Riemann proof passes multiple independent Lean kernel checks; humans can't read it

churchkey · x · 2026-11-22

Update on the Riemann Hypothesis proof: MIT's Andrew Sutherland exported Alpöge and Claude's full proof and ran it through independently written kernels—nanoda (Rust) accepted it in 28 minutes; con-leche confirmed all 1.47M declarations and the five theorems (including RH itself) match Mathlib-only statements; lean4lean passed all 175,187 modules; Lean's own kernel replay passed with zero rejections. A Lean bug is essentially ruled out since every checker would have to share it.

The quoted author's take: the wildest part is the Clay Institute wanting a two-year human review for a 1.7-million-line pull request nobody on Earth can actually read—"we're just waiting for a compiler." No readable proof exists yet.

Original post →

More from Fun

Fun channel →