The thing is: it already is the year of the Linux Desktop for me. I don't care about 'OS market share' or how many people use something; I have no control over them.
I also don't care about "OS-maxxing", either--quibbling over 'Wayland', or which OS has the best window manager, arguing about 'gaming', etc.
What I do care about? Freedom and control. Linux gives me that.
So my desktop? It's Linux. The Year of Linux on the Desktop arrived for me years ago. And it can be that year for anyone, anytime. Today.
I love Linux and think it's better than ever to have Linux as your daily driver, especially thanks to the work of Valve with Proton, but I'm gonna be real and say "the year of the Linux Desktop is a personal journey" is a retrofit. The "Year of the Linux Desktop" meme came out of Slashdot in the 90s/early 00s where people were insisting that Linux was due to overtake Windows as the way an average, non-technical user interacted with a computer.
Of course, this did turn out to be true... in the form of Android, which is maybe the most monkeys-paw-curling way YotLD could possibly happen.
Agreed, mostly. To me the effectiveness of window managers is a bellwether of the control aspect. So, IMO, if you compare them on control or on the quality of window managers, you'll get the same result. Linux has ended up with many window managers (effectively catering to various styles and needs), while macos, for example, has a one-size-fits-all approach with no REAL multiple-desktops (Spaces is a joke and a toy). As a result, I can easily manage 50-100 open terminal windows on my linux box, but on macos 5 is too many (and if I use Spaces to fit a few more, I can't get to the terminal I need in less than a second. In linux, even with 100 xterms, I can). So, a macos product manager might probably ask me why I would want 100 terminal windows, to find me some alternative. In linux no one asks me that question, and hence I can have what I want.
> What I do care about? Freedom and control. Linux gives me that.
I think a lot of this comes down to what we're looking for out of an OS. For example, it is orthogonal to what I care about most: "Provides a low friction interface between my body and arbitrary software"
Relevant: I do think about the freedom and control aspect about computing; I (personally) tie it to the software or hardware design instead of the interface. Or in some cases, the use of creative software. (DAW, CAD, document writers etc)
The object-level discussion is interesting, but I disagree with the premise to such an extent it feels like a moot point. It feels like the article doesn't play out the line to its logical conclusion.
Why would agents want GUIs made for humans? It's already the case that, like everyone who's good at computers, agents want a terminal and good APIs, not some ad-ridden crap.
If anything, AI is a reason why it will never be the year of the linux desktop but also it doesn't matter anymore, because if the higher-order bit of productivity is defined by AI, then my tmux+vim is as good as your Visual Studio.
You probably don't need tmux. The utility is really when you're remoting into machines and want to keep your session (or are too lazy to use nohup or disown)
Your terminal should split panes for and do tabs. Ghostty is my preferred but use whatever. And fwiw, even if your terminal sucks vim can do this all for you too (:term), so you don't even need to leave vim.
> vim is better than your Visual Studio.
FTFY ;)
Side note: just because you can live in the terminal on Linux doesn't mean GUIs can't exist or are even second class citizens. The real beauty is being able to have both. You can have a platform that is usable for most people while not fucking over power users. Wild concept, I know
My terminal splits panes (which I don't use), but what if I want to open two terminals that share the same set of splits? Can't. But tmux can!
What if I want to SSH back into my desktop (because I'm on a laptop or whatever) and grab something from my desktop terminal? Can't. But tmux can!
Vim splits and the vim terminal are poorly implemented. Technically, yes, they work. But you'll run into a lot of issues. I know, because a few years ago I went down the same path: Why do I need tmux, when I have vim!? ... I quickly learned why I needed tmux.
I agree with your side note: plasma+kitty+tmux and a few support scripts:
(please don't criticize my scripts; these were never meant to be shared, and it's a disaster, but it works for me)
I have this script (https://doc.xn0.org/tmuxedkitty-newwindow.sh) bound to WIN+T; it opens kitty, and either creates a new tmux session if there isn't one or attaches to the existing session and creates a new pane.
Then, I have my insane (I understand I am insane, but it works for me!) tmux config file: https://doc.xn0.org/.tmux.conf
> but what if I want to open two terminals that share the same set of splits?
You want clones? I'll admit most terminals can't do this (some can), but I'm struggling to see the use case. What's the advantage of having 2 windows displaying the same information?
> What if I want to SSH back into my desktop
Agreed! That was the explicitly stated usecase where I said tmux was for[0]
> Vim splits and the vim terminal are poorly implemented.
Completely fair and I avoid for exactly those reasons. But they're still handy in a pinch and they're good to know about
BUT tmux is also poorly implemented. Start trying to use sixel (or kitty graphics) in your fzf previews, yazi, or whatever you're displaying things with. This is a big pain point.
> please don't criticize my scripts
Do you want friendly comments? All code sucks so I'll not going to call you dumb or anything. But do upload somewhere so I don't have to download 0x0.st is perfect for this usecase.
> Using titles
Your terminal doesn't do titles? What terminal are you using?
