Research

QuEra, Harvard and Anthropic report AI-designed circuits with lower resource needs

Researchers from QuEra Computing, Harvard University and Anthropic used Anthropic's Claude, paired with an automated evaluator, to design circuits for eight building blocks of quantum algorithms. In a preprint (arXiv:2610.22367, not yet peer reviewed), they report lower gate counts and qubit requirements in seven of the eight. Reported results include a 12-fold cut in Toffoli gates for a Reed-Solomon decoder (5.1 million to about 440,000), and a 9.9-fold cut in a combined metric for a Coulomb potential (Toffoli gates from 6,355 to 1,140; logical qubits from 252 to 142). A fixed-point Gaussian function saw a 39-fold cut in the combined metric (Toffoli gates from about 24,000 to 1,583; logical qubits from 362 to 139). The team also reports a new reversible modular inversion approach, with one elliptic-curve point-addition design needing about 17 times fewer gates than the prior fewest-qubit design while using about 12% more qubits, found in a 53-hour search.

Why it mattersThe combined metric multiplies Toffoli count by peak logical qubits, so 39x is not a 39x speedup, and the figures are resource estimates for individual building blocks, not hardware runs. QuEra says lower algorithm requirements can expand what a given fault-tolerant machine could handle, though the effect on complete applications will depend on how the components are used.

Read the full story at The Quantum Insider →

Summarized by AI from the linked article.