685: Pool Party

You know how it goes. Someone on your team figures out a process, but six months later, a new hire asks, how do you actually do that? If that sounds familiar, check out scribe.how.com. 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 captured 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 onscreen 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.com. And mention Linux 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. 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, how we are pooling the storage around our LAN into one giant useful network drive. We have finally cracked it. Plus, our thoughts on the new Steam frame. Then we're going to round it out with some great boost picks and a lot more. So before we get to that, let's say time-appropriate greetings to our virtual lug. Hello, Mumble Room. Hello. Thank you so much. Hello, Brent. Strong room today. Nice. Nice having you guys here. Thank you. And of course, you can always join our Mumble Room. We've got details on our website. Hello, everybody watching live to making it a Tuesday on a Sunday. And good morning to our friends over at Defined Networking. Go meet Manage Nebula. Defined.net slash unplugged. Nebula is my network hero. When it just comes to something as foundational as a mesh network, something I'm going to build on top of for a long time, I want to own the whole stack. I really like Manage Nebula because it's a great way for you to get started. 100 hosts, absolutely free, and now they are throwing in a lighthouse as well. Plus, you can run your own lighthouse. So talk about something super robust, under your control, no big tech login. It's really great. And it is secure. It's fast. It's built to scale. It uses less network, less CPU than some of the competitors out there. It really is a different VPN experience. Go try it out. 100 hosts for free. You're really going to like it. For anything from a simple host to host to a large global mesh network, you're going to be impressed. Defined.net slash unplugged. And thank you to Defined for sponsoring the Unplugged program. Well, as you may have heard last week, we're officially going to Texas for Texas Linux Festival 2026. We hope you are or will consider doing it too. Yes, November 6th through the 7th. We'd love to see you there because we'd love to have a little meetup. And we are asking if anybody out there wants to throw some sats or fiat our way via the boost page. We are asking because the budget's tight. We are grateful for what we have, but a little bit more and we can maybe do a little meetup. But I think that'd be really great. Boost.jupiterbroadcasting.com And if you can be there, reach out. Let us know. We do have that Texas Linux Fest Matrix chat room. I think that's going to start warming up again. And there may even be an old Telegram group out there rattling around as well. Perhaps we should dig for that. But we'd love to see you there. And then we're going to try to put together some coverage. So if you can't make it, you get a taste of what we think is one of the really special fests out there. So that's also probably why we're mentioning it. So look forward to that coverage in the future in a way that only we do. Before we get into the storage situation that I'm very excited about, I think we should just mention for posterity that we now have the official details on Valve's Steam Frame. And it is available for pre-order starting at $10,000. It's a deal for that. Yeah, I wish. I wish that was a deal. I wish I was that rich. No, $1,059. Which is probably actually slightly cheaper than I was expecting. Of course, if you want to bump that storage up to terabyte, it's more like $1,300. Yeah, but the base is $256. But I was prepping myself for $1,300 being the starting price. Just with the supply chain, what it is. Right. That's just it. It's both more than you'd like and realistically maybe a little better than you hoped. Brent, have you ordered yours? Not quite yet. Not quite yet. I think Jeff's picking me up one though. Right, Jeff? Oh, okay. Yeah. Oh, I wish. You'd have pre-ordered one though, right? Which one? I did get in the reservation queue. I have not gotten email to actually purchase that device yet. You're in the line to get in line? No, I'm in the line to purchase, not the line to get in line to purchase. Oh, there's another one above that. Oh my gosh. It's a several queue system. So supplies are limited. There's a reservation queue, which means that you will get the chance to purchase once it's ready to be purchased, whenever that may be. And then there is the waitlist queue, which means that you're in the next batch essentially. So I'm in the reservation. I'm just waiting for that email to say, give me your money and I will give them my $1,300 for the one terabyte kit. Oh, that's exciting. Okay. Okay. I think the rest of us are in the NV queue at this point. Yeah. Yeah. There is another queue and you're right. The funny thing is, Jeff, is I mean, I assume you're going to use it to game, right? That's why you're getting it. Yeah. Yeah. Mainly gaming. I mean, it's Linux on your face, so I'm going to have as much fun as I can with this thing. I'm so excited for that. See, for me, that's the Linux on the face thing is the really big thing is a persistent workspace and workstation with your apps, your environment, and you just put it in your backpack and then you just slip it on your face. If you're on an airplane, you're in a car, you're sitting at home on the couch. To me, that just seems really powerful. And you combine it with something like a Bluetooth keyboard and you could really get some work done. I've been experimenting just as this came out, I decided I'll fire up the old MetaQuest again. And I've been experimenting. And there's a lot of Android apps that are pretty good now. There's pretty good terminal apps. There's a lot of the work apps you need and there's decent browsers. And you combine that with something like Sunshine and Moonlight. And it's a really compelling work environment. It's not quite there yet. And it's not as open as I would like. And of course, it's attached to Meta. So that's why I'm very excited from a productivity standpoint about the frame. Right. Having some other option out there that is even more just like exactly what you want. But I think if you're skeptical of what I'm saying, I invite you to check out Launch84. We'll put a link in the show notes. My last quest. I decided to kind of do like an exit interview with the Oculus and where they're at with their Horizon OS and all of that. And I really do think there is something there, especially if you spend a lot of your day in a terminal and a web browser and a couple of chat apps. You get a lot of work done that way. Yeah. It's nuts what these things are capable of. I mean, it's because it's the applications. It's the environment. It's your state in this headset. They just slip on and you just write back to