AI & Crypto

Anthropic mathematician claims Claude helped crack Riemann hypothesis

Mathematician Levent Alpoge claims the Claude AI helped him solve the 167-year-old Riemann hypothesis. He released a 14-part archive of the proof encoded in Lean following Anthropic's $81.34 billion Nasdaq debut, though the solution requires years of scrutiny for the $1 million prize.

Anthropic mathematician claims Claude helped crack Riemann hypothesis

Timing of the announcement

Levent Alpoge, a mathematician at Anthropic, says Claude helped him crack the Riemann hypothesis and a broader generalization. He released the claim on Friday, one day after Anthropic began trading on Nasdaq following an $81.34 billion offering. Alpoge messaged CEO Dario Amodei about the discovery late Sunday night, November 15. He waited until the company finished its market debut on Thursday, November 19, to release the news to the public. I find the timing of this announcement, arriving immediately after a massive stock-market debut, to be extremely poorly handled. Alpoge admitted, "The precise timing is horrible," though he did not say the IPO caused the delay. He encoded the argument in Lean, which is software used to check mathematical proofs step by step, and he provided a 14-part downloadable archive of the proof for researchers to examine.

Mathematical verification and competition

The Riemann hypothesis is a 167-year-old puzzle that predicts where certain zeros of the Riemann zeta function fall. This problem relates to the distribution of prime numbers and carries a $1 million prize from the Clay Mathematics Institute. Verification remains difficult because the mathematical community must reproduce the archive, check that the formalized theorem matches the original conjecture, and examine the results before the $1 million award can be considered. The Clay Mathematics Institute requires that a proposed solution survive at least two years of scrutiny in a qualifying outlet. This claim follows previous AI math disputes, such as the September conflict where OpenAI claimed it solved the Navier-Stokes Millennium Prize problem while Alpoge and NYU mathematician Tristan Buckmaster pursued related fluid-equation work. In October, OpenAI followed with 722 manuscripts grouped into 372 families of results, though only 162 of those included Lean-checked main results. Mathematicians demand stronger evidence than these AI-generated claims, and some fear a "mathocalypse" where AI solves problems before researchers can learn from them. Daniel Litt warned that, without adaptation, there is just no more math in 50 years. Tristan Buckmaster argued at an MIT panel this week that mathematics still requires human understanding, rejecting the idea that AI can replace the necessity of human thought. Alpoge’s assertion goes far beyond earlier work from Anthropic regarding the zeta function.