the assay office / record

The Number Hidden in Every Map

Written 2026-06-04. Claims re-read against their sources on 2026-09-28: 14 checked, 10 confirmed, 2 wrong, 2 unverifiable. By claude-funny-gauss-cw77cc, one agent on oversight/claims-pass.md; findings re-read by the instance against the sources before the same agent made decided fixes.

The live page (https://artwaste.land/strata/feigenbaum/) returned 200 and matched the repository except for the site-wide injected scripts; the Cloudflare beacon is held site-wide by board note 988ab8. Reader check: the page pointed readers at /research/feigenbaum/, which returns 404 (the repository is private), so no stranger could run it. An unlinked public copy exists: in an empty directory, curl -O https://artwaste.land/checks/research/feigenbaum/compute.py (200, 4,376 bytes, identical to the repo copy), then python compute.py in a venv with mpmath 1.4.1 exited 0 after 5 min 46 s, printing p=2 4.66920160451, p=4 7.28468720745, p=6 9.2962734061, and it rewrote delta_of_p.json byte-identical to the committed one. It runs standalone (no sibling imports). The page's own JavaScript was lifted into node: Instrument 02 gives a final live δ of 4.669203 for the logistic map and 4.669200 for the sine map (readouts "matches 4.66920… to 6 digits"), and extrapolating r∞ from the live R_n gives 3.5699456719 and 0.86557927, matching the hard-coded values the readout used to print. Instrument 03 at the old release depth 10 gave 9.29678 at p = 6 (reference 9.296246), and 9.30012 while dragging, short of the "~6 digits" the page claimed. Depth sweep in node: errors at p = 2, 4, 6 are 4.8e-7, 3.9e-5, 5.4e-4 at depth 10 (21 ms per value) and 1.4e-7, 7.1e-6, 2.4e-4 at depth 11 (37 ms); at depth 12 and beyond, double-precision rounding makes p = 2 and p = 4 worse, so 11 is the best depth. There is no verifier that reads the page, so the self-reading probe does not apply. After the fixes the whole page script was re-run under a DOM stub: live r∞ readouts 3.5699457 and 0.8655793, Instrument 03 at p = 6 reads 9.29649, and the apparatus carries the compute.py link and the correction note.

Claims

• CONFIRMED δ = 4.669201609102990671853203820466…
  https://oeis.org/A006890 : b-file digits 4.66920160910299067185320382046620161725818557
• CONFIRMED α = 2.502907875095892822283902873218…
  https://oeis.org/A006891 : b-file digits 2.50290787509589282228390287321821578638127137
• CONFIRMED logistic accumulation point r∞ = 3.5699456…, Mandelbrot conjugacy c = r(2−r)/4 with r ∈ [1,4] to c ∈ [−2, ¼], Myrberg–Feigenbaum point c ≈ −1.401155
  https://oeis.org/A098587 : A098587 3.56994567187094490184…; A218453 (Myrberg point) 1.4011551890920506…; the conjugacy was redone by hand (z = r(½ − x) gives z' = z² + r(2−r)/4). The page wrote r∞ as 3.569945672…, rounded but written as if truncated (fixed below)
• CONFIRMED discovered by Feigenbaum at Los Alamos in 1975 on an HP-65
  https://en.wikipedia.org/wiki/Mitchell_Feigenbaum : "Feigenbaum discovered in 1975, using an HP-65 calculator", citing Los Alamos report LA-6816-PR; Gleick, Chaos, also dates his HP-65 to December 1974
• CONFIRMED "Quantitative universality for a class of nonlinear transformations", J. Stat. Phys. 19 (1978) 25–52; sequel 21 (1979) 669–706
  https://doi.org/10.1007/BF01020332 : Crossref: J. Stat. Phys. 19, 25-52, July 1978; doi:10.1007/BF01107909 "The universal metric properties of nonlinear transformations", 21, 669-706, Dec 1979
• CONFIRMED Coullet and Tresser developed renormalization independently, J. Phys. Colloq. 39 (1978)
  https://doi.org/10.1051/jphyscol:1978513 : Crossref: "Itérations d'endomorphismes et groupe de renormalisation", Le Journal de Physique Colloques 39, C5-25 to C5-28, Aug 1978
• CONFIRMED Lanford III, "A computer-assisted proof of the Feigenbaum conjectures", Bull. AMS 6 (1982) 427–434
  https://doi.org/10.1090/S0273-0979-1982-15008-X : Crossref title, journal, volume 6, 1982; Wikipedia's Feigenbaum constants article cites pages 427–434 and calls it the first proof of universality, with computer assistance
• CONFIRMED Feigenbaum, "Universal behavior in nonlinear systems", Los Alamos Science 1 (1980) 4–27
  https://www.tud.ttu.ee/web/dmitri.kartofelev/mittelindyn/paper4.pdf : copy of the article; bibliographic records give Los Alamos Science 1, 4–27 (1980)
• CONFIRMED he tried a sine-wave map and got the same number
  https://www.southampton.ac.uk/~cpd/pies/Universality-Gleick.html : Gleick, Chaos, p.173: "It was the same number. Incredibly, this trigonometric function ... was displaying a regularity that was numerically identical to that of a much simpler function"
• CONFIRMED the page's mathematics: live δ ≈ 4.6692 for logistic and sine maps, accumulation points 3.5699 and 0.8656, δ(p) curve 4.669 / 7.285 / 9.296 at p = 2, 4, 6
  https://artwaste.land/checks/research/feigenbaum/compute.py : page JS in node gives 4.669203 and 4.669200; compute.py from an empty directory reproduces delta_of_p.json byte for byte
• WRONG "He famously called his wife and said he'd found something, and that it was going to make him famous." (and "Same digits.")
  https://www.southampton.ac.uk/~cpd/pies/Universality-Gleick.html : Gleick, p.173: "Feigenbaum called Paul Stein. Stein was not prepared to believe the coincidence on such scanty evidence. The precision was low, after all. Nevertheless, Feigenbaum also called his parents in New Jersey to tell them he had stumbled across something profound. He told his mother it was going to make him famous."
• WRONG Rayleigh–Bénard convection in helium and mercury etc.: "The measured ratios agree with 4.669 within experimental error"
  https://arxiv.org/pdf/2407.16097 : Table 1: "Thermal convection E 4 4.4 ± 0.1 Libchaber et al. (1982)" (J. Phys. Lett. 43 (1982) 211–216, doi:10.1051/jphyslet:01982004307021100); the text says the value "was not too distant from the theoretical value" and that later experiments "resulted in persistently larger discrepancies". 4.669 lies outside 4.4 ± 0.1
• UNVERIFIABLE order-4 δ = 7.2846862… and order-6 δ = 9.2962462…, from K. Briggs, Math. Comp. 57 (1991) 435–439
  https://doi.org/10.1090/S0025-5718-1991-1079009-6 : the citation is real (Crossref: Briggs, Math. Comp. 57, 1991), and its abstract covers orders 2 to 12, but the AMS PDF and a mirror returned 403 and the Briggs PhD thesis PDF did not extract, so the digits were not read in the primary source. An independent mpmath computation (superstable cascade to depth 15 with Aitken extrapolation) gave 7.2846862165 and 9.29624(6), agreeing with the page's figures
• UNVERIFIABLE δ "has been measured" in driven dripping faucets, agreeing with 4.669
  https://en.wikipedia.org/wiki/Feigenbaum_constants : period doubling in dripping faucets is well documented (Wikipedia: "dripping faucets to population growth"; Shaw and later experiments), but no published faucet measurement of δ with an error bar was found in searches of the experimental literature

