Saturday morning I set an alarm for 6:45am, which is not a thing I normally do on a weekend, to watch a rocket blow up on purpose-ish. SpaceX's second Starship integrated flight test lifted off from Boca Chica a little after 7am my time, and I sat on the couch with coffee that I made too weak because I was rushing, and watched the whole thing on the livestream like it was a sporting event. It kind of is one, honestly.
Quick recap for anyone who wasn't staring at a stream at dawn: this is the same Starship stack that blew up in April, except this time it actually worked, mostly. All 33 Raptor engines on the Super Heavy booster lit (April had several out before liftoff and lost more on the way up), the vehicle cleared the tower without shredding the concrete pad into a crater this time because they finally built a water-cooled steel plate under it, and then it did something I hadn't seen SpaceX pull off before: hot-staging. That's where the second stage ignites its engines while it's still attached to the booster, using the thrust to shove itself away, instead of coasting apart first. Looks completely unhinged on camera, this bright pulse of fire mid-separation, but it worked on the first try.
Booster didn't make it. It was supposed to do a boostback burn and come down near the Gulf, and instead it went out in a fireball a couple minutes after separation. Nobody on the call seemed shocked or even that bothered, which tells you something about how these test flights are actually scored internally. Meanwhile the upper stage, the actual Starship, kept going. Kept going past where April's flight ended (that one only got about 39 km up before the whole stack tumbled and had to be destroyed). This one got out past 100 km, into space by the usual definition, still burning, cameras cutting between earth curvature shots and a countdown clock that just kept climbing. Then about eight minutes in it went quiet. Lost telemetry, and it sounds like there was a fuel leak that triggered the automated flight termination system somewhere over the Gulf of Mexico. So still not a full success. But last time they lost the whole plot in under four minutes without stage separation ever happening, and this time separation worked, stage two flew on its own for several more minutes, and they got real flight data out past the part of the mission that actually matters for reentry work later. That's a big jump in six months for a vehicle this size.
I know the instinct from outside the space nerd bubble is to see "it exploded" and file this under failure, and I get why, it does look like a failure if you're used to how NASA broadcasts frame every anomaly as a crisis. SpaceX just runs a different playbook — fly hardware, break it, find out exactly where it broke, build the next one. It's expensive in hardware and cheap in schedule, which is a trade I actually think is the right one for a vehicle nobody has built before, even if it means watching fifty million dollars of rocket turn into a fireball on a Saturday morning feels a little insane to type out loud.
Small tangent because I can't help it: this blog turns twelve years old in a couple days, first post was November 2011, and I genuinely cannot get over how different "watching a launch" was back then versus now. In 2011 you got a NASA TV feed with a guy in a headset narrating in a monotone and maybe a grainy Ustream mirror if you were lucky. Saturday I had four camera angles, real-time telemetry overlays, a chat full of actual propulsion engineers arguing about raptor throttle behavior in the replay within twenty minutes of the anomaly. Whatever you think of the company or the guy running it, the media layer around this stuff has gotten unbelievably good.
Anyway. Not going to pretend I know when flight three happens, permits and reviews take as long as they take, but if the pattern holds it won't be another seven months this time. I'll probably set another alarm.