Firefox Makes All the Mistakes, More Wayland Work  ## Document Notes ## Highlights - **Firefox Turns Toward AI As Core Focus** - Mozilla's new CEO plans to make Firefox a "modern AI browser," shifting core focus to AI integrations. - Nick argues this duplicates existing offerings and wastes Mozilla's differentiation potential. Transcript: Nick So Mozilla has a new CEO, Anthony Enzo DeMeo. He replaces Laura Chambers at the head of Mozilla, and he published a little blog post that set the tone. He says that trust is the one issue in tech and browsers that they want to fight for and that he wants Mozilla to become the trusted software company, basically making Firefox a platform That lets you connect to a bunch of trusted solutions. He says that Firefox and Mozilla can be what people decide to choose, decide to use, and to get to that result, he has three main ideas. The first one is that people must have control over how things work, whether it's about privacy, data use, AI. He wants everything to be clear and understandable, and he wants people to always be able to turn off all the AI stuff. Second, he wants monetization of Mozilla to be fully transparent so people know who's funding what, and that comes back to the trust issue he mentioned. So for now, this is all well and good. Those are good goals that Mozilla currently already achieves. But third, he wants Firefox to become a modern AI browser, a sentence that immediately kills any other effort that they want to make on other areas, at least for me. He says Firefox needs to support new and trusted software additions, thinking about AI stuff, and he says that this will remain the anchor of Mozilla, as in Firefox will remain their Core product. Mozilla will thus invest in AI over the next three years, and they'll try to find other revenue sources than search, as in the Google search deal that they have. After the very much expected and deserved outcry from the community, Mozilla did confirm that they will be implementing a kill switch for all of these AI features to ensure users can Disable everything. And they said that the AI stuff would be opt-in, but the developer saying that also said that just clicking a toolbar button could be considered an opt-in. So that doesn't really mean much, which is why it's good to have that AI kill switch, so you can be sure that none of those features appear anywhere for you. ([Time 0:01:40](https://share.snipd.com/snip/aae60bdf-b1f4-4b9b-97ec-4e5bf381d5eb)) - **Use Forks Or Extensions Instead** - If you dislike Mozilla's direction, use forks or remove vanilla Firefox from your systems. - Rely on extension developers to build AI features instead of centralizing them in Firefox. Transcript: Nick This will be there in 2026, so I would assume by the time they start tacking on that AI stuff to Firefox, you'll already be able to disable it, but it still means Firefox will be laser focused On adding AI tools which will connect to known providers. So the usual suspects, OpenAI, Gemini, and anything else, that's what they are adding to Firefox, a way to connect to those which in my opinion is a huge huge mistake because that's already What everyone else is doing. And to add insult to injury, the new CEO also said to The Verge that Mozilla even considered banning ad blockers from Firefox as this could potentially bring in 150 million dollars of Revenue if they did. The only thing that apparently prevented them from doing that is that it felt off mission. Not that it feels wrong to take away agency or control, not that it feels absolutely wrong to prevent users from doing whatever they want in their browser and to stop them from using actual Useful tools. No, it just feels off mission. So literally the minute Mozilla's mission changes and becomes more focused on, I don't know, striking deals to display ads online, then I guess banning ad blockers becomes on-mission, And they're going to start doing it. This is the stupidest crap you could have ever said as the CEO of Mozilla, and he says it on his first day. Wonderful. And then, of course, we have the entire AI issue. This, for me, is enough, combined with all the previous blunders of Mozilla, which have been very well documented. This is enough for me to delete any copy of vanilla Firefox from any of my systems and to never recommend anyone use Firefox, pure Firefox, again. Forks will be free of all that garbage. Hopefully, they stay free of it. The problem isn't necessarily the AI integration. You can turn it off. The problem is Mozilla is focusing on doing the exact same thing everyone else already does months late in a worse way, not in a more ethical way, not in a more open source way. It is a useless focus if you're doing it this way. It's exactly like bringing Firefox VPN or Firefox Relay or whatever else. You're copying existing products and not doing them better. The only differentiating factor is that you're putting the Mozilla or Firefox logo on it and people