Fields Medalists Sign Open Letter as OpenAI's Math-Proof Race Sparks Attribution Fight

Skynet Chance (+0.05%): The described dynamic — labs spending tens of millions on inference to beat human researchers and publishing unverified proofs — incentivizes secrecy and erodes the external verification culture that safety oversight of advanced AI depends on. Loss of independent human ability to check and understand AI-produced reasoning is a direct precursor to losing meaningful control over AI outputs.

Skynet Date (-1 days): Competitive one-upmanship between frontier labs over prestige benchmarks, plus emerging incentives to hide methods from rivals, accelerates capability racing while degrading transparency. That combination modestly pulls forward the window in which capable systems operate beyond reliable human scrutiny.

AGI Progress (+0.04%): That Fields Medalists feel compelled to respond collectively implies AI systems are now producing proofs of 'important' open problems at a level that materially competes with elite human mathematicians, a strong signal of advancing abstract reasoning. The caveat is that OpenAI's headline proof remains unverified, limiting how much credit the claim deserves.

AGI Date (-1 days): Marathon large-scale inference runs yielding novel mathematical results suggests test-time compute scaling is converting directly into frontier reasoning capability, a lever labs can push harder immediately. Competitive pressure between OpenAI and Anthropic on this axis further compresses expected timelines.