Giorgio Parisi, the Nobel laureate physicist, and Francesco Zamponi have solved a mathematical puzzle that emerged a decade ago from their research on jamming physics, the study of how particles jam together in materials. The pair used Claude, an AI assistant made by Anthropic, to crack the problem rather than developing an entirely new theory.
Jamming describes the transition point where materials like granular substances or colloids shift from flowing freely to becoming locked in place. This phenomenon appears across physics, from sand piles to cell biology. Parisi and Zamponi had identified a mysterious mathematical relationship buried within jamming systems but struggled to explain its origin for years.
The researchers fed Claude information about the mathematical structure of their findings and the underlying physics. The AI system helped them identify patterns and connections they had missed, ultimately revealing how the relationship emerged from first principles. By leveraging Claude's ability to process and synthesize complex information, they bypassed the need for months or years of traditional theoretical work.
Parisi won the 2021 Nobel Prize in Physics for discoveries about disorder and fluctuations in physical systems, work directly relevant to jamming. Zamponi, based at the École Normale Supérieure in Paris, has spent years studying the theoretical foundations of jamming in amorphous materials.
The breakthrough demonstrates how large language models can serve as tools for scientific discovery beyond their typical applications. Rather than replacing human intuition, Claude functioned as a collaborative partner that could rapidly explore mathematical terrain and flag promising connections.
This result carries limitations. The puzzle solved was mathematical rather than experimental, and the AI worked within frameworks the researchers already understood. Claude did not independently conceive new physics or generate novel experimental predictions. The system succeeded at pattern recognition and synthesis within known domains.
Still, the collaboration signals a shift in how theoretical physicists approach stubborn problems. As AI systems become more sophisticated, their role