AI made me a manager, not a better maker

A few days ago @levelsio posted something that stuck with me. He said he now does "insane amazing stuff never possible before" — and feels less achievement than ever. "It feels like being a manager and not a coder." Then, almost as a footnote: "still a net positive (I think)."

That "(I think)" is the honest part. I've been sitting with the same feeling.

The output is real. I ship more in a week than I used to in a month — more code, more content, more of the small things that used to rot in a backlog. Dominik Sobe wrote the same ledger out loud: codes more than ever, ships faster than ever, does more marketing, even makes a bit more money — and then pivots to a feeling he can't quite name. The productivity went up and the satisfaction didn't follow it up the curve.

I don't read that as a complaint about AI. It's a description of a job change nobody put in writing.

When you run agents all day, you stop being the person who makes the thing. You become the person who decides what's worth making, points the work in a direction, reads what comes back, and sends it around again. That's not coding — it's managing. Reviewing, correcting, deciding. The typing was never the hard part, but it was the part that felt like doing the work. Take it away and the job is all judgment, and judgment doesn't hand you the small hit of "I built that."

Which is why one line from Zara Zhang has been rattling around my head: "Taste isn't valuable because it's impossible to copy. Taste is valuable exactly because it defines what everyone else chooses to copy."

That's the whole thing. If everyone's output goes up at once — and it is — then output stops being the differentiator. What you choose to make, what you keep, what you cut, the direction you point ten agents in — that's the scarce part now. Lenny Rachitsky described the same shift for product managers: the best ones aren't coordinating people anymore, they're prototyping with real code, querying data through MCP, running coding agents directly. The role moved from managing consensus to exercising judgment at speed.

Thariq's writing process is the clearest picture of it I've seen: do the engineering, talk to a bunch of people, brainstorm and research with Claude, write the post, give a talk or two, rewrite it, rewrite the intro, wake up at 6am and rewrite again, then post. The AI is in there — but it's one station on a line where every other station is human judgment. The machine drafts; the person decides, reframes, and decides again.

So here's where I've landed, and I'll hold it loosely. AI made me faster and it made me a manager, and both are true at once. The win is real — I'm not romanticising the days of doing everything by hand. But the thing worth protecting isn't my typing speed, which stopped mattering. It's taste and judgment — the deciding — because that's the part that didn't get automated, and if Zara is right, the part that was quietly the point all along.

The tools got faster. Deciding what's worth making is still the job. Maybe it always was.

Posts that prompted this: @levelsio · Dominik Sobe · Zara Zhang · Lenny Rachitsky · Thariq.

More on how I actually work with AI agents — on LinkedIn.