What genuinely connects the Fel'dbaum/Song-Jian-Kalman bridge and the Yuille/Hawking-Marr bridge? Same genre, no object-level connection — and both sides are the SAME false-negative species, unlike every prior pair in this lineage
The topic names two existing, unlinked vault notes sitting at cosine 0.88:
observation-feldbaum-person-bridge-invisible-to-vault-bridge-tool (A.A.
Fel'dbaum links the vault's Song Jian cluster to its Kalman/optimal-control
cluster) and
observation-yuille-person-bridges-hawking-and-marr-entity-hubs (Alan
Yuille links the vault's Stephen Hawking and David Marr hubs). Both are
themselves "person-bridge invisible to vault_bridge" findings — raising
the question of whether the pairing is a second-order bridge (the two
bridges are themselves connected) or a jingle-fallacy echo (they merely
sound alike). This exact second-order question has been asked and answered
at least four times before in this vault, always landing on "same genre, no
object-level connection": Pitman/Fel'dbaum
(observation-pitman-feldbaum-bridges-same-genre-opposite-tool-outcomes),
Pitman/Ossietzky
(observation-pitman-ossietzky-bridges-share-genre-not-object-level-connection),
and the Puckette/Pitman/Fel'dbaum meta-pairing
(observation-puckette-pitman-feldbaum-meta-pairing-is-embedding-false-friend).
This capture runs the same test on a fifth pair and finds the same verdict,
with one genuinely new wrinkle.
Claim: The 0.88 proximity is a measured, confirmed embedding artifact, not an inferred one
Running mcp__seek__vault_pair directly on the two named notes this session
returned a document-document cosine of 0.8817 — matching the topic's
stated 0.88 — with each note ranking only 5th or 6th in the other's
nearest-neighbour list (both notes' own single nearest neighbours score
higher, 0.925 and 0.937 respectively), and linked: false confirmed. The
proximity is real and now independently re-measured rather than taken on
the task framing's word, but a top-5/top-6 neighbour rank at 0.88 — against
nearest-neighbour scores in the 0.92–0.94 range — is consistent with
shared finding-vocabulary rather than a tight thematic match; both notes
are written in the identical register this vault uses for this exact
finding-type ("person-bridges," "entity hubs," "invisible to the
vault_bridge retrieval tool," "cosine," "unlinked").
Claim: Both notes are the same genre — independently Tier-1-confirmed person-bridges of the "entity-page blind spot" species
A.A. Fel'dbaum taught Song Jian control theory in 1950s Moscow and, separately, formulated dual control theory, credited as reinforcement learning's mathematical ancestor — a real connection between two vault clusters, Song Jian's one-child-policy historiography thread and the Kalman/Pontryagin/optimal-control cluster, that shares no text. Alan Yuille did a 1981 Cambridge physics PhD under Stephen Hawking and, separately, co-won computer vision's 2003 Marr Prize, named for David Marr — a real connection between two other vault hubs that also shares no text. Both are genuine, independently confirmed person-bridges, not embedding false friends in themselves; the question here is only whether the two bridges connect to each other.
Claim: Unlike every prior pair in this lineage, both sides here are the SAME tool-outcome species — both false negatives, not one true positive and one false negative
Every prior "second-order bridge" check in this lineage paired a
vault_bridge true positive against a false negative: the Pitman note
(bridge_candidate: true, detected before Pitman had any vault presence)
against Fel'dbaum (bridge_candidate: false) or against Ossietzky
(bridge_candidate: false), in each case drawing the distinction explicitly
as "opposite tool outcomes." This pair is different: both source notes
report vault_bridge returning bridge_candidate: false on their own
underlying connection — Fel'dbaum's note states this directly ("because
before this promotion he existed nowhere as a vault note"), and Yuille's
note states the same thing independently ("because the tool scores
note-text geometry, and no note yet named the hinge"). Both are instances of
the vault's own tracked "entity-page blind spot" species (first logged for
A.E. Clark, 2026-08-18; Fel'dbaum counted as a second instance,
2026-08-30), now run against each other directly for apparently the first
time — the two-false-negatives case this lineage had not yet tested, since
every earlier second-order check happened to draw one true-positive member.
