Adam Kovac, writing for Scientific American, reports on an amateur astronomer who used AI agents to pull a possible exoplanet out of public NASA data. Pavel Rabtsevich, a product manager based in Spain, ran a custom machine learning setup over more than 126,000 target stars observed by the Transiting Exoplanet Survey Satellite. It flagged TIC 4206066, a star about 116 light-years away in Orion that dims for roughly two hours every three days.

Digging through the TESS archives turned up 22 more dimming events. If confirmed, the planet would be a super-Earth about 1.4 times the radius of our planet, with a surface temperature near 1,000 degrees Fahrenheit. Rabtsevich is careful about the claim. Some 2018 dips line up with moments when TESS fired its thrusters.

He says the models wrote the code, ran it and checked each other’s work, while the questions and decisions were his. The signal was not entirely new: a TESS detection filter caught it in 2020 and rejected it for lack of evidence.…