sometimes we use llms at work, when there is a deadline.

its just garbage code.

none of the ai generated code stays in the codebase.

its just waste of time.

FOMO garbage.

more ai generated garbage you have, ai becomes even more worse, even with the most expensive models...

then you have a garbage codebase you cant even work on manually. and have to rewrite everything...

1000 line files, so much shit code even the tooling fails at some point. like you hit OS watch file limits in dev mode.

then you have to get in there and remove everything one by one. and rewrite them...

"last time you used llm, you made it fast":

yeah last time codebase was understandable, so ai was able to duck tape in some stuff. or last time we just made a small interface, we weren't working on this big project...

"just create bad code we have to publish on this date. just make sure there are no bugs"

in what world "bad code" means "faster"? there is no such correlation...

yes llms are useful like if you change something and wanna refactor the codebase to use the new method. and some other stuff... doesnt mean we can use it everywhere...

its just garbage... i hate it...

it just creates the illusion of "fast", but then later down the line you spend even more time fixing and debugging things.

im so tired...