LINUX Unplugged - 688: Eyes on Deck

Episode Date: October 11, 2026

We've turned our Steam Deck into a security monitoring system; Monitor your property, identify people, pets, and vehicles with detailed daily intelligence briefings. All local.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.Hostinger: Create your website today go to Hostinger.com/UNPLUGGED and use code UNPLUGGED for 10% off now.Support LINUX UnpluggedLinks:Web Boost — Send us a boost via sats or USDJupiter Uplink💥 Gets Sats Quick and Easy with Strike📻 LINUX Unplugged on Fountain.FMConnecTen Internet: promo Jupiter35YouTube: Clanker Therapy 2: What We Have Been BuildingAudio: Clanker Therapy 2: What We Have Been BuildingBootleg: Clanker Therapy 2: What We Have Been Building | Members StreamClanker Therapy 3: What Have You Been Building? - Livestream Oct 16, 2026Texas Linux Festival 2026Frigate NVRblakeblackshear/frigate — NVR with realtime local object detection for IP camerasFrigate DocumentationFrigate ReleasesFrigate Object Detector ConfigurationUltralytics YOLO11 Documentationllama.cppOllamaUltralytics YOLO11ultralytics/yolo11 - GitHub — Ultralytics YOLO11 discovery and quickstart for detection, segmentation, classification, pose, OBB, and tracking.cage-kiosk/cage: A Wayland kioskTomBursch/kitchenowl - GitHub — KitchenOwl is a self-hosted grocery list and recipe manager. The backend is made with Flask and the frontend with Flutter. Easily add items to your shopping list before you go shopping. You can also create recipes and add items based on what you want to cook.dahjah/muse-vm-hacks — Basic repo with what you need to go from a bare chat interface with your muse agent to an ssh-accessible VM with the nix package manager and full tailscale access (muse ships with a stipped down versionPick: wtfi2 — Live, visual network-path diagnostic that pinpoints exactly where your Wi-Fi connection dies — and how to fix it. Rust + ratatui TUI.Pick: SpaceScan — Fast disk-usage explorer for Linux: sunburst chart, keyboard-driven table, and text/CSV/JSON reports. Accurate like du, safe cleanup, 10 languages.Pick: Vermouth — A native C++/Qt Linux launcher for Windows, Steam, GOG and retro games. 2.3 MB installed. Zero bloat. KDE-first.Pick: RAWmakase — 📸 Free, fast, Lightroom-compatible RAW photo editor for Linux, macOS, and Windows.ArtCraft — Open Source Tools and Models for Makers, Builders, and Creatives (aka open source Adobe clones)storytold/lightcraft - GitHub — An open-source non-destructive photo editor written by scratch in pure Rust

Berkeley Life's nitric oxide support has kind of become my secret weapon for road trip durability. I just got back from another trip two full days on the road, hours of driving, hauling gear, loading and unloading weird meals I shouldn't be eating, long stretches sitting in the car, and then getting out immediately being useful and then get back in the car. And you know what? I felt great. It's been the pattern for me while using Berkeley Life. I have felt like I've got more staying power through these long demanding travel days. Oxide is something your body already produces naturally, and it plays an important role in circulation, cardiovascular wellness, energy, and just overall vitality. Berkeley Life's nitric oxide support is clinically validated with peer-reviewed research showing a 378% increase in nitrate levels and a 28% improvement in circulation. And one of the coolest things I think about Berkeley Life is that you can actually check your own nitric oxide levels at home. There, no starter kit includes a supplement and the test strips. so you can test your levels before you start and then keep checking them along the way.

That combination really works for me. Publish science, something I can measure myself, and a product that has now been with me through multiple demanding trips. Go to BerkeleyLife.com and use the promo code Linux15 to get 15 to get first order. That's B-E-R-K-E-L-E-Y-L-I-F-E dot com. Use the code Linux 15 to get started. These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, or cure, or prevent any disease. 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, just how smart can your security cameras get? Even the cheap ones, if you kept everything local and didn't rely on the cloud. I took my steam deck on a little bit of a security duty tour, teaching it how to look at what happened around the property and the farm and then report back all locally tell you about that then we're going to read it out with some great picks and boosts some feedback and a lot more it's going to be

a lot of fun so before we get to all that you know what we got to do we have to say time approves to our virtual lug hello mumble room hello hello hello hello hello from a lot they're still going that's a good crew we got in there today and so hello up there in the quiet listening and everybody watching live it's giving it a live vibe we appreciate that. And go say good morning to our friends over at Defined Networking. Defined.net slash unplugged. Go meet Managed Nebula from Define Networking to decentralized VPN built on the open source nebula platform that we love. Optimize for speed, simplicity, and industry leading security. When it comes to your core infrastructure, I want to own my network. That's why I love

Manage Nebula from Define networking. It creates a fast, decentralized mesh where my machines connect directly instead of routing everything through a central server or control plane. I can run my own lighthouses, use their managed lighthouses, real control, and you don't sacrifice performance. Manage Nebula makes it super easy to deploy and get started. Get started with 100 hosts for free. No credit card required. Go to define.net slash unplugged and redefine your VPN experience. Define.net slash unplugged. Thank you to Define. Keep an eye out on their blog, too. More to come. Well, we're letting you know ahead of time, Clanker Therapy 3 this Friday, October 16th, 12 p.m. Pacific, 3 p.m. Eastern, 1900 UTC, 7 p.m.

Over there in the UTCs. You're so thoughtful with your time zones. Mm-hmm. So for this, what have you been building? Tell us what your setup is. Ask us any questions. We'll play your uplink. So go to uplink.jupydrobycasting.com. Choose what have you been building, which I think is the default. And leave us a voice note and we'll play those and answer them. Plus, we'll cover some new projects. I have got some really fun stuff to share.

and we may even have a random Brent show up and tell us about his current workflow, which both Wes and I have been dying to hear about. So it's going to be a good clanker therapy. Details are at meetup.com slash jup broadcasting. Meetup page not required for the stream, just for your information. It'll be this Friday.

And we have the details at jupiterbroadcasting.com slash calendar. The thing to take away is uplink.jubterbroadcasting.com. Leave us your voice note so we can answer those on Friday's stream. And then last but not least, Brent wanted to make sure everybody is aware Texas Linux Fest is just around the corner. Coming right up. Texas Linux Festival, 2026, Brentley. He's going to be there.

We're going to be there. Looking forward to it. I've been on like my slowly ramp up the meat diet in preparation for, you know, what is usually an overload. So thank you, everyone. Yes, I'm very much looking forward to it. We are in the early wrapping our head around meetup phase, but I think we'll probably have a meetup announced pretty soon for when, you know, we're all. down there. And again, just super thankful for Tech Dev for making it possible. No kidding.

We've had some more support come in this week via boost. We'll talk about that later in the show too. But November 6th through the 7th in Austin, Texas, as we record, 25 days, 13 hours, and 48 minutes until it's time. And then we gotta go, boys. We gotta go. So we're gonna be flying this time. There won't be a race, but it's still gonna be a lot of.

Well, I mean, there still is a race. Who gets that first? Wait, you booked my flight. Oh, no. I don't know. You might beat us. The wife booked him, so I don't actually know. I don't even know when my flight is. I paid her off, so it's all good. I wanted to talk about this for a long time.

We've maybe touched on it briefly. I am a big fan of the Frigate NVR project, an open source network video recorder for IP cameras. And it's really easy to get set up, and you can use it with all types of different cameras if you use it go-between like go-to RTC. So I've got multi-vender cameras, Wi-Fi, Ethernet, all different kinds that come into Frigit. And it talks to your cameras over RTSP or ONIV or OMVIF.

And it's really great. It is, it gives you a bird's eye dashboard where you can see what's going on around your home. You can have a record certain events and play them back. It can communicate with Home Assistant so it can tag people and pass that along to Home Assistant. So Home Assistant can take actions like alerting and automation. It has some really great built-in support for detecting objects, which is really cool. It supports zones and mass. You can zone out certain things.

And it has a great UI for going back and finding the things that happened if you're looking for like a particular incident. It seemed just like a great open source project, too. I think it started in 2019. But it's now at the point, right, where there's a person who didn't start it who has more commits than the founder. Both are still very active. So it's just one of the, you know, like it's grown into a community of folks who obviously use it, but are also there maintaining it and adding new features and sharing data and all that. And I imagine, you know, there's a lot of us out there.

