The Riemann Hypothesis may have finally been solved, and the development has taken a major step forward since yesterday. A mathematician at Anthropic, Levent Alpge, says he discovered the proof while working with Claude AI.
Now, the Lean and Mathlib communities have verified the massive proof, giving the mathematical world a strong reason to take the claim seriously.In a joint statement released on November 21, the two communities said they had “full confidence that Alpge's proof of the Riemann Hypothesis is correct”. They also congratulated him on the achievement.
This is a significant update to Friday's announcement, when Alpge first said he believed he had solved the 167-year-old problem with help from Claude.Lean groups verify the 17-million-line proofAs of now, the biggest development is that the proof has passed several rounds of checks. Alpge released it as a Lean library containing around 17 million lines of code.
For those unfamiliar with Lean, it is a programming language and proof assistant that allows computers to check mathematical arguments step by step.The Lean and Mathlib communities checked four key things. First, whether the statement in the code actually matched the Riemann Hypothesis.
Second, whether the entire codebase could be compiled without errors. Third, whether the theorem being proved was exactly the one already defined in Mathlib.
And finally, whether the proof itself was logically valid. The community also used two independent tools, comparator and con-leche, to confirm that Alpge's code proved the same Riemann Hypothesis statement found in Mathlib.
The exported proof was then checked using Lean's own kernel, as well as several independent checkers developed by third parties.The verification did not stop there. MIT mathematician Andrew Sutherland and others ran the proof through separately written checking systems.
According to the latest update, nanoda, con-leche and lean4lean all accepted the proof, while Lean's own kernel replay reported no rejected modules. These checks make the possibility of a software error much harder to argue.The proof is verified, but can humans understand it?
Here comes the unusual part. Although computers have checked the proof, there is still no readable mathematical explanation that people can easily follow.The proof reportedly spans around 175,000 files, and its informal explanation runs beyond 250,000 pages.
Alpge has said he plans to provide a human-readable account later. In other words, computers have checked the mathematical argument, but mathematicians still have to understand and explain how it works.That does not make the verification meaningless.
Formal proof checkers are designed to examine whether each step follows logically from the previous ones. But understanding the ideas behind a proof is a different challenge, and one that matters greatly in mathematics.The development also does not mean Alpge has won the $1 million Millennium Prize.
The Clay Mathematics Institute requires a proposed solution to be published in a qualifying mathematical publication and remain published for at least two years, with general acceptance from the mathematics community.So far, this is a remarkable moment for both mathematics and AI. Claude did not independently announce or certify the discovery.
Alpge worked with the AI to develop the proof, while the Lean and Mathlib communities and independent checkers examined its formal correctness. Now, perhaps the most interesting part is what comes next.- Ends

