Claude Science Produces First Complete Ultraviolet Map of the Sky

Brice Ménard, an astrophysicist at Johns Hopkins University, used Anthropic's Claude Science to produce the first complete ultraviolet map of the sky. The result, reported by The Decoder, was assembled by AI agents that downloaded imaging data from multiple space missions, calibrated the data, and filled unobserved regions through inpainting. Predictions averaged about ten percent deviation from actual measurements.

What did the pipeline do?

Three operations ran inside Claude Science. AI agents downloaded imaging data from multiple space missions into a single working set. The system calibrated the data. The system filled unobserved regions through inpainting, an estimation approach that infers missing observations from surrounding context.

How close is the reconstruction?

Predictions deviated from actual measurements by about ten percent on average. Ménard reports the figure as the working tolerance of the AI-driven reconstruction, and the one researchers should treat as the working ceiling for this class of AI-driven output.

Why is this a first?

No full-sky ultraviolet map has existed before. Past UV imaging arrived mission by mission, leaving the global sky surveyed in fragments rather than as one image. Claude Science stitched those fragments into a single unified UV map. The reconstruction marks the first time the entire sky has been rendered in ultraviolet light as a single image.

How did Ménard frame the work?

Ménard described the project as research that would not have been completed without AI. He positioned the work as an example of what AI agents can deliver on a multi-mission observational workflow — the kind of project that previously stalled between data release and usable product.

What does the result mean for Anthropic?

Claude Science now holds its first verified astronomy result. The full-sky UV map gives Anthropic a reference point for follow-on work in other wavelengths and higher-resolution reconstructions.

Ménard's ten-percent tolerance now gives AI-driven sky reconstruction a measurable benchmark for follow-on work.