don't care. This is not the focus. People want a Firefox that is faster, that is more compatible, that has more interesting features. If you want to add AI, develop extensions or let extension developers make it, and they will complete that integration for you. They'll make it faster and they'll make it better than you. You don't need to focus on this. Focus on differentiating yourself from all the existing AI browsers. You're just copying everything that everyone else does. Again, this hasn't led to any market share gains ever. It's still not going to work. You're still driving Mozilla down the drain. It's insane to watch and really disheartening. So hopefully Firefox Forks will stay free of all that garbage. ([Time 0:03:46](https://share.snipd.com/snip/fc6dccd8-2104-43ee-a573-dfadb291349b)) - **Servo Engine Shows Useful Progress** - Servo released progress but remains far from replacing major browsers as a complete UI is still needed. - Engine improvements (context menus, video controls, parallel CSS parsing) make future browsers plausible. Transcript: Nick Unfortunately, Servo is still making a bit of progress with the release of version 0.0 which of course is not at all ready to provide an alternative to Firefox or Chromium-based browsers. But the work is still interesting. The engine now supports context menus, so right-click menus for links, images, and various other web page contents, meaning you can start implementing some better functionality In a future browser using Servo. It also now supports the controls for HTML5 video, so you can play, pause, and all of that stuff. It supports compressing and decompressing streams as well. The Servo Embedding API that lets you add a Servo-powered web view is also now improved, letting each web view set its own context. So I guess one webview crashing doesn't kill all the other webviews in the system. And CSS is now parsed in parallel with scripts and the page layout for better performance. Is not really enough to make it usable as a daily driver. ([Time 0:06:50](https://share.snipd.com/snip/e01498f8-be76-477a-a16f-780381d386a6)) - **Validate Wayland Mirroring With KScreenDoctor** - Test screen mirroring and scaling under Plasma 6.6 before ditching X11 in 2027. - Use KScreenDoctor to add custom modes and detect available mirroring options. Transcript: Nick Now there's good news for Plasma users, as KD 6.6 will fix a few annoyances with Wayland, and they probably should because they're going Wayland only in about a year. So 6.6 will improve how screen mirroring works under Wayland, where previously it was sort of unreliable. It would sometimes work properly, sometimes not. Sometimes the image would be displayed on only one of the displays instead of being mirrored. Sometimes it wasn't to the right size or the right scaling. In short, it just was very hit or miss, and that is a super important big use case for anyone who uses Plasma to display any sort of slides in a meeting, for example, or in a classroom. I personally ran into that problem when I was teaching. For example, the laptop display would just be fully black instead of being active. Also, scaling factors will be better selected by default when connecting to something like a TV, likely to avoid having extremely small interface elements at 4k, a super small mouse Cursor that you simply cannot use at all. Again, very useful for displaying slides in conference rooms or classrooms. To help with all of this, the KScreenDoctor command line tools now lets users add custom screen modes and they will inform users that screen mirroring is available and possible. ([Time 0:08:07](https://share.snipd.com/snip/f67a8eab-b5a2-442c-b8fe-14f05f68e973)) - **Wayland Upgrades Target HDR And Color Workflows** - Wayland protocols 1.47 improve HDR and color management for professional workflows. - These changes enable better color-proofing and make Linux more viable for video/photo editing. Transcript: Nick And still on Wayland, the new version of the Wayland protocols was released this week, version 1.47 which again improves the color management protocol and HDR support. They have removed some restrictions on various formats to let certain movies play in HDR. The previous behavior was apparently too strict in detecting various formats, which resulted in certain movies not playing in HDR when they could have. They also added new properties that let compositor state that they don't want the colors to be adapted to the display specifically so things can be rendered exactly as they were intended, Something that apparently will be useful for color proofing applications, so maybe stuff like color calibration tools. Now this is stuff that won't be useful for a lot of people but this is the kind of work that makes Linux a more viable platform for professional applications around a color grading workflow, Around video editing, around photo editing. ([Time 0:09:44](https://share.snipd.com/snip/9f852858-81c2-4e8f-b5a2-dfd33aec02bc)) - **Don't