Consolidation run — 2026-07-28 (run 6)

Branch: claude/consolidate-2026-07-28. Workflow: wf_c4dced48-cbd — 25 agents, 25/25 completed, 0 dropped, 0 errored (clean run; no salvage). Seats per .claude/environment-register.md (2026-07-26): all subagents Opus 5 — adjudicators (claim-promotion-reviewer, claim-refuter) at high effort, citation-verifier at medium; main thread Fable. Seat config re-confirmed against the user's run-6 kickoff directive ("opus 5 mostly … fable is expensive"); the register already codified exactly that, so no change was needed.

Run scope

  • C1 — subtraction: claim-retirement-queue-2026-07-28.md, generated with --staleness-days 60 (run-4/5 precedent; the same-day---force trap that bit run 5 does not apply — this is a first run of its date). 99 rows, all keep / review (staleness), 0 concrete-defect rows — nothing to apply. Third consecutive all-keep queue.
  • C2 — candidate triage: skipped by scope. All 130 candidates carry a prior consolidation triage line (run 5 discharged the last never-triaged 10), and no ingest since run 5 has changed any candidate's state, so every KEEP promotion condition is still unmet. Re-triaging unchanged candidates is the skill's own named failure mode ("KEEP without a promotion condition → C2 just relabels the backlog"). C2 re-opens when an ingest or repair moves a candidate's state.
  • C3 — supported jury: the 5 oldest never-juried supported-eligible live claims, of 75 mechanically eligible. All five were created at live on 2026-05-05 in the audit Phase-8 ninth run and had no Status History activity since (beyond the 2026-07-11 lifecycle backfill). Cap 5 of 5 consumed.
  • C4 — backlog batch: skipped (usage conservation per the kickoff directive; the ~15 remaining v1.2-pilot speculate survivors are the next batch).
  • Parked ledger: read at C0 — empty; nothing to unpark. Nothing parked this run.

Status changes

None. 0 promotions, 0 demotions, 0 retirements. All five juried claims stay at live, confidence unchanged at medium. The refuter panel returned 15/15 REFUTED under unanimity default-deny, carrying the all-time tally to 131/131 REFUTED across runs 1–6.

Because no status moved, there is no veto row this run — a veto surface exists only where a status changed. The reviewable actions here are the citation repairs and Counterpressure writes listed below; to reverse any of them, say revert <slug> and the entry returns to its pre-run text.

Claim Verifier 5-test Refuters Outcome
mp-teleology-non-naturalizable 3 PASS / 0 UNC / 2 FAIL HOLD (CONCERNS t2, t5) 3× REFUTED blocked, stays live
non-philosophie-as-empietement-of-refused-world 3 PASS / 0 UNC / 0 FAIL REQUEST-REVISION (CONCERNS t5) 3× REFUTED blocked, stays live
chaui-aristotle-inversion-as-mp-revolution 4 PASS / 1 UNC / 0 FAIL HOLD (CONCERNS t2, t4, t5) 3× REFUTED blocked, stays live
topology-from-piaget-not-heidegger-not-lacan 4 PASS / 0 UNC / 0 FAIL HOLD (CONCERNS t5) 3× REFUTED blocked, stays live
vie-du-lien-prior-to-being-toward-death 2 PASS / 2 UNC / 0 FAIL HOLD (CONCERNS t2, t5) 3× REFUTED blocked, stays live

Blocked promotions — where each objection now lives

Every REFUTED objection was written into the entry's Counterpressure (the blocked-promotions-must-strengthen-the-register rule). Fifteen objection bullets were added across the five entries, each tagged (Consolidation 2026-07-28…) with its lens.

A structural pattern worth naming: all five entries had recorded their own promotion preconditions at creation on 2026-05-05, and all five preconditions remain unmet 84 days later. Three of the five 5-test reviewers reached the block by that route independently. This is the gate working as designed — but it also means these five were never close, and the pre-gate's "mechanically clean" signal does not see self-set preconditions. Worth considering as a future pre-gate check.

Citation repairs applied (Rule-16 class)

Every repair below was independently re-verified on the main thread against raw/ before applying — the skill's documented reviewer-false-positive rule. All were confirmed; none was a false positive.