You would like some home security. You would like maybe monitoring what's going on with your pets. You know, like I have this setup for my chicken coop, for example, because I just want to know. And it actually has been really great. There has been times where we can't find a chicken. And so you can go back to the Frickett and pull a section of video to find out. Well, was the chicken in the hen house last night? Did the chicken make it in last night?

Oh, yep, there's blue. She's roosting. Okay, so she's just out and about right now. She's not that lost. You know, just sort of peace of minds. And then you can, there's an API here, right? So if you have an agent sitting in front of this, you can say, go get me 30 minutes starting at 1 p.m.

or something like that. Give me an active chicken count. Yeah, that kind of thing. So Frigot is really great, and there's a lot you can do with it out of the box. And I just think we probably haven't talked about it much because it is kind of a niche use case, but anybody that has some IP cameras that doesn't want to rely on the cloud or pay for a third party or have a third party

get access to that information or if you already have home assistant and you want to integrate with that and add some automations, you can also do MQTT support. It's really pretty powerful. It's got a very nice architecture too, right? Like it really prioritizes latency.

There's ring buffers and shared memory going on to keep things efficient and trying to, they do some optimization around memory management to prevent things like accidental out of memory, things kicking in. It's quite impressive. And then, of course, like, you know,

there's like the actual data channel versus like the metadata emit fence that get written out. Yeah. And if you get it with something like a USB coral accelerator or another type of coral accelerator, it's pretty impressive the hardware you can run it on. I run it on my O'Droid H3 at home that's doing 15 other things, including jellyfin and media management and all kinds of other activities. And Friggott's just one of the things it does.

Yeah, it's not totally destroying your system. Yeah. So it's not like even a dedicated system. It's just, and it's not a particularly powerful box. I do have a coral accelerator, which, you know, does the detection lift. Right, and that's where a lot of that architecture comes around, right? It's making it's like, how do you make it really cheap to do the quick evaluation? And then for the things that actually trigger the first layer, you can then do more heavier filtering.

So you don't have to heavily filter every single frame. The coral thing, I think I just have to mention, I really think it, this went from useful to actually real-time event-driven, like notifications about somebody is in the drive. driveway with this license plate or stuff you can actually act on right away. So adding a coral, I think, really made it much, much more useful. And then you essentially get cheap motion watching and cheap event watching where the CPU doesn't get that busy and it offloads all of that to that processing device. Coral runs a quantized neural net and can do like bounding boxes like, this is a dog and it puts a little box around the dog.

It can label, so it'll say this is 80% confidence that this is a dog. So you'll get an image with a bounding box around the dog. This is 80% confidence that I've identified a dog here. And that is really useful because all of that is happening locally within milliseconds. So you can get a real alert sent to your phone through Home Assistant as it's happening. You played around with the coral last night. Yeah, I was because it's a neat little bit of technology. There are limitations.

Like you can't really do anything that needs attention. So like LLMs aren't going to work there. But there's a bunch of audio applications because that's even sort of lower bandwidth all the times than doing something like frame processing. Hmm. Hmm. One limitation or concern I had was that it uses a compiler from Google that kind of hasn't really been developed much since, like, 2021 and is closed source. Interesting.

So I spent a while having an open code bang against that. Yeah. And it was able to build a little compiler for itself that seems to be basically byte identical to what Google's been doing. No kidding. Mm-hmm. Including building a YOLO model. No. Yeah.

That runs on the coral? Yeah. Right now it's kind of basic and it's still trying to optimize it. But it was able to at least get it compiled after some surgery through its compiler. You are blowing my mind.

So you built your own compiler and now you're getting the YOLO V9 model? Is that what it is? Because the YOLOO models are pretty small. Yeah. I tried a few of them and it got various percentages

because it was basically measuring how much would actually stay located running on the TPU and not get pushed back to the CPU. Yeah, sure, sure. Yeah. Huh. That is interesting. Basically trying to figure out, like, what more could you do? Audio detection? Just with, like, better tech on the side we control since it's just kind of been frozen for five years.

Like glass breaking, if you were using this in a security context, would be kind of interesting. Dogs, barking. It sounds like there's already stuff out there for identifying bird calls, so that's pretty neat. You could use it for bird detection. Yeah. That's fun. I wonder if you could do it with the amount of birds around my place. So, and then there are some things you can offload.

There's other, obviously, if you've got like an Nvidia GPU, there's other ways you can do offloading. I, I didn't, I have a coral, and I don't, I didn't create my own custom compiler, but I did want to do more with the images than Frigget has built in. So Frigget will do license play, it'll do human. It could even do like package detection to a degree and all that with the coral and all locally and just within milliseconds. But it can't, it can't really describe a scene. it can't monitor events throughout the day and kind of compile together a more comprehensive. This happened, then this happened, then this happened. But an LLM could, but you'd have to be a maniac to send your home footage up to a cloud LLM like, you know, I don't know, Google Gemini or Glawto.

Yeah, you'd really want them to have vision now. They're just accurately describe everything going on. Not only would that be... Not only would that be impractical from a bandwidth standpoint for what I'm thinking of. Right. But it'd just be an absolute privacy nightmare. And then I'm sure it would consume, because you're just doing vision constantly, I'm sure it would consume a fair amount of tokens.

It would get expensive. All of the dynamics of that change, if you could do all of it locally, and if you essentially can run image recognition locally unmetered. But I don't really have any hardware that's capable of doing that. Well, until I realized, I think it was tech dev that reminded me, Jeff's probably pointed out a thousand times, I've got a steam deck sitting right over here. a sneaky hidden GPU

with a decent CPU and a decent amount of RAM and it's like it's a shared memory model with it's already optimized for this kind of thing yes it's not a lot of memory

but it's continuous shared memory which these LMs love yeah what are we working with on that Jeff do you remember I think it's 16 gigs it's not a ton that's but it's you can upgrade it

too I think yeah it's 16 gigs you can't upgrade the RAM but you can change how much a GPU gets in the BIOS. So you can tell it, like, give me half. I think half, maybe only up to four.

I'd set mine to four gigs for GPU, though. So I started thinking, if I could take my Steam Deck and get this running a local model, this changes the dynamics of the privacy situation, of the economics, and all of it. So first thing I had to do was get on the Steam Deck, right? So I jump in desktop mode. And there's nothing really set up in desktop mode. So I'm using the pads to, like, mouse around and getting console up and then using the virtual on-screen keyboard, which I've never been a huge fan of, to type in with the trackpad.

And, you know, of course, I do the right thing. And I go over to OpenCode's website and I grab the Curl Shell command, like a gentleman. Pace that right into my terminal on the Steam Deck and just let her rip, right? You know, just pull it down and let it rip right to shell. Pipe that, baby. And sure enough, it installed, I think, in my home directory. Like, the OpenCode installer detected that it was an immutable, OS, and it, like, detected that you need to set everything up in under my home directory.

Nice. And then I closed console and reopened it and typed an open code and Bob's your uncle. There we go. And then I just used one of the free models. And I just said, turn SSH on, do a couple other changes. Open this baby up. Yeah, and I hit enter. And I did all of that.

And then I set up, like, an SSH key for myself and stuff. And then went over to my main box and just said, go inventory the heck out of this thing, because we're going to bang on this. and so it, you know, S-SH'd in, did a full inventory of the OS, figured out, you know, Steam Halo or whatever. And did the whole thing. So we kind of created like this inventory and documentation. So we knew what we're talking about. We knew how to S-S-H in there and reserved an IP for it and D-H-CP. So that way I could, you know, always talk to the same IP.

Like a gentleman. Right. I got to have these things organized, Wes. It can't be just random IPs. Well, because this is becoming an infrastructure pretty quickly. Yeah. I think it's this thing going to be just on all the time. I'm like, I'm going back to the dock.

I'm going to put this thing in a dock. I'll just turn the screen off. and I'll just let it rip. So got them all loaded up, but like, you can't just go get any model. It needs to work on that hardware. It needs to work with the resources you got.

