The Coe Lab
← Back to Blog

What Even Is an OS Now? When AI Dissolves the Boundary Between Users and Programmers

September 26, 20266 min read
AIoperating systemsfuture of computingsoftware developmenttechnology

AI is quietly dissolving the wall between users and programmers. When anyone can conjure apps in plain English, the operating system's core purpose — partitioning applications from strangers — starts to look obsolete.

There's a question that's been nagging at the edges of the tech world for over a year now, and nobody quite knows how to say it out loud: what even is an operating system anymore?

It sounds like a philosophy 101 prompt, but it's actually the most practical question in computing right now. And a recent essay by Thomas Ptacek (of Fly.io fame, now departing for a new venture) crystallized the issue better than anything I've read this year. His argument is simple and devastating: AI is not just changing what we do with computers — it's changing what computers are.

The Death of Prefabricated Software

For the entire history of personal computing, software has been something made by experts and consumed by everyone else. You bought it off shelves at Egghead Software. Then you downloaded it from the internet. Then it lived in the cloud. But the fundamental relationship never changed: a priesthood of professional programmers built fixed-function applications, and the rest of us used what they shipped.

That model is dissolving. Ptacek describes it with a clarity that's hard to argue with: his Mac menu bar is now cluttered with apps he conjured for himself, using English as his programming language. Not toy apps — real, functional tools that solve his specific problems. And what he did on his Mac, any power user can now do on theirs.

This isn't a prediction about some future capability. It's happening right now, today, in 2026. The boundary between programmer and user is being erased in real time.

Why the OS Exists — And Why That Reason Is Fading

Here's where it gets interesting. Why do modern operating systems exist? Not simply to manage hardware resources — plenty of embedded systems do that with runtimes that barely qualify as an OS. The core purpose of a modern operating system is to partition different applications off from each other and carefully control how they communicate.

That architecture makes perfect sense in a world where you're importing all your software from strangers. You need sandboxing, permissions, process isolation, and security boundaries because every app on your machine comes from a different developer with different incentives and different trust levels.

But what happens when most of the software on your device has the same provenance? When you conjured it yourself, or someone you're on a first-name basis with did? When every application is malleable, subject to growing new limbs at any moment on the whims of its creator?

The entire partitioning model starts to look like overkill. Not wrong, exactly — but designed for a world that's fading.

The World of Apps for One

Ptacek asks us to consider what computing looks like when most applications have an audience of just one or two people. Not mass-market software — bespoke software, conjured on demand to solve idiosyncratic problems:

  • Getting an accurate weather forecast for your specific neighborhood when the nearest weather stations miss the local microclimate
  • Knowing whether to take the highway or the side streets to get home right now, based on real-time conditions
  • Checking if there's a meeting happening right this minute that you're supposed to be in

These sound trivial. That's the point. They're normal-life problems nobody would have built serious applications around, because the audience was too small to justify the effort. But when the cost of building an app approaches zero, the economics flip entirely.

Today, almost all software comes from expert strangers. Tomorrow, strangers will ship just the building blocks — and users will assemble them into solutions for problems no one else has.

The Phone That Doesn't Run Apps

Ptacek's most provocative claim is about hardware. Every mainstream phone on the market in 2026, he points out, was planned in 2023. In 2023, computers were still devices designed to run deterministic fixed-function applications produced by professional programmers. The soul of those phones is stuck in the 1970s — the same PDP-11 model of prefab applications, just beautifully miniaturized to fit in your pocket.

His new venture is building a phone that isn't designed to run fixed-function applications at all. It won't be for AAA mobile games or TikTok. Instead, it builds apps for you, on the device, whenever you want. You ask it for whatever you need, and it conjures the tool.

It sounds absurd. It also sounds inevitable.

The Second-Order Effects Nobody Is Talking About

We're all so bumfuzzled by the first-order impacts of AI — it writes code, it generates images, it answers questions — that we haven't considered the second-order effects. If most software becomes bespoke and ephemeral, what happens to:

  • Software distribution? App stores become building-block stores, not application stores.
  • Security models? If you wrote the app yourself, who are you sandboxing it from?
  • Developer economics? The market shifts from selling finished products to selling capabilities and components.
  • Platform power? The company that controls the best on-device AI model controls everything — not through a walled garden, but through the conjuring engine itself.

These aren't hypothetical concerns. They're the architectural decisions that will define the next decade of computing. And right now, most of the industry is still optimizing for the old world — making better prefab apps, faster app stores, more polished fixed-function tools.

The Unreasonable Future

Ptacek ends with a line worth repeating: if you think all of this sounds unreasonable, that's fine. But we're betting that AI is about to do unreasonable things.

He's right. Every major shift in computing looked unreasonable right up until it was obvious. The graphical user interface. The web browser. The smartphone. Each one dissolved a boundary that everyone assumed was permanent. The boundary between users and programmers is the biggest one left, and it's the one AI is uniquely positioned to erase.

So what even is an OS now? It's a question the entire industry is going to spend the next five years answering. And the answer is going to look nothing like what we have today.

The operating system of the future isn't a platform for running other people's software. It's a conjuring engine for your own.

Related Posts

Strata: When a 125B AI Model Ran on a Gaming PC at 100 Tokens Per Second

A new open-source tool called Strata lets you run Qwen 3.8 Flash Next — a 125-billion-parameter model — on an ordinary gaming PC with an RTX 4090. Nothing leaves your machine, and it's faster than you can read.

Oct 5, 2026• 6 min

When AI Agents Spend Your Money While You Sleep: Why Hard Budget Caps Are Becoming Non-Negotiable

AWS and Google Cloud finally launched hard spending limits in the same month. It's not a coincidence — it's a response to AI agents that can rack up thousands of dollars before you wake up.

Oct 4, 2026• 6 min

When Utah Banned VPNs: How a Court Stopped a Law That Demanded the Technically Impossible

A federal judge just blocked Utah's unprecedented anti-VPN law, ruling that lawmakers cannot mandate perfect geolocation — a technical impossibility. The case reveals a deeper problem: when legislation outruns engineering.

Oct 3, 2026• 7 min