  1. mp-teleology-non-naturalizable — false footnote attribution (2 sites). Two Evidence bullets asserted that Halák's fn. 127 cites Beith, Birth of Sense, 38–41 "as corroborating the non-naturalizable reading". Fn. 127 (raw 2423) cites Thompson and Sheredos; Beith is at fn. 132 (raw 2441) — whose text moreover disclaims the corroboration ("it does not consider how organismal dynamics is dealt with within the autopoietic and organizational approaches themselves"). The extraction note had fn. 127 right for Thompson; the drift was introduced at claim-authoring.
  2. mp-teleology-non-naturalizable — two mislocated quotes. The "two symmetrical errors" quote sits at raw 2395 / printed p. 136, outside its bullet's stated 2417–2455; the "primordial establishment of the problematic" quote occurs only at Ch 2 §2.2 / p. 122 / raw 2081, not at the cited Ch 5 §5 range (the extraction note had this one right). Both split out.
  3. non-philosophie-as-empietement-of-refused-world — the title-level corrective was unanchored. All three Evidence bullets cited Ch. II §1; the passage carrying the interdisciplinarity denial (naming Lévi-Strauss, Guilbaud, Benveniste, Lacan) is at Ch. II §2, raw 887. Verified: "interdisciplinarité" and "Lévi-Strauss" each occur exactly once in the volume, both on raw 887 — outside every range the entry cited. A fourth bullet now anchors it.
  4. chaui-aristotle-inversion-as-mp-revolution — pointer transposition + a structural misattribution. The Chauí-2009-p.45 quote was cited at raw 4378 (bullet 2's anchor) but is at raw 4398, fn. 14 at 4404. More consequentially: the PhP 174 sentence the claim calls its anchor is not in the Freud/Marx note — it closes the sexed-body chapter at raw 2676, and the note opens at raw 2678. The AdV bullet, the only one with no locator at all, gained one.
  5. topology-from-piaget-not-heidegger-not-lacan — off-by-two. The promiscuité de l'Être formula is at raw 1622, not 1620; corrected in Evidence bullet 3 and in Payoff item 3, which repeated it.
  6. vie-du-lien-prior-to-being-toward-death — commentator's prose quoted as MP's, plus an unlocated genre misattribution. Two sentences introduced as "MP's response" are Saint Aubert's own unquoted expository prose at raw 1191, in Chapitre iii — one full chapter outside the cited Ch. V range. And the entry's only V&I anchor (the only thing making this a two-source claim) carried no locator at all and called the passage a working note: the principe des principes is main text of V&I Part 2 "Interrogation and Dialectic", Lingis p. 84 / Fr. p. 117, aimed at Sartre. Recorded on the same bullet: the short formula the Claim field quotes is not in V&I at all — it exists only at SA raw 1499 under fn. 818 ("NPVIf [172]v"), an unpublished feuillet not in raw/, so the entry's title-bearing quotation is uncloseable in-corpus. The mislocation traces upstream to wiki/.audit/synthetic-layer-2026-05-05-ninth-run.md.

Cite-back repairs (concept pages)

  • wiki/entities/marilena-chaui.md — stale label (candidate) on a live claim; synced.
  • wiki/concepts/sein-zum-tode.md — source misattribution, 3 sites (frontmatter sources:, body line 64, Sources line 78): the page named saintaubert-2021-etre-et-chair-ii as "the textual locus for the vie du lien contestation of être-pour-la-mort". Verified: that extraction note contains no occurrence of "vie du lien", "être-pour-la-mort", or even the substring "mort". Corrected to SA-2006 Ch. V with raw locators, and the body gained SA's own subordonne-not-repousse qualification. updated: bumped (citation-interpretation-bearing, Rule 3). Caught by treating the reviewer's single cited site as an instance rather than the extent — run-5's lesson.

New pagination oracle

raw/Merleau-Ponty_Séglard_2003_Nature_course_notes.md folio markers run +16 against printed English folios. Printed Nature 152 = marker {168}; printed 155 = {171}. Main-thread verified by mapping markers {166}–{174} against the axolotl/retrospective-illusion and two-errors passages. Without this, a future verifier reading {152} in that file would wrongly report the cited passages absent. Recorded inline on the claim's Nature Evidence bullet, and joins the Derrida / Malraux / CPP / V&I oracles.

Caps consumed

  • C1 status changes: 0 of ≤15
  • C2 candidates triaged: 0 (skipped by scope)
  • C3 supported promotions: 0 of ≤5 (5 of 5 jury slots used)
  • C4: not consumed

Remaining backlog (re-trigger signal)

  • C3: 70 supported-eligible live claims still unjuried (75 at C0 minus the 5 juried; the 5 juried re-enter the pre-gate after the 14-day settle window, and only once their objections have been addressed).
  • C2: dry until an ingest or repair moves a candidate's state.
  • C4: ~15 v1.2-pilot speculate survivors awaiting a batch.
  • 67 live claims remain mechanically blocked on unanchored Evidence bullets — unchanged this run.

The register is not near dry on the C3 side. Another /consolidate run is the downstream trigger.

Files touched

  • wiki/claims.md — 5 entries: Evidence repairs, 15 Counterpressure objection bullets, 5 Status History prose lines, 5 Updated: bumps; Summary table regenerated by normalize_claims_table.py --write
  • wiki/index.md — Claims section + counts + one Concepts Updated cell (regenerated; not hand-edited)
  • wiki/concepts/sein-zum-tode.md, wiki/entities/marilena-chaui.md — cite-back repairs
  • wiki/.audit/consolidation-pregate-2026-07-28.md, claim-retirement-queue-2026-07-28.md, consolidation-2026-07-28-verdicts.json, consolidation-2026-07-28-adjudications.md, this digest
  • Baseline fixtures refreshed by refresh_baselines.py --apply (Rule 19); 590/590 tests pass; lint baseline gate green (906 pages / 303 claims)

Consolidation — 2026-07-28, run 7 (third run of the date)

