22-spike Spike 6 ADD-2 T-NEW tensions sweep CLOSED 2026-05-03

Outcome

outcome-VALIDATED on strict reading per refined-prompt v3.8 (Candidate 6 PLAN-FILE PRE-RESOLUTION promoted to LOCKED). Kill-condition NOT-MET: 15 T-NEW candidates surfaced (kill-condition was 0 tensions across all 17+ spike outcomes; 0 would have been suspicious — 15 within plan §4 Task 6 budget of 5-15).

What was swept

Source-row inventory (26 T-files + 6 cumulative-state + arch-state §§ + Tier 2):

  • 12 ε.ι T-files (S1, S2, S2.5, S2.6, S2.9, S2.10, S3, S4, S5, S6, S7, S8) at ~/off-github/library/projects/inherit/T-spike-eps-iota-*.md
  • 14 22-spike T-files (12 from dispatch inventory + 2 newly-found by parallel sessions): refined-prompt-v3.7 / mq-meta-q-status-review / phase-2-angles / zeta-q5 / zeta-q15 / rt214 / rt216 / rt218 / rt210-212 / vpc-peering / q016 / q017 + S6a-inherit-v2-merged-owl + S213-S217-pin-yaml-and-sssom-rows
  • 6 cumulative-state docs at ~/off-github/library/projects/inherit/*-cumulative-state-2026-05-02.md
  • Tier 2 pgvector library index T-file
  • arch-state §11/§12/§13/§14 + §4 amendment registry partial verification

174 honesty caveats extracted via Python yaml.safe_load to /tmp/spike-6/honesty-caveats-extracted.md.

15 T-NEW candidates — disposition

5 PROMOTE-TO-A-N at Rich-adjudication time (pre-staged at A-143..A-147 in T-file body §4):

A-NTopicTier-pairCost-row
A-143Faith-tradition Phase-1.5+ wave-per-pillar (vs current Phase-1 commitment)T1↔T2↔T3 cross-cutting-£240-450K/yr if approved (5 pillars × £80-150K each Phase-1.5+ deferred)
A-144VC issuance 2-stage: Phase-1 partner-firm-DID + Phase-1.5+ testator-DID at SI commencementT1↔T2↔T3Combine with A-133 Will-VC-issuance-pattern pre-staged at S5 closure
A-145G-IRI-VERIFY allow-list lifecycle codification at A-130 (BFO 2020 PURL-flakies)T1 (CI gate authorability)Documentation-only
A-146Multi-jurisdiction-customer Will-update + trigger fan-out Year-2+ deferT1↔T2↔T3-£40-80K/yr Phase-1 if approved (conflict-of-laws engine deferred)
A-147SSSOM 1.1 spec-evolution reconsideration trigger (curation_rule forward-incompatibility)T1↔T2Combine with A-131 ε.ι formal landing at Q-005 lock-time; ~£500/~1d down-rev cost if 1.1 retires curation_rule

1 ALREADY-ABSORBED at A-99/A-100/A-101 lock 2026-04-26:

  • T-NEW-S6-05 Texas + Florida + RON terminology drift in Q-015 caveat #11 + Q-016 caveat #5 (both misframe Texas+Florida as “Phase-1.5+” when A-99 already locks them as Phase-1; Phase-1.5+ framing is for cantonal-Erbenbescheinigung-equivalents and additional state-level US jurisdictions BEYOND Texas + Florida). Surface-for-Rich at Q-015/Q-016 lock-time as terminology consistency check.

1 PROMOTE-TO-R-NEW for Spike 7 risk register update v1.2:

  • T-NEW-S6-12 Hindu pillar cultural-alignment risk MEDIUM-HIGH (T49 evidence; £80-150K/yr per pillar run-rate). Spike 7 should ALSO inspect S6-14 (SSSOM 1.1 forward-incompatibility) + S6-03 (G-IRI-VERIFY allow-list) for additional R-NEW promotion candidates.

8 DOCUMENTED-NOT-PROMOTED as operational follow-on / non-load-bearing for arch-state lock:

  • S6-06 LinkML annotation surface separation (covered by feedback_linkml_annotation_surface_separation)
  • S6-07 Cedar 4.10 DateTime workaround via Long epoch_days (richard-task #220 Catala-side)
  • S6-08 G-CCO-OWL-LOAD Phase-1.5 candidate (operational discipline)
  • S6-09 i-ζ:FinancialInstrument collision with cco:FinancialInstrument (covered by ω.η Q-004 + Q-003 locks)
  • S6-10 Postgres provisioning blocker (admin-debt; richard-task #209 covers)
  • S6-11 bge-small-en-v1.5 multi-language gap Year-2+ (richard-task candidate)
  • S6-13 Plan-defect detection rate ~55% per spike (methodological evidence supporting v3.7 → v3.8 Candidate 6 LOCKED — already promoted)
  • S6-15 Anthropic Managed Agents pricing reconsideration (covered by S18 4-trigger framework)

Tier-pair pattern + root cause

Dominant tier-pair: T1↔T2↔T3 cross-cutting on jurisdiction scope (S6-01 + S6-02 + S6-04). Secondary T1 commitments on alignment-axiom + CI gates (S6-03 + S6-14).