work. It's not a whole desk. It's not a particular machine. Yeah. But it's going to be a while, I think. But people will get these and they'll start building. That'll be something. And Jeff, I hope you'll take a look at the desktop experience and give us a report when you do finally get it. Absolutely. We'll do so. Don't think it'll quite meet the needs for you. Yeah. Not yet. Not yet. I've heard you talk, yeah. But it could be built. Exactly. It's Linux on your face. And we already have some decent implementations with Plasma doing virtual desktops in terms of sending it to a different device. So with Wayland and everything like that, I don't see it being too long before we have a bunch of desktops around your head and having something similar than what you get with the Quest. I'm very excited. I cannot wait to get my hands on it. You know quite a bit about The Frame. Since it's just running SteamOS, can you boot straight to the desktop? It doesn't do the back and forth like the deck does. It boosts you into a virtual environment and you can have the desktop floating next to your main environment. So there's no rebooting it or logging out. It's all in one. It's pretty nice. So it'd be running like GameScope on one side using the resources for the environment that you'd play games in. And then the desktop could be rendered next to it. Yep. Just a floating window, your entire desktop, whatever size you want right next to your game or whatever other thing you got going on in the main environment. Do you know for the desktop environment, it would also do the foveal rendering, the render what I'm looking at in high resolution and the other stuff in lesser resolution? I don't think it will do the foveated rendering. I don't think it needs to. No, it won't need to. And the reason I say that because it's not rendering 3D graphics necessarily like it would for a game and foveated rendering versus foveated streaming are different. Now, I suspect that if you're streaming your desktop with specifically with Steam streaming from your main computer, it will be doing foveated streaming in that case. So one thing that I do want to try is streaming a very, very high resolution desktop, quote unquote, like, you know, like a 4K desktop from my desktop and then zooming in having a giant, one giant monitor with tiling. And then the foveated streaming should handle the performance on that. But even then, I don't think the performance is going to be a big deal with their dongle, the streaming dongle they've got going on. So it's cool. It's exciting. I mean, I really, really can't wait to get my hands on it. And the flexibility of being able to stream a large desktop or run the desktop directly on the device is going to be powerful. And then when people work on dedicated work environments for it, you know, I'm sure eventually they'll replace the entire big picture experience and it'll be, you know, some custom thing that, I mean, there's just going to be so much that happens with this, I imagine. It'll just take a little while and we have to get in our hands. The seeds are here just waiting for things to bloom. And I mean, what do you think of the price, Jeff? Was it just about where you expected? Was it a little bit more than you wanted to swallow? Yeah. I mean, I called 1300. I wasn't sure if that was going to be for the 256 or the one terabyte. I'm glad it's for the one terabyte. Micro SDs are unfortunately at 200 bucks for the one terabyte cards. So yeah, I'd splurge for that. It's obviously soldered on. So unlike the deck, you can't replace it. You know, it's where I expected it. Performance wise, we'll see. Everybody, you know, everybody compares it to the, to the meta quest three because it basically is a meta quest three in terms of hardware, in terms of battery life, in terms of displays. A lot of people have complained about the LCDs and this and that, but. Yeah. I don't know. You know, right now I can work two to four hours a week in it. Yeah. Okay. And given that you can just put a battery pack in your pocket, I don't think it's a big deal if you're gaming. My headset, my headset has a replaceable battery. And so I can, and I can swap it out while I'm live. I can just, and then I just plug it into USB-C and it charges. It's very comfortable. And Valve claims we'll be getting that for the frame later on as well. A whole new kind of head strap with a removable battery and a more steam index like headset. That's kind of why I was brought back to the quest and wanted to try this out recently because it's sort of, it's a lot in a lot of ways what I can expect from the frame. But some of the workflows are a little more mature around. It has an MCP and a CLI and all of that. So I think the frame will get there. And I just think it's like for travel, for work, for people that have small spaces they need to be in. It's going to be so great. And then it's going to game too, by the way. It's going to game. Yeah. Then you also have a really great gaming experience right around the corner. Well, congrats, Jeff. I hope the queue doesn't last too long. Maybe, you know, maybe you'll know something by Christmas. It's probably going to be around next year. No, don't say that. No, I hate it. I hope not. It'd be amazing if they arrived in time for the holidays. That'd be really great. Don't prove us wrong. Yeah. I'm really hoping it'll be like within this month. They've claimed they'll be shipping before summer. So, you know, if you're in the reservation, hopefully we'll get it soon. Very good. Thank you, Jeff. Thank you. All right. So, Brett Lee, I have talked about, I want to see if you can recall, but I have talked about the ability to take spare storage on the various computers around the LAN and try to pool it into one large disk. Like a virtual, mm-hmm. Mm-hmm. Especially with storage getting as expensive as it is, right? It's like it's getting expensive to go out there and buy hard drives. And if you've got machines sitting around with a terabyte, so this machine has like 200 gigs free here. This machine has two terabytes free over here, right? You know what I mean? You have a lot of stranded spare capacity, but, you know, being able to design and build from scratch the right sort of NAS or cluster is prohibitively expensive. And I have a feeling you know exactly how that feels, Brett. To me, this feels like your long-term plan to just sip away some of Wes and I's storage for your own works. I hadn't thought of that, but that is a great idea. You boys want to join the pool? What do we get? What do we get? One giant jellyfin chair. One giant distributed jellyfin chair. So this week, RustFS went 1.0. It's an open-source, S3-compatible, high-performance object storage system. And you can think it's all like MinIO, Ceph. There's some other. There's a lot. Garage. Yeah, there's some solutions out there. This one's nice because it's an open license. It just hit 1.0. They've put about two years of work into building something. Well, more than that, almost three years of work building something really impressive. Yeah, it's MIT, fresh implementation, Rust, of course. It supports more than just S3. It can also support WebDAV, FTPS, other things that... Yeah, some of those I think like SFTP is maybe compile time. Some of them you have to like turn on. They're not there by default, but that kind of stuff is growing. 