Researchers at Cornell’s CHESS facility and partner institutions have linked AI, high-performance computing and scientific instruments in a system that can direct experiments in real time. The collaboration has won a 2026 HPCwire Editors’ Choice Award, but the more interesting prize is the working feedback loop behind it.
Watch Desk analysis
What happened
During a demonstration at CHESS, researchers used high-energy X-rays to examine stresses inside a metal sample. Data travelled from Ithaca to computing resources at Oak Ridge National Laboratory, where an AI-driven active-learning system analysed the results and chose which measurements would be most useful next. Instructions then went back to the CHESS instruments automatically.
The collaboration brings together CHESS, Oak Ridge, the University of Tennessee, Knoxville, the University of Utah, NIST, the National Science Data Fabric and ViSOAR. Cornell says the system connects equipment and computing resources hundreds of miles apart. Its account describes a demonstration on structural materials, not a general-purpose autonomous laboratory ready for any experiment.
Why it matters
Many scientific experiments involve repeated measurement and analysis: collect data, decide what to test next, then do it again. Automating part of that loop could help researchers make better use of instrument time and large datasets, while connecting specialists and facilities that are not in the same building. The infrastructure may also be adaptable to other research areas, including quantum materials and biological systems, though those are future possibilities rather than demonstrated results here.
Our read
This is a satisfying example of AI doing a specific job in science: choosing the next useful measurement, then sending that decision back to the instrument. The important advance is not simply that a model analysed data, but that computation and equipment were connected into a working loop across institutions. The next test is whether researchers can make this approach accessible and repeatable beyond the initial demonstration.
What to watch
- Whether the team publishes further results on the accuracy and speed of the measurement choices.
- Which additional instruments and research areas join the infrastructure.
- Whether researchers can bring their own AI models to work with CHESS and partner computing resources.
Discussion spark: For autonomous experiments, should researchers prioritise faster measurements, or keep every next-step decision under direct human control?
Sources and evidence
- Source update (18 November 2026, 19:22 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.