This quote from the article really sums it up:
And to be clear, I don’t care whether you typed the code yourself. I care whether you understood it before you shipped it. I care whether you can explain why the bug happened, why this fix is the right fix, what the model might have missed, and what would make you roll it back.
I have this feeling that a lot of people are against AI-generated code, even if you use it in this “responsible” way - but maybe I’m seeing an exaggerated view? I’m sort of wondering if I should start using AI models in my own work, including open source stuff, but I don’t want it to be like I’m just having the AI do everything for me and I’m worried that others will disregard the code as slop. I’ve seen a lot of frustration with new open source projects that are largely AI-generated, but then again, we’re seeing a lot of those because it’s a useful tool that lowers the barrier for a lot of work. I’m conflicted, anyone have thoughts?
even if you write code using an LLM and understand every line there is no way for anyone to detect whether you used AI or not
I wouldn’t say that. I think you can definitely tell if something is AI generated. If nothing else, you can see it via the git history, with the amount of code written in a short amount of time.
Don’t underestimate the speed afforded me by my vim skills ;-)
that’s not proof that someone has used AI but I haven’t seen a real AI slop repo so dont know
Oh you can. But my autistic brain is really good at pattern detection. And LLM generated code just doesn’t feel human. I can’t explain it other than it has the same vibe and formatting as examples in a textbook.
Also, lazy vibe coders. You can literally see the boundaries of chunks of code where they move from prompt to prompt. It’s like gluing newspaper articles together.
What’s with this trend of everything in monospace?
This has to be one of 3 things:
- Their bug/code base is really trivial
- They are not good at their job and missed important things
- The Claude code they are using is really different than the one I am using
Because I was experimenting with LLMs since gpt 3.5, still use them on and off because FOMO and they are still shit for anything that’s not trivial
