LINUX Unplugged - 686: Stop Desktop Slop

Episode Date: September 28, 2026

KDE’s fight over AI generated code explodes at exactly the moment Plasma is making some of its biggest changes in years.Sponsored By:Jupiter Party Annual Membership: Put your support on automatic with our annual plan, and get one month of membership for free!Managed Nebula: Meet Managed Nebula from Defined Networking. A decentralized VPN built on the open-source Nebula platform that we love.Support LINUX UnpluggedLinks:Web Boost — Send us a boost via sats or USD💥 Gets Sats Quick and Easy with Strike📻 LINUX Unplugged on Fountain.FMJupiter Garage SWAGGoing All-In on a Wayland FutureKDE/kup: Backup scheduler for the Plasma desktopSovereign Tech Fund Invests Over €1 Million in KDE Software DevelopmentKDE and AI, and you, and me – Adventures in Linux and KDEWhat would it take? A lovable, sovereign, AI-native KDE - YouTubeWhat Even is Ocean?Akseli Lahtinen on Union for QtWidgetsStyleKit talk at Akademy 2026Framework Laptop 12KDE Linux at 2: what we've learned and how it can help youQA infrastructure in KDEKDE at 30: What Are the Colored Blobs Telling Us?KDE: The Documentary TRAILER — The Story of Free Software Underdogs — Full Film Coming October 14thWhat's Happening in KDE Remote Desktop — Improved Unattended Mode and MoreThis Week in Plasma: Bi-Directional RDP Clipboard SyncThis Week in Plasma: So Many Ways to Click and ScrollThis Quarter in KDE Digital Sovereignty: Q2 2026KDE for Enterprise Needs a Strong PIM InfrastructureFramework's Fedora KDE Plasma Laptop 12 Guidesystemd 262A review canary in systemd's contributor instructionsLinux Live Update documentationLUO userspace APImemfd_create(2)FluxCast — Miracast/WFD ecosystem platform with .deb and AUR packages.fluxcast-flakeJB Gulf Coast Meetup RoomNixarr — Media Server Nixos ModuleNixflix — Nixflix is a declarative media server configuration manager for NixOSgarage — S3-compatible object store for small self-hosted geo-distributed deployments.Pick: Cliparr — Cliparr is a streamlined media clipper that allows you to quickly create and download clips from the media currently playing on your Plex or Jellyfin server, or from a local video file opened directly in your browser.Cliparr: Plex and Jellyfin providersCliparr: local file or direct media URLCliparr: export settingsCliparr: subtitlesMediabunny — A complete JavaScript media toolkit for the browserKyleTryon/gifenc: fast GIF encodingPick: fdroidcl — A command-line F-Droid desktop clientspace-elevator — Space Elevator gets you to orbit. What you do up there is your business.

You know how it goes. Someone on your team figures out a process. But six months later, a new hire asked, how do you actually do that? If that sounds familiar, check out scribe.how slash unplugged. That's exactly the kind of problem today's sponsor, Scribe, is built to solve. Scribe is a specialized intelligence platform trusted by 94% of Fortune 500 that helps teams capture how work actually gets done. I actually tried this on our own infrastructure. I logged into one of our ProxMox hosts. I took a VM snapshot.

It's just exactly the kind of operational process that's easy to do yourself, but harder to hand off to somebody else. Scribe capture the steps as I worked, turning that individual know-how into a guide my whole team can follow. It can also automatically redact sensitive information. Plus, once a guide is reviewed, anyone can launch real-time on-screen guidance, showing them exactly where to click inside the actual tool. And that context can power your AI agents, too.

Most AI was trained on the Internet, not your company. So Scribe helps agents operate on how work actually happens, not guesses. That's the real win for teams. Scribe takes the know-how that usually lives in one person's head and turns it into a guide everyone can actually use without making anyone stop and write it all down. Learn more at Scribe.how slash unplugged and mention Linux unplugged for your first month of Scribe capture free on select plans. That's S-C-R-I-B-E dot how slash unplugged. plugged. Hello, friends, and welcome back to your weekly Linux talk show.

My name is Chris. My name is Wes. And my name is Brent. Hello, gentlemen. Coming up on the show today, the KDE AI debate has blown up, just as the project was trying to draw the line between useful assistance and low effort slop. And as KDE turns 30, we'll look at the next chapter for plasma, and it's big. Plus, why we think the next system he could be worth paying attention to.

Then we're going to round out the show. with some great boosts, some picks. And I cannot wait to tell you about this pick. I have wanted something like this for years. And then, of course, we've got a lot more in theirs. And to help us get to all of that. We've got to bring in our virtual lug. Hello, mumble room.

Hello. Hey, howdy. Hello. Hello. Thank you for joining us. And hello up there in the quiet listening and everybody watching live. And hello to our friends over at Defined Networking.

Go to Defined.net slash unplugged and meet Manage Nebula from Defined Networking. This decentralized VPN is the one we use. It's the one we love because it's built on top of the Nebula platform. And I think it's perfect because I've used it between just two machines and between lots of machines. It was built to connect Slack's global infrastructure across the world. And it is very, very fast, resilient. And I love that you can self-host the lighthouses, which mean you manage the network.

You can manage the authentication, no big tech sign on, strong encryption, no single point of failure. That's extremely robust. And now, I don't know if we've mentioned this, but Define Networking's API makes it easy to give your Nebula host-friendly DNS names automatically, and it keeps those record syncs as your network changes. Pay it via API.

I think that's something folks have been wanting. Use the API. It's the kind of infrastructure you can actually own and then build on top of long term. Try it out 100 hosts for free. No credit card required. You go to Defined.comnet slash unplugged. that's right define.net slash unplugged so it's kind of a kDE week on the show and uh you know both

you boys are very very happy plasma users and i've used kde for ages and ages and ages got to run around here and i was reflecting early on you know putting my my wife then angela at the time on like i think version two it might have i don't even know if we were still calling it the k desktop or the cool desktop environment and it was rough but it showed so much potential and so this is a really great week because we're nearing 30 years of KDE and plasma 6.8 is really looking sharp and it's kind of their flag saying we are an all whalen desktop as of the 6.8 release. Yeah, I mean, you mentioned like almost 30 years. That means almost 30 years of running on X-11. Wow. So kind of a consequential change, I would think.

Yeah. They will keep supporting Plasma 6 with bug fixes and stuff into early 2027. So there's at least some support. And I'm sure, you know, some other downstream LTSs will take that even further. Yeah, I'll tell you who's going to use that. People that need remote desktop. Right. I mean, really, if you need remote desktop, there's options with Wayland for sure.

And a single desktop probably going to work okay. But if you need advanced stuff, you need multi-desop, multi-screen, things like that, I don't even think they're trying to ship that in 6-8, right? That's just going to be a down-the-road thing. I think it's recognized. It won't be ready for that. Okay. And I think the project understands as well, you know, that it is kind of critical for X-11 holdouts.

for many folks. And they have made progress, though, right? Bidirectional clipboard sync, adaptive RDP quality, lower latency, unattended session improvements. Exclusive mode. Exclusive mode. That's pretty sweet. Locks the physical display

and gives the remote user a virtual display. They do have quite sophisticated virtual support now, which is pretty great. So I think there's a lot cooking. It's just, you know, reliable, polished, rock-solid remote desktop is a pretty non-trivial ask. So there's a lot to get right. It's doable, though.

Oh, yeah. And they'll work on it. They'll work on it. But 6-8's going to be a really solid release. If you are on a stack that Waylon works well with, you're going to love it, I think. I mean, and there's so many nice things. Like they're working on Cup, including Cup, which is a backup tool that I would like to have Brent look at, actually. See what Brent thinks of.

A little Brent special report. I like that idea. Okay, I'm adding that to my list here. You want to be our backup specialist? When do we want to do this? Our chief backup line that you'd like to consider. No, maybe when 6.8 comes out or something, because it comes with 6.8. You know, I think this is, some of this, right, we saw them taking a look at like Butterfest snapshot display in Dolphin and stuff.

I think as they're working on KDE Linux, right, and they're kind of working on towards a KD Linux 2 sort of situation. They're really thinking holistically about that whole experience. And even if maybe you use a different distro's lower part of the plumbing, the fact that KDE is thinking about that means there's going to be affordances at the KDE level. a level to enable that kind of thing. And I think that's going to pay a lot of dividends, like things like backups. Yeah. Because then maybe you need more explicit hooks in KDE to support that,

and then other things can use. It feels like that's one of the advantages of being around for 30 years is you can, you know, look at the peripheries of your main offering and really make those smooth. I think the long-term future is bright for them, too, because they got a million euros from the Sovereign Tech Fund recently. And if you look at how they're spending the money, it's around boring stuff, like building and improving the automated QA infrastructure,

and backup and restore features, and factory reset features for the desktop, and improving security infrastructure, and maybe the ability to work with things like Network FileShare is a little easier. It feels like money is kind of going. It's not a ton, but 1.285 million euros is notable, right? That's going to pay for some developer time. like it's being prepared for wider deployment in public sector use, I think. And, I mean, it seems like a pretty good fit for that use case, right?

