@jenniferplusplus@hachyderm.io
2026-09-14 13:34 UTC
Replies (4)
-
@krig@goto.liten.app 2026-09-14 14:09
@jenniferplusplus@hachyderm.io It's kind of a funny coincidence how the same people who insist that the code is really good are the same people who proudly declaim that they don't even read code anymore
-
@FandaSin@social.linux.pizza 2026-09-14 14:06
@jenniferplusplus@hachyderm.io Right? I'm kinda "forced" to use "AI" at work. I use it only as "autocomplete". It's not that bad at "one-liners" = finishing the line when I've started it. Sometime even if() is made well enough, but when it gets to more lines (for loops, whole functions, ...) it won't do that great job. I find way to optimize the code that "AI" is proposing, or have to completely change it.
-
@chefx@defcon.social 2026-09-14 19:47
@jenniferplusplus@hachyderm.io Well have you seen the code actual humans write? Especially proprietary code? I think it's like that. And AI can write code that works. It's not good code, but it works. And that is what many people care about, that the code works.
-
@zbrando@social.vivaldi.net 2026-09-14 19:55
@jenniferplusplus@hachyderm.io I'm not a programmer but for how I understand how LLMs work I could say: - if LLMs output bad code it could be because they are trained with most of the code available now, which I'm turn could be bad code. - LLMs are an augmentation of the programmer' skills; if the programmer produce bad code because is unskilled the LLM will output bad code faster for him.