AI Specialist Finds Possible Exoplanets Using Claude Code in Two Weeks

Two possible exoplanets orbiting a star 116 light years from Earth were discovered by an artificial-intelligence specialist using a coding bot in about two weeks, Nature reported.

Pavel Rabtsevich, an artificial-intelligence specialist based in Spain with no formal astronomy education, fed telescope data from more than 120,000 stars into Claude Code, an AI coding tool made by Anthropic, according to Nature.

The bot zeroed in on TIC 4206066, a star a little smaller than the Sun and located roughly 116 light-years from Earth, after scanning observations collected by NASA’s Transiting Exoplanet Survey Satellite, Nature reported. The system identified two separate cycles of dimming, with one roughly every 3 days and another every 11 days, Nature reported.

“I didn’t take the models’ first answers,” Rabtsevich told gizmodo.com. “I pushed them, round after round, to improve their own methods, and checked what they gave me against other models.”

Scanning Data for TIC 4206066

Rabtsevich directed the AI agent to sift through archival NASA Transiting Exoplanet Survey Satellite data from late 2025 and early 2026, using a popular planet-detection algorithm to search for regular dips in stellar brightness. The tool flagged TIC 4206066 for showing repetitive patterns of dimming every 3.18 Earth days and every 11.13 Earth days, Nature reported.

Further checks by Rabtsevich using older data from 2018 and 2020 recorded the exact same frequency of dimming, Nature reported. He calculated that the primary candidate is a super-Earth about 1.4 times larger than Earth, while the second signal points to a tentative second candidate roughly 2.2 Earth radii across, Nature reported.

Story on The Morning Science
Photo: themorning.science

Rabtsevich ran 74 tests with Claude Code to challenge his hypothesis and even had the model correctly predict dimming occurrences across separate datasets, gizmodo.com reported. He subsequently utilized an open-source validation algorithm written by University of British Columbia exoplanet researcher Michelle Kunimoto to vet the signals, Nature reported.

Verifying Exoplanet Candidates

While citizen scientists have spent decades leveraging publicly available tools from NASA and other researchers to hunt for worlds outside the Solar System, advanced AI capabilities have lowered the barrier to entry for non-specialists, Nature reported. This newfound accessibility has sparked concerns among professional researchers.

“There’s been a huge, overwhelming surge of AI-driven papers, which tend to have [planet] candidates that are not the best quality,” Kunimoto told Nature. Because some amateur hunters do not fully understand the underlying algorithms, they struggle to validate models properly, leading Kunimoto to note that “the systems we use to disseminate science are being overwhelmed.”

On September 25, Rabtsevich published his findings on Zenodo as a preprint, Nature reported. Although the study has not undergone peer review, professional TESS astronomers completed an independent analysis of the signals and added them to their official list of planet candidates, Nature reported.

Targeting Space Telescope Observations

Key Details of TIC 4206066 Exoplanet Candidates
Parameter Candidate One Candidate Two
Orbit Period 3.18 Earth days 11.13 Earth days
Estimated Size 1.4 Earth radii 2.2 Earth radii
Confidence Level Stronger evidence Tentative

Rabtsevich submitted a formal request to MIT asking that the space telescope be positioned to observe TIC 4206066 during its next operational window, gizmodo.com reported. The request was approved, scheduling the satellite to measure the star’s brightness every two minutes from October 31 to November 26, gizmodo.com reported.

AI Specialist Finds Possible Exoplanets Using Claude Code in Two Weeks
Photo: gizmodo.com

On October 7, 2026, Rabtsevich publicly shared that he used Claude Code to locate the planet candidate around the target star, Nature reported. Earlier in the week, he published the exact timetable of expected dimming events for the ongoing November observation window, noting on X that any deviation from the predicted schedule would also be shared publicly, gizmodo.com reported.

Questions About AI Planet Hunting

How do amateur astronomers find exoplanets using AI?

Amateurs feed massive public archives from space telescopes like NASA’s TESS into artificial intelligence coding assistants such as Claude Code, prompting the tools to isolate periodic dips in stellar luminosity, according to gizmodo.com and Nature.

What is the false positive probability for the TIC 4206066 candidate?

The preprint reports a false positive probability between 0.03 and 0.04 using TRICERATOPS, a standard evaluation tool, though the author emphasizes this does not fully close the case due to unruled companion star assumptions, Nature reported.

From October 31 to November 26, the telescope was due to measure the star’s brightness every two minutes, gizmodo.com reported.