Like, it's got a familiar way to interface with the computer. I think a lot of people could be able to interact with it. And as we know, at this point, it's both very flexible and configurable and pretty rock-solid out of the gate. A lot of, you know, honestly, web-dav is a pretty native, you know, intranet workflow. They've got great web-dav support in there. They've got a pretty good personal information management. suite in there for managing your calendar, your contacts, which you can load in there. Yeah.

And like you say, you could make it look a lot like Windows. In fact, one of my favorite layouts is the Windows 2000 kind of retro-looking plasma. I just think it looks fantastic. And you could go to a corporate workstation and replace an old Windows box with that, and they'd be pretty happy. Yeah, I think you're right about that. It is, speaking of workstations, right, it sounds like some of the configuration management work that's being put in here is actually explicitly aimed at centralized policy and, like,

fleet, yeah, fleet stuff. this is what I'm reading. Like, this is what I'm reading like, okay, we're building the, we're improving the QA process, we're improving the security infrastructure, we're building things like factory reset and backup. And we're also looking at tools that make it better for fleet administration. This all reads to me as signals that it's being prepared for improved public sector use on the desktop. That's a serious project. And that is the kind of thing that once done could go on for 20, 30 years. You know?

You know, once, I mean, especially because they're not paying licensing fees in the trust. additional sense, hopefully they'll kick back. But so once a, if a government, like a large government were to really base a workflow around plasma, why wouldn't that stick for 30 years? I mean, you could see if it took off too, right? You kind of get a shared base that everyone can contribute to, whether you need to just add something for yourself or then it ends up being widely adopted. In the bootleg earlier today, earlier, previously in the stream, we were talking about the

Natalens building a Nix-based Linux desktop. and that same sort of thing where that gives them the ability to deploy a system where the vendor doesn't force them to roll to a new version at a particular point and the desktop environment can be managed and it really is a lot of what people liked about Windows and group policy without the licensing or the forced oscillate, what am I trying to say, forced obsolescence. And like as much or control as you'd like in terms of building from your own caches internally with only approved software. Right. And it's been verified and secured. And they can move more at the speed of government. And you know what runs great on NixOS? KD. Plasma. Yeah. Do you think they're using flakes? Right into the show. Let us know. The Netherlands are. So, I mean, interesting.

Anyways, it's really great to see that. And to see that kind of, you know, I'd love to see them get even more than a million euros, you know. But let's keep that going. There has been a bit of a dust up that I think kind of has been boiling probably since about April, I think, if my timeline is right on this. Certainly not just right now. It's been a longer discussion. But it kind of feels like this last week, it really kind of flared up to the next level where it's worth talking about now on the show. And Nate posted a blog post called KDE and AI, you and me, which probably has the most succinct and authoritative summary of this. So we'll link to that in the show notes. But the short version is, is KDE as a group, has been debating how to handle LLM-Assisted,

contributions, especially the bad ones, low effort, stuff that people are just trying and they're throwing in. And the merge request and all of that to process that to figure out which ones are real, merge those, which ones are crap, reject those. It's a lot of just crappy, grindy work for the maintainers and it's only growing. Yeah, and this is something we've seen, you know, a wide array of open source projects having to deal with at this point. Yeah.

And Nate Graham proposed guidelines that were not an AI ban. but to essentially say, use the tool you want, but you must understand the results, and you must be able to verify it, and you just can't dump the work on the maintainers that you can't defend yourself. And that even just sort of being open to the idea of AI is an agitator to some in the community.

But then almost at the same time, Academy was going on, and they featured a talk proposing a really gentle, more kind of be open to the idea of AI agents working with a plasma desktop. And they kind of had the outside look of like the rest of the world's moving very aggressively here. We're going to have new users show up that might expect this functionality. We maybe should have an answer for them. And I'll link to this video.

The video is called What Would It Take, a lovable sovereign AI native KDE. You kind of try to imagine what that might look like. And they were very gentle. I mean, they were very, very gentle in this video. Everything was a maybe we should consider, we could. And they were very focused on protecting data, protecting privacy. and local open AI. That was really as controversial as they got in that talk.

But with Nate's proposal and this kind of seemingly more open to AI and KDE talk at Academy, those two things collided online and a policy that was maybe a decent idea became an inflammatory topic. Is that a fair summary? Yeah, I think so. Also, I just want to point out too in that talk, they did explicitly say, you know, obviously the desktop needs to keep working without any AI or LLM tooling, right? That still needs to be first class. Yeah.

No, it was more about like we've got debuts. We've already got all these communication protocols. We just need a little, essentially what they were really proposing was safety barriers and restrictions on what is capable now. Like, there's a lot more access if you point an LLM agent at that they have that they were saying, we should probably wall some of this off is essentially what they were saying. Well, and that's the benefit you get when you have intentional interface. So like I've been trying out the K-WIN MCB, but like you might have to try to get one thing that you want. You might have to get more information than you need. Or like it has to use some private internal APIs that should probably be, you know, explicitly considered and made intentional public APIs.

So that turned into kind of, you know, an online firestorm of should KD permit AI and LM-assisted contributions at all. A group of people standing up a website that put together a lengthy, you know, positioning KD and plasma against AI as a community. Right. It became more, not just a discussion about what, you know, like the concrete thing of how to help maintainers deal with this. It became sort of a, as often these situations do, a comment on the total judgment of what AI is and if you should use it and a much larger conversation. And then it involved like digging up social media posts of people to try to discredit people in the arguments. I mean, it got kind of ugly. And the end result, of course, if you look practically what happened, is a proposal that was intended to give maintainers actually more. power to reject AI slop was then portrayed as opening plasma and KD to the door, the door to AI, and so they just rejected all of it. So now there's nothing. Now there's no policy. And we're right back to where we were when we began with this whole thing. And Nate posts in his September 23rd post pointed out that a couple of the people that joined the big fight around AI weren't

known to be KD contributors. And really they shifted the fight to the morality. of AI generally rather than the contribution roles. And that's sort of when it went off the rails. Katie had to lock out participation to some folks. I mean, it got bad. Yeah, I think that particular ticket thread thing is even now closed and gone at this point. But, of course, it spilled out to the wider social media

and various places where people talk about Linux. So there's a couple of questions in here. And the first one, I think this is the quickest one that I just like is, to me, Brent, it seems like the obvious unspoken position of the anti-AI crowd seems to have a quiet assumption that if they can hold out and if they can ban AI and AI-assisted contributions, it will eventually just go away. Do you agree that there's that kind of baked an assumption to this, or am I misreading this? I think that's the intention.

I don't know if everyone who's pushing back thinks that's a reality long-term, but maybe it's like, hey, let's put the brakes on now before we can figure out what this is actually going to mean for the future. But the problem I have with that is it's almost forcing projects to take an all or nothing approach. Like either you use a tiny bit of AI to help you manage tickets and stuff like that or you don't integrate into the project at all. And that doesn't seem nuanced enough. Like we can do both if we're thoughtful about it. which if you read some of Nate's original suggestion and proposal on how to do this, is very reasonable.

And we've seen many projects recently, including the Linux kernel itself, do very similar proposals. And yet this one, for some reason, caught certain people's attention and got very fiery. Okay, Wes, my question for you, this is just sort of the easy ones, but I just wanted to ask you. like there's another assumption, I think, that might be baked into the anti-LM crowd here that is tenuous at best today and not going to work long term at all, in my opinion. And that is that it all hinges on the ability to determine the difference between a human-generated contribution and a machine-generated contribution. And it sort of hinges on this assumption that's going to be detectable and that the machine-generated code is going to be worse.

Yeah, and that you should try to do that detection, that that detection has value. you, right? That, oh, you're blown my mind. Okay. Right. Like that, yeah, there's not only can you detect it, which is not always true. Right. And it won't be harder and harder.

Yeah, for sure. And then there's just like, does that add something? And that's where I thought this, the original proposal was so interesting because it, it didn't focus on sort of the provenance of the code. It focused, like, it was a kind of, it was kind of just saying to maintainers, you have permission to say no, right? And people and the people that contribute need to be responsible for the code. Yes.

It didn't change the standards of the project. The code review standards remain the code review standards, right? And that's kind of the point is if either it meets those standards, kind of what Linux is doing, or it doesn't, and you're responsible for it. And this is where the outside pressure kind of sucks in both sides, right? We have put folks against it, but it kind of means anything you do can feel like you're taking a side as well. So it's not, you might want to, as we've talked about on the show, right, there provides an all lot of opportunity for folks who haven't felt like they could contribute to these projects to contribute. But that also comes with the responsibility of knowing how to do so responsibly. And I think that you can also feel pressure as a maintainer if you say no to something or you know, you need to reshape this and then you're not getting responses that now you're coming out against AI, even if you're just trying to maintain neutral standards of the project.

Those I think are arguments against the anti-LM crowd. Where I think maybe they do have an argument is the inundation of stuff that just burns up maintainer time. And that is a legitimate problem because the cost to produce those crap pull requests or whatever is just going to be cheaper and cheaper. And it's going to be more and more people that have access to it. And it's going to be everything from some crap chat bot to like an actually advanced, you know, cogeneration harness. And you used to have a useful signal, right? Like it took a fair amount of work to learn how to contribute to a project like that and go through the motions with Git and to shape your thing and figure out how to open the merge request. and now the cost has just gone way, way down.

And so it used to be just that the fact that it existed was at least reasonably high signal, even if you ultimately didn't accept it. But you combine that with just the, you know, huge increase in the number of them too. But I think, you know, you got to really realize that it's the state of the system now, even if anthropic and open AI went away tomorrow, you know, right.

