A Personal Journey - Building My Own AI OS

#linux #opensource #ai #programming

Disclaimer: This is more of a read on a personal journey then a tech piece.
-------------------------------------------------------------------

Hey devs! On today’s episode of mad science done in a basement, I’m diving into compiling my own Linux kernel and starting the long road toward building an AI-powered operating system.

Why Even Bother?

Earlier this year, life threw me a curveball — a house fire forced me out of my home. While crashing in a freezing garage (0 degrees celsius on average), I picked up a project to keep myself sane: hacking together a custom OS based on TinyCore Linux.

The goal? A portable USB-bootable OS that detects a machine’s hardware and picks the right small-scale AI model to run (e.g. Tiny-LLaMA, Nous-Hermes). CPU-only AI isn’t practical for coding, but it can handle lightweight tasks like scheduling, file organisation, or small automations.

That project maxed out, but it planted the seed: what if I went further?

The Big Idea: An AI-Native OS

Instead of bolting AI on top (like Copilot), I want an OS built from the ground up with AI integration in mind. Not just a gimmick — but something scalable to the machine’s capabilities, whether it’s a potato laptop or a GPU rig.

TinyCore was a good first experiment, but its limitations sent me toward something more ambitious: Linux From Scratch (LFS).

Where I’m At Now

Right now, I’m deep-diving into LFS and kernel compilation. I’ve got just enough info crammed into my three brain cells to be dangerous.

I’ll be documenting the journey here — partly to track my progress, partly in hopes someone with real kernel experience drops wisdom in the comments.

Open Questions I’ve Got:

What pain points should I expect early when compiling my first kernel?
In my TinyCore project, I had to swap to CMake a lot. Should I expect that with LFS despite Make being the standard?

Looking Ahead

This project might span years (unless burnout wins). University for CompSci is starting soon, so this will weave around that. But the vision is clear:

A proper, full-scale OS with AI baked in natively instead of duct-taped on after the fact.

I’ll share progress updates as I go. If you’ve been down the LFS/kernel rabbit hole before — drop a comment. I’d love to learn from your pain.

// Ghotet