Claude Science Helps Build First Complete Ultraviolet Map of the Sky

Anthropic's Claude Science has been used to create what The Decoder describes as the first complete ultraviolet map of the sky. Johns Hopkins astrophysicist Brice Ménard led the project, using AI agents to handle data-heavy work across multiple space missions.

The system downloaded data, calibrated it, merged it, and filled missing regions with inpainting. The result is positioned as a practical example of AI supporting scientific work that might otherwise remain unfinished.

Ultraviolet light is difficult to map from Earth because the ozone layer blocks it, meaning measurements must come from space. The article notes that NASA's GALEX mission covered about two-thirds of the sky but left out bright star-forming regions.

A full ultraviolet map matters because UV observations reveal dust lit by starlight, including clouds around young stars and rings left by stellar explosions. Filling those gaps gives researchers and educators a more complete teaching and reference view of the sky.

Claude Science coordinated AI agents to combine data from multiple space missions and reconstruct missing areas using inpainting. In this context, inpainting means a model learns from available observations and uses that pattern to estimate gaps.

The Decoder reports that the system's predictions averaged about ten percent deviation from actual measurements. That figure makes the project notable not as a replacement for observation, but as a way to turn incomplete archives into usable scientific material.

Ménard sees the ultraviolet map as evidence that AI can help complete tedious scientific projects that researchers may have postponed. The key implication is workflow: AI agents can take on repetitive data gathering, calibration, merging, and reconstruction tasks.

For scientists, the immediate lesson is to look for stalled projects with fragmented datasets and labor-heavy preprocessing. For readers, the story shows AI's value in research is not only discovery, but also making existing data easier to use.