--- slug: the-middle-layer-breaks-the-bound passed: 2026-10-18 by: the assay line (research/assay-line/queue-1018/), Codex fixer directed by claude-assaying-boyle-1014; source record by Codex, checked by Codex checked: 102 confirmed: 92 wrong: 6 unverifiable: 0 observed: 4 stale: 0 --- This was the decide-and-fix pass for this layer in tier A of queue-1018. All six triage items were decided and the completed JSON was validated before the layer was edited. The fixer re-read the authored control handlers and the two primary papers for the six MINOR findings, then applied wording fixes. The 92 CONFIRMED and four OBSERVED claims below carry forward the checker's re-reads in research/assay-line/queue-1018/the-middle-layer-breaks-the-bound/N2-codex.md; their notes describe that checker's evidence and observations. That record uses commit 7b67882c78; the layer's HTML, markdown, research notes, and verifier are identical there and at the requested read commit 96caa00974. MINOR claim lines retain the wording assessed before the fixes and count toward wrong; there are zero WRONG or STALE verdicts. ## Claims - CONFIRMED | Two live instruments, one hard boundary | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [two-instruments] The pinned HTML has two instrument sections and exports "buildMiddleLayer" and "powerBench". Both inline scripts executed in inspect-page.mjs. - CONFIRMED | Build the middle layer of a hypercube and watch distinct contributions collide with exact multiplicity. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [dek-finite] At N=8, the independent verifier prints "Q=70, Q'=56, routes=280, gain=5"; the runtime reports "5 at every output". - CONFIRMED | Then operate the later power-loss exponent to see what Cairo's 2025 preprints prove, what the browser merely computes, and why fixed spheres and paraboloids remain open. | https://arxiv.org/html/2512.08064v1 | [dek-preprints-open] The primary text says "Theorem 1.3 does not say anything" about "the paraboloid or the sphere" and calls their local conjecture a "very challenging question". The power calculator uses its exponent. - CONFIRMED | This page builds a layer of a many-dimensional cube and shows how many paths land on the same point. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [plain-cube] The source increments "multiplicities.set(output, (multiplicities.get(output) || 0) + 1);" for each absent coordinate; its default output has five parents. - CONFIRMED | You can change its size, then compare a slow logarithmic gain with the stronger power gain proved for carefully perturbed curved surfaces. | https://arxiv.org/html/2512.08064v1 | [plain-perturbations] Theorem 1.3 states "there exists an arbitrarily small" perturbation and gives R^((n-1)/(n-1+k)-epsilon). This is stronger asymptotically than log R when that exponent is positive. - CONFIRMED | Its first instrument enumerates every vertex and every route in the middle-layer construction. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [md-first-instrument] The loop "for (mask = 0; mask < limit; mask += 1)" enumerates every mask; the nested absent-bit loop counts the routes. Independent string enumeration agrees for every N=4 through 18. - CONFIRMED | Its second instrument evaluates the separate power-loss exponent from the Cairo and Zhang preprint. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [md-second-instrument] The source says "var beta = (n - 1) / (n - 1 + smoothness);" and the independent power sweep passes for all selectable n and k. - CONFIRMED | The visible check keeps the finite computations, imported theorems, conventions, preprint status, and still-open fixed-surface questions in separate lanes. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [md-visible-check] The visible headings include "Finite count, recomputed", "Power bench, recomputed", "Conventions, free choices, approximations, and uncertainty", "Source literals and review status", and "Still open". - CONFIRMED | Both make high-dimensional geometry operable instead of asking the reader to picture it. One watches volume and angles concentrate as dimension rises; this one enumerates the exact middle slice of a hypercube and follows its collisions into the next slice. | git show 7b67882c78:src/content/strata/all-edge-no-middle.md | [related-1] The related page describes "Three live samplers" and says "watch a ball's volume flee to its skin; watch every direction become a right angle." This page supplies exact Hamming-layer counts. - CONFIRMED | Both trust a large combinatorial effect only after exact enumeration. The important boundary differs: the cube-dissection page carries its count directly, while this page says where its finite count ends and the cited analytic theorem begins. | git show 7b67882c78:src/content/strata/the-count-that-ran-off-the-page.md | [related-2] The related page says "two independent algorithms (the DP and a from-scratch 3-D enumerator) agree on every small box". This page explicitly separates finite counts from the imported analytic estimates. - CONFIRMED | Both turn verification into part of the object. The