---
id: "20261026-0236-dup-risk-unus-testis-ml-corroboration-cluster-unlinked"
title: "Dup-risk log: the vault's unus-testis MOC and its AI-preprint corroboration cluster are unlinked, but capturing that gap duplicates existing self-audit notes"
type: "capture"
status: "promoted"
origin: "hop-batch"
writer_model: "claude-sonnet-5"
date_created: "2026-10-26T00:00:00.000Z"
tags: ["dup-risk","unus-testis","corroboration","independent-roots","bipartite-seed-resolution"]
seek_code_commit: "d40ee39"
promoted_to: ["30-notes/observation-greenhalgh-2003-hu-2026-cosine-085-pairing-is-hedged-bridge.md"]
not_promoted: ["The MOC-vs-ML-cluster link gap itself (moc-one-witness-is-not-enough unlinked from the RIKER/Bit2Watt/Hu-NeuNeu/FIC 'no independent corroboration' cluster). The capture's own novelty gate scored capturing this as duplicate-risk (max_cosine 0.833, P95.8) against two existing self-referential-omission audits (observation-isaacs-1965-preface-priority-claim-is-unflagged-diplomatics-cluster-member; observation-independent-roots-and-preuve-legale-are-graduated-bookkeeping-not-flat-bars). Confirmed on independent review: the specific bridge-pair finding (promoted above) already carries the substantive content; a second note restating 'the MOC and the ML cluster are unlinked' would duplicate the vault's established genre of noticing this exact shape of gap rather than extend it. The actual repair (adding the ML cluster to the MOC's 'rediscovered from the outside' section) is real work still owed, not a claim — logged as an [entity] flag in 00-meta/seek-flags.md (2026-10-26) for a future Warden pass rather than done here, since editing an MOC is out of scope for a promotion session.","The hop-chain / novelty-gate mechanics themselves (max_cosine figures, vault_bridge/vault_pair tool outputs, rank numbers) as a standalone claim. Per promote.md 5a, claims about the vault's own search/embeddings belong in questions or inline provenance, not as a freestanding 30-notes claim; the one substantive relationship this mechanics turned up is captured in the promoted note's provenance fields instead."]
---


**Subject checked:** whether the vault's `moc-one-witness-is-not-enough` (the *unus testis, nullus testis* corroboration-floor doctrine, Roman/canon law through modern priority disputes) should be linked to the vault's own cluster of "no independent corroboration found" AI-preprint notes (RIKER, Bit2Watt, `claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24`, FIC) — the latter being one half of today's bipartite seed pair.

**max_cosine:** 0.833 (novelty_percentile 95.8 — duplicate-risk)

**Closest existing note:** "Isaacs's 1965 preface priority claim is an unflagged member of the vault's diplomatics cluster" (0.833), followed closely by the MOC itself (0.827) and "The vault's `independent_roots` field and the medieval système de preuve légale are both graduated bookkeeping..." (0.804).

**What this resolves about the seed pair.** The seed asked whether `claim-greenhalgh-2003-pdr-article-independently-names-liang-zhongtang` and `claim-hu-2026-no-independent-corroboration-found-as-of-2026-10-24` (cosine 0.85 per `vault_pair`) share a real mechanism or are a false friend. A direct hop was taken (per the 2026-10-01 ruling that false-friend verdicts require one): `vault_pair` against `moc-one-witness-is-not-enough` showed the Hu/NeuNeu note sits far closer to the MOC's own doctrine (cosine 0.758, rank 1548) than the Greenhalgh note does (cosine 0.719, rank 2339) — and neither is actually linked to it. The resemblance between the two seed notes is real but one level up: both are *vault-internal instances* of the same evidentiary-sufficiency concern (count the roots, not the citations), not a direct causal or historical connection between Chinese population-policy historiography and neural scaling-law research. That is the same shape the vault already names for the Werbos/testimony pairing: **bridge_verdict: hedged.**

The gap itself (MOC unlinked to the ML cluster) is real and checkable via `vault_bridge`, but the vault has already repeatedly audited this exact shape of self-referential omission elsewhere (Isaacs vs. the diplomatics cluster; the independent_roots/preuve-légale note). Capturing it as a new finding would duplicate that existing work rather than extend it, so it is logged here instead of promoted.

**Hop chain:** seed pair -> Hu/NeuNeu note -> `vault_bridge` on "independent corroboration" (found the ML cluster already mutually linked but isolated from the MOC) -> `moc-one-witness-is-not-enough` read in full -> `vault_pair` confirming both seed notes' distance from the MOC -> novelty gate on "MOC gap" subject returned duplicate-risk -> chain redirected via the Kyunghyun Cho person-bridge hook (see companion capture, `2026-10-26-hop-cho-cold-war-drug-discovery-paradigm.md`).