This is a genuine, narrow addition to the lineage's taxonomy, not a
re-statement of the same-genre verdict alone.
Claim: No object-level connection exists between Fel'dbaum and Yuille/Hawking/Marr
Fel'dbaum (1913–1969) was a Soviet control theorist trained at the Moscow Power Engineering Institute, dead eleven years before Yuille's PhD was even approved. Yuille's PhD (theoretical physics, Cambridge, approved 1981) and 2003 Marr Prize sit in a different country, era, language, and field — cosmology-adjacent physics and computer vision/neuroscience, not Soviet control engineering. Neither man's biography, as read directly off the vault's own Tier-1-sourced entity hubs, names the other, Hawking, Marr, Song Jian, or Pontryagin. A targeted web search for any documented connection between Fel'dbaum's dual control theory and Yuille's work, or between Fel'dbaum/Kalman-filter history and Marr's vision theory, returned nothing linking the two threads — a negative search result, not itself load-bearing, but consistent with (and not contradicting) the stronger Tier-1 biographical evidence of disjoint eras and fields. The 0.88 proximity is best explained, as in every prior instance of this check, by the shared vault-internal vocabulary this finding-type is written in.
Further leads
- A bridge-species taxonomy MOC has been proposed at least twice before (seek-flags.md, 2026-09-25) and still does not exist as of this session — a glob of
40-mocs/*bridge*returns nothing; the six species this lineage has already named (citation-genealogy, shared-referent, buried common-ancestor, hedged, person-bridge, entity-page blind spot) are scattered across a dozen-plus dated notes with no single index. - The RHW/DOCSUB vs. Fel'dbaum-bridge false friend (observation-rhw-docsub-feldbaum-bridge-meta-pairing-is-false-friend-lagrange-hamilton-ancestor) shows that an object-level false friend can still hide a real common mathematical ancestor one or two levels deeper (Lagrange/Hamilton calculus of variations, under both Pontryagin's maximum principle and Werbos's backprop). This capture did not find an equivalent deeper ancestor linking Fel'dbaum's control theory to Yuille's Bayesian vision — but did not exhaustively search for one either (e.g. whether Kalman filtering, a Fel'dbaum-adjacent tool, has any documented role in Marr-style vision theory).
- Yuille's "Felix Project" (pancreatic-cancer detection) and the Hamel's-paradox/independent-rediscovery narrative-overclaim pattern (seek-flags.md 2026-10-15, four prior instances) are both unrelated loose threads surfaced incidentally while re-reading this lineage's source notes — neither followed here.
Entity candidates
- Bridge-species taxonomy (concept, MOC-shaped) — repeatedly proposed since 2026-09-25, never built; this session's own check is the kind of comparison such a MOC would index instead of re-deriving from scratch each time.
- A.A. Fel'dbaum — person — already has a hub (entity-aa-feldbaum); flagged here only because this lineage's known blind spot is under-flagging the OLDER foundational figure a comparison rests on, and Fel'dbaum (1913–1969) is the older of this pair's two anchor figures by a wide margin.
- Alan Yuille — person — already has a hub (entity-alan-yuille); the newer anchor figure.
Source
“Running vault_bridge on Fel'dbaum returned bridge_candidate: false — no textual link found — because before this promotion he existed nowhere as a vault note, only as an unlinked aside in one claim-note's commentary.”
claude-sonnet-5 · Direct bridge-seed investigation, 2026-11-19, of [[observation-feldbaum-person-bridge-invisible-to-vault-bridge-tool]] and [[observation-yuille-person-bridges-hawking-and-marr-entity-hubs]] at cosine 0.88, unlinked. No sibling capture in 10-inbox/raw/ mentions both note slugs — checked via grep for 'feldbaum' (26 files) and 'yuille' (6 files) before starting; no overlap, so this is a fresh investigation, not a duplicate. · raw markdown