The Engineering Quantum Systems (EQuS) group at MIT has published a joint case study with OpenAI on the use of an AI agent that autonomously performs the measurement process of superconducting qubits by integrating GPT-5.6 Sol and Codex.
NewsMassachusetts Institute of TechnologyGPT-5.6 Sol
MIT Successfully Automates Qubit Measurement Using GPT-5.6 Sol
This article is a translation. Read the Japanese original
Qubit calibration requires a vast number of measurements and continuous parameter adjustments based on those results, a process that was traditionally managed manually by researchers. In this validation, the AI agent autonomously selected measurement parameters, operated hardware, analyzed data, and determined the next measurement to be performed via experimental software. The research group confirmed that the agent could complete standard measurement sequences without continuous monitoring by researchers.
According to the research team, the introduction of the AI agent allows time previously spent on monitoring measurements to be allocated to more advanced tasks such as experimental design and result analysis. However, it was reported that challenges remain in interpreting ambiguous physical phenomena, such as when signals are weak and noisy, requiring guidance from researchers to find appropriate parameters.
Sources
- How GPT‑5.6 Sol helps run quantum computing experiments （Hacker News Frontpage, 2026-09-09）
- OpenAI Technical Case Study