22-spike Q&A-formulation suite — Spike 3 A2 Phase 2 angles v-viii framing CLOSED outcome-VALIDATED 2026-05-03

Summary

7th 22-spike to close (after Spikes 1 + 12 + 18 + 22 + 2 + 21). Group 1 strict-order WRAP-UP COMPLETE — Spikes 1 + 2 + 3 all closed in same Batch A SPIKE-RUNNER session 5 (Spike 1 closed 09:50 by other session; Spike 2 closed 11:38; Spike 3 closes 11:50). FIRST non-meta-recursive spike to apply v3.7 disciplines prospectively (Spikes 1 + 2 were meta-recursive on plan-file-pre-resolution).

Outcome: outcome-VALIDATED on strict reading per refined-prompt v3.7 (kill-condition ≥1 of 4 angles structurally unable to be evaluated under AHP NOT-MET — all 4 angles produce ≥2-point spread across 7 architecture options).

4 NEW κ.θ angles framed against T82’s 4 SOTA pillars

Plan Task 3 expected the 4 NEW angles enumerated in 2026-04-30-architecture-rederivation-design.md — the design-spec doesn’t enumerate them by letter (refers only to “8 Shape X angles in sequence”). Spike 3 framing decision: derive 4 NEW angles from T82’s 4 SOTA pillars beyond Bass + ISO 25010:

AnglePillarT82 §
vBEA 2e Architectural Fitness Functions (Phase-4 readiness)§3 — continuous evaluation vs Bass point-in-time
viContinuous Architecture Practice (CAP) Deferral Discipline§4 — principle-3 “delay design decisions until necessary” + principle-4 “architect for change” per Erder+Pureur+Woods CAP 2021
viiMCDA SOTA 2024-2026 Sensitivity Discipline§5 — PySensMCDA 8-mode + COMSAM multi-criterion + weight stability intervals + robustness coefficients
viiiAI Architecture Evaluation 2025-2026§6 — AgentArcEval + TRiSM + CSS + TUE + KDD 2025 GenAI Eval

Framing decision rationale (T-file §1.1): T82’s 4 SOTA pillars chosen over Bass PALM 11 business-goal categories + κ.θ Hybrid utility-tree META-frame’s 6 layers extended (META-frame REVERTED in Full Phase-2-3 Reversion 2026-05-01T20:45 — relying on it would inherit the very methodology the reversion repudiated).

4×7 score matrix (PROVISIONAL pending ζ-Q2 re-ask)

AngleB JSON SchemaC SemWebD KG-canonicalF MaximalG+ Commercial-scopedGraph-RAG-NativeG+ + Phase-1-graphRAG
v BEA 2e fitness functions4334525
vi CAP deferral discipline4325434
vii MCDA SOTA sensitivity3345445
viii AI architecture eval2344555
Sum (unweighted)13121318181419

Directional preference: G+ + Phase-1-graphRAG (19) > G+ Commercial-scoped (18) = F Maximal (18) > Graph-RAG-Native (14) > B JSON Schema (13) = D KG-canonical (13) > C SemWeb (12) — matches system-reminder ξ.η ★ 4.55 + ξ.ε 4.60 ζ-Q2 framing.

AHP-readiness check (kill-condition resolution)

AngleScore rangeSpreadUnique valuesAHP-ready?
v2-534YES — strong differentiation
vi2-534YES — strong differentiation
vii3-523YES (BORDERLINE) — meets ≥2-point min
viii2-534YES — strong differentiation

Borderline angle vii recommendation: ahpy weight-stability-interval handling per T82 §5.3 at richard-task #205 AHP session — Option A (use weight stability intervals to define wider acceptable weight range; ahpy supports per-criterion weight ranges) preferred per T-file §3.2.

v3.7 disciplines applied (Spike 3 is first non-meta-recursive)

  • Step 6e MATURITY VOCABULARY (Candidate 1) — outcome-VALIDATED on plain reading
  • Step 6e ARCHITECTURAL-LAYER-OF-FAILURE TAGGING (Candidate 2) — applied (clause_layer: measurement for qualitative scoring)
  • Step 6e STRICT-VS-SPIRIT READING PRE-FLIGHT (Candidate 3) — APPLIED to resolve borderline-angle-vii outcome to VALIDATED
  • Step 6e SAMPLE-SET + STRUCTURAL-CEILING (Candidate 4) — APPLIED retrospectively
  • Step 6e PLAN-FILE URL/IRI/FILENAME PRE-RESOLUTION (Candidate 6) — VALIDATED EMPIRICALLY (SECOND defect-class detection in Batch A; first was Spike 2)
  • Step 11 LOGGING-CONTRACT CLOSURE (Candidate 8) — APPLIED (this T-file authoring + cascade WITHIN SAME session as Spike 2 closure)
  • Step 13 SUB-MODE LIFECYCLE MANAGEMENT (Candidate 5) — NOT APPLICABLE (no TRANSIENT sub-mode)
  • Step 14 PARALLEL-SESSION SAFETY (Candidate 7) — APPLIED (Read-immediate-Edit pattern on plan + arch-state given concurrent Batch K + Batch D sessions)

