Inspiration

Modern operating systems still assume humans should micromanage every click, app, and workflow. We asked a simple question: what if the OS understood intent instead of commands?

What it does

IDOS is an intent-driven operating system where users describe what they want in natural language, and AI agents reason, plan, and execute actions across the system autonomously. Fewer steps, more outcomes.

How I built it

I designed an agentic architecture where a central reasoning model interprets user intent, breaks it into actionable steps, and invokes system-level tools. The system streams decisions in real time and is powered by Gemini.

Challenges I ran into

  • Translating vague human intent into safe, deterministic actions
  • Managing agent decision loops without infinite execution
  • Handling API limits and securing keys for a public demo

Accomplishments that I am proud of

  • A working agentic OS prototype, not just a concept
  • Real-time intent → decision → action flow
  • Clean modular architecture ready to scale

What we learned

  • Intent clarity matters more than raw intelligence
  • Agent systems need strong guardrails to stay predictable
  • Streaming responses dramatically improve perceived intelligence

What's next for IDOS

  • Multi-agent collaboration
  • Persistent memory and user preferences
  • Deeper OS integrations and plugin ecosystem

Built With

Share this project:

Updates