[0] I'll also admit Claude code is another use case. But that is because it is so poorly written not because the terminals suck. I absolutely believe Dario when he says Claude does most of the coding... it shows...
Just because tmux doesn't work for you doesn't mean it can't be useful for someone else. I for one really appreciate having the same interface and keybinds across several devices and I've never felt a need to look elsewhere.
I've always interpreted "Year of the Linux Desktop" as a personal journey, like Hot Girl Summer. It's not about the year that there's a watershed and suddenly everyone is rolling custom distros, it's about an individual's journey with discovering and trying Linux. Every year can be the year of the Linux desktop if you believe!
Interesting explanation of a subject I had no knowledge of! I'm familiar with browser accessibility trees, but I've never thought about how operating systems do it themselves.
From the outside view, I still wouldn't make any bets with 100% certainty about the future of anything to do with computers.
If you grant that there is some chance that the trends of programming models' capabilities will continue for another few years, then there is some chance that software and its bottlenecks will be completely transformed. A rapidly overhauled accessibility tree for linux? A good-enough computer use model that doesn't require accessibility trees at all? A world of bespoke, personalized operating systems? All of these things (and many more) seem like outcomes with non-zero probabilities.
It can be in Spain. Since three years ago, all the public administrations webs can be used correctly from Linux. You don't need any privative software any longer. Also it is gone for good the "Works only with IE", so unfortunately common some years ago.
Yeahhhh…this is not really how Linux works, though.
Most of LLM world is kind of anti-linux right now because the most popular LLMs are walled off by these huge companies and hella expensive. At some point, a nerd will realize they could hack together a surprisingly ok homebrew version of what everybody else is using, and do. Then a company realizes that they can build a brand on the anarchist, grassroots vibe of the homebrew thing, and capitalize its development (software development, but also community development, which is brand development for the company). Now, it’s much later, but the open source thing is competitive, and popular for being open-source.
At one point I got interested in why Red Hat handed over tens of millions of dollars in stock to Linus leading up to their IPO, in exchange for…nothing specific. Nominally it was a gift of appreciation, but handing out random gifts is somewhat opposed to maximizing shareholder returns. It’s because Linus controls merges to the Linux kernel and doesn’t have to care about Red Hat, and the board wanted him to care at least a little bit. They were stuck between “people trust our business because it’s built on this populist OS” and “this populist OS is mostly controlled by a guy who doesn’t work for us.” It’s hard to have one without the other.
I’m glad Apple is taking accessibility seriously, and I wish accessibility worked better on Linux, but I don’t think Linux is ever going to make developers “do their homework,” because the community wouldn’t trust a Linux like that. If the author is right, it’ll happen because “AI for the People, Inc.” builds a business on it and sponsors the work.
This article doesn't make sense to me. The terminals are the *nix accessibility stack for all practical purposes and agents can use this interface just fine. What am I missing?
There is and will always be a 1-3% minority in every aspect: rich, deeply educated, talents, tall, etc etc. It's the right tail of the Bell curve. In OSes, this is Linux.
I personally don't find a need for "agents" to use my Desktop. If the agents need to access data, they seem to manage perfectly fine with other APIs. I'm not going to switch to macOS just so that agents can click buttons on a UI for me.
I would like to see a non-big-corp-controlled (e.g. Open source) OS that is focused on single-user systems. (Personal /"Desktop" computers) ABI compatibilty, no sudo or permissions; "just works". Schedule software, provide a GUI, threads, memory allocation etc. But get out of the way; no complicated user system; no delicate balance of text config files scattered throughout a file system.
Currently, OSS (etc) OSes are synonymous with Linux; I don't think I will ever see eye to eye with the Linux design philosophy; too many compromises which prioritize servers, multi-user IT systems; embraces scattered state across the FS etc.
You might like FreeBSD. It’s unapologetically unix. Jails + root on ZFS is arguably what Linux tried to move to, but (due to too many compromises) will probably never achieve.
15.1 (due out next week) is their first big laptop focused release from the recent grant money. They have a list of current laptops that they test with for you to buy.
Barring that, Devuan Linux is not bad either, and is still a cohesive system.
I appreciate the rec! Hah; networking is indeed one of the things I think it would be good for a GPOS to have (e.g. fits with threads, allocator etc). Also interfaces for the MB's RTC for datetimes etc... Some day?
>This is not the kind of gap a community closes by writing better software. It is the kind of gap that takes a decade of full-time employees auditing every label in every default app, a market mechanism that punishes you when you don’t, and a centralized review process to enforce it from above.
We have seen a bigger push to get everything using XDG config dirs in recent years, and also getting everything working on Wayland. These to me seem similar, other than that this accessibility standard would be even more niche, and if it was stated upfront to be made with AI in mind, I think there would be resistance.
Personally I do not want to let an AI tool run loose on my machine, but I do like having ways to script and automate stuff. I like Sway's IPC and that every keybind is also a command you could run. So the explanation of Apple's accessibility stuff sounds cool. I wish I had something like Unity's HUD where I could use a search to select from any depth of graphical menus in a given program instead of having to poke around by hand. If the accessibility standard were like that and allowed more stuff to be done from the CLI more easily, that would be great.