It needs to be capable of the thing you're trying to do. I mean, even in the last three weeks, there has been just this embarrassment of riches of quantized and smaller and smaller models that can do things that I thought, well, what we're seeing, what we're seeing, and it's something I kind of thought we'd always see,

I think, is these models to be effective at a very specific task, don't need the entire internet in their weight. So they don't need to have a, you know, 10, 20, 30 billion parameters, maybe 1 billion, maybe 4 billion parameters. And if they're just doing a very focused task, they don't need all of Reddit's thoughts on recipes and, you know, car reviews and everything like that. Yeah. And we have the ability to distill down now, right? So we can take like a broadly capable model that's developed that skill and then sort of try to train a smaller model that can just capture the stuff we want from. And what's awesome is this is where all the

free stuff's at. All this really cool, hyper-focused, mission-specific, small enough you can run on real hardware. It's all open source. It's, it's, this is where, and this is where the most fun is at. It's a fun, wacky, creative sort of like. Crazy wild frontier of just models that you do very hyper-specific things. It's really fun. So I had a plethora. Like a treasure chest. Yeah. Yeah. And so job one was throw as many at this thing and see which one's producing. What actually works. Because there's a lot to choose from. So that is what I did next. Finally found a way to use your Steam deck.

What model did you load on here? Because I'm assuming that's not the easy part? You know, I did have an embarrassment of options, Brent. I really did. It was nice. So I call my deck fish deck. Oh, my goodness. I figured, you know, we'll go fishing.

So I ultimately bounced between two models. I kind of narrowed it down to two, probably within about 25 minutes. And it was a couple of options of different Quinn models and YOLO. And YOLO 11N is pretty new. And it won in what I did is I took a couple of images from Frigate that I knew what was in them as test images, like a batch of six test images, and would run them through the different models and then see what their results are, since I knew what the images were. And Yolo just won the most often there. people were at a 65% accuracy with 88% confidence

and cars it could detect with a 91% accuracy but the other thing it was a good job of like bouncing between CPU and GPU so if I kept it warm the first image was like 50 milliseconds to recognize but then it would just be you know 10, 15, 20 milliseconds

that's not bad for an image all local right all local now as soon as I finish this up a brand new Quinn came out that might actually do this even better I mean this is how fast this is moving because I did this on this week and then Saturday or something, a new model came out. There's a couple of extra false positives. I think if I were sending it through a more sophisticated model, some of the things would be a little bit better. Right. Like, for example, it's

frequently detecting the chickens as stuffed animals. I mean, this is true. It calls them fluffy stuffed animals. Fluffy stuffed animals sitting on a shelf is how it describes them. Have you customized any of the sort of, you know, system prompt or other context you inject in around what to expect in the images? Yeah, okay, so I am starting to do that now, but initially I had not in the raw testing. But now that I've kind of landed, I'm trying that to see if I can...

So are you going to push towards more stuffed animals? That's a good question. That's a good question. Describe as if a fantasy novel? The other thing that throws it is it's getting... Okay, so I should probably describe the system. So what happens is there's a collection that's pulling these images from the frigate API,

and it's storing them into a queue on the deck. and then the deck runs every so often and just batch processes them and then it cleans up and since they're coming from Frigot if the coral has assigned like a bounding box with a label that comes through to the model and so sometimes the model

just cheats and reads what Frigot has labeled it as and calls it that so Frigot labels my rooster boss as a dog so every time now Yolo sees boss because Frigot 's labeled that's a dog no that one there that's a dog I know that one. You're like No, buddy, that's not a dog. That's a rooster. That's a rooster right there.

So it goes in and it does this batch processing. And then the thing that I think is probably going to be a little weird to some people, but this is what I'm building, is a nightly summary report that comes to me in a web dashboard that summarizes the day's events around the RV and the farm and the chickens with a top-level summary, which is in progress, and then a couple of key moment snapshots where I can see a description and a snapshot with a button to, you know, make a correction. And I also get like a three-line sentence summary sent to me in telegram of like, this is what happened today.

The chickens did this. We saw Levi do this today. And that kind of stuff. But to get there, I had to kind of put cameras all around the property. And we do have a couple of cameras inside our home. Now, the reason why I did cameras inside the home is we live in an RV that sometimes is in places where you need to know if somebody entered your home. Maybe you're parked at a Walmart.

You know, your home, if you think about this, you put your home in a parking lot or at some random campground where you leave and you're not home, you want to know if somebody came in your home or not. So we have a couple of cameras inside the home in the living room and a couple of interior places, not everywhere, except when Brent's there,

then we put them everywhere. And so we're able to kind of proactively turn those on and off when we're there or not using home assistant. But it does require you have cameras everywhere. And Wes is like, I don't know, that's a little creepy. He said you're a bit of a creep. And your family probably hates you.

I don't think that's what I said. I'm not sure that's what I said. Brian, is that what you heard? I kind of heard. That's what I heard. Oh, yeah, definitely. It's the subtext for Wes. You've got to pick up on.

No, I'm not, I like what you're doing. I think it's a neat setup. And I think it's a lot of fun places you could take it. But there was sort of a lizard brain part of me that it's always watching. Yeah. And I think, you know, we just live in a world where there are fewer and fewer unmonitored spaces. You've talked a lot about, like, flock cameras on the launch, of course, but, you know, everywhere, you know, they go into any old store. They have more and more fancy camera setups and more and more people out there do have sort of things like ring cameras.

And there's nothing wrong with that. I'm not trying to impugn that. It's just sort of the collective part means you spend more time in a state where you know that you're being observed. And then so if that also is sort of in your friend's houses or in your own house or in, you know, someone else has set it up and you're kind of always around it, it just keeps a little part of the brain active and you don't, I don't know. I think this becomes also much more of a consideration when you think about where I could take this already. So very quickly within probably about 10 minutes, I could have a system where I say, how many times today did you see Abby? Or how many times did you see the chickens? What did the chickens do all day today?

Map the movement across the whole day. Yeah. And that I could probably do in the next 10 minutes. And then with probably six months of work and just learning to use this system, I could probably evolve this into where did I put this box? Where did I set this down? You know, those kinds of things. Historical locations of things and how that's changed. Now that's a whole level of monitoring to get there.

Right. And it means cameras and all kinds of places. and it would be completely inconceivable if this was going offsite to a cloud to router. Completely inconceivable. It's something you can begin to kind of think about. But I think the family acceptance factor and your comfort level,

like so for us, like there's no cameras in the kid spaces and there's no cameras in bedrooms or bathrooms, obviously. Yeah. We do have a camera in the living room. That activates, though, when we leave. So the kids know that if Hidia or I leave, and if they're not on the Wi-Fi, there is a possibility that the cameras could be on

when we're not home. That seems like a good way to do it. But we had a conversation around that. And when I, so a couple of years when I put them in ago, because I mentioned at the beginning about being concerned about the RVs security. Yeah. We had a conversation about, you know,

there's a real difference between privacy that you have control over and privacy that you're essentially handing over to a company that you trust. And so that was the first time a couple years ago, like we had a really concrete discussion about cloud versus private and local. And so that was something that, like,

I had to get family buy-off on before I could put that camera there. But the majority of the cameras are on the outside around the farm, which they do show up on from time and time, but we all know about them. And that, to me, is even more useful because I want to know when people came and went, but mostly I want to know if, like, coyotes show up and stuff like that. Yeah, absolutely.

But I do have to, like, if I want to take it further, I have to give consideration to that. Or, like... I think it is that part of it is the capability thing. It's like, it's not like I wouldn't go in the space, or, you know, whatever, like, I don't really care. But it just felt like the combination of how available it is and how powerful it's just sort of growing. And that from the outside, it doesn't really change, right?

Like, what was just sort of one level of monitoring can become a whole other thing. Yes, yes. And, like, I have, I trust you. I'm not worried about what you're going to do with that. That is a really interesting aspect of this, all these capabilities, is these things that have just been doing one thing for all these years and how capable of so much more with this tooling. And the thing that's crossed my mind in this experiment, because what now it's been about three days? When did I start doing this? Yeah, sometime this week.

