I’ve spent the past year integrating AI into things I build. Not as a curiosity. As a real dependency.
Here’s what surprised me most: the bottleneck moved. It used to be implementation speed. Now it’s clarity of thought. If I can describe the problem precisely, the code often writes itself. If I can’t, no tool in the world helps me.
What changed
The ratio of time I spend thinking versus typing flipped. I now write more in natural language — specs, constraints, edge cases — than I do in code. The code is downstream of that.
I also notice I’m less afraid of starting things. The cost of a bad first attempt dropped. You can throw it away and try again with what you learned. That changes how I explore.
What didn’t change
Good taste. Nobody’s automated that. Knowing when something is almost right but not quite — that’s still entirely human work. The AI will confidently produce something plausible and slightly wrong, and you have to catch it.
Also: the value of understanding the system underneath. I’ve seen engineers treat AI as a black box generator. They move fast until something breaks, and then they’re lost. The fundamentals still matter.
The uncomfortable part
Speed creates pressure to skip thinking. When you can produce something in minutes that used to take days, the temptation is to produce more of it, faster. That’s not always the right answer.
The best thing AI gave me wasn’t velocity. It was space — to think differently, to try more angles, to be less precious about any one approach.
That’s the part I’m still learning to use well.