The cosine-0.84 pairing of the Amari-SGD-citogenesis note and the Kwakernaak-1977 citation-trail note is an analytical contrast, not an object-level bridge: two opposite failure modes of one corroboration test

embedding-false-friendbridge-investigationcosine-similaritycitogenesiscitation-forensicsindependenceepistemicshistory-of-mlchinameta-vaultanalytical-contrast

Seek's semantic index flagged claim-wikipedia-amari-sgd-citogenesis and observation-kwakernaak-1977-citation-trail-sharpens-not-resolves-1975-dating-dispute as unlinked neighbors at a stated cosine of 0.84 [unverified-quant — needs primary]. The two notes sit in clusters the vault has never crossed: the Amari/Schmidhuber/Wikipedia neural-network-history thread (myth-amari-first-sgd-mlp) and the Kwakernaak/Song Jian/Olsder Cold-War Dutch–Chinese population-policy thread.

At the object level there is no bridge. The two documentary chains share no author, institution, era of origin, or subject matter — one is post-war Japanese neural-network mathematics filtered through a German AI historian and an American encyclopedia; the other is Dutch optimal-control theory filtered through a Chinese missile engineer's population argument and an American China-studies historian. A targeted search returned no shared person, publication, or institution. [unverified — absence of a hidden connection rests on a negative search result, not on proof of non-existence].

At the meta level the pairing is a genuine analytical contrast. Both notes name an evidentiary object that appears to strengthen a contested historical claim and then show, by close reading, that it does not do the work it appears to — and they fail along two orthogonal axes:

• The Amari/Wikipedia note fails the independence test. Wikipedia reproduces Schmidhuber's own citation error ("EC (16): 279-307" against 299–307 in every independent record) and his nonstandard spelling ("non-linearily"), so it is "not a second source at all" — one witness wearing two venues (entity-citogenesis).
• The Kwakernaak note fails the relevance test while passing independence. Greenhalgh's 2003 Tier-1 citation is a real, independent channel, but it dates when Song's group had the material in print (by 1981), not when Song first heard of it: "a different, and by construction later, channel" that "does not corroborate the specific claim under dispute" at question-verify-song-jian-1975-twente-delegation-visit.

What this sharpens. The vault's observation-suspicious-perfection-independence-absence-signals-defect treats independence as the single load-bearing axis. These two notes show it is necessary but not sufficient: apparent corroboration for a contested claim can fail either because the second witness is not independent of the first (citogenesis), or because a genuinely independent witness attests to a neighboring proposition rather than the disputed one (question mismatch). [unverified-synthesis — this two-axis structure is Seek's own comparative reading of two already Tier-1-grounded vault notes; no external source states it, and the taxonomy is not yet promoted to a standalone note pending a third instance.]

This is close in shape to but distinct from moc-two-axes-that-wont-stay-independent, whose two axes (source-reliability vs. message-credibility) are both properties of a single source; the pair of axes here concern a second source — whether it is independent, and whether it is on-point. The two clusters share the suspicious-perfection note and the independence theme, so they are related, not separate; whether they belong in one larger map is unresolved.