Physical Superintelligence AI System Solves 18-Year Quantum Computing Conjecture with Verified Lean Proof
Physical Superintelligence uses an unreleased Anthropic model to autonomously prove the 18-year-old Aaronson-Ambainis conjecture, formally verified by the Lean proof assistant.
Physical Superintelligence (PSI) has announced that an autonomous AI research system powered by an unreleased Anthropic reasoning model has successfully resolved the Aaronson-Ambainis conjecture, an open foundational question in quantum computing since 2008. The system derived the complete mathematical proof over three days of computation, with every logical step formally audited and accepted by the Lean interactive theorem prover.
Settling an 18-Year Open Problem in Quantum Complexity Theory
First formulated in 2008 by Scott Aaronson and Andris Ambainis, the conjecture posits fundamental constraints on the computational power of quantum algorithms over unstructured or random inputs. The newly proved theorem establishes that quantum computers cannot achieve superpolynomial or exponential speedups on random boolean functions, demonstrating that quantum advantages fundamentally require underlying algebraic, geometric, or periodic mathematical structure, as seen in algorithms like Shor period-finding.
Autonomous Mathematical Synthesis Powered by Anthropic and Formalized in Lean
The breakthrough was led by David Turturean and the technical staff at Physical Superintelligence, who engineered a closed-loop reasoning pipeline built atop an unreleased frontier Anthropic reasoning model. Over a 72-hour autonomous run, the system navigated complex combinatorial pathways, synthesized novel lemmas combining Boolean function analysis with probabilistic combinatorics, and translated every mathematical statement into the Lean proof language, guaranteeing absolute formal correctness without hallucination.
External Academic Audits and Frontiers in Machine-Led Scientific Discovery
The machine-generated proof was audited internally by theoretical physicists and mathematicians including Jim Halverson and Phiala Shanahan, while independent researchers from Stanford University led by Li-Yang Tan and Guy Blanc reviewed the formalization. The achievement represents one of the most significant demonstrations to date of frontier AI autonomous capability to resolve long-standing open problems in theoretical physics and mathematics, pointing toward an accelerated paradigm of automated scientific research.