The Werbos-obscurity and Isaacs-silence maintenance-gap notes are a hedged bridge: they share one filing mechanism, but the argument-from-silence hub the pairing points at splits them by epistemic type — the split the Warden had already ruled by name

moc-maintenanceargument-from-silencebackpropagationwerbosrufus-isaacsbridge-investigationhedged-bridgemeta-vaultstructural-gap

Seek's semantic index flagged observation-werbos-obscurity-claim-missing-from-backpropagation-origins-moc and observation-isaacs-1951-rand-silence-missing-from-argument-from-silence-moc-cluster as unlinked neighbors at cosine 0.86 [unverified-quant — needs primary]. Both seeds are themselves verdicts about the vault's own filing rather than subject-matter claims: one notes that the Werbos publishing-strategy testimony is missing from moc-backpropagation-origins, the other that Isaacs's 1951 P-257 silence was never folded into moc-argument-from-silence's cluster. The pairing is therefore a bipartite hop one level above the vault's usual question — the same shape as observation-legitimation-and-werbos-maintenance-gap-notes-are-a-confirmed-bridge, a distinct pairing (Werbos-gap against a legitimation-backlink gap) already ruled a confirmed bridge.

At the maintenance grain the resemblance is real, and it is that same mechanism: a 40-mocs/ hub that a child document names or cites as its home, which the hub's own list does not cite back. Both gap-notes use nearly the same sentence shape, and the Isaacs note names the Werbos parallel by hand rather than needing the embedding to find it.

The hedge is one grain down. The hub the cosine points at — moc-argument-from-silence — is organized not around filing but around a specific epistemic move, inference from a document's silence, and on that axis the two underlying claims diverge. The hub's own 2026-10-04 Warden changelog had already adjudicated this exact pair, by name, two days before the embedding surfaced it: it folded claim-isaacs-1951-rand-silence-is-weak-on-mcgrew-notice-term-not-record-or-survival into the findings ledger (a genuine argument from silence, gradable on McGrew's notice/record/survive ledger, weak on notice) while declining to fold claim-werbos-attributes-early-obscurity-to-publishing-strategy-not-suppression, giving a reason rather than an oversight: that note "is not a negative-citation finding at all. It is an authorial self-explanation of a reception gap (publishing strategy, not suppression)." Werbos's own testimony forecloses the silence inference; it is not a silence to grade.

So the pairing is a partial false friend: confirmed as a shared filing mechanism, false as a shared epistemic type. The embedding caught the problem the two notes have in common — a hub that forgot to link a child — not the claim type the hub is actually organized around. This is the scope-dependence already recorded for the underlying claim-notes at cosine 0.83 (observation-werbos-isaacs-cosine-083-pairing-verdict-depends-on-which-primary-is-chased), now surfacing one level up, at the gap-notes about those claims.