Torvalds strikes a necessary balance between AI utility and human accountability, ensuring the kernel doesn't succumb to hype-driven mediocrity. Wayland’s superior gaming benchmarks finally silence the skeptics, marking the definitive end of the X11 era.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
Torvalds talks A.I. and Linux, Wayland beats X11, Age verification getting worse - Linux Weekly News
Added:Yes, this video is going up on a Friday, not a Saturday, because I've got to go play with my little plastic miniatures at a Warhammer tournament. And yes, I do look like I'm going golfing because of laundry day. I do need to wash up some clothes, so I don't look like that. I assure you, I do not golf. Anyway, in this week's episode of the Linux and open source news show, we've got Torvalds speaking up on AI. First to say that maybe it's not that ready, and second to say that it is absolutely here to stay in the Linux kernel. So, we'll talk about both statements, the second one being a little bit more controversial, I think. We also have Wayland being better than X11 at gaming, proven by benchmarks at Phoronix. We also have the age verification situation getting worse and worse with a pretty chilling analysis of which countries are trying to do what, and that's bad.
Anyway, we also have this message from our sponsor, which is good. This video is sponsored by Squarespace. You have heard about them by now. They are your all-in-one solution to create, host, manage, build, [music] improve, deploy your website. What separates them from the pack is their very easy tools to build your website. You have pre-made templates for specific pages that you need, pre-made themes, and still all the customization options you might want to [music] arrange the layout to have things exactly how you like. You don't need to know how to code to handle all of this. It's just all drag and drop.
You select a template, you slide your images in, you rearrange the elements on the page, and it just works. And once your website ultimately grows, you can also handle multiple contributors with full access to permissions for each user and each feature. And your login also lets you manage multiple websites if you extend your presence online even more.
You also have a very capable mobile application for Android and iOS to receive notifications for your sales, to manage payments and inventory, to view all the metrics you need. Squarespace is just a fantastic solution to get started with your website and to make it grow as time goes on. So, the link, as usual, is displayed here, squarespace.com/thelinuxexperiment.
You can click it in the description.
You'll get a 14-days free trial, and once you're ready to start, you also get a 10% discount on your first website or domain purchase.
So, let's begin with Torvalds and AI. He talked a little bit about Linux, AI, and more at the Open Source Summit in Mumbai. He said that AI is pressuring the usual Linux kernel workflow because, of course, it is finding more and more what he calls interesting bugs, and this, apparently, is stressing some people out in the community. Torvalds also said that he likes the current release schedule with a new version every 9 to 10 weeks without anything flashy as a headline, just steady improvements. And even though AI looks to be pressuring things a little bit, he's not looking at changing that release schedule. He also said that the issues that stress him out are personality issues, not code issues. He says that code is relatively easy to fix, but personality isn't. And he admits he's caused some of these problems himself in the past. Torvalds went on to say that he is no longer a programmer these days, that he rarely submits his own code. He's not sentimental, either, when it comes to hardware, saying that, basically, if no one else uses something in the kernel, he will be ruthless with removing that because there's no reason to have that maintenance burden. On AI, specifically, he said that recent comments he made that stated AI would give a 10 times productivity boost were not scientific, that this was just a number, as he said, pulled out of his ass, and he said that he hopes AI creates more productivity than it takes away, but that currently they're seeing more junk produced by AI than actual useful code. For him, an AI-generated bug report simply is not enough. A human needs to add a proposed patch. The AI-generated patches are seen as not good enough, being band-aids more than real fixes. Torvalds said that he does use AI for prototyping his own toy project, but he says that these tools are not up to Linux kernel level patches right now. This might give you the impression that Torvalds isn't completely sold as AI on the tool, but no, he is completely in cuz he made another statement. This came out after some discussion on using Sashiko AI to do some code review on the kernel. A developer pointed out that maybe the kernel should follow the first two guidelines that the Software Freedom Conservancy issued on AI. Those guidelines being that the free and open source community should support, not just tolerate, people who reject LLMs.
Basically saying that people who want to avoid AI should be supported and assisted, I would assume in the form of let's not force anyone to use AI or to use AI-generated products. The second point is that every contributor should be allowed to decide if they want to use AI or not, and that no one should be forced to interact with it.
Someone said that this first statement, let's not force anyone to use AI, is very anti-AI, which I don't think it is, but Torvalds said that yes, it is anti-AI and it is not the position of the Linux kernel. He said Linux is not anti-AI, AI is just a tool and a useful one, and that anyone who doubts that hasn't used it. He also said that it's a painful tool and that the job is now to make sure that these tools are less and less painful and more and more useful.
