# Spec-driven development is not Waterfall's revenge

Published 8 July 2026

[Originally posted on LinkedIn](https://www.linkedin.com/feed/update/urn:li:activity:7480584725795880961/)

![Cartoon titled "Revenge of Waterfall": a waterfall with arms and a knife lunges at a startled developer in an office.](https://ovchyn.me/api/media/file/spec-driven-development-is-not-waterfalls-revenge-1.jpg?prefix=production)

"**Spec-driven development is the revenge of Waterfall**."

I can see the worry behind it: write the full spec at the start, let an agent build everything from it, and you are back to big design up front, the rigid plan-it-all-first sequence Agile spent twenty years trying to escape.

But what was Waterfall's actual problem (according to me)? 🤔

It was a **looooooong feedback loop**. You'd spend 3 months on the whole spec, then 2 months designing the architecture, then review all that and repeat it twice, then 12 months of development that easily turned into 24, and only at the very end discover you had been building shit the entire time.

Now you write that same spec, throw it at the machine, and see the result in two weeks. ⏱️ You still learn your spec was shit, only the learning arrives while you can still do something about it, and for a product team that is where the real money sits: cheaper experiments, faster course-correction, far less spent building the wrong thing.

What is Scrum? A little piece of spec, a little piece of architecture, a little piece of code, and a quick result. **Same Waterfall, just squeezed into two weeks** (well, usually four weeks).

So the actual feature here is that **now we can do in two weeks what used to take a year**. Which seems completely fine to me, and honestly that's how it should be. 🙃