Outcome: outcome-VALIDATED. Plan §2 Task 1 explicitly defines no kill-condition for methodological-authoring spikes; outcome-VALIDATED is the default.

Why this memory exists: Spike 1 is the FIRST of 22 spikes in the Q&A-formulation suite per 2026-05-03-22-spike-q-and-a-formulation-suite.md v1.6+. Per protocol A locked 2026-05-03T~10:00 BST: SPIKE-RUNNER session owns full closure cascade per spike (8 steps from plan §1.6); REFERENCE session runs the closure verifier + reports RED gap count. This memory captures the substantive findings + decisions so subsequent spikes (S2 onwards) and the suite-close audit-record (Task 24) have a single grep-able reference.


§1 The 8 v3.6→v3.7 candidates absorbed

Per feedback_kill_condition_maturity_vocabulary_v3_7_consolidated.md (consolidated 2026-05-02 BST):

#Candidatev3.6 location → v3.7 mutationSource spike
15-outcome × 2-TRANSIENT-sub-mode maturity vocabularyStep 6e SPIKES → NEW MATURITY VOCABULARY 5-outcome × 2-sub-mode table replacing 3-outcome listS4 (outcome-VALIDATED-WITH-METHODOLOGICAL-SUBSTITUTION) + S2.9→S2.9b (outcome-VALIDATED-WITH-PROVISIONING-NOTE)
2Architectural-layer-of-failure tagging on kill clausesStep 6e SPIKES → NEW clause_layer: field with substrate / tooling / provisioning / measurement tableS2.9 (provisioning-layer kill-clause distinct from substrate)
3Strict-vs-spirit reading as PRE-FLIGHT (not just post-hoc)Step 6e SPIKES → NEW clause_priority: + spirit_test: fields at spike-prompt-authoring timeS5 (first spike to apply pre-flight discipline)
4Sample-set + threshold-relative-to-structural-ceiling specStep 6e SPIKES → NEW sample_set: + structural_ceiling: fields on numeric-threshold kill clausesS2.6 (BFO 2020 Continuant+Occurrent split → 90%-IC kill clause structurally unreachable for full-CCO sampling)
5Same-session follow-on for retiring TRANSIENT sub-modesNEW Step 13 SUB-MODE LIFECYCLE MANAGEMENT (3-step: introduce TRANSIENT → revisit-within-same-session → drop-or-graduate)S2.9→S2.9b (~30 min retirement precedent: numpy-substitute → pgvector-original after provisioning unblocked)
6Plan-file URL/IRI/filename pre-resolution checklistStep 6e → NEW PLAN-FILE pre-resolution checklist before plan-write commit4 plan-defects in 24h on 2026-05-02 (S6 AM-CDM URL kebab-case 404; S5 FIBO IRI module-path 404; S4 T-file slug mismatch; S-number swap typos)
7Parallel-session Edit-tool stale-Read patternNEW Step 14 PARALLEL-SESSION SAFETY (ALWAYS Read-before-Edit + serial-vs-parallel dispatch tradeoff)S5 (~80 min Edit contention overhead with parallel S2.9b + S8 sessions)
8Logging-contract closure includes TRANSIENT sub-mode follow-onStep 11 LOGGING-CONTRACT CLOSURE → extended close-out to require TRANSIENT sub-mode retired or confirmed within same sessionS2.9b same-session follow-on extending S2.9 logging-contract

Diff stat: 506 → 618 lines (+112 / +22%); 12 → 14 numbered steps; Step 6e sub-clauses 4 → 9 (+5); Step 11 sub-clauses 1 → 2 (+1); §3 Evolution history sections 9 → 10 (+v3.7); cost-benefit summary tables 3 → 4 (+v3.7 incremental cost); CHANGELOG rows 7 → 8 (+v3.7 row at top).

§2 Reconsideration trigger NOT activated

Plan §2 Task 1 reconsideration trigger: “If a 9th candidate surfaces during authoring that materially changes the suite’s already-locked disciplines (e.g., a different maturity vocabulary that re-classifies 3+ prior spike outcomes), pause + ask Rich whether to (a) absorb into v3.7 + retroactively re-rate prior spikes, OR (b) defer to v3.8 + leave prior ratings stable.”

Triggered? NO. Tier 2 pgvector retrieval at /tmp/spike-1/prior-evidence.md (4 queries × top-10 results; sims 0.685-0.802; clustered around S2.9 + S2.6 + S5 + S8 + S2.10 + S3) confirms all hits cluster under the 8 already-consolidated candidates. No 9th theme surfaced.

