AI doesn't replace the engineer

"AI will replace developers" — I didn't believe it. Now I do. But not the way it's being sold.

Let me be precise, because it's easy to slide into hype here.

To carry a product you used to need a team: at least two backend devs, two frontend, a devops. Hands. A lot of hands, because there was physically a lot of work.

Today the same volume is handled by two or three. Sometimes one — with a team of agents. (Yes, "one" already comes with burnout. But that's a separate topic.)

Here's what actually happened: AI didn't eat "developers." It ate the execution layer — the part that needed many identical hands. Knock out CRUD, rewrite a component, bring up infra, grind through the routine — that's delegated to agents now.

And what did NOT get replaced — and isn't being replaced yet:
— thinking
— making decisions
— seeing where the system will break
— designing and holding the architecture in your head

The engineer didn't go anywhere. They moved up a level: from the one who writes code by hand to the one who runs a team of agents that writes the code.

And this is where the new skill lives. Not "being good at prompting." Planning, decomposing, scaling, verifying. Not being the hands — conducting.

So the thesis is simple: AI doesn't replace the engineer. AI replaces the need for five of them, so the work can be led by one — the one who can think.

This is my experience, not dogma. Tools change weekly — verify what's current for your own task.