GPT-5.6 Sol Is Running Quantum Computing Experiments While Researchers Sleep

Description and use

GPT‑5.6 Sol autonomously runs quantum experiments, cutting characterization time from days to hours and enabling researchers to focus on design and analysis.

Problem

Researchers at MIT’s EQuS group needed to perform routine, time‑consuming characterization of superconducting qubit chips, which traditionally required days of manual measurement and analysis.

Outcome

The AI agent reduced routine characterization from days to hours, allowing one graduate student to run experiments on multiple chips in parallel and freeing researchers to focus on experimental design and analysis.

Workflow

• 1. Connect Codex to the lab’s experiment coordination software and provide measurement‑specific skills.
• 2. Codex selects measurement parameters based on design targets.
• 3. Codex sends microwave pulse sequences to the dilution refrigerator to operate the qubit chip.
• 4. Codex collects and analyzes the resulting data to identify qubit transition frequencies.
• 5. Codex calibrates control and readout pulses and measures coherence times.
• 6. Codex chains subsequent measurements, using each result to inform the next step.
• 7. Researchers review results and intervene only when signals are weak, noisy, or unexpected.

Evidence

OpenAI published a fascinating case study today: MIT's Engineering Quantum Systems (EQuS) group is now using GPT-5.6 Sol — harnessed to Codex — to autonomously run superconducting qubit experiments.

Source

When signals were clear, Codex completed standard measurement sequences with minimal researcher intervention.

Source