At the end of an earlier recording for Moonshots with Peter Diamandis, Philip Johnston mentioned almost in passing that he was planning to send a spacecraft to Alpha Centauri. Returning to the show, he explained the constraints he had set himself, and why they were the hard part: "we really want this thing to actually launch."
Those constraints were four. The probe should cover at least 99% of the distance to Alpha Centauri within 80,000 years — a figure Johnston describes as the one that minimizes fuel, with both longer and shorter journeys costing more. It should launch within three years. It should carry a payload of at least one kilogram. And it should cost less than $15 million to design, build and launch, "because we're basically funding it."
Then came six months of not being able to find a route.
Why the sun runs out
Johnston is founder and CEO of StarCloud, an orbital data center company, and he wanted to fly the mission on hardware he already understood: solar-electric propulsion with a gridded ion thruster, the same sort used on the StarCloud One satellite. An ion thruster is, as he puts it, "essentially a mini particle accelerator" — some are small enough to fit inside a 10-centimeter cube. It flings individual xenon atoms out of the back at very high speed, because propellant that leaves slowly is propellant wasted.
The catch is the power source. Solar panels collect less energy the further a spacecraft gets from the sun, so a probe heading outward needs ever larger panels to keep thrusting. Past roughly Jupiter's distance, Johnston says, the returns become very poor and the panels become impractically large.
His team tried Jupiter flybys and slingshots toward the sun. Two people from NASA's Jet Propulsion Laboratory looked at the problem for a few weeks and could not produce anything that worked. Six months of feeding the problem into AI tools produced nothing either.
One of the show's panelists, Alex, suggested he talk to PSI, the AI physics startup he had co-founded. Johnston, unimpressed by the idea that a young AI company would beat his JPL contacts, did not reply for a week. PSI's Matt Pines followed up and asked for the mission specifications. Johnston sent them, he says, mostly to keep Alex happy.
Slow down to go faster
A week later, PSI came back with a trajectory that ran against intuition. The spacecraft spirals out from Earth orbit into orbit around the sun. Then, instead of pushing outward, it fires retrograde — against its direction of travel — slowing itself so that it falls inward toward the sun. It repeats that at the furthest point of its orbit for about five years, five retrograde burns in all.
Only then does it start thrusting at perihelion, the closest point to the sun. PSI calls the maneuver the perihelion pump. It buys two things at once. Close to the sun, sunlight is intense, so the same thrusters can run on far less solar panel mass. And thrusting while moving fast extracts more energy from a given burn — the Oberth effect — with the spacecraft moving fastest exactly at perihelion.
The mission's FAQ puts numbers on it: retrograde thrust lowers the closest solar approach to 0.42 AU, the craft leaves the solar system at 23.64 kilometers per second, and the trip takes roughly 77,500 years.
An overlooked move, not new physics
Johnston is careful about what was and was not discovered. Firing thrusters at perihelion to exploit the Oberth effect is not new. What surprised him was the way the probe gets close to the sun in the first place. "We were anticipating having to get lower the perihelion through orbital flybys, which is how it's been done in basically every other NASA" mission, he says. Reversing the thrusters and simply slowing down is "the simplest and cheapest and kind of most obvious way to do it. But for some reason, it just didn't occur to any of us."
He says the trajectory has since been checked by a number of trajectory specialists, some of them formerly at JPL. And he concedes the point about human capability: give a group of astrophysics PhDs a billion dollars and five years, he says, and he is sure they would have found it. "It's more that this was done in a week."
What the week consisted of
Pines describes PSI's side as close to unattended. The company has an astrophysicist on staff, but the work was "entirely hands off" apart from prompting the system and preparing the final document's graphics and plots. "We certainly didn't load the dice," he says, describing "almost minimal human steering" before the system returned a trajectory that met the mission's constraints. He puts the total human involvement at "maybe five or six hours of human time over the course of that week."
The computational side was larger. Johnston first guessed tens of billions of tokens; Pines gives a total of about 10 billion tokens, depending on how input and cached tokens are counted. Tokens are the units of text a model processes, and that budget covered Monte Carlo simulations the system designed and ran itself, three-dimensional trajectory models, and an optimization that had to juggle cost and launch windows alongside the orbital mechanics.
PSI ran the campaign on the open-source version of its own tooling, released some months earlier, because Fermi Explorer is structured as an open project. PSI's own site describes a public-benefit corporation where human scientists direct research campaigns carried out by virtual physicists, with independent verification built into its definition of a discovery; it lists Pines as CEO and Alex Wissner-Gross as chief scientist and chief of strategy, and reports more than $58 million raised in a seed round led by Breakthrough Energy Ventures. The mission and the funding were announced the same day.
One of the hosts noted that the ratio — roughly 10 billion tokens of thinking for about 100,000 tokens of final output — matches what he sees on unrelated high-end engineering work in his own building.
A hundred kilograms, mostly fuel
The spacecraft itself is deliberately ordinary. It is about a 100-kilogram smallsat, of which roughly 60% is xenon propellant, using off-the-shelf gridded ion and Hall-effect thrusters. It rides to low Earth orbit as a secondary payload on a commercial launch — Johnston cites about $500,000 for a Falcon 9 ride share — then spirals out over roughly a year and a half, spending about seven kilometers per second of velocity change to reach solar orbit before the retrograde burns begin. Johnston says he thinks the whole thing could be done for $10 million, though the published target remains under $15 million, a figure the FAQ says includes operations. It also favors flight-proven components and sets the launch before the end of 2029.
The destination is where expectations have to be recalibrated. "We expect to miss by quite a large margin," Johnston says. Alpha Centauri sits roughly 260,000 AU away — an astronomical unit being the Earth–sun distance — and covering 99% of that leaves a closest approach of about 2,600 AU, out in the region of a star's comet cloud. The probe is expected to carry retroreflectors so that something there could detect it. By then it will long since be, in Johnston's words, "electronically dead"; the FAQ expects communications to end relatively early, after which the craft loses solar power and coasts on its own.
First to leave, last to arrive
That is the premise Johnston embraces rather than apologizes for. The mission hopes to be "the first to leave Earth for another star, but also the last to arrive at another star." If propulsion improves even modestly over the coming centuries, later craft will overtake it; by the time Fermi Explorer coasts past, he imagines Alpha Centauri has already been settled for millennia. The mission video the panel screened extends the thought outward across the next five billion years — a conservative outer bound, one panelist said, and Alex agreed that better propulsion was a matter of five to ten years rather than a century.
The name points at Enrico Fermi's question about why, if intelligent life should be common, we see none of it. The FAQ frames cheap interstellar departures as a way to probe what stops civilizations from leaving home.
Spacecraft have left the solar system before — Voyager 1 and 2, Pioneer 10 and 11, New Horizons — but the panel's claim for this one is that it is the first aimed at a specific star. Asked about Breakthrough Starshot, Yuri Milner's proposal to push gram-scale sails with ground-based lasers, the hosts said flatly that it had died. One argued it depended on propulsion technology, particularly ultra-high-power lasers, that was simply not ready; Starshot's own concept page describes a kilometer-scale laser array potentially reaching 100 gigawatts, along with unsolved problems of interstellar dust and returning data from a chip-sized transmitter. Fermi Explorer, by contrast, involves "essentially no new technology."
Fermi Explorer is structured as a nonprofit, but Pines has a bet running with the hosts that at $10 to $15 million a mission there is a latent commercial market — that every small government and organization that wants to throw a probe into deep space will now be able to afford one. The panel's parting joke was a live podcast from the launch pad, whenever 2029 arrives, and the challenge one of them set for PSI: help launch this probe, then build something that catches up with it.