TT Claude Memory

project_zeta_q1_kappa_theta_utility_tree_locked_2026_05_01

Properties1
descriptionFirst Q re-asked under refined-prompt v3.0.1 post-Full-Reversion 2026-05-01T20:45. A-126 (re-issued) registered in arch-state v3.11. κ.θ = Bass-canonical utility-tree + per-criterion 2-3 ASR scenarios + AHP pairwise-comparison-derived weights + ISO 25010:2023 9-char sub-mapping + per-criterion BEA 2e fitness functions + LLM-evaluable ASR per arXiv 2506.00150v1. 10 META-criteria; weighted sum 4.47 vs κ.γ 4.20 (margin 0.27 sensitivity-driven; 5/8 above 0.20 threshold; 3/8 borderline 0.17-0.19). κ.γ identified as strong fallback. 5 NEW richard-tasks #205-#208. NO new SKOS scheme; NO new CI gate; NO new risks; NO cross-module primitive count change. Anchoring-bias mitigation worked — PRE-REFINED-PROMPT Option C read for shape only; fresh re-derivation produced 10 META-criteria with #10 NEW. WR-criteria-frame-2026-05-01.md v1.0 persisted.

5 min read

Lock: κ.θ Hybrid utility-tree + LLM-evaluable ASR + AHP weights + ISO 25010:2023 sub-mapping + BEA 2e fitness functions

Locked at: 2026-05-01T22:00 BST per Rich confirmation

Margin: κ.θ 4.47 vs κ.γ 4.20 = 0.27 above 0.20 threshold but sensitivity-driven (5/8 perturbations clean; 3/8 borderline 0.17-0.19)

8 options scored: κ.α frozen 1.96 / κ.β reweight 2.76 / κ.γ re-derive flat 4.20 / κ.δ ISO wholesale 3.46 / κ.ε product-derived 2.96 / κ.ζ AHP-formal 3.45 / κ.η Rich-defined TBD / κ.θ ★ 4.47

10 META-criteria + weights (sum 1.00):

  • #1 QA coverage breadth — 0.13
  • #2 Bass 4e canonical alignment — 0.10
  • #3 ISO 25010:2023 alignment — 0.08
  • #4 MCDA-SOTA alignment — 0.08
  • #5 AI-architecture-evaluation 2025 alignment — 0.08
  • #6 Library-grounded confidence — 0.10
  • #7 TT product-need fit — 0.15
  • #8 Phase-2 closure timeline fit — 0.10
  • #9 Acquirer-narrative defensibility — 0.10
  • #10 Cross-Q comparability + sensitivity-perturbation viability — 0.08

κ.θ frontloads T82 §9 prescriptions: utility-tree (T82 §9.1) + per-criterion 2-3 ASR scenarios (T82 §9.2 / Bass 4e Ch 3.3 6-part template) + per-criterion BEA 2e fitness function (T82 §9.3) + MCDA SOTA sensitivity discipline downstream (T82 §9.4) + AHP pairwise-comparison-derived weights (Saaty 1-9; Springer AHP+ATAM+25010 confirmed canonical practice) + ISO 25010:2023 9-characteristic sub-mapping (5 direct + 5 X-ability per Bass §14.3) + LLM-evaluable ASR scenarios (arXiv 2506.00150v1 Jun 2025 SOTA).

Strong fallback identified: κ.γ (re-derive flat 10-criterion META-frame from primary sources; 4.20 weighted sum; ~10 days less upfront work than κ.θ; T82 §9 enhancements deferred post-lock; lower acquirer-narrative defensibility 4/5 vs κ.θ’s 5/5). If Phase-2 timeline binds harder than scorecard models OR utility-tree authoring exceeds 1-week time-box by 50%, fall back to κ.γ.

**5 NEW richard-tasks 205-208**:

  • #205 — AHP pairwise-comparison weighting session with Rich (90 min)
  • #206 — Utility-tree authoring (1 week time-box)
  • #207 — Per-criterion 2-3 ASR scenarios authoring (1 week per criterion; ~10 weeks parallelisable)
  • #208 — Per-criterion BEA 2e fitness function pre-author (Phase-4 prep)

Phase-2 timeline impact: ~3 weeks for #205 + #206 + #207 (parallelisable); ~£12-15K Paul-time-equivalent. Pays back across all 27+ remaining Phase-2-onwards Qs.

