AI takes over a critical “craft” task in quantum computing, beating human experts on laser control

Quantum computing startup QuEra Computing has disclosed that it successfully handed one of the most delicate parts of running its machines—the alignment and stabilization of control lasers—over to an AI agent built with Anthropic’s Claude. In neutral‑atom quantum systems, even tiny drifts in laser position or frequency can ruin computations, so specialists typically spend significant time manually retuning the optics.

In QuEra’s tests, the Claude‑based agent ingested sensor readings from the quantum hardware, iteratively updated its own control logic, and locked the lasers back on target. According to the company, the AI restored the system in seconds instead of the minutes a human expert would need, while holding the beam more stable than manual tuning. The work is part of the broader Model Hardware Standard (MHS) initiative and is slated to expand to other subsystems such as cryogenics and vacuum control. For the AI ecosystem, this is a high‑profile example of large models moving from software‑only workflows into the operation of extremely sensitive physical infrastructure—promising big efficiency gains but also raising questions about failure modes and accountability when something goes wrong.