Space · Thursday, 20 August 2026

01 · Briefing · what happened

The rescue craft could not hold still, so NASA let its telescope go

Space 7 min 25 sources

A refrigerator-sized robot flew up to save a NASA observatory and lost the ability to point itself. Two of its three steering wheels stopped turning, and the mission it was sent to rescue will now burn up.

2 of 3

steering wheels lost

on the rescue craft, three weeks after launch

$30m

paid for the attempt

against a telescope worth about $250m

9 months

from contract to launch

a first-of-its-kind craft usually takes years

190 miles

Swift's altitude

low enough that thin air is dragging it down

At a glance

NASA and Katalyst called off the rescue of the Swift space telescope on Wednesday.
The rescue craft, Link, lost two of its three reaction wheels and began tumbling three weeks after launch.
Reaction wheels are flywheels a spacecraft spins inside itself to turn, using no fuel.
Controllers slowed the tumble from 9 degrees a second to 1.47, but never regained enough control to grab Swift.
Swift is expected to fall low enough to burn up before the end of the year.
Link will still fly close to Swift and practise the approach, which is the mission's remaining value.
Satellite repair is growing regardless: a Northrop Grumman robot is en route to re-engine a 17-year-old satellite.
The Pentagon is separately paying three firms to study removing satellites that cannot cooperate.

Forces in play

Orbit decay High↑

A busy Sun has puffed up the upper atmosphere, dragging Swift down faster than planned. Reentry is expected before the year ends.

Repair market Building↑

Northrop Grumman is flying a robot to bolt a new engine onto a 17-year-old satellite, and the Pentagon is paying three firms to study satellite removal.

Schedule pressure High→

Katalyst built and launched Link in nine months because the Sun set the deadline. NASA says the timeline forced engineers to accept risks they normally would not.

Launch availability Building↑

Quilty Space calls satellite operators an industry in panic over getting rides, even with about 270 orbital launches a year.

Flight rate Easing↓

SpaceX passed 100 missions this year and launched two rockets 38 minutes apart, so getting to orbit itself keeps getting cheaper and faster.

In play Katalyst Space Technologies — built Link in nine months; lost two of three reaction wheels NASA — paid $30m for the attempt and has now let Swift go Swift Observatory — 22-year-old gamma-ray telescope, falling out of orbit Northrop Grumman — flying a separate repair robot to a satellite 22,000 miles up Defense Innovation Unit — paying three firms to study removing uncooperative satellites

How it unfolded

2004 Swift launches to hunt gamma-ray bursts
Sept 2025 NASA awards Katalyst $30m to boost Swift's orbit
3 Jul Link launches on a Pegasus XL
Late Jul Two of three reaction wheels fail; Link tumbles
11 Aug New flight software uploaded; rendezvous still planned
19 Aug NASA and Katalyst call off the boost

Where this points

Link will still fly close to Swift for a practice approach, so watch whether that demonstration succeeds. It is now the only return on the $30m, and it is what the next servicing contract will be judged against.

Full briefing

NASA and Katalyst Space Technologies called off the rescue of the Neil Gehrels Swift Observatory on Wednesday [1][2][3]. The reason was not the grab, the fuel, or the orbit. The rescue spacecraft could not reliably control which way it was facing.

What happened

Katalyst’s Link spacecraft launched on 3 July on an air-launched Pegasus XL rocket [1][3]. It is about the size of a refrigerator, with two solar wings, three robotic arms and three xenon-fed electric thrusters [2]. The plan was to fly up to Swift, catch it, and push it into a higher orbit.

About three weeks after launch, Link began to tumble [1][3]. Two of its three reaction wheels stopped working [1][2]. Reaction wheels are the heavy flywheels a spacecraft spins inside itself to turn without firing an engine. Link’s small cold-gas thrusters, used for fine pointing, were only partly working too [1][2].

