Login
You're viewing the front-end.social public feed.
  • Aug 17, 2026, 2:30 PM

    Need some advice/opinions/ideas:

    This account's purpose is to promote self-hosting, community hosting and managed hosting, in order to encourage independence from giant tech corporations.

    But some of the FOSS that these self-hosting platforms run on are now being generated by LLMs. It's making FOSS dependent on corporations and unethically created software, when the purpose of FOSS is to be independent and ethical.

    I'm not sure how to proceed as so many FOSS projects are allowing LLM 😞

    💬 10🔄 6⭐ 0

Replies

  • Aug 17, 2026, 4:16 PM

    p.s. Wondering if the whole concept of FOSS needs to be redefined to include a rule about not using LLMs. It's difficult to see how FOSS is meaningfully free or libre if its maintenance depends on planet-destroying plagiarism services controlled by giant corporations.

    💬 6🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 17, 2026, 2:40 PM

    @homegrown I really appreciate that you're taking this question seriously. A neutral and transparent approach could be simply indicating if a project includes LLM-generated code or what their policy towards LLM contributions is. This way, your followers can take their own, qualified decision. Unless, of course, you prefer to not propagate AI-related projectes at all. It's your account, and you can decide what to publish with it. Just be transparent about it.

    Thank you for your suggestions. I'll stay along either way.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 17, 2026, 4:33 PM

    @homegrown

    my position is that it's been irredeemably infested & that work needs to go under a different banner

    perhaps "post-open source"

    perhaps "ethical source"

    perhaps "permacomputing"

    perhaps some combination of the above, and some new things

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 17, 2026, 4:35 PM

    @homegrown I think adding something in about "hand-coded" is accepting the framing of the corporations pushing this stuff. Claiming other people's work as your own with no credit given is already against the FOSS ethos. It doesn't matter if you use a neural network to shuffle the chairs around a bit first.

    💬 0🔄 0⭐ 0
  • Aug 17, 2026, 5:24 PM

    @homegrown Eh, just because LLMs are involved in a project's development at some point, it doesn't mean that continued use of LLMs is required to maintain the project. There are many projects that have accepted LLM contributions, or in some (fewer) cases have been entirely vibe coded, but would be completely fine from a future maintenance perspective if LLMs suddenly stopped working or became available.

    The only ways I can think of in which LLM use would make a project non-FOSS are if the project actually includes an LLM - that is, the model weights are part of the "source code" - or if LLM output is held to be under copyright protection as a derivative work of its training data, but that seems unlikely. Maybe there are some other cases I'm missing, but hopefully only niche ones.

    So, my take on it would be, if you want to avoid promoting projects that have used LLMs in their development, that's totally fine, but in that case you'd probably be limiting yourself to a stricter criterion than just FOSS, and it would be only fair to your audience to be clear about that.

    💬 1🔄 0⭐ 0
  • Aug 17, 2026, 5:29 PM

    @diazona

    What happens when LLM use displaces and deskills contributors? This is an issue that pilots noticed, to the point where they had to bring in mandatory manual flying periods to prevent deskilling due to automation (and autopilot is a lot more reliable than LLMs).

    Also, regardless of skills etc, LLMs are unethical to use due to their planet-burning and plagiarised nature. This alone would make me suspicious of projects that use them, as they seem to accept unethical contributions.

    💬 1🔄 0⭐ 0
  • Aug 17, 2026, 6:06 PM

    @homegrown I don't think any of that changes my argument. Like, it's far from a given that LLM use in a project is going to displace and de-skill contributors. And even if it does, de-skilling often has as much or more to do with how the contributors choose to use LLMs than anything done at the project level. That is to say, people can use LLMs in a careful way that improves their coding ability, or in a thoughtless way that replaces it. (I have personal experience with both 😅) Both extremes and everything in between get lumped under the heading of "allows LLMs", but it doesn't necessarily make sense to treat them all identically. And only a small fraction of this LLM use is going to clash with the definition of FOSS. Like, even if LLM use is displacing and de-skilling contributors, there's a big difference between that and making it so a project is fundamentally unmaintainable without LLMs.

    And if you want to avoid promoting projects that use LLMs because of the environmental impacts and your concerns about plagiarism, you can certainly do that; it's just that, again, that goes beyond the definition of FOSS. (Except for the copyright argument I brought up in the previous post.) I could offer an argument as to how many people who use LLMs don't see those ethical issues in the same way - not just "people don't care about ethics", it's nuanced - but it doesn't really make a difference to my main point, which is that what you're considering is a more restrictive criterion than just FOSS and should be described accordingly.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 18, 2026, 10:30 AM

    @homegrown a large proportion of FOSS has always been produced by capitalist companies with often questionable morale. How did you manage to come to terms with that? This is the way you'll come to terms with AI. Boycott the most egregious cases, live with the rest.

