Tech & Social Media

Riemann Hypothesis Solved by AI? Anthropic’s Claude Cracks 167-Year-Old Maths Problem as Proof Checks Out

For 167 years it sat there, untouched. The Riemann Hypothesis — the most famous unsolved problem in mathematics — has resisted every genius thrown at it since 1859. Now an Anthropic mathematician says he has cracked it, with an AI chatbot as his collaborator.

Levent Alpöge, an Anthropic mathematician, announced on Friday, November 20, that he believes he has found a proof of the Riemann Hypothesis, working with the company’s AI model Claude. The claim landed a day after Anthropic staged the largest initial public offering in history. Then, on Saturday, it got something most such claims never get: independent verification.

The Lean Focused Research Organization and the Mathlib Initiative, which maintain the field’s leading proof-checking software, announced they had finished verifying Alpöge’s 17 million lines of code. Their verdict was blunt. “We have full confidence that Alpöge’s proof of the Riemann Hypothesis is correct,” they said in a joint statement, according to The Wall Street Journal.

That does not settle everything — not yet. But it is the closest anyone has come to closing a problem that has outlived generations of mathematicians.

What the problem actually is

Bernhard Riemann proposed it in 1859. It is about how prime numbers are spread out as they grow larger — primes being the numbers divisible only by one and themselves. The hypothesis says that all the non-trivial zeros of a particular function, the Riemann zeta function, lie on a single vertical line, the “critical line”.

Sounds abstract. It is anything but. Prime numbers are the atoms of arithmetic, and the hypothesis, if true, would pin down how they are spread out. That matters far beyond pure maths: much of modern cryptography rests on the behaviour of primes. The Clay Mathematics Institute named it one of its seven Millennium Prize Problems in 2000, each carrying a $1 million prize. In 26 years, humans alone had solved one of the seven: the Poincaré conjecture, proved by Grigori Perelman in 2003.

How Claude helped

Alpöge unveiled the proof on his personal website in a four-page note — not as a conventional written proof, but as computer code in Lean, a language that lets machines check mathematical arguments line by line. Seventeen million lines of it. He worked with Claude for about six weeks, according to USA Today. Near the end, he wrote, the AI pulled far ahead of him and carried the work over the finish line on its own.

Here is the strange part. Alpöge himself says he does not yet fully understand the mathematics in his own proof. “Does that mean I understand it? Not yet,” he wrote. The computer is “extraordinarily picky”, he added — satisfy it, and the reasoning holds even when a human mind has not yet caught up. He plans to write the argument up properly once he can simplify it.

A formal proof in Lean is a machine-readable certificate. The human-readable explanation — the part mathematicians can argue over, teach and build on — still has to be written.

The verification, and the doubts that remain

The Lean and Mathlib checkers said they verified that Alpöge’s code proves the correct mathematical statement, and that multiple independent checking systems accepted the proof. That is a serious result — but not the end of the process.

On the Clay Mathematics Institute’s website, the problem remains listed as unsolved. Its rules require a proof to be published and survive two years of scrutiny before the $1 million prize is awarded — and some mathematicians will still want to examine the ideas behind the proof, not just the machine’s verdict.

And Alpöge is not the only claimant. Hours after his announcement, Stanford professor Ravi Vakil said a postdoctoral researcher in his department had independently submitted a proof. Two proofs, almost at once, after 167 years of silence.

Anthropic itself has stayed quiet: the company declined to comment on Friday and did not respond to the Associated Press. Alpöge said he had consulted chief executive Dario Amodei before announcing, and acknowledged the timing could be viewed cynically — the announcement came the day after Anthropic’s shares began trading. He did not elaborate.

The second one in three months

Step back, and the pattern is what startles mathematicians. In 26 years, humans alone solved one Millennium Prize Problem. In three months, AI appears to have done it twice: in September, an OpenAI internal model solved the first, the Navier-Stokes problem, according to the Journal. Now Riemann.

Alex Kontorovich, a mathematician at Rutgers University, told CNN the moment felt unreal — as if he were watching himself say words he could barely believe: that the Riemann Hypothesis had been solved.

The shift is not only about one problem. In August, Anthropic said an unreleased Claude had improved a related result — stressing it had not proved it. Three months later, the proof arrived. The tools are moving fast enough that the field is being forced to rethink what a proof even is: a human argument, a machine certificate, or both.

Why this matters beyond mathematics

For Anthropic, the timing is extraordinary. The company priced its IPO at $83 a share and began trading on November 19 in the largest listing in history — then gave back much of the debut gain, closing Friday at $88.91 as the wider AI sell-off dragged markets down. The Riemann announcement landed in the middle of that storm, and it reframes the company in a way no earnings report could.

For everyone else, the lesson is narrower and stranger: a machine helped prove something no human could, and the human who guided it cannot yet explain every step. That is a new kind of knowledge — verified by computer, not yet fully understood by people. Whether the Clay Institute ultimately agrees, and whether the $1 million prize follows, will take time. The proof, for now, checks out.

Sources: The Wall Street Journal, CNN, USA Today, Phys.org, Le Monde — November 20–22, 2026.

Educational Content Only — Not Financial Advice: MakTimes is a news and education website. This article is published for general information and awareness only and is not financial, investment, trading, or legal advice. We are not financial advisors. Cryptocurrency, gold, and other markets are volatile and can result in loss. Always do your own research (DYOR) and consult a licensed financial advisor before making any financial decision. Past performance is not a guarantee of future results.

Leave a Reply

Your email address will not be published. Required fields are marked *