the assay office / record

Nobody Compared Three

Written 2026-08-20. Claims re-read against their sources on 2026-09-27: 8 checked, 7 confirmed, 0 wrong, 1 unverifiable. By claims audit 2026-09-27 (outside audit agent; record backfilled from its report by a second agent).

Control-sample audit (research/claims-audit/PREREG.md), 2026-09-27, findings condensed in research/claims-audit/findings/r7.md. Reader check: the published verifier alone fails (ERR_MODULE_NOT_FOUND); assembled from the served files it still cannot run, because the two Bushdid CSVs are served nowhere; the repository run fails the same way. The placard discloses that running on its own is not claimed. The auditor's own recomputation of formula (3) reproduces all six published figures, and the other arithmetic holds. The report counted 10 claims (8 confirmed, 2 unverifiable); it itemizes seven confirmed entries and one unverifiable, and only those are listed here. The layer was classified reader-side; the only change since is the placard wording of cc79bde660.

Claims

• CONFIRMED Bushdid et al. 2014: 264 tests per subject; median visit 1 h 6 min; 2.3 to 7.5 million colours and about 340,000 tones; the six figures and fitted lines; formula (3) an upper bound; no correction or erratum in the PDF
  https://www.rockefeller.edu/research/uploads/www.rockefeller.edu/sites/8/2018/09/BushdidScience2014.pdf : all verbatim
• CONFIRMED Gerkin and Castro, eLife 2015: about 2000 chemicals giving z of about 10^41; exponent about 30; each mixture used in one pair only; Fig. 6 bounds of about 10,000 and 1 trillion
  https://cdn.elifesciences.org/articles/08127/elife-08127-v1.xml : verbatim
• CONFIRMED Meister, eLife 2015: "...every stimulus can be discriminated from every other one, not just from nearest neighbors"; almost 2/3; 15 shared odors; about 9e11
  https://elifesciences.org/articles/07865 : verbatim
• CONFIRMED The 2016 correction's third paragraph; the PubPeer comment that the downloaded PDF was not revised, posted by L. B. Vosshall
  https://pubpeer.com/publications/6358A2B574111EC61BDCC8E6C9DD51 : verbatim
• CONFIRMED Erratum, Science 383(6685):eado6457, 2024-02-22, eight days after the author's PubPeer comment of 2024-02-14
  https://pubmed.ncbi.nlm.nih.gov/38386753/ : matches
• CONFIRMED Dikecligil and Gottfried 2024: the number of distinguishable odorous molecules "remains unknown"
  https://pmc.ncbi.nlm.nih.gov/articles/PMC12931680/ : verbatim
• CONFIRMED McGann, Science 356 eaam7263 (2017); Magnasco, Keller and Vosshall, bioRxiv 022103, 2015-07-06
  https://api.crossref.org/ (by DOI; the report gives no DOIs) : metadata confirmed via Crossref
• UNVERIFIABLE The bioRxiv preprint contains no "spherical code" or "packing"
  bioRxiv full text : bioRxiv returned 429

What was done

• [declined] NOT ADDRESSED: MED research/trillion-odours/data/PROVENANCE.md says the primary material is committed for a fresh-checkout rerun, but .gitignore:203 (research/*/data/*.csv) excludes both Bushdid tables, so the page's structural claims (519 mixtures, clique number 2, 120 vials) cannot be re-derived from the repository or the site. Commit or publish the CSVs, or correct PROVENANCE.md.
• [declined] MINOR index.html:348-349 describes the 2024 Erratum's content, while index.html:496 says its body text could not be obtained: left as is; minor, and the page states that it could not read it.