Back to news
researchAnthropic2026-11-07

Anthropic unpublished model used to prove the Aaronson-Ambainis conjecture in three days

Physical Superintelligence said it used an unpublished Anthropic model to complete a proof of the Aaronson-Ambainis conjecture in three days, a problem unsolved since it was posed in 2008.

On November 7, it was reported that according to Business Insider coverage that morning, AI helped crack a long-standing major problem in quantum computing. Physical Superintelligence announced that the Aaronson-Ambainis conjecture in quantum computing had been successfully proved using an unpublished Anthropic AI model. The conjecture, posed in 2008, had remained unproved and studies the conditions under which quantum computers can achieve computational speeds far beyond classical computers.

PSI built its own research system using Anthropic models and completed the proof in just three days. According to the company, the effort means one of the most fundamental questions in quantum computing has been resolved. The research system that proved the conjecture was developed by PSI engineer David Turetsian, and used an Anthropic model that has not been released, though which specific model remains undisclosed.

The result shows that for quantum computing to achieve certain exponential speed advantages, it must rely on special regularities or structures present in the problem, confirming a long-standing conjecture among researchers that quantum computers do not deliver huge performance gains in all cases, while still delivering large speedups on some tasks. AI-assisted mathematical research breakthroughs continue to accumulate, with OpenAI having announced in September the resolution of a 90-year-old Navier-Stokes equation problem.

Anthropic量子计算猜想证明Physical Superintelligence数学