About three days, I've been getting these dashboard reports. And each night I'm kind of tweaking it a little bit. But across my mind, it's like, what if I wanted to walk around in my underroos? And I'm doing it in a space now that I've turned on for this testing. I don't know if I'll always leave it on in the space, but for these three days I've had it on. I'm like, what kind of summary am I going to get? Am I going to get a snapshot of myself in my underroos if I walk around, you know, with, you know, that guy walks around in messy RV.

But it's interesting. unflattering. Like, you're the architect of the system, and you are questioning whether you should walk around naked in your own space, right? Right. I mean, that's why I don't know. Again, that camera would not normally be on. So it's not normally something I have to think about. I've turned it on just to see how this process goes. And there's been the flip side, too. Like, last night, we were just having a chill moment with like some music playing and the lights were kind of dim. And Hadee and I were chilling on the couch. We all had a great day. And Abby was in the kitchen. making cookies. She's got this great cookie recipe that she makes from scratch. And it's just a everyday moment. But then when I got the summary and I was looking at this morning, I was like, oh, that was actually, that was a really sweet moment. I was a nostalgic for a moment that happened last night that at the moment I don't think I took in how nice it was just to be a simple moment of all of us together with, you know, our daughter making cookies while we chill on the couch. Like that's great. Totally. And then to see it this morning, like, oh, yeah, that was really nice.

I'd actually like to save that. And I thought, I should build in a save feature. I want to save that photo. Aw, good idea, yeah. So there's a flip side to it, too, where it's like, I don't remember anything, and every day blurs together, and I lose all of these moments. I just don't hold them at all. And I noted when I vlogged for like 10 minutes, 16 years ago, how much more of those moments I retained because I could watch them back. So, like, we have more of a memory of those moments.

An automated moment capture. Yeah, and if I could recreate something that's more like that, and less the creepy, that would be the direction I'd like to take it. Yeah, it gives you a little more agency over your own life, stuff data you're already generating that you do want to have access to. And it makes me wonder, what else could I use that deck for? Like, if I'm not using it for this,

or when it's not doing batch image processing. It's a networked GPU inference provider at this point, more than a deck. It's not huge, it's not powerful, but there's so many great little models now. Like, I just saw a really small, fantastic image generation model. Totally, yeah. We should try that. Music generator, you know? I mean, just all kinds of things that I could do on that deck now, along with this.

And it's more exciting than playing video games on it for me right now. I was going to say, have you wired it into production in a way where you have to take, like, a downtime in order to play some video games? It's only been a few days. We're on Kaya watch. High Warning today. Sorry, kids. Actually, you know what I'd really like to get cameras for is, first of the first. well, it's going to be bird season, but we've got two albino deer on the property, and they

show up in the evenings, and it'd be really neat to just capture them. Yeah. Yeah. So it was pretty fun. I tried out Moon Dream 2 as well, which great works great. Quen 2.5 VL3 billion parameter, which worked pretty well. But the winner was YOLO.

The AMD APU wasn't as useful, as I was hoping. A lot of the stuff ended up working on CPU a lot better, so a lot of times I was just doing CPU-loaded stuff. Nice. Yeah. And then, bringing it all together. That dashboard was nice. Put a little SQLite database behind it. And then those images get saved with kind of like a 30-day cleanup retention period. You know, we just don't give enough love to SQLite.

It really sure is handy. Persnickety in the chat here has a question for you. Why not use the hacks forget add-on for this system that you built? Well, this goes way beyond. So you're getting, not only you're getting like a, you can have it detect anything you want, right? But you're also then getting sort of this rich context summary that ties together like the chickens the chickens did this and then they went out and did this and then they went out and did this in a summary and then a hyper summary on top of that that brings it all together

to where I was like yeah I can track I already do track dogs packages and faces with frigate but this is such a whole other level where tomorrow if I decide okay I want to know when this happens I can just go to it and ask it and it has it's a it's a little tiny model with vision that can just go find that stuff so it's like it's way more flexible inventory almost that you can tap into And all of this feeds home assistant, right? So home assistant is a key piece that kind of glues it all together. Home assistant is helping with the, like, if it knows where we're at, like all the location assigned stuff. So wind cameras should be active or alerting is done or triggering an agent for something that's more urgent.

So there's another tier in here where if it observes an image that is an urgent thing, like a security issue, like a coyote or some rando with a van on the property, then it can break out. and take a different path using Home Assistant and send me a telegram with the image attached through... Or I'm sorry, not a telegram, a push notification. So I get a push notification alert that uses Home Assistant for that. That's pretty nice.

It's very good. So this just sort of sits on top of that and gives me even more flexibility. And again, like if I had to send it all up to Claude or whatever to do this or to Gemini or to Chat Chippity or whatever... It'd be a nightmare. Yeah, and it'd be a major bandwidth suck.

But because this is running locally on the deck, already on the land, it's just like such a why not situation. There's so much more you can do. So this is my question to you guys out there is, is anyone else trying this kind of stuff on their decks? Are you using your deck for anything weird? LLM models, you know, self-hosting.

What have you found the deck is great for beyond just gaming? As a Linux computer. Yeah. It's a good little Linux box. So I want to know what has worked and hasn't worked, boost.jupiterbroadcasting.com. You can tell us about that. Got a little feedback here from Kenneth.

Hey guys, just heard episode 687. That's the last one. And wanted to share some gold regarding the Apple and Google TV topic. Go buy a Raspberry Pi 5 and a case with NVME SSD support. Write Librelek to the SSD and boot it up. Then you go and add Jellyfin as a repo and install Jellyfin. Then in Jellyfin install the Cody's sync plugin. Liberleck can do a muster.

much more. I have the older pie 4 and an argon 1m.2 case, but with some overclocking and higher memory to GPU, everything seems to work just fine. Well, Kenneth, this is the thing. And it really upsets me to no end. It looks like the just use a Raspberry Pi and OpenElek or Cody guys were right all along. I've tried to fight it for so many years. and they've they've just had this it's just worked for them for all these years they haven't these problems and I just didn't want to hear it especially I suppose a lot of folks I need a Netflix app where it falls down is that if you do want a proprietary thing

I don't need a Netflix app I just need a TV that works you know Kenneth is right and you know how I know I've capitulated because guess who was in Cody last night playing around with Cody seeing what I could make. Yep. Yep. I don't know you anymore. However, I did it on a Google TV. And this is this quick rant.

I'll try to make this quick. This is really disappointing because it was one of those moments where it's like, I was like, hey, why, if you want to watch TV in bed tonight, you know, we'll do a little bit of it because the kids are out and about. And I was like, well, just chill out in bed

and watch some TV. She's like, yeah, we haven't done that in ages. Like, you're right. We haven't. So I'll go fire up the old Shield TV right now. And this is the good one. I'll fire up the Shield TV. right now. I'll make sure all the updates

are done so it doesn't prompt us to do updates and I'll make sure the jellyfin works so we don't have any issues with jellyfin and whatever we want to do, YouTube, Netflix, make sure it all loads make sure it all go into the Play Store. I'll make sure all the updates are done and then we'll call it good. So the first thing I do is, you know, Nvidia still

bless their hearts updating this thing 25 years later. So I started download, of course it's like 7 gigabytes or 3 gigabytes, whatever. So I go start that in the background and I go do all the other little app updates. By the time they're done, the the download is update and it's ready. Would you like to restart?

Yes. It restarts and for 15 minutes, I'm at just the Android boot loop screen, which is Android's in there flashing, and I realize that it has failed to update at that point. I look it up and I'm like, 15 minutes, like, yeah, 15 minutes, you're probably screwed. So I have to, after an update,

which is always a good idea, unplug the device. That's always what they recommend you do. I think they tell you to do it in red letters. They always say, be sure to disconnect power. A random time during this process. So I'm like, boy, I sure hope they have an AB, boot environment, boom, I pull the power, plug it back in.

It boots up. Updates still available. So I go ahead and do the install again. This time it works fine. Okay. So good. All right. We're all set.

We're good. Everything's been logged in. Fresh update. The updates are done. Turn the TV off. Comes around time. Okay, let's go watch some TV. I turn the TV on.

Go to the Jellyfin app. Head enter. Jellyfin crashes. Oh. Okay. Click it again. Jellyfin comes up. completely logged out.

