Charlotte 
/Cinny 
charlotte
Show content
instead of using this piece of software made by someone who has used LLMs have you considered using software by a fascist instead who has shipped binary blobs in the source code releases?
also that screen reader of yours. some maintainers use llm. don’t use it. no i will not provide a functional alternative
4
6
24
Charlotte /Cinny
charlotte
Show content
the second one is especially egregious since unlike some other software there is no ideologically pure version and the author sure as hell didn’t do anything about that. literally just screaming down from ivory towers
1
0
5
Show content
1
0
0
Charlotte /Cinny
charlotte
Show content
also berating professionals for evaluating new developments in their field honestly sounds a bit anti-intellectualist which is especially funny given how the same people will scream about cognitive decline
1
0
4
Botch Frivarg
deetwenty@todon.nlShow content
@charlotte it is kinda weird since from my discussions with people one of the few places LLMs have some kind of use (mostly as fancy auto complete, or during a "rubber duck" kinda debugging session) is in programming, does it add as much value as has been pumped into this whole "AI" thing? Probably not. And would still need to be used with care and understanding (I.e. won't replace actual software engineers anytime soon), but it does seem useful in those cases nonetheless.
1
0
0
Show content
@novenary @charlotte I'm kinda glad dtolnay went mask-off, because none of his libraries are actually the right thing. Sure, it takes me a few extra minutes (or hours) to find the right thing, but that significantly increases the quality of my code.
It's also interesting that the people who make those kinds of design flaws in their software tend to be really into fascism? Maybe there's a common way of thinking (arrogance?) responsible for both.
1
0
0
Charlotte /Cinny
charlotte
Show content
@deetwenty honestly i’m fine without using LLMs for raccoding directly. I have used LLM autocompletion for years and tried out “vibecoding” for a bit and while instructing the model to modify files works okayish, “tool usage” seems to just be broadly broken with local models which are the only ones i’m gonna use
I think the unwillingness to try out and use new tools is a recurring theme in swdev at least and any moderately popular alternative/new tool is met by pretty much the same vitriol by usually far more fringe group of software devs. There are reasons why new software development is still being done in C for example, even if there usually is no need to do it in that language
0
0
3
Show content
all those tools and conventions we invented to make better software are for nerds anyway right? may as well quit rust and go back to writing assembly by hand, who needs a compiler to prove correctness (/s)
1
0
0
Show content
@charlotte like I get the motivation behind it all, but when you've set the bar such that you're implying "don't use the linux kernel" while your readme has "why is this important? TBD" then what are we even doing?
0
0
0
Show content
@charlotte very funny how "cancel culture" really only ends up upholding the status quo, because the status quo is immune to it
0
0
0
Show content
@novenary I've been fighting with dyn-clone for the past two days, and I've finally concluded that it's entirely unusable for my use-case (cloning an unsized newtype). You'll never guess who wrote it!
I thought I'd caught them all, but evidently not. I'm going to have to roll my own, here (looks like the two alternatives on crates.io have copied dtolnay's flawed design), but I can't wait for CloneToUninit support in derive_more.
0
0
1