I am trying to build a local, private, AI-first operating system and want to get y'all's thoughts.
Hi! Long time lurker here. As the title says, I am trying to build a local, private, AI-first operating system and want to see if there is any interest.
Running local models right now is great, but the AI is basically trapped inside a chat window or a terminal. It doesn’t know what files you are working on, it can’t organize your folders, and it can’t actually automate your computer. It’s just an isolated app. We have Claude Cowork, but you are bound by Anthropic/Claude.
I want to take the AI out of the chat box and bake it directly into the desktop environment.
I’m calling the project Theia, built on Linux for ARM chips. I would say there are three main benefits:
- OS-level access: The model can actually interact with your system, manage your files, and automate daily tasks natively.
- Total privacy: Microsoft Copilot has system access, but it tracks you. Theia runs 100% offline. No cloud, no telemetry, no forced updates.
- None of the Windows bloat: No more asking you to make an account just to sign in to Windows. No more wasting RAM and all the bloat associated with Windows. No more ads or Copilot being shoved into your face.
Before I lock myself in a room for a year to code the core of this, I want to see if this is something the local LLM community actually cares about. Do you guys prefer keeping your models separated as apps, or do you want an OS built entirely around them?
I put up a quick page to measure real interest. The page is https://theia-os.pages.dev/
Would also love to get your guys' thoughts in the comments section below!