Updated
Updated · Livescience.com · Jul 16
Parisi and Zamponi Use 40 Claude Prompts to Prove a+b=1 in Jamming
Updated
Updated · Livescience.com · Jul 16

Parisi and Zamponi Use 40 Claude Prompts to Prove a+b=1 in Jamming

3 articles · Updated · Livescience.com · Jul 16

Summary

  • A July 1 paper says Giorgio Parisi and Francesco Zamponi used Anthropic's Claude to produce a publishable proof that a+b=1, resolving a jamming-physics problem that had resisted them for more than a decade.
  • The pair had first seen the relation in a 2014 numerical study of jamming, where exponents a and b describe how contact forces and tiny gaps scale at the critical transition from fluid-like motion to a rigid disordered state.
  • Claude first reproduced the 2014 result, then generated the core analytical idea after about 40 prompts; the researchers corrected errors, but said the key insight was valid and came directly from the equations.
  • Zamponi said the episode changed his view of what AI can do in theoretical physics, though his current work on hard hyperspheres still suggests human guidance remains essential even when AI speeds coding and drafting.

Insights

A Nobelist's AI partner solved a decade-old physics problem. Are human scientists becoming obsolete?
Was the AI’s groundbreaking proof a flash of digital genius or just a clever mathematical trick?
If AI can mass-produce scientific proofs, how will we measure the value of human discovery?

AI and Physicists Crack the Jamming Puzzle: Analytic Proof of a+b=1 Unifies Decade-Long Theories

Overview

In July 2026, a major breakthrough in physics was achieved when Giorgio Parisi, Francesco Zamponi, and the AI model Claude solved a decade-long puzzle about the jamming transition. For years, physicists struggled to find an analytic proof for the key relation a+b=1 in the full replica-symmetry-breaking theory. Many believed the solution would be extremely complex, but the collaboration revealed a surprisingly simple answer. This success not only unified different theoretical approaches to jamming but also showcased how human-AI teamwork can tackle deep scientific problems that had resisted explanation for years.

...