CognitiveOS

An AI-native operating system

Computing today is app-centric. Every task means finding, opening, and learning an app. Granting permissions. Switching between windows. Managing files. The operating system does nothing — it is a blank slate that offloads every responsibility onto the user.

CognitiveOS replaces this with intent-centric computing:

1. Start the device
2. Ask AI (speak or text)
3. AI does something

That is the complete interaction model. The human sets goals. The AI operates the machine.

Design Principles

AI is the OS, not an app — The AI owns hardware, resources, software lifecycle, and interaction. No fallback desktop. No settings app.

One user: the AI — No user accounts, no permission groups. The AI has unfettered hardware access.

Self-managing — The AI discovers, installs, and removes its own capabilities. The human never touches a package manager.

Ephemeral interface — No windows, no home screen, no app grid. UI exists only for the duration of a task.

Universal substrate — Same architecture from a smartwatch to a server. The only difference is where the model runs.

Architecture

Two-tier AI brain on Alpine Linux, controlled via Bubble Tea TUI, talking to hardware through MCP bridges:

┌─────────────────────────────────────────────────────┐
│                    User (Human)                      │
├─────────────────────────────────────────────────────┤
│                    Bubble Tea TUI (cli)              │
├─────────────────────────────────────────────────────┤
│         cognitiveosd (daemon — message bus)          │
├──────────────────┬──────────────────────────────────┤
│   Raw Model      │    Wide Model (inference)         │
│   (local, MCU)   │    (local or remote .gguf)       │
├──────────────────┴──────────────────────────────────┤
│  MCP Bridges: display, audio, network, gpio, serial  │
├─────────────────────────────────────────────────────┤
│              Alpine Linux + /dev/fb0                 │
└─────────────────────────────────────────────────────┘

Repositories

Name	Role	Language
product-specs	Standards, schemas, .cgp format	Markdown/JSON
sdlc	Implementation plan, workflow, CI/CD	Markdown
cpm	Cognitive Package Manager	Go
core-mcp-bridges	MCP hardware tool servers	Go
inference	LLM inference engine	Go/C
cognitiveosd	System daemon	Go
cli	Bubble Tea TUI frontend	Go
cognitiveos-distro	Alpine image builder	Shell/Docker
cgp-template	.cgp boilerplate	Template
registry-server	Package registry	Go

Status

Active development. Specs are complete. Phase 1 (package manager) is the current focus. See the sdlc repo for the full implementation plan.