The first full-colour images from the James Webb Space Telescope came out this week, and for once the thing dominating tech is genuinely worth the noise. The deep field is extraordinary, thousands of galaxies in a patch of sky you could blot out with a grain of sand at arm's length. But the picture I keep coming back to isn't the image. It's the project behind it.
Webb is the machine that was never going to ship. Years late, billions over budget, an unfolding mirror with hundreds of single points of failure that had to deploy perfectly, a million miles from anyone who could fix it. By every metric a software team would recognise, it was the project you'd have killed at the last review. And it worked, on the first try, exactly as designed.
I work in a world that has more or less decided the opposite is virtue. Ship small, ship often, fail fast, fix forward. And for what I do, that's right. If I deploy a bad config I roll it back in ninety seconds and almost nobody notices. The whole discipline is built around the cheapness of being wrong. Webb is the inverse: a system where being wrong once, at the wrong moment, ends everything, and there is no rollback because there's nobody out there with a spanner.
You can't iterate your way to that. There's no canary deploy for a sunshield the size of a tennis court unfolding at L2. The only tool you have is to be right the first time, which means testing on the ground until you've exhausted your own paranoia, and then testing some more. The famous "344 single-point failures" line gets quoted as a horror story. Read it the other way: somebody enumerated all 344, and engineered each one until they slept at night. That's not recklessness, it's the most disciplined thing I've ever heard of.
What it reminds me, on a week when I've been moaning about a flaky test suite, is that "move fast and break things" is a strategy, not a law of nature. It's the correct strategy when breaking things is cheap and reversible. It's negligent when it isn't. The skill isn't picking one camp and sneering at the other, it's knowing which world you're standing in. A web service and a space telescope are not the same problem, and pretending the same dogma fits both is how you get outages on one end and, on the other, a very expensive paperweight a million miles away.
Most of us will never build the irreversible thing. But it's good, occasionally, to be reminded that it's possible, and that the people who do it aren't slow or timid. They're operating in a world where wrong is fatal, and they got it right anyway. The pictures are the reward. The engineering is the story.