I really like these 2 posts from @conradirwin.bsky.social and @jimniels. They manage to put in words one of my struggles with AI coding tools.
The distinguishing factor of effective engineers is their ability to build and maintain clear mental models.
Why LLMs Can’t Really Build Software
- With LLMs, you stuff more and more information into context until it (hopefully) has enough to generate a solution.
- With your brain, you tweak, revise, or simplify your mental model more and more until the solution presents itself.
One adds information — complexity you might even say — to solve a problem. The other eliminates it.
Just a Little More Context Bro, I Promise, and It’ll Fix Everything
I find this is especially true when maintaining big and complex codebases that I know well. When tacking a new task, I don’t start from scratch. I’ve built mental models, my view was challenged, and I’ve updated my model at a result. And again. And again. I went through hundreds of engineering loops with that codebase. I can put some of that baggage into words and feed it to the LLM, but the loop stops there. LLMs cannot iterate through a loop on their own. They cannot find the most optimal implementation on their own.
At least for now.
#AI #development