Pavel Rabtsevich spent weeks analyzing public astronomical data from NASA's Transiting Exoplanet Survey Satellite, relying on artificial intelligence models like Anthropic's Claude Code and OpenAI's Codex to sort through measurements of over 126,000 stars. Lacking formal astronomical training or a physical telescope, the 28-year-old product manager directed multiple AI systems to hunt for stellar dimming patterns that typically signal an orbiting exoplanet passing across a star's face.
By iteratively refining the models and running dozens of validation tests, Rabtsevich focused on TIC 4206066, a star located approximately 116 light-years away from Earth. The data revealed a consistent dip in brightness every 3.18 Earth days. Historical datasets from 2018 and 2020 confirmed the recurring pattern, leading him to calculate the presence of a planet roughly 1.4 times the size of Earth with scorching surface temperatures around 1,000 degrees Fahrenheit. He also flagged a secondary, less certain dimming cycle occurring every 11.13 days.
After submitting a detailed report to MIT, Rabtsevich secured approval for TESS to re-observe the target sector between October 31 and November 26. While institutional researchers note that AI models primarily excel at filtering vast candidate pools rather than final confirmation, the democratization of powerful language models allows amateur researchers to process galactic-scale datasets that once required extensive manual spreadsheet labor.
Comments