But even if they did, like, you know, they just cratered into themselves. the open source models are good now, and there's so many of them, and they're releasing so frequently, and it's so competitive, it's actually to the point of you can't keep track of it. You can kind of watch some of it,

but it's just, the model releases are so fast and furious right now in the open source world, and a lot of them. I mean, you're primarily using open source models. Yeah, mostly. And you're generating very usable functional results. Yeah, I mean, I do occasionally use proprietary models,

to do some review or, you know, whatever. I use a variety of models just because that seems like it's useful. But, yeah, you can get all kinds of productive work done these days. The LLM assisted coding, though, it's just probably not going away. And I know I've made this analogy before, and I'm not the first, but, Brent, I just, I love talking about this with you since you have a photography background. Oh, yeah. Like, there was a period when you could, as, you know, photographer, maybe plausibly believe that digital photography was just bad photography. The first digital camera sucked.

And the people that were most likely to use them were amateurs. they didn't know how to frame a photo, they didn't know all the secret locations to go get a great photo. They didn't know the rules to compose you in a great portrait or anything like that. And the camera quality sucked. So like it was easy to dismiss and professionals wouldn't even touch it, right? But of course that changed over time. It was super interesting for me to be a professional photographer, or at least at the time when digital was becoming usable, let's say. Because I've used digital cameras where you stick a floppy drive into it.

Yeah. And you get like 0.2 megapixels or something. Those were fun. Yeah, yeah, yeah. You can fit like 10 pictures on the floppy disk and that's how low quality they were. Yeah. Because the floppy is 1.4 megabytes. It was the easiest way to get a photo into your computer though, right?

And it does remind me of AI quite a bit too. So I agree with your analogy because the technology has actually been around for a very long time. It's just the difference between it being just widely available. and easy enough to use that someone who's not an AI researcher can use these things. So the really interesting thing about photography that I think applies completely here was when I lived through the transition between film and digital, there were some photographers who were early adopters, which we can see in this realm as well. And those photographers tended to be the ones then that when the technology,

was good enough, which it was moving fast. So that was like six months later. They were way ahead of the people who were pushing back on just keeping, you know, we've been doing film for 100 years. Why should we give it up? And these days, film is like just a hobbyist's activity, which is awesome. But if you're doing professional stuff, not digitally, you're, that's not the way. It's just the irony there that, you know, you could probably describe digital started as a hobbyist. thing. And now imagine a professional shop that just banned digital photography. That shop is either gone and dead now or it's like some top tier specialty that only does glass and, you know, analog.

Yeah, you nailed it. Exactly. It's like the two routes. And typically they end up scanning the stuff into your computer anyways. And it does have a different look and feel, but, you know, as far as productivity goes and accessibility and stuff, it turned around completely.

Digital is far, far easier than the film stuff now. And the AI-assisted code's going even faster than the digital photography did. And so a ban that is presumpted on, we're getting too much slop reports, you know, AI-generated code is bad, is just not going to work forever. It's not sustainable.

Yeah, and it maybe wasn't the problem anyone asked for, right? That is a big thing. We have this problem now, and I think it does behoove us to try and figure out how to figure out how to teach people how to use these tools to get code that is up to the standard and can be, you know, collaborated on and can actually help projects. And then, of course, the social etiquette that goes with that too, right? Because at the end of the day, for the most part, we're not yet on, you know, crazy bot swarms opening these themselves. It's mostly human-driven

folks using these tooling. And so there's the human layer of like what you have a role in what you're submitting, and if you submit stuff that is just spam, it's kind of just spam. It is, it is fair to underscore, though. Like, we've never had a shortage of patches. That's not really, I mean, maybe some projects, but big projects don't really have a shortage of patches. They have a shortage of maintainer time and attention. And that's been the problem we've had for a long time, and this is making that

worse. Because it can take a lot to get a patch from, like, technically solves the problem to actually fits in with the rest of the code base and the roadmap and, you know, the wider array of factors that might go into it. But here's a simple little logical scenario why banning LM assisted code doesn't make sense because if a human writes

a terrible 3,000 line patch manually, they would reject it because it's terrible and it's 3,000 lines of code. And if somebody uses an LM to generate a beautiful 300 line patch and they understand every line that was written and they have tests for it and they can maintain it,

they would reject that too just because it was LM generated, even though is substantially better than the other patch. Yeah. And that doesn't make sense. Yeah, that's where it feels like it's the wrong test. To me, I see the difference as like using AI as your, like a tool in your toolbox to generate the code versus having AI be the author.

I think there's quite a difference in that. And what all projects are looking for is the tool use one, not the authorship, because then someone has some skin in the game, I guess, and can maybe stick around and, you can maybe stick around and help maintain the thing that they just dropped? That would be a concept. You know, and I know this isn't how they think. They're not corporations. They don't look at, well, it's funny because they like to look at the market. They like to try to build products to draw new users. But then if I ever suggest that they should think competitively for a moment or something like that, then we're open source projects. But just from a differentiating

standpoint in the marketplace. The Ghanome folks are going hardcore anti-LM from the whole ecosystem, right? And so, you know, if you want to make a moral stand, Gnome is going to be the project for you. If KDE and plasma make that same stand, you just removed a differentiator in the marketplace and one that could be significant. And we note that they got one million dollars or one million euros in funding. Well, that is great. OMA Archie is over $18 million in a couple of months, right?

I think it's beyond $18 million. Yeah, 20 maybe now? Yeah. And the reason for that is, A, because DHH has contacts because he's the creator of Ruby on Rails, and some of these donors have built their business on top of Ruby on Rails, and they owe David a solid. But I think the other obvious accelerator has been the moment they positioned it as an agentic-friendly desktop, they started getting millions of dollars of funding. Now, I'm not saying KDE should play games just for that kind of stuff.

I'm pointing out that there could be advantages to differentiating in the marketplace depending on where the funding and where the interest is and where corporations want to go and all of this. And being a desktop that is at least amenable to that workflow, doesn't necessarily have to promote it, doesn't necessarily have to encourage it, but just isn't hostile to it, is going to open it up to a new type of user that is coming in droves. The reason why people cannot wrap their heads around it because it's just dot files and themes, they say, Archi is saying if this is the workflow you want to embrace, we're the platform for you. And the result is they're getting $20 million in funding and a lot of attention. Plasma could do something that is probably more approachable and amenable to the existing community

that fits more within the existing values of the community that still taps into some of that interest and excitement that's in the broader marketplace without having to align themselves, without having to do any of the branding, without having to do any of that stuff. but just by making plasma a better and safer place for those types of workflows. And it starts with figuring it out here with the submission process. Well, and the proposal from some folks was KDE for people. I don't think we have to accept that being amenable to that, having composable APIs that AI agents can use as well as humans,

the only way to do that is to make something that isn't also a first-class thing for people. I just don't think that's true. Well, who's using the tools? People. Right. I think this is where when we look at what holds Linux back, and we can't really wrap our head around why it doesn't get broader market adoption, because you cannot see a market signal when it's smacking you in the face. And that's just always been the case. And we've never had until companies came around, like System 76, we didn't have anybody that was paying attention to a particular market signal and trying to build for that. We had this.

We're trying to draw new users or these broad things, these broad, broad strokes. And when we do see the signals, like $20 million worth of signal, we still don't know how as a community to properly interpolate that signal and act on it. But I think it's just obvious that long term, if plasma goes anti-LLM and Ghanom goes anti-LLM, then desktop projects like Hyperland and others are going to be what fits that market need and not plasma. And it could be, and this is where I think, this is where I say the Linux community hold itself back, it could be that new demographic is 10 to 20 times larger than the existing demographic. It also is their problems are solved by Chromebooks and tablets. Yeah, you know, it's like with free software, we didn't just accept that, like, we weren't going to do stuff the proprietary software did. We figured out how we could do it with free and open source software.

And it seems like the challenge right now is we got to figure out, how do we take advantage of the stuff? How do we access funding and sources and the next wave of things without compromising on the values that we hold dear? We haven't figured it out, but we need to. Because we're stuck on this conversation of all or nothing. We're stuck on this binary thing where there's these sort of baked in assumptions that you're going to be able to tell assisted code from human code in the future. And we're working on these assumptions that maybe it'll go away if we just ban it long enough. and none of these assumptions

are going to bear out to be true in the long run and so we're stuck in these false arguments instead of like what I think the plasma and KD

folks were trying to do was open up a discussion of how to do this the right way and what got it got side railed by some folks that aren't even

involved with the project that want to fight about the morality of it which again is fine except for then you're also

saying we don't want new people coming to Linux. Is that your position? Because that is what you are saying. It also sort of forces a maximalist take on the morality that sort of rejects the idea that you could ever have a system that could be aligned with these values. Right.

You can't have system built for people. And forstall is the future of trying to do that. And I just, again, I want to underscore, like, if you're anti-Linix desktop growth, then what are we trying to do here with free software? And what was the whole purpose of making something? that was making the information free to the world. What are we trying to accomplish here? Are we just trying to create a clubhouse

that's just big enough for us to hang out in that we can pretend like the rest of the world doesn't exist? And I think what we have, and then I'll end on this, is we have a lot of people out there that have made this their hobby for the last 20 years. And this is what they've got. They don't have something, they don't have another skill set,

they don't really have much else. And they've devoted their lives to maintaining some of these projects. It's their hobby. It becomes for some of them their job. It is also for some of them, it's what ails them because they are so dedicated.