Torvalds also said that the social angle of open source is important, but is a side benefit, not the entire point of the project. And he concludes by saying decisions will be made on technical merit, not fear of new tools. There will not be any religious reasons invoked here and that Linux will not be a I quote social warrior project. So, the tune seems to change with every statement cuz at first apparently AI is not up to snuff to write bug fixes, but apparently it's also extremely useful to the point that you can't even discuss the fact that maybe some people don't want to use AI generated code.
I don't know. It feels kind of weird.
The social aspect that he talked about, I think that comment is pretty insane.
Of course, when with open source for your project, the main focus is that social aspect of sharing code, sharing the repo, letting people contribute, making that a public property, a social benefit to everyone of having access to that product. This is why you went with open source. If not, then you could have used any proprietary license you want.
So, saying it's a side benefit, no, it's not. It's the very point of why you went with that license. I I don't really understand that. On the AI part, I do understand the statement though. AI is garbage for 99% of use cases, but for development, it is undeniably useful.
You can look at all the projects that have been vibe coded and have accomplished things that humans have failed at for years. Rewriting X11 in the course of a few months, that has been done with AI twice now. Fixing some software that would never run with wine and making them run in a few weeks worth of time using an LLM, that has been done now. You can't really dispute the fact that AI for development is very useful.
I hate it. I hate how it's built, but it is an undeniable useful and I don't see a project like the Linux kernel depriving themselves of of that. But, the comments on the social aspect of open source, come on, man, that is ridiculous.
Now, we have more proof that Wayland is indeed ready even for gaming because Phoronix did their usual benchmarks comparing gaming performance on KD Wayland and on KD X11. This was done on Plasma 6.7 on CashyOS, and the test was uh probably along the lines of, "Hey, since X11 will be removed in the next version of Plasma, is gaming on Wayland better or actually worse?" So, will that forced transition on KDE uh harm people who want to game or not? The end result is that Wayland either beats X11 for gaming or that they're so close to X11 that there is virtually no difference.
Certain games gave much better performance on Wayland, like Zonotic, Dirt Rally 2, or the Gravity Mark benchmarks. The only place where things were closer was in Cyberpunk, where both sessions tend to perform the same, but the frame rates measured there were pretty low, so I would assume uh Wayland gives you a little bit more headroom, but when you're completely uh strapped for resources, there is virtually no difference. As Phoronix concludes, the Wayland section is ready to take over X11 for gaming, no problem, and is even generally better at gaming, which is not surprising because this session is the one receiving the most work, optimizations, and improvements these days, where X11 sessions are just basically abandoned by every desktop.
Still, that's good to know. It wasn't extremely exhaustive testing in terms of games. Uh I would have like to see uh more games tested, not just a benchmark, like more big-name games running with Proton, but it still seems to show that, yes, we can safely ditch X11 and use uh Wayland and maybe XWayland while Proton gets updated to be Wayland native, which will solve some issues, like for example, uh latency, uh which is a little bit higher with XWayland, but should be actually lower on Wayland than on X11 once XWayland gets the boot. Now, let's talk about age verification again uh because Proton, the the company, not the gaming tool this time, is raising the alarm on that. They now have an interactive map that you can visit showing which countries are introducing age verification laws and how they are implemented. And this is much more alarming than I thought. It's not just a few states in the US and a few countries. It's 49 US states that are talking about age verification, 18 EU countries, so more than half of the EU, China, Australia, Brazil, and Turkey, which is already a massive portion of the world's population. They are in various states of implementation and decision, of course, and you can check for each state or each country what they're trying to achieve here, but it is bad. Proton is calling this the death of privacy online as these measures obviously encourage mass surveillance and create more failure points where this data could be accessed by malicious actors through the age verification services that are being improperly secured as we've seen in the past or even government-based solutions that have been known to be hacked as well in many, many countries. Obviously, terrible idea to implement any form of age verification, especially if you have a centralized database of everyone who logged into anything cuz you're just pointing every hacker to the biggest, most juiciest morsel of digital data ever created. Insane idea to try and build that. We all know it's for potential mass surveillance. You're just opening the door up for more totalitarian governments to abuse that.
It is insane that this is even being discussed and that it's not being fought harder. Anyway, let's go to some much more fun stuff. First, the Cosmic Desktop got its nice frosted glass interface this week. This is an optional thing that you can turn on to give you the usual translucency plus blur effect in Cosmic's own applications.
