There’s a new AI that’s officially started dreaming up its own theories about how physics works and then testing them out.
March 25, 2026
Original Paper
The FERMIACC: Agents for Particle Theory
arXiv · 2603.22538
The Takeaway
Unlike standard AI that just processes numbers, this 'agentic' system can reason through complex theoretical physics and create new hypotheses for how particles interact. It then validates its own ideas against real-world data, effectively acting as an autonomous scientist capable of discovery without human intervention.
From the abstract
We present the FERMIACC, a scaffolded reasoning model built on OpenAI agents designed to autonomously generate and quantitatively validate theory hypotheses for high energy physics data at scale.