Same-day mechanics (user-authorized by the run-7 /consolidate invocation). Run 6 consumed the date key this morning (its commit 5dff3f4 sits unpushed on claude/consolidate-2026-07-28). Per the run-5 precedent: this section APPENDS to the run-6 digest (run 6's content above is untouched); the pregate was regenerated in place and the retirement queue regenerated with --apply --force --staleness-days 60 — the scaled window re-passed explicitly, per the run-5 Codex-P2 trap; run-7 verdict artifacts carry a -run7 filename suffix (consolidation-2026-07-28-run7-verdicts.json / -run7-adjudications.md, generated via gen_adjudications.py's argv path parameterization with the provenance field set before generation); the run branch is claude/consolidate-2026-07-28-run7, stacked on the unmerged run-6 commit so one PR will carry both single-run commits; every run-7 Status History tail carries run 7 plus the workflow id (wf_4199fe6a-fa7). User scope directives: C2 skipped ("no c2 in backlog"), C4 mandated ("run the cfour with opus 5 agents," effort delegated to the maintainer). Veto protocol unchanged: say "veto <slug>" — reverts to the pre-run state.

Run scope (fixed at C0)

Register at C0: 303 entries — 23 supported, 142 live, 130 candidate, 3 contested, 5 retired; normalize_claims_table.py --check in sync. Parked ledger read — empty; nothing to unpark.

  • C1 — subtraction: queue regenerated at the 60d window (--force, same-day): 94 rows, all keep/staleness, 0 concrete-defect rows — sixth consecutive clean queue. Nothing to apply; 0/15 cap consumed.
  • C2 — skipped by user scope, coinciding with run 6's durable rule: the candidate backlog has been dry since run 5 and no ingest has moved any candidate's state.
  • C3 — the run-7 process innovation: a main-thread self-set-precondition screen before batch selection, implementing run 6's structural finding (self-set preconditions are invisible to the mechanical pre-gate). The eligibility frontier was read entry-by-entry in Updated: order and every entry whose own Status History or Counterpressure records an undischarged supported-precondition was deferred with its condition named, instead of consuming a jury slot it was pre-doomed to fail. Screened out (10): anticipation-retroaction-as-temporal-signature-of-life (requires corroborating Halák work or a Whitehead ingest), taddio-secret-science… (speculative-confidence entry whose own record requires uptake/H_synth engagement), chapsal-1960… (its 05-09 reviewer marked conditions (a)/(b) as the supported-preconditions), being-is-power… (family unification + Chouraqui-2014/2016 verification), erde-to-copernican… (self-mapped route-unification, both route-partners mechanically blocked), and the whole 05-16 audit-14 cluster — conquering-language… (PoW p. 141 raw-check + operative-equivalence adjudication), mp-1953-anticipates-1955… (1953→1955 action-typology bridge), mp-principal-condition-1945… (PoP-1945 footnote re-check + Brussels-1946 adjudication), sartre-too-antithetic… (AdV Ch. 5 passage-level juxtaposition), pbp-as-middle-term… (PoW+PbP- together adjudication). Jury batch = the 5 oldest screened-clean: bn-freedom-not-applicable-to-historical-man, museum-stiftung-against-hegel-and-malraux, sartre-childhood-as-philosophical-genealogy, whiteness-as-institutional-habit (all 05-09 thirteenth-run Layer-2 creations, which lack precondition lines), hegel-1807-philological-source-of-sache-selbst (05-19; its only stated condition was for live and was discharged by the recorded 05-19 raw-verification promotion). Cap 5 of 5 consumed.
  • C4 — speculate-signoff batch 4 (2026-07-28-batch4, second consolidation-embedded batch): 6 survivors from the 15 remaining v1.2-pilot pool, selected by remaining-ranking order + seam coverage + pair integrity — promiscuite-as-topological-anerkennung-§671 (motif, rank 15), vie-du-lien-constitutive-mp-heidegger-tension (aporia, 18) with its paired rival vie-du-lien-vs-tode-different-registers-no-rivalry (aporia, 30; batch-2 paired-rival precedent — and a built-in register-drift test, since their baseline live claim vie-du-lien-prior-to-being-toward-death was heavily repaired at run 6 this morning), indirect-vs-direct-continuous-counter-traditional-method (aporia, 19; its rival #32 deferred to batch 5 for the cross-batch consistency test), kosmotheoros-disavowal-as-positive-structure (subtext, 21), bachelard-mediates-schelling-to-mp (genealogy, 23). All four seams covered. No convergent finding → juror-C not spawned (6 agents). The two formerly Knight-blocked survivors are now unblocked (Knight 2024 ingested; IDL-1 sunset) and deliberately queued for batch 5. Outcomes: see the C4 section below.
  • Seats. C3 via workflow wf_4199fe6a-fa7 (25 peer-isolated Opus 5 seats: adjudicators high, verifier medium, per .claude/environment-register.md) — 25/25 clean, 0 dropped, 0 errored, 11.1 min, ~2.34M subagent tokens; third consecutive fully clean adjudication workflow. C4 via the speculate-promote-jury team with every seat on Opus 5 including the lead (user directive; run 5 had run the lead on session top-tier). Effort determination (delegated to the maintainer): the register's adjudicatory high profile is the determination of record; mechanically the Agent-tool team spawn path exposes no per-seat effort override, so seats run at the harness's subagent default — accepted deliberately, procedure fidelity (persistent team + broadcast debate) outranking the effort knob; the main thread (Fable) supervises ratification.