> If you use a Mac and open the Accessibility Inspector tool that’s built into the system (you really should try it), you can see a second version of the computer, hiding inside the first one. The first version is the one you look at: windows, shadows, rounded rectangles, a little bouncing icon in the Dock from Slack announcing that you are falling behind.
Now use that Accessibility Inspector tool inside Slack (an Electron App) and you'll be welcomed to a deeply nested tree of unlabelled objects.
Codex's computer use came from OpenAI's acquisition of the Apple Shortcuts team, whose institutional knowledge allowed them to exploit all sorts of undocumented macOS APIs, not some virtuous accessibility* stack. With 99% of work happening on the web anyway, it IS fair to say that it's not the year of the Linux desktop, or any desktop, because the desktop doesn't need to exist at all.
*macos26 introduced a multitude of accessibility regressions that have real-world impact on humans with disabilities, let alone AI
It doesn't mind what Chromebooks have under the hood. They don't make the Year of the Linux Desktop closer rather the another Decade of Chrome browser.
What is the metric for this claim? A based-linux desktop enviroment used more than MacOS and Windows? Maybe the same AI-corps will be able to expend money creating the gui api for their agents. It shouldn't be a problem for them. What can stop them to do that?
This article pre-supposes that the primary way AI agents will do tasks for users will be through through usage of desktop applications instead of documented APIs. While desktop app usage could be very important during a transition period of agent-computer use, I think it makes far more sense that agents will standardize on the protocols that are already being developed, like MCP. An explicitly documented API will always be better for a machine to use than clicking around and navigating an interface for humans.
"On Linux under Wayland" is a big part of the problem. On X11 a significant part of missing "GUI-exposed-as-api" is present.
If we concede (and I think otherwise) that we need a FOSS operating system and desktop experience to be fully on par with competitors and offer agentic-first options, I think that an open-minded developer (or one that can afford to run a fairly good LLM on local machine), presented with the problem, can see evidently that said roadblock doesn't exist: X11 can stop being a maze, or thousands of Wayland apps can be forked to make them expose an API, the FUSE filesystem kind of API.
I don't care much about agents though, I sure see as potentially useful some desktop assistant, and that is that.
Wayland is a major regression for accessibility. People have been raising that for years, but always shouted down by the brigade with the motte-and-bailey seesaw between “Wayland replaces X11” and “Wayland is just a protocol”.
There's too much splintering in the community now and the incentive drawbridge for FOSS feels like it's closing. Many of the older maintainers who carried projects for decades are starting to step away due to age, burnout, or simple exhaustion.
It feels nowadays that if someone genuinely puts the effort into untangling a codebase, fixing the long-standing issues and navigating the maze of legacy paths, the backlash and politics around it leaves the project deflated and unappreciated. The reaction around Xlibre shows how hostile these situations can become. Personally, I still prefer Xorg in many ways, even if Wayland is technically the newer direction.
FOSS was a powerful ideological concept in the 90s when most software was proprietary and corporate-controlled. The 2000s felt like real growth and experimentation. Today, a lot of it feels fragmented, cynical, and increasingly institutionalized. Another problem with FOSS is that projects usually end up in one of a few states, a strong “dictator” model where someone drives the direction through sheer effort and resources,
a loose community model where everything gets patched together by committee or eventual corporate stewardship.
Or someone gets frustrated enough to fork a project, but then these forks are often treated socially as hostile when that was the core ideology of FOSS.
A lot of modern “FOSS” infrastructure are effectively: corporate-funded, corporate-prioritized or community-accessible rather than community-directed. It's free in the sense of that you can download the source, compile it yourself but for yourself to contribute unless you have the resources, is hard and time consuming. So you end up following the path it's taking.
If the corporations benefiting from these ecosystems consistently reinvested back into the communities maintaining them the FOSS landscape would look very different if not more healthier, more sustainable, and far less bitter.
AI does not need a “desktop” at all for itself. So, what this article is talking about is AI driven user assistance on the desktop. And, for that, the limiting factor is what desktop the user wants to use.
Coding agents run well in a Linux VM and you can run Linux in a VM just about anywhere. A coding agent can be bundled with tools in a Docker file, or it can apt-get lots of useful tools if it needs to. They don’t need a desktop or desktop apps. Why go through an accessibility tree when you can make http requests?
So I expect that we will see more and more Linux VM’s. Maybe it will be like Sqlite, ubiquitous but hidden?
For me the Year of the Linux Desktop is every year since around 2000.
I dual booted Windows since 95, also tried Mac OSX on $job but nothing comes close to the peace of mind of using Linux.
I have lived through spotty hardware support (fixed), install editing too many files (fixed), no games (fixed) and several other problems, but even in the worst of times it is a software that respects you as a user.
Agree with OP. Not because of the accessibility API argument but because of the "small things" like Microsoft office, drivers, the sound not working out of speakers but working with headphones. These small problems have gone unfixed for years, or have become worse, and is the main reason why a non tech person won't transition to Linux.
