OpenAI announced a proof addressing a long-standing question about the Navier-Stokes equations from fluid dynamics, and released both a conventional human-readable proof and a Lean 4 formal proof alongside it, according to John Cook’s analysis.
The formal proof represents a significant shift in the economics of mathematical verification. Historically, formalizing mathematics has been extraordinarily labor-intensive. In a 2005 estimate, researchers Henk Barendregt and Freek Wiedijk found that formalizing a single page of undergraduate mathematics textbook material required approximately 40 hours of work. By this measure, formalizing OpenAI’s 166-page research paper would theoretically require around 132,800 person-hours if scaled for research-level density.
OpenAI’s approach, using AI to generate the formal proof, completed verification in Lean in just 17 hours. According to Cook’s analysis, this represents a cost reduction of roughly four orders of magnitude—the formalization process consumed approximately $1 to $2 million in compute resources, equivalent to around 10,000 human mathematician hours.
This dramatic efficiency gain extends beyond academic mathematics. Cook notes that formal verification has applications in security policy consistency, smart contract verification, and mission-critical algorithm validation. Previously, the prohibitive cost of formalization made these applications impractical for routine use. The cost collapse now makes it feasible to iterate on specifications and incorporate formal verification into development workflows.
Other recent mathematical results have also been accompanied by Lean 4 formal proofs, suggesting this approach is becoming more common. Cook emphasizes that the shift changes not just cost but also how mathematical work is reviewed and validated. Instead of re-deriving proofs in prose each time, reviewers can now compare checked formal artifacts directly.
Comments on Cook’s analysis note that while the achievement is significant, the practical bottleneck may shift rather than disappear. Formal verification confirms “did I implement what I wrote down?” but does not address “did I write down the right thing?”—the specification gap that requires human mathematical judgment. However, cheaper iteration on proofs makes exploring specifications faster and more feasible than before.
Key facts
- OpenAI released a Lean 4 formal proof alongside its Navier-Stokes proof, verified in 17 hours
- Historical estimates suggested formalizing OpenAI’s 166-page paper would take approximately 132,800 person-hours
- The formalization used roughly $1-2 million in compute, equivalent to about 10,000 mathematician hours
- This represents roughly a 10-fold cost advantage compared to the theoretical person-hour estimate
- Formal verification has potential applications beyond mathematics, including security policy and smart contract validation