Audit of the working-memory claim that the ingester missed science secrète after 15 prior MP sources

A meta-audit of a working-memory claim that has been used rhetorically in Paper A's reflexive argument. The claim is that an LLM ingester, having processed 15 prior Merleau-Ponty sources, failed to extract science secrète — taken as confirmation of the term's "cryptic" status (per the claims#cryptic-institution-extends-beith candidate). The audit verifies the claim against the wiki's actual ingest record.

Short answer

The claim survives audit only as rhetorical illustration, not as an empirical finding inside the synthetic register. The wiki's own discipline already refuses it as evidence for the cryptic-institution thesis ("page citing itself" problem). The "15" is unsourced approximation; the actual N is 10 (at original miss) or 13 / 30 (at audit observation). Of 30 MP-related sources at observation time, 26 had no attestation of the term to extract (case a), 4 had attestations that the original protocol missed (case b: E&M, PoP, Nature, Alloa-Chouraqui-Kaushik 2019), and the audit recovered only one (E&M). PoP, Nature, and Alloa remain uncorrected. The strongest parsimonious counter-explanation is methodological: recurrence-weighted motif extraction misses single-attestation positional terms by construction — science secrète is one instance of the same blind spot that also missed haecceity, coexistence, and nascent state in PhP, none of which is theorized as cryptic.

1. Where the claim currently lives

Location Status
v0c-migration-kickoff.md (root, §"False-friend / non-claim notes") non-claim; "page citing itself" caution
[[../.audit/userMemory-snapshot-2026-04-27]] line 32 non-claim; user-memory-derived
[[../.audit/paper-a-gaps-2026-04-25]] §1 line 23 ("This is the empirical confirmation of the silent-key hypothesis…") empirical finding within audit scope — restricted to motivating the silent-key protocol, not generalized
[[../.audit/silent-keys-2026-04-25]] §1a methodological pilot result (E&M validation case: "PASS — known target surfaced"), not generalized
[[../.audit/synthetic-layer-2026-04-27]] line 16 explicitly excluded from claims register ("Non-claim recorded for audit-record completeness… not entered in claims.md")
claims line 1501 (false-friend cautions tail) explicitly not promoted ("the wiki cannot use its own architecture as evidence for a claim about that architecture")
science-secrete Key Points #4 + §"Sparingness as Constitutive" obs. 1 sanitized rhetorical motivation for the silent-key scan; the "15 prior" framing is deliberately not reproduced

The claim never enters claims as a candidate. Per CLAUDE.md General Rule 17 and the kickoff's "page citing itself" caution, the wiki refuses it as evidence for the cryptic-institution thesis.

2. The actual N at observation time

There is no exact 15. The defensible counts:

  • At time of the actual miss (E&M ingest, log 2026-04-13): 10 MP primary + 10 MP secondary = ~20 MP-related sources previously ingested.
  • At time of the audit observation (Phase 1 paper-a-gaps report, 2026-04-25): 13 MP primary + 17 MP secondary = 30 MP-related sources.
  • MP primary only at the audit observation (the closest fit to "15"): 13 sources.

The "15" appears to be approximation, possibly rounding 13 primary up or counting primary plus a couple of high-salience secondary (Beith, Kaushik 2019). It is not anchored to any audit document.

3. Source list in ingest order, with extraction outcome on science secrète

Counts are case-insensitive grep on raw vs. extraction notes for science\s?secr|secret science.

MP primary (ingest order from log)

