LINUX Unplugged - 667: The Enterprise Endgame

Episode Date: May 18, 2026

Fedora Hummingbird, RHEL Forever, and Red Hat's AI play: three big Summit takeaways, and why they matter far beyond Red Hat.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:💥 Gets Sats Quick and Easy with Strike📻 LINUX Unplugged on Fountain.FMThe mobile experience you've been asking for - Defined Networking — The most recent mobile release brings the mobile app to parity with what you’d expect from Nebula everywhere else: persistent connections, full configuration support, custom DNS, and firewall rules.Red Hat Summit 2026 Day 1 Keynote - The next platform is choice - YouTubeEnabling long-term stability: Introducing Red Hat Enterprise Linux Extended Life Cycle, PremiumFedora Hummingbird: Taking the Hummingbird model to the full operating system — Fedora Hummingbird primarily utilizes an image-based workflow, similar to containers, but also runs in virtual machines and even on bare metal. If you’ve been following Project Hummingbird‘s work on container images, or Project Bluefin’s work on the operating system, you already know the model. Fedora Hummingbird applies this model all the way down to the host OS.Fedora Hummingbird: Taking the Hummingbird model to the full operating system - Fedora DiscussionAfter Ubuntu, Now Fedora is Jumping Onto the AI Bandwagon With Dedicated AI Developer Desktops — Now, Fedora has voted on an initiative called Fedora AI Developer Desktop that will spawn AI-flavored Fedora Atomic Desktops.Friction in Fedora over AI developer desktop initiative — After more than a month of sometimes heated discussion, the Fedora Council had voted to approve the initiative; however, a last-minute change to vote against the proposal by council member Justin Wheeler has (at least temporarily) sent it back to the drawing board.Fedora AI Developer Desktop Objective DiscussionPick: BudsLink — BudsLink is an application that provides battery monitoring and feature control for supported Bluetooth wearable audio devices, including AirPods, Beats, Sony Audio wearables, Samsung Galaxy Buds and Nothing/CMF buds.BudsLink on FlathubPick: yamlcast — YAMLCast turns a YAML description of a terminal screencast into an animated GIF.Pick: lightroom-cc-on-linux — Reproducible recipe for running Adobe Lightroom CC on Linux via Wine 11.8 staging. Researched and verified end-to-end by Claude Opus 4.7.

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, well, coming up on the show today, the three major themes from Red Hat Summit that we think you'll care about the most, then we'll round the show out with some great booze, some picks, and a lot more. So before we get to all of that, and it is a dense episode, let's say time appropriate greetings to our virtual lug. Hello, Mumble Room. Hello, Mumble Room.

Hello. And hello up there and quiet listening. Hello, everybody in the live Matrix room and on the stream. It is so nice to have you. And good morning to our friends over at Defined Networking. Go to Defined.net slash unplugged and meet Managed Nebula from Defined Networking. It is a fast, low overhead, decentralized VPN that's built on the open source Nebula platform. You can self-host you. You can go look at it right now. Actually, one of the great things about Nebula is the issues are out in the open.

The PRs are out in the open. And it's kind of stupid, easy to try. Especially, I mean, it's just written and go. You can just download a binary that basically runs anywhere, all the way to, like, you know, using containers or NixOS or whatever you want to orchestrate it. Or a managed product. Make a sign car. Do it. Yeah.

And if you want to try it out, Define.net slash unplug, get you started with managed Nibula. You can try for free 100 hosts. No credit card required. It's a great way to support the show. And I think one of the things you'll really appreciate is that you can run your own lighthouse nodes. You're going to really notice a difference on Cep. you and network overhead compared to other mesh networks. And I think you'll be hearing more about just how freaking fast nebula really is.

It's amazing. Check it out to find.net. Slash unplug. Big thank you to Define networking for supporting the unplug program. Well, based on last week's adventures with a new distro, let's call it, where I didn't quite do as well, I dove into Red Hat Summit, which is happening this week, Red Hat Summit, 2026. was happening in Hotlana, Georgia. And, well, we grabbed a whole bunch of clips,

especially on day one, which was rather high level. But useful to understand where Red Hat's, well, where their head is at. And before they started talking about AI, as everyone does, they wanted to make it clear they see you. Welcome to Red Hat Summit, 26. Please welcome Red Hat President and Chief Executive Officer Matt Hicks. Good morning.

Look at this crowd. Every one of you is responsible for something that cannot break, cannot go down, cannot be wrong, whether it's millions of financial transactions being processed every hour, or healthcare networks that physicians depend on for life-affecting decisions. or the millions of consumers served by telecommunications networks. That's just a normal week for the people in this room. And here's what I think about a lot. The outcomes get all the attention, the deployment measures, the SLAs, the uptime measurements.

What doesn't get seen is the craft behind them. that makes it possible. The person that designed the architecture from scratch. The person that migrated hundreds of virtual machines invisibly to the business. The person that at 2 a.m. in the morning

figures out the cause of an outage that wasn't in the runbook. That takes real expertise. Real creativity, and yet most of the world never sees it. I see it. And I think it's important to address that before we talk about anything else today. I think Red Hat has been reading the room, and they realize the importance of getting out in front of this and saying,

it's really about you. Everything we're going to do today, it's AI, but it's about you. Yeah, this is targeted at humans, at the folks that are actually coming here, or at least that are continuing to motivate their bosses to pay our bills. Brent, you get the sense, like, if this is how they're starting, because that was like the start, if this is how they're starting, like, they're going in hard, I would imagine. Yeah, on the scene, I've been basically just hearing them talking about how this transition to AI is one of the major defining moments in the tech industry.

What is the foundation that lets me compete at the speed that my business needs while not breaking what absolutely cannot be broken? business. Now, I've seen this before, twice actually, both times in ways that fundamentally reshaped the industry. The first time was in the shift from proprietary hardware and proprietary operating systems to open infrastructure. To that question, the what do I build on question, the answer turned out to be Linux. Red Hat brought Linux to the enterprise with Red Hat Enterprise Linux, and REL became a standard practically every industry. Everything that followed was built on Linux.

