This week: Anthropic expanded its AI for Science program to 20 foundations, institutes, and research projects, from the team behind the Hubble and Webb telescopes to a complete wiring map of a fruit fly's brain, all getting free Claude credits.

Elsewhere: How the credits actually reach scientists, which open projects are on the list, and what Anthropic didn't say about the size of the deal.

The New Prime Signal

Most AI announcements this week were about bigger models and bigger enterprise deals. This one is about handing the tools to people who study fruit fly neurons and rare diseases. On October 28, Anthropic announced an expansion of its AI for Science program, which "gives free Claude credits to researchers working on high-impact scientific projects," to 20 scientific foundations, institutions, and research groups around the world. Somewhere a grad student who has been rationing API calls like the last slice of pizza just sat up a little straighter.

What happened: Anthropic started the program in May 2025 with free API credits for individual researchers. Now it is working through organizations. Six science funders and societies will pass Claude credits on to researchers: the Alfred P. Sloan Foundation, the Simons Foundation, Schmidt Sciences, Renaissance Philanthropy's Open Source for Science Fund, the US National Science Foundation, and CAS, the American Chemical Society division that indexes the world's published chemistry. Six institutions will use the credits in-house, including the Space Telescope Science Institute (which runs science operations for Hubble, James Webb, and Roman), the Wellcome Sanger Institute, the Francis Crick Institute, Imperial College London's life sciences department, EMBL, and the Gladstone Institutes. The rest are specific research projects, covered below.

Why it matters: A lot of science now runs on giant shared datasets, and the bottleneck is often people-hours, not ideas. Anthropic says scientists are using Claude "to generate hypotheses, design experiments, swiftly analyze complex data, and communicate findings," and that it has supported "hundreds of projects across biology, medicine, physics, chemistry, climate science, and mathematics." It points to two earlier partnerships: with HHMI Janelia it built the Model Hardware Standard, "a shared specification that lets AI agents safely operate lab instruments," still in research preview, and the Allen Institute is building teams of AI agents that Anthropic says are "turning months of manual analysis into hours." The part we like most: many of these projects are meant to produce things everyone can use, not just the grant winners.

The fine print: Anthropic didn't say how many credits are involved or what they are worth, so there's no way to judge how big this really is. It also says "many of these collaborations are still in their early stages," so for now this is a list of partners, not a list of discoveries. The claims about what Claude does for scientists come from Anthropic itself. And it's worth remembering this is also good business: get the world's top labs building their workflows on your model, and they tend to stick around.

Still, this is the kind of AI news we'd like more of. Free compute for fly-brain maps, climate models, and rare disease databases beats another chatbot that writes slightly better marketing emails. The real test comes later: whether these partners publish open datasets and tools people actually use. We'll be watching for that, not the press release.

The credits reach scientists three ways: through funders, through institutions, and directly to a handful of projects.

How it works: Funders like Sloan, Simons, and the NSF "will distribute these resources to ambitious researchers," so scientists apply through those organizations, not to Anthropic. Institutions like EMBL and the Crick use the credits inside their own labs. And six projects get credits directly: FlyWire at Princeton, which produced the first complete wiring map of an adult fruit fly's brain; LEAP at Columbia, which mixes machine learning with physics to improve how climate models handle clouds, oceans, and land; AstroPath at Johns Hopkins, which borrows calibration tricks from astronomy surveys to image tumors cell by cell; gnomAD at the Broad Institute, an open reference of human genetic variation pooled from "over 800,000 people"; Tau Ceti, a community project building a shared library of math in Lean, a language that lets computers check every proof; and Australia's Centre for Population Genomics, which works to include the country's multicultural communities in genomic resources. Yes, someone put an astrophysicist and a pathologist in the same room, and it worked.

Try it: If you're a researcher at an institution, Anthropic's original program page says you can apply through its application form, and the new post mentions a discounted Team plan for scientists. If you're just curious, two of these projects are open right now: browse a fly's brain wiring on FlyWire, or look up a gene on gnomAD. Neither requires a PhD to poke around, though both may make you want one.