C3 — Supported jury (5 juried, 0 promoted, 0 demoted)

All five blocked; every entry stays live at medium, all seat-named repairs applied in-run, all 15 refuter objections recorded on the entries. Refuters 15/15 REFUTED → 146/146 all-time across runs 1–7.

Claim Verifier 5-test Refuters Outcome
bn-freedom-not-applicable-to-historical-man 2 PASS / 1 UNC / 1 FAIL REQUEST-REVISION 3× REFUTED blocked, stays live
museum-stiftung-against-hegel-and-malraux 3 PASS / 1 UNC / 0 FAIL REQUEST-REVISION 3× REFUTED blocked, stays live
sartre-childhood-as-philosophical-genealogy 2 PASS / 1 UNC / 0 FAIL HOLD 3× REFUTED blocked, stays live
whiteness-as-institutional-habit 2 PASS / 1 UNC / 0 FAIL REQUEST-REVISION 3× REFUTED blocked, stays live
hegel-1807-philological-source-of-sache-selbst 3 PASS / 0 UNC / 0 FAIL HOLD 3× REFUTED blocked, stays live

hegel-1807 produced the jury program's first fully clean verifier sheet (runs 4–7) and was still blocked — the block is thesis-scope ("source" vs attestation; an inheritance conjunct added at claim-authoring that the extraction-note seed never carried), not anchor quality. The pattern of runs 1–7 is intact: reviewers rate the cores sound at live; supported bids fail on cross-source anchor quality, undischarged self-set preconditions, and Counterpressure that concedes rather than defeats.

Rule-16 findings (prominent, per the escalation triage)

  1. bn-freedom — fabricated quotation, repair-class (the run's headline finding). Evidence bullet 1 presented "Like all heroes, this hero who had not chosen this risk takes it as his own" as a verbatim manuscript pp. 147–148 quotation at a fabricated anchor ("Pass 2a entries 19–20"; the note has entries 1–10). Main-thread re-verified repo-wide: the string exists nowhere in the corpus except the Evidence bullet itself; the raw p. 147 paragraph's actual sentence is the mother-analogy ("she took the risk, and consequently he is hers and she is his") — the fabricated quotation re-cast it in hero vocabulary at claim-authoring (the extraction note's Part D seed carries no such quote). Adjudicated repair-class, not halt-class: the work is real and ingested, the locator is real, the provenance chain is fully identifiable — the halt pattern (invented citation to an unread work, the Faul incident) did not occur; runs 4–6 handled the same paraphrase-as-quotation class as repairs. Struck and re-quoted from the verified raw passage with the reaffirming clauses ("which is what I still claim … Is there, thus, another kind? No") restored.
  2. museum-stiftung — inverted counterpressure + unanchored specifying glosses. The Bishop bullet asserted the opposite of the chapter (Kaushik criticizes Bishop, raw 3387/3397; error inherited from the extraction note's Because-leg compression); the Claim's two parenthetical glosses occur nowhere in ch. 12 ("triumph of creation" nowhere in the volume), and MP's Signs assigns reconciliation-in-the-museum to Malraux.
  3. sartre-childhood — edition false friend + two fn-18 misattributions. The Signs bullet hung French Signes (Gallimard 1960) folios pp. 34–35 on the English McCleary wikilink (whose front matter is an Introduction, not a Preface; English location ≈ pp. 25–27, offset-derived); fn. 18 concerns the word envelopment only, yet was cited for genealogical softening in both sartre-childhood and bn-freedom. The unanchored "ENS-modeled" coinage contradicted the manuscript and the entry's own homes; the 1953-downstream sentence had no anchor (one exists at raw 169 and was added) and no inference marker.
  4. whiteness — Foucault misattribution. Counterpressure said Ahmed "engages but does not concede" Foucault; Foucault occurs nowhere in ch. 10 (main-thread grep of the chapter range) — the volume's Foucault engagements belong to Flynn ch. 8 and Chouraqui ch. 9.
  5. hegel-1807 — paraphrase-as-quotation. "Hegel's claim to total synthesis is the Museum of philosophy" was set in quotation marks; the phrase occurs nowhere in the Signs raw (MP: "Hegel is the Museum. He is if you wish all philosophies, but deprived of their finiteness and power of impact") — the Faul form with real work/page/sense; repaired with the actual sentence.

