--- id: "2026-09-26-mubayi-zarankiewicz-astra" url: "https://postcutoff.com/e/2026-09-26-mubayi-zarankiewicz-astra/" as_of: "2026-11-19T22:11:00+01:00" date: "2026-09-26" date_precision: day category: science importance: 3 confidence: high status: [Event confirmed, Awaiting review] verification: Preprint only sources: 1 editor: human editor human_review: null version: "2026-10-04" --- As of: 2026-11-19 22:11 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-09-26-mubayi-zarankiewicz-astra/ # GPT-6 Astra finds a proof improving the Kővári–Sós–Turán bound Full title: GPT-6 Astra finds a proof improving the Kővári–Sós–Turán bound: diagonal bipartite Ramsey numbers b(t,t) = O(2^t) (Mubayi) On Sept 26, 2026 Dhruv Mubayi posted a proof that every bipartite graph with N vertices per side and at least N²/2 edges contains K_{t,t} once N ≥ c·2^t. That improves the classical Kővári–Sós–Turán requirement of order t·2^t and cuts the diagonal bipartite Ramsey upper bound from Conlon's O(2^t log t) to O(2^t). The abstract says "The proof was found by GPT-6 Astra": an 8-page proof the author first found incomprehensible and then rewrote with the model. ## Key facts - arXiv 2609.32937 (math.CO), Dhruv Mubayi, 'Sharper Zarankiewicz and Diagonal Bipartite Ramsey Bounds', Sept 26, 2026 - Result: density ≥ 1/2 bipartite graphs contain K_{t,t} when N ≥ c·2^t; hence b(t,t) = O(2^t), improving Conlon's O(2^t log t) - AI declaration: 'The proof of the main result was found by GPT-6 Astra after several guided prompts by the author. Astra first produced a proof that was only 8 pages long and was incomprehensible to the author'; later write-ups are largely the author's - Mubayi had earlier prompted Astra to extensions of the Erdős–Sós work (see [2026-09-03-erdos-sos-conjecture-proved-gpt-6-astra](https://postcutoff.com/e/2026-09-03-erdos-sos-conjecture-proved-gpt-6-astra/index.md)) ## What happened Mubayi, a leading extremal combinatorialist, reports that GPT-6 Astra found a proof removing the log factor from the best bound on diagonal bipartite Ramsey numbers. The improvement over the 1954 Kővári–Sós–Turán counting bound comes from a new method that he says "will probably have further applications". ## Why it matters It is a quantitative improvement on a classical bound in a central area, credited outright to a model in the abstract. That kind of disclosure was rare before September 2026. ## What is disputed or not yet verified - Verification: Unrefereed preprint ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 118 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 88 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 179 days after its cutoff - Grok 4.7 (training cutoff May 2026): 118 days after its cutoff ## Sources 1. [arXiv 2609.32937: Sharper Zarankiewicz and Diagonal Bipartite Ramsey Bounds](https://arxiv.org/abs/2609.32937) (arxiv.org, paper) ## Changes - 2026-10-01 (filed): Created ## Related - 2026-09-30: [Summer 2026 flood: dozens of named conjectures settled on arXiv with disclosed AI help](https://postcutoff.com/e/2026-09-30-ai-assisted-conjecture-wave-summer-2026/index.md) - 2026-09-16: [Chvátal's 1972 conjecture proved (Chang–Liu–Liu, ChatGPT-assisted), then a GPT-6 Astra 'proof from The Book' and a Codex-built Lean formalization](https://postcutoff.com/e/2026-09-16-chvatal-conjecture-proved/index.md) - 2026-09-03: [GPT-6 Astra proves the Erdős–Sós conjecture with a short counting argument](https://postcutoff.com/e/2026-09-03-erdos-sos-conjecture-proved-gpt-6-astra/index.md)