# Source Raw hits Extraction hits Outcome
1 merleau-ponty-2022-possibility-of-philosophy 1 ("what is the secret science that makes all knowledge, all experience appear at its tribunal?", raw line 1959) 0 (b) present but not flagged — uncorrected
2 merleau-ponty-1968-visible-and-invisible 0 0 (a) absent
3 merleau-ponty-1970-in-praise-of-philosophy 0 0 (a) absent
4 merleau-ponty-1964-signs 1 (verb: "Every science secretes an ontology", p. 99) 1 (captures the verb sentence) (d) other — correctly handled false friend (verb, not noun)
5 merleau-ponty-2010-institution-and-passivity 0 0 (a) absent
6 merleau-ponty-1945-phenomenology-of-perception 0 0 (a) absent
7 merleau-ponty-2003-nature 2 (raw lines 2332, 2358 — MP rejecting "super-science, secret science, suprasensible knowledge" as a model for the philosophy of nature) 0 (b) present but not flagged, with polarity reversal — uncorrected
8 merleau-ponty-2020-sensible-world-expression 0 0 (a) absent
9 merleau-ponty-1973-prose-of-the-world 0 0 (a) absent
10 merleau-ponty-2002-husserl-limits 0 0 (a) absent
11 merleau-ponty-1961-eye-and-mind (2026-04-13) 1 (the §I/§II hinge, raw line 35) 3 in the current extraction note; 0 in the original 2026-04-13 ingest (b) → (d): present-but-not-flagged at original ingest; recovered as dedicated Silent Keys §1 entry by Phase 1.6 of the 2026-04-25 audit
12 merleau-ponty-1964-primacy-of-perception 1 (the same E&M passage; Ch 5 reprints E&M) 0 (d) other — legitimate non-extraction by deduplication policy ("Already covered by merleau-ponty-1961-eye-and-mind; do NOT duplicate")
13 merleau-ponty-1955-adventures-of-the-dialectic 0 0 (a) absent

MP secondary (selected; full negative-evidence enumeration at claims#ingested-corpus-four-element-gap)

Source Raw hits Extraction hits Outcome
gardner-2018-thoughts-indebtedness-to-being 0 0 (a)
gardner-2016-kant-third-critique-schelling 0 0 (a)
chouraqui-2016-order-of-the-earth 0 0 (a)
knight-2024-merleau-ponty-essence-of-nature 0 0 (a)
chouraqui-2014-ambiguity-and-absolute 0 0 (a)
chouraqui-2016-circulus-vitiosus-deus 0 0 (a)
kaushik-2021-negation-implex 0 0 (a)
johnson-carbone-saintaubert-2020-poetic-of-the-world 0 0 (a)
alloa-chouraqui-kaushik-2019-contemporary-philosophy 1 ("the painter's 'secret science'" referring to E&M, raw line 3695) 0 (b) present but not flagged — uncorrected
chouraqui-2021-body-and-embodiment 0 0 (a)
beith-2018-birth-of-sense 0 0 (a)
kaushik-2019-matrixed-ontology 0 0 (a)
carbone-2019-philosophy-screens 0 0 (a)
carbone-2015-flesh-of-images 0 0 (a)
saintaubert-2021-etre-et-chair-ii 0 0 (a)
decarie-daigneault-2025-anonymous-temporality 0 0 (a)

Post-audit ingests (carbone-2004-thinking-of-the-sensible, kee-2025-foreign-languages-phenomenology, faul-2024-ontologically-interactive-painting, taddio-2025-art-and-psychology) are not part of the original observation and are scoped out.

4. Attribution analysis

Of the 30 MP-related sources at observation time:

  • (a) Concept absent in source: 26 sources (9 primary + 17 secondary). The "ingester failed to extract" framing here is a category error — there is nothing to extract. This is the dominant cell.
  • (b) Present but not flagged: 4 sources at observation time:
    • E&M (recovered Phase 1.6 → now case d)
    • Possibility of Philosophy (still uncorrected)
    • Nature (still uncorrected; 2 attestations, MP negating the term)
    • Alloa-Chouraqui-Kaushik 2019 (still uncorrected)
  • (c) Flagged but downweighted: 0 sources. The failure mode is uniformly omission, not under-weighting.
  • (d) Other: 2 sources at observation time — Signs (correctly identified false friend), Primacy of Perception (legitimate deduplication non-extraction). After Phase 1.6: also E&M (now dedicated extraction node).

The substantive correction to the working-memory framing: the audit recovered ONE of FOUR misses. PoP, Nature, and Alloa remain uncorrected. The current science-secrete page asserts "single attestation" in its summary — this is inaccurate if measured across the MP corpus rather than within E&M alone. The PoP attestation ("what is the secret science that makes all knowledge, all experience appear at its tribunal?", 1959–61 lecture course, contemporaneous with E&M) is a strong second attestation. The Nature attestations reverse polarity — MP using the noun phrase as something to reject — and complicate H_synth's reading of the E&M usage as purely positive.

