Anthropic's Claude Designs More Efficient Quantum Circuits, Reducing Gate Count by Up to 17x

According to Anthropic’s official blog on October 16, Anthropic collaborated with Harvard University researchers to enable Claude to autonomously design quantum circuits. In eight fundamental quantum circuit tasks across fields such as chemistry, optimization, quantum simulation, and number theory, Claude’s designs required fewer qubits and gates than any previously published best solutions.
Among these, elliptic curve point addition is the core step in quantum algorithms that break elliptic curve cryptography. Claude has for the first time implemented a reversible modular inverse operation without storing branch history. Compared to previous designs with the fewest qubits, the new scheme uses 12% more qubits but reduces the number of gates by approximately 17 times. In the DQI optimization algorithm, which is believed to offer super-polynomial quantum speedup, Claude redesigned the most costly decoding step, reducing the number of gates by about 12 times compared to Google’s prior approach.
Research suggests that such improvements could significantly reduce the resources required by quantum algorithms, potentially accelerating the deployment of practical quantum computers.