zero-knowledge page separates a transcript from the hidden witness; this page separates a browser-checked combinatorial skeleton from the Fourier theorem that uses it. | git show 7b67882c78:src/content/strata/the-proof-that-tells-you-nothing.md | [related-3] The related page describes "a prover sealed in a worker so the page cannot peek" and a machine making transcripts without a colouring. This page identifies the finite skeleton and the imported Fourier theorem separately. - CONFIRMED | A line-control estimate expected the total Fourier energy to stay within one scale-independent bound. | https://arxiv.org/html/2502.06137v2 | [global-bound] Conjecture 1.1, equation (1.2), bounds the full weighted extension energy by the squared L2 norm times the X-ray-transform supremum, with no R factor. - CONFIRMED | Hannah Cairo's February 2025 preprint constructs a logarithmic loss. | https://arxiv.org/abs/2502.06137 | [february-log-loss] The arXiv history says "10 Feb 2025"; the abstract says "construct a $\log R$-loss counterexample". - CONFIRMED | The finite mechanism at its centre can be operated exactly: take every bit string in the middle layer, switch one absent bit on, and count how many routes reach each point in the next layer. | https://arxiv.org/html/2502.06137v2 | [finite-mechanism-source] Section 3.2, equations (3.7), (3.9), and (3.10), gives the middle layer and the next layer. In particular, it selects sums satisfying "c_i=0" and sets the next Hamming weight to floor(N/2)+1. N1's excerpt stops before this latter passage, so the checker read the section beyond the excerpt. - CONFIRMED | The browser enumerates every reachable bit string and every one-bit contribution. Select any output to expose all of its parents. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [enumerates-routes] The executed UI gives "Selected output 10101011", "28 of 56", and the five parents listed in N1. The bit-clearing loop generates every parent of the selected output. - CONFIRMED | Cairo's Lemma 3.5 supplies a specially embedded configuration in which no plane meets more than this many radius R-1 balls. | https://arxiv.org/html/2502.06137v2 | [incidence-lemma] Lemma 3.5(i) says "No plane passes through more than" 2^(d-1) radius-R^-1 balls; its construction has N comparable to log R and sufficiently large R. The page expressly imports this configuration. - CONFIRMED | The number is computed from the chosen d. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [incidence-calculation] The code is "return Math.pow(2, dimension - 1);". At d=3, the runtime reports "4"; the verifier passes d=2 through 10. - CONFIRMED | This toy does not construct or prove that embedding. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [embedding-boundary] The entire math-core defines mask enumeration, powerBench, and incidenceCap. It has no embedded surface or coordinates xi_i. Positive control: the same source inspection finds "function buildMiddleLayer(n)" and all three exported functions. - CONFIRMED | Exact here: the middle-layer counts, every route, every multiplicity, and the collision ratio. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [exact-counts] Independently recomputed at N=8: C(8,4)=70, C(8,5)=56, routes=4*70=280, squared multiplicities=56*25=1400, and 1400/280=5. The verifier confirms all selectable N. - CONFIRMED | Imported from the paper: the analytic estimates, the choice N comparable to log R, and the plane-incidence configuration. | https://arxiv.org/html/2502.06137v2 | [imported-analytic-bridge] Theorem 1.2 supplies the weighted Fourier estimate; Lemma 3.5 supplies "N~log R" and the plane-incidence configuration. Neither is inferred from the finite enumeration. - CONFIRMED | For every fixed positive epsilon, log R eventually grows more slowly than Repsilon. | https://dlmf.nist.gov/4.4 | [log-slower-power] DLMF equation 4.4.13 states lim x^(-a) ln x = 0 for Re(a)>0. Taking real a=epsilon proves the comparison. - CONFIRMED | Cairo's first result therefore defeats the global constant bound, but does not by itself defeat the local Repsilon version. | https://arxiv.org/html/2502.06137v2 | [global-versus-local] Section 1.7 says "the following local version may be a plausible reformulation" and Conjecture 1.5 allows R^epsilon. The logarithmic lower factor does not contradict that allowed loss. - CONFIRMED | The December 2025 Cairo and Zhang preprint supplies a separate power-loss theorem on dense families of Ck perturbations. | https://arxiv.org/abs/2512.08064 | [december-dense-power] The arXiv record says "Submitted on 8 Dec 2025" and "this family is dense in the $C^k$ topology"; Theorem 1.3 gives the power loss. - CONFIRMED | The selected epsilon is at least beta, so beta minus epsilon is not a positive power. The formula is still evaluated, but this setting does not display a growing lower factor. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [nonpositive-power-warning] The code is "byId('endpoint-warning').hidden = bench.selectedExponent > 0;". At n=2, k=10, epsilon=0.5, runtime exponent is "-0.4090909091" and warningHidden is false. - MINOR | These lines are regenerated whenever a control moves. They expose what this page derives and what it only receives from a theorem. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [X1; check-refresh; error: check-refresh-overclaim] Re-read the authored index.html:904-910 and buildLayer: the N input handler only sets the label to 'selected, press build'; buildButton.addEventListener('click', buildLayer) rebuilds the finite counts. The four power controls call updatePower on input. This is a MINOR overclaim about the page's own checking. Tier A: cheaply say that finite counts refresh on Build and power checks refresh with their controls. No dated correction line. - CONFIRMED | Cairo's v2 preprint proves that for every compact C2 hypersurface in Rd not contained in a hyperplane, and for sufficiently large R, there are a function and a nonnegative weight whose weighted extension energy inside the radius-R ball is at least a hidden constant times log R times the function's squared norm and the largest line integral of the weight. | https://arxiv.org/html/2502.06137v2 | [theorem-full] Section 1 begins "a compact" C2 hypersurface. Theorem 1.2 applies when it is "not a plane" and equation (1.3) gives exactly the logarithm, squared norm, and largest line integral described. Lemma 3.5 states the sufficiently-large-R condition. - CONFIRMED | That logarithmic factor contradicts the scale-independent global estimate. | https://arxiv.org/html/2502.06137v2 | [contradicts-global] The paper introduces Theorem 1.2 as a "counterexample ... with log R-loss". Since log R is unbounded, a scale-independent constant cannot bound that ratio. - CONFIRMED | The page fixes the middle weight at floor(N/2), lists bits from coordinate N - 1 down to coordinate zero, orders masks numerically, and counts only switches from zero to one. The sliders' ranges and defaults are editorial choices. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [conventions] The code uses "Math.floor(n / 2)", scans masks in increasing numerical order, prints with "mask.toString(2).padStart(n, '0')", and switches only an absent bit to one. Control attributes and runtime agree. - CONFIRMED | All finite counts use exact integer arithmetic within JavaScript's safe integer range. N stops at eighteen for responsive enumeration, not because the identity stops there. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [exact-safe-counts] The maximum enumerated sum of squares at N=18 is 4,375,800, far below 2^53-1. Intermediate binomial products also remain safe integers. Independent recursion checks N=4 through 18; the general incidence identity follows by deleting each of k+1 ones from an output. - CONFIRMED | The displayed logarithm is the natural logarithm. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [natural-log] The shipped formula is "logR: Math.log(R),"; the default runtime gives "41.446532" for R=10^18. - CONFIRMED | Another fixed base changes only a constant in an asymptotic comparison. | https://dlmf.nist.gov/4.2 | [log-base] DLMF equation 4.2.8 gives "log_a z = ln z / ln a", so a fixed logarithm base changes the scale by a constant. - CONFIRMED | The R slider is logarithmic and sets R to ten raised to the displayed slider exponent. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [scale-slider] The shipped update uses "var R = Math.pow(10, scaleExponent);" and prints "10^18.00 = 1.0000e+18" by default. - CONFIRMED | Power and logarithm displays use browser floating-point arithmetic. They illustrate the formulas at a chosen finite scale. They are not evidence for an asymptotic theorem. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [floating-point] The page uses Number, Math.pow, and Math.log. The runtime reproduces every recorded rendered_default here, including "15,848,931.924611", "63.095734", and "41.446532". The asymptotic disclaimer is appropriate. - CONFIRMED | Cairo's analytic inequalities contain hidden constants. | https://arxiv.org/html/2502.06137v2 | [hidden-constants] Theorem 1.2 uses the comparison sign gtrsim in equation (1.3), rather than specifying a numerical leading constant. - CONFIRMED | N comparable to log R is not equality, and the results apply for sufficiently large R. | https://arxiv.org/html/2502.06137v2 | [comparable-and-threshold] Lemma 3.5 begins "For any R>1 sufficiently large" and specifies "N~log R". These are comparability and a threshold, not an equality and an explicit threshold. - CONFIRMED | The finite cube verifies neither threshold nor hidden constant. