{ "source_url": "https://github.com/twinheterodox/beta4/blob/main/beta-intervals/review/revision-review.txt", "report": "Independent AI review of the revised interval paper\n18 November 2026\n\nThis public edition retains the mathematical checks on the revisions after\nthe first audit. The reviewer examined the changed passages, their\nsurrounding proofs, the source differences and the cited results, and\nfound no further correction necessary.\n\nDecay with the published normalization\n\nIn the fifteen-value paper the fifth numbered equation of Section 4,\n(4.5), gives the certified upper bound\n\n A = 745969417189431202995707 / 92003136067626388684800.\n\nThe proof following (4.6) bounds the normalized large-prime contribution\nby n(A+o(1)) for precisely the isolated-pole exponents used in the\nrevised manuscript. Rounding A upward gives 8.10809. After subtracting\nG log(n!), the small odd primes contribute o(n) to the upper bound\nand 2 contributes at most 15n log 2+o(n). The global clearer for the\npowered form is D_n^q. Thus each of these terms enters the logarithm\nwith a factor of q.\n\nEquation (4.9) bounds the global maximum of the same phase by\n-18835112315041/10^12 = -18.835112315041. The revised -18.83511\nrounds upward. The derivative estimate for x>=3 in the same source\ngives the claimed tail factor (3n/t)^(3qn). For fixed q this absorbs\nthe fixed polynomial from the Stirling error on t>3n. On the finite\nrange the derivative bound contributes only a polynomial factor.\nThe (n!)^(-qG) exactly cancels the factorial contribution from D_n^q.\n\nWriting x=6931472/10000000, I checked with exact rational arithmetic\nthat sum(x^j/j!, 0<=j<=9)>2. Since x>0, this implies log 2