Post B55ArS73BVHvH6f4LY by phnt@fluffytail.org
(DIR) More posts by phnt@fluffytail.org
(DIR) Post #B550QGubimlJAzF3ho by SuperDicq@minidisc.tokyo
1 likes, 0 repeats
@hj@shigusegubu.club I would get on this if I were you and get Pleroma added as the alternative instead of GoToSocial. https://codeberg.org/small-hack/open-slopware#social-media
(DIR) Post #B552iR5gockdrkDMfY by phnt@fluffytail.org
0 likes, 0 repeats
@SuperDicq @hj Pleroma backend is slopware according to their definition. Don't tell anybody. *shush*
(DIR) Post #B552iRJrxu5sZiqgls by lain@lain.com
0 likes, 0 repeats
@phnt @SuperDicq @hj frontend too
(DIR) Post #B5579uRsAhXSYKwjuC by SuperDicq@minidisc.tokyo
0 likes, 0 repeats
@phnt@fluffytail.org @hj@shigusegubu.club How so?
(DIR) Post #B55AlesmnFZB8uE6ls by icedquinn@blob.cat
2 likes, 0 repeats
@SuperDicq @hj > instead of firefox, use [project 100% dependent on firefox upstream code]these developers are joke people, they are very funny
(DIR) Post #B55ArS73BVHvH6f4LY by phnt@fluffytail.org
1 likes, 0 repeats
@SuperDicq @hj Pleroma backend doesn't explicitly forbid LLM assisted contributions. Which is enough to be on that list as you can see in the social media table. Can't speak for frontend since that is mostly hj's realm. And besides that, there are some small chunks of LLM code in the backend already.The reason why GTS is the only alternative is because they explicitly refuse contributions if they were in any way done with LLMs.
(DIR) Post #B55ArSIOVKMVqHy81o by SuperDicq@minidisc.tokyo
0 likes, 0 repeats
@phnt@fluffytail.org @hj@shigusegubu.club Yeah I personally do not subscribe to this definition of "slopware". I don't see an issue with small sections of code (that you would otherwise copy from stackoverflow) being generated by LLMs.As long as the LLM code is properly understood and verified by humans it will not make a difference if a human wrote it without LLM assistance.Still don't want to encourage people to use LLMs for software development yet though, as they are all still very much proprietary.I would consider projects to be "slopware", when developers are just generating code and not even actively trying to understand the output, and just pushing it as long as it "works".
(DIR) Post #B55B6mA4vSBRnbNtxY by SuperDicq@minidisc.tokyo
1 likes, 0 repeats
@icedquinn@blob.cat @hj@shigusegubu.club Librewolf does its best to remove as much LLM generated stuff as possible from the Firefox codebase, but I'm going to assume that this is a losing battle as they can push out more slop faster than you can remove it from your codebase.
(DIR) Post #B55BL1v58NRej5APJY by mkljczk@fediverse.pl
0 likes, 0 repeats
@phnt @SuperDicq @hj they should add iceshrimp.net and snac2 to the list then, listing gotosocial as a misskey alternative is a joke
(DIR) Post #B55BL2B2B4CnWYd9BA by SuperDicq@minidisc.tokyo
0 likes, 0 repeats
@mkljczk@pl.fediverse.pl @phnt@fluffytail.org @hj@shigusegubu.club They seem to have a very biased preference for projects who are explicit and vocal about excluding LLMs.
(DIR) Post #B55BovB6lOHiVNus3U by icedquinn@blob.cat
0 likes, 0 repeats
@SuperDicq @hj @phnt its funny when they include projects where the policy is basically 100% inspection, because that's what the high performers who stayed that way actually do.i do think the economic centralization is kind of stupid and normies should be pushing against that (but they never really do) and the transformer cult is deeply deluded, though
(DIR) Post #B55CKbp4GJDtk3iW0m by mkljczk@fediverse.pl
0 likes, 0 repeats
@SuperDicq @hj @phnt that’s why i mentioned two projects with explicit ‘AI’ policy
(DIR) Post #B55CKcCoq0DovipU24 by mkljczk@fediverse.pl
1 likes, 0 repeats
@SuperDicq @hj @phnt someone should start drafting AI_POLICY.md in pleroma-meta so there's a chance it gets merged in 10 years
(DIR) Post #B55CKcQI1uztbV8F1s by SuperDicq@minidisc.tokyo
0 likes, 0 repeats
@mkljczk@pl.fediverse.pl @hj@shigusegubu.club @phnt@fluffytail.org I honestly think there's no need to write an "AI Policy".If you submit a pull request and it is slop code, you're not going to merge it. That's basic fucking common sense and that shouldn't have to be a "policy".
(DIR) Post #B55CZeJ5JB0nn2veC0 by icedquinn@blob.cat
1 likes, 0 repeats
@SuperDicq @hj @phnt > instead of hise use dpfthis list includes things that aren't even alternatives. they seriously suggest "just use some C libs" as an alternative for an entire sampler DSP> don't use ruby because matz used claude> [laster in list] prefer this site generator written in rubynot even internally consistent:ablobcatgooglytenor: thanks for the morning joke i'm gonna get coffee ahah
(DIR) Post #B55HBKnegNcs6xS9su by phnt@fluffytail.org
1 likes, 0 repeats
@mkljczk @SuperDicq @hj Backend should also adapt frontends CoC to be on the same line.