Publish Unreviewed AI-Generated Extensions** - Avoid submitting GNOME extensions with large AI-generated code blocks that you don't understand. - Use AI for completion only and ensure you can explain and maintain the code before publishing. Transcript: Nick Gnome has decided to ban AI generated code from all Gnome extensions. Their new extension review guidelines state that if there is any clear sign of AI being used to write that code, the extension will not be approved. This apparently comes after the team noticed a whole lot of extensions with what they call unnecessary code patterns, which is basically the AI generating lengthy code instead of Optimizing or finding a more compact way of doing things or just creating methods and functions that don't do anything. One of the reviewers says that he spends six hours a day reviewing 15,000 lines of code, which of course is not sustainable as a volunteer, and during this review process he found tons Of try-catch code blocks that were not needed in a lot of new extensions whose developers then confirmed that they used AI to write them. So what exactly will trigger an extension being rejected? Large amounts of unnecessary code or a coding style that varies too much inside of the same project which is a telltale sign of using AI because it doesn't really have a competent coding Style if you don't teach it first. APIs being used that don't exist will also do the same thing. AI tends to invent calls to stuff that is just not real, so that's obviously a problem. Comments in the code that are clearly LLM prompts, that's also what certain AIs do. Any other clear indication that AI was used to generate large swaths of the code base will trigger an extension being rejected. You can still use AI for code completion, for learning tool, they are just trying to avoid AI being used as the sole creator of the code and so they don't waste a ton of time reviewing something That their author doesn't necessarily understand. They just want to know that the person submitting the extension masters what the extension does, is able to fix problems, and is able to answer questions. If not, they just are not going to waste any time reviewing it. Totally reasonable. As always, AI, use it if you want, but you need to understand what it does and you need to understand how to fix the problems that it implements. ([Time 0:10:48](https://share.snipd.com/snip/f0e4b686-f30c-49cc-bdc0-47a231aa64ed)) - **Rust Graduates In The Linux Kernel** - Rust is now a supported language in the Linux kernel rather than experimental. - This broadens contributor appeal and eases development of drivers like those for Apple Silicon. Transcript: Nick Rust in the Linux kernel is no longer considered experimental. It is now firmly ingrained as one of the languages that you can use to write code for Linux, as in the Linux kernel, alongside C. They're thus removing the experimental tag from all of that Rust integration, and apparently there was no pushback at all on this move, which seems to indicate that the Linux developer And maintainer community has now made peace with Rust after five years of sometimes pretty heated arguments around its inclusion. Of course, C is still everywhere in the kernel, it's still the primary language for most of the codebase, it is not being replaced at all, Rust has its use case, in Linux's case mostly Drivers, but C also has its place and it's still a predominant one. ([Time 0:12:57](https://share.snipd.com/snip/8d556197-34c6-4752-b67f-e77a38f398e8)) - **Mint 22.3 Beta Adds Cinnamon Improvements** - Linux Mint 22.3 beta adds Cinnamon 6.6 with a configurable menu applet and improved input handling. - Nick plans to cover it in a video if the release timing aligns with his schedule. Transcript: Nick Linux Mint 22.3 got its first public beta. The new version comes with Cinnamon 6.6, which gains a new menu applet that is highly configurable by the user, but it also has much better input method handling and an improved on-screen Keyboard, both things that should also help with the Wayland session being more complete. It's still not going to be the default for Mint 22.3. It might be in Mint 23 based on Ubuntu LTS 26.04, but that's not guaranteed. In the settings, you can also configure Thunderbolt support. You also have more options around how you can tile windows. Mint 22.3 also comes with a new system information tool that gives you access to plenty of very well formatted, very nicely placed information about your hardware and the drivers that That hardware relies upon. Think KDE system information tool but with a much better layout because the KDE one seems to just be displaying info straight from a terminal where Mint actually parses it and displays It in a more legible way. There's also a new system administration tool plus the ability to pause file operations in the file manager, also in the time shift backup tool. You also get per app notification