Comment on Storyden: A forum for the modern age.
mnemonicmonkeys@sh.itjust.works 1 day agoIt’s like the difference bewteen publicly-traded and privately-owned companies. Publicly-traded companies are guaranteed to enshittify due to fiduciary duty. Privately-owned companies could enshittify, but it’s not guaranteed
Shimitar@downonthestreet.eu 1 day ago
I like the analogy…
But at the end of the day better to have something that starts shitty but can be improved over time rather than nothing.
If it stays shitty, then it’s just shit and won’t go anywhere.
mnemonicmonkeys@sh.itjust.works 1 day ago
And how is it supposed to get less shit if it’s vibe coded? The programmer didn’t learn anything when making the initial build
Shimitar@downonthestreet.eu 1 day ago
Maybe other devs join and improve the code, or maybe the original del learns and improve, or maybe he was already good bit had zero time and after a while goes back to improve the vibe coded project…
I prefer to judge by the results rather than by the means used.
mnemonicmonkeys@sh.itjust.works 4 hours ago
Why would experienced devs join a vibe-coded project that barely works to overhaul it? Most of the time it’d be quicker and easier to start a new project from scratch with the same goals but coded competently.
You’re bending over backwards just to come up with a plausible-sounding excuse for vibe-coding. But unfortunately, “plausible-sounding” isn’t the same as “plausible”.
Vibe coding might be “good enough” for a tiny private project for someone who doesn’t know how to code and isn’t interested in learning. But LLM’s aren’t good enough to build usable architectures.
And all publicly available human-created code was already trawled years ago, and the creation of more is slow. LLM’s can’t get significantly better at coding by just throwing more data into their training because there is no more data left that isn’t already in the training.
lambalicious@lemmy.sdf.org 1 day ago
The era of not judging the means ended in, like, 1066 or smth. Vibe-coded stuff for example means burning down a couple of forests just so the AI can propose where to place a semicolon, long before a “result” is even visible.