# Full-Sky Ultraviolet Map (2026)

> Source: https://aiwiki.ai/wiki/full_sky_ultraviolet_map_2026
> Updated: 2026-10-10
> Fact-checked: 2026-10-10
> Categories: AI Research, AI for Science, Anthropic, Data & Datasets
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Full-Sky Ultraviolet Map (2026)." aiwiki.ai, 10 Oct 2026. https://aiwiki.ai/wiki/full_sky_ultraviolet_map_2026
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

The **Full-Sky Ultraviolet Map (2026)** is an astronomical map produced by Brice Ménard, an astrophysicist at Johns Hopkins University and researcher at [Anthropic](https://aiwiki.ai/wiki/anthropic), with [Claude Science](https://aiwiki.ai/wiki/claude_science). It combines ultraviolet observations with estimates for missing regions. Anthropic announced the project on October 8, 2026, as an example of [AI for Science](https://aiwiki.ai/wiki/ai_for_science).[1]

## Observations and coverage

NASA's Galaxy Evolution Explorer (GALEX), launched on April 28, 2003, observed ultraviolet light to study the history of star formation.[3] Its observations supply most of the map. GALEX avoided ultraviolet-bright stars to protect its detectors, leaving gaps around stars and in the Galactic plane.[1]

| Property | Specification |
|---|---|
| Map date | September 16, 2026 [2] |
| Grid | Galactic coordinates; approximately 1.7-arcminute pixels [2] |
| Predicted sub-degree morphology | Approximately 28% of the far-UV sky and 27% of the near-UV sky [2] |

## Claude Science workflow

Ménard directed Claude Science agents to locate public surveys, download observations, correct instrumental differences, and combine the datasets. The pipeline harmonized GALEX, Swift-UVOT, and FIMS/SPEAR observations and incorporated stellar flux estimates from Gaia.[1]

Human inspection remained necessary. Ménard noticed faint circular footprints from individual GALEX observations after two agent reviews had missed them. Claude then corrected residual atmospheric glow across roughly 38,000 observations. The project went through more than a dozen versions over several days.[1]

## Statistical prediction

Missing diffuse emission uses [gradient-boosted regression](https://aiwiki.ai/wiki/gradient_boosting) with 90 features from dust, Hα, Gaia starlight, and position. It is not a language model or image generator. A separate layer adds measured or predicted ultraviolet fluxes for about 119 million Gaia stars.[2][4]

## Validation and scientific limits

Provenance weights identify contributing layers. Held-out prediction errors were typically 12-14%, increasing farther from observations. Predicted resolution is roughly 0.5-1 degree, much coarser than pixel spacing. Filled regions cannot yield new-source discoveries or independent dust measurements.[2][4]

The project had not undergone human peer review; its validation reports came from the system that built it.[2]

## Access and licensing

The project website provides the visual map and a draft technical manuscript. On October 10, 2026, it said FITS and HiPS download links would be added. It specifies CC BY 4.0 for data products and MIT for code.[2]

## References

1. Ménard, Brice. ["The missing map of the sky"](https://www.anthropic.com/research/the-missing-map-of-the-sky). Anthropic, October 8, 2026.
2. Ménard, Brice. ["The full sky in the ultraviolet"](https://menard.pha.jhu.edu/uvmap/). Project website, October 2026. Accessed October 10, 2026.
3. NASA Science. ["GALEX"](https://science.nasa.gov/mission/galex/). Updated August 1, 2023.
4. Ménard, Brice. ["Full-sky ultraviolet maps from harmonised GALEX, Swift-UVOT, FIMS/SPEAR, TD-1 and Gaia data"](https://menard.pha.jhu.edu/uvmap/docs/uvmap_manuscript.pdf). Draft technical manuscript, October 2026.