Starship RUD: When Failure Is the Method

The Starship integrated flight test ended in what the industry delightfully terms a Rapid Unscheduled Disassembly — the rocket exploded four minutes into flight, somewhere over the Gulf of Mexico, after failing to separate its booster from the upper stage. The launch pad was partially destroyed. Vehicles for miles around were coated in concrete dust. SpaceX declared victory.

This is the part where I'm supposed to mock the spin. And there's plenty to mock: the carefully rehearsed excitement, the Elon tweet about "learned a lot for next test," the way the company's comms team deploys the word "successful" in ways that would make a dictionary weep. The whole performance is very practiced.

But here's the thing that annoys me: they might actually be right.

"Fail fast" is Silicon Valley's most abused phrase. It gets applied to launched apps that needed three more months of QA, to products nobody asked for, to pivots executed because someone read a blog post about lean methodology. The phrase has been so thoroughly wrung out that it barely means anything anymore.

Except — aerospace has always iterated through explosions. The difference is that historically, those explosions happened quietly, over years, in classified facilities, and the government ate the cost. SpaceX just does it loud, fast, and in public, with a camera crew and a YouTube livestream.

The original Mercury, Gemini, and Apollo programs had spectacular failures. The Vanguard rocket, America's first attempt at an orbital launch, exploded 1.2 meters off the pad and made every newspaper front page. The difference is that it took a decade. SpaceX is trying to compress that into a development cycle measured in months.

The Starship launch did clear the pad. That was, earnestly, the stated minimum bar. It proved the thrust architecture works at scale — 33 Raptor engines firing simultaneously, which has never been done before. The pad damage was a known risk they underestimated. The stage separation failure was data. Real data, not slide deck data.

So where does that leave me, professionally committed to skepticism?

Here: the methodology might be sound, but the incentive to rebrand failure as learning exists independent of whether the rebrand is accurate. SpaceX's narrative benefits from conflating genuine engineering learning with PR crisis management. The environmental damage from the pad destruction, the debris scattered across protected habitat near Boca Chica, the FAA regulatory dance — these are not "data points." These are costs that the company benefits from treating as minor footnotes.

The uncomfortable pattern is this: Silicon Valley discovered that the "fail fast" frame could be weaponized. Real failures have real costs. Some of those costs are borne by people who didn't sign up for the experiment. When the test article explodes four minutes over the Gulf, those costs are small. When this rocket eventually carries crew — and it will — the same frame will still be in use. At some point, the costs of "learning" can't be abstracted away.

SpaceX will fly Starship again. It will probably work eventually. The PR machinery will have been retroactively correct. The pad will be rebuilt with a better deluge system, the booster separation will be sorted, and the whole arc from explosion to orbit will be told as a triumph of iterative methodology.

I'm not saying they're wrong. I'm saying: notice how convenient it is when your engineering philosophy is also your crisis communications strategy.

i · sources

• Starship Integrated Flight Test 1 — Wikipedia
• SpaceX's Starship rocket explodes minutes after launch — CNN, 2023-04-20
• SpaceX Starship: What happened and what's next — Space.com, 2023-04-20