Think of it as Yoga for the brain, we are going to stretch a bit.
Is Omarchy the Next Giant Killer? Will It Replace Windows, macOS, and Ubuntu?
The operating system of the future won’t tell you how to use your computer. You’ll tell it what you want, and AI will build the computer around you. --YNOT!
Forty-nine years ago, I started working with computers. I was 12 years old. It was 1977.
The Apple II had arrived. The Commodore PET had arrived. Personal computing was suddenly becoming a real thing. I couldn’t afford either one.
But I had a bicycle. And a few miles away, I had a Radio Shack.
So I would ride my bicycle to the local Radio Shack, walk inside, sit down in front of one of their computers, open a BASIC programming book, and start typing.
There was no Internet. No Google. No Stack Overflow. No GitHub. No ChatGPT. No YouTube tutorial explaining what I had done wrong. You typed the program from the book or a magazine.
Then you typed: RUN
And hoped something happened.
The Radio Shack TRS-80 I was playing with could have as little as 4 kilobytes of RAM and an 8-bit Z80 processor running at roughly 1.77 MHz.
Not gigahertz. Megahertz. And not four gigabytes of memory, not four megabytes. Four kilobytes.
Today, the icon for Microsoft Word probably uses more memory than that entire computer had available.
You could save your program to cassette tape. Technically. But cassette storage was so painfully slow and unreliable that sometimes I would simply print the program out, take the paper home, and type the whole thing in again the next time I came back.
That was computing. And I loved it.
Then Something Happened
Over the next 49 years, I watched computers transform from strange little boxes understood by a handful of hobbyists into machines that would have looked like science fiction when I was a kid.
Today I have machines with close to terabytes of RAM.
I have many hundreds of terabytes of storage.
Processors with dozens of cores. GPUs capable of trillions of mathematical operations.
And increasingly, I barely use the keyboard for some of the work I do.
Because I don’t always do the work anymore. Agents do it.
That is the change I think many people still haven’t fully understood.
AI isn’t merely another application running on the computer.
It is starting to become the operator of the computer.
I Don’t Program the Way I Used To
Recently I wrote a replacement for QuickBooks in about five hours.
I built something resembling a Quicken replacement in around two.
Not prototypes scribbled on a napkin. Working software.
I couldn’t have done that traditionally in anything close to those time frames.
And it isn’t because I’ve suddenly learned to type 500 words per minute.
It is because I’ve changed the organizational structure of software development.
I describe. I design. I inspect. I test. I make decisions.
And agents do enormous amounts of the implementation.
I now have agents building software. Agents testing software.
Agents reviewing software written by other agents.
Agents debugging other agents’ work.
I’ve even reached the point where I am building agents to manage other agents.
One human can increasingly operate more like a small software company.
Sometimes I feel like I’m doing the work of ten people.
In certain situations, perhaps dozens. That doesn’t mean ten programmers suddenly have no value. It means the unit of leverage has changed.
The tractor didn’t make farming disappear. It changed how much farming one person could accomplish.
AI is doing the same thing to intellectual work.
And nowhere is that transformation happening faster than software development.
And Now AI Has Reached the Operating System
This is where Omarchy gets interesting. Because I don’t think Omarchy should be viewed simply as another Linux distribution. There are hundreds of Linux distributions.
The world doesn’t desperately need number 301.
What interests me is the philosophy behind it.
Omarchy is being developed by David Heinemeier Hansson — DHH — the creator of Ruby on Rails and co-owner of 37signals.
It is built around technologies including Arch Linux, Hyprland, and Quickshell, creating a highly configurable, keyboard-oriented Linux environment.
But the important word isn’t Arch. It isn’t Linux. It isn’t Hyprland.
The important word is: Agentic.
Omarchy explicitly treats AI coding agents as first-class participants in the operating environment. Its current documentation includes direct support for tools such as Claude Code, OpenAI Codex, OpenCode, GitHub Copilot CLI and others.
That’s a very different starting assumption.
For forty years, operating systems have basically assumed this: Human → keyboard/mouse → application → operating system
The emerging model looks more like:Human → AI agent → applications + operating system + other agents
That is a much bigger change than replacing one Start menu with another.
Windows Is Still Designed for Yesterday
Let’s be honest. Windows carries decades of history with it.
Some of that history is necessary. Microsoft has one of the hardest compatibility problems in computing. Millions of businesses depend on software written years or even decades ago.
That compatibility has enormous value.
But it also means Windows increasingly feels like layer after layer of old architecture, new architecture, legacy interfaces, cloud services, accounts, subscriptions, telemetry and Microsoft’s latest idea of how you should use your own computer.
