OpenAI says external experts guided two results and advised releasing the math output for researchers to explain.

OpenAI released three science results on GitHub and described a new academic review process. The set covers local smoothing for the Euclidean wave equation, a proof that OpenAI says refutes the Exponential Time Hypothesis, and a physics existence proof not released in full. OpenAI used external experts as shepherds for the computer science and physics work and released the math model output directly.

OpenAI’s October release included hundreds of manuscripts with claimed solutions to long-standing math questions. The new approach draws on named shepherds and an independent computer science group led by Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams. An AGMAI panel of math advisers recommended yesterday that AI companies avoid giving selected people first access to output.

Operators and builders should treat these outputs as research artifacts that need independent validation before they enter products or workflows. The AGMAI panel’s criticism of early access highlights reputational and fairness risks for labs that route results to favored reviewers. OpenAI’s decision to publish model output directly also raises the bar for teams that need reproducible proofs.

An independent computer science group will examine and explain the Exponential Time Hypothesis proof, while the physics information remains unreleased in full. The AGMAI panel wants future AI labs to use scholarly repositories, Lean-formalized proofs, problem-selection details, and funding for researchers who study the results. Builders should watch whether OpenAI adopts that advice or keeps its shepherd model.

What matters

• OpenAI added proofs in math, computer science, and physics to GitHub with a new review process.
• Builders should track the review model because access and proof validation shape trust in AI results.
• Watch the independent computer science group led by Impagliazzo, Paturi, and Williams explain the proof.

Why it matters

Watch the independent computer science group led by Impagliazzo, Paturi, and Williams explain the proof.

This GenAI News article was prepared in original wording using reporting and materials published by The Verge. Source reference: https://www.theverge.com/ai-artificial-intelligence/1019143/openai-eth-kaseya-science-math-physics.

Drafted by the GenAI News review pipeline.