Eighty thousand years in transit — and that’s the hopeful scenario.
That figure belongs to the Fermi Explorer Mission. The nonprofit, which is announcing itself today, plans to send a spacecraft toward Alpha Centauri before the end of 2029. The star system lies 4.4 light-years away, and should everything go to plan, the probe may need as long as 80,000 years to arrive.
The path it will follow, though, didn’t come from the team’s engineers. An AI system produced it.
The trick is going toward the sun first
What the AI came back with runs against intuition. Rather than speeding away, the spacecraft brakes, allowing its orbit to dip nearer the sun than Mercury ever travels. At every close approach, it lights its engine. Down there the solar panels soak up four times as much light, and thrust applied while moving fast yields more energy than the identical burn performed elsewhere.
Since the engine fires only in the sun’s vicinity, the panels don’t need to be large. Smaller panels keep the spacecraft light. A light spacecraft is what keeps the $15 million budget intact.
Every one of these maneuvers is already known; stringing them together this way is not. According to a paper that hasn’t gone through peer review, the Fermi team had never weighed the idea.
A year of failing first
For a full year, Johnston and his colleagues searched without success for a route that would carry a small, solar-powered craft to Alpha Centauri on a budget of just $15 million. Power was the obstacle they could never get past. Supply the spacecraft with enough of it and mass piles on, and extra mass means extra fuel burned.
Johnston eventually raised the problem on a podcast run by Alex Wissner-Gross, a physicist and cofounder of Physical Superintelligence. Wissner-Gross suggested feeding it to the lab’s system, Get Physics Done — open-source software that accepts a physics research question, splits it into smaller pieces and determines which simulations to execute, drawing on AI models such as Anthropic’s Claude or OpenAI’s GPT.
Seven days later, the trajectory existed. Johnston didn’t see it coming.
Three days, a billion tokens, one babysitter
According to Matt Pines, PSI’s cofounder and CEO, the system ran largely unattended for three days and consumed a billion tokens. Largely, but not entirely: a staff astrophysicist at PSI nudged it toward the mission’s constraints, requested a cost breakdown and better charts, and combed the output for mistakes.
“The fact that it came up with an entirely different mission profile, one that was creative and not one [the Fermi team] had considered—that was the more surprising aspect,” Pines said.
He doesn’t oversell the limitations. What the model lacks is the judgment and taste a human researcher brings. It can’t reliably tell which questions matter or which avenues deserve effort, which leaves it prone to burning time on dead ends or never branching out to alternative approaches.
“I don’t think we’ve yet figured out how these models can internally represent something like that,” Pines said of research judgment.
PSI is launching today on $58 million in funding, a round led by Breakthrough Energy, the climate-oriented investment group started by Microsoft cofounder Bill Gates.
The last attempt at this never left the ground
Back in 2016, billionaire tech investor Yuri Milner unveiled Breakthrough Starshot, an interstellar effort to send humanity’s first spacecraft to Alpha Centauri. The plan called for powerful lasers to accelerate miniature probes to a fifth of light speed — quick enough for a 20-year journey. Milner committed $100 million to a proof of concept.
Ten years on, nothing has flown.
“We didn’t want to do another Breakthrough Starshot,” said Philip Johnston, cofounder and president of the Fermi Explorer Mission. “We’re dead set on something actually launching.”
The $15 million behind Fermi, presently supplied by individual private donors, amounts to a fifteenth of what Milner staked on a proof of concept alone. Fermi closes that gap by abandoning any ambition of speed.
“We are not constraining ourselves to doing it in a human lifetime,” Johnston said. “Let’s just figure out the way to get to another star.”
The scale problem, in Voyager units
Roughly 25 trillion miles separate Earth from Alpha Centauri. Voyager 1 ranks among the fastest objects humanity has ever sent up, and since 1977 it has crossed under 1% of that gap. Traveling at that pace, the journey would run past 70,000 years.
Fermi’s spacecraft is slated to carry a payload of at least one kilogram: scientific and artistic cargo, messages, and a duplicate of the Golden Record — the gold-plated disc of Earth’s sounds and images that NASA fixed to its pair of Voyager probes in 1977 as a note to any future finder.
Johnston assumes someone will overtake them. “We’re pretty confident that we will not be the first to arrive,” he said, on the reasoning that spacecraft technology will keep advancing. The arithmetic isn’t subtle: an engine built a thousand years from now that is merely 20% faster than today’s would deliver a later probe to Alpha Centauri upwards of 10,000 years before Fermi.
What launching would prove about the silence
The mission also takes aim at a question Enrico Fermi raised in 1950. Hundreds of billions of stars fill the galaxy, and most are far older than the sun. Even a civilization inching its way from star to star could cover the entire galaxy within a few million years — a blink relative to the galaxy’s age. So evidence of intelligent life, if it exists, ought to have turned up already.
Two convenient explanations present themselves: getting to another star is simply too difficult, or nobody else has cared to try. The moment Fermi’s probe lifts off, we turn into a civilization that is both able and willing to reach another star — and neither explanation survives as a reason the galaxy remains unexplored.
The remaining options are grimmer, Johnston said. Perhaps life resembling ours is vanishingly rare. Or perhaps intelligent life is widespread but usually dies off before it can spread.
“One of those reasons could be that once you hit superintelligence, that for some reason is self-destructive,” Johnston said. “Maybe in the next 50 years, there’s some great filter that we do not pass through. That all intelligent civilizations, for some reason, do not pass through.”
An odd sentiment, coming from someone whose path to the stars was charted by an AI.