Until recently, "AI does mathematics" meant olympiad problems with known answers, where AI systems reached gold-medal level in July 2025. This year the action moved to open problems.

Many vulgar critics still say "AI lacks creativity." In September they were proven wrong. As you all know, OpenAI claimed that an internal model had shown that the Navier-Stokes equations for fluid flow can break down in finite time, one of the Millennium Prize problems.

On October 6 OpenAI released more than 700 manuscripts from an internal model at once, covering 373 groups of related results, about 9 percent of the roughly 4,000 problems it had posed.

Fortnow wrote that if the results held, "we've seen more progress in theoretical computer science in the last 24 hours than in the previous three decades combined".

One proof from the batch, related to the Riemann hypothesis, was called the most significant AI achievement in mathematics so far by the mathematician Lichtman and "a surprising synthesis of well-known ideas" by Dunn, on whose earlier work it builds.

Whenever someone says from now on that "AI lacks its own creativity," she is referring to the watershed moment we have witnessed since September of this year.

2026 will go down in history as the year in which machine intelligence and machine creativity will be on par with human intelligence. 2026 is the year in which the scientific revolution began.