Third-party apps will need to update to the latest version of lib cosmic to be able to support it, which I would assume, since there are not many of them, uh this won't take too long. This lets you control how how blur you want, how much translucency you want, and also onto which parts of the interface you want to apply this effect. It can be window backgrounds, it can be panels, it can be applets, it can be all of those, you decide. So, it is pretty flexible, and at least in screenshots, it looks pretty damn good. It is not enabled by default for now. From what I've seen, it also doesn't seem to apply to GNOME applications either, even if you enable the theme GTK apps toggle. This is all handled through libcosmic specifically.
And of course, since KDE apps aren't yet themed in Cosmic, you're probably going to end up with something quite incoherent in terms of look, but as the selection of Cosmic app grows, this will be less and less of an issue, I guess.
And alongside its frosted glass interface, Cosmic also got some cool updates in version 1.3. This includes power consumption metrics in the system monitor for AMD and Intel GPUs. You will also see your VRAM usage for these GPUs.
Nvidia GPUs are also a bit more reliable with proper suspend support. The Cosmic panel can be always hidden if you never use any of what it has to offer. You can also decide to keep the current style of a panel when a window is maximized, so it doesn't have to be opaque, for example. Animations received better frame pacing, too, so they look smoother. You can use AVIF images as wallpapers. There are fixes for multi-GPU systems, for full-screen apps, for multi-touch gestures on trackpads, for external displays. Mouse pointer themes should also work better now. The file manager should detect compatible file types when it's opening a file picker. The Bluetooth applet now shows previously paired devices, and Cosmic moved to a Rust-based network manager client. The application launcher now lets you run applications using the discrete GPU if you have one. There are more fixes for most of the desktop's components. It is a pretty big update.
Now, they even published a long list of everything they added and or fixed in the past 7 months, and it does show that this is a desktop that is moving really really quickly. They haven't changed many important giant features of the desktop. This will very likely be for version two, maybe in Pop!_OS 26.04, but in terms of adding features, applications, and fixing bugs, they've done a stellar job. So, props to them.
Cosmic is really nice. I think it was last week that we talked about a vibe-coded Y11 server, which was a replacement made in Rust, uh vibe-coded through an AI uh for X11. But, we now have another replacement for X11. This time, vibe-coded in assembly. This is apparently a single file without any external dependency on Mesa or anything else, and the project is called Frame.
The developer says it's stable enough for them to daily drive it and to only occasionally yell at it. They say it's about 20,000 lines of code compared to the 4 million lines or thereabouts for X11. And the developer also said that this was all due to his buddy, Claude, which never gets tired, isn't opinionated, and also taught him a great deal about the entire Linux stack. Frame is also the entire graphics engine. So, apparently, you don't even depend on Mesa drivers or anything. It talks directly to the Linux direct rendering manager, which I would assume is good enough for basic drawing pixels on screen, but probably not good enough to do more graphics-intensive stuff.
There's apparently a lot of protocols to implement still around that, but the project looks pretty fun, at least. Now, of course, this isn't meant to be run by everyone or packaged by distros, but it is open source. So, you could theoretically check it out provided you can read assembly, which I think will limit the pool of potential contributors to that project, obviously. Now, I think it does open an interesting line of thinking around AI, which is, at some point, will these tools allow anyone, whether you know how to code at all, read code, or or not, will this allow those people to just create anything out of the blue? Like, "Oh, you know what?
Uh every app I find that does this specific task, they suck. I don't like how they do it. I want mine." And then you just ask an AI and with three or four questions, you get something that works exactly how you need it. It might not be pretty code, it might not be shareable, it might not be secure, but you don't care. You're running it on your own computer, potentially in a sandbox or on your phone, and it's just for your use case. If that happens to exist at some point and you don't even need to understand a single line of code to use it, then I think AI will have a very interesting place in opening up computing away from big tech companies, and I think that's pretty neat. That doesn't solve the ethical problems with AI or the ecological problems with AI either, uh but it would be interesting to follow at least. Now, GNOME OS, the official GNOME distro that is still for now more meant for development than for day-to-day use by users, they're testing a new tool to test experimental applications and system changes. GNOME OS, if you don't know, is an immutable distro, so this makes uh testing things out and changing system components maybe a bit more annoying, but they do want at some point to become a distro that users could use on their regular PCs for day-to-day work. So, the tool that they're introducing is called the GNOME OS developer tool suite. And it is work that is funded by the German Prototype Fund. The goal is to have a relatively simple way to create system extensions for