AI's Riemann hypothesis proof independently checked within a day: "the chance that anything is fishy is approximately zero"
First seen on X 13 hours ago@superalesha ♥ 3,48822 views
"The Lean proof has been compiled and independently rechecked by the Lean kernel with only the three standard axioms. Honestly the chance that anything is fishy here is approximately zero." That was MIT mathematician Andrew Sutherland, speaking to Scientific American on Nov. 20, less than a day after Anthropic mathematician Levent Alpöge claimed he and Claude had proved the Riemann hypothesis.
According to Scientific American, Sutherland and other independent researchers each compiled Alpöge's Lean code without errors. Lean is a language in which a computer checks every logical step of a proof, and an error-free compile is usually taken as confirmation. The magazine says mathematicians are now "almost certain" the proof is real.
One step remains. Kevin Buzzard of Imperial College London wrote on the Lean community chat that a "comparator run" is still pending. It checks that the code does not quietly redefine any symbol in the statement and that it compiles under several independent kernels, and typically takes hours. On his blog he noted that the first question is whether the theorem stated in the Lean code matches the actual theorem. There is still no human-readable paper: Alpöge says a write-up would exceed 250,000 pages, and an X post saying the Lean proof is comparable in size to all human-written formal math drew 960 likes.
Alpöge made a second announcement the same day. He and NYU professor Tristan Buckmaster say they have, with AI's help, solved the "real" Navier-Stokes problem. OpenAI's September solution relied on a finely tuned stirring force that many experts considered a loophole; the new claim is blowup without that force. Buckmaster posted an unpolished write-up, saying that "after last time" they "couldn't risk waiting." The certificate for that result has not been released.
Skeptics were loud too. A post arguing that Alpöge himself called the written-out version unreadable "slop," and that the news landed the morning Anthropic's stock slid toward its IPO price, got 3,500 likes. Anthropic shares fell 9.1% that day. "Claude solved the Riemann hypothesis and $ANT stock is down 10%," Nabeel Qureshi wrote, concluding the market sees AI as "normal technology." The hypothesis, posed by Bernhard Riemann in 1859, concerns the distribution of primes and carries a $1 million Clay prize.
Comments 0
Select a sentence in the story (long-press on phones) to comment on it.
“Everyone thinks Claude just solved the 167-year-old Riemann Hypothesis, until you actually open the PDF.”