That left the three low-thrust plasma engines as the only way to control which way the craft faced, while it circled Earth at close to 5 miles a second [2]. Controllers slowed the tumble from 9 degrees per second to 1.47 degrees per second using those engines [1]. On 11 August they uploaded new flight software and said they expected to reach Swift in late August [1].

It was not enough. Neither NASA nor Katalyst has said exactly what tipped the decision [1]. Link is still alive, and will now fly close to Swift to practise the approach without attempting the capture [1][2][3].

What it costs

Swift launched in 2004 and orbits roughly 190 miles up [4]. It is a one-of-a-kind machine: it spots gamma-ray bursts and swings its instruments onto them within minutes, in X-ray, ultraviolet and visible light [3]. A busy Sun has thickened the upper atmosphere, dragging Swift’s orbit down faster than expected [1]. Without a boost, NASA expects it to fall low enough to burn up before the end of this year [2][3].

The money was small by space standards. NASA paid Katalyst about $30 million for the attempt [1][2][3]; Swift itself cost around $250 million to build and launch [4]. The schedule was the hard part. A first-of-its-kind servicing spacecraft normally takes years; Katalyst had nine months, because the Sun set the deadline [2].

NASA is not framing it as a waste. “This is not the outcome we were working toward, but it does not change why this mission was worth attempting,” administrator Jared Isaacman said [2][3].

The repair business keeps growing anyway

One awkward detail. A Scientific American feature published a day before the cancellation still described the rescue as live [4]. It quoted NASA’s astrophysics director Shawn Domagal-Goldman: “we still do see a pathway to success here” [4]. That is how fast this moved.

The same piece lays out the other half of the picture. Northrop Grumman’s Mission Robotic Vehicle launched last month toward Optus D3, a nearly 17-year-old communications satellite parked about 22,000 miles above the equator [4]. Using two ten-foot arms, it will bolt an electric “jet pack” onto a satellite that was never designed to be touched, buying at least six more years [4]. Optus D3 reportedly cost around $150 million and a replacement would cost about twice that [4]. DARPA’s James Shoemaker says there are roughly 20 to 25 servicing opportunities a year in that high orbit [4].

Governments are buying too. The Defense Innovation Unit and Space Development Agency have picked D-Orbit, Firefly Aerospace and Katalyst for a “deorbit as a service” study [5]. The targets are satellites that cannot cooperate with an approaching vehicle [5]. Starfish Space already holds a $52.5 million contract to remove up to seven spacecraft [5]. And the Australian company HEO signed a deal to borrow Planet’s SkySat cameras, when they are not looking at Earth, to photograph other satellites [6]. HEO’s Will Crowe put the logic plainly: “You just need to have cameras in the right place at the right time” [6].

Overhead, the traffic keeps thickening

SpaceX flew its 100th mission of 2026 on Tuesday, its 97th Falcon 9 of the year, with 73 of those devoted to Starlink [9]. The network now holds close to 11,000 working satellites [9]. On 15 August the company launched two Falcon 9s 38 minutes apart from opposite coasts, cutting 27 minutes off its own record [10]. For scale, it flew 138 times in 2024 and 170 in 2025 [9].

That abundance has not reached everyone. Research director Caleb Henry of Quilty Space describes satellite operators as “an industry in panic” over launch availability [7]. That is despite an average of about 270 orbital rockets a year over the last three years [7]. AST SpaceMobile says it has ten launches booked and 12 satellites up, out of about 45 needed, and is not counting on Blue Origin’s New Glenn returning this year [8]. New Glenn has been grounded since an explosion on 28 May [8].

Elsewhere in the industry: Ship 40, the first Starship upper stage to survive an intact ocean splashdown, finally reached calm water off Christmas Island after 24 days under tow [11]. Astra Space is seeking $250 million at a $1 billion valuation, having been taken private in 2024 for $11.25 million after repeated launch failures [20]. Intuitive Machines booked a $600 million order for three large communications satellites [23]. Virgin Galactic pushed its next commercial flight to February, having not reached space since June 2024 [24]. And an industry report notes the United States launched more than 3,700 objects in 2025, roughly ten times the 2019 figure [25]. The supply chain underneath was built for the older, smaller market [25].

