Users points out in comments how the LLM recommends APT on Fedora which is clearly wrong. I can’t tell if OP is responding with LLM as well–it would be really embarrassing if so.
PS: Debian is really cool btw :)
You must log in or # to comment.
Seems like everyone taking a left turn into crazy ville ……
This is a good approach.
- It’s an mcp server, a “bridge”. A standard way LLMs could talk to your system. It’s not an LLM. It doesnt mandate an LLM. It doesn’t tie you to a specific LLM
- It’s optional. Don’t use it, or don’t install it. No harm done. Even if it’s installed and running, if you don’t use an LLM with local access, no harm done.
- Even the increased attack surface is not a big deal since it is local, optional, and focuses on reading statuses rather than executing actions
- It’s an open standard. If you decide to use it with an LLM but don’t like the results, try a different LLM
This is a community for criticizing LLMs. There is, therefore, no positive interest in a bridge for LLMs. This community is optional. Don’t subscribe, or don’t read it. No harm done.
Yeah, I suppose so. I find plenty to hate about the way companies and too many individual apps talk about LLMs. I really hate that it’s one of the metrics my employer looks at. I always hate how wasteful speculative bubbles like this are.
But maybe this isn’t the place for me since I see some good use cases and appreciate the few times someone does it right.
- I hate how companies like Google and Microsoft are putting LLMs everywhere, making things worse for everyone, forcing everyone to deal with it
- I find it strange that there’s an LLM in my car but it Carrie’s a decent conversation and it can’t actually do anything
- I like this approach of building the mcp server. Keeping it read only, keeping it optional, and leaving the actual LLM open. I also like Apples approach where they seem to care about privacy, about executing on device, and about taking their time to put it in useful places
Yeah, most people, myself included, don’t like AI on principle, but there are valid use cases for it, and not having the capability of integrating with AI tools is going to be a dealbreaker for someone.
That said, I’ve heard MCP is a bit of a shitshow of a standard and is woefully inefficient.
Yeah it is. I have no idea how you more efficiently do that task, but there’s got to be a better way
That’s exactly why I migrated from Fedora to Arch, I want a distro with little to none corporation influence.
To me, the real Linux experience is Arch.
I didn’t want to wait for some bad decision to be made. Time is showing I took the right decision.
Now I’m thinking of doing the same because the entire reason I chose fedora was to take advantage of their institutional support
But if those institutions are going to sell the floor right under me to ai then there’s no point
Check out Debian as well! (or lmde) I have recently switched from Arch to Debian (with sway) to have some more stability. Love it so far, I don’t bother about configs or having by dependencies anymore and can just focus on actually doing some work. Arch has a really active community though with a big repository (which quality varies a lot) and nice documentation.
I might try Debian testing branch first because, frankly, I do not want to deal with arch linux…I just want to use my computer
Check out PikaOS, it’s pretty much pure Debian, but for gaming, makes updating drivers and managing games easy
IMO there are two main Linux camps, and most users fall somewhere in-between. Rolling OS lovers who want to tinker (eg Arch). People who want stability over everything (eg Debian).
The only truly wrong answer is paying for RHEL.
Same. I was on Fedora Atomic but the stench hardened so I jumped over to nixOS a few weeks back. Glad I did.
i’m not ready to go to arch yet since i’m not comfortable enough but i’m curious; is arch based in us or outside? i know mint is based in eu/ireland. i do wish i went with mint debian instead of mint ubuntu but next time
It’s international, but there are a lot of Europeans: https://archlinux.org/people/developers/
CachyOS has its roots in the Polish Arch community if I recall correctly. It’s much less daunting, and I’d highly recommend it.
Check out Garuda. It’s a really stable, well maintained version of Arch. I’ve been running the same install of it for over 9 months with only one screw up that garuda’s own health tool fixed for me.
It’s been a great way for me to learn Linux and Arch, and it rivals Bazzite in my own gaming benchmarks.
thanks! ill look at this
its also the best final fantasy eikon.
