AI
OpenAI describes Codex agents running routine measurements on MIT quantum chips
OpenAI reports that MIT researcher Beatriz Yankelevich connected GPT‑5.6 Sol through Codex to laboratory software for superconducting qubit experiments. On an uncalibrated six-qubit chip, agents selected measurement settings, operated hardware and analyzed results using researcher-provided instructions. The account suggests agents can reduce supervision of routine scientific work, allowing researchers to concentrate on experiment design and interpretation. Weak or noisy signals still required more effort and sometimes expert guidance; the account provides no quantified comparison of time savings.
Editorial summary · Publisher article
What matters
- Agents identified transition frequencies, calibrated control and readout pulses, and measured information retention.
- The group reportedly now uses agents regularly for routine measurements, including extended unattended runs.
- Novel experiments receive narrower agent goals and more researcher direction.
Context & caveats
- OpenAI published the account; the reported benefits are not independently validated.
- Experienced researchers may find the best calibration settings faster.
Why this made the edition
Substantive first-party account describes a concrete laboratory workflow and its limitations, although productivity gains are not independently measured.
Source material