OpenAI cracks 10 unsolved math problems, and a Fields Medal winner at Oxford is grappling with what comes next

OpenAI's announcement that it produced solutions to 10 long-standing mathematics problems, some of which had resisted resolution for decades, is drawing serious attention from inside the discipline. James Maynard, a University of Oxford professor who holds the Fields Medal, told The Verge the development sent him into a sustained period of "soul searching" about the future of his field.

The ten problems in focus

OpenAI's disclosure covers 10 distinct problems, not a single result. Some had confounded academics for decades, which makes the scope of the announcement unusual by the field's own standards. The technology works the way generative AI does when producing text or images, or when proposing ideas in science and medicine: it learns patterns and connections from large volumes of existing material, then uses those patterns to generate new outputs. Here, the outputs are mathematical solutions.

Maynard's year of grappling

Maynard holds the Fields Medal and works at Oxford, placing him among the field's most active senior researchers. His comments to The Verge came days before OpenAI's most recent disclosure, meaning the "soul searching" had been accumulating before the latest results arrived. His framing of the discipline is direct: mathematics has historically been slow-moving, and is now hurrying to adapt. That hurrying is a response to pressure from outside the field, not a pace it chose.

What to watch

The immediate question is whether OpenAI publishes the specific problems and its methodology for external review, giving the mathematics community concrete material to assess. Until that release, the 10-problem figure is the only verifiable number from the disclosure. Maynard's reckoning arrived before the latest print. The 10-problem count is what stands on the tape.