OpenAI Unveils Three Major Scientific Breakthroughs With New Academic Collaboration Model

OpenAI has published three significant AI-generated scientific discoveries spanning mathematics, theoretical computer science, and particle physics. Unlike its previous mass manuscript drop, the company now employs a structured collaboration model: external expert panels — termed "shepherds" — will guide the examination and explanation of each result before broader dissemination.

Mathematics: Local Smoothing for the Euclidean Wave Equation

The first release presents a proof of Sogge's local smoothing conjecture for waves in all spatial dimensions ≥ 2, using an auxiliary partial differential equation. The derivation also yields the Kakeya maximal and dimension conjectures and Stein's spherical restriction estimates across the conjectured range. OpenAI is releasing the raw model output directly, inviting the mathematical community to develop their own explanations.

Computer Science: Refuting the Exponential Time Hypothesis

The second paper claims a deterministic algorithm that exactly counts satisfying assignments of fixed-width CNF formulas in subexponential time and polynomial space — specifically refuting the Exponential Time Hypothesis (ETH) for 3-SAT. A verification working group led by Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams will independently examine the proof. Materials are public immediately at the group's request.

Physics: Rigorous Existence Proof for 4D Quantum Field Theories

The third result, not yet fully released, is described as a mathematically rigorous existence proof for interacting quantum field theories with gauge fields and fermions in four dimensions — including Quantum Chromodynamics (QCD) and a unified completion of the Standard Model. Nobel laureate David Gross will lead a shepherding group including Michael Douglas, Arthur Jaffe, Igor Klebanov, and Nikita Nekrasov.

Why it matters for GPU / AI infrastructure: These breakthroughs demonstrate frontier AI models tackling problems that have resisted human experts for decades. Training and running such systems requires massive, sustained GPU clusters with high-bandwidth interconnects — precisely the infrastructure AiGpu provisions for research labs and enterprises pushing the boundaries of AI-driven scientific discovery.