Artisanal Coding Is Dead, Long Live Artisanal Coding

10x more productive!!!!

It's always the same type. The slop machine told them how brilliant they are, they believed.

I love how all of his PRs are torn apart as AI slop, and he is still posting about it in public as if it was some kind of success story.

I'm not sure a human would this delusional, maybe its a troll of some sort?

This guy is essentially an AI psychosis victim at this point.

They recently opened a pull request on ocaml compiler...

I have no problem with LLM generated PRs in my repo, I merged one the other day and it was very helpful. What I do have a problem with is two things:

0. Make the PR reviewable. That means small, logically distinct PRs, not one huge PR with a bunch of stuff in it.

1. You are 100% completely responsible for the code. I gave the maintainer some feedback on how to add a few lines of comments, they gave that to the LLM, which changed an unrelated path in a pre-commit hook. That's unacceptable, I don't want to babysit your LLM because you can't be bothered to review its output.

2. If I talk to you, I expect you to talk to me. I asked the author a question with a simple answer, and got four pages of LLM ELI5. If I wanted to read four pages of text, I'd open Anna Karenina.

You might notice that the above requirements don't have anything to do with LLMs. I expect them whether a person wrote the PR or an LLM. It's basic etiquette.

I don't think OSS contributors suddenly went crazy and started being rude, but I do think that LLMs allow people who have never contributed to OSS before to start doing it, before they know the rules of OSS etiquette. I'm not sure that's a net positive, but I hope we'll all learn.

Previously flagged https://news.ycombinator.com/item?id=45914635

Likely because of this: https://news.ycombinator.com/item?id=46039274

reads like slop