Claude Science builds a complete ultraviolet map of the sky
Event confirmedAwaiting review
The takeaway
On Oct 8, 2026 Anthropic published a post by Johns Hopkins astrophysicist Brice Ménard (also an Anthropic researcher) describing how many Claude agents working in Claude Science assembled a full-sky ultraviolet map.
Status
- Claim
Event confirmedAwaiting review
- Our reporting
- High confidence
- Verification
- Unverified claim
- Importance
- 2 of 5
- Last verified
- 9 October 2026
Your AI and this story
- GPT-6 Astra161 days after its cutoff
- Claude Opus 5.5100 days after its cutoff
- Gemini 3.8 Flash191 days after its cutoff
- Grok 4.7130 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 100 days before it.
Key facts
- Published Oct 8, 2026 on Anthropic’s research blog; author Brice Ménard, ‘an astrophysicist at Johns Hopkins University and a researcher at Anthropic’
- Inputs: NASA GALEX (about 38,000 observations covering roughly two-thirds of the sky), NASA Swift, the Korean FIMS/SPEAR instrument, the European TD-1 satellite, and ESA’s Gaia and Planck catalogs and maps
- Output: an all-sky map in the far-UV (~154 nm) and near-UV (~232 nm) bands; about a third of the sky is ‘inpainted’ (predicted), using more than 100 million Gaia stars and Planck dust data
- Check: in deliberate occlusion tests on regions with real data, the predicted UV was within about 10% of the measurements; Ménard also inspected the outputs by eye and caught artifacts such as atmospheric (airglow) contamination
- Process: ‘many agents working in Claude Science’ over several days of iterations, individual runs lasting hours; Ménard gave high-level instructions and reviewed the results. The post names no Claude model version
- Ménard: ‘For once, I was able to pursue a project like this without sacrificing time I would otherwise have devoted to research.’ Before, ‘the only UV map I could show was full of holes’
- The map is published as an interactive viewer at menard.pha.jhu.edu/uvmap; no paper, preprint or peer review is mentioned
What happened
Anthropic’s research blog published “The missing map of the sky” on October 8, 2026 (Anthropic1). Brice Ménard, an astrophysicist at Johns Hopkins and an Anthropic researcher, describes a long-wanted project: an ultraviolet map of the whole sky. GALEX, the main UV survey, left about a third of the sky unobserved. Claude agents in Claude Science merged GALEX with other UV missions and trained a model that predicts UV brightness from Gaia’s catalog of more than 100 million stars and Planck’s dust maps, then used it to fill the gaps. When regions with real data were hidden on purpose, the predictions matched the measurements to within about 10%.
The work took several days of iteration. Ménard gave the direction, checked the outputs and caught problems such as atmospheric glow in the data; the agents did the rest. The map is online as an interactive viewer.
Why it matters
It is a small but concrete example of agentic AI doing the data-heavy “infrastructure” work that scientists often postpone for years. It is not a discovery in itself, and it has not been through peer review. It is one of a series of Anthropic posts in autumn 2026 showcasing Claude Science on real research problems.
What is disputed or not yet verified
| Verification | Author’s occlusion tests (~10% accuracy) and visual inspection; no paper or peer review yet |
|---|
Sources
3 sources from 3 sites. Numbers match the chips in the text.
3 sources: 3 primary
Primary
- Anthropic: The missing map of the sky (Using Claude Science to produce the first complete map of the sky in UV light)anthropic.com, official
- Interactive UV sky map (Brice Ménard, Johns Hopkins)menard.pha.jhu.edu, code
- GALEX mission data (MAST archive)archive.stsci.edu, docs
Changes
- Filed (Anthropic research post)