1.0 also added some features that are just nice to have like erasure coding, self-healing nodes, rebalancing of the pool, bucket replication. Just stuff that's just quality of life you want to see before you really commit to something like this. And it landed in the 1.0. So that kind of caught our attention at that point. Yeah, it really seems like they're trying to signal, you know, okay, this can actually maybe be used now. If they didn't, you know, it wasn't rushed out after three months. They've spent, you know, more than two years sort of working and iterating, which makes sense because you really... You're going to build everything else on top of this. So it needs to work. So why object storage? Because I think when we say S3, people think Amazon. I mean, a lot of people probably think Amazon. So why would a person want to run object storage in their home lab or on their corporate LAN or something like that? What would be your pitch? Well, it turns out that you don't always need a file system. A file system comes with a lot of things, right? They kind of grew out of trying to be able to put stuff on disk and use it in a non sort of tape storage based way. We have random access. What can we build? You know, fat in the file allocation table. And obviously there's a lot of great stuff about having a file system. We love file systems here on the show. But these days, especially as we try to, you know, distribute, we have distributed computers, distributed systems. The rise of the cloud. It turned out that we still needed to store files, blobs, backups, media, artifacts, all kinds of stuff. But we didn't actually need all of the traditional properties of a file system. And so if you take off some of those, those end up being like fewer constraints that you need. And you can then optimize the core parts that you're doing, which is like, you know, being able to put and get something by like an object ID and maybe have it distributed, maybe have it replicated, all that kind of layer. You'll also find like a lot of the apps that you can self-host these days, they have an S3 storage option. It's becoming very, very common for an S3 storage backend. Yeah. And it's been that way for a while, right? Like S3 has been here for a long time now. We're even already running our production Nextcloud that way. Yep. We have been for a long time. Mm-hmm. Which is great because you don't really have to allocate space in the traditional way. It's great at storing blobs. You know, it's particularly good at that too. So there's some advantages if you have like images and media files and things like that, which we have a lot of. So it's been, and it's been fairly robust. Any problems we really have had have been more in the application implementation layer. Never really an issue with the S3 storage itself. And it can turn a pile of otherwise separate storage into one programmable storage service, which is what I was interested in. So I wanted to build a little test environment to see how viable this was. So I set up four headless VMs, each with their own 50 gigs of disk to store. And then I set up a fifth VM that acts as kind of like a client that runs Jellyfin and a front end to those headless NIC servers, the little VMs. Because what I was picturing is four or five systems around my LAN, 100 gigs over here, 50 gigs over here, a terabyte over there, that kind of setup. So I was just creating a couple of VMs, few VMs, and a master client server that sat in front of that. And that VM, that master server that ran Jellyfin, used Rclone to connect a RustFS over S3 and mount and expose what was a collection of my storage system. So that was the test setup. And the goal was, with this distributed object storage across multiple machines, would they behave well enough for something like a 4K Jellyfin file to stream? And this is an example, right, where Jellyfin isn't necessarily designed around object storage. So you use an Rclone to bridge the gap and make the S3 object storage dial back and show up as if it were part of the file system. Yeah. And one large disk, too. Not a bunch of individuals. I don't want to mess with that. Right. And I wanted to support one of them at least going offline because on a home LAN, that's very likely that one of these nodes is going to turn off and you don't want the whole thing to blow up. Right. If you sort of turn this into a big distributed system and you don't add redundancy, then you've built yourself a super brittle waiting to fail state. Which is going to be so frustrating when you go hit that play button. So that's what the goal was. I'll tell you how it went on the other side. Well, I don't mean to be this guy, but it was very easy to do this via Nix because RustFS ships an upstream flake. There's even one in Nix packages, too, I believe. That's so great. But if you want a more fancy cluster like we were testing out, then the upstream flake is the good path. It was really nice for just deploying a bunch of VMs and testing this to see if it was viable because you didn't have to have a custom installer, though they do offer one. Or I didn't even have to set up a container. It was really nice. So, yeah, I set up just some Nix-based systems using their flake, enabled the service, and I pointed it at the VM's persistent storage. And then, you know, something I've been doing more is using something to manage the credentials, too, which I really like. Instead of, like, baking them into the Nix store somewhere on your file system, I've been using, like, SOPS Nix. So that's been nice because there is some credential management you have to do. Yeah, it can use SystemD credentials, which I also really like as a way. So SystemD handles making sure that everything's provided securely. You don't have the fault, the downsides of using nvars and that kind of thing. So it's nice to see that support because it means it plays nicely with ship secrets your own sidecar way, use SOPS Nix or age Nix. It just plays right in with all of those. It does. And so that part was quite easy. That whole setup was surprisingly smooth. Where it went wrong was in my own assumptions. I did not properly read the documentation, I will admit. I skipped right. You know, because they always have, like, the splash page for, like, the documentation splash page. And then, like, all the stuff you want is usually, like, in a subsection because I just want to know the one thing, right? So I jumped down. And then later on when I went back and read, like, the front page and I scrolled down, they had this