Linus Torvalds, the top-level maintainer of the Linux kernel, has made a clear official stance that Linux is not an anti-AI project and that AI tools should be evaluated based on technical merit rather than social or philosophical objections. While acknowledging that AI tools cause pain points such as flooding security mailing lists with duplicate bug reports and generating hallucinated reports, Torvalds emphasizes that the solution is to make LLM tools help maintainers rather than ignore them. The Linux kernel community has implemented policies requiring human oversight when AI-generated bug reports are submitted, and maintainers are encouraged to use AI tools while being responsible for verifying their outputs.
Install our extension to search inside any video instantly.
Fork Around and Find Out
Added:[music] Hello friends and welcome back to your weekly Linux talk show. My name [music] is Chris. My name is Wes. And my name is Brent. Well, hello gentlemen. Coming up on the show today, Lionus put his foot down on AI in Linux kernel development.
We'll cover his blunt message he had for developers and then we're going to talk about reworking the note stack. Brent and Wes are coming at me with a whole new hot stack and then I'll do a quick remote Linux desktop showdown. I went through some of the more recent remote Linux desktop options. Came away with what I think is a clear winner. Then we're going to round it out with the great boost. I guess the boost the pics and more. So before we get there just sort of save me mumble room time appropriate greetings everybody. Hello, virtual dog.
>> Hello. Hello there. And hello, Brent.
>> Hello.
>> A strong crew today.
>> I'm gonna pull on them. I'm gonna get through it. I'm gonna get through it.
The other thing that's going to help me is my friends over at Defined Networking. Go to define.net/unplugged and meet managed Nebula from Defined Networking. It is the easiest way to build a fast, resilient, private network on the open- source Nebula platform. It is rockolid infrastructure. And now every new network includes a free managed lighthouse. So you could just get connected. your first machines. I think they're using IPv6, too. Real easy, smooth. And of course, you can still run your own light houses to have your own control, too. So, you can kind of take advantage of theirs if you want.
Use only yours or do a mix of both.
That's really powerful. Owning your own infrastructure at the network level is ultimately important and powerful. So, you can get started with 100 hosts for free. This is not like anybody else. No credit card required. Go to define.net/unplugged.
Check out a totally different way to do MeshVPN. Go to define.net/unplugged.
And a big thank you to Define for sponsoring the Unplugged program.
[music] Speaking of sponsors, we probably should have addressed this earlier, maybe a week or two ago, but this is sort of we don't really think often to talk about the mechanics of the show, but then we publish it and we go, "Oh, we should probably talk about this." So people that have been listening for the last few months know that we're experimenting with dynamic ads and we're kind of doing that in place of when we don't have an ad that's been directly sold like a a deal where I'm coming on and I'm doing the host read. They don't you know it it it my preference is the host read ads.
Um but the ad world has changed and dynamic is the way it goes because the buyers prefer it that way is the customers prefer it. So, what we're going to experiment with is shortening the pre-show or like the pre-roll that you hear if you get a dynamic ad. We're going to I think we're going to We haven't done this yet, but I think we should shorten that to 60 seconds. And then um Roger in [clears throat] the midroll ads, it'll be like ads that are in the show.
Those will be about 90 seconds because that's what the typical length is. And then there'll be ads at the end of the show. And it does help the show if you'd listen to those. However, if you are tired of ads, we do have the member version which offers the same fully produced show by Drew but with just no ads and you also get the bootleg if you like uh which typically doesn't have ads or it has like a shortened version of them. And um we're going to try to dial this in. We do like your feedback on this. We're paying attention and it's new. Like I >> Yeah, very new for us as well.
>> I resisted this for a very long time.
So, I'm now paying the tax of not really paying attention to this and I'm learning on the go on the fly. We we have been really kind of tight really since the beginning of the year, tighter than we really even talk about publicly.
And so, we're trying to find the right balance here to make this work. We're going to adjust some of it and um we encourage your feedback. We encourage you to listen because it does help the show or uh sign up for a membership if you're sick and tired of it. And uh if you have any questions, shoot it in and we'll try to answer. But that we just wanted to mention it. We probably should have mentioned a couple of weeks ago as we began experimenting more with this.
Uh and it's a it's a learning process.
So thank you everybody. We really haven't had any complaints or anything like that, but we just wanted to talk about it now before it became like an elephant in the room. And something that's exciting to see because I'm hoping we're going to make it is the Texas Linuxfest call for papers have reopened.
>> Yeah, that's right. Just a short little window here. Uh it'll close on July 22nd.
>> Oh yeah. Um, but if you missed the first one, you got a shot now. They have also released the first round of approved talks too, so you can kind of see what is already there if you need to calibrate off that. And specifically, they're looking for things in observability, automation, infrastructure, databases, or software development.
>> That's pretty great. I'm hoping we can make it. We don't really have all our plans in yet, but we always try to make that one.
>> Yeah, I believe it's at the beginning of November this year.
>> Okay. Okay. All right. Well, there you go.
>> Get that on your calendars and we'll try to join you.
So, this week, Lion has made a lot of news. Um, before we get into the bomb that he dropped on the Linux kernel mailing list this week, I think we should set some context because, you know, the dogs got to say something about it, right? Like, [laughter] I was on a roll, too. I was just getting going.
[clears throat] But I know Char feels very strongly about this and she doesn't want anybody interrupting the show. So, she's got to she's got to play security, right? She doesn't want anybody interrupting. Think about it. Um, all right. So, I'll I'll start again. Um, this week, Lionus made news with a bomb he dropped on the Linux kernel mailing list. And I think to really kind of wrap our head around this, we should go back a couple of weeks because what we're seeing right now are the headlines and the takeaways from this one post in the mailing list that has a lot of other stuff in it, too. But Lionus has recently been on a bit of a world tour actually uh sitting down with Durk and talking about his thoughts on AI. And what we have is a couple of series of clips that are over the last uh about 60 days. And you can actually hear Lionus' thoughts evolve in that 60 days. And this is about 60 days, the timeline before this big bomb that was dropped on the mailing list that everybody's talking about this week. So we wanted to go back to that because it started it really started with Lionus being a lot more kind of contemplative on the pain points of AI and all this. So he sits down with Dirk Hondle often at these open source summits and there was just a couple that happened recently and some of these clips our members have already heard in the bootleg because we played this a few weeks for them but now it's relevant to this week's episode. Plus there's been new stuff since then. So we're going to catch you up on all of it. So let's start with this. Sinus starts with what he thinks right now is the most the most um painful aspect of AI development.
>> Again, I have a hard time seeing everybody because I have the lights in my face.
>> But how many people here use actually AI for coding?
>> Yeah, pretty much every >> uh so [sighs and gasps] I have a lovehate relationship with AI.
I actually really like it from a technical angle. I love the tools. I find it very useful and interesting, but it is definitely causing pain points and and the big pain points in in Linux traditionally and I suspect in most projects have not been so much the code itself, but has been when you're forced to change how you work and people get into a rut. people know this is this is how I work and and uh this is how I like working and then something comes along and for me it was like long time ago around 2000 I had to change how I worked because I did not scale anymore for the project as Linux was growing and that was I still remember it being one of the more painful episodes in kernel development and it was literally 25 years ago. Um, and I think we're seeing some of the same effects now with AI where it kind of forces people to get out of their comfort zone. So, we've had a lot of social issues with the AI generated bug reports, but also code and and there's uh it's not a Lenus doesn't scale like it was 25 years ago. Now it's people don't necessarily scale and it takes a while to learn how to use the tools in a maybe responsible isn't the right word but in a way that actually works with the community and works with the other developers and we are definitely seeing some of the po pain one of the pain points recently has been the security mailing list where when people find bugs with AI I a lot of the time I mean whe a lot of bugs can be security issues and and I've said for years decades that a security bug is a bug but the other side of that is a bug is often a security bug and you may not even realize it [snorts] and and people think that when they find a bug with AI the first reaction seems to sometimes be let's send it to the security list because this may have security implications and the security list for the colonel was overrun by duplicate reports by tens hundreds of people running AI often the same AI tool >> and getting slightly different results often with the same underlying cause but we were flooded by people sending sending bugs and then you have this list with fairly few people on it because it's all supposed to be super secret and uh and we spent all our time just forwarding these reports to to other the other developers who knew that area better. And we made the policy change that basically if you find a security or any bug with AI, you should basically considered to be public and uh [clears throat] just because if you found it with AI, 100 other people also found it with AI. Uh and obviously that when it comes to things that really are security issues, you may not want to make the exploit public, right? Don't go and say to the world, hey, this is how.