MCP servers are fucking bizarre too. I expected them to be like a normal API server, not realizing how little LLM developers wanted to do anything but make more chatbot. The default implementation is to read from stdin and write to stdout and have it launch a process. There’s “streaming http” as well. It’s all so that it can have a fucking “chat” with the server involved. 🤢
Do not hand these things unsupervised system access, they will do bizarre bullshit and ruin your system.
Et tu, Brute?
gpt-oss 20B
See all the errors in that rambling wall of slop (which they posted and didn’t even check for some reason?)
Trying to use a local LLM… could be worse. But in my experience, small ones are just too dumb for stuff beyond fully automated RAG or other really focused cases. They feel like fragile toys until you get to 32B dense or ~120B MoE.
Doubly so behind buggy, possibly vibe coded abstractions.
The other part is that Goose is probably using a primitive CPU-only llama.cpp quantization. I see they name check “Ryzen AI” a couple of times, but it can’t even use the NPU! There’s nothing “AI” about it, and the author probably has no idea.
I’m an unapologetic local LLM advocate in the same way I’d recommend Lemmy/Piefed over Reddit, but honestly, it’s just not ready. People want these 1 click agents on their laptops and (unless you’re an enthusiast/tinkerer) the software’s simply not there yet, no matter how much AMD and such try to gaslight people into thinking it is.
Maybe if they spent 1/10th of their AI marketing budget on helping open source projects, it would be…
I have been using gpt-oss:20b for helping me with bash scripts, so far it’s been pretty handy. But I make sure to know what I’m asking for and make sure I understand the output, so basically I might have been better off with 2010-ish Google and non-enshitified community resources.
Yeah, that is a great application because you can eyeball your bash script and verify its functionality. It’s perfectly checkable. This is a very important distinction.
It also doesn’t require “creativity” or speculation, so (I assume) you can use a very low temperature.
Contrast that with Red Hat’s examples.
They’re feeding it a massive dump of context (basically all the system logs), and asking the LLM to reach into its own knowledge pool for an interpretation.
Its assessment is long, and not easily verifiable; see how the blog writer even confessed “I’ll check if it works later.” It requires more “world knowledge.” And long context is hard for low active parameters LLMs.
Hence, you really want a model with more active parameters for that… Or, honestly, just reaching out to a free LLM API.
Thing is, that Red Hat’s blogger could probably run GLM Air on his laptop and get a correct answer spit out, but it would be extremely finicky and time consuming.
The comments on the blog post are eviscerating the author. Literally nobody asked for this, and they’re making DAMN SURE they know that.
And now the comments are closed. Embarrassing.
i mean the entire concept of red hat is corporations profiting off of open source
That’s a little disingenuous.
Redhat markets support and professional responses in the form of adjusted builds to CVEs on an enterprise distro of Linux.
I mean, yeah, they absolutely technically profit, and open-source is the root of their classic product, so you’re not wrong there; and I’m not saying so. You’re framing it as if they do nothing but take the source, schlepp it out, and collect protection money and keep it themselves.
This kind of characterization completely sidesteps an important truth about open-source devel: everyone involved with projects like the kernel, for instance, work at a shop like redhat and are paid to do the open source work.
Let me say it again: this corp who is profiting off of open source happens to also be donating killed, paid labour to open source. Look at the leaders of the kernel dev effort and you’ll see google, arm, rh, Intel, and suse, all paying for the labour to keep it going.
Stop characterizing them with language suggesting they’re slimy leeches. They make bad decisions a-plenty, but the biggest thing their involvement provides is continuity and lifestyle security for people maintaining what you hold dear.
Condemn them for their stupid decisions - Ansible, Systemd, IBM buying redhat and accidentally speeding up their enshittification because #ibm, etc. But given how little the foundations support this project (the only one I looked at: I’m only slightly less lazy than the person who didn’t even do this) I suspect their cozy little arrangement where they make bank and you get free software to play with is benefitting you more than you’ll admit.
I love how a red hat product forgets about proper packaging. Pips? Containers? Fuck off. RHEL has an E in it, ya fuckwits.