big warning there that you want to use at least a minimum of four storage nodes. Not the three that I had initially set up. Yeah, I was looking at it. It can technically go all the way down to two. Yeah. But four is, like, what they bless for production and it's what a bunch of their test infrastructure actually tests. It's sort of what they provide guidance on upgrade paths for. So it's really the, like, blessed setup. Or more. In my opinion, if that is a deal breaker for you, that's probably something to flag. It's not the end of the world, but you're probably not using a blessed setup. Now, for me, that was just one more VM. And I also, I do have plenty of systems on my LAN to do that. I probably have more than five. And I believe part of the reason is, like, you don't get necessarily more node. Like, it doesn't protect you from more node failures than three. But it does mean that as you're rebuilding things after one that you're in a more resilient setup. Yeah, it's really about redundancy. Yeah, operations. Yeah. So, and let's actually talk about that for a moment. So I took, in total, I had about 200 gigabytes of raw storage to work with. But when you consider the redundancy, what I got usable out of that 200 gigs was 91.7 gigabytes. Yeah, and you did, like, a racer coding sort of parity raid style setup. Yeah, and so I definitely ate it in the redundancy department. So that is very true. But the resiliency worked. I pulled the plug on a node and the system kept working. And then when you plug the node back in, it rebuilt and it caught up. And so, yes, it does mean you lose a lot of space. But if you have the space around your LAN and you have the nodes and you want to be able to survive a machine going out, it could be worth it. Yeah, and that's part of the neat stuff, right? It's because it can sort of distribute. You know, it can parity it up. It can make the various chunks, distribute those chunks across the cluster. And then you only need the right, depending on the level, that number of chunks again to be able to read from across the cluster to be able to get the file back. The thing I was happy about is the idea. So I wasn't sure how this would work exactly on the implementation side because since Jellyfin doesn't have native S3 support, I had to trick Jellyfin into using the S3 storage. So RustFS presented one distributed S3 pool. And then the R clone mount exposed all of that as one single mount point. Like I had like a slash mount Rust media. And then in Jellyfin, I presume you can just then do the normal media library configuration path. It just looks at it like it's a file system path. Nice. So that made it surprisingly easy. And I could get sequential read speeds around 360 megabytes to 430 megabytes. So I'm sorry, 360 megabytes to 430 megabytes per second across that pool, which I was pretty happy with. But there was definitely, though, a hidden cost in how I had R clone set up. And this is, I think, the second gotcha. First gotcha is you've got to have four nodes to really be stable. The second gotcha is you're going to need to be careful how you configure R clone. Because R clone is going to need to do a lot of caching. And depending on how you configure that, you can eat up a lot of disks. You can spend a lot of time caching. There is a sweet spot that I played with where I could kind of get that right. And I could store the cache in a specific place, which is also very useful. But that was an area where it took a one gigabyte client disk filled up and a two gigabyte upload. So it's just like it's the caching doubles. It is a problem. You need to plan for the caching if you're going to go with this setup. Obviously, the idea would be that Jellyfin just had native S3 support. But there's a lot of applications that don't. And so I do think you can make it work. Yeah, this is kind of a test for that side of things, right? There's plenty of applications that do, but for the things that don't, can it still work for those? Or are you sort of blocking yourself from using those apps? And there could be a better way to mount an S3 file system. Yeah, R clone was just sort of the thing at hand here for some quick tests. And we used it a bunch of places already. But I feel like, you know, what I was able to kind of validate in the testing is spare disks spread across multiple machines were turned into a usable media namespace that had redundancy when one of the machines went down that I could bring up Jellyfin and I could play a 4K movie across that. And that was really the goal. And it works. And it speeds the same. I mean, you have to try it again off, you know, not virtual links and the real network links and stuff. But seemingly, it speeds them might be totally workable. Yes. And I think if you use any kind of infrastructure management tool like Ansible or Nix or whatever, it's really easy to just deploy this across a few of your machines. Yeah, I think that stuck with me when trying it. It was like it was really easy to get going. Sometimes these distributed system tools need a lot of like bootstrapping and setup. And like it takes a while to get the whole system online. And that was not the case here. Yeah. Also, very friendly for agent deployments. If you did want to tie it into an OpenClaw or a Hermes agent, really straightforward. Between just the way you set it up and it also includes a comprehensive command line client. There's a lot there. Plus, these agents already understand S3 and all these things. Like, you could have the agents manage the file system, manage health checks. It's really pretty extensive what you could do if you wanted to. It just sort of plugs right into an agentic workflow if you wanted to have them manage your storage. If you're crazy enough to do that. I thought that was, you know, interesting that if something's coming out today, where does it sit in that? Looking forward in that way. And they're looking forward. They're looking forward. So RustFS hits version 1.0. And with the right set of tooling, there's a lot of ways to crack this. But with the right set of tooling, you can truly do storage pooling that supports failover on one of your devices. And that I was very pleased to see. It is worth noting it does not have the complete AWS S3 API. It doesn't have some of the same sort of strong consistency things around, like, write and then read ordering. In particular, there is no, like, consensus mechanism or raft or that kind of thing. Instead, they lean more to, like, eventual consistency. And so it works really well for blob workflows. In particular, where you're doing, like, a write once. You're not having multiple different actors all trying to write to the same key or use that for consensus or something. Instead, you want something where you're just, like, one process is going to dump it on there and then lots of stuff is going to be able to read. Because if you do have some sort of, like, network partition or, you know, a