OpenAI published over 700 AI-generated manuscripts claiming solutions to open math problems. A math blog collected over 100 responses from researchers, ranging from fascination with new ideas to existential fears and grief over a lost way of working. Fields Medalist Hugo Duminil-Copin wrote, "I am paralysed."
SynthePulse Insight · AI deep reading
Version 1 · 1 source
On October 6, 2026, OpenAI released more than 700 manuscripts at once, claiming to solve hundreds of unsolved mathematics problems. More than 100 responses collected by the mathematics blog Proofs and Prompts show that some results shocked experts, but the manner of release, paper readability, missing authorship, and career prospects triggered stronger doubt and grief than the results themselves.
On October 6, 2026, OpenAI released more than 700 manuscripts at once, claiming to solve hundreds of unsolved mathematics problems. The mathematics blog Proofs and Prompts subsequently collected more than 100 responses from researchers, ranging from intellectual fascination to deep grief over the disappearance of a way of working.
Reports say it is still too early to assess the mathematical substance of the content OpenAI uploaded to GitHub. Some papers have been withdrawn, while others have been criticized for being difficult to read. Even so, several experts said the results impressed them.
Ben Green called some results "absolutely astonishing to me" and said the release solved about three-quarters of the research objectives of the European grant project he received that June. Roman Sauer recognized techniques he had developed over many years in OpenAI's solutions to two completely different problems, and said, "I was stunned."
Alvaro Lozano-Robledo called October 6 "possibly the most important day in the history of mathematics so far," while also noting the model's limitations. OpenAI aimed at the Riemann Hypothesis but obtained only a weaker variant of it. Lozano-Robledo said, "The quasi-Riemann Hypothesis is huge! But it is not the Riemann Hypothesis."
Hugo Duminil-Copin, winner of the 2022 Fields Medal, wrote that OpenAI claimed to have solved every major unsolved problem he had publicly discussed in his career, and said, "The shock is immense. I am paralyzed."
Doctoral student Tristan Humbert, upon learning that OpenAI claimed to have proved the core problem of his doctoral dissertation, felt angry after revising his research plan and reading the paper, calling it "basically unreadable garbage."
Terence Tao expressed mixed feelings, saying that the AI-generated proofs introduced "clever new ideas," but that he felt "deeply dismayed" by the absence of people who could discuss, present, and teach this work.
Henry Wilton criticized OpenAI for withholding author names and failure-rate information, arguing that the lack of authorship means "mathematics is no longer a human endeavor."
Enrico Fatighenti pointed out that researchers applied different standards to AI-generated papers than to human work, and that the community was forced to interpret and explain the AI system's papers while credit still went to OpenAI.
Johannes Schmitt argued that these research papers are a secondary byproduct of OpenAI's efforts to improve its models, rather than primarily aimed at advancing mathematics. This interpretation is only a personal view; the report does not state OpenAI's official motivation for releasing these manuscripts.
Fields Medalist Peter Scholze called for patience and warned that these systems might find ways to break widely used encryption methods. This warning is his personal, non-expert judgment and has not been confirmed.
Reports say that since mid-2026 breakthroughs have been accumulating, first the Unit Distance Conjecture, then Navier-Stokes, and then the ten solutions demonstrated at the release of OpenAI's Astra model.
Danny Calegari said personal curiosity has always guided his project choices, that OpenAI's release did not affect his current work, and praised a mathematical animation soundtrack created in collaboration with a musician, with code written by Anthropic's Claude.
The mathematics community remains divided over the long-term impact of AI-generated papers and the role of human researchers. The report is based only on researchers' responses collected by Proofs and Prompts; it provides no independent mathematical verification, nor a specific list of solved mathematics problems or per-paper verification results.
This article is based on THE DECODER's report on responses collected by the mathematics blog Proofs and Prompts. That OpenAI released more than 700 manuscripts, that Proofs and Prompts collected more than 100 responses, and that some papers have been withdrawn are confirmed facts; the specific evaluations by mathematicians, Scholze's encryption-risk warning, and Green's claim of about three-quarters of research objectives are all source claims and have not been independently verified. The report provides no independent mathematical verification, and the specific number and reasons for withdrawn papers are not stated.
The core tension of this release is not whether AI can produce mathematical results, but the manner in which it enters a community that depends on discussion, teaching, and authorship. Some results shocked experts, but withdrawals, poor readability, failure rates, and missing author information have left the mathematics community feeling grief and unease more strongly than shock. The report says assessment of the mathematical substance is still premature.
Primary report
Primary source