And this, by the way, I want to make it very clear. This is not just about the Linux kernel. When you find an exploit for any piece of software, don't be that guy or that gal who then crows about it publicly and say, "Look, hey, I could bring down uh this big company or whatever." uh just let people know what the problem was and don't necessarily tell them exactly how to make somebody's life really miserable on a Friday afternoon, right? Uh so we are seeing some pain points with AI and uh and uh it's not that the technology is bad and I I've personally really enjoy it but but there's like the social issues that it causes have have yet to be all worked out. I would say >> the conversation kind of goes on they talk a bit more about some of the tooling. uh in a more recent conversation that happened uh after that uh Lionus touches on the he expands a little bit more on some of the pain points they've had with LM development.
>> I I want to point out for the record we made it 26 minutes into this conversation without talking about AI and LLM. I think >> but I think this will be the only talk in this conference that gets to that.
But of course we should talk about this because we did the same session or totally different but same setup session a month ago in Minneapolis. And one of the things that a lot of people picked up on was that on stage you said that you see LMS as just a tool which I completely agree with. And you made this interesting statement that you say it offers an incremental efficiency of a 10x and a lot of people said, "Oh, Linus thinks that LM's make everything 10x better." And so I was curious what that 10x was not scientific, >> right? That was pulled out of my ass number obviously just to clarify >> that was two questions down.
Okay. So, so this was a a intuitive number that you came up with to rephrase what you just said, but so how much of this true added value that LLMs provide? We just talked about this. The core code gets more eyeballs, but the LLMs can apply the same scrutiny, the same algorithm, the same logic testing ability to all of the code. How much do you think that the LLM really create productivity advancement for not you personally but for the Linux kernel community?
>> So right now we're at kind of the point where hopefully it creates more pro productivity than it takes away.
uh we certainly saw more junk being generated by LLMs than we saw useful code up until the like early this year basically and even now uh it's taking a lot of resources from developers because you get these bug reports I mean I don't know how many of you have seen this pattern in your own work. Uh, I bet it happens or I know for a fact that it happens outside the kernel too where you get bug reports that look entirely valid and it actually can take quite a lot of effort to figure out that no that was that was just hallucination.
And uh and it can actually be a huge drain on resources when it takes humans a lot of effort to figure out that hey this machine generated report was not true.
>> And so Lionus goes on to say that he does think there is some long-term good.
Uh and he's he kind of articulated this uh in the previous chat as well. Don't >> get me wrong, this may sound negative to a lot of people, but I actually I I think it's all very good. uh when AI finds a bug in any source code uh that is a short-term pain but long term is you found a bug we fixed it the the end result is better for it so I'm actually this is part of why I actually really enjoy AI and again I want to stress that the conflict is not that AI is bad the conflict is then that there are some social choke points and social pain points that that come with this new tool that we have 35 years of code that it sometimes finds issues that we did not find >> and and it is going to take us some time to to just work through the the new issues that are found. And so I'm actually very positive about this whole thing. I think finding bugs is great because the real problem is all the bugs you didn't find.
>> I also picked up that you seem to note that there has been some improvements in some of the bigger and more important areas.
>> At the same time, what has definitely happened in the last couple of months is that um we are getting a lot of really good bug reports. Uh most of the good ones require more than just the LLM though.
>> [clears throat] >> They they require a person that then goes through the report that was generated by the AI and make sure it was valid and then also acts as the gobetween between the developers who fix the problem. And we've we've had to push back quite a bit on on people in the sense that if you find a bug with an LLM, it's not enough to just ask the LLM to to make a bug report and then throw it over the the fence to us. Uh we want to see a suggested patch. We want to see the human who ran the LLM act as a kind of back and forth that we have somebody who we can talk to and and discuss the issues with because sometimes the patch is a simple oneliner that is obvious.
But quite often it ends up being hey we didn't think about that issue and we will need to reorganize some code and and fixing this properly is going to take a bit of effort. And then you really want to have the person who sent the original report involved in the developers coming back and saying, "Hey, [clears throat] this is a suggested fix. What does your LLM think about this?"
>> Okay, I think that gives us kind of a more expanded view on Linus' thoughts around AI. So everything kind of shifted on July 14th in a Linux media mailing list discussion and uh it wasn't like a like a discussion on a conventional release or anything like that. It was um I guess a thread that concerned connecting a patchwork of open- source reviews like a different a set of that have been like agentically reviewed like a whole patch set.
>> Yeah. So they're connecting an ex some existing tools, right? So there's a bunch of tools in the kernel so that when you submit a patch there's automated review of it. Tests get run, things get built and then you get feedback sent to you and also to the mailing list maybe to the maintainer that says like hey the program found these things you know kind of like linting or running a basic set of unit test or something you know various versions of analyzing the software automatically to see if like it is you know before a human even has to get to it to try to lower the maintenance burden. Okay, >> especially when you consider like Linux tree maintainers, their job is ingesting a giant stream of potential patches and trying to make sure that they're doing so in, you know, a well done, rigorous >> way. And so shash sashiko >> sashiko. Okay.
>> Yeah. So it's not a proposal. It's operational already. It's just this is more about the more plumbing on the like actual you know connecting it more to the ex like >> making it more permanent fixture of the ongoing development process. Right now it's been a tool sort of more on the side >> and so this is a discussion around that.
Yes.
>> And Lionus decides to chime it.
>> Well, it starts um it starts because people are talking about let's add this to the thing and then someone um >> uh objects.
>> Yes.
>> Yeah.
>> So, uh Laurent uh a developer uh comes in to and we can uh got the dudes here, but uh >> yeah, developer comes in to basically say like, "Shouldn't this be an opt-in thing? Isn't that the normal default of what we do?" And then that gets into a whole discussion around is that a viable model and then ethical concerns get brought in and it becomes a question of you know if I if if I don't want to be made I don't want to be forced to use an AI tool and then the whole debate expands around there.
>> Yeah. And then that's when Lionus jumps in.
>> Yes.
>> Yeah. Linus writes, "I realize some people really dislike AI, but this is an area where I'm willing to absolutely put my foot down as the top level maintainer. Linux is not one of those anti-AI projects. And if somebody has issues with that, they can do the open- source thing and fork it or just walk away. AI is a tool and like other tools we use and it's clearly a useful one. It may not have been that clearly even just a year ago, but that no longer is in question today. There are other questions around AI like what the economy of it will actually look like in the end. But quote, "Is it still useful?" That's no longer one of those questions. Anybody who doubts that clearly hasn't actually used it. Yes, it can also be somewhat of a painful tool both for maintainer workloads and from it just keeps finding embarrassing bugs standpoint. But the solution is not to put your head in the sand and sing la I can't hear you at the top of your voice like some people seem to do. The solution is to make sure those LLM tools help maintainers. That's his italics.
Help maintainers instead of just causing them pain. There is no question on that side. We're not forcing anyone to use it, but I will very loudly ignore people who try to argue against other people using it. And no, AI isn't perfect, but Christ, anybody who points to the problems with AI had better be looking in the mirror and pointing at themselves at the same time because it's not like the natural intelligence is always all that great either.
>> The Colonel project has been and will continue to be about the technology.
Sure, the social angle of working on open source is important and often very much a motivating part of the project.
But in the end, it is a side benefit, not the point of the project. Again, his emphasis, not the point of the project.
This is not his emphasis. Some kind of social warrior project. Never has been, never will be. In the colonel community, we do open source because it results in better technology, not because of religious reasons. And so we make decisions primarily based on technical merit, not fear of new tools.
That is a pretty strong clear stance.
What's your reaction to that, Brent? Is that stronger than you expected after hearing the comments from Lionus?
>> Well, I think it's more clearly laid out here. Like I I I found in his clips on stage, he was a bit more exploring with the crowd, like saying, "Oh, yeah, yeah, we're using it. These are the experiences the project's having." But here he clearly states both his personal kind of view but also just a very clear stance from the project and I think a stance that you haven't seen very often in other projects >> especially not you know written in Lionus like words as he does. Um, but the thing that really hit me here is one of these last sentences.
In the kernel community, we do open source because it results in better technology, not because of religious reasons. And we've seen a couple projects in the last few years fall victim, maybe that's a strong word, but fall victim to some infighting in the community just because of differences in philosophy of how the project should be run.
So in a way having Lina say yes this is exactly how it is and this is how the project believes uh this should be implemented is a very clear voice about you know how everyone should be following and there's not much room for uh pushing back I think yeah we've seen Lionus take uh an I don't know what the right but we've seen Linus take a a technical over um religion argument when it came to GPL2 versus GPL3 Mhm.
>> It's been a consistent position for him for for a long time.
