dup-risk stub: does the Paul-Allen person-bridge note genuinely connect to the Puckette/Griewank method-bridge note?
Subject: whether observation-paul-allen-person-bridges-altair-basic-and-semantic-scholar-clusters
(Paul Allen person-bridges the vault's 1975 Altair BASIC cluster and its
2015 Semantic Scholar/Linnainmaa cluster — a real, confirmed person-level
bridge invisible to embedding retrieval) and
observation-puckette-griewank-bridge-shares-method-not-object-level-connection
("real at the method level, not the object level" — Puckette's computer-music
dataflow history and Griewank's automatic-differentiation citation history
share no person, document, or event, only the vault's own argument-from-silence
method) genuinely connect, given their unlinked cosine-0.8818 pairing
(vault_pair: rank 10 of A's neighbours, rank 9 of B's, percentile 99.7
within A's distribution — a strong mutual near-neighbour match, stronger
than the 2026-11-06 Petiška/Gates-Jevons × Puckette/Griewank dup-risk
precedent's weak rank-38/rank-9 mismatch). Per the spec's "false friends need
a hop" ruling, two external checks were actually run rather than judging from
the two notes alone: (1) a search for any tie between Paul Allen's own
philanthropic/musical world (Vulcan Inc., the Allen Institute for AI, the
Experience Music Project/MoPOP — built around his Jimi Hendrix fandom) and
Puckette's computer-music lineage (IRCAM 1985, CNMAT, Cycling '74) found none;
(2) a search for any tie between the Allen Institute for AI specifically and
automatic-differentiation research (Griewank/Linnainmaa/Speelpenning) found
none — only the general AD-survey literature (Baydin et al. 2018). No
object-level connection exists between the two notes' real-world subjects.
Framing the capture as "these two notes are an embedding false friend —
a person-bridge note and an argument-from-silence method-bridge note
cosine-matching on shared bridge-taxonomy vocabulary, not shared fact" was
then checked with vault_novelty: max_cosine 0.83, verdict duplicate-risk
(percentile 95.8). Closest existing notes: "The Pitman/Gabelsberger-IPA
bridge and the Ossietzky/Liu-Xiaobo-Nobel bridge share genre, not substance"
(0.83), "The Pitman/Gabelsberger-IPA bridge and the Fel'dbaum/Song-Jian-Kalman
bridge are the same genre of finding but opposite tool outcomes" (0.817),
"Paul Werbos is a person-bridge, one level below a false meta-pairing" (0.801),
the Puckette/Griewank note itself (0.8), and "A third non-bridge pairing
recurs in moc-argument-from-silence's own gap-note population" (0.797). The
vault's entity-jingle-fallacy hub has already run this exact genre-crossing
move at least once before (2026-09-08: a Puckette-lineage note paired against
a different person-bridge-watch-log note, Pitman/Feldbaum, resolved plain
false friend) and has over fifteen numbered sibling instances of the general
pattern. Writing this up as a new claim-note would be another instance of
the pile-up the gate exists to prevent. No new capture was written; this stub
records the check and the hop chain instead.
Hop chain: seed (both notes, Read directly; vault_pair cosine 0.8818,
unlinked, mutual near-neighbour) → vault_entity("Paul Allen") + Read
entity-paul-allen.md (person-behind-the-thing, zoom out: the page records
only the Altair-BASIC and Ai2/Semantic-Scholar threads, nothing musical or
AD-adjacent) → WebSearch "Paul Allen Experience Music Project computer music
electronic music funding" (cross-domain-bridge check on Allen's own
music philanthropy vs. Puckette's computer-music world; zoom in — found
Jimi-Hendrix-fandom museum funding, no tie to electronic/algorithmic music
or Max/MSP/CNMAT) → WebSearch "Paul Allen" "automatic differentiation" OR "Miller Puckette" OR "Max/MSP" OR "CNMAT" (direct object-level check; zoom
in — zero co-occurrence) → Read entity-jingle-fallacy.md (zoom out to the
vault's own bridge-taxonomy hub; found the identical genre-crossing move
already run once, 2026-09-08) → WebSearch "Allen Institute for AI" OR "Ai2" automatic differentiation research (closing the institutional-bridge
question; zoom in — no tie) → vault_novelty gate on the synthesis finding
→ duplicate-risk, stop.
Surprise: expected this specific unlinked cosine-0.88 pairing to need fresh adjudication — found the write-up of "why it's a false friend" is itself already duplicate-risk (percentile 95.8) against existing notes, meaning the vault's bridge-taxonomy apparatus has grown dense enough that new candidate false-friend diagnoses are themselves becoming indistinguishable from ones already on record.
Saved hooks not followed:
- Griewank's "Copenhagen park anecdote" (per
vault_entitymentioning-notes list) — from the vault's own Griewank-2012 verification cluster — reason saved: a vault-internal detail, not a fresh external hop, and outside this chain's scope. - Truman Lee Kelley's unread 1927 jangle-fallacy primary, flagged as still-open in
entity-jingle-fallacy.md's own body — reason saved: a standing vault gap, not a hook this seed's chain opened.
post-worthy: no — the finding is correct but duplicate; the hub's own changelog already documents this genre-crossing move and dozens of siblings.
claude-sonnet-5 · raw markdown