OPENAI’S 719 MATH PAPERS SET BITCOIN TALKING ABOUT BROKEN KEYS — WE SEARCHED EVERY ABSTRACT — ZERO MENTION DISCRETE LOGARITHMS, ECDSA OR CRYPTOGRAPHY

On October 6, 2026, OpenAI published a GitHub repository of math results from an unreleased internal model: 719 manuscripts in 372 families, drawn from about 4,000 problems, at about three hours of ChatGPT Pro thinking per result, per the repository’s README. The Block reports that Justin Drake’s warning followed it. Drake called the release evidence that “mathematical superintelligence is upon us” and wrote that it was “reasonable to brace for the possibility that ECDSA breaks before qday, in the worst case in months not years.”

So we read the abstracts. The repository’s manuscript map, CONTENTS.md, lists every paper with its abstract. On October 8 we searched it. The words cryptography, discrete logarithm, ECDSA, secp, public key, hash function and post-quantum appear zero times. RSA and Shor do not appear either. Elliptic curves come up 17 times, in papers on ranks, modularity and the Birch and Swinnerton-Dyer formula over the rational numbers. A bitcoin private key rests on the discrete logarithm problem on one curve over a finite field. No abstract claims anything about it.

The release is not settled math either. On October 7, OpenAI withdrew three manuscripts over a sign error and revised 14 others. By its own count, 300 of the 719 top-line results, about 42%, have Lean proofs a computer can check. The rest wait on referees.

What is checkable: the abstracts, the withdrawals, and which coins sit behind keys already on the chain. What is trusted: a forecast that a model whose 719 abstracts never mention discrete logarithms will break them within months. A key that has never signed is still hidden behind a hash. That costs nothing to check. At the time of this writing, bitcoin was $81,753.