>> Wes, what is your reaction here to really framing this as look, the usefulness is proved. If you don't think these things are useful now, then you're just clearly not using them. Um, >> it came up again in the thread. Some folks push back on or, you know, we're saying like, you know, we shouldn't compare these things. You know, these existing tools we have are deterministic. You can debug it. These LLMs are a black box and they can hallucinate. And then Lionus agreed that they were different, but he kind of flips it and he says, "You are correct.
The AI generated reports are usually much superior and more flexible. They actually find real non-trivial problems."
>> Yeah.
>> So there's sort of even more. Um I will say it's worth noting that like this is a big thread with lots of maintainers.
It's not just these folks, right? So there's um like there's uh Roman Gushkin who's a Google kernel engineer who built Shashiko and was sort of, you know, working on it. Lauren Pinchart who sort of started the objection part of this a Linux media contributor um and then Lionus but Ted you know a whole bunch of regular kernel contributors are in there and that's where it's becoming you know like they've been talking about around does does this particular tool work even um about a 20% false positive rate uh 80 to 90% true positives on real patches um and only two maintainers seemingly in the whole thread reported real output problems. Everyone else seems to is using it and likes it. Ted, Steven Rostet, John Cara, Bottomley, Bonzini, they all say it catches real bugs and that the noise is manageable. So we're seeing, you know, it's important to remember that the colonel is like Lionus and then a whole bunch of Lieutenant Ma maintainers, right? And so what's happening is different maintainers are already enabling this and thinking that they may have different rules around do you have to use the tool or not but they are already seemingly to commit that like this is worth it for them as a tool that they want to see used in the code they're maintaining. Yeah, that decision's been made. Um, and you know, so the debate becomes, should we be using AI assisted development in free software? Um, and Lionus had an analogy.
He had a vegetarian analogy. He says being vegan because you think meat is murder is fine, but vegans don't expect the rest of the kernel community to stop eating meat.
>> I think it comes down to like I can understand the position of I don't want to be forced to use this tool that I disagree with. Right.
>> Yeah. And it's not mandatory. And well, so that's where what use I think matters a lot in the structure of what is happening here, right? So it's like I think folks can have sympathy for your boss makes you use a tool and you're supposed to hit a velocity target and you got to ship ship, you know, like there's all kinds of stupid stuff that is just But um in in this case, that's not not what's happening, right? Like what's happening is an automated tool that is open source. I don't know what inference provider they're using or whatever, but it's a Rust tool. It's open source. all of the prompts are there that you know you can the normal open source process sort of applies there >> and then it's finding to at least by the people that are going to ultimately decide if your code is accepted real bugs that humans don't find that you know are quality and then it's sending it to you. So it's not you have to use it to make your code.
>> It's not you have to have a certain tool on your system that you authored the commit with. It's >> you may be forced by the maintainer to respond to the output of this tool. Yes.
>> I don't want to have to sort through this slop that has stuff that's false positives and I can understand that.
>> But the flip side is where it gets harder because what happens if the tool finds a real bug that you now refuse to engage with, >> which is going to become more and more likely as they improve, as they get better, >> which I think connects to Lionus's point. Firstly, I think folks like there is very much a sort of a moral and a social and an altruistic sense to open source. I don't want to negate that at all. That's some of the stuff that drew me me to it. But there is also this technical side, but it's also sort of almost a game theory side, right? It's like a strategic bet on how to build software that you yourself need, right?
Like Google and Microsoft are not altruistic entities. They're doing it because there's a strategic thing and that's sort of a bet and that ties into sort of it's also a methodology to develop software as much as it is a philosophy around software and that is the side I think Linux is trying or Lionus is trying to uh articulate and then he's also saying that we have methods to adjudicate things like you know the kernel community does have a code of conduct now so we can sort of talk about intrapersonal conflict things like that and of course there's the there's the whole system system of maintainership is set up to adjudicate code quality on that sense, but there is no system in place to evaluate individual ethical claims.
>> Once once that's in there, like how do you >> that's an individual thing.
>> There is no system in place to do that.
There's colonel is not set up to do that. So we can't >> right >> it that that's where a lot of this is coming from. I think it's it's really a mechanism thing.
So when you look at the data, go ahead, Brent.
>> Touching on that, it makes me remember uh to remember a bunch of uh Russian nationality >> I was just thinking about >> developers were booted off the project for >> kind of shady reasons.
>> I just as the sanctions or whatever went in during the the Ukraine war.
>> Mhm.
>> Yeah. And we were kind of arguing, well, maybe it's not the project's place to say who can and can't contribute uh because it should fall on more technical merits than it should on social ones.
This this is another example of there's these sort of social lines. Yeah.
>> And that one I think you could argue was partially imposed on them, right?
>> In a sense.
>> Mhm. So there's some research out there that shows that going by numbers that are reported by developers. I think it's interesting that there is sort of this false positive rate. It's like an LM thinks it found a bug.
>> There's also in this specific tool a case where because it sees patches, sometimes it might find real bugs but in code that wasn't what you just added or touched.
>> So there's some noise that makes sense to Yeah.
>> But it looks like like a lot of like the so the rate of is about a 20% false positive. But what is interesting is this like what they're developing like what the developers call like gray zone nitpicky small theoretical stuff that nobody really kind of tends to care about. And when you look at the stuff that's like the real important stuff, the hit rate is more like 80 90% of true positives for the actual critical things that get submitted. So it's an interesting thing is if you look at what's coming in according to self-reported numbers by these developers, the stuff that is important tends to be accurate and the stupid stuff that's just like lowhanging stuff tends to be where the slop is at.
>> You can tell they put >> challenge is really separating the two, isn't it? and tuning it and seeing if it'll be interesting because this will in effect provide some real data out there if you want to mine it to see how this changes over time. Uh I did think there was some signal here in just seeing that like there you know um Chris Mason the butterfs developer has been contributing to some of the the like the prompts that are getting used here and then um Jonathan Corbett at LWN took a look and called them some of the like the the prompts that get fed into this 11stage super it's it's quite the process um called them some of the best documentation on Colonel Patch review ever written. He he expressed a little, you know, sadness um around it >> around that like we couldn't write it until we were writing it for a machine.
>> Uh but seem to think that there was a lot of really good stuff in there.
>> Huh. That is funny how that works, huh?
>> Here's the question I have for anyone in the audience who is on the side of we shouldn't be using AI in free software.
I'd like to hear from you because I think a that's a point that should get voiced more, but b I'd like to know where you stand on if if an LLM assisted tool finds a legitimate security issue, submits it to a project that is anti-AI, anti-LM, what should that project do? Should they just ignore the security patch? Should they go try to recreate it from scratch?
That's a ton of work. That's probably more work than going through their submission. So, I'd like to know if you're out there and you you're not really pro- AI use and free software, how would you solve that little particular problem that's obviously going to come up, what do you do there?
What happens if a real bug is found?
Somebody sends in a patch that's AI assisted or LM generated and that project, do they refuse to engage with it? What should they do? I don't know what the options there are for them because that seems like that's going to be a thing is projects that close the door to this, two things happen. Hm.
[clears throat] Number one, they're going to get folks that are submitting bug reports that may be actually useful in some cases that they're going to ignore.
But number two, and I've talked about this before on the show, and I'll just so I'll make it brief, is they're not going to participate in the dialogue.
And the state of free and open-source AI is at a critical junction right now.
>> Sure is.
>> And it could really use our community out there advocating for what we believe in and what we want to see. And instead, a lot of our community is, like Lionus says, putting their heads in the sand.
As a result, a lot of money from Anthropic and Open AI and others has knives out for open-source AI. And we're going to see a concentrated effort as the open source AI models catch up to the frontier models, quote unquote. They have such a financial position to protect that they're going to turn to lobbying efforts and things like that. And it's going to be like the '9s again with free software, but now in the intelligence field. And only this time, we don't have a stallman out there. We don't have a free software foundation out there for AI. In fact, for the most case, they're all just ignoring and rejecting it. And so, ironically, they're going to end up with worst case scenario AI because they're not participating in the dialogue. And I I just I've said it before so I don't need to go on about it, but I I hope I hope some people have be have been realized or come to realize this since Lionus made that post because I think what Lionus is trying to at least saying for the Linux kernel is the decision is made. These are useful. Now it's on you if you want to use it or not, but from a project standpoint, we're not debating the usefulness anymore. It's a tool and our sovereign in sovereign individual developers on their own prerogative are implementing and using these tools because they want to and that is now the official state the official position of the Linux kernel. You have to wonder if that does move some people and maybe get some people more actively involved in the free software side of AI. I do think I even saw in that mailing list thread some folks saying, you know, like I had folks who had their own ethical concerns but also sort of were forced to reach a conclusion of like well at this point, you know, not using it is worse than using it even with those concerns.