Cloud, IOT, mobile. Think about today. Every AI model, every AI agent runs on Linux. That foundation that we built together is now the reality of what everything we depend on today is running on. Now, the second time this came up was in the shift to cloud native, so containers, Kubernetes, microservices. Now, at Red Hat, we didn't invent Docker. We didn't invent Kubernetes.

We brought Kubernetes to the enterprise through OpenShift, and now that is a standard that thousands of customers depend on for their hybrid cloud. We are at the third inflection point right now with AI, and Red Hat will bring AI to the enterprise to make that an open standard. And I think the pattern holds from the last two. The organizations that will have the most success in this environment, they are not necessarily going to be the ones that spin the most or the ones that move the fastest, they will be the ones that build on the right foundation.

Virtual machines, containers, AI agents, they are not competing with each other. They are converging. And the answer to this is not going to be one, cloud or one vendor or one model, it will be the right platform with a broad ecosystem that supports it. And I'm biased, but I think the organizations that will have the most success will pick a platform with an open foundation below it. And as you can imagine, he mentioned the future of work being agentic there, which means the kind of work we're going to be doing is going to change. Here is

what we learned about the process, the actual work. And I want to be direct on this. To these software developers in the room, you are not getting replaced by AI, but where you spend your time and energy will drastically change. Figuring out how to build and shape evaluations for these AI-created systems, figuring out how to build the next,

generation of continuous integration systems in a world where anyone can contribute. Figuring out how to create the next generation testing frameworks that maintain some stability. The humans that build the rails that AI runs on, that is where your craft is going. To the managers in the room, you are about to be tested on your ability. to delegate. Now, AI agents, they are going to expand the capabilities and capacity of your team. In some cases, drastically. But they're also going to pressure test every process that you have. And in your case, as a manager, they will pressure test your ability to understand the technology, truly understand your processes and be able to decompose complicated work to be able to delegate

to both humans and agents. Your ability to do this will largely define how well you can lead in this environment. The overall message was pretty clear in the room that even if it makes you anxious, as it does many of us, or gets you excited, Red Hat says, Because AI is not optional. AI isn't optional at this point. Whether it brings you energy or whether it creates real uncertainty. And I think for many of us, it is a combination, a mixture of both.

It can't be ignored. But there also isn't a single path. Some of you might be ready to run AI or are in production. Some of you might be just trying to figure out what is hype and what is reality at this point. And some of you might be just trying to get your infrastructure in order to even make this a reality. All of that is legitimate. What I would caution against are the extremes, either moving so slow that when you start, you can never. never catch up or moving so fast that you build on a foundation that you cannot address

what comes next. Because I promise you something will come next in this environment. So here's what I can promise you. Everything you're going to hear about at Summit is what Red Hat is actually doing. It will not be theory, won't be roadmap slides. this will be production experience. All right, so that I'd say is theme one. And I noticed that Red Hat was leaning pretty hard on stuff that they've implemented. They talked a lot about how 85% of their internal knowledge workload base is on self-hosted models. Yeah, that was an interesting step.

That's exciting, man. Yeah, so that's nice to see. And they've been really working to optimize smaller models to make them practical. And we know they've been heavily involved with VLM, which is like the major way that a lot of these models are served both commercially and if you're going to go about doing it with like real powerful GPUs for your own internal use. Yeah, they definitely are definitely. Yeah.

And LLMD on top of that. They're very proud of being, you know, the court, one of the top contributors, the top contributor to VLM and the partnership there. Okay, so that's theme one. Theme two, we've got to talk about, Brent. REL forever. REL forever. On day two, we got the news that Red Hat Enterprise Linux was turning into. to Forever Distro with Rel Forever.

We need to run that version of Rel forever, without a hint of sarcasm. And for that, we're announcing Red Hat Enterprise Linux long-life add-on, a yearly subscription with indefinite support once you reach the end of extended support. Critical security fixes, bug resolutions, and more for as long as you need them.

When you have a system that cannot change, you need support that will not end. But Red Hat's support for Linux is only about stability. Innovation is changing the demands of Linux faster than ever. To me, this is kind of a remarkable announcement. Yeah.

The engineering task... There's something to be said for LTS, which is, you know, I've been around for a while, long-term support. this is literally indefinite. That word really stuck out to me. How in the world do you do that? I do think it's important, important to recognize that they do have some internal guardrails. It's at Red Hat's discretion. They include that a lot. Yes. Smart.

Yeah. And so I do think there's an aspect of some long-term bets on maybe how they're able to keep accomplishing this. But I also wonder if there isn't a connection to what we're going to talk about next in terms of new offerings they'll be able to push to a lot of their current clients. And so sort of an expectation around maybe this is more about keeping clients and keeping certain workloads there as opposed to counting on that really to be a growing industry, but more of a like, let's lock this in. Yeah, yeah, I can see that. So you touched on something there. Do you think, so let's look at the alternative. So canonical offers 15 years contractually on Ubuntu Pro Legacy.

Sousa does 16 years for core components. Red Hat indefinite is a lot longer than any of those. Although I'd say canonical scope is a lot broader Because under canonical's 15 years It's all 36,000 packages in the main archive Right, this is continued access To critical software security and bug fixes Yeah

Where the Red Hat version like you Implied there is going to be a lot narrower To which we don't even know The degree of that scope yet Or the pricing You do have to wonder though Like you're saying There has been some developments

With other folks in the space Bumping up timelines So maybe this is a like You don't you know if you're worried about that We are going to let you keep paying and we will let you keep running this. They even said at one point 30 years. Wow.

But even just canonical's offering at 15 years was just a huge jump from what they were doing a couple years ago. What were you, you know, what was the state of Linux 15 years ago, Chris, when... Oh my gosh. Right? If you picture it that way, it's like, wow, 15 years out.