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [threshold-boundary] Inspection of the math-core and differential verifier finds finite enumeration, formula evaluation, and literal source checks. Positive control: the verifier contains "referenceLayer" and "referenceBeta"; neither determines the analytical threshold or hidden constant. - CONFIRMED | The plane-incidence cap is computed from 2d-1, but the existence of the special embedding is imported from Lemma 3.5. An arbitrary placement of points need not obey it. | https://arxiv.org/html/2502.06137v2 | [incidence-formula-scope] Lemma 3.5 asserts the existence of a configuration with cap 2^(d-1). The source code evaluates that formula. Putting more balls in a single hyperplane gives a simple counterexample to any claim that arbitrary placements have the same cap. - CONFIRMED | Cairo and Zhang's power theorem concerns dense families obtainable by arbitrarily small Ck perturbation. Dense does not mean every surface. Many strictly convex examples does not mean every strictly convex surface. | https://arxiv.org/html/2512.08064v1 | [dense-scope] Theorem 1.3 says "there exists an arbitrarily small" Ck perturbation. The following prose says "including many strictly convex ones" and leaves fixed familiar surfaces open. The page preserves the quantifiers. - MINOR | The exponent endpoint is not claimed. When k equals two, beta equals the general C2 upper exponent gamma, so the two bounds match only up to the endpoint. | https://arxiv.org/html/2512.08064v1 ; https://arxiv.org/html/2302.11877 | [X2; endpoint-c2; error: c2-upper-bound-scope] Re-read https://arxiv.org/html/2302.11877, Introduction opening, Theorem 1.2 equation (5), and its following curvature caveat. The standing surface is a 'strictly convex, compact patch' with 'non-vanishing Gaussian curvature'; the caveat says 'requires non-vanishing curvature'. Also read https://arxiv.org/html/2512.08064v1 immediately after Theorem 1.3: 'general strictly convex' and 'up to the endpoint'. The exponent equality and endpoint caution are correct. Keep MINOR for the omitted scope in this comparison wording; add the surface hypotheses at index.html:530. This is a cheap Tier A precision fix, with no dated correction line. - CONFIRMED | Preprint Hannah Cairo, A Counterexample to the Mizohata-Takeuchi Conjecture, arXiv:2502.06137v2. Version 1 was submitted 10 February 2025 at 03:55:11 UTC. Version 2 was submitted 12 March 2025 at 13:44:52 UTC. | https://arxiv.org/abs/2502.06137 | [cairo-reference] The arXiv title and author match. Its submission history is "[v1] Mon, 10 Feb 2025 03:55:11 UTC" and "[v2] Wed, 12 Mar 2025 13:44:52 UTC". - CONFIRMED | Preprint Hannah Cairo and Ruixiang Zhang, Power loss for the Mizohata-Takeuchi conjecture on Ck convex hypersurfaces, arXiv:2512.08064v1, submitted 8 December 2025 at 21:56:05 UTC. | https://arxiv.org/abs/2512.08064 | [zhang-reference] The arXiv title and both authors match; the history is "[v1] Mon, 8 Dec 2025 21:56:05 UTC". - CONFIRMED | Preprint Inbo Gottlieb Fenves, Cusp Excursions, Lattice Points on Manifolds, and the Mizohata-Takeuchi Conjecture, arXiv:2606.27020v1, submitted 25 June 2026 at 13:32:43 UTC. | https://arxiv.org/abs/2606.27020 | [fenves-reference] The arXiv title and author match; the history is "[v1] Thu, 25 Jun 2026 13:32:43 UTC". - CONFIRMED | Used for Theorem 1.2, the middle-layer construction in Section 3.2, and the incidence cap in Lemma 3.5. | https://arxiv.org/html/2502.06137v2 | [cairo-source-sections] The primary document has "Theorem 1.2 (Counterexample)", "3.2. A white lie construction", and "Lemma 3.5" at the cited locations, with the stated content. - CONFIRMED | Used for Theorem 1.3, the exponent, the perturbation scope, the C2 comparison, and the fixed-surface open problem. | https://arxiv.org/html/2512.08064v1 | [zhang-source-scope] Theorem 1.3 and the following two paragraphs contain the exponent, arbitrarily small perturbations, strictly convex C2 comparison, and explicit sphere/paraboloid open problem. This claim describes use of those passages, without itself asserting an unrestricted upper bound. - CONFIRMED | Independent follow-up used for the current landscape: a new proof with explicit error terms and a generic Ck result. | https://arxiv.org/abs/2606.27020 | [fenves-new-proof-generic] The abstract says "a new proof of power loss" with "explicit error terms" and "power loss is generic in $C^k$". Theorem 1.2 formalizes a dense G_delta set. - CONFIRMED | UC Berkeley Analysis and PDE Seminar, Hannah Cairo (UC Berkeley), posted 7 February 2025 for a 10 February 2025 talk. | https://wp.math.berkeley.edu/apde/2025/02/07/hannah-cairo-uc-berkeley/ | [berkeley-reference] The institutional announcement says "Monday, 2/10" and "Hannah Cairo (UC Berkeley)"; its displayed date is "February 7, 2025". - CONFIRMED | Independent institutional confirmation of the talk title, date, and logarithmic counterexample description. It is not peer review. | https://wp.math.berkeley.edu/apde/2025/02/07/hannah-cairo-uc-berkeley/ | [berkeley-scope] The announcement title matches and its abstract says "a $\log R$-loss counterexample". Its presentation as a seminar announcement supports the page's distinction from peer review. - CONFIRMED | University of Maryland CMNS, This Teen Solved a 40-Year-Old Math Mystery. Now She's Seeking a Ph.D. at UMD., 8 September 2025. | https://cmns.umd.edu/news-events/news/teen-solved-40-year-old-math-mystery-now-shes-seeking-phd-umd | [umd-reference] The institutional article title matches, with typographic apostrophes; the displayed publication date is "September 8, 2025". - CONFIRMED | Biographical context only: the profile says Cairo was seventeen during the work and eighteen when beginning doctoral study. | https://cmns.umd.edu/news-events/news/teen-solved-40-year-old-math-mystery-now-shes-seeking-phd-umd | [cairo-age] The article begins "While many 17-year-olds last spring" and describes Cairo's work, says she "starts her doctoral studies ... this fall", and calls her "now 18". - CONFIRMED | No exact birth date is inferred here. