A file-storage feature worked in development, then gave me problems in production. I spent days trying to work out why, and I leaned heavily on AI to help.
That didn't get me to the answer. I kept following suggested directions, changing things and finding that the next explanation contradicted the previous one. I eventually worked out the problem myself.
The frustrating part was how readily I'd handed over the thinking. A confident explanation can make you feel like you're moving forward, even when you're just going further down the wrong path.
I tried several models during that work. I wanted one of them to find the thing I was missing. In that particular case, I think I would have got there faster if I'd spent more time working through the problem myself first.
I'm talking about my experience with one difficult piece of work. There are places where I find AI useful, and my way of using it has changed. But the debugging session made me pay attention to how I was deciding which explanation to trust.
The code still had to work in our production environment. I still had to understand what was failing. Repeatedly asking for another answer wasn't a substitute for that understanding.
When I get stuck now, I want to keep that responsibility with me. Use the assistance, question the explanation and keep checking it against the problem you're actually seeing.
You can lose a lot of time if you let the tool choose the direction while you just keep making its changes. I certainly did.