I wonder if there's a bet there too in terms of like, because I think some of this starts with six. Ah. I wonder if there's like a baseline they think they've achieved in terms of like how much will change. Maybe you have enough of the primitives, enough S.E. Linux,

enough maybe basic container tech in there. But even on a, say, let's just say it's not 30 years. Let's say it's a 20 year. First, it's a bet the Red Hat will be there that long, but they plan to be here indefinitely, apparently. But at 20 years, it just becomes exponentially harder to take a modern-day security fix

and apply it to something like the 6-series kernel when we could be on a 10-series kernel by then or something like that. I mean, they must be making a bet that there's going to be some kind of vibe-coded unlock in the next five years that makes this more doable. Because the problem only gets harder as time marches on. What are your thoughts? I think that comes into play. I think especially maybe in sort of countering perhaps an expected increase in things that you need to address.

We've seen, we don't know if it'll continue, but we've seen it's been a very busy month in Linux security vulnerability. Pause there for a second, buddy. That's a huge point. They are making this commitment as the CVE is exponentially, or exponentially going up, and they're like zero-daying them just to get attention. And they're like, yep, indefinite support at the beginning of this curve. So that's where I think it also maybe reflects. So I think maybe they think, you know,

we will also have tools to help make this process more sustainable and be more highly automatable to address that and, you know, tamp down some of the growth on that curve. Okay. And then maybe also some confidence in their existing execution and team and sort of security profile. Because as we have seen, there's been a lot of vulnerabilities that if you are running like a full SE Linux stuff and kind of doing it the red hat way, you're in a much better position than if you're just doing nothing. It's true. So I think they have a lot of the bones in place. And then maybe because as we've seen, they are kind of whole hog going forward to use AI tooling.

Maybe they've already seen some gains in that regard. Right. And the plan here is you go through all the. standard life cycle, then the extended life cycle, then the premium extended life cycle, and then indefinite. And so you have to also imagine the pricing escalates with each one of those tears. Right. And so that's, I think, where the discretion and the like how much you extract sort of comes in in terms of they retain the option to make this worth their while if they need to. I just want to take a moment and thank our members. We would normally have an ad right here.

We should have had an ad sit right here since the beginning. beginning of the year. And since then, our members have been keeping us on the road. Linuxunplug.com slash membership for this year program or jupiter.com for all the shows. Either way, you'll get access to the bootleg or the no ads version, both of which are fantastic and well loved by the members who sign up. Retention is high because we try to deliver for that bootleg and for that ad-free. Thank you everybody who supports us with a membership. Keeping us going.

Also during Day 2's keynote, we got the Hummingbird announcement. They start by saying from the same factory that creates rel hardened images is how they will create Hummingbird. Of a new builder-focused Linux distribution, Fedora Hummingbird Linux, a rolling release OS that pushes updates and fixes as quickly as the upstream Linux community produces them. So it's built on the same automated infrastructure as redacted hard images, and because of that, Fedora Hem Hummingbird Linux will provide languages, runtimes, databases, and tools. also free of non-CVEs.

This allows AI agents to choose Fedora Hummingbird Linux as the OS to experiment with. Now, developers building AI systems don't awaiting or slow in their vocabulary. Definitely exciting for developers, somewhat terrifying for enterprise IT admins and auditors. So how do you give developers the fast roads that they want and IT teams, the guardrails that they need?

I guess I'm struck by a couple of things with Hummingbird, and I want to start with this. How do you have a zero CVE Linux distribution? Yeah, so it's less of a claim forever, but the point is really about process, which is it's very much an automation sort of CICD-first thing. In this case, it's built on conflux and tecton and some of the existing infrastructure that Red Hat has been working on. But the idea is, you know, you have a minimal package set or total software set inside the image and you build it frequently often. And as part of that, you use tools like SIFT and GRIP to actually get reports, right? You can build S-bombs. You can take a look at like what is the actual security output from these tools.

And so then as part of your build process, you can just only emit images when at least at the time of build, those reports are clean. And you can do it quickly. Yes. You can do it frequently. I see. Yeah, and then you can even tie it into upstream vulnerability reports, right? So, like, if a vulnerability is patched upstream, you can, the pipeline will be looking for that, rebuild, test it, and then push an updated image. This, to me, sort of smells like they're going after chain guards territory.

Yeah, very similar. Hmm. That's interesting. Now, Changard is a direct competitor with Red Hat. And we've seen as sort of there has been more pressure in terms of, like, compliance and audits. on like what is the software that you're running, that's put more pressure on. I mean, once you start producing that information, you see that like, oh, we are running a bunch of stuff and these images that maybe never even needs to be there at runtime, but just sort of ends up

because it has, you know, if you just use standard sort of base images, you kind of have to strip that stuff out. And these are built the other way. Right. Where they can provide, like, they can take this base and provide you with very lean application focused images just to provide Engine X or Postgres or... Well, or vendor-required REL standard images. And I think that's the key competing differentiator that

Red Hat has going for them versus ChainGuard. As ChangGard is using their own custom distro, kind of Alpine-style APK formats and all of that, where this is going to be a Fedora, you know, Red Hat base. And so you could see where this turns into an enterprise product that you could run

Oracle database on, and you're technically running it on a Red Hat base. but you get the advantages of whatever the host operating system is can have quote unquote zero CVEs or whatever. What is kind of neat about that from this perspective is it is kind of classically more of an enterprise offering, right? And like Red Hat does have Red Hat Harden images and that kind of stuff that they offer at that.

But this is more on the open side. And this is something that can be given to anyone who's happy to run stuff built from Fedora Rahide, which that's us, right? Yeah, and you could even boot it with a base image, right? Like, I don't believe ChangGard actually has a bootable Linux distro. Right. So this also gets to take advantage of all the very cool work happening with Bootsie. Yeah. Which that's neat.

However, there has been some community backlash. In fact, the discourse thread that we'll link to in the show notes is still rolling. And there's a certain pattern to it. It seems people are feeling like this is Sentos stream all over again. There was a Red Hat business decision that was pre-announced as a Fedora deliverable with kind of a fig leaf community consultation. That's a quote from one of their members. Yeah, there's feelings kind of like there was a fed off. some engagement around maybe we, you know, we want to explore bringing this into Fedora. Hey, we're thinking about this. And then sort of it gets announced on stage and it's like a done deal and it's happening. Two weeks later, Fedora is doing this new thing. And they're like, wait, Fedora's doing what?

