According to the release:
Adds experimental PostgreSQL support
The code was written by Cursor and Claude
14,997 added lines of code, and 10,202 lines removed
reviewed and heavily tested over 2-3 weeks
This makes me a bit uneasy, especially as ntfy is an internet facing service.
Am I overreacting or do you all share the same concern?
Oh goddamn it, I’m using this and don’t have an alternative lined up
If using ntfy for UnifiedPush: https://unifiedpush.org/users/distributors/
What is your concern? If it’s a generic “AI”, then I can assure you tha pretty much every software has AI code in it already. Heck, Linus is accepting PRs where AI has been used.
AI is useful. It produces useful code.
Like creative writing, it won’t produce something novel. But man, 75% of code is just boiler plate. AI can do a lot.
That does not absolve anyone of committing crap code. Put your name to it. Own it. Take the consequence of delivering shit code or great code, no matter how it was written. Don’t let AI be a crutch. But you’d be god damn fooling not to use it, where it’s right.
Massive changes made by robit in what has been a pretty stable utility for years is (obviously?) my main concern. It’s absolutely a crutch, and seeing a dev lean on it like this gives me the same feeling Coach must’ve got seeing his star player limping into the big game on a real one. If dude wants to check out and let the machine run his project fine, but I’ll be looking for something someone still cares about and works on.
I think you’d be a fool to use it. At this point it’s subsidized by their need for training data/desire to manufacture dependency, but that won’t be the case for long. It’s expensive, detrimental to your skills, and damaging to both our planet and society. It centralizes and gatekeeps access to information, the most powerful resource of all. “Treat it like an inexperienced dev” managers say, while it replaces their opportunities to gain experience. How are they supposed to even tell great code from shit when everything they’re exposed to has been run through the averaging machine?
I saved your comment for the added arguments against AI.
There’s a big difference between “AI was used in some capacity” and “Entirely vibe coded”
Of course. And when I hear “vibe coded”, I hear someone starting with “make me a cool app” and going from there, with zero understanding of the technical architecture.
If you have a thorough, deeply thought through technical spec, then AI can write a great amount of tests up against that spec, say, and you’ve got a fantastic base for TDD.
I honestly feel like a lot of the downvotes are people thinking AI means “clueless programmer having an AI do its work for you”. Many highly productive, deeply technical developers use it every day.
Idk man by the sounds of it, the AI implemented the entire back end change, adding 14k lines of generated code. The dev doesn’t even seem confident with his own testing. Sounds like it’s closer to the vibe-coded end of the scale to me.