I caught myself doing something embarrassing the other day. I hit a bug, sat there for all of about ten seconds, then alt-tabbed straight to Claude. Not because I was stuck. Because asking was easier than thinking. And the second I noticed it, I felt a bit ill.
ChatGPT and Claude are genuinely brilliant. They debug, they rubber-duck, they talk me out of bad ideas at 11pm. I’m not here to tell you to throw them in the bin. I use them every day and I’m not stopping.
But there’s a difference between using a tool and leaning on it so hard you’d fall over if it vanished. And I think a lot of us, me included, have crossed that line without noticing.
You know the move. You hit a wall, and instead of reading the error, instead of opening the docs, instead of sitting with the problem for a minute, you paste it into a chat box. Fixed. Magic. Onto the next thing. It feels like productivity. It looks like productivity. But something’s getting skipped, and that something is the part where your brain actually does the work and gets better at it.
I’ve seen the threads on Reddit. Developers worried they’ve gone soft. People who used to reach for a regex or a tricky bit of state logic without thinking, now blanking on it and feeling like a fraud. I used to read those posts and think, not me. Then I watched myself ask an AI to write a for loop because I couldn’t be bothered, and the smugness evaporated.
Here’s the thing that bothers me. We already ran this experiment with phone numbers. I knew thirty of them by heart growing up. Today I’d struggle to recall my own partner’s without checking. Not because I got dumber, but because I stopped needing to, so the wiring just went away. Skills you don’t use don’t sit politely in a drawer waiting for you. They rust.
The difference is a phone number is harmless to forget. The ability to reason through a problem you’ve never seen before is the entire job.
Picture the moment it actually matters. You’re in a technical interview, or a live incident, or a code review where someone asks “why did you do it this way” and there’s no chat box to paste into. That’s when you find out whether the skill is yours or whether you’ve been renting it. I don’t want to find out the hard way, and I doubt you do either.
So I’ve made a few rules for myself. Nothing dramatic.
I try to sit with a problem before I ask. Even five minutes. If I crack it, great, it stays mine. If I don’t, I ask, but now I actually understand the answer instead of just pasting it.
When I do use AI, I make it explain, not just solve. If I can’t describe why the fix works after reading it, I haven’t learned anything, I’ve just moved a black box from their server to my codebase.
And once in a while I’ll do a chunk of work with the AI tools closed. It’s slower and a bit humbling, and that’s exactly why it’s worth doing. It tells me which muscles have gone weak before an interview or an outage does.
None of this is anti-AI. The tools won. They’re staying. But there’s a version of this where we end up as really fast copy-paste machines who can’t function the moment the API goes down, and there’s a version where we use the tools to get sharper. The gap between those two is just whether you’re still doing any thinking on your own.
I’m aiming for the second one. Some days I manage it. Most days I catch myself alt-tabbing on autopilot and have to drag myself back. That’s the honest state of it.