Anthropic has assembled the first comprehensive ultraviolet map of the sky using coordinated autonomous agents running through its Claude Science initiative.
Astronomers have struggled with gaps in ultraviolet observations for decades. Earth's ozone layer absorbs UV radiation, forcing instruments into orbit to capture the wavelengths. NASA's Galaxy Evolution Explorer, or GALEX, orbited Earth to survey the heavens but managed to record only roughly two-thirds of the celestial sphere. That mission deliberately bypassed intensely bright star-forming sectors to safeguard its sensitive sensors. The omissions left researchers without a continuous global view of starlight scattering off interstellar dust, stellar debris rings, and clouds surrounding newborn stars.
Automated data pipeline
To resolve the blank spaces, Claude Science deployed agent workflows to manage the raw archival workloads. Johns Hopkins astrophysicist Brice Ménard outlined the methodology on Anthropic's website, explaining that the AI agents retrieved observation files across several space flight archives, calibrated divergent instrument measurements, and synthesized the feeds into an aligned framework.
Where spacecraft never gathered readings, the agents deployed inpainting. The model examined surrounding observed patterns to predict and fill blank coordinates across the celestial sphere. During benchmark checks, the synthetic predictions strayed by an average of roughly ten percent from actual validation measurements.
Unlocking stalled research datasets
The resulting visualization is intended as instructional material for students and researchers. More broadly, the exercise is a demonstration of automated coordination tackling projects that human researchers typically abandon due to sheer logistical friction. Assembling heterogeneous datasets across decades of space missions usually demands months of repetitive data cleansing that research teams lack the labor budgets to pursue.
Ménard suggested that academic teams routinely postpone large-scale compilation tasks due to these administrative hurdles, noting that agent pipelines could clear out long backlogs of fragmented observational archives. Anthropic plans to use the ultraviolet mapping results to show how autonomous agents can operate beyond coding and document synthesis, handling multi-step physical science workflows.
ADFiled by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.