>> It's like that that game theory scenario.
>> Yeah, >> I know it sucks. It's the welcome to the world of technology is this is not the first time we've been here.
>> Oh. Oh. Well, what's a little dynamic ad spot? Do we have anything? You want to keep moving? You ready to sort of set us up and then you boys jump in?
>> Yeah.
>> You got swag and stickers and stuff if you like Apple.
>> That's true. I think we Yeah, I think we're probably going to pull that soon, right? Oh, it's when 72 ships. Yeah, we should mention that at the end of the show cuz this is all getting cut.
[laughter] >> All right, here we go. Coming back.
>> That's true.
>> Thank you, Drew. Coming back in a two.
>> Well, that's strange. The title of this little section says notes with Wes and print. [laughter] Hey, >> and that's because it's been a what we say multi-week journey of reassessing what notes mean to us in our life and times.
>> Oh boy.
>> And we touched on this a little bit in the members feed last week and I think it went into a great bit more detail than we even expected. We were just like I think Chris was like, "Hey, I heard you talking about notes recently." And then there went 10 minutes worth of really interesting discussion. So, but it's been another week and Wes, you've been diving into notes. I've been diving even more into my notes. So, where should we start? Should we give like maybe a detailed summary of what we were talking about last week? I wanted to start if you guys don't mind because I think the real story is why this all started was because you had a Next Cloud disaster >> and I also recently experienced a Nextcloud disaster. [clears throat] So, I thought maybe you go first. Oh, mine really quickly. Uh, I was viewing a very important text file. I clicked it just to view it in the text editor. I didn't edit, didn't save, and it wiped the file out just by viewing it. And then it immediately closed and saved that. And it was a very important file. I had it in backup, but just viewing it wiped it out. It turned out to be like a little like um a little bug where like, you know, basically >> some kind of race condition.
>> A race condition. Yeah. And um wiped out one of my most important notes. And I realized right then and there, I'm not using Next Club for notes anymore. But I want to know what happened to you. And I think cuz that is what caused you to spiral into redoing your entire note setup. [laughter] Oh my gosh, [clears throat] why do I do this? Uh, so I've been using Nexcloud. I looked at the calendar for about like something like 10 years now and notes for that amount of time. and it's been the number one application I would say definitely on Nexcloud that has been important for me over the years but even more broadly it's probably the ecosystem I use the most every single day so if you took away a bunch of tools of mine it notes would be the one thing that like is a capture of my knowledge the place where I think about things the place where I keep you know tricks and tips and stuff so some people don't use notes as much is that but I'm a note fiend so you know kind of like tabs and uh what I have noticed happening recently is a slight erosion of my trust in the sinking process of my notes over the last many months I would say um but it's just like keeps suggesting so it wasn't as drastic as your experience Chris but for me it's these little tiny paper cuts and these little micro warnings I kind of see them as because I would have like one note over here that' be like, "Oh, I thought I wrote a little detail there and it's kind of missing on my laptop and then I go to use my phone and it, you know, I'm out in the world and I want to refer to something that I left on there 30 minutes ago and all of a sudden it's not syncing and it hasn't synced for many hours." And so the things I was relying on having with me at all times are >> just kind of not there. And so you got to find some other way to get it. Um, so I had enough of those experiences that my otherwise deep trust in this note system, as PJ will definitely um, back me up on, has I've been trying to convince him to use this thing forever because I trusted so much in it, has now been eroded to the point where I was like, well, I think I have to solve this a different way because I I looked for, you know, was it a config set? Was it some weird files I had in my infrastructure that are causing problems? Is it a bug I can report or something like that? And for months, I've been looking and trying to try different methods and >> play around and I came to no clear explanation. So, >> you really given this a shot? You were trying not to move.
>> Well, 10 years.
>> Yeah. No, I mean for 10 years and all of a sudden, you know, a key binding goes away and you just can't live without it.
>> Mhm. So, so this is after a long journey how you reluctantly ended up with Obsidian.
[laughter] >> Don't give all the sauce away, right?
And and I don't even know if that's the final landing.
>> We'll get there. But you started flirting with Obsidian. You were a little nervous because it's not free software, but it is Markdown Bay. So, you taking a look at it.
>> It's Canadian. So, >> yeah. All right.
>> Am I right? Uh I I was basically looking for a couple things maybe absolute needs. One is markdown backed because I I I can't do databases. I don't know how those work.
>> Um but also I'm using it on various devices, right? And I want to use Kate on my desktop and I want to use something that's just like fairly fast on the phone. So it has to have a phone >> right >> phone fast way to use it. that that's super important for me. Maybe like 50% of the importance. Um, but also like a really good search that's surprisingly difficult to find in some of these apps.
So, I scour the landscape trying to find, you know, something that's open source, something that's quite mature, that's still developed actively, something that, uh, you know, hits all these other points. And, um, once you put all those filters on and there's not much out there. And turns out Obsidian kind of falls in that category quite nicely and I had hesitated to even look at it or even consider it. Meanwhile, everybody else is raving about it. So, I reluctantly >> Everybody's doing it.
>> Everybody's doing it.
>> I know. Well, I contrarian, you I don't want to be everybody, but turns out, >> you know, they're kind of right.
>> I mean, there's a reason why people like it. However, it did provoke feedback from listener Kai, right?
>> Indeed. Dear Love Team and specifically, dear Brent, [laughter] >> I'm a semi longtime listener and core contributor. Thank you.
>> Thanks.
>> I just finished finished episodes 675 where you were considering Log and Obsidian. But it sounds like you've missed that Logique has moved away from Markdown files and are switching to a graph database approach.
>> Oh, >> the Log [clears throat] Markdown version, OG as they call it, will only receive security updates in the future.
Oh, I am currently also reconsidering my note-taking application. Obsidian just doesn't feel right for me. And I currently I'm I'm currently taking a closer look at Silver Bullet. The beauty, it runs on your server in a centralized place, and you access it from your devices by connecting to the server. The files are cached and synced locally, so you can use it even when there is no internet connection, though they are starting to work on local applications as well. I haven't tested it in depth yet, though I believe that you and I share the same taste. So maybe this is something you could check out as well.
>> You know, Wes, I believe you and I came across Silver Bullet not too long ago.
>> Yes. In our kind of looking around for various note-taking applications for show notes things.
>> Yeah, for show notes. We were looking at this as a potential way to do our show notes.
>> Yeah. So it's MIT licensed, server hosted, browser based. Your data stays on your machine, is plain Markdown files, live preview, markdown editor.
>> I'm liking this. birectional links, a built-in query language, SLIQ, and a Louis scripting environment that lets you program your notes, which is very cool. Run queries, build widgets, automate workflows, all from inside your pages. And they the pages or the plugins run in sandboxed web workers, so you can have extensions uh that can be secure without sacrificing power.
>> Okay.
All right.
>> So, it to me it seems a little complicated, but maybe not when you actually look at what it's doing. Silver bullet. Uh, by the way, silverbullet.mmd if you'd like to check it out, listener.
>> Ah, that's clever.
>> Yeah, Wes, >> I go ahead, Brett.
>> I kind of felt it landed somewhere in the middle. Like, it it's extremely powerful if you know what you're doing.
You don't have to use those features.
Um, it has some nice easy wins like birectional wiki links and those kind of things that I haven't been using that were kind of like on my mandatory list even though that's not what I was using.
>> Built in to-do items is nice too.
>> Yeah, it has all these like well thoughtout quick like access features.
Uh there's there's a nice video where um one [clears throat] of the main developers goes through kind of a tour and uh you could tell they're passionate about it and there's some cool stuff. Um I thought of Wes right away with some of the uh power of using like scripting languages specifically to just make your notes programmatic. And that seems like a wonderful thing for us nerds around here in this room.
Um, >> yeah, it almost makes it closer to something like um like a Jupyter notebook or something, which is interesting.
>> Reminded me a lot of uh textbased accounting.
>> Mhm.
>> I like that comparison.
>> Okay, so here's >> I found what I want to know, Brandt, is what is it not? Well, it doesn't sound like it it worked for you. It doesn't sound clearly not using it right there on that.
>> There's like a big butt hanging here and I want to know what it is. [laughter] Well, the big one here is >> but it's only a web app. Like it runs in a browser.
>> And you want a mobile app. You want you want a mobile app, too.
>> Yeah.
>> He's a big mobile app guy. You know that.
>> And you know, is this the kind of thing people pin? Is this the >> Yeah, I guess you could you could you could pin a tab on that. Yeah, you could definitely pin a tab on that.
