NASA squeezed Treasury, vibe coded, and broke the mold in bid to save Swift

An attempt to save NASA’s $500 million Neil Gehrels Swift Observatory from dropping out of orbit fell short, but the behind-the-scenes machinations required just to make a rescue possible deserve recognition, government and commercial space officials said in a recounting of the mission.

The rescue mission was developed by Katalyst Space Technologies, working under a $30 million contract awarded by NASA last September. NASA gave Katalyst nine months to build and launch a satellite to capture Swift flying some 200 miles above the Earth and boost it into a higher orbit.

Katalyst’s rescue spacecraft, known as Link, successfully launched July 3 and completed some of its initial checkouts before a malfunction caused it to spin out of control a few weeks later. NASA announced in August that Link would not be able to reach Swift as intended, and the rescue mission was aborted.

Still, designing and building a satellite, finding a rocket to launch it, and getting it to work in the harsh environment of space was a remarkable achievement. It would usually take several years to design, build, and test a satellite of Link’s size and complexity from scratch. Link weighed nearly a half-ton at launch, with electric thrusters, robotic arms, and large deployable solar arrays.

In the end, though, the Link satellite did not reach the Swift Observatory. The failure was bad news for scientists who rely on data from Swift, which has unique capabilities combining sensitivity to gamma rays, X-rays, and visible light with the agility to quickly turn toward astronomical targets. This has made Swift the go-to observatory for detecting gamma-ray bursts, the most powerful explosions in the Universe.

But Swift was going to fall out of orbit anyway. The spacecraft launched in 2004 and has far outlived its original design life. Swift does not have its own propulsion system, so it is unable to raise its orbit without help, leaving its fate at the mercy of aerodynamic drag in low-Earth orbit. Increased solar activity in recent years led to an increase in drag at Swift’s altitude, hastening its demise.

Racing against the clock

NASA and Katalyst officials are reviewing lessons learned from the mission to better prepare for the next time they need to rapidly call up a satellite rescue mission. First and foremost: Don’t wait until the last minute, said Shawn Domagal-Goldman, director of NASA’s astrophysics division, in a meeting last week of the National Academies’ Committee on Astronomy and Astrophysics.

“The way that the timeline worked out, we didn’t have the ability to open up the doors fully to proposed solutions. We went with the teams we had on [contract] already. That is not a regret of the team we ended up with. It’s just I think I prefer the more open solutions.”

NASA commissioned 30-day studies from three teams—Katalyst, Starfish Space, and a joint proposal from Cambrian Works and Astroscale—in August 2025 to show how they planned to rescue Swift from destruction. The teams eligible for the studies were limited to partners NASA already had on contract for technology development. Under federal acquisition rules, it would have taken months or longer to start a brand new procurement and bring on a new provider.

The studies led to NASA’s selection in September 2025 of Katalyst to try to rescue Swift. Officials knew the odds were stacked against them.

Katalyst quickly put out orders to suppliers for all the parts required to assemble the Link spacecraft. In some cases, Katalyst found its suppliers couldn’t deliver in time, so it decided to build parts itself. Engineers also had to decide what to test on the spacecraft before handing it over to the launch provider, Northrop Grumman.

NASA’s strategy was to tell Katalyst what to do, but not how to do it.

“On paper, it was five requirements,” said Kieran Wilson, principal investigator for the Link mission at Katalyst. “In practice, it boiled down to do no harm and boost Swift. It was not specified how that was going to happen. That allowed us, in turn, to put engineering judgment ahead of exhaustive process, which allowed us to move much more quickly.”

The Link spacecraft attached to Northrop Grumman’s Pegasus XL launch vehicle.

Credit: NASA/Ron Beard

Working inside a factory near Denver, Katalyst completed the Link spacecraft design at the end of last year and had the satellite ready for final prelaunch testing in April.

“We didn’t have a systems engineering organization,” Wilson said. “We didn’t have the traditional structure that you would see for most places that are trying to do something like this—a very complicated mission.”

The successful launch in July was followed by several weeks of in-orbit checkouts. Remarkably, it looked like the Swift rescue mission might actually succeed in reaching its objective, but a series of malfunctions in late July left the spacecraft spinning.

It’s always a valve

“We experienced a failure of a reaction wheel switch on our power system,” Wilson said. “This was something that we developed in-house. The reaction wheels themselves were from Rocket Lab, which is third party. They’re great. We would fly them again, absolutely. But it was the control electronics for those, particularly the regenerative braking circuit. That experienced a fault, and the transistor on that essentially shorted, and that caused it to overheat and render that circuit inoperable.

The spacecraft used cold-gas Reaction Control System (RCS) thrusters in combination with reaction wheels to control the spacecraft’s pointing.

“We implemented a bunch of software changes on the spacecraft and operational changes in order to mitigate the cause of that issue, which was kind of breaking too hard, too fast, too often, and that worked for a few weeks, and then we ended up having a series of events that started with an unresponsive RCS valve, causing the spacecraft to spin up,” Wilson said.

Those events prevented Link from rendezvousing with Swift and completing the reboost, and Katalyst had to forego its final incentive payment from NASA.

This image of NASA’s Swift observatory was captured at a distance of 12 to 15 kilometers by Katalyst’s Link satellite.

Credit: Katalyst Space Technologies

Some of the Link spacecraft’s most complicated elements, like its robotic arms and plasma thrusters, worked as expected. For a time, ground teams at Katalyst thought they might be able to use the plasma thrusters to regain control of Link’s pointing as it zipped around the Earth at nearly 5 miles per second.

“We were able to do some pretty significant orbital maneuvers,” Wilson said. “We did do a whole bunch of interesting things from the GNC (Guidance, Navigation, and Control) side in order to try to keep going even after we had lost two wheels, and those are what enabled us to get within about 10 kilometers [of Swift].”

