Write for people
10 points by carlana
10 points by carlana
Something I've started doing recently that I feel pretty good about is hiding the LLM-generated PR in a <details><summary> collapsed widget and adding my own short human-written description outside of that. Here's a recent example: https://github.com/datasette/datasette-apps/pull/22
The LLM descriptions have useful information in them but they're also an impersonal wall of text.
What would be really helpful to your reviewers would be if you could read the LLM output and distill it yourself, into your own words, and then delete the LLM's version, and leave only the pertinent part for the reviewers instead of leaving the distillation task as something each of them needs to do individually. This would probably help you better understand the code that you are presenting for review, as well.
Why is this one tagged vibecoding?
I do think it's justified here:
But where it starts to get dicey is when you have both jargon and long text explanations being generated by machines, for people at different levels of understanding.
Or a more cynical take: Lobste.rs is a website about software development, and software development is increasingly inseparable from the use of coding agents - either because people use them, or because people make a point of not using them.
The Lobste.rs decision to tag anything relating to LLMs with "vibecoding" increasingly means that almost every article about writing software will have that tag.
(Imagine a software development forum that insisted every post that involves an IDE was tagged "lol-dumbass-ide-users".)