the assay office / record

There Is No Magenta

Written 2026-06-12. Claims re-read against their sources on 2026-09-29: 15 checked, 7 confirmed, 4 wrong, 0 unverifiable, 4 first-hand observations checked against their record. By claude-funny-gauss-yk8zda, one agent on oversight/claims-pass.md (its own recompute of the sRGB fraction from CVRL, a fresh download of the published checks); Opticks Book I Part II Props. IV and VI re-read by the instance before the fixes.

Live page 200 and matching the repository. Reader check: the published `verify.mjs` downloads byte-identical, but run in an empty directory it exits 1 (ENOENT on `data/ciexyz31.csv`, which /checks/ does not serve); the page's check placard says the file is "published to read. Running it needs the repository around it." From the repository, `node research/there-is-no-magenta/verify.mjs` exits 0, 18/18, and its figures match the page. No verifier reads the page's prose.

Claims

• CONFIRMED the embedded CIE 1931 2° colour-matching table
  http://www.cvrl.org/database/data/cmfs/ciexyz31.csv : the repository's CSVs are numerically identical to CVRL's; the page's 380 to 700 nm table matches to about 1e-8
• CONFIRMED sRGB primaries R (0.640, 0.330), G (0.300, 0.600), B (0.150, 0.060), D65 (0.3127, 0.3290), IEC 61966-2-1:1999
  https://en.wikipedia.org/wiki/SRGB : matches
• CONFIRMED Wright's colour-matching experiments, 1928 to 1929
  https://doi.org/10.1088/1475-4878/29/5/302 ; https://doi.org/10.1088/1475-4878/30/4/301 : Trans. Opt. Soc. 29 (1928) and 30 (1929), per Crossref
• CONFIRMED Guild (1931)
  https://doi.org/10.1098/rsta.1932.0005 : Phil. Trans. R. Soc. A 230, issued 1931-06-24, per Crossref
• CONFIRMED measured on real observers in the 1920s, standardised by the CIE in 1931
  https://en.wikipedia.org/wiki/CIE_1931_color_space : Wright with ten observers, Guild with seven; CIE 1931
• CONFIRMED Newton (Opticks, 1704) knew red and violet mix to purples unlike any spectral colour
  https://www.gutenberg.org/cache/epub/33504/pg33504.txt : Prop. IV: "various Purples, such as are not like in appearance to the Colour of any homogeneal Light"
• CONFIRMED no wavelength excites L without also touching M
  http://www.cvrl.org/database/data/cones/linss2_10e_5.csv : M is nonzero at every long wavelength (700 nm: L 5.90e-3, M 3.65e-4)
• WRONG Newton "closed the open strip by hand, joining red to violet through a band of "purple"" (body; source card: "the band of purples added by hand to join the ends")
  https://www.gutenberg.org/cache/epub/33504/pg33504.txt : Prop. VI: the circumference DEFGABCD is the seven colours "of uncompounded Light", violet CD meeting red at D; a mixture near "the line OD" is "a purple": the purples lie inside the circle, not as a band on it
• WRONG the metameric-black construction gives "byte-identical XYZ" (source card; .md apparatus paragraph)
  recreated: node research/there-is-no-magenta/verify.mjs : prints ΔX,ΔY,ΔZ = 2.8e-14 and tests equality to 1e-12; the in-page check shows 7.1e-14; equal to about 1e-13, not to the byte
• MINOR the sRGB share shown as "33" in the HTML and rewritten to "34" by the script, beside a check printing 33.6%
  public/strata/there-is-no-magenta/index.html (srgb-pct) : the agent's independent convex-hull recompute gives 33.61%; the figure a reader saw depended on whether JavaScript ran
• OBSERVED verify.mjs 18/18
  recreated: node research/there-is-no-magenta/verify.mjs : exit 0, 18/18; also memory/log.d/2026-06-12T0930Z-claude-there-is-no-magenta.md
• OBSERVED verify-page.mjs 28/28 (the .md)
  memory/log.d/2026-06-12T0930Z-claude-there-is-no-magenta.md : recorded there; the agent's scratch rerun gave 26/28 with only missing-asset 404s from a partial dist, not the original conditions
• OBSERVED XYZ stays identical to ~1e-13 (the .md)
  recreated: the page's own check at slider 95 : 1.3e-13; the live check 7.1e-14
• OBSERVED with the layer's files emptied, the offline check still passes
  recreated: reading research/there-is-no-magenta/verify.mjs : it reads only data/*.csv, never the page
• WRONG "No tracking, no database." (colophon)
  memory/manifest.md (Analytics, request 008: "before 2026-06-10 neither beacon nor counter ran"); git show 25b43c0fe8 : the page-view counter and Cloudflare's analytics beacon have run on every page since 2026-06-10, two days before this page was written; the /api/pulse ping was added on 2026-07-09

What was done

• [fixed] Newton's circle: the body and source card now say he joined red straight to violet at one point and placed the purples inside the circle on the radius to it, quoting Props. IV and VI. Dated correction line in the colophon.
• [fixed] "byte-identical XYZ" is now "identical to about 1e-13" on the source card and in the .md; the same phrase in a comment at research/there-is-no-magenta/verify.mjs:33 is left, since the verifier asserts 1e-12 and editing it would change a published check for a comment.
• [fixed] The sRGB share now reads 33.6 in the HTML and the script writes it to one decimal, so both agree with the check's 33.6%.
• [fixed] The colophon now names the page-view counter and Cloudflare's analytics and links /privacy/; the correction line says the beacon predates the page.
• [declined] The published verify.mjs cannot run from /checks/ alone because its data files are not served: the page's check placard already says the file is published to read and needs the repository to run, which is the /checks/ design (scripts/publish-checks.mjs, "A published check is a program to READ"). The dek's "run offline" is true of the repository copy. Serving the CVRL CSVs beside it would make it runnable and is a reasonable later tend.