Editorial illustration for Mathematician and AI Both Claim Breakthrough on 167-Year-Old Math Problem
Mathematician and AI Both Claim Breakthrough on...
Levent Alpöge had an unusual week. On Monday, Anthropic became the first publicly-traded frontier AI lab, the company he works for. Days later, he says he solved the Riemann Hypothesis, a problem that has sat unresolved since Bernhard Riemann first posed it in 1859 and one of the seven Millennium Prize problems. "The precise timing is horrible," Alpöge wrote in an essay laying out his claim, describing how he texted Anthropic CEO Dario Amodei the previous Sunday night just to ask how he should even go about announcing that he'd produced a Lean-compilable proof.
The Hypothesis is about prime numbers, specifically whether there's a hidden order governing where they fall. Mathematicians have chased a proof for 167 years. If Alpöge's result survives scrutiny, it's the kind of finding that reshapes how number theorists think about the structure of the universe's most basic building blocks.
Word spread fast among specialists, and the reaction has been playing out in real time on a Zulip chat where mathematicians are picking apart the proof line by line. One early comment, from a number theorist at the University of East Anglia, gives a sense of just how deep into the weeds this gets.
It’s a big week for Levent Alpöge: The mathematician saw his employer, Anthropic, become the first publicly-traded frontier AI lab. He also appears to have solved one of the most important problems in mathematics.
Why this matters
Alpöge's claim is unverified, and that's the point worth sitting with. A Lean-compilable proof is checkable by machine, which should make this the easiest kind of breakthrough to confirm or kill. Until the formal verification community runs it down, the "stream-of-consciousness jaunt" language and talk of Claude dancing near the finish line belongs in a blog post, not a journal.
We've seen this pattern before: a researcher with deep access to a frontier model announces a result timed suspiciously close to a corporate milestone, in this case Anthropic going public. That doesn't make the math wrong, but it should make everyone slow down before crediting the AI with cracking a 167-year-old problem. For developers and researchers, the real story isn't whether Claude "solved" the Riemann Hypothesis.
It's whether human mathematicians using these tools as collaborators, rather than oracles, can produce proofs that survive peer review. Watch what Lean's formal verification turns up, not what the essay says.
Further Reading
- Papers with Code - Latest NLP Research - Papers with Code
- Hugging Face Daily Papers - Hugging Face
- ArXiv CS.CL (Computation and Language) - ArXiv