Fields Medalist Timothy Gowers: GPT-5.5 Pro Produced PhD-Level Math Proofs — Research Faces 'Crisis'

The Experiment

Timothy Gowers, 2014 Fields Medal laureate and Professor at Cambridge University, published a striking blog post detailing his experience using ChatGPT 5.5 Pro on unsolved mathematical problems from his own research. The model produced what he describes as PhD-level mathematical proofs in roughly one hour — with minimal human prompting.

What Made It Remarkable

Gowers was not testing the model on known textbook problems. He fed it problems from his active research where the solutions were unknown. The model navigated the problem structure, identified an approach, and produced a proof that Gowers evaluated as publishable in a mathematics journal. He concluded: if this is the current level of ability, mathematical research will face a crisis in the near future.

The Crisis Warning

Gowers' concern is not that AI will make mathematics trivial, but that the rate of progress is outpacing what mathematicians anticipated. He distinguishes between AI solving routine results versus genuinely open research problems — and suggests the latter threshold has now been crossed.

The blog post and associated Twitter thread sparked extensive discussion on both Hacker News and Reddit, with mathematicians divided between those seeing this as a transformative moment and those who remain skeptical about the depth of understanding behind the AI's output. Gowers himself acknowledged that AI collaboration may become a new methodology for mathematical research.