Implication: prior spike outcomes (S1 VALIDATED, S2 MITIGATED, S2.5 VALIDATED, S3 MITIGATED, S2.6 MITIGATED, S4 VALIDATED-WITH-METHODOLOGICAL-SUBSTITUTION, S5 VALIDATED, S2.9 VALIDATED, S2.9b VALIDATED, S6 VALIDATED-WITH-NOTE, S7 VALIDATED-WITH-NOTE, S8 VALIDATED, S2.10 VALIDATED) STAND as recorded under v3.6 + ad-hoc post-S3 codifications. v3.7 disciplines apply prospectively from Spike 2 (MQ-* status review) onwards. No retroactive re-rating.

§3 §13 conflict resolution

Plan v1.6 §2 Step 7 said: “If §13 doesn’t exist yet, Spike 1 creates it with the sub-heading ### §13 Methodological / open-tension closure rows”. However arch-state §13 ALREADY EXISTED as ## 13. Pre-spike infrastructure for forthcoming 22-spike Q&A-formulation suite (created 2026-05-03T09:15 BST by the Tier 2 pgvector library index pre-spike infrastructure task). Plan v1.6 was authored at 09:35 BST without awareness of this.

Resolution: APPENDED Spike 1 row to existing §13 alongside the Tier 2 row. §13 title left as-is — semantically Spike 1 IS pre-spike methodological infrastructure for Spikes 2-22 (those use v3.7 disciplines explicitly). Did NOT create §14 — suite-relatedness is the stronger semantic match.

§4 Registry-version-collision observation

Spike 1 implementer subagent surfaced an informational concern: arch-state Changelog table now has TWO **3.30** rows (today’s at line 897, 2026-05-03T09:50; historical at line 922, 2026-05-01T17:00) and TWO **3.29** rows. Verified pattern is INTENTIONAL — line 917 marks the Full Phase-2-3 Reversion 2026-05-01T20:45 BST (**3.10** [FULL PHASE-2-3 REVERSION 2026-05-01T20:45]); after that reversion the version counter restarted from 3.11 and incremented through to today’s 3.30. The arch-state frontmatter version: field is the SINGLE source of truth for “current arch-state version” — set to “3.30” today. Historical rows preserved in changelog table for transparency.

No fix needed. Pattern is established. Future closures continue incrementing 3.31, 3.32, … from today’s 3.30.

§5 Cross-cutting compliance

  • feedback_logging_contract_closure_within_same_session — T-file authoring + arch-state §13 row + Q-cascade row (N/A for Spike 1) + MEMORY.md entry + this memory file + active-work-log update + plan §0 checkbox + plan CHANGELOG + plan lastmod bump ALL within ~70 min wall-clock from session start to closure verifier
  • feedback_paste_safety_for_terminal_handoffs — Write-tool used for all multi-line content in this session; no inline printf-to-file; no curl-through-symlink; no heredocs
  • feedback_test_theories_immediately_when_tabled — no new theories tabled during authoring that needed spike-validation; reconsideration trigger NOT activated
  • feedback_confront_richard_tasks_at_creation_time — no new richard-tasks created (Spike 1 has no richard-task disposition)
  • feedback_kill_condition_strict_vs_spirit_reading_via_outcome_MITIGATED — N/A (methodological-authoring spike has no kill-condition)
  • feedback_surface_alternatives_before_collapsing_synthesis_to_baseline — N/A (no kill-condition to hit)
  • T-file: ~/off-github/library/projects/inherit/T-spike-refined-prompt-v3.7-authoring-2026-05-03.md v1.0 (off-github commit 34924c2; 27,970 bytes / 187 lines)
  • Modified spec: ~/testatetech/docs-strategy/docs/superpowers/specs/2026-04-29-multi-phase-audit/refined-end-of-turn-directive.md v3.7 (docs-strategy commit 09712bb; 506 → 618 lines)
  • arch-state row: ~/testatetech/docs-strategy/docs/superpowers/specs/inherit-v2-architecture-state.md v3.30 §13 row Q-and-A-S1 (same commit 09712bb)
  • Plan: ~/testatetech/docs-strategy/docs/superpowers/plans/2026-05-03-22-spike-q-and-a-formulation-suite.md v1.7 (this commit pending; CHANGELOG row above v1.6 row)
  • Tier 2 prior-evidence: /tmp/spike-1/prior-evidence.md (4 queries × top-10 results filtered to eps-iota-t-file)
  • Consolidated 8 candidates: ~/.claude/projects/-home-richardd-testatetech/memory/feedback_kill_condition_maturity_vocabulary_v3_7_consolidated.md

§7 NEXT

Spike 2 (ADD-4 MQ-* meta-Q status review) — strict-ordered after Spike 1 per §2 Group 1. Spike 2 re-rates MQ-000..MQ-005 status fields against v3.7 disciplines. First spike to use v3.7 disciplines explicitly.

Per Rich-directive 2026-05-03 BST “no rush — 1 window at a time”: Spike 2 dispatches in a future session (re-paste the SPIKE-RUNNER handoff prompt; the §0 plan checkboxes + active-work-log entries are the resume anchor). This session may close after the closure verifier reports.