GNOME OS. GNOME OS is image-based and these system extensions can just layer applications, subsystems, newer drivers, additional components, whatever you want on top of the very stable image. So, if something goes wrong, you just disable the extension and you boot your system as is. You haven't changed a single thing on the stable system that you run. There is a proposal to create test center, which would be the hub to manage all of these extensions instead of using flatpaks, which Of course, flatpaks work well to test a graphical app. You can have an experimental flatpak, and it's not going to change your system. But, the problem is when you want to test a command line utility or a system feature, this doesn't really work. Test center would then let you just click a link from the developer. It would add that experimental app to your list of system extensions. It would install it. And then, when you don't want it, you can just uninstall it with a little trash can icon. This would be the same workflow for system extensions, for applications, command line utilities, whatever. It's basically what Apple already offers with TestFlight on iOS, for example. Developers can also mark these experimental builds as experimental. They can set an expiration date, so no one keeps using stuff that is completely out of date. Now, currently, it is just a proposal. The team is looking for feedback on it, but I think it is a fantastic idea. Not just for testing stuff, but also to fix the main problem of immutable distros, which is if you try doing something that hasn't been specifically thought of by the distros maker, then you can't do it.
You have to jump through so many hoops that it virtually becomes insane, and you just would be better off with a basic package-based distribution instead of an immutable one. This would solve those sort of issues in a nice, graphical, easy-to-use way, especially if you could have a centralized repo of system extensions. That would be pretty neat. Something that amazed me this week in the Linux runs on everything sort of trail. A madman managed to make Linux run on a Sega 32X. Now, if you're too young to know about this thing, the 32X was an add-on, a sort of mushroom that you slotted into the cartridge port of a Sega Genesis or a Mega Drive, as we knew it anywhere in the world that wasn't the US. The Mega Drive was the competitor to the Super NES or Super Nintendo at the time. It gave a pretty good showing in the US and Europe, and it was a fantastic console that I still own today. But, the 32X was a miserable failure designed as a stopgap to introduce 3D before the release of the Sega Saturn, of the PlayStation, and the original Xbox. It sold very poorly. It was a disaster, and it helped sink Sega in terms of producing consoles. Anyway, Linux runs on that little failure now, and it's even able to make use of its two CPUs. Those are mighty Hitachi chips running at 23 MHz with a whopping 256 KB of extra RAM added to the 64 KB that the Mega Drive already had. And also, 256 K of video RAM. Linux can run on this Frankenstein device with parallel processing using the various cores. They couldn't really run it [clears throat] on the measly RAM that the device has, but they used a flash cartridge onto the cartridge port to add 4 MB of RAM, which lets the Linux kernel be stored in memory and boot. Note that this is a CPU configuration that has no CPU cache, no concept of synchronization between CPUs.
The CPUs don't even know each other exist. It was meant to be programmed by hand, where tasks would be manually placed on one CPU or on the other. But, the madman who developed this managed to use the original Genesis CPU as the load balancer that makes the two CPUs in the 32X talk to each other. It is insane stuff to me, and it's really funny cuz it's a device I learned about and that I actually wanted to buy, but my mom said no cuz it was way too expensive, and there were virtually no games on it. And now you can run Linux on it if you want.
It's not going to be usable. You're never going to run a graphical desktop.
It's going to be terminal only, but it boots, and that's actually insane for that sort of stuff. It made me laugh. I wanted to share.
Now, let's talk a little bit about Firefox because they're going to change to a faster release schedule. A new version will be published every 2 weeks now instead of every month. This is something they're experimenting with.
This might not stick in the long run.
And they also say that this will not force developers to work double time.
Things will still only ship when they're ready. This doesn't change the fact that Firefox already has and will still have hot fixes and minor versions in between those major versions every 2 weeks.
Nothing is required from users, either.
You will just get updates twice as often, but not necessarily with more features than before. This affects the desktop and Android versions and will start in September. Shouldn't change much for users. It just means you'll get your feature maybe a little bit earlier than if you had to wait a full month when it's ready. But that's about it. For distro maintainers and packagers, though, this potentially means double the work. Uh if your packaging workflow is not fully automated, then you're going to have to do twice as many packages as you did before, which obviously is going to add some extra workload for a bunch of distros, notably those that don't, for example, use the Firefox snap if they're Ubuntu-based but repackage it for Firefox forks as well.
