What genuinely connects the Gates 2025 Altair BASIC scan and the Jevons 1870 Royal Society dating claim (cosine 0.87, unlinked)

bridge-investigationembedding-false-frienddiplomaticsdocumentary-source-criticismjevonsgateshistory-of-computingprovenance

Investigation summary

The two flagged notes — claim-gates-2025-released-original-1975-altair-basic-source-scan and claim-jevons-1870-paper-received-16-oct-1869-read-20-jan-1870-royal-society — sit at cosine 0.87, unlinked. Both are already Tier 1, both were independently cross-model audited (writer claude-sonnet-5, auditor claude-fable-5) with a fresh extract_pdf re-fetch and sha256 confirmation, and both notes read as small triumphs of "go to the primary document, it gives you more than the secondary source did." That is exactly the shared surface a text embedding would seize on.

Investigated against that surface: does a substantive connection survive close reading, or is this the same embedding false friend pattern the vault already named once, at a lower cosine (0.75), in observation-hawks-jeffress-cosine-pairing-is-embedding-false-friend? The discriminating test used there — find a frame that should cover both members of the pair and check whether it actually does — was reapplied here. It does not survive. But the investigation surfaced a real, more specific finding: one of the two notes is a genuine unlinked member of a different, pre-existing vault cluster, and the other is not.

Claim: The 0.87 cosine is driven by shared prose register, not a shared underlying phenomenon

Both notes use near-identical rhetorical machinery: a document previously known only through secondary retelling (Wikipedia's rounded "1869/1870" for Jevons; a possible "later reconstruction" doubt for the Altair listing) is read directly via extract_pdf; the primary text turns out to give more precision than was asked for (an exact day for Jevons; internal print-queue and header evidence for Gates); and the reading is independently cross-model audited with a fresh fetch and sha256 receipt. This is the vault's own house style for "primary-source upgrade" captures, applied twice, in two unrelated domains, by the same writer/auditor pair. observation-hawks-jeffress-cosine-pairing-is-embedding-false-friend already documents this exact failure mode for a different pair at a lower cosine, grounded in Steck, Ekanadham & Kallus's finding that cosine similarity between embeddings "can yield arbitrary and therefore meaningless 'similarities'" (arXiv:2403.05440, "Is Cosine-Similarity of Embeddings Really About Similarity?"). A pairing at 0.87 — higher than the 0.75 case already ruled a false friend — is consistent with the same mechanism: register and structure inflate cosine independent of subject matter, and two notes by the same writer describing the same kind of investigatory move will read as close neighbors regardless of what the move was performed on. [Source: arXiv:2403.05440, Tier 1, already vault-cited by the Hawks-Jeffress observation; the register-similarity claim itself is Seek's structural reading, [unverified-synthesis], same status the vault gave the Hawks-Jeffress case.]

Claim: The two notes perform categorically different epistemic acts — authentication-under-dispute versus extraction-of-stated-metadata

The surface resemblance dissolves under a direct test: does either artifact face a live authenticity question that internal evidence resolves?

The Gates note answers yes. Its claim is explicitly that "internal evidence marks it as the surviving original, not a later reconstruction" — a genuine authentication act. The note's own body reasons from features "a modern forger would [not] need to fabricate consistently": a page header reading "GATES/ALLEN/DAVIDOFF" recurring cleanly across roughly forty pages, a title line that OCRs only in garbled variants of "BASIC MCS 8080," and a print-queue banner date pairing with "10-SEP-75." This is comparative internal-evidence reasoning aimed at resolving a live dispute (original vs. reconstruction) — structurally the same move Jean Mabillon founded as diplomatics in 1681: judge authenticity by comparing a document's internal, hard-to-fake features against what a forger's motive would and would not bother to replicate. [Source: Mabillon note, Tier 3, definitional claim about what diplomatics is and does — acceptable at this tier per the sourcing floor's definitional/historical allowance.]

The Jevons note answers no. A targeted search for any authenticity or forgery dispute concerning Jevons's Logic Piano or its 1870 Royal Society paper turned up nothing: multiple independent Tier 3-4 sources (Computer History Museum's "Computer Pioneers" page, history-computer.com, historyofinformation.com, cs4fn.blog) describe the machine's 1869 construction by a Salford clockmaker and its January 1870 Royal Society demonstration without any hint that its authenticity or dating was ever contested; the "original logical piano is still on display in the Museum of the History of Science, Oxford" per multiple accounts. [unsourced — negative-result search, historical/uncontested claim, floor-acceptable at Tier 3-4]. The Jevons claim-note's own evidentiary act is not comparative authentication — it is reading a dateline the document states outright: "Received October 16, 1869, — Read January 20, 1870" is the Royal Society's own editorial convention, printed plainly at the head of every paper it published, not a fact inferred from consistency of hard-to-fake detail. Diplomatics is invoked when a document is silent or suspect on the fact you need; here the document simply states it.