There might be a so-called "Year of the Linux Desktop", but it'd require Microsoft either doing something so disastrous that people cannot use Windows, or pivoting away from NT.
I don't see why AI agents need to use the GUI very much? If anything, all the major advances with AI agents have been in CLI domains that Linux is perfectly well adapted to. Besides, surely AI agents could just contribute code allowing them to use Linux, no?
I live in the terminal and that means I can speak the same language as agents[0], but I'm human and also enjoy GUIs and TUIs.
There is no reason a machine can't be usable by normal people AND make power users happy
I don't know why we believe we're have to fuck over power users. The computer as a machine (which includes your phone) doesn't work without that. The problem is we've been creating walled gardens and saying "power users don't matter, they're a small percentage of users". The percentage of the userbase is a meaningless statistic. It's a proxy for other things (usually revenue).
But your power users are the ones who write programs, fix bugs, and do free work that makes your product more valuable. Apple/others would still make money on every app even if they were to stop taking a cut of sales and stop charging money for developer licenses. They make money by selling the fucking hardware and shipping the OS. They create the environment, they create the default base experience. But no one buys a smartphone or computer without installing new software. Being able to do that is the reason these products are do great.
That's the real reason Linux is starting to win. Because Microsoft and Apple have lost touch. They got greedy, egotistical[1], and myopic.
[0] people keep asking me if it's worth learning now with AI. Bash and terminal knowledge have only become *more* valuable, not less. Think of it this way: do you want a manager with an MBA or one that's been an engineer?
[1] Apple is famous for blaming the user and not recognizing the design is bad. "People wouldn't buy it if it was bad". Well guys what? People won't buy it if it's bad AND there's another option. That's the variable that changed: options
A great write-up, with a fair description of Linux. It cites facts and gives some explanations. And it brings good news: My KDE Plasma Linux Desktop running on X11 is completely safe from being hijacked by agentic robots. I run Wayland only for the few games that don't work on X11, then I switch back again because Mathematica does not like Wayland.
Linux is an OS that lets me do what I want when I want it and importantly _doesn't_ do a whole lot of shit that I _don't_ want it to do, as opposed to macOS or WinDOS. That includes robots wreaking havoc on my desktop.
Being able to run the software you want to run is very important. But not being forced to run software you don't want to run is even more important. And neither Microslop nor Apple respect the wishes of their users in these regards.
>There are many reasons for this. Drivers. Games. Adobe. Microsoft Office. Battery life. The thing where you close the lid of a laptop and open it again later to find that it passed into the good night.
The last one is a huge problem for Windows as well. Its due to Microsoft discontinuing support for S3 sleep mode, which in turn, caused motherboard manufacturers to discontinue S3 support in the BIOS. Which means its no longer available even if you install Linux on the laptop since it requires firmware support to work. You can still find laptops that support S3 sleep if you really look hard enough. Or buy a Mac.
My first instinct was to just not open the article based on the headline. But I thought, "what if there's a good point that I, as a Linux user, should be aware of?"
It was worse than I imagined it would be. I now deeply regret giving this article a click.
Basically, it's all about how AI can use Mac OS features.
Eh, the point is interesting, but I'm not sure it's not solvable. Beyond that, I'm quite hopeful at linux breaking out in a big way in the next couple of years via chromebooks. My theory is that we'll start seeing a hockey stick graph of ai-found/exploited windows zero days, and in response we'll see a dramatic acceleration adoption of chromebooks. Voila, YotLD.
It has many advantages but also serious issues. On the latest Ubuntu, I started downloading a game from the App Store during a background OS update and it locked up so bad I had to run the terminal with a hot key to salvage it. That’s not a real desktop experience.
...I wasn't expecting the argument to be that Linux interoperates poorly with AI Agents lol.
I think the author is actually on the right track at first then dismisses it with: These are "why a person did not switch to Linux last" and not "why the desktop, as an institution, will continue to belong to Apple and Microsoft". You can absolutely get to the root cause of the former and find foundational issues that explain the latter.
The thing is: it already is the year of the Linux Desktop for me. I don't care about 'OS market share' or how many people use something; I have no control over them.
I also don't care about "OS-maxxing", either--quibbling over 'Wayland', or which OS has the best window manager, arguing about 'gaming', etc.
What I do care about? Freedom and control. Linux gives me that.
So my desktop? It's Linux. The Year of Linux on the Desktop arrived for me years ago. And it can be that year for anyone, anytime. Today.
I love Linux and think it's better than ever to have Linux as your daily driver, especially thanks to the work of Valve with Proton, but I'm gonna be real and say "the year of the Linux Desktop is a personal journey" is a retrofit. The "Year of the Linux Desktop" meme came out of Slashdot in the 90s/early 00s where people were insisting that Linux was due to overtake Windows as the way an average, non-technical user interacted with a computer.
Of course, this did turn out to be true... in the form of Android, which is maybe the most monkeys-paw-curling way YotLD could possibly happen.
