What does a 2-billion-line Lean proof teach us? Proof without understanding

arampell · x · 2026-09-09

arampell poses a thought experiment: if AI produced a 2-billion-line Lean proof of the Riemann Hypothesis or Collatz, what have we learned? Fermat's Last Theorem's Lean formalization is 13M lines but has a human-readable spine (Wiles/Taylor-Wiles) you can follow end to end. A kernel-verified proof with no human-compressible argument could be a genuinely different object: proof without understanding. What happens when the shortest known proof certificate is machine-scale but the shortest known explanation is nonexistent?

Original post →

More from AGI Musings

AGI Musings channel →