Open Source Software Community - it's about ethics in Code of Conducts

  • 🔧 Site instability resolved. You can report double-posts and broken attachments. For bigger issues, use the Technical Grievances thread.
    🇵🇦 Nuestro primer dominio localizado está en español en kiwifarms.pa. Our first localized domain is on Spanish on kiwifarms.pa.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Destacado
Ferryman dijo:
My only worry is that this adds an extra burden on reviewers to scrutinize sloppy code, and Ludo is a softie that I'm not sure will be willing to pull commit access to people who just pump out slop. He's been pretty ardently against AI so maybe it will be different if shit packages start getting submitted.

Another idea is to chunk the big "Guix" channel into smaller ones, one for "pure", no-AI core code, and a second "guix-modules" o algo that would contain packages with slightly laxer quality control rules. Both of these assume that the FSF won't put the kibosh on AI code in their upcoming decision.
Presiona para expandir...
Personally, as someone who actively uses LLMs for writing banal guix packages, I think that AI use should only be permitted there, must be marked, and relegated to smaller packages. That, and some translation work, seems to fit the actual capabilities of AI models when it comes to real productivity.

Anything more than that is inviting an entire Ganges' river worth of trash code into the operating system. Guix is on too good of a path right now to invite in problems, even if its for the realization of more gains.
 
Personally I think the whole open sores / free socialism model was a mistake from the beginning, and AI is being used as an excuse to avoid admitting that... because what is the functional difference between bad code written by AI and bad code written by a jeet, a chink, or a tranny? The correct answer is FUCKING NONE. If this leads to projects rejecting outside contributions then it can only be a good thing. Unfortunately this will probably also lead to retards cranking out slopcode via Clod and trying to pass it off as their own work, so nothing much will change unless more rigorous programming practices are enforced that aislop-artists have real difficulty employing (and if they do manage to employ, thus reducing the incidence of error, then MISSION FUCKING ACCOMPLISHED).
 
Ferryman dijo:
I want your guys' opinion on something. The Guix mailing list has been conducting a GCD (Guix Consensus Document) debate as of late regarding the use of AI/LLMs in Guix System. Its a pretty even split between people wanting to either outright ban AI contribs, minimize them to only allow code reviews & AI assistance, or hold people that use AI at a higher degree of responsibility for code quality. All three seem sane in my eyes, but one thing that doesn't is this obsession with ecology and the carbon footprint impact of AI. I find it absurd that this gets brought up constantly and it honestly feels very subversive in the sense that it throws a wrench in normal discussion for the purpose of distraction or conversational hijacking. Sure, data centers are messing with a lot of nature and a lot of people, but if anyone really cared about ecology they'd be actively pressuring every nuclear power to dump their entire payload on India. Am I in the wrong thinking of the ecology angle as a red herring?
Presiona para expandir...
Things you can do to address ecological concerns other than not using AI:
  • Use smaller models and local LLMs instead of data center-provided massive LLMs. You can get a lot of mileage out of simple local models as a glorified code completer or to write boilerplate stuff, even develop whole complex features granted you have to guide it further, and That's A Good Thing because it lowers the risk of producing a massive chunk of software you don't understand at all.
  • If you're used to building resource-hungry Electron apps, use AI to help you build native software and understand pipelines. More energy developing the app? Certainly. Is it offset by the energy saved running the app if you got a big userbase? Also yes.
  • Use AI as an extra pair of eyes to find stuff that may already exist that does the stuff you need and adopt it, maybe even contribute to it. Many things aren't easy to find with a simple Google search, especially these days when all big search engines are enshittifying search to drive people to their LLMs instead.
 
Ecological concerns are nothing more than moronic virtue signal, resources exist to be used as they have no value outside of being used, the only consideration when it comes to resource usage ought to be cost as it indicates the point at which their usage is no longer beneficial. There is nothing wrong with burning coal, killing fish, chopping down trees etc in order to fuel shitty jeetware electron app as long as the benefits of it out value cost of the materials needed to do so. I recomend all enviromentalists to follow logical consequence of their beliefs and reduce resource usage to minimum by roping themself.
 
