The end of mathematics?

Are we witnessing the end of mathematics?

Can a machine make mathematicians obsolete? That question has hung over the field since OpenAI released more than 700 AI-generated proofs on October 6, claiming solutions to at least 372 long-standing problems, according to a Morning Star column.

The short answer is no, though the profession is now being forced to ask what it is for.

Why it feels like an ending

The release followed OpenAI’s September 8 claim that its AI agents found a counterexample in the Navier-Stokes problem, one of the seven Millennium Prize problems. The column says OpenAI used 10,000 agents and took under 88 hours. The proof is written in Lean, a format that is logically checkable but almost unreadable to people.

That gap matters most. If proofs arrive faster than anyone can read them, mathematics risks containing results no human understands. Tristan Buckmaster, who was working on the same problem, called it a “Deep-Blue Kasparov moment.” Chess survived Deep Blue, but it was never the same game afterwards.

Why it isn’t one

Little has been verified. No human has confirmed the Navier-Stokes result, and the Clay Institute has two years to rule on it. Within a day of the larger release, OpenAI withdrew three files over errors and amended 19 more.

There are also open disputes. Buckmaster and Levent Alpöge said their recent advances were leaked to OpenAI days before its announcement. Both had used OpenAI’s Codex tool, and it is unclear whether the company’s system could access their data.

Many mathematicians also argue the goal of the field is understanding. A letter first signed by 25 Fields Medallists said AI firms’ approach harms mathematics, and a new Association for Human Mathematics accused OpenAI of disregarding scientific norms.

What to watch

The Clay Institute’s verdict, independent checks of the 372 claims, and how journals treat AI-generated proofs will show whether this is an ending or a turning point.