Starship SN8 flew 12.5 km and then exploded. It

I just watched a rocket belly flop from 12.5 kilometers.

SpaceX flew Starship prototype SN8 this afternoon. It launched from Boca Chica, Texas, climbed on three Raptor engines, shut them down in sequence, tipped on its side, and fell. Belly first. Controlled. Like a skydiver spreading their arms.

The belly flop is the key maneuver. It’s how Starship is designed to slow down when entering an atmosphere, whether Earth’s or Mars’s. Instead of pointing engines-down and burning fuel to brake, the whole ship acts as an air brake, falling sideways, trading speed for drag. Then, at the last moment, it flips upright and fires the engines to land.

SN8 did the belly flop perfectly. The fall was controlled. The aerodynamic surfaces steered the ship through the descent. The flip was violent and beautiful, the ship rotating from horizontal to vertical in seconds.

Then it came in too fast. The landing burn started but didn’t produce enough thrust. Something with the fuel header tank. The ship hit the pad at speed and exploded in a fireball that turned the Boca Chica sky orange.

I should have been disappointed.

I was ecstatic.

Why an explosion can be a success

The primary goal of SN8 wasn’t to land. It was to test the belly flop. The subsonic aerodynamic control. The engine relight. The flip maneuver. All of those worked. The data SpaceX got from this flight would have taken years to collect in wind tunnels and simulations.

Elon tweeted “Mars, here we come!” from the test site. The ship was still on fire.

NASASpaceflight had live coverage and the comment section was celebration. Not in spite of the explosion. Because of what the explosion represented. SpaceX builds fast, tests fast, fails fast, and learns fast. SN9 is already in the assembly bay. It’ll fly in weeks. Not months. Weeks.

Compare this to traditional aerospace. A failure like this would trigger an 18-month investigation, a congressional hearing, a redesign cycle, and another 18 months of testing before anyone tried again. SpaceX’s timeline is build, fly, explode, learn, repeat. In weeks.

What I saw

The footage from the onboard camera during the belly flop is the most extraordinary rocket video I’ve ever seen. The ship is falling sideways, and through the camera you can see the ground slowly getting closer, the sky tilting, the engine bells visible at the bottom of the frame. It looks exactly like a person falling through the air and looking down.

And then the flip. The engines ignite and the view swings violently as the ship rotates. Sky, ground, sky, ground, and then engines-down, pointing at the pad, falling fast, too fast.

Orange flash. Static. End of feed.

I watched it four times. Each time, I noticed something new. The way the aerodynamic flaps moved independently during the descent. The green flash of the engine relight (copper combustion, I think). The shadow of the ship on the ground in the last frames before impact.

The speed

Here’s what I keep coming back to.

SN8 was built in a tent. Not a clean room. Not a proper factory. A collection of structures in a Texas coastal town that were barely there two years ago. The ship is stainless steel, welded by hand in many places, looking more like a water tower than a spacecraft.

And it flew. To 12.5 kilometers. Did a belly flop. Did a flip. Almost landed.

SN9 is already assembled. SN10 is being stacked. SN11 is in fabrication. SpaceX is iterating on rockets the way software companies iterate on apps. Ship it, see what breaks, fix it, ship again.

This is what I meant two years ago when I wrote about Starhopper looking ridiculous. The early versions look silly. They look unfinished. They explode. And then one day, one of them doesn’t explode, and suddenly you have a ship that can go to Mars.

We’re watching that process happen in real time, in public, streamed live to anyone with a browser.

The future keeps exploding on the launch pad. And they keep building the next one.

Related thinking: