Automating Iterative Experimental Workflows

Quantum computing research involves calibrating superconducting qubits—a process requiring hundreds to thousands of preliminary measurements. These measurements are interdependent: each result informs the next, and physical drifts require constant, adaptive decision-making. By connecting AI agents (specifically GPT-5.6 Sol via Codex) to laboratory control software, researchers can automate these repetitive calibration sequences. The agent is provided with specific 'skills'—instructions on how to run, evaluate, and refine experiments—allowing it to autonomously identify transition frequencies, calibrate control pulses, and measure coherence times.

Balancing Autonomy and Human Oversight

While AI agents excel at well-defined, routine workflows, they face limitations when dealing with ambiguous or noisy physical signals. In such cases, the agent may struggle to select optimal parameters, requiring intervention from an experienced researcher. The current best practice is a hybrid model: the agent handles the heavy lifting of standard characterization, while the researcher monitors progress remotely (e.g., via mobile) and provides guidance when the AI encounters unexpected physical behavior. This setup allows for 'overnight' experimentation, significantly increasing the throughput of chip characterization.

Scaling Research Through Agentic Infrastructure

Beyond simple measurement, the integration of agents into the lab environment enables a shift in researcher focus toward higher-level tasks like experiment design, data interpretation, and theoretical planning. Researchers can deploy multiple agents to work on distinct problems simultaneously—such as measurement, theory, and chip design—by leveraging the agent's ability to write, test, and modify control code in real-time. This infrastructure allows for a more iterative research cycle where the AI acts as a force multiplier, handling the execution of long-running tasks while the human focuses on the strategic direction of the research.