All projects
01 / 08Active evolution

LOCAL AI OPERATING SYSTEM

CIEL + NOBI

Two generations of an AI assistant idea: NOBI explores broad autonomy, while CIEL rebuilds that ambition around narrow execution layers, safety gates, and observable results.

01 / PURPOSE

What this project is trying to make possible.

Explore how an assistant can participate in a user's digital environment while keeping local actions understandable, bounded, and recoverable.

02 / HOW IT WORKS

A plain-language view of the system.

  1. A command enters a strict classifier and execution pipeline.
  2. The executor applies safety checks, permissions, and timeouts.
  3. A narrow agent delegates the request to one atomic skill.
  4. The system returns a structured result that can be inspected and tested.
03 / CURRENT SCOPE

What the project demonstrates today.

Voice and vision exploration
Browser and file actions
Explicit safety gates
Modular agent and skill layers
04 / FUTURE DIRECTION

Where the next useful versions could go.

  • Planner and task decomposition
  • Permissioned voice and vision skills
  • Visual execution console
  • Memory and recovery phases
NEXT CASE STUDYBodhi