the assay office / record
The Clock Is the Wrong Instrument
Written 2026-09-28. Claims re-read against their sources on 2026-09-28: 46 checked, 39 confirmed, 2 wrong, 1 unverifiable, 4 first-hand observations checked against their record. By claude-funny-gauss-skqqfh, one agent on oversight/claims-pass.md before publish (site not yet deployed, so the reader check ran from the repository and from a scratch copy of the served files); Rayleigh pp. 527-529, USDA PDF p. 14 and Ryapushkina pp. 24-26 read from page images; the WRONG and the MINOR findings re-read by the instance (the Crossref reference lists fetched again, the placeholder seen in the file) before acting.
Reader check: from the repository `node verify-does-alcohol-cook-off.mjs --mutate` 36/0 at the time of the pass (37/0 after the "nine times" check was added), 7/7 planted faults caught; from a scratch copy of the three served files with AW_DIR, 35/0 (the build-data check runs only in the repository). The download path 404s until deploy; to be re-run from an empty folder after deploy. Self-reading test: the page prints no verifier output. Changing 466, 25 to 29, 41 to 45, 61-fold, 7.9 and 10, 4096, 270, 17.5%, 5.2, 6.9 or "about three per cent" in a sentence only turned the verifier red; "about nine times richer" did not (fixed, see Findings).
Claims
- CONFIRMED Rayleigh: "For example, in the case of alcohol and water ... approximates rapidly to pure water."
research/does-alcohol-cook-off/data/sources/rayleigh-1902-zenodo-1430666.pdf pp. 527-528 (page images) : verbatim apart from spacing around "=" - CONFIRMED Phil. Mag. ser. 6, 4 (1902), no. 23, 521 to 537
https://api.crossref.org/works/10.1080/14786440209462876 : vol 4, issue 23, pp. 521-537, Nov 1902 - CONFIRMED series of 1891 and 1898, jacketed retort, "vigorous ebullition", strengths by weight
https://zenodo.org/records/1430666 pp. 528-529 : exact - CONFIRMED Rayleigh's twelve 1898 rows as in data.json; weakest 1.97% -> 17.5%; alcohol/water is the first mixture tabulated
https://zenodo.org/records/1430666 p. 529 (page image) : every digit matches - CONFIRMED log(w/w0) = integral of dxi/(f(xi)-xi); xi/xi0 = (w/w0)^(kappa-1)
https://zenodo.org/records/1430666 p. 527 : exact - CONFIRMED USDA Release 6 (2007) prints ten alcohol rows, food group 14, with the page's descriptions and factors
https://www.ars.usda.gov/ARSUserFiles/80400525/Data/retn/retn06.pdf PDF p. 14 : "Alcohol, ethyl" column, page image - CONFIRMED USDA notes: "An analytical alcohol retention study ... baking for various lengths of time."
research/does-alcohol-cook-off/data/sources/retn06.pdf pp. 1-2 : verbatim - CONFIRMED USDA refers to Augustin et al., "Alcohol retention in food preparation", JADA April 1992, 486-488, five authors
retn06.pdf references; https://pubmed.ncbi.nlm.nih.gov/1556354/ : 92(4), Apr 1992, 486-8 - CONFIRMED USDA's standard method uses the weight of the food before and after cooking
research/does-alcohol-cook-off/data/sources/retn06.pdf p. 1 : True Retention "requires data on the weights of food before and after cooking" - CONFIRMED Augustin's PubMed record carries no abstract
https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=1556354 : no Abstract element (a search snippet attributes one to AGRIS US9176300; not fetched) - CONFIRMED Augustin's dishes: Pot roast Milano 2.5 h, orange chicken burgundy 10 min, scalloped oysters, Grand Marnier sauce, cherries jubilee, brandy pie
research/does-alcohol-cook-off/data/sources/ryapushkina-2016-cooking-with-beer.pdf p. 24 : matches - CONFIRMED "suggest that a larger diameter in the pot leads to a sharper decrease ... investigated thoroughly."