>> He'd [music] rather open new windows like some kind of entropy enthusiast.
Um, now I what I'm wondering is are you willing to hear the log sec sales pitch because I think Wes, you've stuck with it, right?
>> Oh, yeah.
>> That's saying something because it's been a couple of years. It's been a few years. Yeah.
>> I don't use it for like absolutely everything, but yeah, I got I got some some real data in there.
>> So, Bren, I mean, would you could we be interested in putting you in a nice log today? Uh, >> I also thought it would be a good um just because there is there is a big change going. I thought it was just a time to check in as well. Like, you know, it's not up front. It's not going to be for everyone. And it's a different it's a very different model. But >> can you tell me about the change in your words and what you think about this this database change they're doing?
>> Yeah, so it's it's one way to think about it is it's more of almost like less of an upgrade and more of a different product.
>> Yeah, it seems like it.
>> Mhm. It's a block level graph database with typed properties. You get data log queries and an AI agent API. It also happens to be a note-taking app. Oh, >> okay. I see. So they're they're pivoting for the agentic future. Well, I mean in some sense, but this was sort of always an evolution because like in the Obsidian model, uh just as an example, not but just sort of like you inherit a hierarchy from the operating system, right?
>> Yeah.
>> You're sort of like you have a folder tree, >> you have markdown in there, you probably have it all organized.
>> Yeah. It's all folders and whatnot.
>> Yeah. And >> which I kind of I do like to some degree.
>> Yeah. That's not that's that's not bad.
I'm not saying anything's wrong with that. Um but this is a different way to think about how you might want to organize things. And so, um, the old version of Log sort of was a mix where it it had markdown files on disk, but it was never a great fit for really what the best like representation and use of what Log SQL was good for was, I thought. So, now with the Log C DB version, the SQLite database, okay, >> I would also argue I get why people don't like databases for storing stuff.
If you are using agents, I think that's perhaps less of a concern because they're pretty good at getting data in and out of things and they have no problem manipulating SQLite for you. Um, so or turning a SQLite dump back into tech structured text files or things like that. But you don't you are no longer a prisoner of the file system. Pages are UUID addressed entities with titles, no folders, no paths. Um, basically the hierarchy is at the block level including create parent child but then any other hierarchy you want. You're not limited to a tree. You build whatever graph you want with the graph primitives that are built right in.
>> Huh.
>> And then of course it does come with an MCP. It comes with the CLI.
>> Okay.
>> Um, of course it has really nice search.
Plus it has um the query language. It includes data log lets you do stuff. So like if you want to write a query that isn't supported by how however the search works as it stands now you have full access to like the underlying data model um it does provide a readonly markdown mirror.
>> Oh that's interesting.
>> So like you don't have to worry about that side of it. You can't edit those files and have it sync back.
>> Oh wow.
>> Um but you can totally read Yes.
>> Yeah you could view. Yeah.
>> It has a mobile app.
>> Yep. Mobile app that's new. It's all very new. Um so a lot of the biggest critiques right now is like it's moving fast. It's a big change and I think maybe the folks behind it haven't super articulated what I think is kind of the exciting vision for for where it could be going.
>> It seems notable that uh in May they raised $4.1 million.
>> Yeah.
>> Wow.
>> So I think they're going to be around for a minute working on this. Um and that knowledge graph seems to be the thing that got them the funding.
>> It is AGPL. Uh so you can pay them $5 a month if you want them to handle the sync. Uh but you can host it yourself.
>> That's pretty reasonable. Um, and the new stuff, which I haven't tried this yet, but the new sync is real time sync.
>> Oh.
Huh.
>> That is really nice. It does come up when I'm moving between machines sometimes.
>> And it is super new. So like some folks with super giant libraries have had some issues importing them, but it does support importing from the previous version.
>> Um.
>> Okay.
>> Yeah. So I think it's sort of like as we were talking about in the in the pre pre the members show we some of the biggest maybe one of the competitors to a personal knowledge system is like the fact that if you have something like Hermes it's building a thing that's like has markdown files that it's indexing to search and injecting and like you know a lot of these agent memory systems or QMDs or whatever are sort of half of >> part of like you know building structure on top of markdown >> and so I think where this is interesting is you log seek is presents another point on that continuum especially like I was trying out last night and the the MCP is really nice it's very well done it's not perfect yet but it has pretty much everything you need and so you can read the files if you want you can read it and do searches through the MCP it's just that the thing that the main thing that changes is the right path goes through the database >> yeah which is I mean for a long-term tool probably probably not a bad idea also I noticed they have a CLI as well >> CLI yep client and a Windows. Do they have a Windows? Yeah. Yeah, they have a they have a Windows client, I think.
Okay. I I got a couple questions here, Wes, since you >> over the years seem to have really >> connected with this ethos, if I say that.
>> What do you find is the perfect kind of information to store in this? Cuz I realized when we were having these discussions about like, well, what is notes even, right? That I was trying to store very different types of information in the same place to have accessible in the same ways. And maybe That's not a very good idea. So perhaps there's a way to store the right kind of information in something like log and store the other kind of information somewhere else.
>> Oh boy, that sounds complicated, Brandt.
>> I know that's why I was leaning away from it. But Wes seems to think this is really good for something. So >> why not something?
>> Well, I think I tend to keep like some stuff like per project stuff tends to live in like a repo that I have for a project. So, I mean, not that there's not notes or thoughts or ideas about that that might be in there, but it tends to be that kind of stuff. It's sort of it's ideas, maybe things I want to investigate, you know, things I've learned about stuff, notes for a future me going through or redoing something.
And I find that I what is really nice is that you can just enter information and then start build adding links to it and like you don't it's very easy to go navigate that. So I don't worry as much about having a strict hierarchy because it was it always felt like I had part of what limited me was classifying it up front like that. And this means I don't really have to. I can just put information in and sort of the structure emerges.
>> Yeah. I do sometimes spend a lot of time figuring out okay where does this live?
What do Yeah. Like classifying it is Yeah. But but I I wondered if like there's some kind of upfront discipline every time you're making a note to try to tag it or tag a block or like can you explain a little bit of the >> I don't know >> like hygiene to make sure the notes actually link and stuff.
>> Yeah, >> that's the thing I need to learn to make this work well.
>> Well, uh I I think you just kind of play with it. Like you could have an orphaned um thing, but you know there's autocomplete on the tags. So, as long as you have as long as you your variance in the tags that you do isn't so crazy and then you can kind of go check and especially now if you had an agent in there, you could go do some sort of a quick audit.
>> Can I ask you about that, Brent?
>> Is that is your intention to have an agent have access or edit these? Well, when we first started this discussion, it wasn't. I think partly because I wanted to have some protected space where this was like blessed information that I had, you know, cross reference with documentation or had used in anger and was like the golden command or whatever it is that solved that problem.
And I feared that letting an agent kind of have access to that would mix the information in a way that I felt was impure somehow. Um, but you guys challenged me and so I've been using agents like crazy to make a like a Carpathy LLM wiki thing and I've been freaking why did you guys get me into this? So for weeks now been playing with this and it's totally amazing. So I'm starting to change that.
So with that consideration, to me it seems like Logac looked at what Obsidian is doing for agents and said, "Oh, we can we can do better than that." I mean, this is with the M see I'm thinking with an MCP and a CLI, you really could truly brain dump and then have the agent start connecting the dots, right? They're actually really good at that kind of context stuff. So you could have it postfacto and you know, obviously propose it before it makes the changes, but postfactor you could say go through and link things. And I've done this.
That's why that's why I can speak to this is I wasn't very good at linking my stuff together like you know some people they go into like Obsidian they have these awesome like I don't know like point graphs or whatever like this incredible cluster of notes mine are just like dots all disconnected >> just like stars in space.
>> Yeah.
>> And so you know I had I was very bad at that but the agents are able to go through and very effectively you know notice the commonalities and link the things together.
>> They're like your librarian almost. I I had a thought on your question too about the like the tagging versus or and as it relates to directories. I think so one version is you can just do the same thing where like you were going to say which where where in the tree does it go could be which is the prim which is the first tag I add. But then the thing you don't have to do is fight about the relative importance. You don't have to be like oh well should it be in this one or this one because you just add both.
[laughter] >> Okay. Yeah. And then the other thing too is like it's sort of nice with if you have an agent in front of it, you could capture notes in Telegram, you could capture them in Kate, wherever you want and then >> the agent could go through and save them, you know, using this database, this block system, the agent could files it for you and then retrieves it for you. The one limitation right now is there isn't a headless MCP, so it requires the window to be running, but there's no I was taking a look at the code and there doesn't seem to be any technical limitation there. It just isn't plumbed up that way yet.