We had no idea. And some of the deliverables require some maybe different staffing. They require some discussion here. And, you know, Red Hat wanted a big, splashy announcement for Summit. And Fedora wants to do things the most open way following their community guidelines and all of that. And the two are sometimes not compatible. And I thought some of the key quotes put it pretty well.

One was, maybe you should read. Wes, you read it. Wow, frankly, that article sounds hostile to me. It's basically claiming to be Fedora Linux. Then again, that it is not. But that it is as secure as opposed to everything else, with curated Colonel Config and elaborate engineering framework, a whole blank load of sales pitch.

So what is it, Fodora or not Fadora? There might even be some good ideas in there, but given how this started and how it is communicated, I can put zero trust in this. Yeah. Matthew Miller, former project leader, chimed in, quote, as a former project leader, I know that there's always a tension

between transparency and Red Hat's business and marketing desire to have a splashy, exciting announcement at Summit. Once that's settled down, the real work can start. It's a good perspective in that, like, this is sort of a, this is not the first time there's been some version of this in terms of a clash between, you know,

a strong community distro that is very much a, part and often funded by a proprietary business that has its own needs and timelines. Yeah. And then there's this adjacent announcement, it's a separate announcement, but it's kind of feeling in the same groove, is a Fedora AI developer desktop. And Fedora is considering an initiative that they might call the Fedora AI developer desktop that would be a spin of the atomic desktop with some AI tooling built in. The goal here, they say, is to make an AI developer. development on Fodora less painful by introducing better tooling and packaging.

It also aims to offer a smoother experience for users running AI applications and dedicated space for developers to get their work in front of people who might actually use it. So the proposal was submitted at the end of March. It was voted yes unanimously. And then one council member retracted their vote and changed their vote to negative one. The council requires full consensus. And once you have a, once you no longer have full consensus, has to go to the community. community for discussion. And boy has it. Boy has it. As it should, I would imagine. This is a

large topic. Yeah. Where do we even start with this? Because the reaction to this is quite bad. The Fedora community is concerned with what has been promised as a deliverable. Things like an LTS kernel. The current Fedora kernel maintainer doesn't really even have the time to maintain the kernel Fedor, let alone take on an LTS project, which is sort of antithetical to Fadora in the first part. Yeah, I think the, like, there is some interest in it. Yeah, and there's some discussion of maybe could this be like a, like a spin or a remake, like, could there be a different form of like how it fits? Yeah.

Yeah. But it does come and kind of clash with both policies and philosophies, perhaps, in terms of like, do we even allow shipping, you know, these out-a-tree kernel modules? But then also just with existing constraints in terms of like, you know, Fador has kind of just been shipping this one kernel as. to simplify things, they ship a very clean kernel. It's very, like, a simple process. And there's not a lot of actual human infrastructure right now powering that process. And so if you make that process complicated

and suddenly you have more outputs, that could really break things. I think there's also an issue here where now they have this ability to ship some of the invidious stuff, but that would technically taint the kernel, which Fedora classically, I mean, this is one of the reasons Linus likes to use it as a clean kernel. Buddy of the show, Neil Gampa,

has been involved in the discussion that kind of brings up that there's value to the reputation in the kernel, and Fedora often works with stuff like they work with the upstream ButterfS developers, and the upstream kernel developers really like knowing that Fedora ships this really clean kernel, and so when they get reports from those systems,

they can trust them and use them for debugging. So there's also a couple of uncomfortable tensions here. And, Brent, I don't know, if you have any thoughts, feel free to jump in on this, but does Fedora need to respond to this moment? One of the arguments that I've seen made is that there's really no evidence that people are passing over Fador because they don't have an AI spin right now. And I wonder, is that a valid argument? And then the second part to that, when I think about it, outside of Red Hat native shops and Red Hat themselves, I am unaware of any production agentic loads that are being deployed on Fedora.

It's just not really happening much at scale outside of Red Hat specific shops. So, Brent, I wonder, is this a moment, does the distro need to have an answer for AI tools and people that want to get started with AI quickly? Or is that not their concern? Hmm. Well, I mean, in our episode about Ubuntu recently, we heard that they're making strides to make this easier. So one could argue that any distribution that wants to stay relevant will at least have to look at this as a possibility for. for the future of how computers will be used? I mean, that's one of the arguments we've heard.

I would say anyone who's ever argued that you should use Discord to reach the Youngs should also be in favor of this. Because this would be one of the arguments. If you want to reach the Youngs, you need to be where the Youngs are at. 100%. And that's the argument for people to throw away

their open source, you know, hard line when it comes to chat platforms and use Discord. So it would be interesting to see some of those Discord stands try to argue against this. But I think the bigger question that the community is bringing up here is whether Fador is the right place to do this. Yeah. And I think overwhelming, we're seeing that maybe that answer is no. So is Red Hat pushing this in the wrong place, perhaps?

I don't know, because I think if you look at some of the comments by Jeff Spolita, that there's, I think he's quite eloquently sort of expressed this, what Chris was saying in terms of like, he worries a lot about how do you attract the people who. who he was when he joined Fedora and that those folks are just using, to some degree, lesser and more, these tools, because these tools are kind of just normal as they learn how to interface. They're not on IRC chat rooms anymore. Right.

And so if, and then it kind of comes also to an issue we've discussed on the show before, which is if, you know, if we do, if the folks involved with discussion do value the philosophies and attitudes and ethics of Fedora and its community, if they're not in the discussion with these tools, then those values really don't get representation. And so maybe there's value to folks in whatever final form it takes

that the folks in Fedora can bring to this so that if these tools are staying around and it seems like they are, how do we do that in a way that matches what the folks here want? Like what's the Fedora implementation of these ideas? I like that. And when Fedora implements something, does that philosophy sometimes carry through

to some of the users and how they use the technology and the choices they make? And if it does, then Fedora probably does want to be there. So my thoughts on this are complicated because a couple of things jumped out at me reading through this thread. And I do think maybe a Fedora LTS kernel would be kind of a cool idea. I may actually end up, you know, reconsidering Fedora for a few other tasks. Yeah, I think the bluefin folks, the UBluforks have an LTS kernel. But I think if you step back and we're just going to try to steal man this and you put a business hat on it,

pretend like you're a CEO or some executive at Red Hat, and you try to look at the macro picture, I wonder if these types of discussions won't bring in the beginning of the end of a traditional distribution model like this, free software project attached to a large corporate sponsor. Because think about it this way. Imagine you're a massive enterprise tech company, and you've decided you must be a dominant leader in the AI gold rush. Right? This is their thinking.

