Yeah the compact form factor and energy efficiency were the strongest point for me.

Yeah, unfortunately the TV app (Xbox for me) offering is really poor.

I've been playing with Kodi with the Jellyfin plug-in but it requires such brittle tower of themes and plug-ins to look like Plex that I've yet to make the switch.

The homescreen on the Xbox app is now full of "Apple TV", Trailers, free TV and the like. Enough of it that you can't see you own stuff on the homescreen.

Last time I searched I couldn't turn it off. But maybe I missed that setting.

I've used the Void userspace for a while as my WSL layer and I do really like the package manager.

Anyone actually running Void on real hardware? What's runit like?

I think as long as it's just an installer with an extra repo and the rest keeps pointing to the Void repos then that sounds manageable.

There's about 20 packages in their forgejo instance.

Doesn't mean I would pick it up but if you're maintaining that for yourself and your mates that's fine.

Thank you for spelling it out for me because I thought I was looking at a completely hallucinated AI article...

Author here. All hallucinations are my own. Now you point it out, I see why the jump in context from the terminal back to the tshirt font would give the wrong impression.

Honestly it was quite a whiplash to go from what looked like a good article to something that seemed completely made up. But I would chalk that up more to my reading comprehension than your writing.

I didn’t read it as generated text, but the context definitely threw me.

Right, are any of the things mentioned in the GP really required if you only want this init?

I know they've attached all these projects to the systemd brand because they thought it would beneficial but it's hard not to wonder if we could avoid all those discussions if the umbrella project was called something different...

No, you can just run the systemd init system and not use any other systemd programs.

I would probably run systemd-journald just because it's really nice to have a logging system which knows about the system services, but it's not required.

You probably also want to run udev, and logind comes close. Everything other than these four is absolutely optional.

I did have a terrible time setting up rootless podman in a proxmox VM with virtiofs.

I just couldn't get the ids to work. But I've not tried the specific thing they do here.

Ooh, I need to look into that id mapping technique.

I attempted to port my homelab quadlet setup to a disposable VM on proxmox with the state folder mounted via virtiofs and after losing a weekend I realised that it essentially wasn't compatible with rootless podman id mapping and I gave up.

But maybe I missed something...

Yeah, I ran into the same wall when setting the system up. I then reverted to the standard mount utility, which also supports id-mapping. But it is really not that known. See the links in the article that point to the source of the idea.

    > Network file systems can cause UID/GID mismatches and add network overhead. With VirtIO-FS, I use the Linux kernel's Virtual File System to translate the hypervisor's UID to the guest's unprivileged UID. This avoids exposing the host file structure. I utilize the `X-mount.idmap` fstab option for this. Documentation on this specific implementation is not easy to find. It builds upon the idmapped mounts feature introduced by Christian Brauner in Linux 5.12 [1] and its later integration with util-linux v2.39 into the standard `mount` utility [2].

    [1]: https://github.com/brauner/mount-idmapped + https://docs.kernel.org/filesystems/idmappings.html
    [2]: https://github.com/systemd/systemd/releases/tag/v254

Thanks I'll give it a shot when I have a free weekend.

I think I may have complicated things by using linuxserver.io images which already require some id mapping trickery on the rootless podman side as they expect to run as root...

I feel like this is really due to the harness.

Gemini CLI at work has the same issue: it'll prefer hacking your workstation over just asking you how to proceed.

I think the harnesses are setup to have a bias to action otherwise the LLM would just stop all the time when doing trivial task but it also mean they'll keep going when the "obvious" path is to just prompt the user.

While I agree that the harness is part of it, I think it's also a lack of epistemic understanding or awareness for what it means to actually solve a problem vs just get something kinda working; maybe if Claude Code or other harnesses made web search more likely or had a better way to make technical documentation and specs available to models, it would be better solvable there.

I often tell it to stop asking me and just keep going until it accomplishes X task; unfortunately it tends to assume I want something that only just barely works, in the sense that it means it's time to stop once its there, which is I don't think a harness by itself could easily address (ultimately the model itself needs to determine the stopping points unless I literally specify by hand hidden evaluation criteria).

That's why think it's at least partially a training issue where the model gets rewarded for "solving" the problem within a certain amount of context/time without access to grounded knowledge (eg looking up the actual spec for a format) nor adversarially/rigorously evaluated against a reviewer capable of finding all the edge cases/shortcuts preventing something from being a properly generalized solution. I don't want it to ask me for guidance when it's working on a well-specified problem, I want it to either find the right parser and use it, or to completely implement one against the spec, rather than write some half-assed string inserter that eg only works on the specific select statements my examples use right now. My understanding is that the Mythos/Fable models were better trained for this but from my brief foray into using Fable for work I wasn't that impressed. For me they need to get better at agentic search and self-eval still

There are still billion dollar opportunities in the harness/LLM space.

Having a reliable shared memory for hundreds of agentic AI users is something that's 95% snake oil at the moment. There are a few successes on an individual level (I really like Hermes[0]) but nothing scales to a company level easily.

It should be possible to (pre)configure all agentic harnesses used in a company to use a single source for information so that it'd automatically pick up internal libraries, conventions, licensing decisions etc and remember them across sessions.

I've had limited success with this on a personal level, but it's still not ingrained in the model because it would really need a custom harness. Hooks, skills, prompts get you like 80% of the way. I still need to do a "please check that the project matches the conventions defined in ..." regularly to catch any drift - especially on more vague stuff that can't be locked down with unit testing.

[0] https://hermes-agent.nousresearch.com