PSI Achieves Five Quantum Research Breakthroughs Using GPT and Claude

Huo Xing Finance reports that AI Physics Laboratory Physical Superintelligence (PSI) has made new progress on five long-standing quantum information problems using models such as GPT-6 Astra and Claude, and has publicly released papers, proof code, and verification materials. During the research, physicists provided guidance and validation, while the models participated in derivations. The team also employed AI agents from Codex and Claude Code to write formal proofs in Lean, which were then checked by computers for logical validity. One study demonstrated that in a specific quantum field, quantum entanglement persists even between widely separated regions, but beyond a certain distance, non-local correlations can no longer be detected via the CHSH Bell inequality. Another study proved that repeatedly applying a certain class of quantum operations will eventually destroy entanglement. These five findings do not fully resolve the original problems; some conclusions apply only under specific conditions, and unresolved questions remain.