700 manuscripts, 48 hours, three withdrawals. The verifier won.

OpenAI recently released a massive repository containing over 700 mathematical proofs generated by an internal AI model, promising to provide formalizations in Lean over time. The release sparked immediate debate regarding academic norms and the role of AI in scholarship. Within 48 hours, three of the results were withdrawn, highlighting the effectiveness of public, mechanical verification. The author argues that this incident demonstrates a positive shift in AI development: rather than relying solely on corporate press releases, the community can now use open repositories to perform independent audits. By shipping the verification logic alongside the AI-generated artifacts, OpenAI inadvertently created a system where errors are surfaced publicly through commit logs. This event serves as a case study for the industry, suggesting that transparency and mechanical reproducibility are essential for building trust in AI-generated outputs, moving beyond mere marketing claims to verifiable, tamper-evident results.