Since 2019, NASA has offered the public free access to an enormous trove of starlight data collected by its Transiting Exoplanet Survey Satellite. Thousands of astronomers downloaded portions of it. Automated pipelines scanned it for repeating dips that signal orbiting planets. But a team of researchers recently fed the entire first-year dataset, all 83 million stellar light curves, through a purpose-built transit search and pulled out 10,091 planet candidates that every prior effort had missed.
Many of those candidates produced only a single dip in brightness as they crossed their host star, a signal type that standard detection algorithms are designed to ignore because it does not repeat on a predictable schedule. The results, described in a preprint posted to arXiv in early 2025, suggest that a vast reservoir of overlooked worlds has been hiding in plain sight in publicly available data.
To put the number in perspective: NASA’s Exoplanet Archive currently lists roughly 5,800 confirmed planets discovered over three decades of searching. If even a fraction of these 10,091 candidates survive follow-up scrutiny, the haul would reshape the known catalog.
Where the candidates came from
The discoveries emerged from the T16 Planet Hunt, a project that reprocessed every full-frame image TESS captured during its first observing cycle in 2018 and 2019. Full-frame images are wide-field exposures the satellite records continuously; they capture light from millions of stars at once, far more than the smaller postage-stamp cutouts that NASA’s own pipeline prioritizes for detailed analysis.
The T16 team built a custom library of systematics-corrected light curves covering 83,717,159 stars brighter than magnitude 16 in the TESS field of view. They then ran a transit search across every one of those curves, flagging signals consistent with a planet blocking a sliver of starlight as it passed in front of its star.
Standard planet-hunting algorithms rely on a technique called box-least-squares fitting, which looks for periodic, repeating dips. That approach works well for planets on short orbits that transit multiple times during an observation window. But planets on longer orbits may cross their star only once in a given TESS sector, producing a lone dip that the algorithm treats as noise and discards. The T16 search was specifically tuned to catch these single-transit events, which make up a significant share of the 10,091 new candidates.
One confirmed planet so far
At least one candidate has already graduated from statistical blip to verified world. The team used radial-velocity measurements from the Planet Finder Spectrograph on the Magellan telescope in Chile to confirm a hot Jupiter orbiting the star TIC 183374187. That confirmation matters because it demonstrates the pipeline can produce real planets, not just instrumental artifacts or eclipsing binary stars masquerading as planetary signals.
The gap between one confirmation and 10,091 candidates is vast. Experience from NASA’s earlier Kepler mission showed that a meaningful fraction of transit-like signals turn out to be false positives: background eclipsing binaries, starspot variability, or detector glitches that mimic the shape of a planetary transit. Scaling the confirmation process to thousands of candidates will require years of telescope time, and most signals will never receive that level of attention.
Separately, NASA has endorsed the broader approach. The agency has documented how deep-learning models applied to TESS data previously identified 370 new exoplanets, confirming that AI-driven candidate identification is now a mainstream tool in the field, not a fringe experiment.
Why ‘impossible’ needs quotation marks
Popular descriptions of these candidates have used the word “impossible,” but no primary source from the T16 team does. What the researchers actually showed is that single-transit events, which are difficult to detect and even harder to confirm, had been systematically overlooked. Whether any of these candidates represent truly exotic configurations, such as planets on extremely long orbits or worlds gravitationally scattered far from where they formed, cannot be determined from transit data alone.
There are also open questions about overlap. Multiple research groups and AI pipelines have been working on the same TESS archive simultaneously. The degree of independent discovery versus duplication between the T16 catalog and other candidate lists has not been publicly quantified. Until cross-matching is complete, the net addition to the known candidate pool could be smaller than the raw number suggests.
Another source of uncertainty is the stellar data underpinning the planet-size estimates. A transit dip reveals the ratio between a planet’s radius and its star’s radius, so any error in the stellar catalog propagates directly into the inferred planet properties. As stellar measurements are refined with data from the European Space Agency’s Gaia mission, some candidates may shift categories, moving from planet-sized to too large, or the reverse.
What comes next for the catalog
The T16 project’s light-curve files are hosted as high-level science products at the Mikulski Archive for Space Telescopes, operated by the Space Telescope Science Institute. The files use FITS format, the standard container for astronomical data, and anyone with an internet connection can download them. That open access is what separates this work from proprietary discovery claims and makes independent replication possible.
NASA’s own documentation of TESS data products independently confirms that the underlying images exist and are freely accessible. The institutional infrastructure lends credibility to the general framework of the discovery, even while the specific candidate list awaits formal peer review.
The real test will not come from the original team’s claims. It will come from independent groups running their own analyses on the same data and, eventually, pointing ground-based telescopes and instruments like the James Webb Space Telescope’s spectrographs at the most promising targets. For single-transit candidates especially, pinning down orbital periods and masses will demand either patience (waiting for a second transit in future TESS observations) or intensive radial-velocity campaigns from the ground.
For now, the catalog is best understood as a large, promising, and unfinished map: 10,091 places in the sky where something blocked a star’s light in a way that looks like a planet. Some of those signals will prove real. Others will dissolve under closer inspection. But the fact that they were sitting in a public archive for years, photographed by a NASA telescope and never flagged, is itself a discovery about how much science remains buried in data we already have.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
