GPT‑5.6 Sol + Codex run overnight qubit calibration at MIT EQuS

· opgehaald 12:09

OpenAI case study: MIT’s Beatriz Yankelevich wired GPT‑5.6 Sol (via Codex) into EQuS lab software; on a clear six-qubit chip it autonomously found transition frequencies, calibrated pulses, and measured coherence — freeing hours of routine characterization.

On 8 Sep 2026 OpenAI published “How GPT‑5.6 Sol helps run quantum computing experiments”: MIT Engineering Quantum Systems (EQuS) graduate student Beatriz Yankelevich connected GPT‑5.6 Sol through Codex to the group’s lab software that drives superconducting-qubit measurements inside dilution refrigerators. Given measurement skills and chip design targets, the agent chose parameters, ran hardware, analyzed returns, and either refined or saved results for the next step. On an uncalibrated six-qubit process-control chip with clear signals it completed a standard calibration sequence with little intervention — identifying transition frequencies, control/readout pulses, and coherence times. Noisy or weak signals still needed human steering. EQuS now regularly leaves agents on overnight/cleanroom runs so researchers focus on design and analysis; Yankelevich notes checking progress from her phone.