RESEARCH · RESEARCH · #4023
Anthropic's Claude claims a Lean-verified proof of the Generalized Riemann Hypothesis; Stanford posts independent human draft
Anthropic mathematician Levent Alpöge announced a claimed proof of the Generalized Riemann Hypothesis produced with the AI model Claude and published as machine-checkable code in the Lean theorem prover; hours later, Stanford mathematician Alessandro Masullo released an independent, all-human draft of a proof that had been reviewed by Brian Conrad. Both claims are under rapid scrutiny—Alpöge's proof is Lean-encoded and thus formally checkable if the formalization matches the intended statement, while Stanford's draft was rushed public and several issues have already been reported.
KEY POINTS
- Anthropic mathematician Levent Alpöge announced a claimed proof of the Generalized Riemann Hypothesis produced with the AI model Claude and published as machine-checkable code in the Lean theorem prover; hours later, Stanford mathematician Alessandro Masullo released an independent, all-human draft of a proof that had been reviewed by Brian Conrad.
- Both claims are under rapid scrutiny—Alpöge's proof is Lean-encoded and thus formally checkable if the formalization matches the intended statement, while Stanford's draft was rushed public and several issues have already been reported.
- If correct, a verified proof of GRH would be a landmark mathematical result and demonstrate AI's capacity to contribute to and accelerate deep theoretical research, but current claims remain unconfirmed and are actively being checked.
WHY IT MATTERS
If correct, a verified proof of GRH would be a landmark mathematical result and demonstrate AI's capacity to contribute to and accelerate deep theoretical research, but current claims remain unconfirmed and are actively being checked.