New pagination oracle + seat-conflict adjudication

raw/Merleau-Ponty and Contemporary Philosophy.md (SUNY 2019) runs {N} = printed + 12, volume-wide. Surfaced independently by three seats (museum scope refuter; whiteness verifier; whiteness scope refuter) against one dissenting seat (the museum verifier's "no offset oracle needed"). Main-thread adjudicated by the volume's own TOC, per the run-4 CPP method: Kaushik printed p. 253 / chapter heading under {265}; Ahmed printed 197 / {209}. The extraction note's entire Pass 2a location column carries the +12 (nine chapters enumerated in the scope objection), as do its five Part D candidates — the in-batch locators were repaired (museum 265–278 → 253–268; whiteness 209–231 → 197–219); the volume-wide sweep (extraction note + being-is-power…, flesh-three-senses…, flesh-panpsychism… and any other dependents) is routed to the next audit Citations phase. Joins the Derrida +10 / CPP +16 / Nature +16 / Malraux non-uniform oracle family. Also recorded en passant: raw Signs = printed folio + ~25–27 (seat estimates diverged; per standing discipline, per-locator verification, never blanket).

Cite-back sweep (no status moved; label/citation syncs only)

Jury-named propagations plus the standard post-apply extent grep (run-5 lesson), which caught two sites beyond the jury's named list — a second verbatim copy of the old sartre-childhood wording on wiki/entities/jean-paul-sartre.md, and the old whiteness wording on wiki/concepts/politics-mp.md. Synced surfaces: sache-selbst.md (stale "(candidate)" labels ×2 → live; GA 14 pp. 73–76 → 75–78 ×2), motifs.md §die Sache selbst (stale candidate label; the settled "Hegel → … → MP" transmission line hedged as disputed; GA-14 range; the "philological source" predicate marked disputed), maurice-merleau-ponty.md (×2 sites), jean-paul-sartre.md (×2 sites), two-ways-of-being-young.md (×4 French-pagination sites), habitual-body.md, politics-mp.md, merleau-ponty-1964-signs.md (Connections locator). Frontmatter updated: bumped on all seven touched pages (citation-interpretation-bearing, Rule 3). Plus one metadata-only backfill (no bump, logged here): raw_file: added to alloa-chouraqui-kaushik-2019-contemporary-philosophy.md.

Status changes

None. 0 promotions, 0 demotions, 0 retirements across C1–C3. As in run 6, no veto row exists because no status moved; the reviewable actions are the citation repairs, Counterpressure additions, and cite-back syncs above — to reverse any of them, say revert <slug> (register entry) or name the page. C4 ratification actions are listed in the C4 section below with their own veto paths.

Audit-routed items (outside consolidation write-scope)

  • The MP&CP 2019 extraction note's volume-wide +12 location column (15 Pass 2a items + 5 Part D candidates) and its Bishop Because-leg compression (raw 3387/3397 counter-evidence).
  • The Sartre-1984 extraction note's footer-convention inconsistencies (its own p. 146 / p. 131 attributions fall one page later under the validated footer convention) and its double-voiced parallel-discovery location (p. 129 vs pp. 138–139).
  • The merleau-ponty-1964-signs page's English-folio pinning for the two-ways passage (targeted print check; offset-derived ≈ pp. 25–27).
  • "Habit" is absent from the I&P source page and extraction note, leaving institution's "Not habit" Key Point wiki-side unanchored (raw I&P 1564 verified this run; needs a Rule-16 anchor on the wiki side).
  • The merleau-ponty-1945-phenomenology-of-perception source page lacks the comet/"trails behind" passage in Key Passages (currently raw-only + body-schema).

C4 — Speculate-signoff batch 4 (2026-07-28-batch4)

Second consolidation-embedded batch; first with every seat on Opus 5 including the lead (user directive; MANIFEST D4). Verdicts: 6/6 converged, all 4-of-4 unanimous, 0 escalations (parked ledger stays empty), 0 standing lead overrides — 1 PROMOTE-CANDIDATE · 4 HOLD-AS-SPECULATIVE · 1 RETIRE-AS-FALSE-FRIEND. Five converged at round 1; survivor 5 (kosmotheoros) at round 2 after a 2-2 split, with the batch's only probe movement (subtext FIRED-NARROW → FIRED-WIDE). 6 of 24 seats changed token Phase 1 → Phase 3, every change attributed to a named peer finding; three of the six moved against the changing juror's own prior argument; 11 seats raised confidence, zero lowered it.

Phase 6 ratification (main thread) applied:

