research-norms

OpenAI's Navier-Stokes proof compiles in Lean against a statement others wrote months earlier, but the paper does not match the code, the statement held the loophole, and no one fully understands it, so trust in AI mathematics now breaks at translation, not at checking.