So check on Wes's GitHub to see next week if we got something [laughter] to consider. I mean, you've got you got me thinking about it. So >> like it might not be for anyone, but I think it's it's a fun way to at least make you rethink some of uh how you think about and relate to the knowledge you're trying to store.
>> Yeah. And if you're looking for a knowledge system, that's Yeah, that is all right. So log L O G S EQ. boots, but it definitely now it maybe also makes it a little clearer because it's definitely now not decisively not competing for I have an existing markdown infrastructure that I want a thing that can upgrade it to a graph that is they're totally separate now. So, okay. Uh we uh we are going to uh cut this out but uh should we come back?
Which one of you would like to take Mike? It's kind of a long one. Do you want to take it Wes?
>> Sure.
>> Since we were just doing a lot of Okay.
All right. All right, we'll come back with Wes in a three and a two.
And we've got some feedback from our dear Olympia Mike. Howdy JB team. I've been quiet since Linuxfest Northwest, but definitely not idle. Lots of stuff has been happening thanks largely to people in this community.
>> First, Nixbook 2605 is officially out with a new installer.
>> Hey, new installer is cool. Yeah, flakes. It's a good question, but um and it that inherited upstream changes that allows new users to choose their file system on install. Hey.
>> Wow.
>> And it has a progress bar that doesn't appear to get stuck at 46%. That's a big upgrade.
>> That's nice. Yeah, I like that.
>> In Nixbook itself, the channel Okay, that answers it. Uh has been bumped to 2605, a new version of Cinnamon, a collection of default free wallpapers, and fixes to Etsy scale, so multi-user systems have more consistency. That's a nice touch. Good. This has all been aggressively tested and working well. If you have a Nyxbook, you'll automatically get this within a week. Or if you want it now, you can just run upgrade and reboot from your start menu.
>> Love that.
>> So cool.
>> Wow. The new installer can be downloaded from nixbooks.org.
>> Also, the Computer Upcycle Project has officially won its first grant as well.
The city of Olympia has chosen the computer upcycle project as one of 60 nonprofits to fund this year and awarded us $15,000.
>> Congratulations.
>> Amazing.
>> Very welld deserved. This will be used to purchase chargers, SSDs, and branded laptop bags along with continuing to fund the development of Nextbook. We're also committed to hosting monthly workshops in Olympia aimed at the tech curious and youth to show them hands-on and how to take that old laptop and turn it into a Nextbook. All of this makes me super excited and happy about the future of Linux and getting more people involved and aware of what's possible.
And of course, I'll urge them all to be listeners of this year podcast.
>> Thank you, Mike.
>> Thanks again to everyone who has donated, supported, or simply cheered on Nickbook and the nonprofit. This community really does make anything possible.
>> That's really great. So sweet. It feels with the price of hardware right now more important than ever that there's something like this out there. And I wonder if it could be a model for other areas because you know Mike's doing a lot, but it' be great to see if like another organization like this could pop up say like on the east coast and then pop up in the UK and pop up in Canada, you know, you know, just sort of sprout.
>> I think that'd be really neat to see.
But in the meantime, Olympia Mike is doing incredible work and I do appreciate the update.
>> We sure do.
>> It's nice to hear. Thank you, sir. And uh keep up the good work and keep us posted on it. And with that, gentlemen, let's do a couple of boosts.
[music] And now it is time for the boost. And Greg the lawyer is back. He's our baller booster with 100,000 sets.
>> Wow.
[music] >> That's not possible. Nothing can do that.
>> Well, Greg the lawyer did. He said, "Love the discussion on making a clanker OS. I've been using Claude to maintain my flake for a while and it's nearly flawless. I now get a dedicated Hermes profile set up for the task. My Hermes runs on a Debian VM. I would love a crank a Clanker stream on how you have yours running on Nyx.
>> Uhhuh.
>> I really combo stream.
>> I feel like I missed the boat. I set up just the most amazing EV camping routing. Like, you know, all the years I've been road tripping, I've been collecting like what my idea of a perfect app would be. you know, wildfire, cellular signal, closures, everything I've ever wanted to know in one thing. And I finally built it. I should have streamed it. Should have streamed it. I could stream a tour.
>> I feel when we maybe get back and we get our feet underneath us, Greg, I I feel like it's going to be time for another Clanker stream. We've probably gone too far, but we just we got stuff coming up.
We're doing a little trip to Canadia up to Brent Landia. So, when we get back from that, I think >> we got to pin some tabs.
>> We're going to start cooking. We're going to pin some tabs and we're going to start cooking on a Clanker stream.
That's the mission. Thank you, sir.
Appreciate that baller boost.
>> BTC is my 401k boost in with 3,998 sats.
>> I like you. You're a hot ticket.
>> Oh no. Seems my lightning node can't reach Brent or the JB node. But here's some more for you.
>> And a catchup boost to provide value while my albby hub was down and streaming stats were depleted.
>> Well, thank you.
>> So, uh, BTC, I saw that boost coming live. I was like, oh, what's going on?
Went over. Yeah, tour was totally down and that node's on tour. Totally offline. A little hard.
>> Yeah. Like, [laughter] oh, every like I wasn't monitoring tour. I was monitoring the lightning demon. I was monitoring the Bitcoin demon. I was monitoring the booth.
>> Not monitoring tour. So, I got that fixed. Thank you very much. BTC is my 401k.
Odyssey. Wester's here with 2550 sets.
>> Put some macaroni and cheese on there, too.
>> Hey, could you guys look into Adesius?
that uh PewDiePie thing he made. I know he's probably a little more jank compared to something like Hermes, but you know, it's still worth looking into since the popular YouTube is building it and deploying it. It might be something worth looking into. It could be fun.
>> Yeah. Uh we've been talking about building an aentic test harness bench.
Something like that. That could be a good one. Thank you, Odyssey. Nice to hear from you. Otterrains here with 5,000 sats.
>> Tough little ship. Little >> I did it. I took the plunge and I signed up for an intro computer programming course. Oh. Until now, I've just been a dabbler in writing scripts. The course is about Python. Nice. It's asynchronous online at my university.
>> Cool. It's well done. It's a good blend of bite-sized chunks of written and video content with homework and quizzes.
How do you guys uh have you guys taken lots of computer programming courses?
What worked especially well for you and what inspired you? Oh man. The only official computer course I've taken, I think, other than like training courses on like here's a crash course on PHP. I did one of those. I did a crash course on Python. Uh but the only actual official training I took was basic.
That's I'm an old man.
>> And yet I've never seen you write basic.
How is that?
>> Well, I wrote some cool basic programs.
Let me tell you what. [laughter] >> QBasic was my friend.
>> If it helps, Visual Basic was my first programming language. So, I hated visual. Okay. So, what about you?
[laughter] >> Uh, yeah. Well, before I ended up getting a physics degree, I was a CS major. So, I took a the first two or three years of most of those classes. I would say >> everything since then's been selftaught.
>> Yeah.
>> Wow.
>> Uh, I think No, I did um I think I did one uh MOO style uh functional programming course and a machine learning course, but >> Yeah, >> that was a long time ago now.
>> Yeah.
>> Um >> Yeah. Back when we called it machine learning. Yeah. Yeah, that's how long ago it [laughter] was. Yeah.
>> Um, what usually helps me is sort of trying to chase the thing that I'm most curious about. Like that doesn't work.
You know, some things you just have to grind through. That's always going to be the case. But, you know, finding things that motivate you, that you're curious, that make it >> feel more like play than like work.
>> And if you can try to keep do whatever you can, sometimes it's worth a little like you don't want to overfixate on your setup and not the actual work. But it is worth a little bit of investment just to try to keep the feedback loop as quick as possible because that helps keep it feeling like play.
>> Yeah, I agree. It's good. That's good advice.
>> I was doing exactly that last night where I will confess that I stayed up all night, never went to bed at all because I was so deep in the flow of like playing and chasing curiosities and like learning all sorts of stuff and breaking things.
>> You still showed up late for the stream.
Like how does that work? like you're up.
>> Well, like >> we all stayed up.
>> I was moving slower and I'm on a different time zone.
>> It wasn't even any squirrel based excuses.
>> I know. It usually got a squirrel based cats or anything.
>> We were all up late last night. I was up late last night. I replplumbed uh I have Blue Bubbles set up >> and I was repling it and then I was working on some of the EV road trip. I was up way too late, too.
>> It's We all know we have a show.
>> Yeah. Your version of late and my version of late.
>> Yeah, that's true. My version of late's like midnight and I'm like it's like 4:00 a.m.
>> Yeah. Yeah. Yeah. All right, I'll take this next one. Marvin42 came in with $42 in Fiat Funk coupons.