WielkiWapiesz2137 dijo:
Ecological concerns are nothing more than moronic virtue signal, resources exist to be used as they have no value outside of being used, the only consideration when it comes to resource usage ought to be cost as it indicates the point at which their usage is no longer beneficial. There is nothing wrong with burning coal, killing fish, chopping down trees etc in order to fuel shitty jeetware electron app as long as the benefits of it out value cost of the materials needed to do so. I recomend all enviromentalists to follow logical consequence of their beliefs and reduce resource usage to minimum by roping themself.
Presiona para expandir...
"fuck you, I've got mine" Idk if that's boomer or jeet mentality, but it's bad regardless.
Using fishing and lumber is especially brain dead take. As with proper management you can get sustainable indefinite resource extraction. As opposed to crashing everything and having to wait as it regenerates, if it even can.
Just because some concerns are overblown, does not mean there isn't some validity to them.
 
Cherenkovblue dijo:
Personally I think the whole open sores / free socialism model was a mistake from the beginning
Presiona para expandir...
It might appear so, but that's projecting the current state of open source onto its entire history. We've seen a huge influx of third worlders who do not follow the social contract; they have no feelings of guilt or duty. That along with the chud and tranny (and more) social media hiveminds is what's causing the deterioration. Where are the new Torvalds' or de Raadt's? Open source has been flooded by spineless goyim.

Cherenkovblue dijo:
because what is the functional difference between bad code written by AI and bad code written by a jeet, a chink, or a tranny? The correct answer is FUCKING NONE.
Presiona para expandir...
Agreed. We must address the sociological problems, not per se the tools which empower the masses of retards. For bigger projects this means gatekeeping. The Ladybird project has done this, rightfully so.
 
Unintelligence Score dijo:
Use smaller models and local LLMs instead of data center-provided massive LLM
Presiona para expandir...
It's really training that uses the most. Running the model itself after it's trained isn't that big of a hit

If everyone switched from dc AI to local AI the same energy use would happen, but split between consumers. Possibly more since weaker hardware has to work longer

There's not much really that the individual can do. Most of what you do to go green(not an insult, I respect people who genuinely want to protect the environment), is overshadowed by large corporations or countries just dumping 10 lifetimes worth of your waste in 2 minutes. The smog in LA or NY or wherever went away after legislative changes, people didn't just decide on their own

We're going to be using more and more energy the more days pass. That's a given, there's no way you can stop that.

Proper solution is nuclear power so we have enough energy to not care. Submarines run on nuclear power and they're just fine, datacenters can have their own nuclear powerplants too and the problem is solved. They take time, but the sooner we start the sooner we can benefit
 
0gh dijo:
If everyone switched from dc AI to local AI the same energy use would happen, but split between consumers. Possibly more since weaker hardware has to work longer
Presiona para expandir...
More local LLM would at very least reduce the need for all these datacenters being built.

With datacenters you get extra energy costs, construction materials, hardware being bought explicitly for them and potentially cycling at a fast speed, so that means the same hardware becomes much more expensive for normies. Every place where the new datacenters are being built are seeing energy costs because some retard is unable to stop himself from asking Grok to make a car wear a bikini.

There's also no fucking way that cloud LLM services aren't the world's worst privacy nightmare (so far) and that users aren't being datamined to hell at a level never before possible.

Not to mention the push to move every computation to the cloud is just a fucking rent-seeking enterprise. Natural resources aside, I would like people, local businesses and particularly myself not see money simply evaporate because I need to write code or do work.

At the moment a lot of money is just being tossed into a black hole of profitabilty, but muldi-billion-dollar investors aren't stupid. Even if we don't see right now how they plan to turn the burning pile of money into profit, you can be sure they have several different strategies to get money out of it, and I am certain we will hate any strategy they eventually put to use.

