Newsletter Subscribe
Enter your email address below and subscribe to our newsletter
[forminator_form id="25163"]

math.mit.edu+1.techcrunch.com+1.scientificamerican.com+1.Independent researchers have compiled Levent Alpöge's computer-checkable proof of the Riemann hypothesis and found no errors. One last audit was still pending as of Saturday, November 21, 2026. Scientific American reported that MIT mathematician Andrew Sutherland and others rebuilt the Lean code "without issues." Sutherland told the magazine, "Honestly the chance that anything is fishy here is approximately zero."scientificamerican.com
The checks came hours after Alpöge, a mathematician at Anthropic, said on Friday, November 20, that he had used the company's Claude model to prove the 167-year-old conjecture about how prime numbers are distributed. He did not release a conventional paper. Instead he published the proof in Lean, a language that checks proofs by computer. New Scientist reports the Lean version runs to 17 million lines.scientificamerican.com
Sutherland's verification page at MIT reports that all 175,187 modules compiled with zero errors. It also says the five main theorems rely only on Lean's three standard axioms, and that independently written checkers, including nanoda and con-leche, accepted the full exported proof. The page says the headline theorem uses the Riemann hypothesis exactly as Mathlib, Lean's standard math library, defines it. "The Lean proof has been compiled and independently rechecked by the Lean kernel with only the three standard axioms," Sutherland told Scientific American. TechCrunch reported that several people in the Lean community's online chat said they had also compiled the solution.techcrunch.com+2
Kevin Buzzard of Imperial College London said a "comparator run" is still needed. The test confirms that the code does not "redefine the meaning of any of the symbols in the statement." In paranoid mode, it also checks that the proof compiles under several independent kernels, so it cannot be exploiting an unknown bug in Lean's official kernel. At his last update, Sutherland listed both the end-to-end comparator run and the Lean paranoid-kernel replay as still running, though one comparator run had already passed its statement comparison and axiom check.math.mit.edu+1
Passing the machine checks would not tell mathematicians why the proof works. Oxford mathematician James Maynard said, "I suspect that the proof will turn out to be correct," but he expects it to take a long time to digest. Number theorist Andrew Granville called Alpöge's four-page summary "garbled". Mathematician Thomas Bloom urged caution. "In proportion to the magnitude of this claim, we should be appropriately sceptical," he told New Scientist.nature.com