On October 16, 2026, Anthropic published “Designing more efficient quantum circuits with Claude”. The work was led by Andi Gu in Anthropic’s STEM Fellows Program, mentored by Alek Kemeny, with Susanne Yelin and Mikhail Lukin of Harvard (Lukin also of QuEra Computing). Claude was given plain-language specifications for eight quantum-circuit primitives, from chemistry, optimization, physics simulation, number theory and numerical computing, plus an automated evaluator that checked each design for correctness and scored it on qubits times gates.
Working in that closed loop with minimal human input, Claude Mythos 5 found designs using both fewer qubits and fewer gates than the best previously published ones on all eight tasks, in some cases by an order of magnitude or more. For elliptic-curve point addition, the core of quantum attacks on elliptic-curve cryptography, a 53-hour run produced the first reversible version of the “divsteps” modular inversion method that needs no stored branch history; Claude proved new algebraic identities and checked the result in Lean. One design in the resulting family uses 12 percent more qubits but 17 times fewer gates than the smallest previous design. For the Reed-Solomon decoder at the heart of the DQI optimization algorithm, switching to Berlekamp-Massey decoding and sizing memory from 403 million simulated decodings cut gates by a factor of 12.
For context, the post notes that the estimated gate count for simulating the FeMoco enzyme cluster fell about 150,000-fold, from 50 trillion to about 340 million Toffoli gates, over eight years of work by many groups.
Why it matters: quantum algorithms are code that can be checked and scored automatically, which suits an AI search loop. Cheaper circuits lower the hardware bar for useful quantum computing, and that includes the bar for breaking elliptic-curve cryptography.
What it does not show: these are resource estimates for circuits, not runs on quantum hardware. The authors note Claude was conservative by default and had to be prompted to broaden its search, and circuit design is only one part of building an error-corrected machine.