Comment by Matl
13 hours ago
Problem is in many places the AI push is to use agents for everything and so instead of writing code now you end up reviewing what the LLM generated for most of the day.
13 hours ago
Problem is in many places the AI push is to use agents for everything and so instead of writing code now you end up reviewing what the LLM generated for most of the day.
I dont have bandwidth to check ai made code
that's okay! nobody had the bandwidth to check compiler generated machine code either so we just stopped doing it
Its a false premise so often repeated. I am really tired. Compilers are ruleset: If you pass it one statement it'll produce the same output all the time. Its easy to verify it works. And if enough people have used it before and found no problems, there probably aren't many. Spoiler: People always find niche bugs.
But an LLM will generate you two completely different things today and tomorrow. Heck, even if you stash its changes and redo the same prompt.
They are NOT comparable at all.
2 replies →
Compiler generated machine code is however going from something (fairly) well defined (higher level representation) to something else about equally well defined (lower level representation).
In that sense it's a lot closer to translating Italian to English than what LLMs are doing.
I trust LLMs a lot more if I have a complete Go program I want them to rewrite in Rust.
Yeah, but compilers actually work reliably. LLMs do not.
1 reply →
I find mostly what I do is think about how to improve the product, what the user wants, and what information the LLM needs and how to get it into the context window.
Sure. Ultimately it's still you who is responsible for the code you ship I assume, no matter how it came about.