The waterfall method doesn't work for software engineering. It never has. And no matter how much we keep trying to make it work, it won't. We've been trying for 50 years and it still doesn't work.
The devil is in the details and the person who doesn't know the details will never get the high-level design right. They might get it close enough to right that it kinda works. Or maybe the entity actually implementing it can hack around the design enough to make it work. But if you don't know the details, you design it wrong.
And, no, unit tests won't save us. We tried that, too. Does anyone remember the 00s? We wrote lots of unit tests. Waterfall still didn't work.
CEOs and managers hate this. We keep trying waterfall because that's how business people want software engineering to work, not because it actually works. They want to build software like you build a building. Someone makes them pretty pictures to look at and approve, someone else comes up with the overall technical design, and then an army of workers build the building.
(And I'm not slamming on tradespeople here. They're also very skilled, far more than most architects and engineers and certainly more than the property developers know.)
But that doesn't work. It's never worked. And it keeps not working.
This post is about AI coding.
bimmûre
Christian Gmeiner
Alice Averlong🏳️⚧️
Aras Pranckevičius
Collabora