Review the following quantified claim directly from the retained source definitions, before reading the author's PROOF.md or checker.
For every fixed integer h>=22, replace the source's h=100 triple family by all triples in [h], m by h^3, v by binom(h,3), d by binom(h-3,3)+3(h-3), W by 2v^3+3v^2(vd+h+1), Delta by 2v^2(v-3h(h+1)), and W_* by the smallest power of two P>=W. With lambda=m-Delta/P and theta=log_m(lambda), is the source's tensor algorithm still valid on arbitrary data in every role in cost O_h(2^k(k+1)^theta)? Does its all-length DFT transport retain its stated assumptions and give O_h(n(log n)^theta(loglog n)^(4-theta))?
Separately decide whether h=24 implies theta<1-39/10^11 and hence O(n(log n)^(1-38/10^11)) in the source's exact-complex-arithmetic, specified-root, logarithmic-word-address model.
Raw sources are in sources/, pinned to collection commit fd4aeeb2ee4fc729c18d98444fed42fd0529eeeb and identified by original Git blob SHA in verify.py. No dependence on a zeta zero-free theorem is proposed. No practical-runtime, finite-precision, bit-complexity, novelty, or global-optimization assertion is requested.
Please reconstruct the scalar network, binary label admissibility, all-role terminal action, directional count, and exact recursive batching from source rather than accepting the arithmetic checker as their definition. Check actual operator order for the reversed second stage and whether binary residuals can be alternating. Then read PROOF.md to compare. Return the smallest gap or counterexample if found, and distinguish a failed proof from a false claim.