    On the last point: you can redefine FOSS to anything you want, attempts have been made already (various "fascists not allowed" licenses, for instance). The question is how to avoid being locked in your small corner.

    💬 0🔄 0⭐ 0
  • Aug 17, 2026, 4:53 PM

    @homegrown use of LLM in software development is inevitable, especially that the quality of generated code improved a lot over time.

    But the most ethical use of LLM is actually in producing FOSS, as the models have been trained on FOSS projects mostly.

    Also, it's ethical to disclose the use of LLM in the project description.

    💬 0🔄 0⭐ 0
  • Aug 17, 2026, 5:36 PM

    @homegrown A similar list of self hosted items’ recently showed all LLM touched projects, which was a real shame to see.

    On the one hand: it’s lovely that people are excited about having made something (even if they don’t actually make it, but hopefully you know what I mean!), but on the other: the hell am I running any of that.

    I think I would keep highlighting those without significant LLM involvement, perhaps with being open to retracting previous suggestions when they go to broken slop

    💬 0🔄 0⭐ 0
  • Aug 17, 2026, 5:52 PM

    @homegrown To be honest, this makes me depressed and wanting to see myself out of this industry.

    I have a very defeating feeling, everything is slop now and the ability to think shines out of its absence.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 17, 2026, 11:34 PM

    @homegrown so basically, to sum it up, "AI is bad because t is bad and your project should die if you use a LLM because AI is bad and did I mention AI is bad and AI is bad because AI is bad".
    thanks for the laugh for the laugh today, apritiate it.

    💬 2🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 19, 2026, 3:15 AM

    @chroma @adisonverlice @homegrown If i may say, People vibecode for valid reason. For example, i vibecode because i tried learning programming languages. I tried 3. It all went to my head, i somehow couldn't understand all that. This is why i use Claude Code to help me. Do i test it? Absolutely. Do i report bugs to it before i give it to people? Yes. But if you want to make software professsionally? I would go AI assisted

    💬 0🔄 0⭐ 0
  • Aug 19, 2026, 6:04 PM

    @spacepup @chroma @adisonverlice

    That's not you programming though, that's the LLM stealing some other programmer's work and failing to credit them or ask permission.

    You could ask a human to program for you if you can't do it yourself?

    I cannot draw, so I asked a human artist to do artwork for my website. They did the artwork, I paid them for their labour, and it created something new.

    💬 0🔄 0⭐ 0
  • Aug 19, 2026, 6:03 PM

    @adisonverlice

    GenAI / LLM is bad because:

    -It is destroying the planet
    -It is stealing people's labour
    -It is plagiarising without permission, compensation or credit
    -It is giving AI companies the ability to control society's discourse
    -It is driving up the cost of computing to the point where ordinary people cannot afford it any more
    -It is deskilling the people who use it to the extent that they become dependent on it to do stuff they could already do before

    💬 1🔄 0⭐ 0
  • Aug 19, 2026, 6:31 PM

    @homegrown again, i'm not saying AI is the end-all, be all of society.
    I think even @alexchapman who calls out AI extreamests and advikates for AI, is stupid, I think he knows there are several problems with AI companies.
    yes, it does pollute the plannet is generate a lot of noise, some of that sound does actually does make people sick. I've had it happen to a friend before, where there are vibrations that can cause headake, etc.
    I also agree that they do take public data, but here's my policy on this.
    a lot of my shit, for example, my essays I published, aren't AI generator, but here's myp olicy. if you publicize it, expect to be rammed by AI, and don't complain because you published it for the world to see.
    if you don't want it scraped by AI, please, do not publish it, keep it outside of the world
    otherwise, any AI will be able to scrape it.
    this is especially true as AI gets more agentic in its functions.
    but you're asically saying generative AI as a hole is bad, which I think is unfare.

    keep in mind some people have to use AI, for example, I am blind, so I have to use generative AI to see images, which uses Gemini nano on my device.
    o wait I forgot, you don't care because AI is bad bad bad bad

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Aug 18, 2026, 6:01 AM

    @homegrown to me this whole line of discussion is similar to some of the fsf people running around and discouraging the use of MIT because only some obscure 40 page AGPL++ 1.23 or later is really free and really open source.

    Sure - carve up the community along another arbitrary line while not applying the same standard to the big commercial actors in your real life.

    💬 0🔄 0⭐ 0