Claude agents stitch together the first full UV map of the sky
A JHU astrophysicist used Claude Science to merge old UV surveys into a full-sky map. About a third of it is modeled, not observed.
The UV sky never had a complete map, and Claude Science has now produced one, though not all of it is observed. Johns Hopkins astrophysicist Brice Ménard, who also works at Anthropic, used it to build a full-sky map in far-UV (154 nm) and near-UV (232 nm). Anthropic published the write-up on Oct 8.
The ozone layer blocks UV, so the data has to come from space. NASA's GALEX (2003–2013) covered about two-thirds of the sky in roughly 38,000 observations. To protect its detectors it skipped the brightest regions, including much of the Milky Way's plane. A team of Claude agents pulled in public data from GALEX, Swift, FIMS/SPEAR and TD-1, then calibrated, cross-matched and merged it. Ménard's own description of his part: "guiding them in the process."
The catch is that about a third of the map is predicted, not measured. That includes much of the galactic plane, the most striking part. The gaps were filled by inpainting, using correlations with visible, infrared and radio data. Tested on real regions that were hidden from the model, the predictions came within about 10% of the actual measurements. UV light from 100M+ stars, estimated from Gaia, was layered on top. Every pixel is labeled measured or predicted. A human also had to catch faint circular artifacts left by individual GALEX pointings, which had passed two rounds of agent review.
Why it matters: AI agents can now do days of tedious data-merging, but the map is a teaching tool, not a new observation, and it still needed a human to check it.
Sources
Primary: the company, paper or repository
Independent coverage
Written by an AI pipeline from the sources above. Methodology · Report an error
Feed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.