indicators, you get better template management, nightlight can now be on at all times if you want that, and there are a few other things. ([Time 0:14:18](https://share.snipd.com/snip/9a23f8ec-5182-4555-8ed2-fe7f2b79f490)) - **Improving Snapdragon Boot Support On Linux** - Fedora changes aim to auto-select DTBs for Snapdragon EFI systems to improve ARM laptop boot experience. - Wider ARM adoption still lags due to performance and power management issues on many devices. Transcript: Nick Now there's good news for ARM support on Linux. As a previous Red Hat engineer, Hans the Good moved to Qualcomm and he's now working on improving support for Snapdragon laptops, something that we recently talked about as being problematic Or even not really working on Linux, depending on the laptop model. His meta changes to Fedora to implement a better out-of experience on Snapdragon-powered devices, notably to make the Fedora ARM64 images boot and detect hardware properly. It's a modified kernel image that includes the necessary device-free files so it can be loaded at startup, it can load the right drivers, where previously users would have to manually Tweak the ISO themselves to add the device 3 files. This should help Fedora support these laptops, at least if the changes are accepted, but it doesn't mean this system on a chip will be well supported. Other distros already have images for these devices, Ubuntu. Even with these images, performance and battery life is pretty spotty compared to x86 CPUs. We also had the example of Tuxedo trying to make a Snapdragon-based laptop work and they're basing their work off of Ubuntu and even with that, it just was not very good. Now, honestly, the only real draw of ARM over x86 for a general user is the battery life. And apparently on Linux, we just cannot reach that because we just don't have good enough support for most of those ARM chips. So right now, you're just better off using an x86 laptop in a lot of cases, unless you can find an actually very well-supported model of an ARM laptop that offers comparable performance And better battery life. ([Time 0:15:55](https://share.snipd.com/snip/f59efef8-e999-4a0c-bae5-a6c091b7c314)) - **Scan And Update Bundled Libraries In Snaps** - Scan snaps with tools like Snapscope to find bundled library CVEs and fix them. - Update bundled dependencies or use maintained runtimes to reduce vulnerability exposure. Transcript: Nick He started a new website called Snapscope, which uses Gripe to scan packages for existing CVEs and vulnerabilities. Gripe can scan a ton of packages for various distros and packages made for specific programming languages as well. The website just lets you search for a package name, an organization, or a developer, and you can see for every single snap package how many open CVEs it has and you have links to all of Them to see what the problem actually is. You also get charts showing recently scanned packages, the highest vulnerability packages as well, and you can just queue snaps that have not been scanned yet for a scan. Generally it seems that most of these vulnerabilities are linked to the bundled libraries that haven't been updated properly by the developers. This is an issue with most of these containerized packaging formats. You can use runtimes and external dependencies that are updated by the makers of the packaging format or by the makers of that dependency. But generally, developers tend to include libraries that either don't have a runtime or that they don't know about inside the package itself. And those libraries tend to just go unupdated for a long time. And so they have vulnerabilities. The sandbox helps a little bit with mitigating that, but it's still a problem. So that's a very interesting tool, not necessarily for users, because I'm sure plenty of people would just never look at that, but for developers, because it just points the finger On this bad practice of bundling tons of libraries and just never updating them at all, just working on the code base for your app, because in a lot of cases, this will result in what you Distribute not being very good for users. ([Time 0:17:49](https://share.snipd.com/snip/b89613d7-524e-44c8-8f1d-845c961dfd8c)) - **Denmark Pilots National Open IT Stack** - Denmark pilots an open-source desktop for 600 traffic authority employees using NixOS and LibreOffice. - The project aims to expand to 15,000 users to avoid vendor lock-in and keep data stored locally. Transcript: Nick They started moving around 600 employees from their road traffic authority to an open source IT system. They expect to migrate 15 000 people if this first pilot works out, and they cite flexibility as the reason to avoid having one supplier setting all the prices, the update policies, And the storage constraints, which just means they want to avoid being stuck with a monopoly, which is what Microsoft is. This open IT system is called SIA Open, and it's based, of course, on Linux, built by