Anthropic's Claude Science creates the first complete ultraviolet map of the sky

Johns Hopkins University astrophysicist Brice Ménard used Anthropic's Claude Science to build the first complete map of the sky in ultraviolet light, a project described on Anthropic's website. Claude Science coordinated AI agents that pulled data from several space missions, calibrated it and merged the datasets into a single map. Missing regions were reconstructed with inpainting, a technique in which a model learns from existing data to fill gaps.

The ultraviolet sky has never been mapped in full because Earth's ozone layer blocks UV light, so the measurements have to be taken from space. NASA's GALEX mission covered roughly two-thirds of the sky but skipped bright star-forming regions, leaving the map incomplete. UV observations trace dust that is lit up by starlight, including clouds around young stars and rings left behind by stellar explosions.

When the reconstructed gaps were checked against actual measurements, the predictions averaged about ten percent deviation. Ménard frames the map primarily as teaching material, and he suspects many scientists have been putting off similar data-heavy projects that AI could now make tractable.

Why it matters: research teams that depend on assembling and cleaning data from scattered instruments may increasingly treat AI agents as the layer that does that work, which can make previously stalled sky-surveys or archive-merging projects feasible. The reported ten percent average deviation is a useful benchmark, but for scientific use a developer or researcher would still need to know how that error distributes across the sky, since poorly covered bright regions are exactly where reconstruction is hardest.