RE: https://ruby.social/@aesthetikx/116189523597378713
Related: Claude code support for #freebsd doesn’t exist either.
Every discussion on this veers off from “llms suck so …or I don’t need llm” but misses: users should have the ability to make their own opinion based on use.
@aesthetikx @bms48 I noticed that when it starts up, the system message at the bottom now basically says that the NPM install approach has been deprecated on all of their platforms. (Mac, Linux, windows).
This would make FreeBSD a second class citizen. Obviously, people have different feelings about whether that’s a good thing or not. I think it’s to the negative.
@benjamineskola @aesthetikx I’ve read your posts elsewhere and you’ve been consistent on this, and I think you to be an honest interlocutor: but you will grant CFCs and lead based paint and petroleum are legally PROSCRIBED outside certain contexts, no?
I will agree, that if LLM clients are proscribed no OS should support them at all.
So this is closer to banning cars (not illegal) and therefore is largely a question of economics and personal agency; or perhaps sustainability-motivated vegetarianism.
@sgharms @aesthetikx I think you have it slightly backwards. CFCs aren’t bad *because* they’re banned; they’re banned because they’re bad. Before they were legally banned, they were still a bad idea, and it was a good idea to avoid using them.
I don’t think ‘it’s not literally illegal to use LLMs’ is a good enough reason to support their use.
We agree that (1) LLM tools + infra are not proscribed and (2) that the badness of CFCs was due to physics not law.
But, should FreeBSD decline to advocate for a given software(s) to be available on its system? I think no. It is a stable, general-purpose OS, and it should ensure that it hosts at a level that rivals the other OS'. It ships with technology that could do bad things, but we leave it to users to not do them.
So I'm advocating that FreeBSD be supported as a first-class platform for ollama and for Claude Code.
@sgharms @aesthetikx I don’t think I’d necessarily say that the FreeBSD project should *ban* such software from being available for its users. (Undecided on this point. If they refused to accept it into ports, I’d be happy; but if they accepted it, I don’t think I’d blame them.)
I just don’t think anybody need make an effort to make it available. I think the fewer people who have access to it, the better.
Basically: what good is actually done by this being available? What is actually achieved by advocating for it to be available on more platforms? etc.
Great. It seems we agree that a ban is not the answer.
As for what actual good is happening, I made you a blog [post] with my experience.
I hope you see why I want AI tools on FreeBSD: porting a desktop package; reporting/resolving driver issues; helping maintain the ports tree. Thoughtful use of AI tools can help.. Advocacy for Claude Code on FreeBSD may help me do more for FreeBSD!
There's also a visibility problem. I don't want *nix culture to treat BSDs as has-beens where they don't get tools/builds (ghostty, claude code, ollama). I was [mocked] on FreeBSD forums for suggesting such lack of toolchain / insularity could lead to irrelevance among the new blood who are used to such tools.
I got into FreeBSD thanks to the Changelog podcast with @allanjude and a desire to avoid the drama of FSF. Given changes in Linux-land, and the ongoing 'laptop' project for FreeBSD, it may yet be the free OS we deserve.
--
[post]: https://stevengharms.com/posts/2026-03-09-my-successes-with-ai-augmented-code-as-of-march-2025/
@sgharms Going back to my original reply: personally, I don’t find “I found this tool useful” to be a very persuasive argument. My point was (though perhaps I didn’t make it very clearly) that I think some of the reasons to be against generative AI are unrelated to whether individuals find it useful.
I think the problems it causes for the environment and society in general, as well as the FOSS ecosystem and community in particular, all are cause for concern even if individuals find the tools useful.
And, personally — I’m just a random user so it’s not like my opinion counts for more than anyone else’s, but I would move away from FreeBSD if the use of these tools became widespread in its development.
There are systemic degradations to work against (environmental, the practice of intellectual inquiry); I, too, remain committed to doing so. But while I don't own a car, i recognize their use for many, and want to see their externalities contained.
And I grant that the recent vim maintainer bots going insane at each other certainly suggests AI enabled FOSS _can_ go very wrong, indeed. If such were to become the standard of FreeBSD commits, I'd likewise prepare to abandon.
If thoughtful technologists don't investigate and frame AI while understanding the lay population and their needs, then that need will be serviced by AI-bros and grifters. Better that those who can lead, do lead (and that requires it running on my platform ;)).
You've been generous and thoughtful throughout and I've felt rewarded by this exchange.
@sgharms The car analogy is an interesting one. I'd strongly advocate for public transport and I think expanding that would be vastly preferable for the vast majority of situations. But not all situations, admittedly: in rural areas it's probably never going to be feasible, and for people certain disabilities a car might remain more practical.
And yet I wonder if LLMs are really comparable? My perception (and I acknowledge that my bias might play a part here) is that the vast majority of the demand for them is essentially manufactured. People use LLMs for things they were perfectly capable of doing by hand a couple of years ago, and some seem to have genuinely convinced themselves that it would be impossible to survive without the LLM.
To my mind, a lot of these users are more like the people who determinedly drive a car in central London despite the costs (individual and systemic) and despite the ready availability of public transport (which in turn is made worse by private motor traffic), not the people who drive because it's their only option.
(But I also try to be conscious of actual justifications: e.g., as an accessibility tool. I would not want to rule those genuinely-necessary uses out; but I just don't believe that those are the majority of cases.)
@benjamineskola @sgharms if I may, a perspective from a long time FreeBSD user who finds LLMs useful. I'm currently running a small (agri) farm after startup/software career, and been using FreeBSD since 5.1.
There's only so much time and energy left after outdoor work, but my curiosity for learning more about the OS and building tools for myself has not waned.
I find LLMs useful to keep tinkering with my home server instead of abandoning the hobby (which happened, and last year I got back).
RE: https://mastodon.bsd.cafe/@grahamperrin/116220642823558416
@sgharms noted, your recent comment: "… we can’t tell users “You’re wrong; you don’t want this.”.
@sgharms I'm confused by <https://github.com/anthropics/claude-code/issues/29374#issuecomment-4031281907> as a _closing_ comment.
What am I missing?
Steven G. Harms
Bruce Simpson, Ph.D.
deSilva
ben
Harshad Sharma
Graham Perrin