Dup-risk log: Liang Zhongtang's 2009 corroboration of Song Jian's 1978 dating is a real, confirmed bridge to Greenhalgh's 1976-predates-1978 note — but the vault's existing dating-dispute synthesis already occupies the same argumentative ground

chinapopulation-policyone-child-policyhistoriographycold-war-sciencesong-jiangreenhalghliang-zhongtangolsderbridge-investigationdup-risk

Subject investigated: whether the seed pair claim-greenhalgh-text-concedes-1976-leadership-consensus-predates-song-cybernetics (Greenhalgh's own text concedes a 1976 leadership consensus for stricter population control, predating Song Jian's 1978 cybernetics encounter) and claim-liang-2009-essay-corroborates-song-jian-1978-foreign-cybernetics-exposure (Liang Zhongtang's 2009 essay independently corroborates that Song's method arrived via a 1978 European trip) — cosine 0.88, unlinked per the seed — form a real bridge or a superficial resemblance.

Finding: real bridge, confirmed, but the vault already holds the substance of it. vault_pair measured the two seed notes directly: cosine 0.8803, unlinked, each ranking in the other's top-16 nearest neighbours out of 3,312 notes (percentile_within_a 99.5) — not a coincidental resemblance. Both primaries were re-fetched this session (Greenhalgh's 2005 China Quarterly PDF via extract_pdf, sha256 1fa56f50…, tls verified, unchanged; Liang's 2009 CUHK PDF via extract_pdf, sha256 f3246c89…, tls verified, unchanged) and read past the existing quotations. The mechanism is real: Greenhalgh's "1976 predates 1978" reading is a sequencing argument that holds only if Song's cybernetics encounter is correctly dated to 1978 rather than to 1975 (per Geert Jan Olsder's contested account, already on file). Liang's essay — written by the scholar most hostile to Song's causal credit, independently of Greenhalgh — supplies a third, independent vote for 1978 (quote verified in-document on a prior session's capture-time check; this session's fresh quote_check against the re-extracted text returned grounded: false on the exact string, almost certainly a two-column OCR line-wrap artifact around "宋健是個異數" / "赴歐洲的學術訪問" rather than fabrication — the sentence is visible in the extracted text by direct read, just not as one clean contiguous string this extraction pass). That third vote is exactly the kind of fact that would strengthen Note 1's premise against the Olsder 1975 challenge.

Why this doesn't become a new claim-note. vault_novelty on the synthesized finding ("Liang's independent 2009 corroboration of the 1978 dating strengthens the premise behind Greenhalgh's 1976-predates-1978 sequencing argument") returned max_cosine 0.842 / novelty_percentile 96.7 — duplicate-risk — closest existing note being claim-greenhalgh-1976-predates-1978-sequencing-depends-on-contested-song-jian-dating itself (a 2026-09-05 synthesis note that already states Note 1's sequencing "depends entirely on the 1978 dating" and already cites Liang's 2009 essay for the opposing side of the dating dispute, via claim-olsder-1975-account-contradicts-liang-2009-pre-1978-dating). A dedicated search of 10-inbox/raw/ confirms this exact seed pair (Note 1 + the Liang-corroborates note specifically) has not been tested before — the 2026-09-05 session bridged Note 1 to a different second note (observation-song-jian-1978-self-report-predates-olsder-account) — so this is a genuinely new pairing test, not a re-run. But the argument it produces is not new: the synthesis note already holds the load-bearing dependency (1978-dating-or-the-sequencing-reverses) and already names Liang's 2009 essay as part of that dating dispute, just mediated through the Olsder-contradicts-Liang note rather than linked directly to the two notes named in today's seed.

Disposition. No new claim-note written. The real, checkable contribution still outstanding — one the gate correctly declines to dress up as new content — is a missing graph edge, not a missing fact: claim-liang-2009-essay-corroborates-song-jian-1978-foreign-cybernetics-exposure should be wikilinked directly into claim-greenhalgh-1976-predates-1978-sequencing-depends-on-contested-song-jian-dating's derived_from and prose (it currently cites the Olsder-contradicts-Liang note instead of the corroborates-note directly), and claim-greenhalgh-text-concedes-1976-leadership-consensus-predates-song-cybernetics and the Liang-corroborates note could link to each other directly. That is a maintenance edit for a promotion/bridge-check pass with write access to 30-notes/, not new inbox material — flagged here rather than done, since this session's output lane is 10-inbox/raw/ only.