And something else has changed. The relationship between the user and the machine.
When I bought a computer decades ago, I felt like I owned the computer.
Increasingly, modern computing can feel like I purchased hardware that gives me permission to participate in somebody else’s ecosystem.
Microsoft decides things. Microsoft changes things. Microsoft adds things.
Microsoft removes things. You adapt.
AI may expose how outdated that relationship has become.
Apple Has a Different Problem
I like Apple. Apple makes extraordinary hardware.
macOS is polished. The integration between Apple’s hardware and software is one of the company’s greatest advantages.
But Apple also has a philosophy. And Apple’s philosophy is essentially: We have thought very carefully about how this should work. Please don’t mess with it too much.
For most consumers, that is actually a feature.
For someone like me, it eventually becomes a limitation.
I want to change things. I want to connect things. I want one agent talking to another agent talking to another server.
I want to rip something apart at 2:00 in the morning because I suddenly thought of a better way to do it.
Apple gives you an exquisitely designed house.
But Apple would prefer that you don’t start moving the plumbing.
Omarchy is closer to handing you the architectural drawings, the toolbox and a robot contractor and saying:
What would you like changed?
Interestingly, Omarchy is now even working toward easier support for Apple Silicon Macs through its recently announced Omarchy M effort.
That tells you something about where this may be headed.
Keep Apple’s excellent hardware. Replace the assumptions about what the operating system should be.
And Then There Is Ubuntu
Ubuntu has been my world for years.
I like Ubuntu. I use it heavily. I use it for servers. I use it on my main workstation.
It is dependable. It is documented. There is an enormous ecosystem around it.
I am not uninstalling Ubuntu from everything tomorrow.
But Ubuntu still largely represents the previous generation of Linux computing.
It is an excellent operating system onto which we are adding AI.
Omarchy is experimenting with something slightly different:
What does the computer look like if AI agents are assumed from the beginning?
That difference may sound subtle. It isn’t.
The smartphone wasn’t simply a smaller computer.
The web browser wasn’t simply another application.
And all the wonderful ways Linux can punish you for forgetting one character.
AI agents collapse much of that knowledge barrier.
You can increasingly say:
I want this window to always open on monitor two. Find what controls that, show me what you’re changing, back up the original configuration, make the change, test it, and undo it if it doesn’t work.
That is profound. The audience for a software modification no longer needs to be one million people.
It can be one person. You.
That principle was also central to the Omarchy discussion I was reading: the combination of configurable open software and agents makes highly individualized computing increasingly practical.
The Computer Becomes Something You Can Talk Into Existence
Think about where this leads.
Today you download an application because someone predicted that enough people would want a particular feature to justify building a company around it.
Tomorrow you may simply describe what you need.
“I want a little panel showing my servers.”
Build it.
“I want this folder automatically analyzed and categorized.”
Build it.
“I want a dashboard showing which AI agents are working, which are waiting for me and which have failed.”
Build it.
“I want this machine dedicated to supervising my other machines.”
Build it.
“I want a new interface around an existing command-line program.”
Build it.
That last example is already happening around Omarchy: agents can often build a personalized interface around existing mature tools instead of recreating the difficult underlying functionality from scratch.
That might become one of the dominant software-development patterns of the next decade.
Don’t rebuild everything.
Let mature infrastructure perform the underlying work.
Then let AI generate the personalized layer around it.
We May Be Watching the Death of the Universal Interface
For decades software companies have spent enormous amounts of money designing interfaces that hopefully work reasonably well for millions of people.
But what happens when the interface only has to work for me?
My accounting system doesn’t necessarily need the same interface as yours.
My development environment doesn’t need to resemble yours.
My operating system doesn’t need the same workflow as yours.
AI makes mass personalization economically possible.
That’s something we’ve never really had before.
And operating systems may ultimately become the most personalized software of all.
If You Don’t See What Is Happening, Look Up
AI has already hit software development harder than almost any other industry.
And I think we are still near the beginning.
If you’re a programmer and you believe software development five or ten years from now will look essentially like software development five or ten years ago, I think you’re a dinosaur looking up at the asteroid and wondering why the sky suddenly got bright.
The keyboard isn’t disappearing tomorrow. Neither is the programmer.
Neither is Windows. But the role of each is changing.
Microsoft knows it. Apple knows it. Google knows it. OpenAI knows it. NVIDIA certainly knows it.
Microsoft is desperately trying to establish its position in the AI computing stack.
And I would not be surprised to see companies such as OpenAI and NVIDIA move increasingly toward complete AI-centric computing systems.
