Omarchy vs AppleSoft

For decades, Windows and macOS dominated the desktop. Every year has been ‘the year of Linux’, the moment the masses would realise they could break free from their digital overlords. The endeavour has failed every time. It turns out that building one of the foundational user experiences of modern life is not easy, it demands a concentration of energy and coordination rarely found in decentralised open-source communities. Now Omarchy has arrived: an opinionated desktop operating system with an explicit mission: expand the circle of Linux beyond its kernel of software savants to a broader congregation of the technically inclined. A worthy pursuit, and an entirely plausible one, chaperoned as we now are by machines of loving grace. The question is whether this open-source OS could find its way into the mainstream, onto the machines and into the lives of the decidedly technically disinclined.

Omarchy is a portmanteau of omakase and Arch Linux. Omakase is Japanese for I’ll leave it up to you: let the chef serve what they think is worth your while. On this take-it-or-leave-it principle, David Heinemeier Hansson built Ruby on Rails, which, I think it’s fair to say, has done rather well for itself. Clearly, the man has taste and a flair for the dramatic. So too with Omarchy, an Arch-based Linux distro: it has an über-geeky synthwave aesthetic, comes with a suite of default apps installed (There’s zero bloat here: Just everything I use, as DHH put it in the manual), and offers a GUI experience based on tiling and workspaces through the window manager Hyprland. Omarchy pursues distinction, definitely not imitation.

‘Linux has been waiting for its big breakthrough for many years, but it won’t happen on its own. There were some geeks who thought that if you just made it technically sound and free, people would come on their own. No matter how user-unfriendly it was. That’s not how it works. People want something cool. Something that does something new — not a Temu version of Windows.’

That’s how DHH explains it to a Danish news site. He’s not wrong. A decade ago I ran Ubuntu for a hot minute, then threw the keyboard out of the window and bought a MacBook. Windows had been annoying at times, Ubuntu was an ordeal. The system needed constant babysitting. Few things flowed. Programs crashed. My precious suite of tools was replaced by inferior, duct-taped garage builds. These interfaces were not retro-suave; they were plain kitsch, hard to love for anybody except their creators. I possessed neither the patience nor the grace to keep the thing alive. It made me sad.

A wave forming

When I installed Omarchy, I was instantly smitten. Everything worked — or, rather, my AI agent could make everything work. Monitor flickering on the boot screen? Ask Claude to fix it. Done. No battery status in the Bluetooth menu? Fix it. Done. The friction is gone, and what remains is the unadulterated joy of making things appear in the world. Homo faber in the virtual. Building stuff is exciting, but I’m not wedded to process for its own sake. Coding is fun, puzzles are fun, doing things the Long, Hard, Stupid Way can be rewarding, but there is a peak on the landscape of creative pleasures that is seeing an idea shift the state of the universe to something slightly better. Omarchy captures precisely that spirit.

Omarchy is agentic in two senses. First, it is an agent-first environment: you’re expected to use your agents to bend it to your tastes, and it ships with an Omarchy skill. Second, the OS itself is being built by agents. After decades of chiselling code by hand in a Vim terminal, DHH has gone full agent: at the opening keynote of the most recent Rails World, he declared to his disciples that handwritten code is dead.

Something like a wave is forming. The Omacom Foundation, the non-profit behind the project, has raised over $20 million through a trickle of grassroots donations and the open faucets of Silicon Valley’s heavyweights: Patrick Collison, Jack Dorsey, Tobias Lütke and others. Meetups are spawning around the world. The community is still dominated by a group of excitable followers, more than eager to throw themselves into the deep end of the OS, but there’s a good chance the circle will widen further still.

OS for the masses

The desktop OS is one of the foundational human-computer interfaces of our times. New form factors keep arriving — most notably has been the phone and watch OS, but voice-only and headset systems are probably not far off — yet the desktop stays. It’s the carpenter’s workbench, the place of dedication, of worship. Since the dawn of the personal computer, the desktop has been dominated by the AppleSoft duopoly: today, Windows is installed on around 76% of desktops, macOS 17%, and Linux on 5% (StatCounter). There’s been much thinking-out-loud about what it would take to make Linux a contender, a somebody, give it some class. The task of creating a system for the normies goes far beyond the wrangling of code and system design and deep into the messy realm of human psychology. That’s murky territory.

