Anthropic reports AI-designed quantum circuits with lower resource demands

Anthropic reported on October 16, 2026 that Claude produced improved designs for eight quantum-computing routines using an automated evaluator to check correctness and score resource use. The work covers building blocks for applications including chemistry, optimization and number theory.

One elliptic-curve point-addition design uses 17 times fewer gates than the smallest previous design while requiring 12% more qubits. That trade-off matters: reducing operations can make a computation less demanding, even when it needs somewhat more quantum memory. Anthropic says the point-addition exploration ran for 53 hours with limited human input.

The result concerns circuit design and resource estimates, not a demonstrated useful computation on a quantum processor. Evaluator feedback, numerical testing and formal verification supported parts of the work, but the announcement does not show that today’s hardware can execute these applications at practical scale. Its significance is evidence that AI-assisted algorithm design could reduce the resources future quantum machines need. Source: Anthropic.