Agreed, mostly. To me the effectiveness of window managers is a bellwether of the control aspect. So, IMO, if you compare them on control or on the quality of window managers, you'll get the same result. Linux has ended up with many window managers (effectively catering to various styles and needs), while macos, for example, has a one-size-fits-all approach with no REAL multiple-desktops (Spaces is a joke and a toy). As a result, I can easily manage 50-100 open terminal windows on my linux box, but on macos 5 is too many (and if I use Spaces to fit a few more, I can't get to the terminal I need in less than a second. In linux, even with 100 xterms, I can). So, a macos product manager might probably ask me why I would want 100 terminal windows, to find me some alternative. In linux no one asks me that question, and hence I can have what I want.
> What I do care about? Freedom and control. Linux gives me that.
I think a lot of this comes down to what we're looking for out of an OS. For example, it is orthogonal to what I care about most: "Provides a low friction interface between my body and arbitrary software"
Relevant: I do think about the freedom and control aspect about computing; I (personally) tie it to the software or hardware design instead of the interface. Or in some cases, the use of creative software. (DAW, CAD, document writers etc)
The object-level discussion is interesting, but I disagree with the premise to such an extent it feels like a moot point. It feels like the article doesn't play out the line to its logical conclusion.
Why would agents want GUIs made for humans? It's already the case that, like everyone who's good at computers, agents want a terminal and good APIs, not some ad-ridden crap.
If anything, AI is a reason why it will never be the year of the linux desktop but also it doesn't matter anymore, because if the higher-order bit of productivity is defined by AI, then my tmux+vim is as good as your Visual Studio.
Your terminal should split panes for and do tabs. Ghostty is my preferred but use whatever. And fwiw, even if your terminal sucks vim can do this all for you too (:term), so you don't even need to leave vim.
FTFY ;)Side note: just because you can live in the terminal on Linux doesn't mean GUIs can't exist or are even second class citizens. The real beauty is being able to have both. You can have a platform that is usable for most people while not fucking over power users. Wild concept, I know
<strongly offended noises>
Everybody needs tmux, especially locally.
My terminal splits panes (which I don't use), but what if I want to open two terminals that share the same set of splits? Can't. But tmux can!
What if I want to SSH back into my desktop (because I'm on a laptop or whatever) and grab something from my desktop terminal? Can't. But tmux can!
Vim splits and the vim terminal are poorly implemented. Technically, yes, they work. But you'll run into a lot of issues. I know, because a few years ago I went down the same path: Why do I need tmux, when I have vim!? ... I quickly learned why I needed tmux.
I agree with your side note: plasma+kitty+tmux and a few support scripts:
(please don't criticize my scripts; these were never meant to be shared, and it's a disaster, but it works for me)
I have this script (https://doc.xn0.org/tmuxedkitty-newwindow.sh) bound to WIN+T; it opens kitty, and either creates a new tmux session if there isn't one or attaches to the existing session and creates a new pane.
Then, I have my insane (I understand I am insane, but it works for me!) tmux config file: https://doc.xn0.org/.tmux.conf
Then, I have my insane zshrc that auto-titles my tmux windows: https://doc.xn0.org/.zshrc
Using titles from: https://doc.xn0.org/tmux-window-titles
I have put way too much thought and time into this...
BUT tmux is also poorly implemented. Start trying to use sixel (or kitty graphics) in your fzf previews, yazi, or whatever you're displaying things with. This is a big pain point.
Do you want friendly comments? All code sucks so I'll not going to call you dumb or anything. But do upload somewhere so I don't have to download 0x0.st is perfect for this usecase. Your terminal doesn't do titles? What terminal are you using?[0] I'll also admit Claude code is another use case. But that is because it is so poorly written not because the terminals suck. I absolutely believe Dario when he says Claude does most of the coding... it shows...
Just because tmux doesn't work for you doesn't mean it can't be useful for someone else. I for one really appreciate having the same interface and keybinds across several devices and I've never felt a need to look elsewhere.
I've always interpreted "Year of the Linux Desktop" as a personal journey, like Hot Girl Summer. It's not about the year that there's a watershed and suddenly everyone is rolling custom distros, it's about an individual's journey with discovering and trying Linux. Every year can be the year of the Linux desktop if you believe!
I like this definition: https://lwn.net/Articles/928581/
> More importantly for the future success of Linux was that the X11 system was ported to it, making 1992 the year of the Linux desktop.
The year of the linux Desktop was the friends we made along the way... fixing our Linux desktop.
I love this. 2026 has been my hot linux summer on the desktop
Agreed. For me it was over a decade ago and the meme stopped being funny. It feels like people bragging about being behind, which is uncomfortable.
Interesting explanation of a subject I had no knowledge of! I'm familiar with browser accessibility trees, but I've never thought about how operating systems do it themselves.
From the outside view, I still wouldn't make any bets with 100% certainty about the future of anything to do with computers.
If you grant that there is some chance that the trends of programming models' capabilities will continue for another few years, then there is some chance that software and its bottlenecks will be completely transformed. A rapidly overhauled accessibility tree for linux? A good-enough computer use model that doesn't require accessibility trees at all? A world of bespoke, personalized operating systems? All of these things (and many more) seem like outcomes with non-zero probabilities.