No server entry. I don't have a keyboard hooked up to this thing. I don't even have my phone paired to it at this point because it's been so long. So like I'd have to like bang out my jellyfin username and password and URL and everything just using the on-screen keyboard. And we bailed. I'm like, that's not what we came here for. No. And that was my Google TV experience last night.

So. Will you try a fan? Yeah. Well, I mean, like the idea is like I thought like, okay, the issue is I'm not using these things enough and I just need to give them a a little TLC, and like I would, a device I use daily, just get it, you know, and then, of course. It was probably the update, probably something like the Jellyfin app was suspended, it rebooted, I don't know, but, yeah. Blumen Strong wrote in, here's a little Kickstarter for using Jellyfin MPV Shim.

It's a cage and a NixOS to have a dedicated Jellyfin client. It's been working for me for at least six months without issue using Jellyfin app to control the client. The key thing is to make the cage wait for Jellyfin server. to be accessible before allowing to start the MPV, the MPV shim. Yeah, Cage is like a kiosk-oriented. Like a sidel. Wayland. No, it's like a whaling capacitor.

Oh. Oh. There you go. Oh, that's awesome. So he is setting up a wailen-based, dedicated kiosk jellyfin. Yep. Well, that's exactly what I asked for. Yep, here in NXOS.

Blooming Strong. Tell you what. Although, I don't know if you notice it at the bottom. I expect a full report from Brent on something for the Texas trip. I was just thinking this morning, we're going to get down to Austin, and I am going to feel like such a jerk for once again not having some solution for this. So we need some hardware, right? What do we want to use for hardware?

We need something small and portable to run this on. Like the deck, you mean? No. Maybe. We'll put Cody on the deck. Yeah, call it good. That's really great, though, Bloom. Thank you very much for that.

Wow. I got to think about that. We've got to think about getting some hardware together for that. We probably have something, but they're all pie-based. It doesn't have to be super small. It doesn't have to be crazy, right? I know Brent wants to use an Android phone. No, I didn't say Android.

Oh. Yeah, it's the phone part. It's the phone he can't find anything else to do it because it's not really a phone. No, I just, I don't want that phone. I want, like, a phone. phone format because it's easy to pack and travel with or whatever. You don't want it to be an Android phone or an iPhone. Oh, just give me Linux, man.

Right? Can't we have that? Don't you have a Linux? Many of the popular Linux phones that are on commercial offer, you know? I mean, you got one. Didn't seem like it worked last time we tried. Oh, yeah. Well, that's because it was how old?

No, I don't even remember. It's been that old. But it was exciting and entertaining, wasn't it? It was that. Yeah. So, stage one success. Sometimes, though, we just want to watch Baywatch or Star Trek. Yeah.

Yeah. And your Klingon is not very good. Yeah, you're right. I agree with that. I got to polish that. So if anyone has hardware suggestions, send it in. And now it is time for the boost.

You could boost it in like our baller booster did. Is everybody sitting down? Yeah. Are you ready? No. Black host is our baller this week. Black host knows we're going to Texas and came in with a, incredible 1,234,567 Satoche's.

Fun will now commence. That is Blackhost. That is very much appreciated. And Blackhost is sending over a couple of sats for the Texas Fund. Just a few. After all, that brisket goes best with more brisket. That's very insightful, Black host. That is true. Nothing washed it down the brisket like some brisket or some ribs.

Thank you very much. we will put that there towards the Texas fund. We got to get that meetup. You got the lean and you got the fatty. Oh, the fatty. I honestly was not initially a fatty guy either. Let me tell you what. We now, with that boost right there, we've got to get our act together on the meetup.

It's official right there. Thank you very much, Blackhost. It's happening. Well, Greg, the lawyer boosts in with 22,200 and 22 cents. This old duck still got it. But this in the barbecue fund. Ha! Very good, you guys.

B-O-O-S-T. Thanks, Greg. there. Barbecue. Not sure boosted in 5,555 Satoshi's. The traders love the ball. I've set up Mews to monitor local Facebook groups and Instagram posts in Kauai to keep track of the rebuild efforts before my trip.

It's great, especially since I don't have an Instagram account. Yeah, that is handy. I feel like if they're going to give you a free compute, why not use their free compute to interact with their services that you don't want to have to use. That's actually, it makes a lot of sense. I think I want to use it to just like scour Facebook marketplace, you know, the facebag space for just everything that I want to buy use because they've destroyed every other used local platform that I used to use.

So screw them. Yeah, just a facebag bot that just goes through, muses through and finds you the deals on the bag. I like the idea that it's a Linux VPS as an interface for Facebook and Instagram. Instagram. And the chatbot on. Yeah, well, you know, it's their own fault. Dajah comes in with 6,8-58 Satoshes. Coming in hot with the boost.

I'm so happy a lot of projects are finally getting out of the AI is good, AI's bad rut. I have ended up somewhere similar for my engineers of work. At the end of the day, they're responsible for their code, period. People have to be making slop. People have been making slop for decades. Anyways. So if they submit a PR with GPL code in it or security flaws or anything else,

it's not mergeable and needs to be sent back regardless if AI was involved or not. The biggest benefit we have seen from this is finally upping our CI game. Nice. I bet that is true for a lot of people. I certainly had up my CI game. So I bet that is true. I was just reflecting on that earlier this week about how I've had to adopt a lot more developer workflows that I've never bothered.

Could never have been arced. You're a software factory now, whether you like to admit it or not. We've also been putting a couple of lines in our agents.md file to have the agent do something trivially obvious to a human, but been surprisingly good at filtering out the AI made this without a human looking at it whatsoever, and automatically sending it back to the contributor to fix and resubmit. Hey, sounds like what system he was doing. Yeah, that's an interesting approach. We talked about that recently.

It's not bulletproof by any means, but it's good enough to catch most people who don't really care about the quality of their submissions. They don't care enough to perform a simple task. I was eating some fries and felt compelled to boot. You follow me. That's a little nod to the bootleg there for one of our songs. Thank you for the boost. Thank you very much. Abe Mante, Boosen, with 6,000 cents.

Hey guys. You've given me so many good picks that I thought I would send one back to you. Kitchen Owl is what we've been using as a self-hosted grocery list. It integrates with home assistant, so we just use our voice. to add items in the list when we run out. That's brilliant. I like the UI, too. It looks like it's got a really good mobile UI.

It also sorts the list by Algo, so when you walk through the store, checking items off of the list, it can learn where the items are. Uh-oh. And orders the list that way for subsequent visits. Stop it. That way you don't get halfway across the store

and have to double back to the dairy section for something that was lower on the list. Cheers. Caesar's Ghost. Thank you. And it is AGPL3 as well. This looks like a heck of a pick. That's a really good pick.

Kitchen owl. Impress. Kitchen owl.org. Thank you, Abe. Get your kitchen owl. Kitchen owl. Bamham 182 sent in a 2,500 set boost. Boy, they are doing a lot with mayo these days.

Regarding Google TV, I found the experience to be a bit better if you enable apps-only mode. It doesn't fully remove the idiocracy-style ad bloat, but it does make the whole experience slightly more streamlined. I was really excited about trying to get a steam machine. to replace my on 4K Google TV, but I just can't justify that cost right now. Yeah. Understandable.

Yeah, it'll be a nice side benefit when you can get the steam machine that it could also possibly do that. Okay, app only mode. I will look into that. I do have to say, on my shield, I have the projectivity launcher, and that I really do think is quite great.

I'm quite happy with projectivity, but you have to replace the launcher to get there. CGBA came in with 10,000 sats. I normally stream my sats, but I forgot to top my fountain wallet off. So here's a boost. Since I'm here, I just want to let you know that I've been experimenting with a lot of self-hosted LLMs and Hermes, thanks to you. Also bringing an old topic back. My password manager of choice is pass.

Why no one has mentioned it? I don't know. It's the standard. I even recently set up a Yuba key to unlock the password. Very nice. Maybe we should take another pass and pass. Especially these days. Keep us posted on the local LMs.

And I'd love to know, C.G, what hardware you're using, too. Oh, hey, another boost here from Daja. Yeah. I noticed this one says open code. Oh. Somewhere in the boost metadata. Oh. Mm-hmm.

