Are you learning anything yourself when using coding agents?
I just started playing Half Sword, at first I just brute-force smash-clicked over and over and failed and failed until it somehow magically clicked in my head and I intuitively understood what to do. In that clear-headed moment, I kind of reviewed my last coding sessions and thought about them: did I actually remember them, or did I even learn anything? To my surprise I couldn’t remember much. As it seems the models get smarter and smarter, I kind of do the opposite — I would say I regress. Sure, I get better at writing specs and giving needed context and whatever, choosing which model for which task. But even this task is getting easier and easier, as there is no deep cognitive challenge here, so it can’t really get written into my brain. Programming per se is kind of cold, isolated work, but at least you had some cognitive stimulus going on, which this agentic coding — or however you call it — does not get to my brain at all.
Do you feel the same, or do you have a different approach to it?
No comments yet.