Astronomers Use Powerful New AI to Uncover 118 Planets Hiding in NASA's Data
New analysis has turned NASA’s TESS mission from a targeted planet spotter into a broad census machine, with AI uncovering undiscovered worlds hiding in plain sight within the mission's data. For one, fresh tools like RAVEN have validated 118 planets and surfaced more than 2,000 high-quality candidates.
Astronomers at the University of Warwick employed the AI-tool RAVEN to examine data from more than 2.2 million stars
and validated 118 new planets, including 31 newly identified worlds,
while also producing more than 2,000 vetted candidates for future
follow-up. Its mega-haul includes ultra-short-period planets that race
around their stars in under a day and rare objects in the so-called
Neptunian desert, where theory expects planets to be scarce.
Elsewhere, another study of TESS data has made massive head ways, too. The T16 planet hunt project reports 10,091 new planet candidates from 83,717,159 light curves, a scale that's hard to overstate. The team, led by Joshua Roth, sifted through first-year TESS full-frame images and identified 11,554 candidates orbiting stars with periods from 0.5 to 27 days; more than 1,000 of those were already known, which means the rest represent a major expansion of the candidate pool.
The authors say the haul more than doubles the number of known TESS candidates, and they confirmed one of them with Magellan: a hot Jupiter around TIC 183374187, a metal-poor thick-disk star, proving the pipeline can pull real planets out of the noise.
While it's cool that astronomers have found more planets, it's how they found them that has changed the rules of the hunt by looking far beyond the bright, obvious stars that have dominated previous exoplanet surveys. For example, T16 pushed into stars as faint as 16th magnitude in the TESS band, a region of the sky that earlier searches largely left unexplored.
The "old" way of exoplanet discovery relied on patience, bright stars, and narrow targeting; the new one depends on machine learning, bulk data, and a willingness to search where the universe has been faint and inconvenient. We now have the computational muscle look hard into places and discover worlds and objects that previously remained invisible.
Therefore, one can say that the most important consequence of these results may be methodological rather than numerical showcases. If surveys like T16 and RAVEN keep scaling up, exoplanet astronomy stops being a game of isolated discoveries and becomes a statistical map of planetary systems across the galaxy.