Now that we’ve entered an era in which agents may multiply human output by a factor of ten, or perhaps a hundred, could there be a path for an OS like Omarchy to reach 10%, 20%, or even 30% of desktops? Switching costs are brutal. People spend much of their time with their operating system; they have predilections, muscle memory, a specific way of doing things and expectations of how things should work. Most people do not want to change their ways. They are fond of their ways. They’ll go to considerable lengths to avoid changing their ways. A switch to a new OS, therefore, has to feel less like a migration and more like moving into a nicer flat on the same street. It seems that Omarchy will also come in a kid-friendly version: no sudo, vetted package library, parental oversight, etc. Could Omarchy also be tweaked to become an OS for the tech-apathetic? Picture a retired property lawyer or a tween Labubu content creator: what would it take for them to feel at home on, say, a friendly derivative distribution of Omarchy that we’ll, for the sake of argument, call Oma-Tomo? Tomo, as in the root of the Japanese word for friend, tomodachi.

First, the basics. Installation and setup should be easy. Monitor, microphone, headset, keyboard, speakers, the IoT-enabled vacuum cleaner — all of it should connect with minimal fuss. Apps should be fast, responsive and robust. Software and hardware updates should arrive with boring regularity and not like an ambush. The OS as a whole should offer reasonable shelter from the pitiless barrage of security attacks. It must feel like a home: straightforward to navigate, easy to maintain, safe. Omarchy would remain the base OS, doing all the groundwork of maintenance, updates and security, on top of which Oma-Tomo would sit as a hand-holding OS for the less confident user.

Holding hands

Omarchy takes some getting used to. As someone who loves the system, I don’t mind, but this is a dealbreaker for most. Here I present a non-exhaustive and rather random collection of obstacles I believe would frustrate the technophobic, with rough outlines of approaches that might ease the pain.

Saturated keybindings. By default, there are so many global bindings that hitting a key combo feels risky: a few unfortunate presses and you’ve enabled scroll mode in the window manager, or zoomed in to 200% with no obvious way to zoom out, and you helplessly press and click things, while windows fly around, as you entangle yourself ever further in the trap you’ve set for yourself. This is a undignified experience. It feels disempowering. A few patterns could help here.

• Keep bindings to an absolute minimum. A handful of menu opens and tiling operations, nothing more. More keybindings could emerge over time as users grow comfortable with the ways of the OS. ‘You just opened [app] for the third time — do you want to assign a keybinding to make it easier to open in future?’
• Introduce a universal undo. Ctrl+Z is one of the most brilliant features of the digital world: where else can you, mere mortal, traverse the decision tree bidirectionally? Wherever possible, let users reverse what they just did: a single global undo that takes back the last action. Accidentally enabled zoom? Ctrl+Z. Undo. You’re now back where you came from. This becomes the fallback people learn alongside every other command, and it provides psychological safety as they punch buttons.
• Guard destructive actions. Anything that destroys data should be blocked or confirmed by default.

Window tiling. The Hyprland window manager is excellent but unusual. The basic commands are easy; the trouble is that the primary interaction runs through workspaces, not the program-first navigation around which macOS and Windows are structured. People are used to dragging windows about with a mouse. A first-run onboarding, and suggestions as you go, would turn a strange concept into an intuitive one.

Show what is running. People expect to reach running programs from some kind of menu. Without much overhead, programs could be pinned or dragged to the menu so they stay reachable from any window, giving a quick sense of everything that is open and a way back to the window you came from.

Make moving a file simple. It is surprisingly complicated: open the file manager in another window, drag files across. An intuitive gesture for this would do a great deal of good.

Above are a few cherry-picks to give an idea of the design challenges that stand in the way of broad adoption. Each is a hard problem in its own right. Solving them all would take coordination, wisdom, strategy, flair — in short, a band of expert designers collaborating at scale inside a loosely connected community. Yesterday, such a design effort would have meant far too many cooks in the kitchen, but perhaps we’re in a new era, in which taste can evolve in unison across time and space.