  • One claims.md entry at candidate/speculative: claims#promiscuite-as-topological-anerkennung-671 (ASCII slug — the survivor's § form aborts normalize_claims_table.py), authored with the jury's six-item pre-registration repair set executed in full (three seats made items 1–3 vote-conditions): rejection axis restored to asymmetric-dominance figures in Claim and falsification condition; authored neutral substitute descriptor (reciprocal mutual-implication); sources re-attributed SA-2023 → SA-2006 (zero promiscuit* on the SA-2023 pair — main-thread re-verified; the IDL-2 wrong-volume class); Hegel slug/status-label corrected; phantom self-referential Evidence bullet struck; homes substituted to recognition-and-institution + master-slave-dialectic; plus G's seventh item (E's internal-resistance discriminator authored in, R's batch-made caveat carried). All four residual objections sit in the entry's Counterpressure; the unrun eleven-attestation bidirectionality check is the named unrepairable killer.
  • One false-friend caution: pensee-de-survol §Open Questions (the s5 RETIRE — verbatim converged text, honoring the three binding wording constraints: section-heading citations, quarantine path in-text, no restatement assertion). (Corrected 2026-07-28 post-ratification review: a second caution was written to sein-zum-tode §Open Questions from the s2 paired recommendation and has been removed. Survivor 2 converged on HOLD-AS-SPECULATIVE, whose mapped Phase-6 action is a SPECULATIONS.md annotation; a concept-page caution is RETIRE's action, and no protocol amendment was sought. The verdict and the annotation stand; see the Errata section below and ADJUDICATIONS.md §"Main-thread errata" E-2.)
  • Six SPECULATIONS.md jury-verdict annotations (survivor pool now 9 remaining of 32).
  • No ingest-dependency entry for s6: the lead's Phase-4 "route to both" extraction invoked a juror's conditional fallback whose condition the team had already resolved — a substantive-override near-miss caught by juror R after shutdown, withdrawn by the lead, and replaced in ingest-dependency-ledger.md by a withdrawal record preserving the four in-corpus anchors (logged as MANIFEST D11 + cohort-fingerprint obs. 6; live evidence for the SKILL's named lead-extraction weakness and its proposed v1.1 fresh-context extraction-checker).
  • Lead-owned quarantine appends stand: promotion-ledger batch-4 triage entry; methodological-queue MQ-9 … MQ-17 (0 from ROUTE verdicts — all cohort observations / panel-routed residues, incl. the two-key coverage gap, the misreporting class, the sign-off provenance gap, and Juror I's Phase-0c page-body scope question).

Run-design findings worth carrying:

  • The register-drift briefing changed an outcome. Survivor 3's Restatement audit was written against the pre-repair vie-du-lien-prior-to-being-toward-death; against the same-morning post-repair entry it near-duplicates Counterpressure deposited hours earlier. Without the C0 briefing the batch would plausibly have registered a candidate restating run 6's own repair.
  • Same-gap-twice trigger 2 fired for the fourth consecutive batch (Test 2 at 6/6), but the diagnosis bifurcated: half anchor-stripping (MQ-8 class), half correctly-cited anchors that argue against the claim — a class citation-hygiene checks are structurally blind to (MQ-12; survivor 4 had the batch's cleanest citations and its worst self-description).
  • Harness adaptation (MANIFEST D9): the harness forbids a teammate spawning named/background agents, so the five seats ran as persistent synchronous subagents resumed by agentId — Phase-1 parallelism via concurrent spawns, persistence via transcript resumption; juror completion notifications bubbled to the session main thread, which relayed each to the lead (run-5 quirk, now with its mechanism explained).
  • Four Harvest nominations travel with the HOLDs (three seats warn a HOLD whose nomination does not travel is a quiet retirement): s2's direct-contrast redraft (SA-2006 + Sein und Zeit); s3's subordination-with-preservation form (SA raw 1507/1497, after the parent's conjunct separation); s4's reprise/lineage-placement form (Madison Appendix II args 36–42, raw 2987–3168; extraction l. 209; nominate at candidate) — three seats call this the batch's most consequential loss if dropped; s6's four in-corpus Bachelard anchors (CPP ch. 2 §III; CPP chunk-B §VI.D — two sites, cited separately; Inédits II extraction l. 486; SA-2006 raw 1618 as context only). All require their own slugs (G's restoration/installation rule); recorded in NEXT.md for the next audit Harvest.

Veto paths for the C4 writes: veto promiscuite-as-topological-anerkennung-671 — removes the candidate entry; veto batch4-<survivor-slug> — removes that survivor's SPECULATIONS annotation; veto caution pensee-de-survol — removes that caution. Each is a single revert on this run's one commit. (The veto caution sein-zum-tode path is spent: that caution was removed at the 2026-07-28 post-ratification review — see Errata below.)

Status changes (run-7 totals)

Slug Change Rationale Veto
claims#promiscuite-as-topological-anerkennung-671 (new entry) created as candidate, confidence speculative Speculate-signoff batch-4 4-0 unanimous PROMOTE-CANDIDATE (round 1); six-item pre-registration repair set executed at registration; four residual objections in Counterpressure veto: say "veto promiscuite-as-topological-anerkennung-671" — removes the entry (it returns to SPECULATIONS.md as annotation-only)

No existing entry moved status: 0 promotions, 0 demotions, 0 retirements across C1–C3; C4 added one candidate. Register: 303 → 304 (23 supported / 142 live / 131 candidate / 3 contested / 5 retired).

Caps consumed / backlog remaining

C1 0/15 · C2 skipped (user scope; backlog dry) · C3 5/5 juried, 0/5 promotions · C4 one batch, 6 survivors (cap ceiling). Remaining backlog: 65 supported-eligible live claims unjuried (70 at C0 − 5 juried; of those 65, 10 carry screened self-set preconditions with named discharge conditions, incl. the whole 05-16 audit-14 cluster — the next clean frontier starts at hegel-arbeit-anchors-capital-as-phenomenology-motif and hegel-mature-recognition-is-forgiveness-not-master-slave); C2 dry until an ingest or repair moves a candidate; 9 v1.2-pilot speculate survivors remain for C4 batch 5 (incl. the two now-unblocked Knight-dependent survivors and s4's deferred paired rival, which batch 5 should adjudicate against this batch's s4 verdict); the ~82-row harvest reservoir plus this run's 4 jury Harvest nominations. The self-set-precondition pre-gate check (run 6's queued improvement) remains the highest-value tooling item — this run implemented it manually as the C3 screen. Next trigger: another /consolidate run.

