AI Doom Scrolling

I decided to pick up Rails again. It's been years, and I wanted to build something from scratch without reaching for the same Next.js setup I've configured a thousand times. The whole point was to learn. Have fun. Remember what it felt like to not know how something works and figure it out anyway.

So I told myself: no Claude. Just me and the docs.

That lasted about ten minutes.

It started small. "Let me just ask Claude to define the data models. I don't want to think about that part." Fine. Reasonable. Then it was the migrations. Then the scaffolding. Then the seed data. I looked up and there was a sort-of-functional GTD app sitting in front of me. I couldn't tell you how the routing worked. I didn't learn a single thing, which was the entire point of sitting down.

Here's what bugs me: it felt exactly like doom scrolling. I opened my terminal with an intention, and then I woke up from a daze twenty minutes later with nothing to show for it except output. Each prompt was a tiny hit of progress. "Oh cool, that worked. Let me just do one more thing." No friction. No natural stopping point. No moment where I had to pause and think. The tool is so good at removing friction that it removed the learning too.

And I get the appeal. I'm a manager. Describing what I want in plain English and getting a working thing back in minutes? That's the execution pipeline I fantasize about with actual teams. But when I delegate to people, I still understand the work. I review PRs. I ask questions. I stay close enough to make real decisions. With Claude, the code showed up and I barely glanced at it. I wasn't delegating, I was just... not in the loop. On my own side project.

That raises a question I keep circling back to. If I'm not writing code, and I'm not learning the underlying tech, what exactly am I doing? Am I still technical? Honestly, maybe not in the way I used to be. In my day job I delegate code already. That's been true for a while. But if that's the deal, then the things I actually own are execution strategy, architecture, unblocking my team. The thinking work. And prompting an agent to build a todo app is not that.

The thing is, I love solving problems. Watching an agent build something doesn't scratch that itch. It scratches the other itch. The restless, compulsive one. Same dopamine loop as Reddit, just dressed up as productivity.

I don't know what to do with this yet. Maybe it's as simple as keeping "AI projects" and "learning projects" separate. Maybe it's about closing the tool the second I notice I've drifted, the same way I try to close Reddit. I suspect the answer is boring and disciplinary and I'll have to keep re-learning it.