█

AI/TLDR

Shtetl-Optimized · 2026-11-20 · notable

Scott Aaronson — 'The stochastic parrots have apparently proved' RH

In short

Scott Aaronson reacts to Levent Alpöge's claim that Anthropic models helped prove the Generalized Riemann Hypothesis in 17.4 million lines of Lean. He bets the proof will stand, but warns Lean's soundness gaps must be fixed.

Shtetl-Optimized · 2026-11-20 · 4 sources checked

Shtetl-Optimized blog share card

A complexity theorist bets the AI-made Riemann proof will stand, and names P vs. NP as the 'Final Boss' left.

What is it?

"The stochastic parrots have apparently proved the Riemann Hypothesis" is a November 20, 2026 post on Scott Aaronson's blog Shtetl-Optimized. Aaronson responds to Levent Alpöge's statement that Anthropic models helped him prove the Generalized Riemann Hypothesis, with a Lean formalization of 17.4 million lines. He has not checked the proof himself.

How does it work?

The post places the claim next to earlier AI work showing the zeta function's nontrivial zeros lie at most 3/8 away from the critical line, which Aaronson says already carried much of the Riemann Hypothesis's number-theory payoff. He then weighs the Lean certificate: Lean has had kernel bugs that allowed proofs of 0=1, and he worries that hard-optimized AI could one day exploit such gaps, though he sees "zero indications" of that now.

Why does it matter?

Aaronson writes "my bet is that the proof will stand", and notes that no AI result with a Lean certificate has been retracted so far, while several without one have. He argues that human mathematicians will still need years to understand and rebuild the ideas, and that P vs. NP now becomes the "Final Boss" of mathematics people care about.

Who is it for?

mathematicians, theoretical computer scientists, formal-verification researchers

Sources · 4 outlets

Tags

  • scott-aaronson
  • anthropic
  • claude
  • riemann-hypothesis
  • mathematics
  • lean
  • formal-verification
  • ai-for-science

← All releases · Learn AI