It can be in Spain. Since three years ago, all the public administrations webs can be used correctly from Linux. You don't need any privative software any longer. Also it is gone for good the "Works only with IE", so unfortunately common some years ago.
Yeahhhh…this is not really how Linux works, though.
Most of LLM world is kind of anti-linux right now because the most popular LLMs are walled off by these huge companies and hella expensive. At some point, a nerd will realize they could hack together a surprisingly ok homebrew version of what everybody else is using, and do. Then a company realizes that they can build a brand on the anarchist, grassroots vibe of the homebrew thing, and capitalize its development (software development, but also community development, which is brand development for the company). Now, it’s much later, but the open source thing is competitive, and popular for being open-source.
At one point I got interested in why Red Hat handed over tens of millions of dollars in stock to Linus leading up to their IPO, in exchange for…nothing specific. Nominally it was a gift of appreciation, but handing out random gifts is somewhat opposed to maximizing shareholder returns. It’s because Linus controls merges to the Linux kernel and doesn’t have to care about Red Hat, and the board wanted him to care at least a little bit. They were stuck between “people trust our business because it’s built on this populist OS” and “this populist OS is mostly controlled by a guy who doesn’t work for us.” It’s hard to have one without the other.
I’m glad Apple is taking accessibility seriously, and I wish accessibility worked better on Linux, but I don’t think Linux is ever going to make developers “do their homework,” because the community wouldn’t trust a Linux like that. If the author is right, it’ll happen because “AI for the People, Inc.” builds a business on it and sponsors the work.
This article doesn't make sense to me. The terminals are the *nix accessibility stack for all practical purposes and agents can use this interface just fine. What am I missing?
There is and will always be a 1-3% minority in every aspect: rich, deeply educated, talents, tall, etc etc. It's the right tail of the Bell curve. In OSes, this is Linux.
I personally don't find a need for "agents" to use my Desktop. If the agents need to access data, they seem to manage perfectly fine with other APIs. I'm not going to switch to macOS just so that agents can click buttons on a UI for me.
I would like to see a non-big-corp-controlled (e.g. Open source) OS that is focused on single-user systems. (Personal /"Desktop" computers) ABI compatibilty, no sudo or permissions; "just works". Schedule software, provide a GUI, threads, memory allocation etc. But get out of the way; no complicated user system; no delicate balance of text config files scattered throughout a file system.
Currently, OSS (etc) OSes are synonymous with Linux; I don't think I will ever see eye to eye with the Linux design philosophy; too many compromises which prioritize servers, multi-user IT systems; embraces scattered state across the FS etc.
EasyOS is a very unlinuxy Linux that somebody here mentioned once when I was having a similar whine.
https://easyos.org/about/how-and-why-easyos-is-different.htm...
Note the remarks about sudo, folders, and GUIs.
You might like FreeBSD. It’s unapologetically unix. Jails + root on ZFS is arguably what Linux tried to move to, but (due to too many compromises) will probably never achieve.
15.1 (due out next week) is their first big laptop focused release from the recent grant money. They have a list of current laptops that they test with for you to buy.
Barring that, Devuan Linux is not bad either, and is still a cohesive system.
Sounds like TempleOS checks all those boxes. You might miss networking though.
I appreciate the rec! Hah; networking is indeed one of the things I think it would be good for a GPOS to have (e.g. fits with threads, allocator etc). Also interfaces for the MB's RTC for datetimes etc... Some day?
>This is not the kind of gap a community closes by writing better software. It is the kind of gap that takes a decade of full-time employees auditing every label in every default app, a market mechanism that punishes you when you don’t, and a centralized review process to enforce it from above.
We have seen a bigger push to get everything using XDG config dirs in recent years, and also getting everything working on Wayland. These to me seem similar, other than that this accessibility standard would be even more niche, and if it was stated upfront to be made with AI in mind, I think there would be resistance.
Personally I do not want to let an AI tool run loose on my machine, but I do like having ways to script and automate stuff. I like Sway's IPC and that every keybind is also a command you could run. So the explanation of Apple's accessibility stuff sounds cool. I wish I had something like Unity's HUD where I could use a search to select from any depth of graphical menus in a given program instead of having to poke around by hand. If the accessibility standard were like that and allowed more stuff to be done from the CLI more easily, that would be great.
> If you use a Mac and open the Accessibility Inspector tool that’s built into the system (you really should try it), you can see a second version of the computer, hiding inside the first one. The first version is the one you look at: windows, shadows, rounded rectangles, a little bouncing icon in the Dock from Slack announcing that you are falling behind.
Now use that Accessibility Inspector tool inside Slack (an Electron App) and you'll be welcomed to a deeply nested tree of unlabelled objects.
Codex's computer use came from OpenAI's acquisition of the Apple Shortcuts team, whose institutional knowledge allowed them to exploit all sorts of undocumented macOS APIs, not some virtuous accessibility* stack. With 99% of work happening on the web anyway, it IS fair to say that it's not the year of the Linux desktop, or any desktop, because the desktop doesn't need to exist at all.
