Anthropic: Claude Designs More Efficient Quantum Circuits, Reducing the Number of Gates Required for Some Tasks by Up to ~17x

According to a report on Anthropic's official blog on October 16, Anthropic collaborated with researchers from Harvard University to enable Claude to autonomously design quantum circuits. Across eight foundational quantum circuit tasks in fields such as chemistry, optimization, quantum simulation, and number theory, Claude's designs required fewer qubits and gates than previously published state-of-the-art solutions. Elliptic curve point addition is a critical step for quantum algorithms used to break elliptic curve cryptography. Claude successfully implemented a reversible modular inversion that does not require storing branch history. Compared to the previous design using the fewest qubits, the new approach requires 12 percent more qubits but reduces the gate count by approximately 17 times. Within the DQI optimization algorithm, which is considered promising for delivering super-polynomial quantum speedup, Claude redesigned the highest-cost decoding step, cutting the gate count by roughly 12 times compared to Google’s prior solution. The study suggests that such advancements can substantially lower the resources needed for quantum algorithms, potentially accelerating the deployment of practical quantum computers.