A little more development time would have given Katalyst a better chance of finding the faulty circuit and valve before launch. Engineers made “gut-wrenching” decisions in the months leading up to launch on which tests to perform and what tests they simply did not have time to do, Domagal-Goldman said.

Wilson acknowledged shortcomings on Katalyst’s side, too, such as “overly light” staffing on the company’s power systems team. The satellite industry supply chain couldn’t deliver critical components for Link’s electrical system, like power conversion and distribution units, on schedule to meet the deadline for rescuing Swift.

“We had to develop a lot of that ourselves, and that meant that the engineers we had were overly stretched in too many different directions,” Wilson said. “There’s a lot of complexity there, and we just did not have enough time or people to test that with the level of care that was necessary. We did do a lot of testing on it, and we didn’t see anything quite like this in testing.”

But more time isn’t always available. Satellite reentries are difficult to forecast, and changes in solar activity can change the density of air molecules in low Earth orbit, resulting in more or less drag on satellites flying there.

“I think next time we’d like to get ahead of it in advance of an unexpected change in the de-orbit date,” Domagal-Goldman said.

What went right

Despite the technical failures after launch, officials pointed to Katalyst’s achievements with the Link spacecraft as showing the way for future satellite servicing and quick-response space missions.

The first step was to get the government to help where it could, and get out of the way where it couldn’t. NASA fell victim to a government shutdown last October due to a funding impasse between the Trump administration and the Republican-controlled Congress. The timing meant the shutdown would create a burden, not relief, for Katalyst. It started five days after NASA awarded Katalyst the contract to rescue Swift, just as the space agency started making payments.

“They needed a huge influx of cash in order to go buy stuff,” said John Van Eepoel, Swift’s mission director at NASA’s Goddard Space Flight Center. “Otherwise, we never would have made it.”

NASA’s procurement team worked in a “fever pitch” to arrange the payments before the government closed its doors, Van Eepoel said.

“That was, I think, heroic and herculean movement with people knowing how to make the procurement system just shake the money free,” he said. “I think somebody was on the phone to Treasury because I was sweating that we weren’t going to make it, and I’d done everything that I could.”

The engineers responsible for operating the Swift spacecraft also came up big in the run-up to the rescue mission. The reboost attempt would be moot if the atmosphere dragged the Swift observatory toward reentry before Katalyst could launch Link in pursuit.

Jamie Kennea, who led Swift operations until earlier this year, said his team at Penn State University devised a way to reorient the observatory to give it a more streamlined aerodynamic profile, minimizing drag and extending its time in orbit. This gave Katalyst a few more months of schedule margin at the expense of suspending Swift’s scientific observations.

“That sounds simple on paper,” Kennea said. “In reality, getting there was extremely hard. We had to use every single tool at our disposal to achieve that goal. That would include rewriting and rewriting software. The use of agentic AI, even vibe coding, helped us out in a lot of cases where we couldn’t figure out how to do things. GPU compute was used extensively to achieve this. What we ended up achieving was essentially extending the lifespan of Swift by several months.”

Wilson said the Link mission came in “within about 5 percent of our predictions” at the time of Katalyst’s bid. “We were within $1.5 of our $30 million mission target, which includes launch, labor, contractors, what have you.”

This shows that a satellite rescue or reboost mission doesn’t have to take hundreds of millions of dollars and years to get to the launch pad, as several government-led demonstrations have done, Wilson said. One example was OSAM-1, a NASA mission that would have attempted to grab onto an aging Landsat Earth-imaging satellite and refuel it.

With the refueling test, NASA’s OSAM-1 demo mission had more scope than Katalyst’s Swift rescue mission, but the agency spent $1.5 billion on the project before canceling it in 2024. At the time of cancellation, NASA officials noted advancements in the commercial industry as one reason for pulling the plug on OSAM-1 and breaking the mold in how they view satellite servicing.

“The industry is at the point where it’s ready to take off,” Van Eepoel said. “And we’re seeing that work out in real time with other companies that are doing rendezvous missions right now: Starfish and Northrop Grumman. They have launched RSGS, and that’s on its way to GEO (Geosynchronous Orbit). So this is just the beginning.”

“Speaking for NASA on its own, this isn’t the last servicing mission that we’re going to try to pursue,” Van Eepoel said. “There are other astrophysics platforms that need help. We’re actively looking at those as we move into some type of a service model for missions.”

Artist’s illustration of Katalyst’s Link spacecraft (left) with NASA’s Swift observatory (right).

Credit: NASA

The results from the Swift rescue mission “represent a floor, not a ceiling,” for what’s possible with a low-cost satellite servicing mission, Wilson said. “We did some really, really cool stuff. We got really, really close. I’m absolutely gutted we weren’t able to get further.”

The Hubble Space Telescope and the Chandra X-ray Observatory are among NASA’s other aging astronomy missions. There have already been commercial proposals to service and reboost Hubble and Chandra, including one from then-commercial astronaut (and now NASA administrator) Jared Isaacman. So far, NASA is not pursuing any of the offers.

Meanwhile, after moving to within 10 kilometers of Swift to test its navigation system, the Link satellite reentered the atmosphere on September 25. Swift has temporarily resumed scientific observations before it falls into the atmosphere and burns up later this year.

The loss of Swift “creates a big hole” in NASA’s science portfolio, Domagal-Goldman said. “We are doing the proverbial running around and seeing what we’ve got on the shelves to see what is available for us to put back up in space to replace the Swift capabilities we’re about to lose.”

Stephen Clark is a space reporter at Ars Technica, covering private space companies and the world’s space agencies. Stephen writes about the nexus of technology, science, policy, and business on and off the planet.