OpenAI scans more than 100 world-class puzzles, but he says: "Math is just getting started".

On September 21, OpenAI threw out an announcement: An internal model that began training on August 28 has already captured more than 100 world-class mathematical puzzles spanning most areas of mathematics.

It can be said that the mood of the mathematical circle has been sinking for half a month. Readers are worried about the decline in degrees, textbooks are worried about the collapse of journals, and those who write half-hour proofs begin to doubt that the craft is still a few years away.

Right in the middle of this low pressure, University of Toronto mathematician Daniel Litt posted a long post on his personal blog titled "A Beginning for Mathematics".

The title is a bit of a self-correction.