OpenAI model's 372 complex mathematical proofs stir the field, as leading mathematicians weep in despair

On 6 October 2026, OpenAI published hundreds of mathematical research papers produced by one of its AI models, causing a stir across the world of mathematics. The papers contained proofs of 372 complex mathematical problems. Some of those problems had been worked on for decades, and others had occupied mathematicians for centuries.

Of the 722 manuscripts published, three were later withdrawn. For many of the problems, on which researchers had spent large parts of their lives, the AI produced correct solutions in a matter of hours or days.

Many of the world's leading mathematicians were astonished by the event, but the astonishment was mixed with anger, despair and a deep sense of emptiness. The question being asked is what will become of mathematicians' long years of work, their research and their future, if a machine can solve difficult problems so quickly.

French mathematician Hugo Dumini-Copin won the Fields Medal, one of mathematics' most prestigious prizes, in 2022. On hearing of OpenAI's success, he was so distraught that he sobbed loudly enough to wake his daughter. His comment on the future of his own research field was: 'They have destroyed my research field.'

Steven Strogatz, a mathematician at Cornell University in the United States who has written several popular books on mathematics, also broke down in tears when speaking about the AI's recent success. He felt that mathematicians had been able to think of themselves as people who discover new things, solve hard problems and give the world something new through those discoveries, and that this identity was now changing. For Strogatz the change is at once thrilling and frightening: at times he is excited by new possibilities, at times he feels lost, and at times he feels deep grief, because for mathematicians a solution is not just a correct answer — it involves long effort, pushing past failure, and the experience of sharing knowledge with other researchers.

One of the hard problems of mathematics is the Navier-Stokes equations, which are directly related to describing the flow of liquids and gases. Researchers had been seeking the answer to an important mathematical question about them for nearly a century. Last month, OpenAI claimed its internal AI model had solved the equation in just 88 hours.

The Navier-Stokes problem is one of the seven 'Millennium Prize' problems announced by the Clay Mathematics Institute in 2000, each carrying a prize of one million dollars. The verification of OpenAI's solution was under way, and the Clay Mathematics Institute has also acknowledged that the model may have solved the problem.

The presentation of the solution drew criticism at first. Many found the proof hard to understand, and some dismissed it as 'AI slop'. But when hundreds of research papers were published at the start of October, the matter was no longer confined to a single problem; with so many solutions appearing at once, mathematicians realised a major change was coming to their world of work.

Kevin Buzzard, a professor of mathematics at Imperial College London in the United Kingdom, noted that in 2022 ChatGPT would make mistakes even when asked to multiply two five-digit numbers. That same technology has now reached a point where it produces solutions to problems representing many years of research in the blink of an eye. Even researchers who understand how AI works find the pace of its improvement astonishing. Leslie Anne Goldberg, head of the Computer Science department at the University of Oxford, also said that, having seen the AI's earlier weaknesses, she had not understood that the technology could improve so fast.

Not all mathematicians have reacted in the same way. Oxford professor Marcus du Sautoy is both thrilled and somewhat anxious about AI. Recently, with AI's help, he disproved a conjecture made by his PhD supervisor in 1986, and he says AI has given his research new life. AI's great advantage is that it can draw on the results of a vast amount of human research at once and find connections between ideas from different studies, though combining human-made ideas is entirely different from a machine creating its own new ideas.

Oxford mathematician James Maynard believes AI may soon become capable of creating major mathematical theories by itself. The possibility is as exciting as it is worrying. In his view, AI can make mathematics progress faster, but there is a large mismatch between the interests of the commercial companies building the technology and the long-term purposes of mathematical practice.

Leslie Anne Goldberg believes the main purpose of mathematics is to increase human understanding. She therefore does not want to be involved in having AI solve problems and publishing the results; as she puts it, she is not a typist, and she sees no point in spending her valuable time merely asking AI questions and relaying its answers. In Goldberg's view, the biggest crisis is for students at the start of research: many PhD students in mathematics spend years working on the very problems through which they train themselves as researchers. If AI solves those problems first, what their research goal will be, and what kind of work will await them after their studies, is now the big question.