That's a lot of extra work. So, we'll see how they manage that. And still on Mozilla, they released a new study to look at how Windows pushes Edge so aggressively to their users. They listed all the usual weird tactics and dark patterns that they could notice in the EU or the larger European Economic Area, in the UK, in the USA, and in India. And it does look like the EU's consumer protections are helping a little bit here in terms of the horrendous stuff that we don't have in the EU, but other places in the world do have. Uh there's a banner pushing you to use Edge.
There's an ad banner injected onto Chrome's download page. So, Google's web page to push users to download Edge.
There are obstructions to making changing the default browser difficult because you have to pick every file type that you want to use with that browser individually. There are dark patterns that try to restore Edge after you move to something else under the cover of fixing settings or syncing stuff. It's terrible. Now, the EU still has a bunch of these horrible tactics like, for example, when you upgrade from Windows 10 to 11, you're going to get Edge by default instead of your own browser.
Copilot opens links in Edge instead of the default browser. User data sharing in Edge is opt out, not opt in. All of that sort of stuff. In the US, all of those tricks are present just like in India, and that is a lot of stuff just to try and avoid people replacing Edge.
And honestly, some of those things feel like they shouldn't even be legal, like actually modifying the DOM, the web page of a competing browser. That is insane.
Imagine if they did the same but just hid the download button. That's the exact same technical aspect. Instead of injecting a banner, you're removing a button. It's the same technical manipulation. This would be illegal. So, I don't see how injecting a banner on Chrome's download page is legal. That that seems insane. Opt out data sharing doesn't really feel right. Reverting to Edge when you click the button saying, "Fix my settings." That also doesn't feel right at all. This should absolutely be illegal, not just in the EU but in the US, too. Now, the problem is, as usual, Mozilla is fighting against the wrong opponent. Edge has like 10% market share worldwide in terms of desktop web browsers. Chrome has 75.
Target Chrome. Fight against Chrome.
Don't fight against Edge. If you take Chrome down, you also take Edge down cuz Edge is blue Chrome. It's literally Chrome with another logo. Now, if, unlike me, you enjoy AppImages, you'll be happy to know that there's an interesting app that makes installing AppImages about as easy as installing applications on macOS. It's called App Manager, and it is a GTK4 and libadwaita app. It will open when you double-click any AppImage file. You will get the application icon, your applications folder, and an arrow pointing towards it. You just drag the applications icon into the apps folder, and it installs the app image in the right folder with shortcuts in your menu, with file associations, all of that sort of stuff.
It can also be opened without an app image. In this case, it acts as a hub or a manager to give you all the options to change launch options on your app images, and the various path that they use, icons. You can perform updates if you input it a a GitHub or GitLab link in there. Basically, you manage your app images in a single place. The main benefit here is that AppImage does not depend on FUSE to mount and run your AppImages and install them. It uses instead U runtime, meaning you will not need to install libfuse on distributions that do not have it, like, for example, Ubuntu by default. And also, generally, FUSE2 is unmaintained, used by a lot of AppImages, which means you're just opening up a security problem if you install that on your system. Now, of course, there are plenty of other tools to do exactly that, but this one seems really well-designed and really handy.
And also, it fixes the most glaring issues with AppImages, which is their reliance on a FUSE, or at least older versions of FUSE. And now you have something that is truly portable because you can just pick up your entire applications folder with all your AppImages, and you can open AppImage Manager on a new PC, import all of those, install them all at once, and you're done, which is, I think, a pretty good workflow. Now, all AppImage lack is a central repository that is not AppImageHub cuz this is full of old and weirdo stuff. You need a little bit more control, vetted stuff with minimum requirements to have good quality app images. That would be pretty neat. Maybe it exists. If it does, let me know in the comments. And this will conclude this episode, but we can't finish without a mention to our sponsor, TUXEDO Computers. TUXEDO makes laptops and desktops that ship with Linux preinstalled. They're based in Germany, but they ship to most countries in the world, and they have a big big range of devices. You'll find some more affordable stuff with small form factor laptops all the way up to giant gaming devices, whether they're laptops, desktops, towers, workstations. You have a lot of control on the hardware you want, on the specs you want to configure. And also, of course, you can change the keyboard layout, you can change the logo engraved on your device, a bunch of stuff like that. And you know that Linux will run absolutely fantastically on those devices cuz they're made for that. I've used TUXEDO Computers exclusively for the past few years. They've never let me down.
They're fantastic devices. I can only recommend them. And as usual, the link is down in the description. Anyway, this will conclude this video. You know where all the usual buttons, links, and comment sections are and why you should absolutely interact with at least one of those items so the channel can keep growing a little bit. And of course, you have all the links to support the show down in the description. Thank you all for listening and all watching, and I guess you'll hear me in the next one.
Next. Bye.
>> [music] [music] >> Mhm.
Related Videos