Have some fries, guys. You sent me down the Muse rabbit hole this week. So this boost was sent via open code using my Muse model via a custom open-a-I compatible wrapper and Albi Hub. Okay. Okay. So Muse is using the VM to drive open code where he has boost CLA set up, connected to his AlbiHub.

Wow! Or I think, you see, maybe open codes talking to and controlling the Muse model? Oh, oh, maybe. Plus, Albihab also has, like, you know, you can, yeah, okay. All running via Nix on the free Muse compute.

V-Zero's on my GitHub, very much a rough proof of concept, but at least gets U.S.H-Vy-Tale and the opening A-Rapper with a single prompt. Beautiful. That's got to be one of the nerdiest boosts we've gotten in a very long time.

Well done, sir. Wow Wow. Muse VMHag And it actually sent the ads. That's incredible. That's, oh, think about the layers there. Think about the layers there. That's amazing.

That's, that's great. I think we might have a live boost we may want to include, or you want to include that. Oh, yeah, coming in hot right now. I think so. I don't know if Wes has the means, but we can continue here if you want to grab them. From bearded tech.

It came in via the web boost. Why don't we grab while Wes pulls that? why don't we grab Tech Geek who came in with $25 just because he says That's brilliant That's good Thank you

I appreciate you I appreciate you Thank you very much Well Wes is working away Did I stall long enough for you to Let's see here Do we have it? I don't know if we do

Well I can tell you the amount And we can read the message in the next one If we want to be fair about it Bearded Tech came in with a space balls boost one, two, three, four, five sats. So the combination is one, two, three, four, five. Thank you, bearded tech. Appreciate you.

And we'll have that message as well. Prosno link? What do you think? Prosonic. Oh, that's way better. That's way better than what I said. They came in with $10. Ten gold blooms out of a possible ten goldblums.

Boosting for Texas and the muse episode. Ignored the corporate agents until the last episode, signed up immediately wondering, like, Like you guys, what can I actually do with all the positive things with this? Also, first time booster! Nice! Like mini, I came from self-hosted, and now I'm here for unplugged. Boost!

Thank you. We sure appreciate it. Put that towards the Texas Fund. Yeah, the idea of just free compute, like, I'm not going to use it for anything super private, but it's also going to be interesting to watch how they try to compete and what they keep rolling out. You know they're not done yet. Whereas the line came in with a free members boost. long time booster

or long time member first time booster there you go I got it thank you very much as you react to the feedback about discoverability of the booster oil

I tried to find it now it's easy so keep up the good work nice all the talk about jellyfin fits my home lab perfectly as I'm

simplifying to a true NAS uh huh the free as X86 box for the this free is a X86 bucks for the TV

I do sometimes just think life would be a lot easier if I just did X86 everywhere right yeah It might not be exciting, but it might also work.

X86 Maxi over here. DeBrown Chico comes in with a member boost. Hey, another long-timer. Long-time supporter since the self-hosted day is here. I get the buzz around Omarchi and it's million-dollar funding. I think it's like $21 million now or something like that. But the creator can be a bit polarizing. Could we maybe pivot the spotlight away from him and focus more on the tech itself,

like the security and privacy of an agent-driven OS? Have we talked about? do you change much? I think we've just mentioned some of the updates, mostly in the pre-show. Yeah, I think that, you know, the comment this might be in response to, and maybe I didn't make the point well enough, is I think what I'm saying is remove the person, even remove the project to just ask yourself, why have they raised $22 million? And why haven't other projects raised $22 million?

What is that? What is the signal? Now, that does require taking the emotions and putting it aside and just asking yourself, what signal could the free software world be taking who struggles to raise money, who struggles with funding every single day? Why was this project within three or four months able to raise $22 million?

They just announced now, and it's just kicked off three years of funding Hyperland development and a couple of other projects, and they have like a full-time staff of like five people now. So I think the question that we have to ask ourselves is, what can we separate a signal from that

and internalize? And that's all. I don't really think you have to talk about DHH much there. I don't think. Yeah, I think you're kind of aligned with maybe what the booster's asking. It may not have been a mid-reaction. It might be sort of trying to express similar things. Yeah, yeah.

I think we're on the same page then. Brett, you want to take DeckBot? Well, DeckBot sends in a member boost here. Working on the back catalog and headed several listeners mention automatic ripping machine. Yes, yes. We are the automatic ripping machine show now. Well, I wanted to pile on and mention that I once got it running on an

8 gig Raspberry Pi 4 running Ubuntu server, a USB hard drive, and a USB Blu-ray drive. DVDs took about a week, and it never finished a Blu-ray. You know, I can almost do a week. It's a little slow, but like, who cares, right? You're just letting it go. Yeah. But never finishing the Blu-ray is a bit tough. That's a bit tough.

What were you saying about X-A-6? You'll fall asleep halfway through anyway. Love X-86. So last comes in. with a member of boost. Hello, J.B. Trying this member boost thingy just to say thank you

for the last 16 years. You gotta be kidding me. You gotta be kidding me. Is that 2010, then? Listening to you literally got me from help desk into college into a wonderful career directory where I have used what taught me

what you taught me at every turn. You guys make a real, tangible difference, never stop. Well, thank you. Yeah, thank you, Silas. You think it's Seattleis? Silas. Thank you, Silas, very much.

I hope we get a chance to shake your hand one day. 16 years is. no joke. That's older than two of my kids. Encoded member Boots in. I've been a listener for two years and giving value back by joining as a member. Hey, thank you. I look forward to listening to new episodes every week. My two cents on Clinker safety. I have mine used trash CLI and not RM and I found it helps. Also stating it in my clanker.mdr.md. Can't wait to see y'all in Austin at Texas Lutzfest. Cheers to that encoded. Yeah, I like that. Trash CLI is a good idea. Okay. Okay. Trash CLI, huh?

Why not? Then you just sort of have a delete queue. You can review it, go delete things, put it on a schedule if you want. It seems a way safer, as long as they use it. But yes. As long as that's a big caveat there. Well, Monty boosted a members boost. Clinker Therapy Volume 2 was great, and I'm glad to hear there will be more coming.

Yeah, this week, Friday. Check it out. This open code V2 with the server client setup and an API for the Clanker is going to change the way I do project. Yeah, buddy. Also, like the idea of having the agent create a project dashboard for tracking progress. Thank you, you guys. Yep, why not.

Yep, could use Zola for that, too, actually. You are onto it, Monty. You are in the right lane. You are in the fast lane, sir. Good luck with OpenCode 2. So worth it, especially if you have a persistent system. You can leave it running on. That's great.

The 4992 comes in. It says, thanks for the show. Thank you for being a member, Nafora. Thank you, everybody, who boosted. We have Jay Moon, who came in with 1,000. 99 set, just below the 2000 set cut up, but wanted to recommend the self-hosted app, nostalgia TV, I think is it?

Yeah, nostalgia TV server on Docker Hub. Yeah, I saw this come in live, and I wanted to pull this one up. I thought this was great. It's a connector for Jellyfin. It generates a programmable schedule, but only plays the files directly from your Jellyfin library on demand

to match the lineup. Nice. That's a really fun idea. Clever. Kind of minimal, clean. It kind of gets me my ersats. Right. But without your ersets.

Yeah. But still gets you the TV guy jumping around thing. I understand that you're hating on Google TV right now, but a dedicated Android app can connect to the server, or even run independently, and works flawlessly on my on-branded 4K Pro, 3 gigabyte of RAM, 32-gig storage box from Walmart for $60. He got cut off.

He said, goodbye with Pro something, and he got cut off there. J-moon, I wanted to pull that up because that's a great tip there, and thank you very much. That might be just an easier route to go in the future. I'm going to definitely look into that. All right. Let's wrap it all up. You know this is going to be a good one.

We had some really great boost coming in. We started off with Blackhost coming in with that incredible baller boost. And, of course, Greg coming in with a nice, big, fatty McDuck and all of that. So thank you, everybody who streamed Sats. We had 10 of you streamed Sats. This has been tricking down slowly, but we still got 20,418. Thank you very much. When you combine it all together, this and you take the Fiat boost to, you convert them to current market Sats.