node goes offline, you can have... Basically, it has to get a lock, a quorum lock on the cluster, and then it can write. But if that lock expires, there's no code right now. There's a feature plan, but there's no code right now to, like, stop. So you can have two different writers both get success and then the later version wins. So one of them just silently has its writes not work anymore. So there are some things like that, but for most traditional, I think, S3-style workflows, it should work just great. I got a couple of questions that I can throw at you boys. Can we do a little thought experiment? I'm curious if, you know, I've been setting up redundant RAID on various hard drives in remote locations recently. And I'm just wondering, could this tool kind of supersede that? You don't need RAID, but you can use something like this for the redundancy. It has some cool modern redundancy in it. And maybe you don't need to necessarily have both? Yeah. I mean, it basically copies, or not copies, but has a very similar implementation of its read Solomon sort of erasure coding stuff to ZFS or, you know, other modern file systems like that. So you do get, you know, per block hashes and it checks that before it reads it out and serves it over the API and that kind of thing. So I think it is designed to store your files durably well. What kind of defense in depth and like, you know, what file systems that you want have under for the servers you're running anyway, I think is probably more the operational question. Okay, next question. When can we expect this in the kernel? Yeah, I don't think so. No? Okay. All right. And then last question. Did you mention what this is written in? I didn't really hear any mic or anything. Oh, did you? Did you catch the? Oh. It's MIT Licensed Rust. Oh. Very fancy. I think also, like, if you're still listening and you're kind of still asking yourself, why would I want object storage? Again, guys, why do I want object storage? This could be the project to play with to figure out why you might want object storage. This is what's really nice about this is this is very accessible to any of us to deploy this. And it's clean. It's simple. And it's pretty easy to get running on just about any distro you're likely to try it on. And it seems like they are trying to hit to this kind of case. Like, there's other projects that kind of aim for, like, geo-distributed versions of this. Or there's, you know, things like Ceph or, like, more file system semantics that are distributed. And this seems to be really targeting kind of exactly what you were hoping, which is be able to use a sort of diverse array of storage on something like a local or enterprise network and turn it into a single pool. Yeah. Now, there's probably a lot of ways you could solve that. And I would like to hear how you have solved it, boost.jupiterbroadcasting.com. I'm not married to this approach. I am kind of thinking about it. I was going to ask, are you keeping it? Yeah. I think, I mean, not this particular setup, but I think definitely at home. And possibly for a new self-hosted Jellyfin server I want to build in the future. That would be an interesting production use case. Yeah. Especially if you had some internet links in there. Uh-huh. Which maybe there's also a better tool for that, but I'm not sure. Boost.jupiterbroadcasting.com if you've got any feedback on that. But I do think this is what I'll be using for that moving forward. And so I'll eventually get some serious, not just testing in the lab, but serious hands-on with it. You know, probably in the pretty near future. I imagine with your new fancy state-of-the-art file system, you're going to be throwing some media files around of various machines. Is that your hope, Chris? You got new plans for this? You know, I am trying out a new box, and I would love some feedback on this from the audience. It's the Xiaomi TV Box S 3rd Gen 4K UHD. It's Google TV. It's got 32 gigs of storage, Wi-Fi 6, HDMI 2.0, $92 US on Amazon. That's not bad. It also takes SD card slots. Like, the Google TV is expensive. The Apple TV is expensive. They're all, you know, the Google TV is almost $200. This is under $100, and it's pretty powerful. Of course, it's the Google TV interface. And so the ask I wanted to put out there, if anybody else has this, and have they been able to reflash this or clean up the slop that's on this thing? Because it's loaded with the Google experience. Let me tell you, it's an experience. But for $92, I'm pretty impressed with the performance. It's not... And it has all the apps that you might need or want, right? It doesn't feel any worse or any better than my Shield, which has been kind of my benchmark for a long time. It feels probably, you know, about as fast. Because as we know, if you under-spec Android boxes, they can be a little slow. Did you consider any other boxes to solve this problem? Like maybe an old cell phone running Linux, maybe? Yeah, yep. I did, but it's for my daughters, and, you know, they just, you know, I don't want to... If it was a TV for me, I would probably actually just get like a Roku or, yeah, like use an Android. Use my Paizel, potentially. But for them, I wanted something that was a little more appliance. And I felt like that was a pretty good price at $92. Because when I look at the market, the only thing that gets close to that would be a Roku or a Fire Stick. And I would, in that situation, I would just prefer to have the Android applications. Because then you get a good Jellyfin app, you get the good Plex app, that kind of stuff. There's stuff like, you know, Mesh VPN apps and... That's the other big one is it has to be on the Mesh network. That was also a big deal, yeah. So I'm going to put a link to it in the show notes if you're curious. I don't have any affiliation. I don't even, it's not even an affiliate link. I don't make a dime. But it is working for me. And if you out there have one like this and you've been able to make the experience better without all the Google crap on there, give me your tips. Help us hack Chris's DVD. Yeah, let me make it, just make it better. That's all I want. I just want it to not be so basically a constant commercial. Because that's what it feels like. But think of the advertisers and think about Google. Yeah, I need to know about their new experiences. Yeah. And now it is time for Le Boos. Speaking of experiences, we got some support this week and we're going to start with Anonymous, who comes in as our baller booster at 75,000 sats. Hey, rich lobster! Ain't too shabby. All right, so Anonymous writes, recently got a Mac SE30. I'll try to figure out how to boost over Apple Talk. That is such a classic machine. I use the crap out of you. I don't know who Anonymous is, but I do know they're my new best friend. I do know that. That is going to be such a fun project. Also, tell me if you're not tempted to get TCPIP working on that thing somehow. Come on, right? Come