Codex, powered by GPT-5.6 Sol, coordinates superconducting qubit tests

MIT graduate student Beatriz Yankelevich used GPT-5.6 Sol, harnessed to Codex, to automate routine measurements on superconducting qubit chips. This allowed her to shift focus from hands-on operation to higher-level research tasks like experiment design and data analysis. Superconducting qubits, cooled to near absolute zero inside dilution refrigerators, require software control, making them ideal for AI automation. Yankelevich found that GPT-5.6 Sol could often complete routine measurement workflows autonomously, saving significant time. GPT-5.6 Sol Coordinates Superconducting Qubit Measurements GPT-5.6 Sol identified how long a superconducting qubit retained quantum information during recent experiments at MIT. Graduate student Beatriz Yankelevich connected the AI, harnessed to Codex, to laboratory software controlling dilution refrigerators and qubit chips; this allowed the system to independently complete standard calibration sequences without intervention. The resulting data revealed qubit resonance frequencies and enabled accurate control, a process previously demanding significant hands-on time from researchers. Codex’s ability to coordinate interdependent measurements stems from the software-controlled nature of superconducting qubit experiments; these qubits, cooled to near absolute zero, are manipulated using microwave signals and analyzed through digitized returns. Yankelevich equipped Codex with measurement-specific skills, detailing how to execute and evaluate each experiment, enabling GPT-5.6 Sol to select parameters, operate hardware, and analyze data. When signals were strong, the AI refined measurements or saved results for subsequent steps, demonstrating a capacity to handle routine workflows. “I can have multiple agents working on different problems at once, and I spend most of my time on higher-level work—interpreting results, devising experiments, planning next steps for the agents, reading, and writing,” Yankelevich stated. The system’s capabilities extend beyond simple execution; GPT-5.6 Sol identified a qubit’s transition frequencies and calibrated the pulses used for control and readout. This level of automation is particularly valuable given the extensive preliminary work required for each qubit experiment, typically hundreds to thousands of measurements, work that Codex is demonstrating the potential to streamline. While Codex excelled with clear signals, challenges emerged with weak or noisy data, sometimes requiring guidance from an experienced researcher to refine measurement parameters. “I’ve built infrastructure to guide agents through several parts of my work—measurement, theory, and chip design—and now it’s really starting to pay off,” Yankelevich explained. The potential for remote monitoring further enhances the efficiency of this AI-driven approach; Yankelevich noted, “I can check in from my phone, see what they’ve done, and steer them if something needs fixing or if I want to explore a different direction.” She also described running measurements for extended periods, even while working in a cleanroom, highlighting the system’s ability to operate without constant supervision. This frees researchers to focus on higher-level tasks, such as data analysis and experimental design, accelerating the pace of discovery in quantum computing. I can have agents running measurements for many hours overnight or while I’m working in the cleanroom. Beatriz Yankelevich, a graduate student in MIT’s Engineering Quantum Systems Group (EQuS) Codex Automates Calibration of Six-Qubit Chips Connecting laboratory software to GPT-5.6 Sol enabled autonomous execution of calibration measurements on a six-qubit chip, a demonstration of streamlined quantum experimentation at MIT’s Engineering Quantum Systems Group (EQuS). The system’s ability to independently complete a series of interdependent measurements, where each result informs the next, proved particularly valuable given the inherent drift in qubit properties and potential for inconsistent results. Codex’s operation isn’t simply rote execution; the AI analyzes results and adjusts subsequent measurements, mirroring the adaptive approach of an experienced researcher. During testing, GPT-5.6 Sol identified the qubit’s transition frequencies, calibrated the pulses used to control and read it, and identified how long the qubit retained quantum information. The benefit extends beyond simply accelerating routine characterization; Codex facilitates a more flexible approach to experimentation. For well-defined workflows, the AI can operate largely autonomously, while for novel experiments, researchers assign narrower goals to the agents and use their coding capabilities to develop and test new control, analysis, and simulation tools. This direct connection between AI agents and laboratory equipment allows for rapid iteration, revising code, testing against real measurements, and completing extended work periods without interruption. The immediate impact, according to the research, is steady progress on experimental analysis and measurements, even without constant human oversight. Experienced researchers may still identify optimal calibration settings more quickly, but the time saved on monitoring allows for broader exploration and deeper analysis. I’ve built infrastructure to guide agents through several parts of my work-measurement, theory, and chip design-and now it’s really starting to pay off. Beatriz Yankelevich, a graduate student in MIT’s Engineering Quantum Systems Group (EQuS) EQUS, the ARC Centre of Excellence for Engineered Quantum Systems, is an Australian research centre established to translate quantum research into practical technologies, talent and broader impact. Headquartered in Brisbane and active since 2011, EQUS fosters collaboration between the University of Queensland, which leads the consortium with joint hardware and systems research, UNSW Sydney focusing on silicon spin qubits, and the University of Sydney, specialising in quantum photonics and control. The centre’s structure encompasses research programmes alongside translation activities, producing resources such as quantum explainers and the Futurum careers platform, as well as public engagement initiatives like a dedicated podcast. EQUS maintains a commercial partnership with IBM Quantum, providing Australian academic researchers with access to quantum computing resources. This focus on both fundamental research and practical application aligns with the work described in this article, where automated control of superconducting qubits, a key area for EQUS partners, is achieved through artificial intelligence. EQUS’s broad remit covers research, translation, policy and public outreach, and it was recognised in 2025 for championing quantum progress and female scientists. Source: https://openai.com/index/codex-quantum-computing-experiments/ More like thisQuantum PhysicsWiring density limits qubit control, Bluefors research showsArtificial IntelligenceDeepSeek used distillation to train its R1 and V3 modelsQuantum Computing Business NewsNew cryogenic platform supports Quobly’s quantum roadmapArtificial IntelligenceUChicago team wins award for AI filter at CERN’s colliderStay currentSee today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals. Tags: