OpenAI posts over 700 AI-generated maths preprints at once, drawing both celebration and complaint from mathematicians

OpenAI posted more than 700 mathematical preprints entirely generated by AI on the software repository GitHub on 6 October, spanning various fields of mathematics. Earlier AI results in mathematics tended to appear as single papers or single-problem demonstrations; this release arrived as a simultaneous batch whose scale a mathematician called “staggering”. The Nature report records the release date (6 October), the platform (GitHub) and the count (more than 700), and quotes University of Connecticut mathematician Alvaro Lozano-Robledo commenting on X.

The batch did not include solutions to any of the five remaining Millennium Prize Problems selected by the Clay Mathematics Institute, though some preprints addressed problems related to them, including the Hodge conjecture on complex algebraic varieties and the Riemann hypothesis on the statistical properties of prime numbers. The report explicitly separates “solving a Millennium Prize Problem” from “working on problems related to them”, noting that OpenAI’s September preprint on the Navier-Stokes equation solved one of the seven problems and another was solved by mathematicians in the early 2000s. The Nature report states this scope directly and quotes an OpenAI spokesperson: “We will leave it to the broader community to assess [their] significance.”

Reactions among mathematicians were divided: some celebrated the solution of longstanding problems, while others took to social media to complain about being scooped, with one physicist calling it a “slopocalypse”. This surfaces the practical tensions around priority, credit and community norms raised by AI-generated mathematical output, beyond the question of technical capability. The report quotes several named mathematicians’ public comments, including Lozano-Robledo’s question about why OpenAI would release hundreds of solutions at once and Anthropic mathematician Levent Alpöge’s remarks about “sad stories” alongside “the most significant moment in mathematical history”.

OpenAI said its goal is to “provide tools” for mathematicians and scientists, adding that “to accomplish this, we evaluate and develop our models against difficult frontier problems in all sciences.” This supplies an institutional rationale for the batch release, framing difficult frontier mathematics as an evaluation and development setting for its models. Nature quotes the OpenAI spokesperson’s email directly in the report.

This summary is based on Nature’s news report and is intended for readers who want to understand the release event, its scope and the immediate community reaction. The report itself is not a technical assessment of the correctness or significance of the preprints’ mathematical content, and the OpenAI spokesperson explicitly left significance assessment to the broader community.

The report does not enumerate the mathematical subfields of the more than 700 preprints, their individual technical depth, or any independent peer assessment of any single one. Readers concerned with the status of a particular problem would still need to return to the original preprints and subsequent community evaluation.