Published Friday, November 20, 2026. This is a developing story — neither proof has been verified yet, and this article will be updated as the checks come in.

An Anthropic mathematician says he and Claude have proved the Riemann hypothesis — the most famous unsolved problem in mathematics, unproved for 167 years and carrying a $1 million prize. He posted the claim on Friday morning, November 20, 2026, with code he says contains the full proof. Hours later, a second, human proof landed from Stanford. Neither has been checked yet.

The hypothesis is a claim about prime numbers — 2, 3, 5, 7, 11 and so on — which look random as you walk up the number line. In 1859, Bernhard Riemann proposed that their spacing is secretly controlled by the Riemann zeta function, and specifically by where that function equals zero. His guess: every interesting zero sits on a single vertical line. Computers have checked the first ten trillion zeros — all on the line. Nobody has proved it for all of them.

“I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” — Levent Alpöge, mathematician at Anthropic

What Was Actually Posted

Alpöge, a number theorist at Anthropic, posted a four-page note on his personal website saying he believes he has solved the problem — and that the same method also settles the Generalized Riemann Hypothesis, a much larger version of it. With it he released the proof written in Lean, software that checks each logical step, split across a 14-part archive. There is no preprint, no journal submission, and no independent verification. The full written argument ran past 250,000 pages — he called it “slop” and did not publish it.

He is unusually candid: he says he is presenting a complete argument before having mastered all the mathematics around it, and that Claude moved faster than he could follow. He directed the approach throughout, but credits the model with pulling ahead near the finish — including a central step built on a long chain of transformations called theta lifts.

The Second Proof That Landed Hours Later

While number theorists were still downloading Alpöge's Lean files, Ravi Vakil — a Stanford geometer and president of the American Mathematical Society — posted a draft by Alessandro Masullo, a Stanford postdoc, claiming the Generalized Riemann Hypothesis by an entirely human route. The draft, titled “The Music of the Primes,” is twelve years of work. No AI was involved at any stage.

Masullo's draft has one advantage Alpöge's does not: time under review. Brian Conrad, one of the field's most respected and most skeptical number theorists, has been checking the argument closely for months, and Vakil says it has held up so far. The plan was a year of quiet polishing before any announcement — Alpöge's post made that impossible, so Vakil released the draft as it stands and asked the community to check it, formalize it, and patch any small gaps.

The two claims cover the same ground on paper: the Generalized Riemann Hypothesis implies the original. But they are not the same proof.

Why Mathematicians Are Taking It Seriously — And Why Nobody Is Celebrating Yet

The reaction has been a strange mix of awe and caution. Andrew Sutherland of MIT independently compiled Alpöge's result in Lean and called it “an historic moment,” adding that “there is much work to be done to understand the mathematical details of the proof.” Hector Pasten of the Pontifical Catholic University of Chile said that if it is correct, it completes “a dictionary between different branches of math where the connections were only partial — something like a Rosetta Stone.”

But a Lean proof has a catch: the compiler checks that each step follows — it does not check that the theorem you proved is the theorem you meant. The verification comes in three stages. First, does the code build? Anyone can download the 14 files and ask Lean to check them. Kevin Buzzard of Imperial College London put it practically: “Let's see if it compiles first.” Second, and decisive: does the final statement actually match the standard Riemann hypothesis? That needs human number theorists, not a compiler. Third, can a human understand it? Until the second check is done, calling the problem “solved” is ahead of the evidence.

THE STORY SO FAR

1859 — Bernhard Riemann publishes the hypothesis in a paper on the distribution of prime numbers. He dies seven years later, at 39, with the problem still open.

2000 — The Clay Mathematics Institute names it one of seven Millennium Prize Problems, each carrying a $1 million bounty.

2003 — Grigori Perelman solves the Poincaré conjecture, the only Millennium problem solved so far. He declines the prize money.

Summer 2026 — Alpöge tweets a Claude-generated solution to the Jacobian conjecture. Weeks later, an experimental Claude improves a 50-year-old zeta bound, raising the proven share of zeros on the critical line from 41.6 percent to 67.2 percent.

November 19, 2026 — Anthropic's shares begin trading on the Nasdaq in a record-breaking debut, opening 34 percent above the $83 IPO price.

November 20, 2026, morning — Alpöge posts his claim and the Lean code. Hours later, Vakil releases Masullo's draft.

WHAT HAPPENS NEXT

In the coming days, outside experts will try to build Alpöge's 14 Lean files and check whether the final theorem matches the standard statement. Number theorists are already picking through the code in real time, and the first reports could come within days or weeks.

Even if both check out, the $1 million is years away. Clay's rules require a qualifying publication, a wait of at least two years, and general acceptance in the mathematics community. The earliest the prize could be considered is 2028 — more likely 2029.

Alpöge says his plan is to understand the argument himself, give talks on it, and write it up properly.

WILL IT BREAK YOUR BANK LOGIN?

No — and this is the question almost everyone asks first. The hypothesis is about how primes are spread out; a proof would tighten the error bars on counting them. RSA encryption, the system behind online banking, depends on a different hard task: taking a huge number that is the product of two big primes and recovering them. Nothing in a Riemann proof makes that factoring problem easier. The real threat to RSA is a working quantum computer running Shor's algorithm.