Overview
How might we design autonomous mode display so people can trust and act on AI output?
When to use
- Essential for AI coding agents, always-on assistants, and automated workflows where unattended execution needs a glanceable, always-visible signal so users never discover autonomy by surprise.
When to skip
- Fully manual copilots that never run without a keystroke.
- Backend jobs with no user-facing surface (use ops dashboards instead).
- Momentary generations shorter than a status animation would help.
Rules
Autonomous runs with no visible indicator.
Indicators that stay on after the agent stopped.
No path from the indicator to pause, stop, or logs.
Decorating marketing screenshots with fake “live” badges.
Evidence
| Product | Implementation |
|---|---|
| Cursor | Agent badges and status while background runs continue. |
| Tesla Autopilot | Persistent driving-assistance state in the cabin UI. |
| macOS | Menu bar microphone and screen-sharing indicators. |
| iOS | Orange/green status dots for mic and camera use. |