on. And then how great would it be if you could get some sort of SSH client? That's probably not going to happen. It's just your thin client. That's what you're able to use to work on. I don't think the cryptography is going to happen on that old thing. You know, we're complaining about computer prices these days. Back then it was sold for $4,369 US dollars, which in today's dollars, $11,350. So not a bad deal? Yeah, that is the funny thing. If you look at 80s prices and value them by today's dollar, yeah, we were paying out the nose back then. That was one of the big controversies, really. So yeah. Wow, that is a great boost. Thank you very much, Anonymous. Hope you have fun. Josh comes in with $24,690. All right. I hoard that which your kind covet. Oh, it's a double batch of 1, 2, 3, 4, 5. Oh, you know what? So we might need to... You might be going a little too fast. He needs to simmer down over there because it's going to get ludicrous. You're going to have to go right to ludicrous speed. But mostly value. We just got a rocket ship and a thanks. Well, thank you very much. Going to put that towards the Texas budget. Greg Vilar came in again this week with a row of Megaducks, 22,222 sets. Megaducks. I'm fine. He's not looking up for old McDuck. Greg also just sent the value. No message on that one. So thanks, Greg. Thanks, Greg. Appreciate you. Get off my TTY, 42 also came in with a row of Megaducks. This old duck still got it. Okay. In the pre-show, you celebrated more Home Assistant config removing from YAML. I can't reconcile that with the promotion of declarative systems, coughed Nix. It seems like HA moving to click ops. Did we celebrate that? I don't know if that's what you were trying to say. Yeah. I am a little torn on this particular issue myself. Like because this has recently just been changed with Modbus, which before had to be all done with YAML, which was a massive pain in the butt in the before time. And now it's funny that they're getting rid of YAML when it's easier than ever to generate YAML. That is, there is some irony there. Home Assistant EA continues, Home Assistant YAML is easier to grasp than Nix and it's easy for an agent to write. What am I missing? I don't think you're missing anything. I suppose agents are at least very good with the API. There is that. I don't, I don't know. I don't, I don't think, and I really respect the Home Assistant team. I don't think they've wrapped their head around the fact that a lot of people are just going to use that as an agent API now. And they're not going to build dashboards. Like they're doing all this work around making dashboards really pretty and easy now. I have a Home Assistant instance. I've never even loaded the dashboard. I've never even logged into it. I just have the agent manage all of it. And if I want to know what the temperature is, I ask the agent. And if I want to know if the light's on or off, I ask the agent. Because it's a chicken coop. You know? It's like, I don't think they've wrapped their head around where this is going for them yet. And perhaps what you're witnessing here get off my TTY is sort of a result of that. Because I kind of agree with you. Although I do like, at the same time, like to see it easier for people to get Modbus working. I'm a complicated guy. Love Down Babylon comes in with 21,337 sets. It's Lurks! Hey, hey Chris. Sir Lurks a lot here. You, Adam Curry, and Dave Jones, and others inspired me to launch my own value-for-value podcast, Love Down Babylon. First episode, out now. A multifaceted adventure through personal healing, technology, platforms, big ideas, world problems, and hope for our future. Here is Elite Satoshi Boost as thanks for letting me sat-svertize my show. Fun will now commence. Oh, congrats, Lurks. Go check it out. Great name. Love Down Babylon. Go search for it on the index. Well, Gene Bean is here with a row of ducks. Ooh. Look at all those ducks. Gene Bean says, Bravo, Tech Dev. Looking forward to hearing the Texas Linux Fest coverage. Yes. Aw. Thank you again to Tech Dev for becoming the elite sponsor of our trip to Texas Linux Fest. That was amazing. This is the way. Gene is here with another row of ducks. You make me want to be a better man. You mentioned Discord, and it reminded me, I'd like to get your take on Armada. The headline of their website is, Encrypted Open Source Discord Alternative. What do you think? This is one I don't think I've seen before, actually. I don't think I have. They have a web client. I like that. Zap Store. Google Play. Okay. Lots of pink. Lots of pink. Lots of pink. Animated Jeff support. You know, it's a little frustrating that the UI already looks better than Element. Like they've already surpassed Element in the UI. One of their top features here that they're promoting is unbannable communities. Well, that is important. Self-hostable at every level. You don't want, because, you know, pretty soon those Canadians are going to get banned from communities, so you don't want that to happen. Self-hostable at every level, you say, huh? Interesting. Okay. Should we put that on our list to check out? Oh, and it is interoperable? With other Nostra apps? I feel like the only thing is, Gene, is we really seem to be at like a platform fatigue. Could you imagine us asking the audience? I like this called Soapbox. Could you imagine us asking the audience to sign up for another chat platform? Some of you would do it. It's a lot. And at the same time, I'd have to be asking myself, why don't we just go back to IRC? That would be, that's now its competitor in my mind, because IRC was so simple. It's easy. Right. There's a risk now that if we make any move, it will just default. Go back to IRC. I miss you, IRC. But thank you, Gene. I think that is worth looking at. Well, Gene sent a couple more messages. Really enjoyed the great Jellyfin information. And FYI, stock Jellyfin is finally in the default Samsung Tizen store. He says he's looking forward to the tutorial on ARM that we promised. Did we promise something like that? Uh-oh. He says, I tried it a while, but unsuccessful. So hopefully it shows what I missed. And last one here. Hold on a second. I feel like this last couple of boos, like you guys are like, I don't remember saying anything about the home assistant YAM thing. I don't remember saying anything about ARM stuff. It's possible we've also forgotten a lot, because that's usually true. It could be the B team. So you just need to give us more context so we can remember what we've promised. Yeah, it must have been Jeff's project. Okay, continue on. And provide proof, too, next time. Thanks. Yeah. Well, it's fun to talk to Jeff. Gene's last row of ducks here. Here's a pick for you. Absorb is a new-to-me audiobookshelf client that works on both iOS and Android that seems to implement the entire API, even the admin bits. Oh, thank you, Gene. Thank you. Absorb. The dev has a buy-me-a-coffee link, and I encourage anyone as impressed as I am to consider using that button. 