>> The answer to the ultimate question.
>> Here's a few bucks to help pay for refactoring Hypervibe. [laughter] Uh, I just found out that the Dendri Nyx idea a few days ago and have been trying to get Hermes to use it. I'm still struggling with the concept and it wants to build everything monolithic or make each setting a file. [laughter] Time for better skill shopping.
>> Oh, fun update. Thank you, Marvin. The dendritic Nick stuff is interesting. I think that might be something I have to poke at more.
>> Thank you. That's a great boost, too.
>> Anonymous boosts in with 25 Fiat fun coupons.
>> Oh my god, this drawer is filled with Froot Loops.
>> Longtime listener, first time booster.
>> Thank you from Austria.
>> Awesome boost.
>> Thank you for the first time boost.
Appreciate that.
>> Jem Le is here with Ten Fia.
>> [music] >> Thanks for all the hard work you guys put into this show week after week. Next time you're headed out east, uh, come have a good barbecue with us in case Kansas City.
>> We would love that.
>> Sounds ideal.
>> No complaints here.
>> Well, someone just needs to get a certain van working.
>> Solid idea.
>> I could do that. I just need a reason, you know, just barbecue.
>> Yeah. Yeah. Mhm.
>> I'm in.
>> I don't think I've ever really properly had it. I My biggest exposure has always been to Texas barbecue. I'm openminded.
We need to do a barbecue tour.
>> Mhm. Mhm. Mhm. High five connoisseurs here with $5 in fiat fun.
>> Everything's under control.
>> How are you all thinking about the secession planning now that there are agents? I imagine my family could get an agent to handle much of the home lab maintenance after I'm gone.
>> Fascinating.
>> Wow. Do you remember that episode we made and then there was like a couple projects trying to tackle this, but we came to the conclusion that it's just a very hard problem.
>> A succession agent.
>> This is cool. um start building. I say >> I mean there's something that >> just imagining a chat GBT thing behind a window they break in case.
>> Yeah. Break in case or like a or Telegram chat thread that's like set up but don't use unless [laughter] >> I mean it gives a whole new meaning to soul.md doesn't it?
>> Oh okay. We got to move on. It's getting dark. Uh Wes, you want to take our first member free booth?
>> Yes.
>> Uh I bear comes in with two member free booths.
>> Oh, by the way, had to reign up for Jupiter party. Not sure what email I used. Oh >> yeah, that's that's happens for sure.
>> Thanks for adding the web boost with USD. I'm not into Bitcoiner sets, but I love the shows.
>> Here's a little utility that I have been vibe coding for Linux based routers and firewalls.
>> I'm using it on my Vios boxes, but it works on any Linux instance using IP chains.
>> Cool. SW top.
>> This looks really good. It's got a great tuy.
>> Python MIT licensed. Oo, very colorful and pretty tuy. a real time router firewall traffic visualizer for the terminal inspired by BTOP. And you nailed it. It looks so good.
>> This is impressive.
>> Okay.
>> And a release 3 days ago. Very nice.
>> That's really sweet. That is so So what we need is uh you need to vibe up a flake.nix. [laughter] We'll give it a shot.
>> Thanks for the support.
>> Yeah, thank you very much and thanks for signing up again. Appreciate that.
Well, Mr. Pickles is here with [laughter] a free members boost.
You should provide API keys to members.
We have three 3090's and use about 2 billion tokens a month.
>> Oh. Oh, 3090s as in invidas.
>> Oh, yeah. Right. I don't even, you know, I just block those out of my mind because >> I also had a pick. Histler, a full text indexer for every website you visit.
[snorts] >> Great for finding thing you read about a month ago but can't recall. Holy smokes.
>> This is like indexes content so you can find information again from the web interface terminal or an AI assistant connected through MCP.
>> I guess it's by Adam Tobber uh as Moo who also made uh CRX maybe not the NG one but the one before that and a bunch of other smaller projects. Wow.
>> Yeah. No kidding. That's such a neat find. Thank you, Pickles. Appreciate that. I'll take the king. King Arthur is back with a member boost because I'm the wiz.
>> Uh, is not a king voted into parliament as one doesn't vote for one's king?
Rather, the lady of the lake bestows Excalibur to the one worthy to rule over Britain.
>> Of course, I do not require a sound clip or a member boost. Only specific boost for those who can bestow one, two, five, three sats to you in the future. Oh, what is that?
12.
>> That's got to be.
>> Does it require a map or just >> I think it's related to the other >> culture. The overall >> it must be. Yeah, we have some homework.
>> Mhm. Thank you, Kane. Appreciate you.
>> LP boos in. I'm just starting to dig in for Nyx OS, but I got confused from the start.
>> It's okay.
>> What's the difference between declarative Nixos versus the community one?
>> Which one is the best for a VPS versus desktop laptop? Oh, you must mean determinate nicks.
>> Ah, that's autocomplete maybe.
>> Uh, okay. So, determinate nyx is Yeah, it's like uh Yeah. Okay. That's like I guess you would call it a dro of nicks, a spin of nicks, a fork of nicks.
>> Yeah, somewhere in there.
>> Yeah. With with some tools and installers that are designed for, you know, enterprise use and whatnot in there. Um, perfectly fine. I think though probably for most people, unless they know they need to terminate Nick, they probably just want to use Nyx.
Yeah, it offers some extra stuff, but you you probably won't need or necessarily even care about a lot of that until you are a more experienced next user or you're trying to use it for some like production sort of purpose.
>> What about Wes? What about if you What about if you want to install it on an existing distribution?
>> Yes. So, for that, the determinst installer is great.
>> I think so, too.
>> Mhm. And then I also It also means like this decision doesn't matter for the most part as much as you think it is.
So, if you're going to install it there, just reach for whatever. As long as you get Nyx, it'll work. Uh, if you're going with Nyx OS, then you're probably just going to get like the standard Nixos installed media, and then you're just going to get the plain upstream Nyx OS anyway. Uh, and either way, they'll all build your next packages.
So, for that, you'll be just fine.
>> All right. The amount of overlap is overwhelming. Like, there's a very tiny slice of notable differences, but most of it as a beginner.
>> All right. Help me break this down.
Dance Tango writes, "Chris, you you considered a dedicated camera, a camera without AI that fixes the image. Is that really something for you?" Do you remember what they're speaking about?
>> Yeah. Um, I think we had a caller.
>> Oh, we were talking about dedicated pocket cameras.
>> Mhm. Or a dedicated like uh music player or something like that.
>> Mhm. Mhm.
>> So, the question is a camera without AI that fixes Oh, the camera without AI that fixes the image. Is that something for you? Well, I would think if you were going for like a pocket camera, you probably want you don't want something manipulating the image, right?
>> You do that after the fact if you want to do it, >> right? Yeah, that's what I would think.
>> How much of a purist are you?
>> That's that is a good point, though.
Like with the camera phones, you're not really getting a true picture. You're getting like an interpretation that's pretty close.
>> You always did. You always did.
>> That's true.
>> Well, uh Paul comes in with our last member boost. Half listened to the bit about analog cameras. 2 years ago, I've started taking pictures of concerts with a Nintendo 3DS.
>> Awesome.
>> It's far more engaging for me than taking them with my phone. If you're interested, you can see them on 3dstagevibes.atat.
>> Stop it.
>> Let me know if you want to know more.
>> That's That is definitely a vibe.
>> So, you were only half listening and yet you sent us some cool photos.
>> That's pretty great.
>> I'm looking at them right now.
>> 3DS stage vibes.
>> They're surprisingly good.
>> Huh.
>> Yeah, for a 3DS, right? It came out better than it is. Yeah.
>> There's a philosophy in the photography world that says uh the best camera you have is the one you have with you. So if you've got a 3DS, you start shooting with >> And why don't you have a 3DS?
>> I've actually got a dead one. I need to replace the battery of I really would like to get it working again. Uh thank you everybody who boosted the show. Uh we had 12 of you stream sats as you listened. You sat streamers. You collectively stacked 25,726 Satoshi's.
>> Not too shabby. Not too shabby at all.
When you bring it together with uh the Fiat boost and all the traditional boosts uh and convert it all to stats, our grand total for this week's episode is 265,638 Satoshi's. We tell you the numbers so that way we're transparent. You know what the show makes. You can decide what you think the value of the show is. We put it out there and we ask if you get value from it, want to keep it going, prevent the fate like the Linux magazines have seen. You can do it at boost.jupitarbroadcasting.com or with a podcasting 2.0 app.
If your boost is above 2,000 sats, we'll read it on the air. You can also just stream sats. You can boost below 2,000.