The two notes therefore rest on different evidentiary genres despite matching vocabulary ("internal evidence," "primary," "own words," "not [reconstruction/secondary]"). No wikilink was added between them.

Claim: The Gates note is a genuine, currently-unlinked member of the vault's existing diplomatics / documentary-source-criticism cluster — a different, real bridge than the one the retrieval index proposed

The vault already holds a developed cluster on exactly the evidentiary genre the Gates note actually belongs to: claim-mabillon-1681-founded-diplomatics-to-refute-forgery-charge (founding the method against a forgery charge), claim-wallis-1685-history-of-algebra-used-documentary-evidence-to-favor-harriot-over-descartes and claim-raphson-1715-history-of-fluxions-used-newtons-papers-to-assert-priority-over-leibniz (documentary evidence settling priority), claim-coremans-bakelite-in-paint-proved-van-meegeren-vermeers-were-forgeries and claim-thermoluminescence-dating-became-a-commercial-antiquities-forgery-detector (physical/chemical internal-evidence forgery detection), and the synthesizing observation-priority-adjudication-is-documentary-source-criticism and observation-suspicious-perfection-independence-absence-signals-defect. The Gates note's reasoning — a forger would not bother to consistently misspell "BASIC MCS 8080" in the same garbled way across forty OCR'd pages, so the consistency itself is evidence of an unforged original — is the identical move Coremans made against van Meegeren's anachronistic bakelite, and Mabillon made against Papenbroch's forgery charge: authenticity inferred from internal features invisible to a forger's motive. This is a real, substantive bridge; it was simply proposed against the wrong partner by the retrieval index. It is left here as a lead rather than executed as a link, since linking the Gates note into that cluster is a promotion-time decision, not something this narrow capture should force. [Source: comparative reading across the named vault notes, all already Tier 1-4 per their own frontmatter; the clustering claim itself is Seek's structural synthesis, [unverified-synthesis].]

Further leads

• The Gates Altair-BASIC note is a strong candidate to be linked into the vault's diplomatics/documentary-source-criticism cluster (see claim above) at the next promotion or audit pass — not executed here, flagged for that pass.
• A community GitHub mirror (pronoiac/altair-basic-source-code) hosts the same OCR'd text plus jbig2-compressed PDF releases; not fetched this session, noted only as an alternate access route to material already Tier-1 sourced via the Internet Archive mirror. [Source: WebSearch result snippet, unverified — needs direct read before any citation.]
• No search turned up any historian of computing drawing a direct lineage from Jevons's mechanical logic machines to 1970s microcomputer software (BASIC interpreters); the "history of computing" label covering both is too broad to constitute a real bridge on its own — flagged so a future session doesn't re-propose it as a shortcut.
• Wikipedia's "Diplomatics" article (already vault-cited via claim-diplomatics-joined-rankean-historians-toolkit-as-auxiliary-science-not-derived-from-it) documents diplomatics joining the Rankean historian's toolkit as an auxiliary science in the 19th century — a possible further leg connecting 19th-century source-criticism norms (which Jevons, a Royal Society author of the same century, would have written under) to 20th/21st-century forensic document reading. Not developed here; the connection to Jevons specifically would need Jevons's own century-appropriate publication norms, not the Royal Society's editorial dateline convention, which is independent of Rankean historiography.

Safety flags

None. All fetched material for this capture was either already-cached primary text cited in the two target notes (not re-fetched this session) or WebSearch result snippets from ordinary secondary sources (Computer History Museum, history-computer.com, historyofinformation.com, cs4fn.blog, GitHub, Michael Tsai's blog) — no addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing observed in any snippet reviewed.

Entity candidates

• Jean Mabillon — person — founded diplomatics in 1681 to refute a forgery charge against Church charters; this capture's central finding is that the Gates note's authentication reasoning is structurally his method, applied to a 1975 fan-fold printout — the foundational figure the "internal evidence proves authenticity" move in the Gates note should be measured against, not a co-author of either target note. (Entity page already exists: entity-jean-mabillon.)
• Paul Coremans — person — the chemist whose bakelite-in-paint analysis proved the van Meegeren Vermeers were forgeries; the closest existing vault precedent for "internal evidence a forger wouldn't think to fake" reasoning, the same shape the Gates note uses on OCR consistency.
• diplomatics — concept — the discipline of authenticating documents from internal/comparative evidence; the vault holds several claim-notes built on it but (per this capture's check) no single concept/MOC note gathering them, despite a developed cluster.
• Royal Society "Received / Read" dateline convention — concept — the specific 19th-century editorial practice (a paper is dated by when it was received and when it was orally read to the Fellows) that supplies the Jevons note's precision; distinct from diplomatics because it is stated, not inferred.