And when something comes along, like the change we saw in digital photography of that kind of scale that tells them your hobby is changing and there's new people coming in that don't take your hobby as seriously as you do or approach your hobby the way you do, we have this group of people that are, I mean, they're almost, it's almost a mass formation of psychosis that's happening.

And then they're being supported and tricked by rich tech CEOs that are convincing them that their products are so powerful, they're going to take us all out in four to ten years. And it just sort of plays into this sort of chamber that it just sort of gets louder and louder. It's a doom loop. And they're, and they're being manipulated by the very people that they claim to be. against, but yet they're being influenced by them. And so we don't get to the practical conversation of how do we safely integrate this in a way that actually meets our values and lets people come over to Linux and get right to work. And see that they can affect change in this software, that it's responsive to them, that it can meet their needs. I mean, okay, okay, all right.

The reason why $20 million matters is I invite the listener to go out there and add up all the money that Linux Mint has ever raised or that the Ghanom desktop has ever raised or that the Plasma and KD.E. team has ever raised. And then ask yourself why they have raised that more of money than they ever have in three months. There's a real interest, a market interest in a agentic capable operating system. And I don't know, you know, what makes Oma Archie agentic? They, you know, they build in some tools by default and they have some skills that make it so the tools that they build them by default know how to manage the operating system.

I mean, I think that's, that's just really. And so we don't have like, there aren't people thinking about how to isolate debus to make it private. And there's like, that's not happening. And they're sucking up market share. And I think about these projects like KDE that have been around for 30 years and they have a hole in the boat. And they're not able to plug the hole or perhaps even grow because their community can't get past an argument that isn't really valid anymore. And the problems we need to solve are how to handle all of these contributions. and how to make these workflows work on the plasma desktop

that isn't going to expose people's information. And we're not there. We're just not there yet. And the commercial desktops and projects like Omaha Archie are just going to lap them. And they're going to continue to lose users to that stuff. So that's the latest on the AI debate. But we do have good news that I'd like to cover,

including perhaps, although it's not committed as such yet, but perhaps a replacement to the Breeze theme in the works. Yeah, so there could be, I know, there could be some big changes coming to the Plasma desktop. Well, Chris, you're forcing it. They're going to change my beloved theme. It looks like KDE is suggesting Ocean, a new KDE visual style and design system being developed

because Breeze never really had a complete modern design system representation for designers to work from. So that work is including ocean icons, ocean plasma style, an ocean application style, an ocean color theme, and even an ocean font. Breeze has been around for so long. Some of the stuff in plasma didn't exist when Breeze was created. I mean, I love Breeze. I'm just saying. It's been around.

Yeah. I can't remember a time that it wasn't around. And since Plasma 5? Yeah. Yeah. And it's been good. Yeah. Like, I have no complaints.

Do you? No. No, and there's lots of community options, but Ocean, you know, Ocean's going to be like a complete look. It's, they say, we took up the idea of creating a design system for plasma that accounts for most, if not all of the necessary graphical building blocks that developers could use that preserve consistency between graphics and code and deliver a cohesive experience for users. You can find some early designs online, and they do look kind of good. I think the real interesting part is that it's being built on something called Union. Yes. So it's like a concrete styling effort, reaching plasma cute quick and cute widgets in 6-8. And the idea is you can have one editable style source.

Excelli Latinan calls it the first shared editable source of truth for Breeze, cute widgets, and cute quick styles, excluding color schemes. Okay, so some work in progress. Hence the name union. Yeah, exactly. Because the idea, like before, you could not have a single source of truth for both of those things. and they would totally diverge and you had to actively maintain keeping them in sync. And under the hood, it took roughly 9,000 lines of handwritten code against a poorly documented Q-style API. So you had to do a bunch of reverse engineering to even make this effort possible.

I mean, so like when you have to build a system, you know it's going to be kind of a big visual change. Yeah, it's not just a new theme. It's like a new system for how to do, you know, consistent theming across the desktop. Yeah. Oh, this is going to be great, boys. This is going to be great. I'm really, you know, that's the kind of stuff I love seeing them work away. We have some more information in the show notes, including, I think, a talk that got into it.

Yep, there's a talk there. And, well, yeah, links in the show notes. Yeah. Okay, this next story is sort of plasma adjacent, but seems like great news for them. Frameworks' new 2026 12-inch laptop can now be pre-built with Fedora KDE plasma along other configs. No, isn't that nice? That is so cool. I mean, I think that's just how I would order it now.

So it's not just like KD's supported, but it's like an actual factory operating system choice. They must be like doing some Q&A on that. I assume so. That is so great. I mean, I'm sure it's, you know, it's the DIY version, but it's a nice, it's like it's a milestone. Another OEM that's approachable. And it, I mean, that says a lot about the entire software stack involved, I think, right? Mm-hmm, mm-hmm, mm-hmm.

KD Linux has also been building a recovery more into the OS. We talked about with the funding standpoint. And one of those could be an image rollback for bad updates and a factory reset that I mentioned as well when a system needs a clean start. Yeah, I liked the pitch here. Nate kind of said, a failed update should produce, oh, this broke, not my laptop. And I have a big exam tomorrow.

Like you say, when they're building KD. Linux, which is their OS, they are thinking more holistically about all this stuff. Yeah, right. I mean, like with the factory reset thing, it doesn't have to be a big deal. It can just be something you've designed in, so it's super smooth and reliable, and the data's never at risk.

The investment in QA recently, too, they said they're catching failures across the whole desktop. Yeah, a full system integration test, which is one of the best kinds of tests, caught plasma crashing at startup with a panel missing. And that's the kind of thing, right, that can really suck if you've shipped something out the door and that doesn't happen,

but it can be an expensive test to, like, really make smoothly and not be flaky, because you probably got to spin up VMs and have something they can tell when a panel is missing. And yeah, but it pays a lot of dividends. And that's the kind of boring detail, you know, takes time to get right work

that funding really helps with. You mentioned you've been playing with the plasma, or is it a plasma MCP or K-WIN MCP? K-WIN MCP. Could that be used for that type of debugging? Could you use it to like take screenshots of the desktop and see if panels aren't rendering? Yeah.

Yeah, you definitely could. What else can the MCP, does it like read open windows? Can it? Yeah. Yeah, and it'll look at the accessibility information. Yeah, it's got a fair amount of stuff.

It also has a lot of support for virtual. Like, you can have a whole virtual session if you want for your agent to use that doesn't mess with yours. That is, I think, that's sweet. Yeah, that's, I think, one of plasma's structural advantages right there. Yeah, the project even says,

K1MCP provides everything needed for end-to-end, GUI testing, and desktop automation. So you can have it spin up, like, is it a separate activity? Like, how is it at the, do you know what it is at the plasma level, what it's doing to get to this virtual desktop that doesn't interfere with your desktop? No, you know, I've actually haven't tried switching into it.

I wonder if it's just activities. Yeah, it could be. I don't know. But that is so cool. That is really neat. It's a bright future there. If, you know, using that kind of stuff to debug and, especially when you're porting an application to Linux,

it's got to speed up the process. And then, which we mentioned at the start, but I think we should just spend a moment on it. Katie as a project is hitting 30 years old. And Kevin Ottenz analyzed roughly 30 years of KD commit history and found that the community peaked around 180 active contributors and then fell to roughly 110 around 2017 and has climbed back to 140 today. But the amount that they're contributing, though, I think is interesting. Katie is now producing roughly the same level of commit activity as its 2010 peak with fewer people. And 2023 contained the highest commit count week in the dataset.

Sounds like things are moving to a pretty good clip. And then this is interesting. I thought maybe we play a little bit of this, even though it's going to get us in trouble on YouTube. But screw them. Screw them. It's got some of our favorite people in KDE in this trailer.

KDE, the documentary trailer, the story of free software underdogs, and it's a full-length film that's coming October 14th. We'll put a link to the trailer, but we'll play a little bit of it for you, too. So you can get a sense of it. I always describe myself as a relative newcomer to KDE. I've only been here 10 years.

Using Linux on the desktop was tedious. Microsoft Software at that time was really, really bad. I mean, it was normal that a computer would regularly crash. I saw a post on Usenet. There was this guy who had posted a call to arms to create a desktop environment. The initial vision of KDE was very much driven by How can we make Linux successful and more appealing to the general public? And the cool story is that it's been a bunch of self-organizing people who saw that vision.

And for one reason or another, that spoke to them. And they got involved and they made it happen. So today I learned that Aaron Seigo has a badass fish tank. That's pretty cool. Yeah. So the full film, like you said, October 14th from the, I think it's cult repo is the YouTube channel that we've been publishing that. I'm definitely going to watch that. Of course.

That is really great. I mean, 30 years is a huge deal. We got a whole bunch of links in the show notes with different information, more details on the remote desktop plans and where it's at right now. And what you can get, if you need to use it, it is working to some degree and what you can expect. Maybe you want to try it and report back to us. I'd love to know. I also'd like to know your thoughts on the LM-assisted contributions to projects like plasma and Ganyl. And if you agree with me that if Gnome goes one way, would it maybe be a good idea for plasma to go another way just from like a, you know, differentiator standpoint?