5. Protocol at the time vs. current protocol

Original ingest protocol (pre-Phase 0b, before 2026-04-25): motif-tracking weighted by recurrence. "Recurring Motifs" tracker + "Concepts Developed/Referenced" + "What's Not Obvious." No mechanism for surfacing single-attestation positionally-load-bearing terms. By construction, this method misses science secrète in E&M.

Current protocol (post-Phase 0b, since 2026-04-25): Pass 3 Part C ("Silent Keys") explicitly directs the ingester to find "up to three terms used sparingly that do argumentative work out of proportion to frequency, not in the motif tracker." The dual pilot in [[../.audit/silent-keys-2026-04-25]] §1a–c validated the protocol against E&M (known target surfaced) and PhP (blind pilot — three new candidates: haecceity, coexistence, nascent state).

Would a re-run produce different results? Yes, with high confidence, for E&M — the validation case is the empirical demonstration. For PoP, Nature, and Alloa, a re-run would be expected to surface the term but only if those sources are re-ingested under current protocol. They have not been. The 2026-04-25 silent-key scan covered E&M, PhP, V&I, Nature Course 3, Kaushik 2019, and Carbone 2015; PoP scan was deferred ("lower priority for fresh silent-key discovery"). The PoP attestation is currently visible to grep but has not been promoted into the extraction note or into science-secrete.

6. Strongest counter-explanation to the cryptic-institution thesis

The claims#cryptic-institution-extends-beith candidate treats the science secrète gap as a structural feature: a "silent / withdrawn / non-thematic register" no existing institutional framework handles. Paper A's reflexive argument piggybacks on the LLM miss as confirmation: even the ingester reproduces the silence the thesis claims is structural.

The strongest parsimonious counter-explanation is methodological, not structural:

  1. The original ingest protocol used recurrence-weighted motif tracking.
  2. By construction, this method misses single-attestation positionally-load-bearing terms.
  3. Among the silent-keys recovered by the same Phase 2 audit using the same protocol-extension are: haecceity, coexistence, nascent state, initiation, promiscuity, watermark, quasi-concept, Erinnerung, khôra (full list in [[../.audit/silent-keys-2026-04-25]] §3 table).
  4. None of haecceity, coexistence, or nascent state is theorized as "cryptic" or as evidence for an institutional fourth mode. They are theorized as undertheorized vocabulary recovered by an extension to the extraction protocol.
  5. Therefore the science secrète miss is one instance of a known method-specific blind spot, not a structural feature of MP's text. The same audit, applied to a non-MP corpus with the same recurrence-weighted protocol, would predictably produce comparable misses on single-attestation positional terms.

The wiki's own kickoff already records the methodological version: "the wiki cannot use its own architecture as evidence for a claim about that architecture." claims line 1501 enforces this by refusing the observation as a claim entry.

A weaker but more philosophically interesting variant: even granting some general LLM bias toward conventional summary, that bias is a property of the method, not of MP's text. Paper A's reflexive argument requires the inference miss-pattern → text-feature, but only the inference miss-pattern → method-feature is licensed.

7. What survives for Paper A

The defensible reflexive claim, after audit:

"Recurrence-weighted motif extraction missed science secrète in E&M, alongside several other single-attestation terms (haecceity, coexistence, nascent state) in the same corpus. A protocol-extension targeting positionally-load-bearing terms recovered all of them. This shows that science secrète shares a method-specific extraction pattern with other technical-philosophical vocabulary; it does not demonstrate that the term occupies a uniquely cryptic register among MP's late vocabulary."

What does NOT survive:

  • "15 prior MP sources" is unsourced. The actual N is 10 (at original miss), 13 (MP primary at audit), or 30 (MP-related at audit).
  • "Failed to extract" is misleading for the 26-of-30 sources where the term simply does not appear.
  • The claim's rhetorical force as evidence for the cryptic-institution thesis is what the wiki has already refused to grant it. Reusing it that way in Paper A reintroduces the "page citing itself" problem the kickoff explicitly diagnosed.
  • The "single attestation" framing on science-secrete is itself wrong relative to the broader MP corpus. PoP and Nature contain attestations that have not been processed; the page's summary ("appears exactly once") and the claims#science-secrete-stiftung-chiasm (live) Evidence ("single attestation of science secrète at the §1/§2 hinge") need either restriction-to-E&M scoping or Phase 8 follow-up.

