This past Friday, OpenAI published a proof for a maths problem that had been unsolved for the past 50 years. Most importantly, that proof was entirely generated by its latest gpt-5.6 Sol model. It has not yet been formally verified, but early expert reaction has been cautiously positive.

What does this mean for mathematicians? Why should we even care? How good are LLMs at maths anyway?

I’m a mathematician who’s been working in data and AI for the past decade, and helped train some of the frontier models at graduate level maths. This article attempts to address these questions.

Roasting welcome.