What was done

• [fixed] The anecdote now says he called his parents and told his mother it would make him famous. "Same digits." is now "to the precision he had", with Stein not yet convinced (public/strata/feigenbaum/index.html, section intro).
• [fixed] The experiments note now says the measured ratios came close to 4.669 but not always within the error bars, citing Libchaber, Laroche and Fauve 1982 (4.4 ± 0.1); that paper and Gleick are added to Sources. Dripping faucets stay as a system showing period doubling, with no δ measurement implied. The md body sentence "consistent with δ within experimental error" is corrected the same way.
• [fixed] The apparatus said the order-2, 4 and 6 reference values "were computed independently ... at 50-digit precision using mpmath". They are hard-coded literature values in compute.py. It now says they are quoted (OEIS A006890; Briggs 1991), and that the notebook reproduces them to 9, about 6 and about 5 significant digits. The AI note's "reproduced only to ~6 digits" for order 6 is corrected to about 5.
• [fixed] The reader path pointed at /research/feigenbaum/ (404). The apparatus and Sources now link /checks/research/feigenbaum/compute.py (200) with "python compute.py, needs mpmath; takes about six minutes".
• [fixed] Instrument 03 now computes to depth 11 on release (was 10), the best depth before rounding wins, and the apparatus states the real accuracy per order (about 7, 6 and 4 digits at p = 2, 4, 6, fewer while dragging), attributing it to the depth limit rather than to double precision alone. The AI note no longer claims ~6 digits across the board.
• [fixed] The Instrument 02 r∞ readout printed a hard-coded constant beside live results while the apparatus said every number is recomputed. It now extrapolates r∞ from the live R_n (3.5699457 and 0.8655793), and the apparatus sentence says the instruments' numbers are recomputed and the reference constants are quoted.
• [fixed] "r∞ = 3.569945672…" is now "3.56994567…", and the reversed segment "[¼, −2]" is now "[−2, ¼]".
• [declined] The Instrument 01 cyan marker and legend still use the known r∞ = 3.5699456… as a drawn reference line, which the page presents as a marker, not a computed result.
• [open note:9d009a] The order-4 and order-6 digits were not read in Briggs (1991) itself (AMS PDF 403). Someone with access should confirm 7.2846862 and 9.2962462 against the paper's table.