Cornell wins $7.5 million for research into visual computing with light. The Cornell-led Deep Nanophotonic Inference Systems project will study foundations for optical-electronic systems that could perform some AI image-analysis computations using light.
Cornell team develops an ultrasound-directed microbubble chip. The Cornell researchers’ chip forms hydrogen microbubbles at electrodes within a three-millimeter area; ultrasound-driven oscillations then direct fluid flow.
Italy’s ENEA outlined its role in SUPER on October 14. The three-year project aims to help Alpine local authorities evaluate fragmented data on buildings, energy consumption and planned investment through shared digital methods.
Argonne National Laboratory and the University of Illinois Urbana-Champaign reported a laboratory demonstration coupling traveling spin waves in a magnetic material to a superconducting resonator. Argonne’s October 13 announcement describes a hybrid approach to one-way microwave signal control.
Argonne National Laboratory1 min read Data included
Duke-led AITC-AIGE, one of five centers selected by the U.S. National Institute on Aging, will receive nearly $25 million over five years. The center aims to advance trustworthy AI research for aging and dementia care.
Duke University School of Medicine1 min read Data included
Cornell researchers propose a credit market for data-center power cuts. Cornell’s proposed Curtailment Credit Market would let data centers facing required cuts pay more flexible energy users to take on part of their reduction obligations.
MIT’s IRIS model uses AI to reconstruct cells’ signaling histories. Researchers at MIT and the Whitehead Institute reported IRIS, a neural-network model that infers active cell-signaling pathways from gene-activity patterns.
Argonne reported on October 12 that 75 people participated in ATPESC 2026. The program ran July 26–August 7 in Illinois, with researchers carrying out hands-on work on the Aurora, Polaris and Frontier supercomputers.
Argonne Leadership Computing Facility1 min read Data included
Brookhaven National Laboratory described its X-AutoMap automation tool on October 14. The lab said roughly 500 particles from an urban air-pollution sample were analyzed in about 10 hours, compared with about 44 hours normally required for manual data collection.
Brookhaven National Laboratory1 min read Data included
The U.S. Department of Energy announced the public release of Genesis Agent Skills for the Genesis Mission on October 15. The package provides reusable instructions and workflows that AI coding tools can use in scientific work.
Google Cloud announced a preview integration between Gemini and its Borderless Lakehouse. Lakehouse catalog federation can expose sources such as Databricks Unity Catalog and Snowflake Horizon inside Gemini, letting users query records in natural language while agents use the same data context without moving the underlying data.
Meta engineers built Forum, a separate app for Facebook Groups, by placing an AI agent inside the planning, implementation, testing and verification loop. The company defined success as production-ready changes that build, pass tests and survive review, rather than the amount of code generated.