GPT‑5.2 Pro discovers new gluon amplitude with AI-assisted proof

What's the deal? OpenAI reported that its GPT-5.2 Pro model derived a new formula for gluon amplitudes previously thought impossible and helped produce the formal proof confirming the result. Human researchers at OpenAI and partner institutions verified and validated the discovery.

Why now? Traditional Feynman diagram calculations are extremely complex for these amplitudes. GPT-5.2 Pro simplified the expressions and spotted general patterns, producing a formula valid for any number of gluons.

What could go wrong? The amplitudes occur under very specific momentum conditions, limiting immediate practical applications. The growing use of AI in theoretical physics also requires rigorous validation and reproducibility standards.

The signal: The result highlights AI's potential to generate fundamentally new scientific knowledge. The method is already being extended to gravitons and other interactions, suggesting a future where models like GPT-5.2 Pro can act as co-authors of theoretical breakthroughs.

In a September 8, 2026 post, OpenAI said a group of agents had produced a solution to the Navier–Stokes Millennium Prize Problem using a next-generation model more capable than GPT-6 Astra. The announcement prompted criticism from mathematician Mehdi (e/λ), who argued that formalising a blow-up case under controlled external forcing with a large agent and GPU setup should not be presented as solving the full problem. The exchange highlights the distinction between proving a constrained result and resolving the complete mathematical formulation, as well as the need for independent expert scrutiny of AI-assisted proofs.

Read more: OpenAI · AI STAT · OpenAI on X · Mehdi (e/λ) on X