Webb Saw a Planet at Alpha Centauri Then It Vanished

In August 2024, the James Webb Space Telescope pointed at the nearest sunlight star to Earth, blocked its light, and photographed something beside it. An object more than 10,000 times dimmer than the star, sitting about twice as far out as the Earth is from the Sun. A gas giant roughly the mass of Saturn at Alpha Centauri A, four light-years away. If confirmed, it would be the closest planet to another star that anyone has ever directly photographed. Astronomers went back in February 2025 to look again. Nothing. They went back in April. Nothing. The planet was gone. On the 1st of September 2026, NASA switched on an instrument built specifically for this problem, and it is heading to the same point in space where Webb already sits. So, we are about to find out which of two things happened. Start with the system, because almost everything people assume about it is slightly wrong. When you hear the nearest star, you probably think of one point of light. Alpha Centauri is three. Two of them, A and B, are sun-like stars orbiting each other. The third is a small dim red dwarf called Proxima Centauri, which sits further out and happens to be the closest individual star to us. You may have heard of Proxima Centauri B, which gets described as a potentially habitable world. That is a different planet, at a different star in the same system, and it has nothing to do with what follows. Proxima already has confirmed planets, at least three of them. Those were found indirectly, by watching the star wobble as they tug on it, rather than by taking a picture. The star in this story is Alpha Centauri A. It is the closest star to us that genuinely resembles the Sun, and along with its companion it forms the third brightest point in the night sky. So, you would expect it to be the easiest place in the galaxy to hunt for planets. It is close to being the hardest. The problem is brightness. Being nearby means being blindingly bright, and a planet beside a star like that is a moth beside a searchlight. Any light it reflects or emits is drowned before it reaches us. On top of that, these stars drift across our sky faster than almost anything astronomers track, so the target will not hold still. Thousands of exoplanets have been found by indirect methods, watching for a dip in brightness or a gravitational tug. Actually resolving a planet has its own separate point of light, right next to its star, is a different order of difficulty. If you'd rather have the measurements than the headlines, subscribing helps this channel keep doing it that way. So, how do you photograph something beside a light that bright? You block the light. The device that does it is called a coronagraph, and the principle is the one you use without thinking when you raise a hand to blot out the sun so you can see what is beside it. In an instrument, it's a mask, positioned in the optical path exactly where the star's image forms, cutting the glare so that fainter things nearby can emerge. Webb has one on the instrument called MIRI, and MIRI is the reason this observation was possible at all. It works in the mid-infrared, which means it sees heat rather than reflected light. That distinction matters. A gas giant does more than reflect its star. It radiates a faint infrared glow of its own, left over from its formation and from the slow concentration of its interior. To a heat-sensing instrument, that glow is a signature you can hunt for even when the object is drowned in visible starlight. Pulling it off here required a custom observing sequence designed specifically for this target, because Alpha Centauri B sits close enough to A that its light contaminated the measurement. The team had to subtract the light of both stars to see anything at all. And when they did, something was there. An object over 10,000 times fainter than Alpha Centauri A, at a separation of about two astronomical units, which is roughly twice the Earth-Sun distance. Webb was designed to hunt the faint glow of the most distant galaxies in the universe, and picking a planet out of the glare of a bright nearby star was never in the specification. It managed it anyway. You could see the scale of what was involved. The team was subtracting the light of two stars from an image at a wavelength where the telescope itself glows with its own heat, to reveal a point 10,000 times weaker than the brighter of them. Any error anywhere in that chain produces a false point of light, and the team knew it. Before we get to what went wrong, being precise about why this particular candidate caused so much excitement matters, think about what direct imaging usually gets you. Every directly imaged planet found so far has been the same kind of object, young, still hot from formation, and orbiting far from its star where the glare is manageable. Typically, more massive than Jupiter, less than 500 million years old, and beyond 10 astronomical units out. This candidate is none of those things. It sits at two astronomical units, which is close in, and Alpha Centauri A is around 5 billion years old, comparable to the Sun. If it is real, it is a mature planet in a tight orbit around a Sun-like star, which is precisely the category that direct imaging has never been able to reach. One footnote here has nothing to do with science, and everything to do with why people care. In the film Avatar, the moon Pandora orbits a gas giant circling this exact star. For decades, that was invention. Then Webb found a candidate gas giant at that same star, at roughly the kind of distance a storyteller would have chosen. Pandora is fiction, and no evidence exists for moons here or for anything else. But it's a strange thing when reality starts sketching in the outline of an old story. So what should you make of a result like that? The honest answer is that a single image, however good, is where the work starts rather than where it finishes. Science requires that a result be repeatable, and the more consequential the claim, the higher that bar sits. The closest planet ever directly imaged is about as