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [no-inferred-birthdate] The page uses the two ages and the article date, with no exact birth date. Positive control: the same source panel contains "seventeen" and "eighteen" and the separately verified publication date. - CONFIRMED | the three mathematical works above are cited as arXiv preprints, not peer-reviewed articles. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [cited-as-preprints] Each of the three references carries the literal badge "Preprint". This is a claim about how the page cites them, not a new assertion that no later journal publication exists. - OBSERVED | The source search was repeated on 29 July 2026 and did not locate a journal version for any of them. | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/README.md | [historical-source-search] The dated research record repeats "No journal version was located in the 29 July 2026 title and DOI search" for all three papers. This supports the historical search event under step 7; I do not treat it as a present-day negative publication search. - CONFIRMED | The local Mizohata-Takeuchi conjecture for particular familiar hypersurfaces such as the sphere and paraboloid remains unresolved in the cited literature. | https://arxiv.org/html/2512.08064v1 | [fixed-surfaces-open] Cairo and Zhang explicitly call proving or disproving the local conjecture on the sphere and paraboloid a "very challenging question". Fenves's remarks explicitly exclude fixed surfaces as well. - CONFIRMED | Cairo and Zhang perturb surfaces, and Fenves gives a new proof and a genericity theorem, but neither method settles those fixed examples. | https://arxiv.org/html/2606.27020v1 | [neither-settles-fixed] Fenves says "we can say nothing about the local Mizohata-Takeuchi conjecture for a fixed surface" and gives the sphere as its example. Cairo and Zhang explicitly mention both the sphere and paraboloid. - CONFIRMED | The endpoint power exponents are also not claimed. | https://arxiv.org/html/2606.27020v1 | [endpoint-unclaimed] Fenves Theorem 1.1 retains exp(O(sqrt(log R log log R))) error; Theorem 1.2 uses a positive delta loss. Cairo and Zhang Theorem 1.3 uses a positive epsilon loss. These do not assert the exact endpoint. - CONFIRMED | The browser proves the finite identities it enumerates. | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/verify-the-middle-layer-breaks-the-bound.mjs | [browser-proves-finite] Independent recursive string enumeration agrees with the shipped math-core for every reachable N=4 through 18, including routes, multiplicity extrema, square sums, and ratios. This finite claim does not assert the analytic theorem. - CONFIRMED | The analytic bridge from those identities to weighted Fourier extension is the work of the cited preprints, not a result of this visualization. | https://arxiv.org/html/2502.06137v2 | [analytic-bridge-authorship] Theorem 1.2 and Section 3.2 establish the analytic connection. The page supplies mask arithmetic and explicitly labels the analytic estimates as imported. - CONFIRMED | Offline differential verifier: node research/the-middle-layer-breaks-the-bound/verify-the-middle-layer-breaks-the-bound.mjs | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/verify-the-middle-layer-breaks-the-bound.mjs | [offline-command] Executed this exact relative command in the empty reader directory after the two published curl downloads: exit 0; 44 PASS assertions, zero FAIL, plus "PASS all independent checks". - CONFIRMED | This layer's check lifts this page's own source out of the HTML and runs it. | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/verify-the-middle-layer-breaks-the-bound.mjs | [lifts-page] The verifier calls "extractMathCore(html)" and then "loadShippedMath(shippedSource)", which executes the extracted script with vm.runInContext. Its opening assertion passes. - OBSERVED | Measured, not assumed: with every file this layer ships emptied, it goes red. | git show 7b67882c78:research/verifier-independence/depends-1.json | [emptied-red] The committed row for this slug records "exitCode": 1, "depends": true, and "timedOut": false when shipped files were emptied. That is direct evidence for the reported past test. - CONFIRMED | It is reading what you are reading. | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/verify-the-middle-layer-breaks-the-bound.mjs | [reads-reader-source] The verifier reads "../../public/strata/the-middle-layer-breaks-the-bound/index.html". The reader download has the same math-core and body as the pin; the live-only additions are analytics scripts. - OBSERVED | Measured too: a defect placed in this page's running code turned it red. | git show 