AI Makes Historic Progress on 150-Year Math Mystery

A computer just did something mathematicians have been trying to do for more than a century and a half.

Anthropic's new AI model made significant progress on the Riemann hypothesis, one of math's most famous unsolved problems. The hypothesis explores how prime numbers are distributed, and there's a $1 million prize waiting for anyone who can prove it.

What makes this breakthrough remarkable isn't just the progress itself. It's how it happened.

An Anthropic employee without advanced math training simply asked the AI to "take a real stab" at the problem. Then they walked away for a day and a half while the model got to work.

The AI coordinated 60 different versions of itself, like a team of researchers working together. It tested 650 different approaches to the problem and generated 31 million tokens of output as it worked through possibilities.

Two of those AI subagents developed the key mathematical ideas. Thirteen others contributed supporting concepts, while 30 attempted new approaches without success. Thirteen served as fact checkers, and two helped write up the findings.

Human mathematicians at Anthropic confirmed the results were valid. The proof was then formalized using Lean, an open source verification tool that ensures mathematical accuracy.

This isn't the first time AI has cracked tough math problems this year. Models have solved multiple Erdos problems, and OpenAI's internal "Astra" model recently proved 10 major mathematical results. Another Anthropic effort even disproved the longstanding Jacobian conjecture.

The news has sparked both excitement and concern among mathematicians. Some worry that AI could undermine the field's tradition of individual researchers taking credit and responsibility for their discoveries.

But others see opportunity. Fields Medal winner Timothy Gowers compared it to astronomy, where stars aren't named after the people who discover them. Maybe math doesn't need individual names attached to every theorem either.

Why This Inspires

This story shows how technology can enhance human curiosity rather than replace it. Someone without deep mathematical training asked a good question, and AI helped bridge the gap between wondering and discovering.

The breakthrough suggests we might be entering an era where some of humanity's oldest mysteries become solvable. Problems that have frustrated brilliant minds for generations could yield to new forms of collaboration between human creativity and machine processing power.

What other unsolved puzzles might fall next? From medical research to climate science, the same approach could accelerate discovery across fields where progress has stalled.

The future of scientific discovery might not be humans versus machines, but humans and machines working together to answer questions neither could solve alone.