Files touched

wiki/claims.md (5 C3 entries: seat-named repairs incl. the fabricated- quotation strike, 15 verbatim refuter objections, 5 SH lines, 5 Updated: bumps; + 1 new C4 candidate entry appended) · cite-back/caution surfaces (11): sache-selbst.md, two-ways-of-being-young.md, habitual-body.md, politics-mp.md, pensee-de-survol.md (+caution), sein-zum-tode.md (cite-back only — the caution written here was removed at the 2026-07-28 post-ratification review), maurice-merleau-ponty.md, jean-paul-sartre.md, merleau-ponty-1964-signs.md, alloa-chouraqui-kaushik-2019-contemporary- philosophy.md (raw_file backfill, metadata-only), motifs.md (§die Sache selbst: 3 sites) · C4 batch dir wiki/.speculative/promotion/runs/ 2026-07-28-batch4/ (59 files: MANIFEST, baseline, interpretation map, 24+24 verdicts, 6 debate transcripts, ADJUDICATIONS, cohort-fingerprint) · promotion-ledger batch-4 triage entry · methodological-queue MQ-9…MQ-17 · ingest-dependency-ledger withdrawal record (no IDL-3) · 6 SPECULATIONS.md annotations · run artifacts (regenerated pregate + retirement queue; consolidation-2026-07-28-run7-verdicts.json with clean-run provenance; -run7-adjudications.md; this digest section) · derived surfaces (claims Summary table 304 rows; index Claims section + counts + 8 page-table cells; overview stats) · baselines (lint baseline 906/304/382/4; claim-status fixtures; pagefind comment; graph SHA) · wiki/log.md · NEXT.md (OPS-015). Test suite: 590 passed / 10 skipped.

Errata — run 7, post-ratification review (2026-07-28)

Six findings on the run-7 branch at head 635e9cb7, delivered out-of-band (no Codex ledger row, no GitHub comment thread). All repaired in one review-fix commit on claude/consolidate-2026-07-28-run7. No claim status moved; the register stays 304 (23 supported / 142 live / 131 candidate / 3 contested / 5 retired).

  1. promiscuite-as-topological-anerkennung-671 — repair item 2 was half- executed. The Claim field carried the restored asymmetric-dominance rejection axis, but the Falsification condition still tested the substitute axis (reciprocal mutual-implication / bidirectionality) while reporting its failure as a rejection-axis dissolution. That is the split-axis defect item 2 named, and ADJUDICATIONS made items 1–3 explicit vote-conditions for R and S — so as registered, two votes reverted to HOLD. Repaired, not relabelled: a source-grounded rejection-axis falsifier was authored for both members (MP/SA at SA-2006 raw 1618/1622 — Piaget's deficiency-verdict retained vs reversed into "sens positif"; Hegel at §§670–671 superseding vs continuing §§186–196), and G's unrun eleven-attestation bidirectionality check is retained in Counterpressure as the separate substitute-axis limitation it always was. Status History and this digest carry the re-open. Entry remains candidate / speculative.
  2. Piaget-formula locator: raw 1620 → raw 1622 (main-thread re-verified against the raw; 1620 is the preceding "réhabilitation ontologique du sensible" paragraph). Fixed on the live surfaces — claims.md (Evidence bullet + E's Counterpressure residual), wiki/concepts/topology-mp.md, wiki/concepts/chiasm.md (an extent-grep catch not in the review list), and the SA-2006 extraction note at three sites (Pass 2c, the Ch. VI locator list, and silent key #3) — closing the item run 6 deferred as outside consolidation write-scope. Batch-4 verdict files, debate transcripts, cohort-fingerprint and interpretation map are left verbatim as the record of what the jury read; the erratum is recorded in ADJUDICATIONS.md §"Main-thread errata" E-1.
  3. Survivor 2's false-friend caution removed from wiki/concepts/sein-zum-tode.md. Survivor 2 converged 4-0 on HOLD-AS-SPECULATIVE; the skill maps HOLD → SPECULATIONS.md annotation and RETIRE → concept-page caution. The panel's reasoning for pairing them is preserved in ADJUDICATIONS.md and not withdrawn, but no protocol amendment was sought or granted, so the write is out of carve-out. Verdict retained, annotation stands. Digest apply-state, veto paths, files-touched list and NEXT.md OPS-015 corrected; wiki/log.md is append-only, so a correction entry was appended rather than the run-7 entry rewritten.
  4. Stale repair provenance in vie-du-lien-prior-to-being-toward-death. Its last Counterpressure bullet still read "Flagged here rather than repaired" for the SA-2021 cite-back; the repair landed at run 6. Rewritten as a completed 2026-07-28 repair to SA-2006 with the evidentiary finding (SA-2021 attests neither phrase) preserved as the reason the attribution moved.
  5. Wording propagation completed. sein-zum-tode.md §Open Questions still said "MP's vie du lien inversion" while the Connections bullet four lines up had been reworded to the sourced subordination-and-SA's-staging formulation; aligned, together with the page's index.md Summary cell, which carried the same stale "inversion" wording (extent-grep catch — the editorial Summary cells are hand-maintained and are a known staleness sink). sache-selbst.md §"Inheritance to the Phenomenological Tradition" renamed to "Later Callers of the Formula (Husserl, Heidegger, Merleau-Ponty)" with a lead-in stating the grouping is by shared rallying-call form, not established transmission; the section's "Husserl picks up the phrase" and the Key-Points label "Phenomenological inheritance" de-settled to match the hedge the jury-blocked claim already forced on the Connections bullet.
  6. Methodological-queue contradictions resolved. MQ-16 both carried a wiki-content finding and said it had been routed elsewhere; the content item (survol absolu under-extraction in the Institution and Passivity ingest, with E's Rule-16 locator abstention and the printed-edition action) now sits in NEXT.md OPS-004, and MQ-16 keeps the process-only routing note plus its generalizable finding (speculate-signoff has no cumulative append target for content residues). The batch heading MQ-9 … MQ-16 and the ADJUDICATIONS artifact-inventory reference both corrected to MQ-17.