(DIR) Post #B55KpTFOzgG9kyGHb6 by madengineering@mastodon.cloud
0 likes, 0 repeats
@SuperDicq @hj @phnt The problem with llm generated code is primarily one of motivation, the people pushing for it are doing so because they don't want to understand, they don't want to verify, they just want to fart this out and be done with it.
(DIR) Post #B55KpTVM2N1IYRj1Si by phnt@fluffytail.org
1 likes, 1 repeats
@madengineering @SuperDicq @hj That is largely an oversimplification with no bearing in reality. Some use LLMs as an assistant or a rubber duck, because they want to verify that what they wrote doesn't have unintended behavior or that it can be simplified while still being functionally same. Some use LLMs to fill in their lack of knowledge on a subject and/or check that they understand it correctly. Putting everyone in two baskets marked as "lacks motivation, too lazy to verify/understand" and "everyone else" is dumb.
(DIR) Post #B55LyKRLD13N08dLbk by phnt@fluffytail.org
0 likes, 0 repeats
@icedquinn @SuperDicq @hj It's open sores alternative.to with about the same quality.
(DIR) Post #B55LyKnfryuy7P5BQ0 by icedquinn@blob.cat
0 likes, 0 repeats
@phnt @SuperDicq @hj realistically if you want a foss sampler without hise you're looking at linuxsampler or shortcircuit.the surge/shortcircuit teams would be considered slop*, though, so you're stuck with linuxsampler.linuxsampler also hasn't been relevant in a decade because nobody uses gigastudio anymore and the one sample editor for it crashes (and upstream never accepted a patch that fixed the save segfault.)* they're in the 100% inspection camp, afaik
(DIR) Post #B55OU2ASsbn9VQzBc8 by SuperDicq@minidisc.tokyo
0 likes, 0 repeats
@phnt@fluffytail.org @madengineering@mastodon.cloud @hj@shigusegubu.club Yeah the people who just make pull requests with code they don't understand are the worst offenders I think. They don't want to put in any effort double checking their code or trying to understand it so they just shove that work onto the maintainer instead as a big "fuck you, go figure out what my chatbot wrote".Not every LLM user is like this and I think there are some legitimate usecases for LLMs, but most people overhype them.
(DIR) Post #B55QFNxRmZw6eR2Wci by Phobos@eientei.org
1 likes, 0 repeats
@icedquinn @phnt @SuperDicq @hj Imagine that, deranged people not being consistent. Shocker that ain't it.
(DIR) Post #B55QImAqCjQccmvBei by phnt@fluffytail.org
0 likes, 0 repeats
@SuperDicq @hj @mkljczk It's the same line of thinking as with CoCs. Trying to codify basic common sense like don't be an asshole and pain in the ass to deal with. Except now it is "don't spam us with irrelevant and broken patches you don't understand". You don't need written rules to justify banning someone for either of those.If it is outright spam, of course I'm not going to touch that. But I'm also not going to approve/merge something that I'm not willing to deal with later. Whether it was written by an 'AI' or not changes nothing in that balance. Besides that, you can never be 100% sure that it wasn't just written by a new contributor that doesn't understand the codebase to the level required to make a proper patch. And things that can't be enforced 100% correctly shouldn't be rules in my opinion.So if it is clearly not written according to some level of standard I'm happy with, I'll gladly offer feedback and advice to the contributor no matter if it was written with an LLM or not. Because then the contributor might come back later with some other contribution armed with more knowledge then before and won't need as much assistance as before. Or might even not need an LLM. If the contributor can't do it to the standard, I can always come back later and fix it for them when I have the time. All I ask is to not be spammed with walls of text that don't make any sense and clearly weren't proof read. That's it.
(DIR) Post #B55QImTz3Yjza9sTUe by mkljczk@fediverse.pl
0 likes, 0 repeats
@phnt @SuperDicq @hj CoCs are mostly about how issues are addressed and not about defining being an asshole an pain in the ass to deal with
(DIR) Post #B55QImjE8svyLR0eFk by cjd@pkteerium.xyz
0 likes, 0 repeats
CoCs are like barred windows - they're not a problem in and of themselves, but when you see them you want to ask "why is this here?"
(DIR) Post #B55QImuZSi0YucJhw0 by phnt@fluffytail.org
1 likes, 0 repeats
@cjd @SuperDicq @mkljczk @hj They are a good reflection of a community. If a community needs codified rules on how to behave and how to deal with those that break the rules, it's probably not a good community in the first place. Moderation is something you shouldn't need much of, if you've cultivated a good community.
(DIR) Post #B5651rdGkjPSkfGykS by mkljczk@fediverse.pl
0 likes, 0 repeats
@SuperDicq @icedquinn @hj yeah you can’t maintain a soft fork like this for too long
(DIR) Post #B5651roy3EldKwkJyy by mangeurdenuage@shitposter.world
0 likes, 0 repeats
@mkljczk @icedquinn @SuperDicq @hj Give me 4M a month and FF will rise again.