Get ready, boys. Are you ready? Get ready for this. we stacked a grand total of 1,363,3,320 Satoshi's. Wow. We are going to Texas. We are going to have a meetup. That's my job this week is to get that sorted.

Buy some extra comments. We're going to need them. Yeah. Let's keep an eye on the meetup page for that. Thank you, everybody. Boost.jupbroadcasting.com is how you can get a message in, send sats or fiat fund coupons, or put your support on autopilot. But this year has been devastating financially, and your support has been the only thing keeping us going because the sponsors are kind of moving on.

We've got a couple of nibbling these days, but we have dug ourselves quite the hole. And without you, we would not be here at all. And we're chugging in now on the 13-year mark. That's no small thing. And there's really only you out there as an audience that cares about a show like this. It's not going to be a mass market thing. Never will be. And we're totally happy as long as you're happy.

And you can let us know at boost.jubitubricasting.com. Wes, you found such a cool network troubleshooting visualizer for the Tui. I was wondering if you were going to find and notice this one. Oh, I noticed this one as soon as it came in. And I'm like, oh, yeah, we're doing this. Because I've actually been kind of trying to figure out an ideal Wi-Fi set up for me. So this is an app that gives you a live visual network path and diagnostics that, I mean, how do you describe it? It's like it shows you kind of the path that you're,

data is taking and where it's hitting issues. Yeah, exactly. It's called WTFI2, or what the F, internet. A live visual network path diagnostic that pinpoints exactly where your connection dies and tells you how to fix it. It's, I guess, actually, a ground-up rewrite of the archive WTFI v1. The old version printed a flat checklist and left you to correlate it. This one draws a fancy diagram, streams every hop live, reasons about the root cause, so the answer is one-line.

Nice. Yeah, it has a very shiny little to-wee going on. Of course, it's written in Rust. Oh, of course. Ratatooie. It's written in Rust. It does seem like a nice place to start to, like, you're on some network, it's not working, you don't want to do a bunch of troubleshooting. Answer a bunch of your first troubleshooting questions that you could run several different commands to try to get all at once.

Is this the year of the to-y? Yes. Seems like. Every week. Well, don't roll out 2027. Every week, we've got another great to-y. Yeah. This is incredible. You know, you were musing about local mail clients.

Yeah. And I was thinking of myself when you're saying that. I think I would want a to TUI if I was going to do that. Interesting. Or I might. I might. Huh, huh. All right. Well, let's break that trend with a full-on desktop application.

Boys, we're going to talk about space scan. And I admit I have a soft spot for these fast disk usage analyzers that actually visualize it in a decent way. And this one's called a starburst chart, keyboard-driven. and it outputs in text CVS JSON as well. So if you want to throw it into something else, it's just really great fast disk usage analyzer with a sunburst chart and colors that means something for space usage and things like that.

Really easy to identify where all your space is either on your whole system or in your home directory. You can measure by bytes or number of files, and it takes pride in getting accurate numbers with support for fuse file systems over the network and things like that as well. Updates live as you go, filters by large things like ISOs, MP4s, things like that.

Also, accurate numbers, real disk usage like DU does, you know, handle sparse files correctly, hard links, count ones, things like that, which is nice. And it says they have what they call safe cleanup. So you can move files to the trash or delete them permanently, always with a confirmation. Space scan refuses to delete anything that is or contains a mounted file system. I do like to let you clean it up too. Yeah, because you're already doing the work. It's why I have to go open some other thing just to go clean up the stuff, which is probably what you were looking for in the first place. I thought we should give a quick mention.

This is a packed pick segment. We should give a quick mention to Vermuth. Mm-hmm. A lightweight Linux game launcher. It's really designed for the plasma desktop, but would probably work on any desktop. It's native C++ and cute. It lets you launch Windows games from Steam and GOG and retro games in a two megabyte app. Zero bloat.

KDE first. Native, yeah, C++. Yeah, seems like if you just, all you're looking for is nice launcher. You're not kind of... What you want is something that works. really well, can use and find all your games, and has a single place to launch them and get out of your way without having to install, you know, games enough are already hundreds of gigabytes.

And there's folks that, you know, they want a plasma native feel when they're on the plasma desktop and this would get to that, too. Now, normally I would spend a little more time on that app pick. It does have a flank, too. Oh, we do have to talk about this app pick. It's kind of mandatory that we cover one of these at least. and this is part of a group of applications that have essentially recloned, re-created the functionality of the Adobe Creative Suite. And this one is like Ra-Machis.

How would you say that? Ramakase. Ramakasi, which is an open-source lightroom replacement. And all of these are moving quick. And it even opens lightroom files. It's a fast, non-destructive raw photo editor. developed for Linux, MacOS, and Windows, written in all of these, are written in Rust.

And the UI, the UI not only looks like Lightroom, but if you were to take Lightroom and put it side by side with this application, it would look almost identical. And if you opened up a project file, all of the sliders, all of the color settings, everything is restored, identical to the way it is restored in Lightroom.

Yeah, seems like that was basically the ask and a whole team of folks, mostly clankers. turn loose to go work on. Although now there's been more contributors getting involved and kind of taken off. I couldn't help myself but to run this thing. And, you know, NixOS makes it so easy. So you just get it going like nothing.

And like, Chris, you asked me like, hey, Brent, did you use Lightroom, you know, when you were a professional photographer? And that was the main piece of software that I used for most of my career because it was, when it came around, like, everybody was using Photoshop, which was, you know, pretty powerful doing digital, you know, corrections and things like that. You should have seen the automations and macros and stuff that us photographers were using. It was, let's say, a little bit clanky. And when Lightroom came around, it just completely changed overnight

the workflow that photographers took. And it was like, you can't do photography without it after having used it. So when I moved to Linux in about 2010, that's what I was looking for. It was like a perfect lightroom replacement. And I used Darktable at the time, which is an amazing project. They do some really cool stuff.

And I use that for many years. But this Ramacase looks and feels exactly like Lightroom. Like a one-to-one twin almost. And Darktable did very close to that, but also had its own features and stuff. that were very advanced and really cool. So there were some advantages there that Lightroom didn't have. But if you just want a one-to-one replacement that is very fast and really cool,

this would be the way to go. And so I've got it loaded up here and I've just been using it. It's like some old photos that I had. It's sweet. It does an amazing job. I have zero. Like I've only used it for 10 minutes. So, you know, bear in mind.

But you should start here. Like anybody who's just starting with Lightroom and hasn't used. it before, don't use Ramacase. You know, there's probably some bugs or whatever, but this is a great little piece of software. This is exactly what I wanted back in 2010. Was like, give me, this would have been huge right back then. Yeah. Give me exactly the proprietary software, but don't give me the subscription and stuff and make it open source so I could, you know, follow my ideals and contribute to like something that everybody can use. And this gets me thinking about, like, is this the beginning of the end for proprietary software?

Well, they're going, I mean, this feels a bit like a statement. This is the quintessential, right? Because all of this, not that they, I don't know if they would have accepted it, but all this clanker time could have been spent on Gimp or Blender or Darktable. Mm-hmm. And again, just as a side comment, it does make me want to revisit that conversation one day that if these projects reject AI development, they're just going to get replaced so fast. So fast. But that's a separate conversation. I think also you have to consider Adobe's probably going to come after these people or try to.

I don't know. It's open source. I mean, this cat is out of the bag. The Photoshop one is crazy too. I think, too, we've already seen sort of fights over folks like relicensing their tests or making tests private. And then this is sort of an output driven, right? Like try to build it and compare the two things and do a clean room implementation just from running the program. They do say it's a clean room.

And I think it shows the power of that approach now that we can just grind against it. There is a double-edged sword to this type of project. And that is, it's moving really fast. The last commit was two hours ago on a Sunday. There's been 21 releases since September 26. There has been six in the past week. Substantial additions are coming in like batch export, experimental local AI masks, improved catalog handling. There's been Intel and Nvidia driver fixes that have landed.

Most of it, though, is actually coming from the original developer. There's been 24 outside contributions that have been merged. There's been plenty that are contributed, but 24 that have been merged. Still early. Yeah, like this one doesn't have a lot of support yet, so there's still something. One other important thing in the ReadMe notes that before the 1.0, the file format could change, like how the edits render could change just as that way. Sorry, Brian, go ahead.