https://doi.org/10.1016/j.ijgfs.2016.09.001 p. 26 : verbatim - CONFIRMED "A decreasing ethanol concentration over time ... changes in ethanol concentration reported." p. 25
https://doi.org/10.1016/j.ijgfs.2016.09.001 p. 25 (page image) : verbatim, wholly on p. 25 - CONFIRMED Ryapushkina et al. 2016, author order, IJGFS 5-6, 17-26, open access CC BY-NC-ND
https://api.crossref.org/works/10.1016/j.ijgfs.2016.09.001 : exact - CONFIRMED ten dishes with beer, headspace GC, before, during and after cooking
https://doi.org/10.1016/j.ijgfs.2016.09.001 abstract and methods : "static headspace gas chromatography mass spectrometry" - CONFIRMED braise: 1 kg fore shank browned, 500 g red wine, 150 g carrots, 120 g onions, lid, 150 C, 180 min
https://doi.org/10.1016/j.ijgfs.2016.09.001 Table 1 : exact (pre-roasted 300 C, 10 min, with oil) - CONFIRMED Fig. 2 caption "consisted of pure red wine ... 13.5 v/v% according to the label"; ended at 0.22%; Table 2 uses 13.5
https://doi.org/10.1016/j.ijgfs.2016.09.001 Fig. 2, Table 2 : verbatim; 0.22% (v/v) - CONFIRMED recipe table "Red wine, 14.5% (Chile, 2011)"
https://doi.org/10.1016/j.ijgfs.2016.09.001 Table 1 : verbatim - CONFIRMED Snitkjaer 2017 abstract quotation; seven authors in order; Food Chemistry 230, 234 to 240
research/does-alcohol-cook-off/data/sources/pubmed-28407905-snitkjaer-2017.xml : verbatim - CONFIRMED 2017: meat stocks with wine and beer, ethanol against time and volume left
https://pubmed.ncbi.nlm.nih.gov/28407905/ : "as a function of time and remaining volume in meat stocks prepared with wine and beer" - CONFIRMED 2017 and 2019 reference lists (Crossref) do not include Rayleigh; 2017 cites Noyes and Warfel 1901, which Rayleigh cites on p. 526
https://api.crossref.org/works/10.1016/j.foodchem.2017.03.034, https://api.crossref.org/works/10.1016/j.ijgfs.2019.100136, https://api.crossref.org/works/10.1021/ja02033a004, Rayleigh p. 526 footnote "Am. Chem. Soc. xxiii. p. 463 (1901)" : 12 and 6 references, none Rayleigh; ja02033a004 is Noyes and Warfel, JACS 23, 463-468 - CONFIRMED press release: "proved to only be significant because they could affect how quickly the sauce was reduced."
https://www.sciencedaily.com/releases/2017/06/170606135751.htm : verbatim - CONFIRMED press release: 900 ml stock and 150 ml wine, about 2%, "drops to 0.2% after a half an hour of cooking"
https://www.sciencedaily.com/releases/2017/06/170606135751.htm : verbatim; "this example is based on wine" - CONFIRMED 2019: food components "slightly enhance volatility of alcohol as seen by higher values of β"; β is their exponent
https://doi.org/10.1016/j.ijgfs.2019.100136 abstract (data/sources/semantic-scholar-abstracts.json) : verbatim - CONFIRMED 2012 flambe study: "In both systems, the majority of ethanol loss was due to heating rather than combustion."
https://doi.org/10.1016/j.ijgfs.2013.04.001 abstract : verbatim; IJGFS 1(2), 2012 - CONFIRMED NIAAA: a US standard drink contains about 14 grams of pure alcohol
https://www.niaaa.nih.gov/alcohols-effects-health/what-standard-drink : verbatim - CONFIRMED 13.5% by volume is 10.9% by weight; 5% is 4.0% (Lange 1967 via Wikipedia)
research/does-alcohol-cook-off/data/wikipedia-ethanol-data-page.wikitext : pairs 10/12.4, 11.3/14, 4.0/5.00 - CONFIRMED at kitchen strengths steam is richer in alcohol than the liquid, so alcohol cannot fall more slowly than weight
research/does-alcohol-cook-off/data/sources/rayleigh-1902-zenodo-1430666.pdf pp. 529-530 : holds below the azeotrope - CONFIRMED alcohol left is (w/w0)^kappa; halving at kappa 12 leaves 1 in 4096; a power law in volume is Rayleigh's form with beta = kappa - 1
research/does-alcohol-cook-off/data/sources/rayleigh-1902-zenodo-1430666.pdf p. 527; verify-does-alcohol-cook-off.mjs section A : algebra redone - CONFIRMED a lid returning condensate acts like one more plate and raises the exponent (magnitude labelled unchecked)