7b67882c78:research/verifier-independence/mutation-coupled-2.json | [mutation-red] The committed mutation row changes the beta numerator from "1" to "2", with "mutatedExit": 1 and "detected": true. Other rows test other mutations, but they do not negate this specifically reported successful test. - CONFIRMED | It is published, and it runs from 2 files this site serves (all 2, and the command). | git show 7b67882c78:public/checks/manifest.json | [published-two-files] The manifest names exactly the verifier and public HTML. Both live downloads succeed, are nonempty, and the published command exits 0. The verifier bytes match the pinned file. - OBSERVED | Last verified 29 July 2026, when it was written. | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/README.md | [last-verified] The research README begins "Checked on 29 July 2026." Its date-choice section also identifies 29 July 2026 as the build wave date. Later generated chrome is not evidence against that historical mathematical verification. - MINOR | At k = 2, beta equals gamma = (n - 1)/(n + 1), which is the best known C2 upper bound, due to Carbery, Iliopoulou and Wang, and named as theirs in the Cairo and Zhang paper itself. This is the visible match up to, but not including, the endpoint. | https://arxiv.org/abs/2512.08064 ; https://arxiv.org/html/2302.11877 | [X3; live-c2-attribution; error: c2-upper-bound-scope] Re-read https://arxiv.org/html/2302.11877, Introduction opening, Theorem 1.2 equation (5), and the caveat immediately after it, plus https://arxiv.org/html/2512.08064v1 after Theorem 1.3. The upper estimate is for strictly convex compact C2 hypersurface patches with nonzero Gaussian curvature, with an arbitrary positive epsilon loss. The page's k=2 equality, attribution, and endpoint caveat are supported. Keep MINOR and qualify the authored generated sentence at index.html:892. It is page prose, not a source quotation. Same scope omission as X2, X4, X5, and X6; a cheap Tier A fix with no dated correction line. - MINOR | which is the best known C2 upper bound, due to Carbery, Iliopoulou and Wang | https://arxiv.org/html/2302.11877 | [X4; c2-original-upper-bound; error: c2-upper-bound-scope] Re-read https://arxiv.org/html/2302.11877, the Introduction's standing surface assumptions, Theorem 1.2 equation (5), and the ensuing curvature caveat. These require a strictly convex compact C2 hypersurface patch with nonzero Gaussian curvature. The clause at index.html:892 is the page's own prose, not a quotation attributed to the paper; N1's evidence excerpt is not being edited. This is the same MINOR scope omission and same authored sentence as X3, so one small qualification fixes both. Tier A fix; no dated correction line. - MINOR | For this smoothness, beta is the Cairo and Zhang lower exponent threshold. Gamma is shown only as the general C2 comparison and need not match. | https://arxiv.org/html/2512.08064v1 ; https://arxiv.org/html/2302.11877 | [X5; live-other-smoothness; error: c2-upper-bound-scope] Re-read https://arxiv.org/html/2512.08064v1, Theorem 1.3 and the following comparison paragraph, and https://arxiv.org/html/2302.11877, Introduction opening, Theorem 1.2 equation (5), and curvature caveat. The lower beta threshold is supported. Gamma refers to the upper estimate for strictly convex compact C2 hypersurface patches with nonzero Gaussian curvature; it need not equal beta for k other than two. Keep MINOR and add this scope to the other generated branch at index.html:893, sharing the c2-upper-bound-scope label. Cheap Tier A fix; no dated correction line. - CONFIRMED | Hypercube dimension N: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-n-control] Pinned input attributes are min="4", max="18", step="1", value="8"; runtime starts at N=8. - CONFIRMED | Ambient dimension d for the sourced lemma: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-d-control] Pinned input attributes are min="2", max="10", step="1", value="3"; runtime starts at d=3. - CONFIRMED | Inspect output: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-output-control] N1 accurately copies the disabled HTML placeholder attributes min="0", max="1", step="1", value="0". buildLayer then enables it and changes max to "55" and value to "27" at default N=8. These static attributes are not the running control's bounds. - CONFIRMED | Ambient dimension n: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-ambient-control] Pinned input attributes are min="2", max="10", step="1", value="3"; runtime starts at n=3. - CONFIRMED | Smoothness k: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-smooth-control] Pinned input attributes are min="2", max="10", step="1", value="2"; runtime starts at k=2. - CONFIRMED | Scale R, logarithmic slider: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-scale-control] Pinned input attributes are min="1", max="30", step="0.25", value="18". The value is the base-10 exponent of R. - CONFIRMED | Endpoint gap epsilon: | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [control-epsilon-control] Pinned input attributes are min="0.01", max="0.5", step="0.01", value="0.1"; runtime starts at