applicable to llms but not restricted to them, capitalism
@lumi
Another thing I don't love is appeals to futility when arguing against adopting a policy against it. It may not always be possible to distinguish slop code from other code, but that's on the person presenting it as their own. All technology is inherently built on trust and cooperation, it'd be silly insist that we shouldn't bother because someone may lie. At the very least, signaling that you yourself commit to not using it is worth something.
Also, a lot of slopcoders are really proud of the fact they don't review their code, and easy to spot.
food
@mayu @lunareclipse
This is good, I also recommend adding a little bit soy sauce or straight MSG and table salt. More soy really helps flavor the other soy
pol, rust, llm
@natty @lumi @marcusxms (dropping cwebber to not spam them)
The term LLM should hopefully better outline what we're actually against, since no intelligence of any sort is mandatory for contributing to Alacritty.
heh
pol, rust, llm
@natty @lumi @marcusxms @cwebber
At least in ghosttty's case, that doesn't really turn me off zig much. It's far from a core component or important library to the ecosystem
re: genai, request
@snugglybun @lumi
why even bother if i'm gonna be replaced by a soulless machine, and then most people won't notice anyway
I have half baked thoughts of encouragement if you'd like to hear them :3
@lumi
"hey, couldn't we lose billions of dollars on that while securing ourselves a nice parachute?"
@natty
If you do find a solution to sandboxing/image library you like, do let me know please!
genai, "ethics aside"
@lumi
ugh, lost my first draft because my computer froze. Fortunately, I was able to take a photo of most of it.
For me the entire technology just misses the point. The main reason to use it for writing, digital art, music, programming, etc. is because you don't care about the process, only the output. But we already have enough to read, look at, listen to, and I'd argue, code to last several lifetimes. The only people who care about churning out "content" are capitalists who don't care about the art or what form it is. And if all you care about is the volume of content, we already have enough.
There are many reasons to write something of your own besides just the output. You may enjoy writing for its own sake, or it helps organize your thoughts, you want to make something new*, or have your own voice heard. Using an LLM to churn out slop doesn't help any of those. There's also an element of "if a human didn't care enough to write this, why would I care enough to read it."
Some with any other art, like playing music. Someone can already listen to any recording of Bach by the best players in the world, or their favorite band, but they still learn to play it themselves, or attend live shows, or write their own music. You can see that tech bros just don't understand the point by insisting that nobody actually enjoys learning an instrument.
Heck, this is part of why people care about sports too I guess. Nobody (very few) watches sports only for the world records; they watch what's live now even if Michael Jordan is retired and his entire career recorded. They haven't yet figured out how to try to replace football with a bot, but I'm sure they would if they could.
LLMs also are not able to average out something into the best. And if you want to make the best (which is not the only reason to make something), you cannot learn without doing, so using an LLM to do it for you will hurt your efforts in learning.
This last part is my worry with programming especially. It's not controversial to say the state of software was pretty bad before LLMs, but adding more code of mediocre quality that people definitely always review the output of (they don't. even if they may intend to) is not going to help things. I'd much prefer to stop the treadmill and actually try improving the quality of software, tooling, documentation, accessibility, security, etc. than using a tool that drowns us in a volume of slop that's impossible to review, made worse by nobody learning any expertise.
*For non fiction this can include reporting on new events, novel research, or a new analysis. LLMs aren't any good for that. For fiction new doesn't mean "100% original", but that doesn't mean the average of all previous works is new or interesting either.
@natty
preferably sandboxed
sandboxed how? I think it'd be neat to have finer grained control within the program of what gets what, but it seems like wasm is the best way to do that without a language that supports ocaps without an escape hatch 
image-rs apparently supports that well though
@ireneista
Neither are probably useful to you, I just thought it might be interesting to know that there's at least someone using SASL PLAIN to 'mitm' the credentials and pass them along to the OAuth service
@ireneista
(from the other thread)
it can in principle be used by something that sits in front of a SASL service and handles the user interaction and takes temporary possession of the resulting token, but we're not aware of anything that actually does that
I don't think I fully understand, but this might be one of these
An OAuth2/OpenID Connect (OIDC) Authorization Server on top of Prosody’s usual internal authentication backend.
modules.prosody.im/mod_http_oauth2
Uses SASL PLAIN for Prosody to get the username and password, and then pass it along to the OAuth2 service
modules.prosody.im/mod_auth_oauth_external
@2something @violet
Shoot (to both statements).
The main reason I like the browser extension is domain matching, so I don't have to worry quite as much if it's a phishing site.
@2something @violet
Gnome secrets looks nice. I like the browser extension keepassxc and others have, does it have something like that?
As for syncing I haven't checked for certain that it does not use AI, but syncthing and syncthing-fork on Android might be worth investigating.
not very ace
@lumi
lock gock 