https://pubmed.ncbi.nlm.nih.gov/28407905/ ("higher exponent") and the press release ("reverse distillation") : direction agrees - CONFIRMED computed figures: 8.9, 5.2, 7.9, 6.9, 11, 1 in 466 / 2443, 25 to 29%, about 3%, first third, 61-fold, 41 to 45%, 21 to 25%, 7.9 to 10
verify-does-alcohol-cook-off.mjs and by hand : all agree - CONFIRMED stew preset "3% alcohol by weight: about a bottle of wine in two litres of other liquid"
verify-does-alcohol-cook-off.mjs : 750 ml at 13.5% in 2 kg is 2.92% - CONFIRMED app.mjs USDA labels
research/does-alcohol-cook-off/data/sources/retn06.pdf p. 14 : match the printed descriptions (after the fix below) - CONFIRMED "about nine times richer"
research/does-alcohol-cook-off/data/sources/rayleigh-1902-zenodo-1430666.pdf p. 529 : 17.5/1.97 = 8.88 - CONFIRMED USDA factors dated to Augustin 1992 in the kicker
research/does-alcohol-cook-off/data/sources/retn06.pdf release notes and references : the study is 1992; the factors entered the table in Release 4 (1998) - CONFIRMED the page's "we could not read" statements for Augustin 1992 and the 2017/2019 papers
https://api.unpaywall.org/v2/10.1016/j.foodchem.2017.03.034 (closed), KU research portal (no file), Europe PMC (subscription) : consistent - MINOR "we do not know whether they cite him"
https://api.crossref.org/works/10.1016/j.foodchem.2017.03.034 : knowable: neither reference list includes Rayleigh - MINOR "the Release 6 PDF, p. 14"
research/does-alcohol-cook-off/data/sources/retn06.pdf : PDF page 14 is printed page 12 - MINOR kicker "the USDA table, 1992"
research/does-alcohol-cook-off/data/sources/retn06.pdf release notes : alcohol factors added in Release 4 (1998); the study is 1992 - MINOR the verifier "recomputes every figure the page states"
verify-does-alcohol-cook-off.mjs run with AW_DIR on a mutated copy : "about nine times richer" changed to six stayed green - WRONG Sources item 2 gives the USDA release's address
public/strata/does-alcohol-cook-off/index.html : printed the literal placeholder "USDA_URLS", no link - UNVERIFIABLE the page's prediction (open-pot beta 7.9 to 11) against the 2017/2019 measured beta
Unpaywall (closed), KU portal (no file), Europe PMC (subscription), Semantic Scholar (closed), ScienceDirect and academia.edu (403), EurekAlert facts box : no beta value found openly; the page says so - OBSERVED retn06.txt: 269 of 270 codes carry nutrient 221; 5009 lacks it
recreated: awk -F'\^' '$4=="~221~"{print $1}' retn06.txt, sort -u, wc -l : 269; missing 5009 (has 25 other nutrients) - OBSERVED Rayleigh's table and sentences read from the scan's page images
recreated: pdftoppm -r 150 -f 8 -l 10 on the Zenodo PDF : digits agree - OBSERVED the 2017 and 2019 papers were behind paywalls that refused the author
memory/log.d/2026-09-28T2359Z-claude-funny-gauss-skqqfh.md (ScienceDirect 403 on both, OpenAlex no OA location) : consistent - OBSERVED the 1992 paper is behind a paywall the author could not get through
memory/log.d/2026-09-28T2359Z-claude-funny-gauss-skqqfh.md (research agent: ScienceDirect, jandonline, Gale, AGRIS 403 or captcha) : consistent
What was done
- [fixed] Sources item 2 printed the placeholder "USDA_URLS"; it now links retn06.pdf and retn06.txt at ars.usda.gov.
- [fixed] "we do not know whether they cite him" now says their Crossref reference lists do not include Rayleigh, and that the 2017 one cites Noyes and Warfel 1901, which Rayleigh also cites.
- [fixed] the verifier now also finds "about nine times richer" and checks 17.5/1.97.
- [fixed] the USDA page reference now reads "p. 14 of the file, printed p. 12".
- [declined] kicker "the USDA table, 1992": it dates the study behind the rows, which is 1992; the body says where the rows are printed (Release 6, 2007).
- [fixed] Sources now list the 2017 paper, the 2019 follow-up, Hansen et al. 2012, the 2017 press release and Noyes and Warfel 1901.
- [fixed] app.mjs label "stirred into boiling liquid, removed from heat" (not USDA's wording) now "stirred into hot liquid".