via ~/off-github/library/extract-pdf-text.sh - Per-brand local design-system CSS — 2026-04-20: each brand has src/styles/{brand}-design-system.css; mhq2 shared system superseded
- Law vs practice split — 2026-04-20: INHERIT (Apache) carries LAW; InheritKit (AGPL+commercial) carries PRACTICE
- Viotti April update — Blaze fix likely + MCP plan via Sourcemeta One + doc generator + launch help offered
- InheritKit concept — Tier 2 AGPL+dual-commercial embedded SDK as INHERIT revenue layer
- Strategy conversations — Prefer tiered framings + honest self-correction over single-answer recommendations
- JSON Schema + JSON-LD combined — 2026-04-20: JSON Schema primary structural + JSON-LD additive semantic; both Apache 2.0
- v2 international + ISO-track intent — 2026-04-20: NOT UK-centric; ISO via OASIS→JTC 1 PAS fast-track 2028-2030
- Use “Ecosystem” not “DADE” in specs — 2026-04-20: confidentiality + strategic flexibility; DADE-dedicated docs are exception
- International English (US-style) for new v2+IK writing — 2026-04-20: forward-looking only; UK English stays default elsewhere
- Cross-reference discipline with versioned pointers — 2026-04-20: versioned bidirectional cross-refs + frontmatter
related: - Stealth discipline — partner-outreach friction cost
- Target lawyer-embracers not rejecters — 2026-04-20: target Innovation/Tech roles + own legal-tech builds; messaging: augmentation not replacement
- v2 Apache commitment DEFERRED until public release — 2026-04-22: code-inherit-v2 stays PRIVATE; Apache irrevocable only at public release
- INHERIT vs InheritKit bias — RE-REVERSED to lean-to-IK — SECOND reversal 2026-04-22: lean-to-InheritKit in close calls
- Three-track partner badging — INHERIT Contributor / InheritKit Certified / TT Partner — multiple per partner per context
- 17 — Long strategic: 8+ specs, InheritKit-first, three-track, Henry+Command Centre
- Session 2026-04-17 pilot — E&W×succession scoping v1.1; Framework v1.2 (FW1-FW5 absorbed); 8 SCRs
- Session 2026-04-17 tax — E&W×tax v1.0; 29 TAX-SCRs; 7 test vectors; FW-TAX-01-04 surfaced for Framework v1.4
- SCR numbering convention — Dimension-prefixed naming (SUCC-/TAX-/TRUST-/POA-/FAITH-); grouped by dimension
- Policy-as-code strategic context — Potential bridge layer between Apache INHERIT + AGPL InheritKit
- API Days Singapore contacts — Donnie Prakoso (AWS) + Erik Wilde (OAI Ambassador, Jentic); agent-readiness gap surfaced
- Lau Pa Sat Report primacy — Primary evidence for INHERIT direction; endorses JSON Schema; 10 enhancements not paradigm shift
- DADE Unfinished Digital Estate report — OIDF 2026-03-03 (Flanagan/Kiser/Saxe); OAuth Token Exchange + UMA + Sovrin protocol anchors
- Licence strategy open questions — v3.4 reopens licence-stack as research question; iText subagent + TT internal-licence architecture
- AI-vendor commercial model — Friday 2026-04-17: high-conviction huge-revenue opportunity; 3 routes; Phase 1 deliverable
- Sourcemeta revenue-share for MCP — Two-tier MCP: A Sourcemeta no-rev / B TT-built 100% TT
- Session 2026-04-17 PaC — PaC architecture spec v1.0 + spike brief; option B recommended; reframed mid-session
- INHERIT architecture mandate — Expanded 2026-04-17: willing to throw away JSON Schema; 8 candidate paradigms evaluated
- Options D and E synthesis — After Option-C books read: D (LPG+RDF+TT-LLM+MCP) and E (AI-native); LOC + acquirer-LLM constraint
- Session 2026-04-17 v4 architecture — v4.0.0 spec + positioning + licence + AI-vendor + DADE: 153pp; JSON-LD 12b; Apache retained
- Session 2026-04-18 v2 scorecard — 4-option scorecard B/C/D/E; D 66/80; D→E phased recommendation as pre-F baseline
- Sequential vs parallel revisions — Parallel for independent research; sequential for refinement revisions with dependencies
- Lau Pa Sat is self-marking — Rich’s scenarios are stress-tests not independent evidence; don’t narrow paradigm evaluation
- Sourcemeta loyalty bias — Treat as current-decade tooling partner not architectural axis
- Premature option-siloing — When Rich introduces a topic, don’t silo as preferred; keep option tree wide
- Recency vs maturity (book recs) — In stock + enough reviews; sweet spot 2024-early 2025; never forthcoming/MEAP
- Wait for grounding material — Don’t produce methodology from training-data priors when authoritative books arriving
- Scope vs commercial reality — Bootstrap capacity + £5m+ TT-sale = architectural constraints
- No current-stack bias for INHERIT v2 — Vercel/Supabase NOT a constraint; evaluate tech on merit
- adoption criteria in v2 scorecards — 2026-04-20: no “cost to migrate” or “developer familiarity”; prefer fundamentals
- Logic trumps hallucinations — LLMs as narrow adjunct; canonical layers Catala+Alloy+TLA+Rego
- No TT training data for INHERIT — No TT corpus, no fine-tuned model; off-the-shelf vendors acceptable for narrow adjunct
- Scotland as second jurisdiction — phase-timing SUPERSEDED
- InheritWills subscription-native + MFI Living Monitoring tier — 2026-04-19: IW ~£10-25/mo; MFI L0 ~£5-10/mo; Richard-task #115
- AI raises the delegation bar — AI subsumes easy-to-delegate; remaining work needs more sophisticated people
- Nova partnership deprioritised — 2026-04-19: do NOT propose Nova as build partner; Rich+Claude is primary; Josh technical lead
- OpenInherit public-release discipline — 2026-04-19: gradual module-by-module + launch-coupled; v6.6 stays as-is; don’t improve preemptively
- Session 2026-04-19 evening — 7hr filing+brand-architecture consolidation; brand-architecture.md v1.5 + frontmatter-conventions.md v1.0