The other really cool thing about this is that they have the capacity to just import your lightroom catalogs. Yeah. I didn't get to test this because I had to scour a very old harddress to try to find. It has a built-in MCP too. And that's a killer feature because you can just migrate quite quickly, assuming, you know, it all works and stuff. But very, very thoughtfully implemented. What's neat, too, I think is, right, we can take this. Like, I've already experimented a little bit with making myself kind of like a version of, you know, kind of like Instagram style filter editor for photos. Just because I didn't need the full stuff. I just wanted basic stuff. But you could now take, you know, they've got versions of a bunch of algorithms. and filters and effects and capabilities that you could also take and stick on mobile

if you want it or render in a web UI or you're not limited to making only using the clone aspect because it's MIT licensed. The big ones I think we still need to see. Will we see an after effects? Will we see Premiere Pro? Will those come? And how will we as a community field?

Because these are open source, but they are essentially competing with Acadian Live or Blender. It also shows, right, that the other capability exists, right? Proprietary folks can clone GPL We have already seen stuff going GPL to MIT with clones. Could you even argue that because they sat on their butts and didn't do this themselves, the market did it for them?

There's that. There is certainly that interpretation. I wonder. This is wild to me. And I've also, I haven't seen these yet, but I've also read online that the, I think it's Word and Excel. I'm not sure about PowerPoint, but I think Word and Excel have essentially been cloned as well from the Office 365 versions of them as local apps that will run on Linux. Looks like VectorCraft. PJs and a few people folks are linking that.

So you're... What's VectorCraft? Is that the officer? Open source vector art tool, so illustrator sort of thing. Wow. Yeah. Everything's getting cloned very quickly. So Merry Christmas.

99% ruffed. Oh, they're point three. Yeah. Vector Craft, Vector Illustrator, an open source. source clean room implementation of Adobe Illustrator rebuilt in pure rust. A fast open source clean room take on an Adobe Illustrator workflow. They're at least

going to have to stop saying Adobe Illustrator. Yeah, that's going to get. Jeff, stop sending me. There's too many. Oh my gosh. It's a, it's a, it's there's, this is a thing, you guys. Ah, effect craft. No. After Effects. Yeah, it's just like

compositing tool. No way. They've done after effects? I go, why not? I mean, the start of it anyway, right? How done is it? Should we just do a Pix show? Looks nice.

Jesus, it's coming so fast. What's also crazy, he's all are supporting Linux. They're all open source. They all support Linux. They're all rust, which is interesting too, but holy crap. Aftercraft.

A lot of contributors. Like on this EffectCraft, 53 contributors right now? Like, they're probably a couple of robots in there, but. Hey, this one supports free BSD too. So the last commit was three hours ago. They can create an application and improve it

faster than their legal department even knows a sign. Oh yeah. Oh yeah. Wow. Totally slot. Just a moment of silence for proprietary software. Yeah. Yeah. Like, yikes. I mean, it's a new world. It is really wild.

Like, that's, Adobe is sort of like the quintessential moat. And my mom has worked in the graphics industry for my whole life. Like I just imagine like telling her that she could just go use a free version of Photoshop now. And it's 95% feature complete at some point down the road or something. Most people only use 5% to 10% of the features anyway. Some people are paying $1,000 a year for Creative Cloud.

Not anymore. And when it's open source and if you're a shop, you clone that down and you clank in your own features. Right. Of course. You clank in your own workflows. You clank in your own automations. And you just maintain your own fork. You can submit stuff if you want.

You can keep it yourself, whatever. This really is going to shake stuff up. And Adobe's going to have to fight this, but it's a futile effort. Because that's all MIT licensed. You can clone it down now. A lot of people have. Like trying to replicate Photoshop is not a new thing, right? We've had the Gimp for decades now.

But this happens like in a week. You know, the Gimp has been around for, decades trying to do this and adding features and stuff. This happens in a week and it's done. You could see why people that have been working on something for 30 years might hate something like this a little bit. I could see that. I'm curious about sort of the engineering side of, you know, what the stack looks like to run the test to make it as automated as possible. Because there's a huge coverage, right?

That's part of the power here is you can not only write the code, but... You can handle some of the commits that are coming in. and actually do all of the edge cases of, like, exploring a UI like that. Yeah. Right? Like, there must have been a lot of that kind of testing. Automated UI testing, I'm sure. Because, like, you are not, the human did not spec out the entirety of Photoshop, right?

Like, you had to discover that and then test against it probably as part of it. I'm just curious about that level of the harness setup. You could test a lot of it automatically, though. But you've got to put in the work to build the stuff to make that all run smoothly. But once you've done that, you could probably apply it to other applications. Exactly. Now you have a, maybe not fully generic, but getting close to something. something like a useful machine to make software from other software.

It's going to be a wild year, boys. It's going to be a wild year. And how long have people wanted Photoshop on Linux? And now it's here and I feel a little weird about it. I don't feel like it's not a total dubbs for some reason. I don't know what it is, but it's both amazing. Like, seem like it was after effects? Like, I just assumed it's already happened.

And I thought that would not happen for a while. It's already happened. Like, my mind keeps getting blown. And now I can have after effects-ish on Linux? Are you kidding me? Like, that could up our game. At the same time, like, I don't know, man. I don't know. I don't know.

It's a little weird. For these projects, I always love to dive into their agents, MD, if they've got it in the GitHub repo. Yeah, that's a good idea. For EffectCraft, I think the first instruction is really important compared to Photoshop and the Adobe software. Number one, never crash. People trust after, oh, geez, that was a Pavlovian. People trust EffectCraft with hours of work. Don't malform.

You know, any extensions, no bad expressions, no bad scripts. Don't break anything. That's the first, yeah, if only the proprietary software I did it in a while ago. It's weird that you can kind of get a sense of the developer ethos through the agents file. Yeah. Number two, no Adobe artwork. Oh, yeah. Yeah, avoid that.

Avoid the Adobe branding. What were you going to say there was? I was going to say on that first one, when you build the harness to sort of do this testing stuff we're talking about, it also works for that, right? because maybe now if you have programmatic control of your whole interface, you can do sort of fuzz testing against it, just trying to manipulate the UI, get into weird situations. That's my job. Yeah, well, Brent, there still need somebody like you at the end of the day.

I bet if you spent too much time with it, you'd probably find a couple of things. Something tells me. I still have faith in you. All right, well, there you go. There's a lot in there this week, and we'd love to hear from you, boost.jupiterbroadcasting.com. We're just around the corner to Texas, and then don't forget this Friday, October 6th,000, 16th, clanker therapy.

Yeah, number three, and we want to hear from you, answer your questions. Meetup.com slash Jupyter Broadcasting for the timing and all that. Yep, and uplink.com for the voice note. Send us a voice note about two minutes long. All right, before we get out of here, Wes, you want to give the power users any pro tips, some secret sauce in the RSS feed? You got anything for us? Yeah, well, if you want to find out what we talked about, or maybe get a summary,

or like find out which section of the show had that pick and what the name was, because you kind of forgot, the VTT has all that. That's true. That's true. There's also chapters, which is like our attempt at a first take.

And actually, really, it's not our attempt. It's Drew's take. That's why it's really good. Because you get Drew's interpretation of what we just did with the show. So that's a little extra metadata and flavor on the phone.

If you got a podcast app that supports chapters, which most of them do, take a look at it. He's usually got a funny one. Even if you don't need to jump that way. Yeah. It's like a little extra.

There's even an MP4 in that RSS feed, if you please, if you want to see a visual of what we talked about. and you can find out when we're live. See you next week. Same bad time. Same bad station.

You can make it a Linux Tuesday on a Sunday. Join us 10 a.m. Pacific, 1 p.m. Eastern over at jb.live.tv. Or wherever you like to watch it in your podcast app on the YouTube. We don't care. We're just glad you're there. That's truly it. A podcast wouldn't be here if it didn't have listeners and viewers and if you didn't tell people about it. So thank you so much for listening to this here show. We really do appreciate you newcomers.

you long timers, and of course, our supporters, the boosters and the members out there. Thank you so much for listening to this week's episode of Linux Unplugged. We'll see you right back here next Sunday, as in Tuesday, on a Sunday.