There’s nothing that wrong with llm code but it’s sad (and probably rather detrimental in the long term) that instead of improving tooling, languages and concepts, the solution somehow is overseeing a probabilistic monkey just hammering it out.
@poni I think the idea is that now you may be able to just make all the perfect tooling for yourself, yourself.

That was my interpretation of https://mariozechner.at/posts/2026-03-25-thoughts-on-slowing-the-fuck-down/ anyway.
Thoughts on slowing the fuck down

Thoughts on slowing the fuck down

@liskin i tried to put together an mcp server for org mode notes i have so i can get like daily summaries of things to do and keep tracking and it was pretty miserable, just another pile of low quality code with insanely low quality random oauth that is open to the internet, instead of one good trusted open source codebase, it feels bad
@poni my LLM autocomplete renamed all usages after i renamed a variable and for some reason this is something trivial that good old IDE tools always struggled really hard with for some reason
@poni we really have grown so stagnant in QoL features huh

but also it doesn't help when you have tons of people who refuse to use anything but vanilla vim or emacs
@katka i actually spent an hour yesterday vibecoding a neovim config and then got some terrible depression
@poni perhaps relatedly, it turns out there's more money in making more of average-quality software than in making existing software better :(