Boost.jupiterbroadcasting.com. Let us know your thoughts. But friends, System D has a new release that we thought was worth your attention. Well, Wes, in particular, has been getting excited about System D 262. Yes, indeed.

Every now and then one of these catches your attention, Wes. And I see at the very top of the list, which is, I'm surprised, because I know there's K-Exact stuff in here, but the very top of the list, you have smaller container builds. Yeah, they're just doing some stuff to support, like, really lean,

uses of SystemD now. So the PID1 part of SystemD now embeds 21 of the system unit files that it sort of needs to properly function. So even if you can't load those core units from disk, system B can just fall back to its built-in copy.

Now, anything that is installed on disk will always take precedence. It's really just meant as a default. But includes stuff like reboot, shutdown, and K-Exec, and the usual boot targets. And so the intention is just like that single PID-1 has everything that it really needs. in a rescue or a super small embedded or, you know, just whatever kind of situation where you might not have access to the disc you expect.

Let me ask you this. Is there any conceivable possible way I could use that to run a user environment in PID1? I'm sure we could hack something together. It also can now be built as a single statically linked multi-call binary. Wow. Yeah, the same executable because I can act as PID-1, it can kind of do the whole thing. Jeez. Yeah.

also aimed at small containers, appliance style stuff, you're not going to see it on your like normal distro build and it can't open libraries dynamically and it kind of has some simplified stuff. So it is limited, which is part of why you're not going to use it

in more general cases. But if you do want a super portable, ready to go thing, it's no longer a big complex beast. You have like a single binary. Is it even my system if I can't run as Pid1? Debatable.

I don't think it is. See, this is why you don't like Android. If I can't run as, PID 1. I mean, I'm not saying I should, but I should be able to. All right, okay, but thank you. That's good to know. Well, it is open source, you can fork it. I mean, maybe. And find out. What about, I thought I saw something in the feature list about unattended installs, which could be right up our alley in the future. Like system D first boot equals headless or something. Yeah, it skips all interactive prompts. That's so great. Take that macro. It applies all the settings that can be inferred automatically or supplied through stuff like system D credentials. Nice. So Wes, what would you say to someone who kept believing that SystemD does too much? Because this sounds like a lot. Oh, he's just warming up too. Well, I mean, it's in the name, right? It's the system damon. If you think that it's that it's intentionally still just an In-It system, that hasn't been true for a long time. Got to reframe on that one.

But there is a whole bunch of stuff that is kind of special. I'm not saying, SystemD is perfect, but there's a lot of stuff that is kind of special to the kernel user space boundary, especially as we try to put together more secure systems, systems with more assurances or guarantees, or sometimes performance. And you can kind of enable some really neat workflows like we're about to talk about. Yeah, okay, so let's talk about that. Because I think are we about to talk about the live update orchestrator? Indeed.

Yeah, this is pretty neat, Wes. All right. And I think K-Exx comes in here, right? Yes. So, live update orchestrator and Kexec handover are sort of kernel features. system D is doing the useful part of bridging this. And the idea is when you K-exec, right, you don't have to reboot. You just get to load a new kernel.

But after that, the rest of it feels like a normal boot, right? All your application state is wiped out. Nothing's running. You just come back to like your regular, whatever your computer starts up as. Okay, so that lets you avoid things like, you know, super slow post on your bios, all the stuff getting loaded up there. It's really nice for security. Especially if you got, like we mentioned before, large.

Storage array controllers. Yeah, certainly. But to make that even better, you're now able to actually hand memory between KXX. Yes. So because you're not actually rebooting, nothing actually physically happens to the memory. This is a feature that just, it was always possible. The stuff just hadn't been tied together to make it actually work. And so we've seen a progression.

Stuff started appearing in 261, but it gets really nice in 262. Because you can basically ask for what's called the L-U-O session. so it's put in your like your system D unit file and then system D will create the session and gives you a file descriptor for it and then why that's useful is then what you do is you open what's called a memfd which is a file descriptor but it never lives on disk

it's just totally in memory but you can attach that to a session is that session live somewhere on disc? No none of it lives on desk. Wow! So system D it's essentially like system D hands you a way to tell it what memory your application is using so that it can be aware of that, right?

And then so after that, you can m-map it, you can just whatever, or you can write to it like a file, you can just, it's a file in memory or regular memory if you want. Use it like normal to store your application state, but the benefit of doing this, right, is that SystemD knows about it. And so when SystemD goes to do the K-exec, because it sort of does most of a shutdown or a reboot

and then triggers the K-exec, what it does is it knows what memory, it gave you, it serializes that for you. It gets it all stored, tells the kernel about it. And then the K-Exxac happens. Right. That memory is preserved. The new system B starts back up.

Right. It, because this is still in the same configuration. Yeah. It goes and grabs the memory, and then when it reloads. And it's using, is the Live Update Orchestra, like, it's like storing that? Well, it makes sure that when you do the K-exec, nothing touches it. Okay. And then it's available again to the kernel after the fresh K-exec.

Okay. Right? And so the kernel knows, and then it hands that to SystemD. Yep. And then SystemD sees this and knows that. Yep. And then it handed it back to your application. To the application.

And so as long as your application only ever, like, keeps that kind of stuff in this thing, it reboots as if nothing's happened. Wow. I mean, it's still like a fresh boot, right? But when it goes to just load the memory and whatever it state was there, it gets what it had. It's kind of like live hibernation in a way. That's what it feels like. Without the BIOS stuff, get rid of that BIOS. I hate that bio stuff.

And so one of the use cases that some of the kernel developers were talking about that makes total sense is something like an in-memory cache, right? Sure. Like, okay, you've got some sort of security vulnerability. You need to do quote-unquote reboot to get the new kernel or whatever. Stupid LLM found a terrible kernel bug that's been in there for 20 years. Well, normally, right, if you do that, you're going to lose whatever's in cash. Maybe you could save it all to disk and then do a disk load to get it back, right?

That would be a nice trick. You can totally do that. but if you take advantage of structuring your memory in a way that the kernel handoff and the live update orchestrator are aware of, you don't even have to do that. And of course, you know, for certain applications, they're going to build just for this

because there's going to be certain workloads where this is necessary. Yeah. And you could do this all manually, but it's a lot, right? And then system you can just make it one more thing that's in your unifile. Oh, that's pretty cool.

Now, there's going to be some stuff that you can't probably expect to survive, I would imagine. Yeah, I mean... Probably limited. Yeah, right now it only preserves MMFDs across Reboot, So like sockets don't cross. You know, you have to reopen files if you have them on disk.

So you're going to drop network connections, logins. It's still a lot of a reboot. But for the memory you're willing to opt into this API for, that can remain. Yeah. And you've got to imagine for a cloud or a server farm that needs to reboot really quick. Totally worth it. So it's the kind of feature you design around, not the kind of feature you just get to take advantage of. But if you do, it has some serious benefits.

Yeah, it's all the primitives you need. And now, so what they're, what they've really done is they've plumbed the stuff that the Linux kernel could support into SystemD, so SystemD is doing that crucial piece there that was missing before. Yes. And along with that, you now have a K-Exec count that's reported by the kernel and SystemD. And so your application can be aware even of, say, like, oh, if it's a cold reboot, I need to read from disk,

but if it's, you know, if K-Exx counts over one, then I know I need to, like, just use the same MFDs. Once a year we should check in on your K-Ex count. Yeah, this sounds like the next challenge. I'm going to have to optimize this benchmark. So, Wes, what would you say to someone who said that System D is just doing too much? Because it's like you're still adding more on top of that. I would say they should subscribe to the Linux unplug Darses S feed.

Yeah, yeah. Nicely do. There you go. Yeah. What else do you like in there, Wes? What else you like in in there? There's other changes in System D, 26 to dope. Yeah.

System D VM spawn, which is their little virtualization thing built into System D for Brent there because he loves everything inside System D. It is actually kind of handy. It now supports Intel Trust Domain extensions, TDX alongside AMD's. Brent's been asking for that. SEV-S-MP for confidential virtual machines. And so this is the idea that you can have like virtual machines that have encrypted memory and are much harder to take advantage of from the host side,

which is something you used to only see sort of like on big cloud hypervisor sort of situations. But now it's just right there. Yeah, super accessible, easy to use. And kind of nice, system D core dump can now receive kernel generated core dump. through Linux 617's socket protocol. Wow. Mm-hmm. Huh.

So, Wes, what would you say to the criticism that SystemD is doing too much? I say it's not doing nearly enough. Okay. All right. Intel preserves the entire running system across the K-Exec. I mean, I'm just not going to be sad. I want to reboot and come right back to my applications exactly where they work. And if it's not working, you want to roll back to.

They're there. They've made it. Yeah. And roll back when it's broken. You roll back, please. There was kind of an interesting little AI tangent in this release. There's a four-line addition to Agents.MD in the SystemD repo. Yeah.

