AI Helps Solve a 30-Year Optimization Problem
pkerger · reddit · 2026-07-17
The author claims that by using a prompting method mimicking OpenAI CDC's proof project, they got a model to provide a solution during a single 148-minute session, which has been formally verified using Lean.
Main Content
- The research subject is a 30-year open problem in convex optimization/complexity theory.
- The author, a UC Berkeley IEOR professor, emphasizes the results have not yet been peer-reviewed.
- They believe this prompt + model collaboration approach holds greater potential for open problems in theoretical computer science and mathematics.
Additional Information
- The author provided a Medium article, GitHub repository, Lean code, full prompts, and build instructions.
- They also mentioned failing in previous attempts with earlier model versions, but succeeding "on the first try" with the new version.
Related event: GPT-5.6 Solves Long-Standing Math Problems(4 posts)→
More from Research
- AI folks, stop saying "biology" when you mean drug development — KevinKaichuang · 2026-10-26
- ostris Pretrains 'Primal VAE': Aggressive 2x-Downsample, 1-Channel Latent Design — ostrisai · 2026-10-26
- ostris pre-training "Primal VAE": 1-channel latents match modern 8x VAEs via patching — ostrisai · 2026-10-26
- Shanghai AI Lab-led team releases 100-hour egocentric wearable dataset for humanoid training — ChuckDBrooks · 2026-10-26
- ConvFill: A 0.6B Talker Buys Time While a Large Reasoner Whispers the Real Answer — andimarafioti · 2026-10-26
- Flow Reasoning Models turn flow matching into iterative reasoners, hitting 99.5% on Sudoku-Extreme — alec_helbling · 2026-10-26