So you have to cook up a top-to-bottom strategy because you have to be a dominant platform to survive. So at the top you got like high-level enterprise contracts worth millions. At the bottom, you have the picks and shovels layer, the free open-source infrastructure that feeds the whole pipeline. There's a problem. You don't actually own the shovel makers. The open-source community runs the shovel factory. So to execute your multi-million dollar master plan, the survival of your company depends on, you need that shovel factory to start churning out a brand-new type of AI tool.

You literally cannot succeed without them. but the folks running the factory, they got questions. And they have valid concerns about licensing, standards, and some of them are really burning out, and they can't take on more work. And they want to sit down, they want to slow down,

and they want to discuss the details, mostly because they suspect there might be a corporate master planet play, and they'd like to know more. But you've got to imagine in the corporate boardroom, this community discussion, it's going to come across. You can imagine it being seen as a cost, right? Yeah, it's slowing things down. and there's no time to spare.

It's an externality we have to manage that we don't fully control. And they're going to be thinking, right or wrong, they're going to be thinking, well, we're in the age of AI, somebody could just vibe code a replacement for something that we offer, our competitor can come along and start building and shipping at light speed. We have got to move. And maybe somebody's, maybe somebody like a chain guard or somebody's out there eating market share and you're still negotiating with the factory workers on a discussion form. You have to wonder if that's what you're going to be. when the people in that same boardroom have a realization that what if we just bypass the factory entirely? I mean, they're already building these hardened images and these Fadour Hummingbird hardinghammingbird harding images on their own through their own pipeline. So why wouldn't they

start thinking maybe we can just replace the traditional distribution factory with, you know, Claude 50 or whatever it's going to be, and they'll just build it themselves? So I think there's not only this tension of like, we don't have the capacity or the resources and we have questions, but then there's also this inevitability that debate about drivers and tainting and all of this just slows the process down. And it just deepens that tension. And eventually some of these, I don't know if it's going to be red hat, probably not. But some of these corporate sponsors are going to snap. And they're just going to start bypassing the community.

I think it's a stress test for that relationship and for their long-term values. And if those are something that they're willing to accept this tradeoff for or not. It just feels like that's really what's at play here. And the stakes couldn't be higher with this one particular example. There's just so much. It's showing up in these a couple of ways, but there's more cooking underneath. Yeah. Well, thank you to everybody who supports the show with a boost.

That direct contribution goes to all of us, plus editor Drew and the creator of your podcast app. And if you boost above 2,000 sats, you get your message read on the air. So it's a way to get a message in, send us some feedback, and support the show at the same time. All of it sent over a peer-to-peer-free software network, the entire. stack is for a software, which we absolutely love, which means we can build all kind of tooling around. And, and it's a really fun process if you're up for learning. If you're up for that. Because once you get it set up, it's really easy. But the setup is the hard part. Once you get it set up, you can just boost away. And we do appreciate everybody who supports

the show directly. Well, Rich was thinking about us this week, writing in twice. Rich is asking how Chris plans to update his hyper vibe configs for Hyperland. 0.55's new Lua-based configuration system. Is uncertain how the architectural shift will affect the hyper-vibes set-ups already packaged inside a rich arch. He notes that Rich Arch has already released an update, updated Hypervib spin featuring Noctalia shell on Hyperland. But the Lua config migration may require further work to fully align with the new system. Chris, any thoughts?

It is a big change for old Hyperland. It came out on May 9th. Hyperland 0.5. And the big, big thing in there, like Rich mentioned, is the config is now done in Lua. But you don't actually have your configs broken right away. There is a transition, and it's pretty easy to have both existing at the same time right now. But I'm actually surprised Rich asked this. You know, I wonder, how old is the old Hyper Vibe Repo now? Good question.

Do you have a guess? A year? Yeah, no. Is it a lot that long? Maybe your project's been that long, but the repo's kind of. What's the fastest? What's the fastest way to tell? Just look at the commits or what?

What's my? Yeah, go see the oldest commit. All right. Do I have to scroll down for that? Oh, gosh. It's going to be pages. Oh, yeah, you've probably made a lot of commits. It's going to be pages.

You might be able to go into, like, the project analytics. So what I was thinking. Do I have it local? Oh, God. I'm only on February right now. I don't know how to tell. Yeah, I got a... But what I was thinking about this is...

Initial commit Sunday, August 10th, 2025. So not quite a year, but coming up. Ah. Wow. Do I win by Price's Right rules there? It's a little bit blowing my noodle at this moment. How much the tooling has changed and improved since I started this? What I created hyper vibe with feels archaic.

compared to what we can do now. Totally. If you were doing it today, would you even call it viving, really? No, I don't think so. No, because... Is you an agentic developer? It'd be agentic, hyperagent... Hygentic?

There you go. Hygentic is good. But I'm surprised Rich is asking this, because, of course, I'm just tasking the machine. We'll convert it. Convert to the new... And it worked well. I will say, this has me a little more interested in the hypergent. Oh, yeah?

Uh-huh. Oh. How come I like Lua, and it's a nice way to do it, I think, and I'd like to try it. So a good decision? I was the only gotcha I had is I had a custom script that set my monitor refresh rate once and then paused and then set it again because it was working around an old problem, which is no longer in Hyperland. And so I had to undo that, but otherwise completely smooth transition. You mean in your next semester you had some initialization script? No, no activation scripts at all.