91% Dart, 3% Kotlin. This looks really good, and I like the stats page, too. I am a huge audiobookshelf user. Use it every night, and use it on long drives. Love a new client. I'll give you a new client that I've been using this week. A little client called Listen. Listen, at least on Android. I don't know if it's available in other platforms, but I've been using that this week, because same as you, Chris, it's your fault. Listen to audiobooks every night, and just a nicer interface. A little simpler. I like it. All right. Listen. Ironically, Google's original podcast app was called Listen. This is L-I-S-S-E-N. No T's around here. All right. Thank you, Gene. Appreciate you. That's some good boosting right there. Daja's here with 4,747 Satushers. Oh, my God. This drawer is filled with broloads. Plus one for Wizard. It also works with audiobookshelf. Oh. Did we promise that, too? No, no, but that's something I'm going to look into. Brent actually already used mine to create an account on my audiobook server when I recommended. I sure did. ADHD is awesome. Super handy for when you want to invite people to all your platforms at once with a single invite code. It also does ROM. Yes, I have ROM. That's impressive. I love ROM. R-O-M-M. I want to talk about it on the show all the time, but I've done an episode on it, and we try not to overdo the gaming stuff, so I haven't mentioned it again, but that's my excuse. R-O-M-M. Love it. It's unbelievable what you can do in a web browser. That is great. Thank you, Daja. Derivation dingus boosts in with a double row of ducks. I like you. You're a hot ticket. It's a personal assistant agent to it. It can display the same data as to-do lists, Kanban, Gantt charts, et cetera. I really like it. If you guys try it, I'd love to hear how you end up using it. I might now that I got the plus one on that. Mm-hmm. That's some good signal there. Is that a promise you're giving out there? I should be careful. I want to try it as well, so. I do too. I do too. All right, Brent, get on it. And I think we got a little more signal here. Oh, okay. I thought Herder was pretty nice, but I love Zelage and really missed some of its features when using Herder. Yeah. But then I found ZJ-Radar. It's extremely similar to Herder, but it's just a Zelage plugin. I set up the alias Zerder, which launches Zelage straight into the radar profile. I honestly haven't launched Herder since, since I got it installed. I highly recommend. There are people listening that have no idea what any of those words mean, but I am very grateful. I've been wondering about this because I have seen this Zelage plugin going around. It sets up a little sidebar, and then you have, like, it's aware of what your AI agent, you know, whatever you're using, your open codes, your clods, whatever, and it's kind of a context aware of when they need your help and stuff like that. But it gives that capability to Zelage. Yeah. To the terminal multiplexer maybe you already use in love. Yeah. Interesting. Because I agree. There's things about Herder I like, but there's things about Herder that feel like it's not designed for a guy like me. Interesting to see, too, because I have used a lot of Zelage, but I have not yet tried a Zelage plugin. So I'm kind of curious on that experience. Ha, ha, ha. Fascinating. Thank you, Dingus. Appreciate you. Well, we got a digital farmer here with a rope ducks. Just has boosting from my new LND node running Nix OS. We got it. That's exciting. Nice work. Well done. Keep on farming. Thank you, sir. MRAM comes in with one, nope, well, I guess one, one, one, one, one sats. There you go. That's sort of what I was trying to find. One gold blooms out of a possible ten gold blooms. Yeah. Okay, there you go. Because I'm the wind. MRAM writes, hi, I listened to episode 331. Whoa, that is a time travel and boost. You are all so exalted over your Arch server. I enjoyed it big time. Wow, we've had the Arch server going since that long. Oh my goodness. No way. That's crazy. You can't use Arch as a server, though. No, I can't. I like the AI stuff, too. Wish I understood it better, though. Keep up the good work. Well, keep on listening, MRAM. You just might. Thank you for the boost. Wow. I love time travel boost. I had no... Dang. Dang. That is a typo, though, right? That little server through every summer, every winter, out in that garage. The cottonwood disaster. Cottonwood, cats, rats probably. I don't know. Jeez. That, I am... I just feel... I'm so proud of that little server, guys. I'm so proud of it. A dude trying stuff boosts in 11,670 cents. I don't understand what the heck is going on here. Woo! See you at Texas Linux Fest. Ah, yes. Had to work the dates into this boost. Oh! Clever. Give me a holler when this goes south. November 6th and 7th. Yep. Awesome, a dude. Looking forward to it. Well, Pixie's here with a $5 fiat boost. That's not possible. Nothing can do that. Putting some money up to get a chance to read Chris's Nix or SOPS config for the media stack. Plus, Jellar experience would be appreciated. I was just looking at that project. I completely blanked on that. I will try to get a version of that. Is this another promise you made? That's a good... I think so. Oh, my gosh. This is why we're keeping us honest. Keeping us transparent and honest. Keeping us accountable. We intend to be. We're just, you know, really fallible. Well, I got really excited with storage this week. I will be honest. Once I saw that 1.0, I'm like, can I build it? Can I build it? I got a little carried away. Thank you, Pixie, though. I need to export all that anyway. So as I sort of just decommission that test box, I'll try to capture that. And then I'll put a link in the show notes when we got it. Sam H. comes in with a free member boost. A time travel boost from 682. I also use Pi with my framework desktop. A simple Pi extension written by Pi helps select models from LlamaSwap and get the right reasoning level options. Prompt processing is slow on this GPU, so the smaller the system prompt of Pi helps things get started faster. Very nice. That makes a lot of sense. Another case where a little bit of a harness can be really helpful for that kind of stuff. Anonymous comes in with a free member boost just to say, great job, guys. I learned something new on every episode. And this one was for 684. I hope that is true. You're ready for this jelly. We definitely try, so I hope that is true. That is one of our goals for sure. Going for the value. Thank you, everybody who supported the show. Members, those of you who boosted below the 2,000 sat cutoff, Mayhem came in at 1,999 sats with a Zubamoof. Okay. When you bring it all together, we had 16 of you stream sats, and collectively, you all stacked 29,371 satoshis. Thank you very much, sat streamers. When you bring it together with our fiat boosts and our regular boosts and average it all at the current sat market