Redefining the Operating System

Date: September 25, 2026

It might seem absurd to say—given that it has been the sole topic of conversation for over a year—but the true magnitude of AI's impact on the world of computing hasn't actually hit us yet.

---

A New Chapter

Before I dive into the theory, a personal update: I am leaving Fly.io to collaborate with Kurt on a fresh venture. I'm mentioning this immediately so you are aware that I am, quite literally, talking my book.

The Illusion of the "Easy" Computer

My introduction to computing was a family friend who sold us our first "computer" (I use quotes because it was likely a VTech Laser 200 clone). It featured a Z80 processor, a chiclet keyboard, ran via cassette tapes, and hooked directly into the television.

At the time, I understood the concept of programming. I spent my days dreaming of the simplistic games I could create, imagining a device that would let me build something akin to the Hitchhiker’s Guide text adventure I played at my father's office.

However, the reality was a cold shower:

Computers were not nearly as magical as my childhood imagination had suggested.

By my mid-teens, an interest in computer security finally pushed me into actual coding. But the more I grasped how software was structured, the more intimidating the prospect of building a full-scale application became. I could write a high-performance network server without blinking, but a word processor?

That was a simple task. That was like being asked to construct a diesel locomotive from scratch.

The Dissolving Boundary

Recently, I've written about my experience creating native macOS apps. It has been surreal to witness AI erase the traditional lines between:

• Backend $\leftrightarrow$ Frontend
• Web $\leftrightarrow$ Native
• Systems $\leftrightarrow$ Applications

While these are massive shifts, AI is dismantling an even more fundamental wall: the divide between the programmer and the user.

Currently, my task switcher and menu bar are filled with tools I "conjured" for myself. My primary programming language is now simply English.

Note: This isn't just a hopeful prediction; it is a statement of current reality. To call it a "prediction" is only to concede to the skeptics in the same way I "predict" the sun will rise in the east tomorrow.

Second-Order Effects: The Death of the "App"

We are so distracted by the immediate (first-order) effects of AI that we are missing the second-order consequences. We must ask: What happens to computing when the target audience for an application is only $n = 1$ or $2$ people?

The Evolution of Software Distribution

I believe we are witnessing the end of the "fixed-function application." Instead of downloading a tool built by a stranger, users will interact with software they summoned into existence to solve a hyper-specific problem.

Examples of "Too Small" Problems:

• "Give me a precise weather forecast for Roscoe Village in Chicago that accounts for lake effect, ignoring the O'Hare station."
• "Don't show me a map; just tell me if I should exit the Eisenhower Expressway at Roosevelt, Madison, or Lake."
• "Scan my calendar and emails: is there a meeting happening right now that I've forgotten?"

These are "normal-life" frictions. In the past, no developer would build a professional app for these because the market was too small. But soon, developers won't ship apps; they will ship the building blocks.

We will still have "megaprojects" like browsers, but they will be outnumbered by millions of tiny apps that use, for example, $\frac{1}{7}$ of a word processor's logic to solve a specific domestic chore for someone who doesn't know what a for-loop is.

What is an OS, anyway?

If you follow this logic, the very definition of an Operating System changes.

Traditionally, an OS isn't just about managing hardware resources (embedded systems do that without being "OSs"). The core purpose of a modern OS is to partition applications and strictly regulate their communication.

$\text{Traditional OS} = \text{Isolation} + \text{Controlled Communication}$

This architecture is vital when you are running software written by strangers. It is completely unnecessary when:

1. The software has the same provenance (you created it).
2. The software is malleable and evolves instantly based on your whims.

The New Hardware

This is exactly what we are building: a platform designed for this new paradigm.

Every smartphone available in 2026 was designed back in 2023. In 2023, we were still building devices for "deterministic, fixed-function applications" created by a "priesthood" of programmers.

I imagine those designers—mustachioed men hacking on VT52 terminals with "More Than A Feeling" blasting on an FM radio—simply scaling and shrinking that 1970s design to fit in your pocket. Today's iPhone is essentially a beautifully machined DEC PDP-11.

Our phone is different:

• It is not designed for fixed-function apps.
• It is not for AAA mobile games (which nobody actually plays).
• It is not a TikTok machine (though it can do that if asked).

Instead, you ask it for what you need, and it builds the app for you, on the device, in real-time.

I know how this sounds. I know I sound like every other startup founder announcing a "revolution" before revealing they've joined some niche clone of Uber. I'm sharing the technical philosophy first to strip away the "startup announcement" scent.

The last year in tech...