Anthropic Says Claude Helped Crack the Riemann Hypothesis
A mathematician at the AI company Anthropic says he used the firm's Claude model to prove a 167-year-old conjecture that has defied the field. Independent verification is still pending.

01What happened
Levent Alpöge, a mathematician at the AI company Anthropic, says he has proved the Riemann hypothesis, a conjecture about the distribution of prime numbers first posed in 1859 and considered one of the hardest open problems in mathematics. It is one of the so-called Millennium Prize problems, a short list of questions carrying a reward for anyone who settles them. Alpöge says he reached the result with the help of Claude, Anthropic's large language model. If the proof holds up, it would be a landmark for both mathematics and for the use of AI as a research partner.
The large caveat is that the claim has not been checked. Proofs of this scale take time for other mathematicians to read line by line, and the verification process remains incomplete. The announcement's timing has drawn scrutiny because it landed just one day after Anthropic's stock market debut, prompting questions about whether the release was meant to coincide with the listing. Until independent experts confirm the work, it stands only as a claim.
02How markets reacted
The debut itself drew real money. Cathie Wood's ARK Invest put about 91 million dollars into Anthropic, buying 934,968 shares across six of its exchange-traded funds on the company's Nasdaq debut. At the same time, the firm trimmed a combined 4.1 million dollars from its Ethereum and Solana staking funds, shifting exposure toward the newly public AI maker.
03What it means · The Griffin Read
An unverified proof claim functions as a capability signal, not science.
The claimed Riemann proof surfaced the day after Anthropic's market debut and was flagged to the CEO the night before. That sequencing makes it read as investor communication as much as mathematics.
Anthropic captures the reputational halo while staying officially silent, and the researcher pre-emptively takes 'unilateral responsibility' for the timing. Upside accrues to the company; downside is outsourced to one individual.
Labs now generate marquee math claims faster than the field can verify them, following OpenAI's Navier-Stokes and mass-Lean pattern. The verification lag itself sustains hype longer than confirmation or refutation would.
Treat this as an unconfirmed narrative lever, not a result. Weight compiling Lean files and named number-theorist sign-off, not the headline or the capital rotation around it.