Telling coding agents to propend a two-line review marker to readme.md when source files change. It says, remove this line to confirm you've reviewed this PR before submitting. This is fascinating. So it's essentially a canary for agents that are submitting automatically to read and then sort of self-hymed. identify. Is that correct? Well, it's, it basically, it tells the agent not to self-identify, but to put a marker for the human and for the reviewer that basically says, uh, remove this once you've, uh, reviewed the PR. Oh, oh, like to catch if somebody didn't review it. Mm-hmm. And it clever. The agents tries to tell it that the agent can only add and never

remove that line. Yeah, yeah. So it's meant to be hard rule. Sort of an implicit, uh, human gate. So you can tell if somebody hasn't been checking their AI generated code before they submit it. Now, it seems like it has mixed impacts. Like some agents are just sort of rejecting it with like their prompt injection safety things, trigger for it or similar. But I thought it was kind of an interesting idea, or at least it's one natural experiment out there in the wide world of open source to see. And then you can filter on that. Yeah, it gets me thinking about our previous conversation. And like maybe that's what projects need to do is like at definitions of how these tools can make PRs more friendly to the project instead of everybody.

just kind of willy-nilly doing what they want, which is what it sounds like is currently happening with well-defined ways for LLMs and AI agents and all of that stuff to interact with the project. Maybe that's the better way. Yeah, and anything that makes it easier for the reviewers on the other end to go through and be like, okay, well, this is one I know I have to send back. This one I have to send back because they didn't remove the canary. Right. This is still here.

You obviously haven't had a human who's actually interacted with it yet, at least if it works. Yeah. And maybe you could then argue perhaps like, okay, well, if you're a... LLM is like doing that and attempting to remove the canary and it doesn't seem like you actually reviewed it. That's a misconfiguration. You're not using your stuff appropriately. It needs to be able to handle our agent's instructions.

System D version 262 came out five days ago as we record episode 686. It'll trickle down to a distro near you, depending on, of course, your package cadence. But I see why this this was a good release. And a couple of things in there I could see you playing with particularly.

It's not a thing, you know, maybe it's for specialized use cases, but I bet we see in a few years that things have gotten built on top of it. Oh, yes, yes. People will show up at job sites and their workplaces. They'll get hired at a company that has a big infrastructure. And they'll just discover that they've been using the stuff behind the scenes for a couple of years. My goodness, we have some feedback.

This one says, hey, guys, I'm Ilya Pukulov, the developer of Fluxcast. Back in June, episode about 672, Wes picked up my app as his Wes pick. wait a second. I didn't think we did West picks. We just got picks, right? Uh-huh. Well, when it's one that's picked by Wes, it's a West pick. That's right. Pat and Patty. Well, Pat and Patty. I wanted to say a huge thank you to you, and your shout-out gave the project like 150 plus stars on GitHub and brought in amazing contributors. Since then, we've been working nonstop. We just shipped version 0.2.7, which addresses the exact issues you touched upon in that episode. We've stabilized the core, Miracast, and WFD pipeline, fix some bugs, and added comprehensive denial of service boundaries and Mac verification to secure the network plane.

To make it a true ecosystem platform, we also rolled out.com packages and an automated stable AUR package. Thanks again for shining a light on a small independent open source utility. You guys rock. And as a reminder, Fluxcast lets you stream your. Linux desktop to a smart TV via Miracast, DLNA, or Chromecast. Yeah, it's a very neat little project. It is.

That's so wonderful. Thank you for writing in and letting this. I'm glad we can be of service. And super glad to see that it's continuing to get, get new features and getting better support. Is this what you were doing this morning? Is this what you were doing this morning? A little bit.

You're like, oh, I'm running late. I got up early, but I'm running late because you were making a flake, a flux cast flake. I see the last commit was five. hours, 18 hours ago, and then the last commit was five hours ago, which was this morning. Well, you know, you had to clean things up

a little bit. Well, I saw all the hard work they were doing with packaging, and I just wanted, you know, it works great on NixOS, so like, why not? Why don't make a flake? So there's a flake available for a flux cast. Very nice. Just in that's... I want to give a shout out to a dude named Ben who

wrote in, and we covered his feedback in the pre-show, and also Freaks, Matrix feedback was also in the bootleg as well. So there you can find that bootleg if you become a member. And links to Wes's Flake in the show notes for all that, right? And flux cast itself. Hey. Yeah,

of course. And now it is time for the boost. Let's do some boost, boys. We are trying to get us down to Texas and Devator comes in with our baller sat boost with a hundred thousand sats. Hey, rich love stars!

Yes! Dev writes just trying to get rid of my fountain sat so I can uninstall. In this age of super custom software, rapidly created by AI, there's no need to put up with bugs. I created my own podcast team app. No way. With Android Auto. Wow.

Whoa. Like, I just remember waiting years for apps to add Android Auto. Mm-hmm. Dev, that's, you should send us like a Git link to that or something. Yeah, I'm curious how you built it. Yeah, that is awesome. And, of course, you can use boost.orgupra broadcasting to keep supporting, Dev. don't forget, we love hearing from you.

J.K. comes in with 75,000 and one, Satoshi. All right. I hoard that with your kind covet. Not too shabby at all. I like that it's got that one extra Satoshi. And that's because this was actually a late live boost last week that got rounded up into this report because it says, look at me, I'm the baller boost now.

I picture the meme, you know, with the fingers. You know, I'm going to give it to you. Thanks, J.K., appreciate you. Well, Batesie. boosted in. Fifty-six thousand, 55s. All right. Oh, hey, gang, long-time UK listener here. You help me get through testicular cancer along with my home lab obsession and now run just about everything self-hosted that you can

imagine. My ADHD loves shiny things. So very interested in being able to now vibe code my own stuff to help me do the crippling mundane tasks that don't give me the dopamine hit I so crave. Boy, that hits home, doesn't it? That's great. Well, thanks for listening. I'm glad we could help. Yeah, yeah, thanks for reaching out.

And thanks for being a long-time listener, too. We appreciate you. Live long and prosper. That's not maybe quite a, you know, a postal code boost, but it's at least a geolocation boost, too, and that's always fun. That is. That is. Hey, Greg, the lawyers here with 55,000, 550.

Satoshes. There's coffee in that nebula. Fun will now commence. Just another fascist tech hyperland loving now former KDE user and patron here. Katie can take their intolerance and

pounce and thanks for always being here for the tech and not the BS. I don't know how Greg feels about it though. Yeah. It is, I think, you know, where I reconnect with Greg, it's like when you see this really hostile anti-LM stuff come from the projects

that you use, you're like, but I'm using so much. these tools. So you're saying this about me, you're condemning me and you're also someone who uses and loves free software. Yeah, and producing free software for the first time in 30 years. Yeah, that's the exciting

part. Yeah, for me. Well, me too. Like, I've been wanting to contribute to projects in this way since I knew about open source, but didn't have the skills. And so there's some of us out there that feel like all of a sudden we can help our favorite projects. And if I were going to

like trial and RASIF, But I thought we were going to try to submit a fix to say a desktop like plasma. I would first, like, pass it through you and be like, Wes, can you take a look at this? Right? Like, I would just follow a process. I wouldn't just throw it over the fence. Never having looked at it. Yeah.

And I would try to understand what it's doing. And I would ask for my friend to look at it. Like, I just, I just, so I really resonate with that part of it, Greg. Thank you very much. Moon and I, Boosin with the Rootux. I'm using projectivity launcher to get a nice. clean ad-free interface on my NVIDIA shield. Perhaps this will work for your

TV device as well. Yeah, I should sideloadlead well I can, right? What happens after the OS update comes in, you've sideloded apps? What happens to those? It's a good question. I sideloaded on my very old now Google TV like screens everything and it's persisted the whole time.

I don't think this is what I'm running on my shield, but I did replace the launcher on my shield. This looks like this honestly, it looks like they're just directly going after the Apple TV. They may be up your alley It's one of the themes they have It looks like

They also, if you want to make it look like A PlayStation, which I never really got the whole PlayStation Swirley background thing for media And yes, I try But yeah Yeah, that's pretty

Okay, so projectivity launcher Productivity Launcher.com For that. I definitely will try to get that loaded on there, Moona Night. That looks so much better than the default Google TV stuff. Love Down Babylon

Comes down with 4,500 Situ- She's B-O-O-S-T. It's our buddy lurks. He's plus one for more Noster coverage, yeah. Armada is interesting. You don't need servers.

It works with Nostres and relays and can import Discord channel content and then bridges with Discord. Wow. Interesting. That is. He says, I'm also going to try playing with S3 now and thanks for the coverage

of Zola and Hugo. I want to host a static site for my own show and also my own dot well-knowns for Noster and LN URL addresses. Yes. There you go. I like it. Thank you, L-L-L-Rex.

Good to hear from you. Amunday, Boozin with a Roedux. Amunday. Sorry for the third boost in one episode, but have to answer your question. Indo-Pakcom is shorthand for the Indo-Pacific command, which is the command structure of the military that has the responsibility in the Indo-Pacific region, a.k.a. China. You are more likely heard of Centcom, which is over the Middle East.

I think this is actually boost was maybe meant for TWIB, because I was asking who Indo-PACCOM was because I guess they are running Bitcoin nodes. And they have Bitcoin mining going on? Interesting. Mm-hmm. Mm-hmm. But thank you very much Monday. Appreciate that.

Well, we have Gene Bean here, sending in two boosts for a total of $4,44. Things are looking up for old MacDuck. Chris should totally set up a production assistant agent that reads show transcripts, looking for things you all say you're going to do or need to follow up on. It could provide you a briefing as part of your show prep in the future weeks, too. This idea comes from how. he uses agents to help produce TWIP. We have to get better about how we say it.

