AI-for-Science Catchup: Developments We've Been Tracking
Generative crystal design moves from research papers to usable tools, Elicit ships Research Agents and a public API, and autonomous lab platforms cross from industry into academic reach.
Monthly digests of new tools, pricing changes, and research developments — focused on what actually matters for researchers, not AI hype.
Generative crystal design moves from research papers to usable tools, Elicit ships Research Agents and a public API, and autonomous lab platforms cross from industry into academic reach.
Argonne extends its battery-materials foundation model toward molecular crystals, ECMWF's AIFS remains the operational benchmark, NOAA's GraphCast-derived models keep outperforming legacy systems, and a 45-year GraphCast hindcast archive opens up new climate research.
GenCast's peer-reviewed Nature results are genuinely strong — beating ECMWF's ensemble system on 97.2% of verification targets. Now NVIDIA's open-source Earth-2 suite claims to beat GenCast too, but that comparison isn't independently verified yet.