consequential as claims get. So the team went back. February 2025, using Directors Discretionary Time, which is telescope time granted outside the normal competitive process because a result is judged urgent enough, they found nothing. April 2025, nothing again. The signal that had been so striking in August simply did not reproduce. One image, standing alone, is a candidate, not a confirmation. And this is where it gets interesting, because a non-detection leaves the question open rather than closing it. The team ran computer simulations of millions of possible orbits consistent with the original observation, and asked a specific question. Among orbits that stay gravitationally stable in a system with two stars pulling on things, how many would have hidden the planet in early 2025? It came back at 52%. So in more than half the cases, that fit the data. The planet would have swung too close to Alpha Centauri A to be separated from it. Lost in exactly the glare the coronagraph was fighting to suppress. A world on that orbit drifts in and out of view rather than sitting still for us. You should really sit with how uncomfortable that position is for the people involved. They have a result they believe in, published in a serious journal, and they cannot reproduce it. Nobody enjoys that, and the temptation to quietly drop it must be considerable. Instead, they published the simulations that explain the non-detection, which is the harder and better thing to do. To break a tie like this, you cannot simply look again with the same eyes. You need a different kind of eye. On August 30, 2026, a Falcon Heavy launched the Nancy Grace Roman Space Telescope from Kennedy Space Center. It is named after NASA's first chief of astronomy, the woman, usually called the mother of Hubble, who spent a career arguing that a great space observatory was worth building. Roman carries two instruments. One is a wide-field survey camera, which is what most of the mission is about. The other is a coronagraph instrument, and it is the reason this telescope belongs in this story. On September 1, 2026, NASA confirmed that the coronagraph had been powered on. The activation began at 7.27 in the morning Eastern Time and completed at 8.22. It is the first coronagraph ever flown with active wavefront control, meaning it carries deformable mirrors that correct optical distortions in real time, adjusting their shape continuously while observing. Nothing like it has operated in space before. And this is why the pairing matters. Webb saw the candidate in the mid-infrared, in heat. Roman's coronagraph works in visible light, in reflected starlight. Those are two entirely different fingerprints of the same object. If Roman catches it, the two measurements together would give something neither could produce alone, constraints on the planet's actual size and on how reflective its atmosphere is. Heat tells you about internal energy, reflected light tells you about surface and cloud. Two telescopes, two kinds of light, one point of doubt four light-years away. And you can see why the timing feels almost arranged. The instrument capable of settling a question that has been open since 2019, powered on eight days after launch, while the argument was still live. That is coincidence rather than planning, but it does mean the weight is now measured in months instead of years. Four things need saying plainly, because coverage of this story has been getting all four of them wrong. The first is that the discovery predates Webb. The candidate at Alpha Centauri A was directly imaged as early as 2019 from the Very Large Telescope in Chile and was reported in 2021 by a team led by Kevin Wagner. That observation was too short to be conclusive. What Webb provided is the strongest evidence to date, which is a different and more careful claim. Second, the observation is not recent. Webb looked in August 2024, the results were announced in August 2025, and the follow-up attempts happened in between. Anyone presenting this as breaking news is describing a two-year-old measurement. Third, Roman's coronagraph is formally a technology demonstration. Its stated purpose is proving that this class of instrument works in space, to inform future missions designed for imaging Earth-like planets. The team has three months of pre-planned observations spread across the first 18 months. Nobody has promised that it will walk up to Alpha Centauri and settle this. Fourth, the candidate remains a candidate. It fits the data, the simulations and a plausible orbit, and it might equally be an artifact of an extraordinarily difficult measurement pushed to the limit of what the instrument can do. Both remain live. One more thing worth knowing about where both telescopes will be sitting. L2 is empty. It's a pointed space where the gravity of the Sun and the Earth combine so that a spacecraft can hold station with almost no fuel. Always on the night side of our planet, always pointed away from the light. Webb has been parked there since 2022. Roman arrives in weeks. Two of the most capable instruments ever built, at the same address, able to take turns on the same star. So where does that leave the planet next door? An object exists, or existed for a few days in August 2024, beside the closest Sun-like star to Earth. Saturn-sized, sitting where a smaller world might hold liquid water, 10,000 times fainter than the star it orbits. It has refused to sit for a second portrait, and half the orbits that fit the data explain exactly why. Roman reaches its position at the second-largest point later this month, about a million miles out, the same gravitational parking spot where Webb already sits. Then roughly 90 days of commissioning. First images are expected in early 2027. Which means the question has a deadline for the first time since 2019. Either the planet drifts back out of the glare and two telescopes catch it in two kinds of light, or it stays hidden and the argument continues. If you want the full picture of the telescope that just switched that instrument on, and what else it has built to find, that video is on screen now.