Errata addendum — reconciliation against the Codex review at the same head

The six findings above were delivered out-of-band. Reconciling afterwards against chatgpt-codex-connector[bot] showed two reviews on this PR: the first on d4fd9580 (four findings, discharged at 635e9cb7, ledger rows 36–39) and a second on 635e9cb7 itself carrying six inline findings that overlap the out-of-band set. Four of the six match errata 1/2/5/6. The remaining two, plus one divergence, are recorded here:

  1. saintaubert-2006-vers-une-ontologie-indirecte.md Core Argument 14 — headline still said "refuses". The Argument opened by asserting that MP's Ineinander ontology "refuses the Heideggerian être-pour-la-mort" while the hedge appended to it on 2026-07-28 said, four lines below, "i.e. subordination, not refusal" (raw 1507). Headline synced. This is the fifth surface of the same over-wording — after sein-zum-tode, ineinander, empietement and martin-heidegger at Codex comment 3665699366 — a quintuple of the contradictory-doc-surfaces class the codex-review skill records as usually a pair or triple.
  2. Batch-4 cohort-fingerprint.md cumulative totals were wrong. The four-batch line read 3 PROMOTE-CANDIDATE / 10 HOLD; a recount from the four committed ADJUDICATIONS.md files gives 7 / 11 (v0 pilot 2 promote / 2 hold; batch 2 4 / 2; batch 3 0 / 3; batch 4 1 / 4). Verified independently before accepting. The other five categories summed correctly. The undercount understated the cumulative promotion rate, which is the one number that section exists to report.
  3. Divergence, recorded rather than silently rejected. Codex asked the batch-4 ADJUDICATIONS summary to "count all three Phase 6 writes", the commit at 635e9cb7 having applied three. The inconsistency is real; the resolution runs the other way. The third write was survivor 2's caution, and survivor 2 is a HOLD — counting it at the accountability surface would ratify the carve-out breach erratum 3 withdraws. The write was removed and the summary line clarified to say the caution was recommended and not ratified. An artifact-anchored threaded reply is posted on the comment; all six findings are reacted 👍 and ledgered.

Errata addendum 2 — third Codex round (head 993c798)

Seven findings, all accepted (one scoped). Substantively: the rejection-axis falsifier authored at erratum 1 tested valence where the axis names dominance — a positive-but-still-hierarchical reading of MP's promiscuité would have passed the test while the axis failed, so the falsifier now tests Piaget's ranking directly against raw 1618. Two further surfaces of already- repaired items were missed by my own extent greps: motifs.md §Ineinander (the fourth raw-1620 site, now the last live one corpus-wide) and the durable SPECULATIONS survivor annotation (still recording the withdrawn sein-zum-tode caution as written). A Rule-18 raw check confirmed that the erratum-7 headline sync, while correcting the verb, carried an angoisse/Dasein conjunct raw 1507 does not support — narrowed, with that contrast presented separately as Saint Aubert's framing. NEXT.md gained a refreshed State Snapshot (scoped: the drift predates this PR) and a done_when gate so OPS-004 cannot complete while silently dropping the routed survol absolu check. The sache-selbst index Summary cell was the second editorial-cell staleness catch of this PR.

Two process notes worth carrying forward. The editorial Summary/Query cells in index.md are a systematic staleness sink — they are preserved verbatim by normalize_index_tables.py, so no build gate ever touches them, and both catches this PR were interpretive assertions that had been de-settled everywhere except there. Worth a standing extent-grep step whenever a claim is restatused or a page's headline relation is reworded. And the pinned graph SHA-256s earned their keep twice: provenance notes naming other pages as [[wikilinks]] register as edges, and the guard caught both before they landed — fixed at the cause rather than by pinning the new state, per Rule 19's maintainer-adjudicates fork.