*macos26 introduced a multitude of accessibility regressions that have real-world impact on humans with disabilities, let alone AI
> because the desktop doesn't need to exist at all.
Which is a really strong argument for most people just buying chromebooks, which run linux.
It doesn't mind what Chromebooks have under the hood. They don't make the Year of the Linux Desktop closer rather the another Decade of Chrome browser.
Or just use your phone, remoted into any machine anywhere
macos26 is just one big UI regression. Ugh.. so much wasted screenspace.
I love Linux. It's my job, and it's my preference everywhere.
That being said, of course it will never be the year of the Linux desktop because Linux is a GPL licensed kernel.
Pretty sure it's been the year of the Linux desktop for 30 years for me...
What is the metric for this claim? A based-linux desktop enviroment used more than MacOS and Windows? Maybe the same AI-corps will be able to expend money creating the gui api for their agents. It shouldn't be a problem for them. What can stop them to do that?
"The Year of the Linux Desktop" isn't a time period, it's the friends we made along the way.
Like the sisterhood of the traveling distro?
This article pre-supposes that the primary way AI agents will do tasks for users will be through through usage of desktop applications instead of documented APIs. While desktop app usage could be very important during a transition period of agent-computer use, I think it makes far more sense that agents will standardize on the protocols that are already being developed, like MCP. An explicitly documented API will always be better for a machine to use than clicking around and navigating an interface for humans.
"On Linux under Wayland" is a big part of the problem. On X11 a significant part of missing "GUI-exposed-as-api" is present. If we concede (and I think otherwise) that we need a FOSS operating system and desktop experience to be fully on par with competitors and offer agentic-first options, I think that an open-minded developer (or one that can afford to run a fairly good LLM on local machine), presented with the problem, can see evidently that said roadblock doesn't exist: X11 can stop being a maze, or thousands of Wayland apps can be forked to make them expose an API, the FUSE filesystem kind of API.
I don't care much about agents though, I sure see as potentially useful some desktop assistant, and that is that.
Wayland is a major regression for accessibility. People have been raising that for years, but always shouted down by the brigade with the motte-and-bailey seesaw between “Wayland replaces X11” and “Wayland is just a protocol”.
There's too much splintering in the community now and the incentive drawbridge for FOSS feels like it's closing. Many of the older maintainers who carried projects for decades are starting to step away due to age, burnout, or simple exhaustion.
It feels nowadays that if someone genuinely puts the effort into untangling a codebase, fixing the long-standing issues and navigating the maze of legacy paths, the backlash and politics around it leaves the project deflated and unappreciated. The reaction around Xlibre shows how hostile these situations can become. Personally, I still prefer Xorg in many ways, even if Wayland is technically the newer direction.
FOSS was a powerful ideological concept in the 90s when most software was proprietary and corporate-controlled. The 2000s felt like real growth and experimentation. Today, a lot of it feels fragmented, cynical, and increasingly institutionalized. Another problem with FOSS is that projects usually end up in one of a few states, a strong “dictator” model where someone drives the direction through sheer effort and resources, a loose community model where everything gets patched together by committee or eventual corporate stewardship.
Or someone gets frustrated enough to fork a project, but then these forks are often treated socially as hostile when that was the core ideology of FOSS.
A lot of modern “FOSS” infrastructure are effectively: corporate-funded, corporate-prioritized or community-accessible rather than community-directed. It's free in the sense of that you can download the source, compile it yourself but for yourself to contribute unless you have the resources, is hard and time consuming. So you end up following the path it's taking.
If the corporations benefiting from these ecosystems consistently reinvested back into the communities maintaining them the FOSS landscape would look very different if not more healthier, more sustainable, and far less bitter.
AI does not need a “desktop” at all for itself. So, what this article is talking about is AI driven user assistance on the desktop. And, for that, the limiting factor is what desktop the user wants to use.
Coding agents run well in a Linux VM and you can run Linux in a VM just about anywhere. A coding agent can be bundled with tools in a Docker file, or it can apt-get lots of useful tools if it needs to. They don’t need a desktop or desktop apps. Why go through an accessibility tree when you can make http requests?
So I expect that we will see more and more Linux VM’s. Maybe it will be like Sqlite, ubiquitous but hidden?
For me the Year of the Linux Desktop is every year since around 2000.
I dual booted Windows since 95, also tried Mac OSX on $job but nothing comes close to the peace of mind of using Linux.
I have lived through spotty hardware support (fixed), install editing too many files (fixed), no games (fixed) and several other problems, but even in the worst of times it is a software that respects you as a user.
Agree with OP. Not because of the accessibility API argument but because of the "small things" like Microsoft office, drivers, the sound not working out of speakers but working with headphones. These small problems have gone unfixed for years, or have become worse, and is the main reason why a non tech person won't transition to Linux.
Since AI is so smart now, these problems will surely be fixed in no time!