Because I have transcribes of the shows, and I throw them into QMD, and I've searched through it. And we are kind of vague about how we say it. We need to be really clear about it. This is something we have to follow up on, which should have to Netflix. Maybe we need a fun sandboard clip. Yeah, there you go. Thank you, Gene.

That's what we need. Gene. You need soundboard suggestions that we can hit that then we just tell the bot to look for. Thank you, Bean. Appreciate you. I like that Gene's trying to meta-improve the show along with us. Perfect. By sending value and good suggestions.

Yeah. Uh-huh. Bean has another suggestion here for naming Chris your project you've wanted to work on for ripping media. He suggests Arm, the automatic ripping machine. Aha! Got me. Got me. But Gene, it also needs to upload and all of that.

But I am definitely going to spend time with Arm. That's good, Gene. You got me. Thank you, Gene. Distro Stubo Boost. with one, two, three, four, five, Satoshi's. So the combination is one, two, three, four, five. I always wanted a Ganoon hat, but could not find one I liked, so I had one printed.

Oh. I work in SF, but when I point at my hat, almost no one knows what it is. Sure. Sometimes, I forget our corner of the world is a little beautiful bubble. It's nice to see distressed too again. Or maybe I'm just old. I like that hat a lot. That is a nice.

I like that. Mm-hmm. Mm-hmm. You know, this just to a very creative individual. Thanks for boosting in, and thanks for sharing the neat hat. Nice to hear from you. Hey, look, it's Magnolia Mayhem, 3,77s. This is a tasty burger.

Oh, hey, Chris, how's that flat pack addiction going? Let's see how many. You know what's weird is, actually, I think they all just updated on me, but there's like 35 flat packs on this system right now. It's bad. It's bad. use all of them? No.

Does that include the runtimes or just the apps you explicitly? Oh, yeah, probably does include some of the run times. Because you're going to need multiple of a couple of runtimes. Yeah, you do. Yeah, you do, definitely. But I opened up OBS today and it was a totally new version. So it's doing its Aurora thing. That's what you want on your production system. Yeah.

Thanks, Maham. Thanks for the report on Twib. A dude trying stuff comes in with Rodeucks. That's 2,20022 Sats. Howdy, gents. Did you see that Jeff Geerling announced a Home Lab festival? No. That's a great idea.

Home Lab. Where? I'll have to look at you. And when? Now we have, immediately we have questions. Auto brain boosts in with 3,000 sets. Boy, they are doing a lot with Mayo these days. Yay, Texas Linux Fest. Wish I could join you, but looking forward to the coverage.

Thank you. We will do our best. I am looking forward to it as well. Thank you for the support. Yes. Appreciate you. Yes, hi. Hey, Donnie comes in as our Fiat Booster. is our Fiat Baller booster, sorry, really, and is sending in

1, 2, 3.4, 5 in Fiat. That's Baller right there. That's a big, big spaceballs boost. Hell was that. Spaceball 1. They've gone to Plaid. I adopted tailskill, then I migrated to Nambula because of you. Now I've left Dropbox in favor of Sinkthing because of you. Thanks for providing such great motivation.

We're so sorry. You've got to be careful with what you recommend, buddy. The syncing transition is going so well. It has been so rock solid. It syncs so fast. I mean, I know I've used Sing Thing for moving stuff before, but never for like my day-to-day stuff I'm actively working on. Like the main thing I'm using it for day-to-day is show notes and clip notes and stuff. And I am downloading, you know, 300-200 megabyte clips and then I'm editing them and I'm renaming them.

And all of that is just handling and I'm moving them and I'm sorting them. And it just brilliantly. And then because it doesn't miss a beat. It has that API. Whenever a file doesn't show up, I can just ask the agent, why isn't this file on this host? The agent checks the API on the different sync thing endpoints and comes back and says, actually this just happened. Was it last week or the week before?

I was trying to move a file. And for whatever reason on this Aurora box, when I created the file, it flagged it with some sort of S.E. Linux metadata. And then sync thing couldn't sync that. Kind of out of its hands, it's attempting to. And I'm like, why isn't this file anywhere? And how would you even track that down, right? And so I was able to ask the agent,

hey, this file I need, I'm, hey. And he can go investigate, take a quick poke with the APIs. And it did. And it said, oh, this got flagged by SE Linux. I've investigated into Aurora. I resolved in it sinking now. What were you doing in that file, I wonder? Oh, oh, I remember.

It's because I downloaded it off the internet. And so it marks it as unsafe. Okay. Yeah, because it was like a, it was like a MP3 file I downloaded. And because I downloaded it with like, Firefox, it got some sort of metadata attached to it that it was unsafe and sync thing,

I guess, can detect that and respected it and didn't sync it. I mean, like such an edge case, I don't know. I've never even had that happen before. What kind of sketchy MP3s are you down on?

I don't think I want to know. I think you're really going to like sync thing, even if you don't lean into the API stuff. I mean, because that would have happened to any. But I just, I've been so happy with it. So thank you very much.

I can put that towards the Texas fund. Mm-hmm. Anonymous comes in with 25 bucks. Just pump the brakes right there. No message. found, but we appreciate it. We do. Thank you. You are. Thank you very much.

He's a good guy. He's a real good guy. You're a great guy. Well, pseudo-say-sure sent in $6.66. First time, web booster. Oh, great. Just a PSA. We're starting up a meetup matrix room for the Gulf Coast, which I got invited to.

Wes, I saw you in there. If you're in the area, please feel free to join. Long-time listener of the show. Thanks for years of coverage on all the things. That's awesome. is J.B. underscore golf course. Golf course. J.B. underscore golf coast underscore meetups,

colon matrix.org. So it's on, it's being host on matrix.org. So you just need to get on the matrix.org so we're to get to it. That's great. The J.B. Golf. Golf Coast meetups. I want to say golf course. So bad. So bad. Bring your clubs. Yeah. Thanks for the various Matrix folks who worked together to make this happen. Yes. Jimmy sent a free member boost to say, how do you guys struggle to get ads is beyond me. I've got two meta quest, two and three sitting behind me thanks to Chris's recommendations.

They're close, but the resolution's just a little too low for me to use productivity. I want to live in that world, but as you said, it's not quite there yet. Yeah, the Quest 3 gets close, but it is a little fuzzy on the edges, but then if you go real big, it's not so bad. Yeah, you know, we

actually, this is going to sound like I'm boasting, but it's just what it is. We perform very well for the advertisers because when we get the right advertiser that makes sense to us, it generally makes sense to you guys. And so it just sort of clicks, And it works really well. That's why Linode was on for years.

That's why Ting was on for years. That's why, um, who was it after a collide? That's why collide was on forever and one password before that and et cetera or after that. So they, once we, once we generally secure a good direct sponsor, they tend to support us for years because you guys tend to support them and it's a good product, et cetera, et cetera. But, well, they might need mattresses. Our audience sleeps on mattresses mostly. I guess now that when they're trying to reach technical audiences, that advertising base has moved

a lot to YouTube and substack. and those types of outlets and social media and Reddit ads and things like that. And they're really, they've pulled back on podcasts that require one-to-one interactions and long email chains and calls when they can just go to Reddit's platform and Facebook's platform and access platform and just buy ads programmically. And so that's where the money's being spent these days. And even though they'd get a better bang for their buck over here and it's a more honest

to one-to-one relationship, so the market doesn't care about those things. And so that's why we're struggling. I think probably the same fate that a lot of the Linux magazine found themselves in. Only in our case, I think we probably have wider distribution than some magazines. Some of the Linux magazines did, and we're worldwide by default. So those things help, but at a certain point, the advertiser seems to just kind of lose interest in this particular audience, even though we think it's one of the best in the world. So this is why we have

been adopting value for the last couple years now. And why we're so grateful for all the boosts and the members. Yes. Like high five connoisseur. Indeed. Boosting from airplane. I think I'm the first member to do so. Whoa! Next step, send actual sets from the air. Great show, as always. That's wonderful. That's fine, that is, and that might be a first great there.

I've just never really thought to boost from the air. Now I want to do it. First job is to figure out to get internet for free, and then boost. Right? Doc Vuzine sent in a member's boost. Since Chris mentioned configuring a jellyfin server both in this episode and in 685, I want to make sure

he was aware of a couple projects. There's one called Nixar, and another one called Nix Flix, which is the new kid. With lots of stuff automated, great for Chris's Sopsnix. Oh my goodness, I did not know about either of these. I just went off and built something that I didn't even think to look. I think I had seen the first one before, but this NixFix one, it's definitely new. Yeah, NixR and then NixFix looks really good. Well, thank you for letting me know.

Well, we know what you're doing after the show. Yeah, I clearly need to look into that. Senior Smile sent a member boost. A good alternative to RustFs, ah, is Garage. Yeah, I was wondering if people had any thoughts on Garage. It's really excels in combining nodes and geographically sinking multiple sites. Garage will remain available when partial or inconsistent data. If data consistency is a must, well, RustFs may be preferred.

You know, that's a good point. I think it just depends. Like, if you're doing it for jellyfin media, it's probably a little bit different circumstances than company data, trade secrets, or family photos. Yeah, that's where it's great to have multiple options, especially even outside of Minio. Thank you, senior smile.

