Rubin Anomalies

LSST × ANTARES

About this site

How it works, where the data comes from, and who made it.

What is LSST?

The Legacy Survey of Space and Time (LSST) is a ten-year optical survey of the southern sky conducted by the Vera C. Rubin Observatory in Chile. Beginning science operations in 2025, LSST captures the entire visible southern sky every few nights with a 3.2-gigapixel camera — generating roughly 10 million transient alerts every night. Each alert represents a source that changed in brightness or position since the previous observation.

The scale of this dataset is hard to overstate. In a single night, Rubin will detect more transient events than all previous optical surveys combined. Most alerts are familiar phenomena — variable stars, known asteroids, supernovae following predictable light curves. But a small fraction behave in ways that don't fit any known template. Those are the ones this site is looking for.

What is ANTARES?

ANTARES (Arizona-NOAO Temporal Analysis and Response to Events System) is an alert broker developed by NOIRLab that ingests the raw Rubin and ZTF alert streams and adds contextual annotations. For each event, ANTARES computes derived properties, applies machine learning classifiers, cross-matches against known catalogs, and tags objects with labels that reflect their behavior — things like nuclear_transient, dwarf_nova_outburst, or LAISS_RFC_AD_filter (flagged as anomalous by a trained neural network).

ANTARES also produces a per-locus anomaly score (0–1) that reflects how unusual an object's light curve is compared to the broader variable-star population. This site queries the ANTARES public search API daily, pulling the most recently updated LSST loci and running them through a local scoring filter.

Methodology

Each locus retrieved from ANTARES is assigned an interest score from 0 to 100 based on a combination of its tags, the ANTARES anomaly score, and its photometric properties. The top six from each daily poll are surfaced here.

Signal Points Why it matters
Tier 1 tag +25 each Rare, specific labels: tidal disruption events, ML anomaly flags, super-luminous supernova candidates, dwarf nova outbursts, and similar high-value signals
Tier 2 tag +10 each Broader labels — nuclear transients, extragalactic sources, fading events — that are interesting when combined with other signals
ANTARES score up to +40 The ANTARES 0–1 anomaly score scaled to 40 points
Outlier classification +20 ANTARES has classified this object as an outlier — its light curve does not match known variable star classes
Amplitude > 2.0 mag +25 Dramatic peak-to-peak brightness change in the r-band
Amplitude 1.5–2.0 mag +15
Amplitude 1.0–1.5 mag +5
Known boring type −40 RR Lyrae, Cepheids, eclipsing binaries, and standard AGN are deprioritized — they are well-understood and common

Only objects scoring at or above 40 are saved. The daily snapshot always shows the six highest-scoring objects discovered in the past 24 hours. If no new objects qualify on a given day, the previous snapshot is retained.

How it runs

1
Poll — A Python background process queries the ANTARES search API every 24 hours, fetching all LSST loci updated since the previous run.
2
Score and filter — Each locus is scored as described above. Objects above the threshold are saved to a local SQLite database with their coordinates, tags, and photometric properties.
3
Summarize — The top-scoring new object from each poll receives a plain-English summary generated by Claude (Anthropic). If the API is unavailable, a deterministic fallback summary is generated from the object's properties.
4
Snapshot — The six highest-scoring objects from the past 24 hours are recorded as a snapshot and served by a Flask web app via a JSON API.

Author

Built by Jason Keller — a developer interested in data pipelines, APIs, and building things that make complex data explorable. You can find more of his work at jaskeller.com.

Disclaimer

This is an independent personal project. It has no affiliation with NOIRLab, the ANTARES team, the Vera C. Rubin Observatory, or any of their partner institutions. Data is accessed via the public ANTARES API and sky imagery via the DESI Legacy Surveys. Nothing on this site should be taken as an official scientific result or endorsement.