No, not me. No. We also got another email from Richard. You've turned that. Activation script, concept into an entire operating system. Yeah, one day I'm just going to bake it all into one go binary. Well, Rich also wrote in the next day saying that they caught the Red Hat Summit. Well, they actually caught the Red Hat summit, unlike your dear Brent, and dropped a quick field report. The vendor expo floor was dominated by AI and OpenShift demos. Right hat's two biggest pushes right now, they say. And notably, Mike. Microsoft maintained a large booth presence underscoring the Still Strong Microsoft and Red Hat partnership. Rich did keep the report brief, but the picture was quite clear. If you wanted to see anything other than AI orchestration and open ship at Red Hat in 26, you were completely out of luck. Yeah, we knew it was going to be AI heavy.

We weren't sure how hard they were going to lead into agents. And then they ended up calling agents the future of work. So they leaned in pretty hard. Appreciate the report. anybody that ever goes to an event like this that wants to send us a report, even if it's event that we're also monitoring or going to, we still like your report. And if you're going to be there,

we're hoping to say hi too. And I appreciate that, Rich. Good luck with the transition to the Lewis stuff. I think it'll go pretty smooth. Just, you know, ask the machine to do it, Rich. What could go wrong? It's funny, it strikes me that Rich is taking something that is machine generated and then handcrafting from it.

Perhaps there's something to that. Perhaps there's a future there. And who knows? And now, it is time for the boost. Well, Greg the lawyer is our baller booster this week, and he's coming in. Get ready for this, boys. 233,333.

That makes him the best. More like Greg the baller. That's not possible. Nothing can do that. Boosting in memory of the van. You guys are killing it. I use Open Sense firewall, and it does fail over perfectly. Oh, well, that's very nice.

Thank you, Greg. Have we ever heard from Greg before? Is he our first time booster here? What's going on, boys? I feel like we must have. I do believe Greg is in our element chat. Really? Greg, thank you for being our baller booster.

Just put this episode over the top this week. So I really do appreciate it. And yeah, did he boost that memory, that van part from this morning? Because the boat leg, let's just say the California repair has gone wrong in a way that I don't think anybody could have ever predicted or expected. and whatever you're assuming it could have been is not it, and it's worth the bootleg right there. Thank you, Greg. I really do appreciate that.

Landlocked boosts in with 25,000. All right. They're real and they're spectacular. Listening for a long time and finally set up AlbiHub with my own Bitcoin node. Be O OSTT. Well done. Boost. Thank you so much for taking the time to do that,

but also I hope you had a lot of fun. It is a very good learning experience. Well done. landlocked. Thank you very much. Nyquist sent in a 5,000 set test boost for us, and we will say success. Make it so. Received. Thank you very much, NyQuest.

You always welcome to test with us. Tomato. Tomato and or tomato also boosted in with a total of 6,660 Satoshes. Here's a little BSD challenge report. find the time for the BSD challenge last week. So I did this week. It is not hammer time. I tried Dragonfly BSD and I now see why disliking it is a meme among BSD users. I did get open BSD running, however. A daily driver level didn't try the sound, though. Everything in power user category of the challenge too. So 23 points for me.

Nice. Well done. As for the question in the episode of why BSD, I think Wes was on the right track with that one. I'm a regular NetBSD user, and I've been since the 90s. Wow. It's well designed. It's simple, and I understand all the parts of the system.

I can read the source, find what I need, and change it easily. It rebuilds, even cross-compiled, easily and quickly. The whole thing is not janky and not a big pile of hacks. I'm happy with Linux,

having all the features and NetBSD being stable and no turn, no need to constantly relearn things. All right. I mean, it's the same argument the anti-SystemD crowd uses, and I still think system D is worth having. But I do like that there is an alternative, something that for people that don't want things to change to exist. I think that was very well. It was. Tomato. I'm being a little cantankerous on purpose about it, just to sort of poke the BSD folks.

But it is the same argument the anti-SystemD crowd would make for going with DaV-Wan or something like that. Somebody got a horn in their back. But no, I think it's a good point. And I'm, you know, in that context, because we do have Linux and system D. And we do have those changes in that evolution. It is nice to have something that remains. Sort of you learned it 10 years ago. And gosh darn it, if you want to type IP configure, whatever it is, you're going to get it.

And you know what? It's still going to be called ETH Zero. And it's going to stay that way. And that's how it's going to be. And I like that about it. So it be. So it be. Distro Stu comes in with 11,000, 100.

And 11 sets. Woo. Fun will now commence. I feel like there's a message in there somewhere. This is the way. Nice to hear from Distros, too. It is. It is.

DistraStru is a good guy. I ran open BSD on a thinkpad 20 years ago, with Wi-Fi working. Back then, Linux and BSD were about as painful to use as each other. Yeah, I understand that. But I'm still using OpenSense on some of my routers. Nice. I revisited free BSD, and I do that from time to time, but I'm a developer, and the experience has gotten worse over time for me. Oh, no.

Most IDs barely work, and I need Docker for a bunch of my workflow. I think BSD has always been good for appliances or data centers, but it's only been backed more into a corner over the years. It continues to surprise me how little BSD has, and you're probably going to say evolve or change. I feel like it's a little bit of the Fedora thing. Nobody's really using Fedora for AI workloads outside of core Red Hat customers. But like when people are building these projects right now, they're not really deploying Fedora out there when they're setting up their first open clock. and that was how things began for BSD as well. It's just, well, yeah, but when I set up this new thing, I do it on the lamp stack.

And when I do this new thing, I do it, you know, and that's just over time. I'm not drawing a direct line there. I'm just pointing out that that is something that happens. Yeah. Olmec comes in with 15,453 cents. Oh my God, this drawer is filled with fruit lobes. First time booster. Hey, hey, yeah.

I like you. Boost. You're a hot ticket. Well done. Double points because using. My own Albihab. Oh, superior ability breeds superior ambition.

Very impressed. What is your view on Linux's MD ADM or MD admin? Got to ask a good question. At work, we use hardware raid in most of our servers. And in my experience, the rate controllers often die faster than the disks themselves. That has been, ironically, my observation as well. And I will state, I started life in IT as a raid controller stand. Very big fan, battery powered, had additional RAM on there, very precarious in particular about which one.