What the telescopes found

New work in Nature Astronomy argues the early galaxies that already puzzled astronomers are heavier still [12]. Masses are inferred from bright stars, then scaled up to account for the faint ones using an assumption borrowed from our own galaxy. Lead author Chloe Cheng compares it to judging a city from its skyscrapers: “a far more numerous population of low-mass stars is concealed by those rare, bright stars” [12]. Change the assumption and the galaxies get more massive, and harder to explain.

Astronomers also reported GJ 523b, a planet 2.5 times Earth’s width but about 23 times its mass [13]. That density is odd. Cores much above 20 Earth masses are usually thought to start hoovering up hydrogen and turning into gas giants; this one apparently did not [13].

And around the young triple star GW Orionis, 1,300 light-years away, ALMA has found a finger of gas pouring into the system [14]. It runs about 0.2 light-years, roughly 1.2 trillion miles [14]. The exchange of spin between that stream and the outer dust ring is tipping the ring over [14]. It may explain why so many planets are found orbiting at odd angles to their stars.

Closer to home

Curiosity has driven into a field of small cracked polygons in Gale Crater, each 4 to 8 centimetres across [15]. It is the first time the rover has seen the pattern from the ground [15]. Similar shapes elsewhere on Mars have been read as old mud cracks, but the team says the cause here is not settled [15]. Perseverance, meanwhile, photographed Phobos crossing the Sun on 13 August, a day after Earth’s own total eclipse [16].

A NASA study in Science Advances found that microbes carried by astronauts could plausibly survive in shaded pockets near the Moon’s south pole [17]. That matters for reading lunar chemistry later: it may get hard to tell what was always there from what we brought.

Two more things are moving. The Nancy Grace Roman Space Telescope was built around spare hardware the National Reconnaissance Office gave NASA for free [18]. It now sits at the Cape awaiting a Falcon Heavy launch, under budget and ahead of schedule [18]. And NASA’s Lunar Reconnaissance Orbiter has photographed a fresh crater on the Moon [21]. A discarded Falcon 9 upper stage hit on 5 August at 5,400 mph, digging a hole about 60 feet wide and under 10 feet deep [21].

What is next

Isaacman says he is “extremely confident” Artemis 3 will fly in 2027, with stacking of the SLS rocket already under way [22]. Space nuclear power is drawing money and attention, though officials warn the fuel supply chain could become a bottleneck [19]. And Link will still fly up to Swift and take its pictures. That leaves one thing the mission can still deliver: data on how to approach a satellite that never expected company.

02 · Lesson · why it matters

Why the only thing a spacecraft can push against is itself

With nothing outside to push against, a spacecraft turns by spinning a wheel inside itself - free, until the wheel runs out of room.

How it works

On Earth, turning means pushing on ground, water or air
In orbit there is nothing outside to push
So the craft spins a heavy wheel inside itself
The wheel turns one way, the craft turns the other
Steady sunlight and thin air keep loading the wheel up
Once the wheel is maxed out, only real fuel can bleed it off

The twist

A spacecraft with nothing to push against turns by pushing on a part of itself, which is free until the part runs out of room.

Where you've seen this

A falling cat

twists its front and back halves in opposite directions to land feet-first, with nothing to push on

A diver mid-air

tucks and untucks to change how fast they spin, without adding any spin

A household budget

you can shift money between pots for a while, but that is not the same as more income

The catch

Moving something internal only buys you room, never new capacity, and the moving part is one more thing that can break.

Full lesson

A machine that could not stop turning

Three weeks after launch, Katalyst’s Link spacecraft began to roll. Not fast. Nine degrees a second, about a full turn every forty seconds. On Earth you would barely call that a spin.

In orbit it was fatal to the mission. Link could not steady itself enough to reach out and grab a telescope. Controllers ground the roll down to under one and a half degrees a second, and it still was not enough. On Wednesday NASA let Swift go.

