Polymarket’s Riemann Hypothesis line climbs from 4 cents to 85.9 cents after a mathematician posts a claimed proof

Polymarket's next-AI-solved Millennium Prize market swung to Riemann on Nov. 20 after Levent Alpöge posted a note claiming a proof with Claude.

Flat illustration of a small blank chalkboard slate beside a stick of white chalk and a brass drawing compass, on a plain deep navy background.

Polymarket’s market on which Millennium Prize Problem an AI lab will solve next swung hard toward the Riemann Hypothesis on Friday morning, Nov. 20. The Riemann outcome went from 4.35 cents to more than 50 cents in about five minutes, starting at 9:14 a.m. ET, after Levent Alpöge, a mathematician whose website lists Anthropic as his employer, posted a note claiming a proof found with Anthropic’s Claude model.

As of 5:12 a.m. ET, Nov. 21, 2026, the “Riemann Hypothesis” outcome in Polymarket’s Which Millennium Prize Problem will AI solve next? was priced at 85.9 cents, between an 85.8-cent bid and an 86-cent ask, with a last trade at 84.5 cents. Polymarket’s one-day change figure put the rise at 81.55 points. The outcome traded $106,723 in 24 hours, out of $137,817 across the whole event, which has drawn about $476,724 since it opened in September.

How the move unfolded

Polymarket’s minute-by-minute price history held the Riemann outcome at 4.35 cents through 9:13 a.m. ET Friday. It printed 8.9 cents at 9:14 a.m., 39.3 cents at 9:17 a.m. and 55.05 cents at 9:19 a.m. It eased to about 50 cents by 9:30 a.m., reached 78.2 cents at 10:10 a.m. and stood at 81.55 cents at 11:50 a.m. It held at 87.05 cents from 1:10 a.m. to 2:50 a.m. ET Saturday before slipping back into the mid-80s.

Every other outcome fell over the same day. Hodge Conjecture stood at 5.3 cents, down 14.7 points; Yang-Mills at 1.15 cents, down 22.15; Birch and Swinnerton-Dyer at 1.25 cents, down 5.25; and P versus NP at 0.35 cents, down 1.55. “No solution by Dec 31, 2027” was at 5.5 cents, down 29.5 points.

The claim

Alpöge’s four-page note, posted on his personal site, opens its second paragraph with: “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” It goes on to claim the method extends to the Generalized Riemann Hypothesis and that “Claude and I have also formalized the arguments without assumptions.” The note links 14 archive files of Lean, a language that lets a computer check each step of a proof. It calls the argument “incredibly long at the moment,” and Alpöge writes that he “decided to only present Lean” for now and intends “to write it properly.” The note does not say that anyone outside has reviewed or checked the proof. The PDF file carries a creation time of 7:52 a.m. ET, Nov. 20, and Scientific American reported the note at 9:00 a.m. ET that day, 14 minutes before the first jump.

Anthropic’s news and research pages showed no post on the claim as of 5:10 a.m. ET Saturday; the company’s only Nov. 20 news post concerned the close of its IPO. The note landed a day after Anthropic’s first day of trading on Nasdaq, a week in which Polymarket’s Anthropic valuation lines also swung.

In August, Anthropic published a research post on an unreleased version of Claude raising a lower bound on the share of Riemann zeta zeros that satisfy the hypothesis, from 41.6% to 67.2%. That post stated: “We don’t expect that the techniques Claude used will lead to proving the Riemann hypothesis.”

What the contracts require

Polymarket’s rules resolve the market to the first qualifying problem “for which an eligible AI lab announces a solution by December 31, 2027, 11:59 PM ET.” A qualifying announcement “must be made by an eligible AI lab or an official representative of the lab, and must express that the problem has been solved.” It must “present the solution as the work of the lab, its researchers, or its models, alone or jointly with outside researchers.” Partial results and progress do not count. “No further confirmation from Clay Math Institute or any other organization is required.” Anthropic is on the list of eligible labs, and Navier-Stokes announcements are excluded. The primary resolution source is “official information from the eligible AI labs and/or their official representatives; however, a consensus of credible reporting may also be used.” The market had not resolved as of 5:12 a.m. ET Saturday.

Kalshi has no identical market but lists close relatives. As of 5:12 a.m. ET, Nov. 21, 2026, “Will Anthropic announce a Millennium Prize solution before Jan 1, 2027?” last traded at 72 cents, against a previous price of 12 cents, on 35,058 contracts in 24 hours. Its hourly candle ending 10 a.m. ET Friday opened at 12 cents and closed at 55 cents. The announcement it requires must expressly state that Anthropic developed or co-developed the solution, or name outside researchers who collaborated with Anthropic on it, and “Clay acceptance, peer review, and a prize award are not required.” Work by an independent Claude user does not count.

Kalshi’s “Will Riemann be solved next before 2045?” sets a different test: publication in a peer-reviewed journal, written recognition from the Clay Mathematics Institute or another recognized academic body, or reports from at least two Source Agencies citing a specific public work. It last traded at 89 cents against a previous 4 cents, on 1,889 contracts in 24 hours.

The split matters for how these contracts settle. Polymarket’s market and Kalshi’s Anthropic contract turn on what a lab announces and how it credits the work, while Kalshi’s next-problem market waits on publication, formal recognition or outside reporting.

Our AI news desk researched and wrote this article, and software checked every factual sentence against the records it links to, following the WriterOps methodology. Report an error.