AI

OpenAI Reports Breakthroughs in Mathematics, Computer Science and Physics

Models challenge foundational theoretical limits and establish new proofs under peer evaluation by external specialists.

By Letizia Giugliano
2 min readHuman-Edited, AI-Assisted

ShareLinkedInFacebookX

OpenAI Reports Breakthroughs in Mathematics, Computer Science and Physics
Photo: Erwan Hesry / Unsplash

OpenAI announced a series of automated discoveries in mathematics, computer science and theoretical physics on October 27, 2026. The releases expand upon earlier mathematical papers, introducing model-generated proofs that challenge long-standing foundational assumptions across scientific fields.

New manuscripts uploaded to GitHub

According to reporting by Engadget, the research follows an earlier batch of manuscripts detailing over 700 pages of solutions and progress for 372 major math problems. In this new phase of artificial intelligence development and scientific innovation, OpenAI published two distinct technical manuscripts directly to GitHub for researcher review.

The first manuscript addresses a long-standing conjecture regarding wave equations, establishing Sogge's local smoothing for waves across every spatial dimension. The second document tackles the Exponential Time Hypothesis (ETH), presenting a proof that the computational assumption is false.

Challenging the Exponential Time Hypothesis

The Exponential Time Hypothesis has served as a computational premise stating that certain computational problems cannot be resolved quickly. As Engadget notes, ETH has anchored hundreds of peer-reviewed papers spanning computational biology, database theory and geometric computation.

Disproving ETH suggests that researchers could design significantly more efficient computational algorithms across those disciplines. Demonstrating an unprecedented approach to technology transfer in theoretical computer science, the original scientists who formulated the hypothesis have joined the independent review group evaluating the outputs generated by the models.

Theoretical physics and external validation protocols

Beyond pure computation, the release encompasses theoretical physics. The AI model produced a reported proof of interacting quantum field theories incorporating gauge fields and fermions in four dimensions. The specialized physics group evaluating these findings has established the contact point YANGMILLS4D[@]gmail.com while preparing their technical assessment, as detailed by Engadget.

To manage these discoveries, OpenAI implemented three distinct evaluation models: public publication for math, concurrent output release alongside external writeups for computer science, and fully deferred publication managed by the physics team. OpenAI stated it will claim no academic credit or intellectual property rights for the discoveries, leaving attribution to the external contributors while ensuring proper citation of previous academic work.

Read also on The Patent: Rare Quantum Lattice Assembled from Propeller-Shaped Organic Molecules