a company called Semaphore. It's based on NixOS, it uses LibreOffice as the office suite, it uses Zulip for chat, Jitsi for meetings, and other open source software around that. It seems to be using Cinnamon as the desktop, judging from the menu logo on the screenshot on their website. Of course, using this system also means that the data will be stored in Denmark, so it complies with EU regulations as well, as using Microsoft products is still a bit touchy on that. You don't really know whether your data is being transferred to the US or is used for AI training purposes or anything else, so storing it at home is much better. It's not the entire country moving to Linux and LibreOffice. At some point, I hope all of these efforts start maybe working on a unified solution instead of having each admin try to implement their own Linux distro or their own version of LibreOffice And trying to maintain that over time. Would be nice to have a country-wide effort or even better, an EU-wide effort that could be adopted with strong providers that you can trust for implementing these solutions. Local EU providers, not Microsoft Cloud, please. That would be pretty neat. So hopefully at some point this takes off, but this is starting to look like a big number of implementations being worked on. Hopefully they work out. ([Time 0:20:14](https://share.snipd.com/snip/c6923e68-e88c-41c4-a997-7f706e92ced8)) - **Asahi Linux Merges Key Apple Support** - Asahi Linux continues upstream work: Apple SMC, hwmon, power events, battery drivers, USB3, and microphones. - These merges improve power management and make M1/M2 Macs far more usable under Linux. Transcript: Nick Hopefully they work out. Now, Azahi Linux is making good progress again, with a lot of their work being included in the kernel 6.18 which was released two or three weeks ago. So of course there's the Apple SMC driver which is a crucial building block to build individual device drivers that work with the usual existing Linux subsystems. There's also the hardware monitoring driver that was accepted in the kernel 6.19 so it's coming soon and it will let temperatures, voltage, and current be read and acted upon using The standard Linux hardware monitoring interfaces, which might help with battery life on Apple devices as well. 6.19 and under review are patches to support the power button and switching the lid so just folding your laptop or not which will trigger some events on the system side this is what you'd Expect from a laptop there's also the battery and power supply drivers being merged so again stuff that will help with power management and battery life usb-c support is also almost There with proper usb 3.0 speed supports, M2 Pro and M2 Max microphone support is now implemented too, and there's even work to make sure that the Anaconda installer used in Azari will Hide the Apple-specific partitions to avoid them being deleted by the user if they decide to go with a custom partition scheme, and which would render the Mac sort of a brick. All in all, it's really good progress on the M1 and M2 versions. ([Time 0:22:06](https://share.snipd.com/snip/e5aabd39-6b23-4af3-ad9c-da18520cd433)) - **NVK Gains Large Performance Improvements** - NVK driver merged compression and large-page support in Mesa, boosting VRAM efficiency and throughput. - Game benchmarks show NVK reaching up to ~92% of NVIDIA driver performance on some titles. Transcript: Nick And finally, the NVK driver is also getting some attention in Mesa 26 because they have merged support for compression. This just lets the driver compress images that are rendered and of course decompress them so things use less VRAM and they can be transferred faster around the system for, of course, A huge boost in performance compared to the previous state of the driver. They've also merged support for larger page sizes, again, to let the driver store information in bigger chunks, which also means fewer operations need to be performed and you get better Performance. What kind of performance? Well, the merge request has a link to a Google Docs which indicates the FPS reached in a few games with these things enabled and it's quite impressive. Atomic Heart, with all of these changes, reaches 67% of the performance of the official Nvidia driver. Deep Rock Galactic is at 63%, Subnautica at 61%, it goes up to 92% for Serious Sam 2017. Basically, it seems that the older the title or the simpler it is graphically, the closer it can get to the official proprietary driver. There are exceptions, XCOM 2, for example, only reaches 44% of the performance of the official drivers, but generally, it is super impressive work. This means that NVK and Nuvo are catching up. And honestly, at 70, 75, 80% of the performance, a lot of people are going to take that trade-off instead of having to tack on an external module from NVIDIA, which oftentimes breaks. ([Time 0:23:52](https://share.snipd.com/snip/c8fce119-0c39-4c4a-b8c0-2eddd21dfaf0))