TOP 15 Data compression Interview Questions and Answers 2019 Part-2 | Data compression | Wisdom jobs
wisdomjobs
281 views•2019-06-28

CTS 158: 802.11w Management Frame Protection
ClearToSend
4K views•2019-02-04

NDSS 2019 Send Hardest Problems My Way: Probabilistic Path Prioritization for Hybrid Fuzzing
NDSSSymposium
496 views•2019-04-02

How realistic is Cities: Skylines?
CityBeautiful
159K views•2019-02-14

GUIs & TUIs: Choosing a User Interface for Your Python Project | Real Python Podcast
realpython
2K views•2025-04-04

The OSI Model - Explained by Example
hnasr
225K views•2019-05-12

Cloud Computing - Introduction
elithecomputerguy
98K views•2019-10-07

From Traveler's Dilemma to Dynamic Routing | Demystifying Networking
IITBombayJuly
5K views•2019-08-04
Trending

WOW! Judge TURNS THE TABLES on Trump in His OWN $10B LAWSUIT!!!
MeidasTouch
197K views•2026-07-23

Playstation NO DISC/NO BUY Fight Is Over...
DavidJaffeGames
4K views•2026-07-23

Steam and Xbox Just Dropped The Hammer On PlayStation
OhNoItsAlexx
9K views•2026-07-23

Americans Confused in Australia for 17 Minutes Straight
IWrocker
17K views•2026-07-23