Designing more efficient quantum circuits with Claude
Anthropic says Claude designed circuits that use fewer qubits and gates across eight computational tasks, with results described in a preprint.
TL;DR
- Anthropic says Claude proposed lower-cost designs across eight quantum-circuit tasks; the accompanying preprint reports improvements in seven of eight.
- The researchers used an evaluator to check correctness and score proposed circuits, and report a new reversible modular-inversion design.
- The work is a preprint and company-led research; the results do not establish when practical fault-tolerant quantum computers will be available.
Anthropic researchers reported using Claude to design quantum circuits for tasks in chemistry, optimization, simulation and number theory. The company said an automated evaluator checked correctness and scored each design by its qubit and gate requirements. [1] [2]
The accompanying preprint reports that seven of eight tasks reduced both Toffoli-gate and qubit counts against prior designs. For the remaining task, the authors describe a new reversible architecture for modular inversion, an operation used in elliptic-curve calculations. [1] [2]
Anthropic’s post says Claude worked in a feedback loop with the evaluator and developed one point-addition design after a 53-hour run with minimal human input. The researchers also report that a redesigned decoder reduced the costliest step in a quantum optimization algorithm by a factor of 12. [1] [2]
The results are presented in a preprint and the company’s research account, and have not yet been established through peer review or independent reproduction. The study concerns circuit resource estimates; it does not demonstrate a completed practical quantum computer or set a timeline for one. [1] [2] [3]
Why it matters
The work tests whether model-guided search can improve quantum algorithms by reducing circuit resource requirements. Its significance depends on independent review and whether the designs translate to usable hardware.
Editor's note
Frames performance claims as preprint and company-reported results; does not infer a timeline for practical quantum computing.