---

Claim. The note A fixed-window unconditional reduction for Erdős–Graham's binomial divisor problem, dated 29 May 2026 and headed as a prepared note with no byline, claims the negative answer to Problem 387 without the Generalized Riemann Hypothesis. Przemek Chojecki posted it in the site's discussion thread on 2026-05-29 and wrote that GPT-5.5 Pro gave them the argument. Its Theorem 2.1 states that for every fixed $B\geq3$ and arbitrarily large $k$ there are a residue class $n\equiv\alpha\pmod{N_k}$ and fixed integers $g_i$ with $g_i\mid n-i$, $g_i\geq B$ and $\prod g_i=k!$, such that no prime $p\leq k$ divides $\binom nk$ or any of the quotients $(n-i)/g_i$; the only analytic input is the Siegel–Walfisz theorem. Combining this construction with the unconditional divisor propositions of the first, GRH-conditional version of arXiv:2605.21221 (its Sections 6 to 10), Theorem 1.1 claims that for every fixed $B\geq B^*$ infinitely many $\binom nk$ have no divisor in $(n/B,n]$, and Corollary 1.2 that no absolute constant $c>0$ works. The note states that it does not recover the moving window of that paper's Theorem 1.4. Its Section 5 refers the Lean formalization of the fixed-$B$ construction to Naprienko's repository, which it cites as [5].

Submission note. Posted to the site's forum by Przemek Chojecki on 29 May 2026:

With some back-and-forth GPT-5.5 Pro gave me an argument for an unconditional covering theorem without assuming GRH. Have a look at this note. This should finish the problem.

Depends on. The divisor propositions of the paper whose later version is the accepted claim Bui, Naprienko, Pratt and Zaharescu 2026, which the note uses without reproving them.

Discussion. Pratt replied the same day that the note looks similar to one of Naprienko's notes. Naprienko wrote on 2026-06-03 that the note rewrites their fixed-$B$ construction, public in their repository since 2026-05-23 (the write-up unconditional_bpz_patch.pdf with the Lean development CoverBPZ), and that this construction completes the qualitative question.

Standing. The note is not refereed and has no outside acceptance, so the claim stays claimed. The problem's standing rests on the accepted claim of Bui, Naprienko, Pratt and Zaharescu.