Fields Medalist Timothy Gowers reports ChatGPT 5.5 Pro solved open mathematics problems autonomously, improving exponential bounds to quadratic and polynomial in under two hours with zero human contribution to the core ideas.

ChatGPT 5.5 Pro solved an open problem in number theory posed by mathematician Mel Nathanson in 17 minutes and 5 seconds, delivering a quadratic bound after Nathanson had only achieved an exponential bound. The model then rewrote the argument as a LaTeX preprint in 2 minutes and 23 seconds.

On a generalized variant, the model improved a bound from exponential to polynomial in 31 minutes and 40 seconds total. MIT student Isaac Rajagopal assessed the polynomial-bound result as "quite ingenious" and "completely original," calling it "the sort of idea I would be very proud to come up with after a week or two of pondering."

Gowers concludes that the bar for mathematical contribution has shifted: "The lower bound for contributing to mathematics will now be to prove something that LLMs can't prove, rather than simply to prove something that nobody has proved up to now." He notes that anyone starting a doctorate today and finishing in 2029 at the earliest will see mathematical research "changed out of all recognition" by then.

Google DeepMind's AI agent Aletheia (built on Gemini Deep Think) showed both breakthroughs and limitations: it independently wrote a math paper and disproved a decades-old assumption, but when tested on 700 open math problems, only 6.5 percent of its answers proved usable.