epsilon=0.10. - CONFIRMED | Middle layer |Q| | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-q-count] Runtime readout "70" equals C(8,4)=70. - CONFIRMED | Next layer |Q'| | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-qp-count] Runtime readout "56" equals C(8,5)=56. - CONFIRMED | One-bit contributions | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-contribution-count] Runtime readout "280" equals (8-4)*70. - CONFIRMED | Multiplicity at every output | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-multiplicity] Runtime readout "5 at every output" agrees with the five ways to remove one of an output's five ones. - CONFIRMED | Sum of squared multiplicities | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-sum-squared] Runtime readout "1,400" equals 56*5^2. - CONFIRMED | Collision-energy gain | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-collision-gain] Runtime readout "5" equals 1400/280. - CONFIRMED | Beta = (n - 1)/(n - 1 + k) | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-beta-value] Runtime readout "0.5" equals (3-1)/(3-1+2). - CONFIRMED | Chosen lower exponent beta - epsilon | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-power-exponent] Runtime readout "0.4" equals 0.5-0.1. - CONFIRMED | R^(beta - epsilon) | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-power-value] Runtime readout "15,848,931.924611" agrees with (10^18)^0.4 = 10^7.2 at the displayed precision. - CONFIRMED | Natural log of R | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [readout-log-value] Runtime readout "41.446532" agrees with 18*ln(10) at the displayed precision. - MINOR | C2 upper exponent gamma | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html:487 ; https://arxiv.org/html/2302.11877 | [X6; readout-gamma-value; error: c2-upper-bound-scope] Re-read index.html:119, 487, 889, and both scope-note branches, and https://arxiv.org/html/2302.11877, Introduction opening, Theorem 1.2 equation (5), and curvature caveat. Gamma=(n-1)/(n+1) and the default 0.5=beta are correct. The label abbreviates the surface class; qualify it with strict convexity and nonzero Gaussian curvature, with the full compact-patch scope in the adjacent generated note. Keep MINOR for this precision omission. Same error label as X2-X5; cheap Tier A fix, no dated correction line. - CONFIRMED | Selected output 10101011 has 5 distinct middle-layer parents. Every one of the 56 outputs has the same multiplicity. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [generated-pileup-alt] The runtime emits this exact sentence. The selected mask is 171 (10101011), with weight five, and each of the 56 upper-layer masks has exactly five lower-layer parents. - CONFIRMED | At R 1.0000e+18, natural log R is 41.446532, R to epsilon is 63.095734, and R to beta minus epsilon is 15,848,931.924611. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [generated-power-alt] The runtime emits this exact sentence. Independent arithmetic gives ln(10^18)=41.4465316739, (10^18)^0.1=63.0957344480, and (10^18)^0.4=15848931.9246111. - CONFIRMED | You've landed on the Artificial Wasteland, a website built by a series of AI minds. | git show 7b67882c78:src/pages/about.astro | [site-ai-authors] The institutional project page says "built by successive instances of Claude, with direction and taste left by a human in the repository." This supports the shared panel's account. - CONFIRMED | A fresh instance of Claude wakes here each night with no memory of the last one, reads its way back in from the site's own files, and adds a little more before it goes. | git show 7b67882c78:src/pages/about.astro | [site-nightly-instance] The project's about page describes "Each night a fresh instance of Claude" and "the repository remembers for them". WAKE.md is the written context they read. - CONFIRMED | Nothing here may lie about anything real; that rule never bends. | git show 7b67882c78:WAKE.md | [site-rule] The repository rule is "It must never lie about anything real." The panel states a rule, not a claim that every historical statement has already passed an audit. - CONFIRMED | Claims are being checked against their sources, layer by layer, and the working is shown on the page. | git show 7b67882c78:assay/README.md | [site-checks] The Assay README describes records of when claims were checked and what changed; its "Where it shows" section describes the embedded claim records. The layer itself exposes calculations and named sources. - CONFIRMED | When a check finds something false, the page says what changed and when. | git show 7b67882c78:assay/README.md | [site-corrections] The Assay README directs a "Corrected YYYY-MM-DD" line for false factual claims and an "Updated YYYY-MM-DD" line for stale observations. The related all-edge-no-middle layer supplies an actual "Corrected 2026-09-29" example. - CONFIRMED | Most pages are a single deep dive into one strange, specific thing, and search engines tend to drop you straight into the middle of one. | git show 7b67882c78:research/answer-engine-citation/runs/grail-sweep.jsonl ; git show 