We still put in our show notes and we read it ourselves and have a nice little chuckle usually cuz they're usually quite nice. But we really appreciate, of course, you can get a membership at linxunplug.com/membership or support the whole network at jupiter.party.
keep the thing going and get the bootleg or the ad free version of the show and have that support on autopilot and then you get [music] access to the free member boost at boost.juptorbroadcasting.com.
Thank you everybody.
[music] All right, so we're doing something a little different for the pick segment and I'm going to make it quick because I'll tell you right up front, Moonlight and Sunshine are the way to go. Sunshine runs on the remote computer, captures the desktop, it encodes using your GPU or CPU if you don't have a GPU, and it's more like a video stream. It's really designed for remote gaming, which means a modern composited desktop actually performs almost natively. You can even do things like use OBS and if you had to, you could even edit video even over a relatively weak connection.
>> And that's kind of where you were playing with some of this, right? The idea of like, well, we've evaluated a ton of different ways to handle remote studio production over the years. Rustes has been a real solid go-to. Um, but on Wayland, it requires someone at the console to authorize the connection, >> which is kind of a no-go.
>> Cannot do that. Um, Meshtastic is a community favorite, but I don't need fleet inventory, file transfer, process management, a dashboard. I don't need any of that. Um, and I also don't need it to solve the proxying, the remote, because I'm going to use mesh networking. So, I've solved the network aspect already.
>> And there's a lot of tools that bundle a bunch of that stuff together. Yes. Yes.
Yes. So, what I really like about Sunshine is that the hardware video encoding, if you have it, even with just a basic Intel laptop GPU is really impressive. It's fine on this on the CPU, too. I have one system here. We have to do on the CPU, but everything works. HDR sound all comes through. So, you can hear the desktop audio as you're in the remote session. Uh, it works with plasma and gnomes animations, hyperlands animations. You can watch videos. It has uh multimonitor support. So all of it's really really nice. Um and of course it has Wayland support as well as X11 support which matters a lot.
>> What's the what's the downside?
>> Well the downside is is that you have to set up Sunshine on one end. You have to set up Moonlight which is just a desktop application and you pair them once and once they're paired once you can connect but you have to the downside you do have to do that onetime pairing.
>> So there's a little bit of setup.
>> Mhm.
>> Especially before maybe you are in a remote situation.
>> But the handy thing is is the pairing UI is a web interface. So you can probably if you're on the mesh network, just bring up the remote web UI, put in your PIN and pair that from the remote system.
>> Okay, >> so there's that's not so bad. So really the downsides are pretty minimal unless you need the other things that a meshtastic team viewer type things offer with file transfer and process management and maybe even user management and inventory. I don't need any of that. Um, so I was really impressed because I wanted to revisit and that I'll actually tell you the strongest contender doesn't come up very often. The strongest contenda is not meshtastic, is not rustle. Those are great.
>> Meshtastic.
>> Yeah, that's what it's called, right?
Oh, I'm sorry. Mesh Central. Thank you.
>> Thank you.
>> I knew there was something. [laughter] >> I mean, video over meshtastic is impressive.
>> That honestly, if you could do remote desktop, even process management over meshtastic, I would be thrilled. No, I meant mesh central. I just got the meshes meshed up. Uh, but I actually think the one that we really should give more credit to and the one that people should consider besides the Sunshine and Moonlight combo, the folks over at No Machine are still cooking.
>> They've been going for so long.
>> They've been going, >> which is why I think folks forget it.
>> Yes, they've been going since the very early days and they really have pioneered hardware H.264 encoded remote desktop streams and it still kicks butt.
I think no machine is better if you need a separate remote session. So if you want to be able to log in and have a desktop session running and have a whole other user session that's in a remote session, >> No Machine's great at that. Sunshine, you'd have to set that up completely.
Sunshine is really for you're taking control of the console.
>> Mhm. [clears throat] >> No machine lets you do both.
>> And it supports USB and device redirection, so you could have printers and certain mouses and keyboards. So it does a lot more nice stuff. Um, but if you just need something quick, you've already got the networking taken care of, like a mesh network or something like that, you don't need a central proxy or inventory or fleet. I really think for the 10 minutes it takes to get Moonlight and Sunshine going again. It's like >> and these are pretty easy. Like I they're now it's been it's been out for a while, right? So they're like pretty available and repos are >> Sunshine is is actively still being developed. Uh, Moonlight hasn't had an update for a while, but it's a pretty stable. I mean, it's just an H.264 streaming client that supports input, so it's not very complicated. There are a few forks out there, but Moonlight, it's packaged in Flatpack. It's packaged in just about every DR now because it has been pretty stable for a while. Same with Sunshine. But Sunshine is still moving. They're still progressing with it. Still getting better and better. And it was really nice. Been a couple years to check in. Um, I did all the remote testing. We'll see though. We're going to go to Canada and next Sunday's episode will be a remote episode. And if we're live, that meant that uh sunshine and moonlight worked. And if we're not live, >> keep on refreshing that RSSP. [laughter] >> Yeah, exactly. Yeah, exactly. Uh it should be good, though. We're We got some good projects planned when we get up to Brent's house. That's going to make I think for a great episode.
>> I got a whole list.
>> Yeah, I know you do. We're looking forward to it. So, of course, we also want to know your feedback on Lionus' position in the colonel now. How you feel about that? And particularly, I'd like to reach out to those of you who are against AI use with free software and what you'd like to see happen if a project receives a legitimate concerning bug that's generated or you know the the patch the fix is generated by an LLM.
What would you like to see that project do? Couple of range of questions there and also your thoughts and feedback on log Obsidian and free software notes in general. I think all of those would really fill out the feedback and and the boosts and we'd really appreciate it.
So, why not join us? See if we make it live. See if it actually happens.
>> See you next week. Same bat time, same bat station.
>> Make it a Tuesday on a Sunday. Join us at 10:00 a.m. Pacific, 1:00 p.m.
Eastern. Of course, we've got chapters.
>> Oh, yeah. And transcripts >> as are all available.
>> Magic feed. Magic uh text files just waiting for you to download them.
>> I don't know if we'll have video next week, but we usually have an MP4 hidden in that feed as well, if you can believe it. If you could even believe it.
>> If you find yourself a fancy client, they can [music] make it even easier.
>> Yeah, they'll just show it to you. Yeah.
How about that? Podcasting 2.0. If you want more show, remember that Lup Lug gets together in the mumble room before we get going. They're cooking all along.
Members, you get the full bootleg version with twice the content clocking at almost 2 hours right now. And then links to everything we talked about today. How to contact us, Mumble, our Matrix chat, all of it at Linux unplugged.com. And this was Linux unplugged 676. [music] Thank you so much for joining us on this week's episode of your unplugged program. We'll see you right back here next Tuesday, which of course is next Sunday.
[music] [music] [music] Heat. Heat. N.
[music] [music] [music]
Related Videos
12:08
Artificial Intelligence
Expanding Stikbot thumbnails
leopoldshorts
2K views•2023-09-24
32:31
Artificial Intelligence
Digital Discrimination: Cognitive Bias in Machine Learning
redmonktechevents2974
4K views•2019-12-18
1:22:05
Artificial Intelligence
Evolutionary Approach to Clustering by Ujjwal Maulik
ICTStalks
279 views•2019-06-26
55:30
Artificial Intelligence
Rose Yu "Learning from Large-Scale Spatiotemporal Data"
networkscienceinstitute
2K views•2019-03-04
28:24
Artificial Intelligence
Stanford Seminar - Generalization through Task Representations with Foundation Models
stanfordonline
4K views•2025-07-14
8:29
Artificial Intelligence
Satellite-Based Wheat Yield Forecasting using GEE & Transformer Neural Network
gisrsinstitute
634 views•2025-06-15
32:06
Artificial Intelligence
Paradigm Shifts in Data Processing for the Generative AI Era: Robert Nishihara of Anyscale & Ray.io
GradientFlow
2K views•2025-01-02
39:48
Artificial Intelligence
How to Build Your Own GenAI-Based Knowledge Management System
2150GmbH
360 views•2025-06-03
Trending
13:38
Law
WOW! Judge TURNS THE TABLES on Trump in His OWN $10B LAWSUIT!!!
MeidasTouch
197K views•2026-07-23
13:25
Business
Playstation NO DISC/NO BUY Fight Is Over...
DavidJaffeGames
4K views•2026-07-23
22:40
Business
Steam and Xbox Just Dropped The Hammer On PlayStation
OhNoItsAlexx
9K views•2026-07-23
16:52
Anthropology
Americans Confused in Australia for 17 Minutes Straight
IWrocker
17K views•2026-07-23