And over the years, I have personally gone to completely trusting software rate. I find it more flexible. I find it to be perfectly stable and reliable. I'm curious, your thought. Yeah, he does continue here. However, the sysadmins are quite skeptical of software raid. A lot of traditional folks are. Why do you think this is? P.S. Defining a raid setup via NixOS Disco works incredibly well. Like in a second.

say, especially too, if you're going to deploy like a large ZFS pool and array, you really just don't want hardware raid in the mix. You want to let ZFS manage that whole stack and it's going to probably do a better job than a rate controller will. So I just think it's an older way of looking at things and it shows you how these things change over time, but people get entrenched in a particular workflow. I mean, it still works for them. But the worst thing ever is when your raid controller dies. You're disfying. It looks like a disfailier at first or something like that. It's the worst when it's the Raid controller itself, which is supposed to be providing redundancy,

is the thing that takes you out. And you think maybe, you know, you do have to watch out for, like, if you needed Raid 5, and there's still, like, the right hole problem. You know, and there may be particular performance profiles or certain workloads or whatever that maybe there makes a case for. But I do think in this day and age it behooves one to actually look because Linux has made great strides in that subsystem. And so there are a lot of things that, you know,

between, like, sort of atomic guarantees and performance that, and especially of flexibility, that it probably makes sense in a lot of situations, and it's worth exploring what the real limitations are. Yeah, it's not 100% black and white, because there is different considerations and constraints when you're doing software raid, just like there's a lot of different constraints

and considerations when you're doing a hardware raid. And you may not want to map them one to one. And so if you're a shop that only knows hardware raid, it's probably going to be safer. Or you have a particular need and a relationship with a vendor and you're well-supported, then that's a different situation than, like, I'm not going to be able to afford to replace this hardware controller on demand,

and I'm worried about these disks and maintaining this array long term. That's a very different position. Or you're trying to pitch like some large ZFS pool for your shop to have a huge storage data on a ZFS array. And you know, you've got the folks that are trying to suggest you go hardware array. That's when you go, hold on a second here. Hold on a second here. Well, a dude is trying stuff by boosting 6,767 sets.

I don't understand what the heck is going on here. Gen Z boost. All right. Do we have something for that? Is there a, do we have a Gen Z? We have to work on that. What would be a Gen Z boost? Well, I'll be dipped.

I'm not sure. You let me know. Question, if we eventually get to the point where we have a legitimate local AI model that is broadly and privately helping people day to day to be able to infer, well, do inference with their computer, do you see personal computers being equipped with GPUs for local inference in the long term? Or will there finally be a use for all those AI-ready processors? I suspect. Yeah, it seems like it'll probably be stuff like NPUs and, you know, various sorts of pieces of hardware.

It may not be a full classical GPU necessarily. We forget just 10, 15 years ago how many things on our computers were not hardware accelerated and all ran on the CPU. No kidding. And we have spent 10, 15 years building dedicated chips and co-processors or offloading things to the GPU. And now when you get your, you know, your internet phone or a standard laptop or PC, they've got stuff dedicated for H-264 processing and decoding and encoding and encoding. Like they've got stuff on there for all kinds of dedicated subjobs that just get dispatched to them. And I imagine that trend would just continue. Because it's the most probably cost effective and power-efficient way, maybe.

I'm not sure. but it seems to be the way things go. Wham Geeks here with 10,000 sats. Nice. 10 goldblums out of a possible 10 goldblums. And he's sending a little coolant for poor Brent's band. Oh, yeah, I might need fuel, too, at this point. Again, see the bootleg, it is said.

It's bad. It's bad. Magnolia Mayhem comes in with 6,300 sets. Oh! I don't have any fries to get back to, so here's a boost. Thank you, sir. Thank you, sir. Another bootleg reference.

Bootleg is coming in today. Thanks for the live boost, guys. Really, between the Boller boost and the live boosts, really put it over from a kind of a standard of fair episode, maybe even not doing so well, to a really strong episode. It just takes a few people, and it makes really all the difference. I do want to pull forward to Moon,

and I boost it in 520 sets, just because they're kind of having fun. You can tell timestamps on boosts after I made a comment that we added that meant picking up on the metadata when it's included. Well, guess I have to boost 420 sets at minute 69. That would be, I'll hold. whole new level of combo and a way to maybe do a discount combo is your boost amount combined

with the time code. But you'd have to give us a hint. We're not going to just notice it automatically. But that could be a way to, that could be a neat combination of numbers and tricks. Well, I mean, it was 420 SAT. So maybe it is at that minute 69. That's, right? I love it. Right. And you know, maybe you double it or something. I don't know. Okay. So there you go. Thank you everybody who support us. Of course, folks also just turned on on the SAT streaming and their podcasting 2.0 app that supported. and they stream stats as they listened to 18 people did that, stacking us, 29,186 sats this week. When you combine that with our boosters, we had a grand total of 342,015 Satoshi's.

Not too bad at all. Thank you everybody who supported the show with a boost. It is a value-for-value podcast. We put it out there right now with very limited sponsorship. The community keeps us going with a signal of value. We got value from the show. You send it back. One of the great things about the boost is an entirely free software.

stack. So we're doing it with no middleman, no PayPal who shut down our account or anything like that. Thank you everybody who supports the show. We oppose. And of course, thank you to our members as well. We really do appreciate you. Well, we got a few picks to cover before we get out of here. And Bud's Link has been making its way around again, so we wanted to pull it forward. It's an application that provides battery monitoring and other features to control different Bluetooth wearable audio devices including AirPods beats Sony audio wearables, Samsung

Galaxy Buds, and the nothing buds as well. And it looks like it's got a pretty slick UI for it as well. Part of God. Yeah, it's nice just because I think this is the kind of thing if you are getting started or maybe using the Linux desktop a little

more. You just want to have. Yeah, exactly. Coming from your MacBook. It feels well supported. It makes us feel like we know what we're doing. We're not stuck in, you know, 1995. I go back to as well, it's a good trend that we keep seeing these boutique, well-built, very purpose-focused apps because that was something that the Mac got 10, 20 years ago, and it started to slow down as the platform gets worse. And now isn't it interesting as those folks have come over to Linux and other people are now here to use these applications?