If that happens, ask yourself something:
Why should an AI workstation primarily be designed around a keyboard and mouse?
Those are interfaces designed for humans controlling applications manually.
But what happens when the primary interface becomes intent?
You explain what you want. Agents determine how to accomplish it.
The screen becomes less of a cockpit filled with controls and more of a place where you supervise what is happening.
That Is Why I’m Testing Omarchy
I’ve only been testing Omarchy for a few days.
Is it ready to replace every workstation I have?
No. Not yet.
When something absolutely must work, reliability still matters more than ideology.
Even enthusiastic early discussions around Omarchy acknowledge the tradeoff: a deeply modifiable Linux environment can give you capabilities unavailable elsewhere, while occasionally requiring you to repair things Windows and Mac users simply expect to work.
That matters. An operating system isn’t useful because it is clever.
It is useful because tomorrow morning you can turn the computer on and get your work done.
But I don’t think that diminishes what Omarchy represents.
In fact, it makes the experiment more interesting.
Because I already know what role I want it to play.
My bot machine. My agent workstation. My AI interface into everything else.
A machine where AI isn’t another browser tab.
It is part of the environment.
And There’s Another Reason I Care About It
Ownership. That word increasingly matters to me.
I want machines I own. I want data I control. I want software I can inspect.
I want local AI when local AI makes sense. I want cloud AI when cloud AI makes sense.
I want to decide.
Running software locally doesn’t automatically mean the AI processing is local — an agent can still send requests to cloud models — so those are separate architectural and privacy decisions.
But the important thing is having the ability to make those decisions.
Omarchy is open. It is modifiable. It is increasingly designed around agents.
And once I install it on hardware I own, that computer is mine to reshape.
That is important to me.
Very important.
Could Omarchy Actually Replace Windows, macOS and Ubuntu?
Tomorrow? No.
Windows has an enormous software ecosystem.
Apple has spectacular hardware integration and an enormous consumer ecosystem.
Ubuntu and the broader Linux ecosystem have years of infrastructure, support and institutional knowledge behind them.
Those things don’t disappear because somebody made an interesting desktop.
But technological revolutions rarely begin by immediately replacing the incumbent.
They begin by exposing that the incumbent was optimized for assumptions that are no longer true.
BlackBerry didn’t disappear because the iPhone had a better keyboard.
The iPhone questioned whether a phone needed a physical keyboard at all.
Kodak wasn’t defeated because somebody manufactured slightly better film.
Digital photography questioned whether you needed film.
AI may force us to ask the same kind of question about operating systems.
Maybe the question isn’t: “Is Omarchy better than Windows?”
Maybe the question is: “Why are we still using operating systems designed around humans manually operating every application?”
That’s a much more interesting question.
The Giant Killer May Not Be Omarchy
And this is the part I find most interesting.
Omarchy itself may not ultimately conquer the desktop.
Something else might.
Microsoft could transform Windows.
Apple could fundamentally rethink macOS.
Ubuntu could become much more agent-native.
A completely new operating system might appear.
The winner may not even exist yet.
But Omarchy may still prove historically important because it is exploring the right question early.
What should an operating system look like when AI agents become first-class users of the computer?
That is the experiment.
DHH’s own project now describes Omarchy as a “malleable OS for the age of agents,” and its development process is itself heavily agent-driven; the project says Omarchy Quattro was built almost entirely with agent assistance.
That is why I’m paying attention.
Forty-Nine Years Later
In 1977, I rode my bicycle to Radio Shack so I could type BASIC programs into a computer with 4K of memory.
I would stare at that machine and imagine what computers might someday become.
I couldn’t have imagined this. A machine with nearly a terabyte of memory.
Hundreds of terabytes of data. Artificial intelligence writing software.
Agents controlling computers. Agents building agents.
And a human increasingly moving from the person doing every task to the person describing, directing and judging the work.
Science fiction didn’t arrive with flying cars. It arrived quietly through a terminal window.
Forty-nine years later, I feel a little bit like that 12-year-old kid sitting inside Radio Shack again.
I’m looking at something primitive. Something unfinished.
Something that occasionally doesn’t work.
But something that hints at an entirely different way of using a computer.
Will Omarchy replace Windows?
Will it replace macOS?
Will it replace Ubuntu?
I don’t know.
But after almost half a century of watching computing evolve, I’ve learned one thing:
Pay very close attention when a new technology doesn’t merely improve the old way of doing something — but questions why we’re doing it that way at all.
Omarchy is asking that question.
And that may be far more important than whether it wins.
