AI Ends the Fixed-Function App — and the Author Is Building a Phone for It

A longtime security engineer argues that the real disruption from AI isn’t dissolved boundaries between frontend and backend or web and native — it’s the collapse of the line between programmers and users. He describes filling his own Mac with dozens of throwaway apps he built for himself using plain English, and predicts that any power user will soon do the same. The consequence is a world of software with an audience of one or two: hyper-specific tools for trivial, personal problems that no professional would ever have shipped because the market was too small.

He pushes this to a provocative second-order conclusion about operating systems. The core job of a modern OS, he contends, is to isolate mutually-untrusted applications imported from strangers and mediate how they talk to each other. That model makes little sense when most of your software shares a single provenance — you — and even less when every app is malleable and can grow new features on demand. If software stops arriving as prefabricated fixed-function products, the sandboxing rationale that defines today’s OS design starts to erode.

The piece doubles as a startup announcement: leaving Fly.io, the author is building a phone designed around on-device, AI-generated apps rather than app-store binaries planned years in advance. He’s candid that this reads like every other grandiose “future of programming” pitch, and holds back the technical specifics to avoid the launch-hype taint. The through-line is that hardware and OS conventions still trace to 1970s assumptions about a priesthood of programmers shipping apps to passive users — assumptions he bets AI is about to make obsolete.