Comment by lionkor
15 hours ago
> LLMs don’t reach for it
Please excuse my language, but that's only the case if you're a complete imbecile and can't write a for loop to save your life.
An LLM doesn't "reach" for anything, because you tell it what to do. You go "use this tool, to do that thing, in this way, here's a bunch of written guidance on how exactly to do that, and if you run into issues ask me". Any other use is glorified copy-paste slop code and will end up with a worse codebase than if you let an egomaniac run amok on a single codebase for 20 years.
You are ignoring that prompting is fractal all the way down to the scale where the human writes everything by hand.
If I say, "start a C++ package skeleton" I'm getting CMake, not waf, make, bjam, or whatever. If I state the build system, then I'm still getting some package layout drawn from the training distribution. If I specify that, I'm still getting snake_case or CamelCase or whatever naming convention, drawn from the training distribution. If I specify that, etc etc.
It is distribution sampling all the way down.
So, yes, literally LLMs ARE reaching for "it".
This brings up the question how much weight these people have, if don't even know what their solution should be based on.