I don't know man I just want to apt-get install local-llm-coding-chat-thing, open a wizard where I check the programming language(s) and task(s) I need it for and have the program determine from my hardware what models fit, pull those and run those properly sandboxed so it doesnt POST the contents of my .ssh folder, other credentials and any other personal data it finds to www.legit.gay when I ask for help figuring out how to cast shadows in an OpenGL scene. I don't like that the only "choice" most people ever see is which billionaire gets to rape their privacy.
 
Has anyone seen that proof-of-work bot detector called Anubis? I went to the UNESCO website a few days ago for research and I was greeted with an anime girl checking that I wasn’t a bot. Apparently it’s made in javashit and is super inefficient, so if you’re connecting from an old device it might take a good second for you to load into a site with Anubis. That’s assuming your browser even supports JavaScript. If you’re UNESCO, you might want to target developing countries, but because developing countries usually run older hardware, they might not be able to connect because of Anubis. Huge L!
Cherenkovblue dijo:
Personally I think the whole open sores / free socialism model was a mistake from the beginning, and AI is being used as an excuse to avoid admitting that... because what is the functional difference between bad code written by AI and bad code written by a jeet, a chink, or a tranny? The correct answer is FUCKING NONE. If this leads to projects rejecting outside contributions then it can only be a good thing. Unfortunately this will probably also lead to retards cranking out slopcode via Clod and trying to pass it off as their own work, so nothing much will change unless more rigorous programming practices are enforced that aislop-artists have real difficulty employing (and if they do manage to employ, thus reducing the incidence of error, then MISSION FUCKING ACCOMPLISHED).
Presiona para expandir...
Not every programmer is a jeet or tranny and AI is making actual good coders worse. This next generation of programmers are mostly brainfucked beyond belief from reliance on AI. I interact with grad students on the regular and even a lot of them don’t know what SICP is. We are losing the art of efficient programming, and that’s been obvious since JavaScript took over the web. AI just accelerates it.
I do think jeetcode is still better than AI, but even if it wasn’t, I still would rather code be written by a human than an AI.
 
check out these cool drawings @dcss made!
photo_2026-06-08_16-23-07.jpg photo_2026-06-08_16-43-02.jpg photo_2026-06-08_17-09-54.jpg photo_2026-06-08_17-09-55.jpg photo_2026-06-08_17-09-56.jpg photo_2026-06-08_17-45-40.jpg photo_2026-06-08_17-45-41.jpg
 
Gilgamesh dijo:
Has anyone seen that proof-of-work bot detector called Anubis? I went to the UNESCO website a few days ago for research and I was greeted with an anime girl checking that I wasn’t a bot. Apparently it’s made in javashit and is super inefficient, so if you’re connecting from an old device it might take a good second for you to load into a site with Anubis. That’s assuming your browser even supports JavaScript. If you’re UNESCO, you might want to target developing countries, but because developing countries usually run older hardware, they might not be able to connect because of Anubis. Huge L!
Presiona para expandir...
Whenever anyone mentions anubis i like to link this blog post by Tavis Omandy https://lock.cmpxchg8b.com/anubis.html. You may or may not know of him, he's a security researcher at google and has found numerous high profile vulnerabilities (such as cloudbleed). It's a good read,

twitter replies are fun too.
 
Última edición:
i don't know if anyone posted this?


Canonical posted an update [1] about their decision to switch to uutils
reimplementation of GNU coreutils [2]. In it, they detail the audit
performed by Zellic, which found 113 issues, 44 of which were assigned
CVEs. They go on to explain that as a result of TOCTOU races, they have
decided to continue using GNU 'cp', 'mv', and 'rm' in Ubuntu 26.04.
Presiona para expandir...

the jokes write themselves really.
 
Destacado
Debes iniciar sesión o registrarte para responder aquí.
Atrás
Top Abajo