Amateur Astronomer Leverages Multi-Agent AI to Detect Hidden Exoplanet Candidate
Public NASA TESS archive analysis surfaces a potential super-Earth orbiting star TIC 4206066.
1AI Pipeline Scans Star Archives
Pavel Rabtsevich analyzed light curve observations across 126,000 target stars using custom artificial intelligence agents. The automated system parsed open data from NASA's Transiting Exoplanet Survey Satellite to isolate anomalous transit signatures. The model flagged star TIC 4206066, situated 116 light-years away in the Orion constellation, exhibiting periodic two-hour dips.
2Physical Profile of the Candidate
Further analysis across historical TESS records revealed 22 distinct transit events linked to this host star. If validated, the object represents a rocky super-Earth approximately 1.4 times Earth's radius. Due to its tight three-day orbital period, researchers estimate the candidate planet maintains scorching surface temperatures near 1,000 degrees Fahrenheit.
3Multi-Agent System Architecture
Rabtsevich deployed several specialized machine learning models to write, execute, and cross-validate scientific code. While the autonomous agents handled raw computational screening, Rabtsevich defined data constraints, candidate selection criteria, and publication rules. Yale astronomer Greg Laughlin noted that large language models like Claude and Codex excel at literature data extraction.
4Signal Filtering and Noise Artifacts
Transit signals often mimic hardware anomalies or astrophysical noise, requiring rigorous vetting procedures. Rabtsevich noted that certain light dips in 2018 coincided precisely with TESS thruster firings, which risked skewing photometric measurements. He emphasized relying strictly on reproducible calculations, open control samples, and verifiable signals rather than raw model consensus.
5Institutional Tracking and Convergence
MIT researcher Pablo Mercader Perez revealed that TESS algorithms initially flagged the star in 2020 but dismissed it due to insufficient signal-to-noise ratios. Subsequent observational sectors provided denser photometric data, enabling MIT deep learning pipelines and Michelle Kunimoto's LEO-Vetter software to independently detect the planet candidate earlier this year.
6Target Scheduled for Re-Observation
Following the open-source analysis, NASA's TESS mission team scheduled TIC 4206066 for targeted re-examination once spacecraft orbital geometry aligns this November. The star now sits on TESS's high-cadence observation schedule to confirm signal authenticity. Rabtsevich also shared preliminary data indicating a possible secondary exoplanet situated in a wider orbit.
7Emergence of Citizen AI Astronomy
The discovery illustrates how consumer-accessible AI frameworks empower citizen scientists to process petabyte-scale public astrophysics archives. Multiple independent researchers are now deploying similar multi-agent architectures against public NASA datasets. This decentralized algorithmic screening relieves significant analytical bottlenecks for academic institutions by triaging subtle transit signals overlooked by standard institutional pipelines.
8Implications for Exoplanet Discovery
Validating this candidate super-Earth demonstrates that mature machine learning workflows can extract verified physical signals from historical space telescope data. As space agencies collect vast telemetry, human-guided autonomous agents offer scalable filtering for astronomical discovery. The upcoming November observations will provide the definitive radial velocity and transit data required for official confirmation.