OpenAI announced three major scientific discoveries in math, computer science, and physics on October 26, 2026, including resolving the Kakeya conjecture in all dimensions and disproving the Exponential Time Hypothesis (ETH), while collaborating with academic experts to validate results.
Mathematical Breakthroughs: From Conjectures to Proofs
In computer science, OpenAI claims to have disproven the Exponential Time Hypothesis (ETH), a long-standing assumption about computational complexity. The AI’s natural language disproof of ETH was formalized in Lean on October 17, 2026, and released alongside a public GitHub repository. Syncai.news noted that the company added OpenAI results to GitHub, including Local smoothing for the Euclidean wave equation, which uses an auxiliary par… Read the full story at The Verge.
New Collaboration Model: Academic “Shepherds” and Open Review
OpenAI adopted a novel approach by involving academic “shepherds” to oversee the interpretation of its findings. These experts, chosen for their expertise, will organize groups to understand and validate the results. For the computer science and physics discoveries, the company recommended releasing raw AI outputs for independent review. Hindustantimes.com noted that the computer science research will be led by a group with UC San Diego’s Russell Impagliazzo and Ramamohan Paturi at the helm. In mathematics, researchers were advised to work through explanations themselves, reflecting a tailored strategy for each field.
The initiative aligns with OpenAI’s $5 million contribution to the Simons Foundation, aimed at supporting workshops and programs to help researchers build on AI-generated results.
Implications for AI-Driven Research
The announcements highlight AI’s growing role in advancing scientific knowledge, particularly in fields where traditional methods face limitations. By engaging academics as “shepherds,” OpenAI aims to bridge the gap between AI-generated research and peer validation. However, questions remain about the long-term impact of AI on scientific collaboration, verification processes, and credit distribution. The company’s approach could set a precedent for future AI-driven discoveries, though its effectiveness will depend on the academic community’s engagement and feedback.