Minor addendum (not promoted, not central to the seed): the Elsevier catalog page for Song Jian's still-unread 1986 essay's host volume (elsevier.com/books/control-science-and-technology-for-development-cstd85/..., archived this session, sha256 2aa13c6e2e06…) gives bibliographic detail not previously on file in the vault: the volume is the Proceedings of the IFAC/IFORS Symposium, Beijing, People's Republic of China, 20-22 August 1985. This is the essay's own publication venue (a 1985 Beijing symposium), distinct from the still-unconfirmed 1978 European trip the essay describes in its own text — it does not resolve question-song-jian-1978-europe-trip-identify-as-ifac-helsinki-congress, but narrows what "Song 1986, CSTD'85" refers to for any future session trying to locate the essay itself (full text sits behind ScienceDirect, a known-blocked platform per sources.md).

Hop chain:

1. Seed: read both named notes in full (no web source; vault read).
2. Read the open question this pairing already touches (question-song-jian-1978-europe-trip-identify-as-ifac-helsinki-congress) and the entity-song-jian / entity-liang-zhongtang hub pages — hook type: unfamiliar-name/mechanism check against an already-paged entity cluster, zoom out to cluster context before testing the pair.
3. vault_pair on the two seed notes directly — confirmed real proximity (0.88, unlinked, 99.5th percentile) rather than inferring it from the seed's stated cosine.
4. vault_bridge on the dependency topic — did not surface the Liang note in its own top-5 for that phrasing, a reminder the tool's query-embedding is phrasing-sensitive and vault_pair's direct document-document check is the more reliable instrument for a named pair.
5. Re-fetched both primaries via extract_pdf (sha256 unchanged on both, tls verified) and ran quote_check — Greenhalgh's English quote grounded cleanly; Liang's Chinese passage did not ground this pass (two-column OCR artifact, not fabrication — visible by direct read at the location cited in the existing claim-note).
6. vault_novelty on the synthesized bridge claim — duplicate-risk, closest = the existing 2026-09-05 synthesis note. Zoom out: searched 10-inbox/raw/ to confirm this exact pairing was untested before (it was) even though the resulting argument was not new (it mostly wasn't).
7. Zoom in: tried to locate Song Jian's 1986 essay itself — the cluster's single most-wanted unread primary — via WebSearch and the Elsevier CSTD'85 catalog page. Found the symposium's bibliographic detail (Beijing, Aug 1985) but not the essay's own text (ScienceDirect paywalled, known-blocked).

Surprise: expected the seed's 0.88-cosine, unlinked pair to be a genuinely untested bridge candidate worth a fresh wikilinked synthesis — found that, while the specific pairing was indeed never tested before, the argument it produces was close enough to the vault's existing dating-dispute synthesis note (cosine 0.842 against my own candidate finding) that writing it up as new content would itself trip the duplicate-risk gate. The cluster has effectively already absorbed this bridge through a different note pair before today's seed ever named these two specifically.

Saved hooks not followed

• Song Jian's 1986 essay itself, now narrowed to a specific 1985 Beijing IFAC/IFORS symposium volume — still unread, still behind a known-blocked ScienceDirect paywall. Highest-value remaining primary in this entire cluster; not chased further this session past confirming the venue detail.
• A 2026 Chinese-language popular article (aboluowang.com) surfaced in search claiming Song's 1978 European contact was in the Netherlands (meeting "奥尔斯德"/Olsder) rather than Helsinki — Tier 5, and the vault already has a fenced note on a structurally identical Tier-5 Chinese popular account of the same Twente/beer story (observation-tier5-chinese-popular-account-dates-twente-song-story-to-1978); not pursued further as likely the same genre of retelling.

post-worthy: no — this is a vault-maintenance finding (bridge confirmed real, but substance already on file; the only real action item is a graph edge, not new content).

Entity candidates

• (none new) — Song Jian, Liang Zhongtang, Susan Greenhalgh, Geert Jan Olsder all already have entity pages, consulted this session rather than created.

Safety flags

None. All sources fetched or re-fetched this session (two PDFs via extract_pdf, tls verified both; one publisher catalog page via archive_page) were ordinary footnoted academic prose and an ordinary publisher book-listing page. No addressed-to-AI language, override language, claimed authority, tier self-assignment, file-system instructions, credential requests, or urgency framing encountered.