Most-recurring root cause (3 distinct spike outcomes converge on it): faith-tradition Phase-1 vs Phase-1.5+ scope tension:

  1. S5 ζ-Q15 directly via T49 cultural-alignment risk MEDIUM-HIGH for Hindu pillar £80-150K/yr per pillar
  2. S22 Q-017 trigger-event detection partial coverage gap for civil-registry triggers Year-2+ across faith jurisdictions
  3. S21 Q-016 Will-update faith-tradition witness-pattern-mapping (Islamic two-male / Hindu pandit-presence / Jewish two-kosher-witness)

Single largest cost-row impact lever: T-NEW-S6-01 (faith-tradition Phase-1.5+ wave-per-pillar) + T-NEW-S6-04 (multi-jurisdiction-customer fan-out Year-2+ defer) approved together at Q-015/Q-016/Q-017 lock-time would reduce Phase-1 by ~£280-530K/yr; aligns with #220 E&W IHT SCOPE-DOWN narrowing-but-deepening for first-mover positioning capture.

Plan-defect detection (v3.8-LOCKED Candidate 6 PRE-RESOLUTION)

Plan §4 Task 6: zero defects detected. 4th post-Spike-1 run with zero defects (after Spikes 1, 10, 12, 18, 21, 22 also clean). Detection-rate update: ~55% across Spikes 2-22 closures so far (5 detected of 11 closed: Spikes 2 + 3 + 4 + 8 + 9; 6 zero-defect: Spikes 1, 6, 10, 12, 18, 21, 22).

T-NEW-S6-13 documented-not-promoted as methodological evidence supporting Candidate 6 LOCKED in v3.8 (already promoted by parallel session same day).

Reconsideration triggers preserved (T-file §8)

  • A-143 — 5y partner-firm cultural-alignment evidence accumulates (Hindu / Sikh / Jewish / Islamic / Buddhist 5 pillars; success → Phase-1 re-table)
  • A-144 — Law Commission Modernising Wills Law 2025 enabling-power SI commencement → testator-facing electronic Will execution becomes legally valid → λ.ε standalone re-tabled
  • A-145 — BFO 2020 PURL fix upstream OR allow-list grows >5% of total IRI universe → revisit categorisation
  • A-146 — 5-firm Phase-1 pilot multi-jurisdiction case-rate >5% of estate-volume → conflict-of-laws engine pull-forward to Phase-1.5
  • A-147 — SSSOM 1.1 final spec publication → curation_rule upstream stable / retired determines κ.δ ★ vs κ.γ down-rev

Cascade artefacts

ArtefactPath / commit
T-file v1.0~/off-github/library/projects/inherit/T-spike-t-new-tensions-sweep-2026-05-03.md (347 lines / 42 KB) — off-github commit f642ee0 (push verified)
Tension candidates/tmp/spike-6/tension-candidates.md (15 rows; working artefact)
Honesty caveats parsed/tmp/spike-6/honesty-caveats-extracted.md (174 caveats from 26 T-files)
Tier 2 evidence/tmp/spike-6/tier2-queries-summary.md (4 queries × 10 results each)
arch-state rowv3.38 → v3.39 §13 row Q-and-A-S6 + Changelog v3.39 — docs-strategy commit 1d6e984 (push verified)
Plan §0 checkboxSpike 6 ticked + plan v1.23 → v1.24 + CHANGELOG row — same docs-strategy commit
INDEX.mdNO bump (Spike 6 architectural-debt closure NOT forthcoming-Q-formulation)
docs-personalNO commit (no richard-task disposition; A-N amendments pending Rich-adjudication land at relevant Q-lock times)
Memory filethis file
MEMORY.mdnew entry prepended above most-recent
Active-work-logSpike 6 CLOSED entry

Methodological observations

  • 13th spike of 22-spike suite to close (after Spikes 1+12+18+22+2+21+3+8+9+10+4+5)
  • FIRST architectural-debt closure spike (Group 3 Spike 6 of Batch C; Spike 7 risk register update STRICT-AFTER same Batch C session inline per plan §4 strict-order)
  • Parallel session DISCOVERED 2 additional T-files (S6a-inherit-v2-merged-owl + S213-S217-pin-yaml-and-sssom-rows) added during the session — subagent’s count of 14 22-spike T-files reflects this; my dispatch prompt’s count of 12 was stale within ~1h
  • Pin-drift hook —fix encountered known nested-paren bug at long-form arch-state lastmod prose; manually rewrote both floor-pins (arch-state v3.39 + refined-prompt v3.8) to canonical clean form "vX.Y (YYYY-MM-DDThh:mm)" per parent CLAUDE.md §6
  • Candidate 6 LOCKED in v3.8 by parallel session same day (independent of this Spike); my Spike 6 closure pin-bumped refined-prompt floor v3.7 → v3.8 to reflect locked discipline
  • Spike 7 R-NEW substrate ready at T-NEW-S6-12 (Hindu pillar) + S6-14 (SSSOM 1.1) + S6-03 (G-IRI-VERIFY) for Spike 7 risk register update next inline this Batch C session