Giorgio Parisi and Francesco Zamponi have resolved a decade-old mathematical conjecture in jamming physics using an unconventional partnership with Claude, an AI language model. The proof appears in the Journal of Statistical Mechanics: Theory and Experiment.
Jamming represents a fundamental transition in disordered systems where particles abruptly become immobilized. This occurs in granular materials, colloids, and other complex systems when density or stress crosses a critical threshold. The mathematical relation eluded physicists despite sustained theoretical effort since its initial proposal over ten years ago.
Parisi, who won the Nobel Prize in Physics for his contributions to disordered systems and complex networks, collaborated with Zamponi at LaSapienza University of Rome on the problem. The team deployed Claude to assist in navigating the proof's logical structure and mathematical framework. The AI system contributed to formulating arguments and exploring potential solution pathways that human researchers had not fully developed.
This collaboration demonstrates a pragmatic application of large language models in theoretical physics. Rather than replacing physicists, Claude functioned as a collaborative tool that helped organize complex mathematical reasoning and identify unexplored avenues within the jamming problem's solution space.
The resolution of this conjecture advances understanding of jamming transitions, phenomena with applications spanning materials science, powder technology, and biological systems. Jamming principles explain how sand flows through silos, how colloids reach percolation thresholds, and how cellular tissues organize.
The work raises questions about the future role of AI in mathematical proof-finding. While the accomplishment itself validates using language models for hypothesis exploration and logical scaffolding, the physicists' expertise remained essential for directing the investigation and validating results.
This outcome suggests that AI excels not at independent discovery but at accelerating research processes where human insight drives the inquiry. For complex physics problems requiring deep domain knowledge and creative mathematical thinking, AI