We're starting to see that type of stuff show up. And it is available as a flat pack, which is nice. You can get it going really quick and easy. It communicates with devices that are using the L2 cap or RFCOM sockets. RFCOM. That sounds old. And it is GPL3 as well. Now, Wes, I believe you found YAMLcast this week.

YAMLcast is a really neat tool. It uses a YAML description of a terminal screencast and converts it into an animated Jeff. Yeah, okay. So under the hood, there's the excellent VHS tool, which is, well, as they say, your CLI, home video recorder, but essentially write terminal gifts as code. So you can do, so you know, you want a gift of like, um, something happened. Sorry, animated Jeff of something happened in your terminal, right? Maybe to demo, like, changes to your product or some new feature you're adding or like a workflow

you're doing. Yeah. The cool AI thing you built, whatever. Oh yeah. Definitely got to show that off. Everybody cares. Well, I send you guys. I said you guys.

I care about your amount of like M.B.4's a gift system. I send it to you because you guys are the only people that care. Right. Only barely cares. Yeah, true. Okay, so great tool So you kind of like write it out But they have this custom tape format they call it

Which is cute and it's part of the name right But it's just like this custom Sort of ad hoc format for them Maybe you want to be able to use something More standard. So this is just sort of a YAML wrapper I see it's able to make It takes that dot tape Well no it lets you write YAML

And then it turns it into the dot tape And then uses VHS to render it So you can just go straight from YAML to No I follow you totally This is not NICH at all guys Not niche at all And so it's VHS that produces the Jeff.

That's correct. Oh my God. Okay, great. Thank you, Wes. What a great find. Uh-huh. And it sort of starts making sense if you go look at like a dot tape file

and then you look at, I mean, we know what a YAML file looks, but if you look at like the example dot YAML here in the repo, you can see, like, if you just want to be able to list, it looks a lot more like something you'd see in CI, which is like a list of bash commands or whatever you're running in YAML that you're kind of used to reading rather than happen to maybe adapt to an arbitrary format.

I'm looking forward to seeing a screencast from you soon, you know? I'm looking forward to it. I hope. So this next pick is pretty interesting. A entrepreneur-newing developer has managed to combine wine staging with a series of patches and DLL files that have been modified to get a reproducible recipe for running Adobe Lightroom Creative Cloud on Linux. You get the full Creative Cloud desktop app working under Wine.

You can sign into Creative Cloud. You can get Cloud-synced photo library photos. There are some dialogues that don't work, but the remove, heel tools, brush, mask, that type of stuff is working. There's a couple things that still aren't totally sorted. If you've got a 64-bit Linux distro with Wine 11.8 staging or newer, Nvidia, AMD, or Intel GPU,

as long as you've got Vulcan drivers working and a valid Creative Cloud subscription and about 10 gigs of free display. Of course. You can actually get Lightroom Creative Cloud now working on Linux. Pretty neat. This is so wild. The Rapper's MIT.

The state of wine these days is just something. else. The Adobe tools coming, I mean, the next unlock is going to be Photoshop and then Premiere and then eventually After Effects. And once we get to After Effects. Like what are the main things we hear about people not being able to move to Linux? One was gaming solved, right? For the most part. Second was Adobe Tools? Well, today that sounds like that solved. So we're running out of reasons. People can't use Linux. I mean, it would really be such an unlock. My mom has been using Photoshop since 1.0. She got access to the beta. I mean, she was involved in a Seattle graphics arts program when they called using Photoshop and things like that. They called it graphic arts. Yeah, yeah. And our Seattle Institute of Art had a program and they had a partnership with Adobe and all of that she was involved with. So to say that she's been using Photoshop for a little while is she's been using Photoshop for as long as any human that doesn't work at Adobe can possibly use Photoshop. And her entire career pipeline is around the Adobe tools. Built around those tools. Yeah.

So she doesn't, she continues year after year to have. She has to buy a new Mac, and she eventually gets forced because of the deprecation of the OS and the updates. And so for her, this would be such a massive, massive unlock. If I could move her to Linux, then all she has to maintain is a creative cloud subscription. Yeah, see, I guess Apple doesn't really offer anything like Ralph Forever, do they? No. And the thing that strikes me is if an individual with Claude Code can do this, then what's stopping Adobe from doing this directly?

and just kind of going the valve route, creating an Adobe proton-style compatibility layer that then they just design their applications to run on top of. We'll get on an Adobe, like let's start spending some money on this. It's also kind of wild, just another reminder of how much at this point Linux is really becoming a Windows runtime. Yeah, yeah, really. Which is, I think, the ultimate future of Windows, inevitably.

If only Microsoft would realize it. Yeah, they will one day, but it's going to take them a while. And, you know, I'm never going to use this. We're not going to use this. But it just seemed like a milestone to get a current creative cloud product, something like Lightroom, running on Linux. Go see what you can port to Linux today.

Well, not port, but, you know, hack together and wine today. Yeah, yeah, tell us. We want to know. All right, that is everything. Links to what we talked about will be in the show notes, which you can find at Linuxonplug.com. Or, of course, over at jupiterbroadcasting.com. Also, I want to encourage you

to make it a special event. make it a Tuesday on a Sunday. Join us. Sunday is at 10 a.m. Pacific, 1 p.m. Eastern. See you next week. Same bad time. Same bad station. Wes Payne, you want to tell them about some pro tips? Metadata, things like that, transcripts? Oh, yeah. Well, we've stuffed a whole bunch of extra features into our RSS feed. So maybe just go look at it, because it's nice to look at on its own.

Can I tell you, if we were like corporate owned, we'd be saying it's the agentic ready podcast. That's right. Because our chapters are in JSON. Our transcripts are plain text that any agent can ingest. Point your MCPs our way. Our podcast is more agentic-ready than your buddy's podcast, that's for sure. Cloud JSON chapters. Okay. That's cloud aided chapter.

We'll never do that again, all right? Thank you so much for tuning this week. Hybrid Cloud transcripts. Oh, yeah, thank you. We'll see you right back here next Tuesday. As in Sunday. Yes, it is Hammer Time.