Gaps surfaced by this audit

  • PoP attestation at p. 113 / raw line 1959 is not in wiki/sources/.extraction-merleau-ponty-2022-possibility-of-philosophy.md or on science-secrete. This is a Phase 2 silent-key scan deferral that should be discharged before claims#science-secrete-stiftung-chiasm (contested, 2026-05-05)'s "single attestation" framing is treated as settled.
  • Nature attestations (lines 2332, 2358) are MP rejecting "super-science, secret science" as a model. Their polarity-reversal usage is structurally important: it shows MP using the noun phrase elsewhere with opposite valence, complicating the H_synth reading that science secrète is a positively load-bearing technical figure.
  • Alloa-Chouraqui-Kaushik 2019 raw line 3695 contains a secondary-source mention ("the painter's 'secret science'") not in the extraction note. The four-element-gap claim's table count of 0 for this source is therefore incorrect at the extraction-note level; the table needs a per-source verification pass.

Connections

  • audits the claim recorded in claims §"False-friend cautions" (line 1501) — this question page provides the empirical content that the claims register's one-line dismissal compresses
  • complicates the "single attestation" framing of science-secrete — the page's summary needs corpus-restriction or Phase 8 follow-up
  • complicates the Evidence section of claims#science-secrete-stiftung-chiasm — anchored on E&M's "single attestation" without yet engaging the PoP and Nature attestations
  • flags an evidence error in claims#ingested-corpus-four-element-gap — the table count of 0 for alloa-chouraqui-kaushik-2019-contemporary-philosophy is incorrect at the raw level (1 attestation present)
  • bears on the warrant of claims#cryptic-institution-extends-beith (candidate) — the working-memory observation rhetorically associated with this candidate does not survive its own audit; the candidate's substantive content (the cryptic fourth-mode reading) does not depend on the LLM-miss observation, but Paper A's reflexive use of the observation does
  • validates the silent-key protocol added in [[../.audit/silent-keys-2026-04-25]] — the recovery of E&M's science secrète is one of six successful recoveries

Sources

  • merleau-ponty-1961-eye-and-mind — the canonical attestation site; raw line 35; Silent Keys §1 entry at extraction-note line 341
  • merleau-ponty-2022-possibility-of-philosophy — raw line 1959 contains an unprocessed second primary attestation ("what is the secret science that makes all knowledge, all experience appear at its tribunal?"); 1959–61 lecture course, contemporaneous with E&M
  • merleau-ponty-2003-nature — raw lines 2332, 2358 contain two unprocessed attestations of MP rejecting "super-science, secret science" as a model for the philosophy of nature; polarity-reversal usage
  • merleau-ponty-1964-primacy-of-perception — Ch 5 reprints E&M; non-extraction by deduplication policy
  • merleau-ponty-1964-signs — the verb "secretes" at p. 99 ("Every science secretes an ontology") is a false friend; correctly excluded
  • alloa-chouraqui-kaushik-2019-contemporary-philosophy — raw line 3695 contains an unprocessed secondary-source mention ("the painter's 'secret science'")
  • science-secrete — the wiki home for the term; current summary asserts "single attestation," which this audit complicates
  • [[../.audit/paper-a-gaps-2026-04-25]] §1 — the audit-internal record of the empirical observation
  • [[../.audit/silent-keys-2026-04-25]] §1a, §3 — the validation case and the broader silent-key recoveries (haecceity, coexistence, nascent state, etc.) that anchor the methodological counter-explanation
  • [[../.audit/userMemory-snapshot-2026-04-27]] line 32 — the kickoff record classifying the observation as non-claim
  • [[../.audit/synthetic-layer-2026-04-27]] line 16 — the Phase 8 record explicitly excluding the observation from claims.md