Paul le fou blog

I have multiple excuses on why I am using AI in my programming. I'd say I used copilot because I hate writing tests and fixtures and it's so good at stubbing dummy data that almost looks right, right? Then there are things that I did countless of times. Things like setting up the application, wiring the environment variables, writing abstraction to make API calls, writing ETLs that just receive a model and modify it slightly to send similar model somewhere else.

We've all been there. I get mad at copilot every time it almost does a right job. This time I finally got something interesting that I really wanted to do. It was an optimization task, one of the jobs were getting OOM error and I needed to think of a solution. When I looked at the code I had a feeling that I didn't have in a while. I had a general understanding of where it should go, but I didn't see the exact solution as of yet. I knew I shouldn't do it, but I was just curious I wanted to prove dumb AI can't see what is the issue here. I just copied one of the function and pasted it without any context just saying what I want it to do.

And you guessed it right. It solved the task. It wasn't exactly right, but it stole the joy of figuring out how exactly I would approach it. After a 40 minutes of manual boring job that I'd actually enjoy AI to do and it was all done. I got zero enjoyment of fixing the issue and I let AI do the most interesting part of the task while I was left to clean up the mess and adjust everything else to work properly considering new solution