I learned to code in the church of agile.
Big Design Up Front was the sin you confessed to. You didn't plan, you sprinted. You shipped a thin slice, watched it break, learned something, shipped the next one. Anyone who showed up with a 30-page design doc got the look. Boomer energy. Waterfall guy.
And honestly? The church was right. For the time.
Here's the thing nobody said out loud, though. Agile didn't win because planning is bad. Agile won because planning was expensive. Writing the doc cost a week. Drawing the design cost a week. Then a requirement changed on a Tuesday and the whole thing was landfill. So you stopped planning and started discovering the design by walking into it. Small steps, because every step you didn't have to throw away was a step you got to keep.
The plan was precious. That's the whole story. You protected it because it cost so much to make.
That number just went to roughly zero.
I sit in plan mode now and talk through a feature for twenty minutes — goals, non-goals, the failure cases I always forget — and out comes a real spec. Modules, data flow, the ugly edge cases. Not a napkin. An actual document I can read and argue with.
Requirement changes on a Tuesday? I regenerate it. Whole thing, two minutes. Wrong approach? Scrap it, redo it, no grief, because I'm not burning a week of my life, I'm burning a coffee.
When the plan stops being precious, the sin stops being a sin.
So lately — and like everything else this year, I mostly just keep wondering what the best practice even is anymore — I've started thinking we may have misjudged the old man. Waterfall wasn't wrong. It was early. It described a sane way to build software that happened to require a tool nobody had yet: something that makes thinking-it-through cost less than building-it-wrong.
We have that tool now.
There's a version of this where it's just waterfall in a hoodie, and the same trap closes on you — you spec the whole world up front, hand it off, and discover the spec was fiction the moment something real ran. That trap is still there. The fix is that the loop got short. You write the spec, it builds in ten minutes, you see it run, you fix the spec and go again. The doc isn't set in stone. It's a sketch you redraw on every lap.
The funny part is what the planning gives back. Half the time I write the spec and realize I didn't understand the feature. The doc wasn't for the machine. It was the best rubber duck I've ever had.
I'm not fully off the leash here — on a brand-new thing I still let it build something rough first and figure out the shape by poking it. Some problems you can't think your way into. You have to see them move.
But on anything I half understand? I plan first now.
Twenty years of being told that was the boomer move, and the boomers might've just been waiting on better tooling.