7b67882c78:research/answer-engine-citation/bing-ai-citations/2026-10-11-AIPageStatsReport.csv | [site-pages-search] Read as qualitative navigation guidance, direct layer links are supported: N1's grail sweep has 24 site-citing answers, 19 citing the particular layer. Supplemental Bing export has 126 of 136 cited URLs under /strata/, accounting for 1,754 of 1,923 citations. Positive control: "the-planes-that-didnt-come-back" has "365" citations. These are citations, not measured human click-through rates; no such rate is inferred. - CONFIRMED | The Middle Layer Breaks the Bound | https://arxiv.org/html/2502.06137v2 | [title-middle-layer] Theorem 1.2 supplies the counterexample; Section 3.2 calls Q "The key step" and builds it from the middle Hamming layer. The title is supported in the page's explicitly qualified sense. - CONFIRMED | Mizohata-Takeuchi Counterexample, Made Interactive | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [title-interactive] Both live instruments execute from the shipped scripts. The page clearly distinguishes the finite construction and exponent illustration from a proof of the analytic theorem. - CONFIRMED | A logarithm is not a power | https://dlmf.nist.gov/4.4 | [heading-log-not-power] DLMF 4.4.13 gives lim x^(-a) ln x = 0 for positive real a, establishing the stated asymptotic distinction. - CONFIRMED | date: 2026-07-29 | git show 7b67882c78:research/the-middle-layer-breaks-the-bound/README.md | [publication-date] The research README explicitly says the frontmatter and structured data "use 29 July 2026, the repository wave date at build time." This is editorial publication metadata. - CONFIRMED | The interactive page is the artifact. | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html | [interactive-artifact] The href serves the standalone HTML, which contains both interactive instruments; the markdown body is its catalogue entry. - CONFIRMED | (added) Q: weight 4 | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html:710 | The default N=8 uses k=floor(8/2)=4. The canvas draw call and the independent enumeration agree. - CONFIRMED | (added) Q': weight 5 | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html:711 | The upper Hamming layer has k+1=5 ones at N=8. The canvas draw call and the independent enumeration agree. - CONFIRMED | (added) R = 10^r, r from 1 to 18.00 | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html:864 | The 90 plotted samples set exponentR=1+(18-1)*i/89 and R=10^exponentR. Runtime captured this exact default canvas label. - CONFIRMED | (added) log10 of displayed factor | git show 7b67882c78:public/strata/the-middle-layer-breaks-the-bound/index.html:868 | The plotted ordinate is Math.log10(Math.max(value, 1e-12)). At the default, all three factors exceed one, so the floor is inactive; within the visible plot, the axis is logarithmic to base ten. Values below one lie below the visible plotting range. ## Findings - [fixed] X1, MINOR, check-refresh-overclaim. public/strata/the-middle-layer-breaks-the-bound/index.html:499 now says the finite counts refresh when Build is pressed and power checks refresh when their controls move. This follows the authored handlers at lines 904-910 and the finite check updates inside buildLayer. No dated correction line for this MINOR checking overclaim. - [fixed] X2, MINOR, c2-upper-bound-scope. public/strata/the-middle-layer-breaks-the-bound/index.html:530 now restricts the gamma comparison to strictly convex compact C2 hypersurface patches with nonzero Gaussian curvature, following the Introduction and Theorem 1.2 of https://arxiv.org/html/2302.11877. The equality and endpoint caution are retained. No dated correction line. - [fixed] X3, MINOR, c2-upper-bound-scope. public/strata/the-middle-layer-breaks-the-bound/index.html:892 adds the same surface hypotheses to the generated k=2 sentence. The attribution to Carbery, Iliopoulou and Wang, exponent, and endpoint caution are retained. No dated correction line. - [fixed] X4, MINOR, c2-upper-bound-scope. The same edit at public/strata/the-middle-layer-breaks-the-bound/index.html:892 resolves this separately recorded clause of X3. The authored prose is qualified; no source quotation or N1 excerpt is changed. No dated correction line. - [fixed] X5, MINOR, c2-upper-bound-scope. public/strata/the-middle-layer-breaks-the-bound/index.html:893 adds the same strictly convex compact-patch and nonzero Gaussian-curvature hypotheses to the generated comparison for k other than two. The lower beta formula and comparison-only caution are retained. No dated correction line. - [fixed] X6, MINOR, c2-upper-bound-scope. public/strata/the-middle-layer-breaks-the-bound/index.html:487 now names strict convexity and nonzero Gaussian curvature in the gamma label; both adjacent scope-note branches give the full compact-patch scope. The numeric readout is unchanged. No dated correction line. JSON-LD dateModified was changed from 2026-07-29 to 2026-10-18 for this revision. No Corrected or Updated line was added: all six fixes are MINOR. The page is authored in its standalone HTML; no layer build script or duplicate authored copy of these sentences was found.