Brice Ménard, an astrophysicist at Johns Hopkins University and a researcher at Anthropic, says he worked with Claude Science to assemble a map of the entire sky in ultraviolet light. His account, published on October 8, 2026, describes a map that combines existing space-based surveys with estimates for regions without direct UV observations.
A UV map of the whole sky
The map combines far-ultraviolet light at 154 nanometers with near-ultraviolet light at 232 nanometers. Ménard’s account says roughly one-third of the sky in the map consists of predictions, including much of the plane of the Milky Way.
Claude’s agents gathered and processed survey data, while Ménard set goals and reviewed the resulting maps. He says the project began “this summer”; the account does not give a more precise start or completion date. Read Ménard’s account.
Why parts of the sky were missing
NASA’s GALEX mission was the largest ultraviolet survey used in the project, according to Ménard’s account. From 2003 to 2013, GALEX imaged about two-thirds of the sky in roughly 38,000 observations. Its observations avoided bright stars and much of the Milky Way’s plane because of the risk of detector damage.
The project also drew on UV observations from NASA’s Swift mission, South Korea’s FIMS/SPEAR and Europe’s TD-1, along with data from ESA’s Planck and Gaia missions. Together, those sources supplied measurements in different parts of the sky and across different surveys.
How the surveys were combined and gaps estimated
Before combining the observations, the team worked to make the surveys consistent: removing glare from bright stars, cross-calibrating instruments, matching image resolution and placing observations in a shared coordinate system.
For areas without UV observations, Claude used a technique called inpainting: estimating missing values from surrounding information. Ménard’s account says the estimates drew on relationships between ultraviolet brightness and visible, infrared and radio observations in regions where UV measurements existed.
The reported test and an artifact correction
To test the estimates, Ménard says Claude was asked to hide portions of regions with known UV measurements, estimate the missing values and compare those estimates with the hidden data. After several rounds of refinement, he reports that the estimates came within about 10% of the real measurements in those tests.
Ménard also noticed faint circular patterns in the processed maps. His account attributes them to residual atmospheric glow in GALEX observations and says Claude traced the issue and applied a correction across all 38,000 observations.
What the finished map includes
The map adds estimated ultraviolet light for more than 100 million individual stars, inferred from visible-light measurements by ESA’s Gaia satellite. Anthropic’s account says viewers can use map layers to distinguish measured pixels from predicted ones and see uncertainty estimates.