value, our total this week was 254,069 satoshis. Not too bad for an average week, but we are trying to get that up a little bit on the average to put that budget towards our Texas trip. Anything there would help. Boost.Jupiter Broadcasting for that, but we are grateful. And, of course, thank you to our members who do put their support on autopilot. All right. Yeah. So thank you, everybody. We really do appreciate it. Of course, if you're a member, too, not only do you support the show automatically, but you do get a free member boost. So you can always send in that. And we like it. Boost.Jupiter Broadcasting.com for that. Thank you. And we appreciate everybody who supports the show. We are still sort of struggling in the sponsorship department, still relying on the dynamic ads and all of that. And so it does make a big difference. And, yeah. You know what I'm realizing? I didn't put my picks in the doc. I was wondering. Hmm. I wonder if I could retrofix that. I wonder if I could retrofix that. Because I really liked those picks. I was going to say I got two backup picks. You do? Yeah. Gosh, mine were really good. And I never, you know, it's because I don't actually write in the web browser. I write. Yeah. You have a working document. That's right. All right. Well, what are your backup picks? And then I'll save mine for next week. Well, I don't know. I was just trying to, I wasn't sure if we had picks. So I was just trying to see what had been, you know, put in our pick. Yeah. Yeah. There's a few options. I could probably find it in there. This is my bad. I didn't put it in there. Okay. Well, someone linked in our chat room, one of our chat rooms earlier today. Okay. Flakelit. Run systemd services from Nix flakes and update them independently of the host system. That is such a West pick. Mm-hmm. That's a strong West pick. Normally, a Nix OS host has to be rebuilt to update one of its services. Sure. Flakelit takes the same approach as how the virtualization OCI container stuff does, but with flakes instead of container images. The host only declares which flake to run and which settings to pass. The machine itself resolves the flake, evaluates and builds it, and switches the systemd units. There's no images, and there's no registry. The units run straight out of the host's Nix store. Well, that's not too bad. Now what we need is a pick that checks our show notes to find out if we didn't put our picks in there. You got a sneaky pick you want to throw in, Brent? Well, I'm going to throw an opportunity here at Jeff Lee seems to have a pick. Oh, PJ pick. All right. Yeah, a sneaky little pick. It's more for the Android side of things, but it involves the good old AI stuff. There's an Android app called Chompus. It is a calorie counting slash health and food counting application, and it can use your local AI. It can use whatever AI you point at. It can use Digni or anything, and it has a nice big registry built in for various foods as well. So you can do things like take a photo of the plate that's in front of your face with some kind of idea, like a hand or a coin, and describe what you're going to eat. And it does a pretty dang good job at calling your AI and saying, okay, you're consuming this many calories. You're consuming this much fat, this much protein, et cetera, et cetera, et cetera. Knock it off. I would have mine yell at me about it. You know better. Stop it. There is a coach thing as well, so you can kind of chat with your AI and do that as well. And I don't try it. I don't use it, but I have tried like, hey, what did last week look like? And it did pull in that data and give me information about the last week of food. But I've been using it a lot with my local hosted AI stuff. I think I'm using Gemma 4 specifically, and it's been working pretty good. All right. We'll put a link to that in the notes, and then I'll toss in one last little extra bony pick here. It's called FF Dash. It's a Rust app, and it's a terminal UI for batch, AV1, and VP9 video encoding. I am becoming a big fan of AV1. And so it'll use hardware acceleration if you got it. The TUI gives you nice real-time progress monitoring, full control over the settings, and it's made to work as a visible encoding dashboard over SSH. Yeah, this looks really nice. I haven't tried it yet, but it's in the queue to try for sure. I love the look of it. And for me, this was a hard requirement that it supported QuickSync, and it does. It supports QuickSync, and it'll also support the NVIDIA stuff if you got it, or the AMD. Nice. Yep. Doesn't discriminate on platform. You got your VA API and your NVEX as well as the QSVs, Wes, which is what you want to see in there. So again, it's called FF Dash, and it is MIT licensed, and it is written in Rust. So there you go. Not the picks I intended, but they were bangers, and they're worth saving. So I'll save them for you. And instead, you got a bunch of nice surprise picks that we didn't even know. Yeah, so stay tuned next week for our real picks. When's the last time we've even left something out like that in the talk? It doesn't happen often. No, it does not. But I'm going to blame Brent. He's our QC guy. Brent's QC guy. That's all right. Now, I will blame you out there for not visiting us on a Sunday and making it a Tuesday because you're missing out. It's a life experience. Well, okay. Maybe I'm overselling it a little bit, but it is a fun experience. You can make it a Tuesday on a Sunday over at jblive.tv, 10 a.m. Pacific, 1 p.m. Eastern. See you next week. Same bad time, same bad station. I mean, we got all kinds of ways to experience it, too. We got audio. We got video. We got, you know, we got the whole mumble experience. You can just be in the chat room if you want. We're on the, like, proprietary platforms if that's more your speed. That's true. Got them over on the tubes. And then we have a handcrafted RSS feed. Beautifully designed. Full of extra stuff like? JSON cloud chapters. Boom. Boom. SVT, SRT, VTT. I made up a fake file format there for a second. Transcripts, though. That's right. Transcripts. Transcripts. That's good. You like those. MPEG layer 4 and layer 3. If we got, if, you know what? Sometimes we might be showing a visual of what we're talking about and you don't even know it because we're so good at the audio show. But it might be in there if you watch the video. A little extra. Yeah. And it's all, of course, linked over at linuxunplugged.com. And this was episode 685. So you could just go to linuxunplugged.com slash 685. Then when you're all done, head on over to Jupiter Broadcasting. Check out the launch for my last quest this week in Bitcoin and others coming soon over there at jupiterbroadcasting.com. But that's it for us. Thank you so much for joining us on this week's episode of Your Unplugged Program. And I hope to see you next Tuesday, which is what we call Sunday. It's a Tuesday on a Sunday. See you next week. See you next week. See you next week.