The thing that broke was not the engine, the arms, or the fuel. It was the part that decides which way the craft is looking.

Everything that turns has to push on something

Stand up and turn around. Your feet twist against the floor, and the floor twists back. A rowing boat turns because the rudder shoves water sideways. A plane turns because a flap shoves air. In every case you take some turning from the world and give the world an equal turn the other way. You never notice, because the Earth is heavy and does not visibly care.

Now take all of that away. In orbit there is no floor, no water, and near enough no air. There is nothing outside the spacecraft to push on at all.

You turn yourself by turning a part of yourself

It spins a heavy wheel inside its own body.

The rule underneath is one of the tidiest in physics: the total spin of a closed system cannot change on its own. Spin a flywheel one way inside a spacecraft and the spacecraft must rotate the other way, so the two cancel. Nothing was added and nothing left. The spin just moved from one part of the machine to another.

A falling cat does the same thing. It has nothing to push on either. So it twists its front half one way and its back half the other, and lands on its feet without gaining any net spin.

This is why every serious observatory uses wheels rather than thrusters. Turning becomes an electrical problem, not a fuel problem. Point at a galaxy, point somewhere else, point back - all night, for years, for nothing.

The wheel fills up

Here is the part that gets missed.

Sunlight has a faint pressure. So does the last thin haze of atmosphere down where Swift orbits, 190 miles up. Neither is strong. Both push the same way, all day, every day.

The wheel has to absorb every bit of that, and the only way it can is by spinning faster. Faster and faster, until it hits the speed it was built for. Engineers call it saturation. The plain version: the wheel is full.

At that point the free trick stops. To empty the wheel you have to hand the spin to something outside the craft, and the only thing outside is exhaust. You fire a real thruster and burn real propellant, purely to get your steering room back. That is why a perfectly healthy telescope can still have a fixed number of years in it. The clock is the fuel it spends on standing still.

Fewer wheels, less room

A spacecraft normally carries at least three wheels, because there are three directions to turn in. Link lost two of its three. Its small gas thrusters were only half working. That left slow plasma engines doing a job they were never shaped for.

And a wheel is a moving part. Almost nothing else on a spacecraft touches anything else. The wheel spins, on bearings, for decades, in a place nobody can reach. It is one of the few components that can simply wear out.

The difference between borrowing and rearranging

Last week this beat looked at how a probe steals speed from a passing planet. Same law, opposite move. In a slingshot the momentum is real and it arrives from outside: the planet gives some up, the probe takes it, and the books balance across two bodies.

A reaction wheel takes nothing from anywhere. It only shuffles what is already inside the box. That is why it is free, and it is exactly why it runs out.

Around a young triple star in Orion, a stream of gas a trillion miles long is tipping a whole disc of dust over. Spin arriving from outside changes a system permanently. Spin moved around inside only changes where it sits.

Where the line reaches

This is not a spacecraft rule. It is the shape of any system with no outside to lean on.

A household that cannot raise its income can still move money between pots, and that genuinely helps, right up to the day the pots are empty. An organisation under a hiring freeze can move people between teams and look responsive for a while. The rearranging is real work and it buys real time. It is also not the same thing as more, and mistaking one for the other is how you get caught out.

The satellites overhead that carry your weather forecast, your maps and your bank’s timestamps are all doing this quietly, every minute, spinning small wheels to hold themselves still. Most will end not because they broke, but because they ran out of room to turn.

Link is still up there. It will fly close to Swift and take its pictures. It will come home with a lesson about how hard it is to hold steady when there is nothing to hold on to. Nobody planning that mission listed the wheels as the thing that would decide it.

03 · Lab · your turn

Hold the Telescope Steady

Rehearse the trade a spacecraft makes when the only thing it can push against is itself, and the wheel it pushes on keeps filling up.

04 · Hope · carry this

Catching a moving telescope with a robot is something we have barely learned to try. The first attempts fail, then they stop failing. That is how every skill arrives.