Benevolent design dictators and the swarm

The $20 million raised to build Omarchy is chicken feed next to the billions Apple and Microsoft pour into developing, distributing and maintaining their flagship OSs. Each year they’ll outspend Omarchy by a couple of orders of magnitude, which means, from first principles, that development of Omarchy would have to be a hundred times more efficient to compete. That would entail removing the friction that emerges in large-scale collaborative efforts. A big enterprise is full of wasteful bureaucracy. Like a space rocket whose mass is mostly the fuel it must burn to escape gravity, the accumulated transaction costs of thousands of employees have historically been the price of building mega-projects. Maybe AI is the anti-gravity sauce.

Loosely organised groups of open-source contributors have created a stupefying amount of value over the years, yet most projects never acquire the polish that makes software accessible. Maintainers of successful FOSS projects skew towards shape-rotators, who tend to take pride in their ability to navigate difficult systems and carry some disdain for the people who do not master the same abilities. VideoLAN is an awesome feat of engineering, built by hundreds of contributors under the suzerain Jean-Baptiste Kempf, but VLC’s GUI certainly isn’t a pretty thing. The most successful open-source projects evolve under a charismatic leader who can argue points of technical minutiae, and that works because code is either right or wrong: it compiles or it doesn’t, it introduces a bug or it doesn’t, it slows a process or it doesn’t. While good design does indeed emerge from some principles that hold true in all contexts, it’s also about taste, that elusive, fuzzy, shape-shifting property. And designers — those who really care about such amorphous things — have been excluded from meaningfully contributing to these projects, for want of engineering flair and because consistent design is hard to draw from distributed collaboration between dozens or hundreds of designers. The lack of a design dictator means slippage, inconsistency — in short, shitty design.

Now there may be another way. Imagine the Oma-Tomo design principles distilled into a set of maxims that an AI design dictator director — the benevolent supremo as a model — can use to gatekeep the pull requests of agent swarms belonging to thousands of design contributors. Consider the aforementioned idea of reducing the keybindings to an absolute minimum: a designer in the community proposes a concept based around increased mouse use and opens a pull request with a design spec to formalise their proposal. However, the Oma-Tomo design principles state that every action must be possible by voice or keyboard alone, so the Oma-Tomo Design Director agent closes the proposed spec. A carefully curated committee of tastemakers guards those Oma-Tomo design principles with their lives and reputations.

On these categorical imperatives, the circle of contributors to the design of Oma-Tomo is expanded by coordinating a swarm of agents belonging to hundreds, or even thousands, of contributors — orchestrating, approving, rejecting, requesting changes, and ensuring the core design maxims hold true in every corner of the system. If such a Design Director could be built, it would deliver an operating system usable by mainstream humans and their agents alike, and a whole new class of contributor, the designers, might finally join the cause. At this point, the orders-of-magnitude capital advantage of AppleSoft may not matter all that much.

Cracks in the duopoly wall

This is a unique moment in Linux history. A rizzy hypeman with both the capital and the technical and organisational skills for greatness has launched a distro that makes many, many people feel excited. It’s hard not to be infected by the pure enthusiasm of DHH — p(bloom) all the way, baby. The stated goal is to expand the circle of Linux to technically minded people, the ones too impatient to sacrifice their scarce hours to reading the fucking manual. People like me. I already love Omarchy. I think many others could too — with a few turns of the knobs.

Another distro based on Omarchy could make the OS digestible for a broader user base by reducing complexity, softening some of the edges and making a couple of apologies for Linux. Of course, building an OS for the masses would pitch it in direct competition with macOS and Windows and their gargantuan coffers, so the open-source collective would have to find ways to scale not just the technical effort but the design effort too. With the introduction of an AI Design Director and a human-curated set of design principles, it may be possible to fill the kitchen with more chefs than ever, and construct a UX as smooth as what people are used to, albeit using different design primitives. Ultimately, there is a full-AI version of all this: something akin to Samantha in the film Her — a voice-first interface that grows through a relationship with the user. Omarchy’s ability to change according to the wishes of its user points in that direction.

Maybe we could be seeing the first rays of light through the cracks in the wall of the AppleSoft duopoly, of which we have all been vassals for a lifetime.