Anchoring-bias mitigation worked:

  • PRE-REFINED-PROMPT prior Q-001 (Option C 14-criterion 4.74) read for SHAPE only (option-set structure A-G; criterion-categories)
  • Did NOT lift prior 14 underlying architecture criteria
  • Fresh META-criteria re-derivation produced 10 META-criteria with #10 (cross-Q + sensitivity-perturbation viability) as a NEW META-criterion not in prior lock
  • Phase-2 timeline as first-class META-criterion at 0.10 (not present in prior weighting)
  • “Continuity-with-v1.9” fully absent per ζ.2 amendment

Convergent conceptual direction; more rigorous structure:

  • PRE-REFINED-PROMPT picked Option C (re-derived flat 14-criterion) at 4.74
  • Refined-prompt picks κ.θ (frontloaded utility-tree synthesis) at 4.47
  • Same conceptual direction (re-derive from primary sources, NOT freeze v1.9, NOT wholesale-ISO)
  • More rigorous structure (frontload utility-tree + ASR + AHP + ISO sub-mapping rather than deferring T82 §9 enhancements post-lock)
  • Refined-prompt strengthens rather than reverses prior intuition

WR file persisted: ~/off-github/library/projects/inherit/web-research/WR-criteria-frame-2026-05-01.md v1.0 — 5 web-research findings; Finding 1 (LLM-supported ATAM, arXiv 2506.00150 Jun 2025) is the small-to-moderate META-criteria framing influence; Finding 2 (AHP+ATAM+25010 canonical) confirms META-criterion #1; Findings 3-5 downstream (Cedar Analysis open-source / SPEF) or confirming.

8 sensitivity perturbations — all preserve κ.θ★:

  1. Phase-2 timeline ↑ 0.20: 4.22 vs 4.00 (margin 0.22; CLEAN)
  2. AI-arch-eval ↑ 0.15: 4.40 vs 4.16 (margin 0.24; CLEAN)
  3. TT product-need ↑ 0.25: 4.37 vs 4.20 (margin 0.17; BORDERLINE)
  4. Drop ISO entirely: 4.47 vs 4.28 (margin 0.19; BORDERLINE)
  5. κ.θ Bass score -1: 4.37 vs 4.20 (margin 0.17; BORDERLINE)
  6. κ.γ TT product -1: 4.47 vs 4.05 (margin 0.42; CLEAN)
  7. Add ζ.2-alignment criterion: 4.48 vs 4.22 (margin 0.26; CLEAN)
  8. Drop MCDA SOTA criterion: 4.47 vs 4.28 (margin 0.19; BORDERLINE)

Methodological observations:

  1. FIRST Q re-asked under refined-prompt v3.0.1 — 10-step protocol executed cleanly (verification + phase + Q-count + phase-fit hard guard + gap identification + research sweep + Q-framing + decision-quality-check + reasoning + reversal-readiness)
  2. Anchoring-bias mitigation worked — fresh re-derivation produced different META-criteria + weighting despite same overall direction
  3. κ.θ as genuinely-innovative SOTA synthesis — prior PRE-REFINED-PROMPT did not surface utility-tree + ASR + AHP + ISO-sub-mapping as a SINGLE option (treated as deferred enhancements); refined-prompt brainstorming widening produced κ.θ as first-class option per refined-prompt step 7 “6-8 GENUINELY innovative options” discipline
  4. 0.27 sensitivity-driven margin honestly flagged with κ.γ fallback identification per feedback_automatic_deep_dive_when_options_feel_imperfect
  5. WR-file persistence — first WR file informing a META-frame Q (vs prior WR-bequest-class for module-authoring Q)
  6. Convergent-with-prior-conceptual-direction validates refined-prompt strengthens rather than reverses

NO REVERSAL of: A-1..A-125 amendment registry; MQ-000 ζ.2 scope amendment; brand-architecture v1.16; master plan v1.10; per-repo BUILD-PLAN.md v1.4-1.6; 2026-04-30-architecture-rederivation-design.md v1.1; refined-prompt v3.0.1; T82 architecture-evaluation methodology synthesis.

NO new SKOS classifier scheme (META-frame doesn’t add domain primitives). NO new A-21 CI gate (no schema/code surface yet). NO cross-module primitive count change (still 27 = 25 §2.2 + 2 §2.3 pending). NO module structure change (still 9). NO new risks (κ.θ Phase-2-timeline risk pre-mortemed at 25% but mitigated via 1-week time-box on utility-tree authoring + κ.γ fallback path).

Phase 2 status post-lock: 1 Q answered/cascaded (ζ-Q1) / 18-30 budget / 3-6% complete; pre-build-start cumulative 1/100-150.

NEXT: ζ-Q2 = architecture-options selection (B/C/D/F/G+/[6th]) — first scored under κ.θ-LITE (the 10 META-criteria + reasoned-argument weights from this Q’s scorecard) as the Phase-2-timeline-realistic interim approximation; full κ.θ utility-tree + ASR scenarios + AHP weights author in parallel via richard-tasks 205-207. Refined ζ-Q2 sensitivity re-run under formal AHP weights only if material rank-changes emerge.

Graph View

Backlinks