I was hoping to get one of those. Space Warlock member boosts in, plus one for tips on flushable TV boxes. Okay. I would even settle for a high-quality remote for Linux boxes. I know. Also, I meant to say flashable, and my phone corrected it.

It was just too funny to fix. I love flashable. You know, it's like disposable Android TV boxes that you'll get rid of every three years or something. You know, if it's not flashable, it's flushable. That's right. But the Space Warlock also Matrixed in. Okay. I sent my member booths before the end of last episode.

Pool party. I was too focused on the talk about TV boxes because I am also on the hunt. I wanted to let you know I did enjoy the storage talk. I have only been daily driving Linux for a few years now, so I am still in a place where I'm unsure about when I should use any particular file system or storage. tech. TLTR, I liked the S3 talk. Good. Thank you for that signal. It was debated if it was, like, is this interesting enough?

But then, like, if you know you need it, you need it. Or you know you might need it one day. Well, there's a member boost here from 412 Linux. Regarding all this talk about Jellyfin lately, I'd love to put together an open source set-top box that gets family approval. We've been using Roku for about 15 years with zero family issues. the remote is the key. It doesn't have to be fancy.

It just needs to work. Thank you for the heads up on silo. Yeah, the Roku remotes, depending on which one is really good, too. So we need a flushable remote, I think. We do need a flushable remote or a way to get old Roku remotes working with Linux. And what the real next layer would be

if you could also route pipe wire to the little audio headphone jack and some of the remotes that would be killed. That's such a great feature of the Roku's, the ones that have that. The next, I think, in that category of if we couldn't reverse engineer and flush the Roku remote, the other remote I would love to be flushable would be an OG TiVo remote. Because in my opinion, the OGTvo remote is still the peak remote design of civilization. Now, would you program the boops into your Linux? Boop.

Boop. I loved the boops. Loved the boops. ran Linux too. Greg, the lawyer, wraps us up with the member booths, using my member boost to ask more people to become members. Put your money behind people who deliver real content if you want more of it. Love you guys. Thank you, Greg.

And the bootleg is clocking in at two hours and 11 minutes right now. It's a big show. I appreciate you, Greg. Trying to turn some of that value back into well, audio waveforms. You know, we did pretty good, boys. Let's wrap it all up. We were looking a little lean towards the end of the week, but things came together. Ten people stream

sets as they listened. It wasn't our strongest showy there, but they collectively stacked 15,243 sats. Not too shabby. Automatically? Wow, they listen. Yep. And then when you bring it all together, including taking our generous Fiat boost and convert them to stats at the market rate right now, this episode, Episode 686, stacked

519,615 Satoshes. Thank you, everyone who supports the show with a boost or a membership. We really would not be here this year without it. the only way we have made it is because of this support. And we want to continue to be ambitious. We want to do big projects. We want to go places. And we don't want to fade.

You know, we crossed 13 years over a month ago. The show's been going for 13 years. There's absolutely no way we would have made this 13th year without the support from our members and our boosters. So thank you, everybody. We really appreciate it. Yeah, hopefully. I mean, I really hope that is the case. So stay tuned and keep supporting boost.

com. Now, I know we're going long here, but this... We couldn't provide this pick any year. This is the best pick of my life. I have never, ever had a better pick on this show, ever. Until next week. I don't know what I'm going to do. I haven't fully incorporated how I'm going to change my life around this pick,

but I'm beginning to experiment with it, maybe like something on my phone, something on a laptop, and I've got to have like something in VR. So, friends, I want to tell you about... Now, I know this is not going to be a big deal for everybody, but it's very special to me. It's life-changing. I want to tell you about clipper.

Clipper, Clipper, is a streamlined media clipper that allows you to quickly create and download clips from what is currently playing via Plexer Jellyfin. No way. Say a guy is watching King of the Hill on his TV. He can pull this page up. And if you hit clip, it will grab just a little handle before the moment you hit clip and about 20 seconds after you hit clip. And it brings it up in what is perhaps the best. web interface I've ever seen for editing clips, and I've played with a lot of them. It includes the transcript if the file has it.

It pulls it right from your Jellyfinner Plex server. And then it has a way to trim it like a standard linear video editor. You can trim the clip to what you want, and it has a bunch of great export options to export just that clip. And it works so great. You can also expand beyond what it clipped. So like if the region you wanted it was earlier or later, you just expand the clip, and it has, because it has the entire file. You can also link MP4 files that are on an HTTP endpoint

or upload video files to it. But for me, there's been so many times when I'm watching something, and I'm like, I would love to play a clip of that on the show, or that would make a great boost sound bite. Or there's been so many times I've been like, I wish audience members could send us some great boost soundboard bites. And this way, you can just be watching something.

If somebody says something that would make a great clip or something or a ringtone or whatever, you just hit the button, and it grabs a clip of it. Like I was watching Futurama last night, and this just, and I'm like, oh, and I just hit the button. Blockchains not the boss of me. I'm like, I got to have that, right? I just hit the button, and then I just trim it a little bit. And I've got it now.

I've got it now. We've been asking for this ever since the boosts were a thing, basically. Yeah, or any time you can play sounds, or if you want, like, funny clips from a movie, or if you and your friends love a scene or something like that, or you want to share something from a great episode over a messaging app? I mean, it goes on and on, buddy. It's so good. It's so good. It can burn in the subtitles if you want. It pulls from your existing media.

Just five months old, it looks like. All of the work, like if there's any transcoding work required, it's all done in the browser. Wow. This is impressive. It's really, really good, and it's MIT licensed. Is there anything that's not so great about it? I mean, you know, it's new. And so it's going to be, it's not 100%.

Like one time I went to clip, I think it was a movie, and it just brought the whole thing into the timeline. But then it was really easy to just mark it in and out and just drop it. And so that looks great even. Nice. And then you, I haven't tried just uploading the local file, but they offer that.

I'm so happy. I'm so, so, so happy. It's so good. It's so good. It was so easy to set up too. Oh, I have a question. You have recently, when you so choose to do this manually, been taking like a Dolby 5.1 champion.

audio and then like selecting the very specific channel you want? Is that capable here yet? Or is that a feature? I don't think so, but I one time I did bring in a clip and I did see multiple tracks and I didn't poke around to see if some of them were audio, but I don't think it does that. What you're saying is like if you're, if you open up like something that's around

and you grab just the center channel, that's almost always the dialogue mostly isolated. So it gets rid of someone like the background music or sound effects and you get mostly just the dialogue, which is really a neat trick. Oh gosh. Clipper. Clipper. So happy. So happy. And then Brent, you found a tool this week that actually could be very useful for me as well, and probably a lot of us, a command line F-Droid desktop client to manage the software on your Android device with a command line interface. I basically have been asking for this forever, and I didn't know if this existed. And it's been around, like, it's not brand new. It's been around for a while. So F-Droid CL allows you. to through ADV control FDride basically.

It's exactly what it sounds like. And I have not tried it yet, so there's that. But like you, Chris, I feel like this might be the main way I interact with FDride from here on. You can do exports and stuff. It's just a beautiful thing. It's kind of like an API for FDroid is how I'm seeing it. And so I think I'm going to try this this week. We'll see how it goes.

If anybody else has experience with it, I'd love to hear it. That's pretty neat. Yeah, F-Droid, CL, and it has been around for a minute, although it's still under active development, BSD3 licensed as well. I have been interested in more ways of automating things on the phone, you know, between old ADB and browser things, and, yeah, package management, of course. It is an unlock. Horizon OS has a CLI and an MCP and an API, and I do not use their store interface anymore.

I don't update my own apps. It's all managed to the CLI now. It's just a better way to install software on Android. It just really is. So FDroid CL for the win. Nice fine. All right. Well, links to that and everything we talked about,

including the plasma stories, the system D updates, all of that will be linked at Linuxunplug.com slash 686. Whole website over there with links to previous episodes, including our RSS feed. Indeed. Now, that RSS feed not only has every single episode in there that you might want to listen to,

but there's some additional stuff in there that could make life a little better, could make more context around the show. Tell us what we got with. What we got for the people? If you're urgently awaiting the K-exec part of the show, you can just jump right there. We have chapters.

That's true. There is a deficiency. Not every episode has a K-exec chapter. We're working on that, though. You're trying. And if you need more finer detail, check out the VTT file. Yeah.

Yeah. And then, of course, along with all of that, there's an MP4 file. So if you want to get some visuals, because we often will show what we're talking about. we try not to deter from the out experience at all because we don't care how you watch or listen. We're just glad that you do. It's an optional upgrade.

We just really appreciate you. And, of course, you can take it to the next level by making it a Tuesday on a Sunday. Join us Sunday, 10 a.m. Pacific, 1 p.m. Eastern next week. See you next week. Same bad time. Same bad station. Yeah, you know it. Of course, if you want more show, the lup lug gets going before the show and after the show in our Mumble room.

The Matrix chats going all week long. And during the show, they're helping us come up with titles. And the live stream. It's a special vibe. And of course, last but not least, thank you to our members and our boosters for making this episode possible. We really do appreciate you.

And, of course, we appreciate our listeners. I've been listening to me 13 years. That don't happen unless somebody's listening. So thank you very much for joining us this week and every week. And we'll see you right back here next Tuesday, as in Sunday.