Plan defects surfaced (SECOND v3.7 Candidate 6 detection)

Plan §2 Task 3 has TWO file-path/wording defects:

  1. Files block references 2026-04-29-multi-phase-audit/2026-04-30-architecture-rederivation-design.md — actual path is parent docs/superpowers/specs/2026-04-30-architecture-rederivation-design.md
  2. Step 1 expected design-spec to enumerate angles v-viii by letter — design-spec doesn’t (refers only to “8 Shape X angles in sequence” generally)

Plan v1.17 → v1.18 patch candidate flagged for REFERENCE session.

Empirical pattern: 2 plan defects in 2 spikes (Spike 2 + Spike 3) confirms v3.7 Candidate 6 captures a real defect-class.

Cascade artefacts

  • T-file: ~/off-github/library/projects/inherit/T-spike-phase-2-angles-v-viii-framing-2026-05-03.md v1.0 (283 lines / 31.8 KB; 8 body sections + frontmatter + CHANGELOG; 9 honesty caveats); off-github commit pending (Batch A closure batch 1)
  • arch-state: v3.35 → v3.36 with NEW §13 row Q-and-A-S3 + Changelog row; docs-strategy commit (Batch A batch 2; pending)
  • phase-2-architecture-options-rescore.md: v2.0 → v2.1 NEW section ”## κ.θ Angles v-viii (added 2026-05-03 per Spike 3)” appended; same docs-strategy commit
  • Plan: v1.17 → v1.18; §0 Spike 3 checkbox ticked; lastmod 12:40 → 11:50; CHANGELOG row added; same docs-strategy commit
  • Memory file: this file
  • MEMORY.md: index entry to be added (above most-recent entry)
  • Active-work-log: CLOSED entry to be added
  • NO docs-personal commit (no richard-task disposition — richard-task #205 unchanged)

Companion pin floor-bumps

  • arch_state: v3.35 → v3.36
  • refined_prompt: UNCHANGED at v3.7 floor
  • predecessor_plan: UNCHANGED at v1.9 snapshot
  • design_spec: UNCHANGED at v1.0 floor
  • outcome_index: UNCHANGED at v1.0 floor

Methodological observations

  1. 7th 22-spike to close, after Spikes 1 + 12 + 18 + 22 + 2 + 21.
  2. Group 1 strict-order WRAP-UP COMPLETE — Spikes 1+2+3 all closed; Phase 2 has methodological foundation in place.
  3. FIRST non-meta-recursive spike to apply v3.7 disciplines prospectively — Spikes 1 + 2 were meta-recursive on plan-file-pre-resolution; Spike 3 detects v3.7 Candidate 6 defect on Spike 3’s OWN plan task wording. Same pattern but applied externally.
  4. T-file pre-flight T82 read-first was load-bearing — saved ~50% effort vs ~25% estimated in plan §0 row 63 (T82 §3-§6 directly enumerate the 4 NEW angle candidates).
  5. NO architectural change beyond §13 row + Changelog row + 1 phase-2-rescore section append; NO master plan / per-repo BUILD-PLAN edit; NO risk register change; NO new SKOS scheme; NO new A-21 CI gate (still 22 per A-130); cross-module primitive count UNCHANGED at 27; module count UNCHANGED at 9.
  6. NO richard-task disposition (richard-task #205 ~90-min AHP pairwise unchanged + remains operational; Spike 3 surfaces angle definitions + rubrics + 4×7 raw scores; AHP weight allocation is downstream work).

Reconsideration trigger preserved

If Rich subsequently disagrees with:

  • 4-angle framing derived from T82’s 4 SOTA pillars (would prefer Bass PALM / κ.θ META-frame extension / different decomposition) → re-frame Spike 3 outcome at next session; plan v1.17 → v1.18 + new κ.θ structure
  • 4×7 score matrix specifics (would prefer different scores under fresh option-set evidence) → ζ-Q2 re-ask under refined-prompt v3.0.1+ will produce locked option-set + scores
  • Angle vii’s borderline-differentiation status (would prefer drop angle vii or expand its rubric) → trivial rubric update + re-score; Spike 3.5 mini-spike candidate

→ Spike 3 T-file is the starting reference for any of these reconsiderations.