I wish somebody would make a Polymarket bet out of this. I'm 100% with the author on this one
I’d take the other side of that bet if i didn’t think gambling was a cancer
Good thing health insurance is so cheap where I live. I really don't mind risking a little cancer for a good bet
I'll bet you gambling isn't a cancer
There might be a so-called "Year of the Linux Desktop", but it'd require Microsoft either doing something so disastrous that people cannot use Windows, or pivoting away from NT.
I don't see why AI agents need to use the GUI very much? If anything, all the major advances with AI agents have been in CLI domains that Linux is perfectly well adapted to. Besides, surely AI agents could just contribute code allowing them to use Linux, no?
Because there is a world of software out there that isn't CLI-based and much of it may never be updated to expose LLM-friendly APIs.
Of course, but how many of those are really relevant for AI agents/couldn't be done through another means.
It matters if it's relevant to the person asking the AI agent for help with what they're doing with the software that they already have.
Weird argument.
I live in the terminal and that means I can speak the same language as agents[0], but I'm human and also enjoy GUIs and TUIs.
I don't know why we believe we're have to fuck over power users. The computer as a machine (which includes your phone) doesn't work without that. The problem is we've been creating walled gardens and saying "power users don't matter, they're a small percentage of users". The percentage of the userbase is a meaningless statistic. It's a proxy for other things (usually revenue).But your power users are the ones who write programs, fix bugs, and do free work that makes your product more valuable. Apple/others would still make money on every app even if they were to stop taking a cut of sales and stop charging money for developer licenses. They make money by selling the fucking hardware and shipping the OS. They create the environment, they create the default base experience. But no one buys a smartphone or computer without installing new software. Being able to do that is the reason these products are do great.
That's the real reason Linux is starting to win. Because Microsoft and Apple have lost touch. They got greedy, egotistical[1], and myopic.
[0] people keep asking me if it's worth learning now with AI. Bash and terminal knowledge have only become *more* valuable, not less. Think of it this way: do you want a manager with an MBA or one that's been an engineer?
[1] Apple is famous for blaming the user and not recognizing the design is bad. "People wouldn't buy it if it was bad". Well guys what? People won't buy it if it's bad AND there's another option. That's the variable that changed: options
A great write-up, with a fair description of Linux. It cites facts and gives some explanations. And it brings good news: My KDE Plasma Linux Desktop running on X11 is completely safe from being hijacked by agentic robots. I run Wayland only for the few games that don't work on X11, then I switch back again because Mathematica does not like Wayland.
Linux is an OS that lets me do what I want when I want it and importantly _doesn't_ do a whole lot of shit that I _don't_ want it to do, as opposed to macOS or WinDOS. That includes robots wreaking havoc on my desktop.
Being able to run the software you want to run is very important. But not being forced to run software you don't want to run is even more important. And neither Microslop nor Apple respect the wishes of their users in these regards.
I'll never read an article with a title like that.
Having read the article, I can tell you that you chose well.
Is HN read-only? All the vote buttons disappeared
>There are many reasons for this. Drivers. Games. Adobe. Microsoft Office. Battery life. The thing where you close the lid of a laptop and open it again later to find that it passed into the good night.
The last one is a huge problem for Windows as well. Its due to Microsoft discontinuing support for S3 sleep mode, which in turn, caused motherboard manufacturers to discontinue S3 support in the BIOS. Which means its no longer available even if you install Linux on the laptop since it requires firmware support to work. You can still find laptops that support S3 sleep if you really look hard enough. Or buy a Mac.
My first instinct was to just not open the article based on the headline. But I thought, "what if there's a good point that I, as a Linux user, should be aware of?"
It was worse than I imagined it would be. I now deeply regret giving this article a click.
Basically, it's all about how AI can use Mac OS features.
It is always the year of the linux desktop.
It's been at least a solid decade of the Linux desktop, it's just not evenly distributed yet.
Eternal September 1991
I like this take too. It’s never ending as more and more install Linux.
Eh, the point is interesting, but I'm not sure it's not solvable. Beyond that, I'm quite hopeful at linux breaking out in a big way in the next couple of years via chromebooks. My theory is that we'll start seeing a hockey stick graph of ai-found/exploited windows zero days, and in response we'll see a dramatic acceleration adoption of chromebooks. Voila, YotLD.
It has many advantages but also serious issues. On the latest Ubuntu, I started downloading a game from the App Store during a background OS update and it locked up so bad I had to run the terminal with a hot key to salvage it. That’s not a real desktop experience.
Wrong and I’ve been saying this for almost a decade now: the Year of Linux on the Desktop is not a global event. It’s a personal event.
It's been the year of the Linux desktop for a while. Someone has been sleeping under a rock.
Lower lift to add accessibility tree as a new feature to Linux desktop environments, vs de-enshittifying MS and MacOS desktops?
...I wasn't expecting the argument to be that Linux interoperates poorly with AI Agents lol.
I think the author is actually on the right track at first then dismisses it with: These are "why a person did not switch to Linux last" and not "why the desktop, as an institution, will continue to belong to Apple and Microsoft". You can absolutely get to the root cause of the former and find foundational issues that explain the latter.