ToolNavs Find Useful AI Tools
Submit Sign in
Back to AI information
Riemann Hypothesis Proof Formally Confirmed by Lean: All 17.37 Million Lines of Code Verified

Riemann Hypothesis Proof Formally Confirmed by Lean: All 17.37 Million Lines of Code Verified

AI information • Admin • • 16 views

The proof of the Riemann hypothesis now has a formal verdict. On November 22, 2026, the Lean and Mathlib community issued a statement confirming that the proof by Levent Alpöge, a mathematician at Anthropic, is correct, and congratulating him on the achievement. The statement closes two days of suspense: the proof published on November 20 is no longer merely "claimed" — it is a conclusion backed by independent machine verification.

17.37 million lines of code, independently run to the same result by dozens of people

Alpöge published the proof on November 20: working with Claude, he found the line of argument and released it as 17.37 million lines of Lean code spread across 175,187 files. Lean is a tool built to check mathematical proofs for errors, and its verification programs are designed so that they provably cannot accept a flawed proof. After the statement was published, dozens of mathematicians and verifiers ran the code independently and arrived at the same result. Andrew Sutherland, a mathematician at MIT, has collected these independent checks on his personal website: all modules compile with zero errors, the five headline theorems depend only on the three standard axioms, and two independently written verification kernels also accept the full proof.

The same day brought another proof into public view: the "purely human proof draft" released overnight by Stanford comes from independent mathematician Alessandro Masullo and Brian Conrad of Stanford University. Masullo spent nearly 12 years writing the 940-page manuscript without using AI, and had planned another year of polishing; news of the Claude proof forced them to publish the unfinished draft early. That human proof is still under review, while Alpöge's proof has already secured machine verification.

Earlier than anyone expected — and it changes what "proof" means

The Riemann hypothesis is one of the most famous unsolved problems in mathematics, describing how prime numbers are distributed. Its resolution arrived far earlier than forecast: forecasters on the prediction platform Metaculus had expected it only by May 2035, and mathematician Terence Tao said just last month that the problem would likely fall only around 2050, through a combination of humans and AI.

But the confirmation has its limits. No human being can read 17.37 million lines of code in full, so trust rests on machine verification rather than human review. In other words, the meaning of "proof" is changing: machine-readable and re-checkable is becoming more important than human-readable. The Lean and Mathlib community's statement is also not the same as journal publication, and mathematicians are candid that truly understanding the mathematics in the proof will take time.

Three things are worth watching next: when journal peer review will be completed, whether the proof can be compressed into a human-readable version, and whether formal verification will become a standard step before major mathematical results are published. Whichever moves first, it will be the next chapter of this mathematical event.

Recommended Tools

More