Long-term exploration. No model service or funded project is implied.
M05Horizon / Mission
Personal AI with a human at the controls
Investigate tools with visible permissions, inspectable work, and a real stop button.
The outcome
What we’re working toward.
A concrete proposal for a personal AI tool that preserves human control.
Scope
- Investigate local and hosted model tradeoffs.
- Make cost, data access, and human approvals explicit.
- Keep original inputs and recoverable outputs.
The finish line
Done means demonstrated.
- Useful problem and risk boundaries defined.
- Privacy and operating-cost model reviewed.
- Prototype proves interruption, review, and recovery.
These are completion criteria, not claims that every item is already complete.
Workshop log