AI and linked outputs
Hoot and Autopilot read short summaries of what is linked to the work in view, never whole contents.
When you ask Hoot about a project, goal or program, it can read a short summary of the outputs linked to it. Autopilot reads its own project's linked outputs the same way and can cite them as evidence.
A summary holds only:
- a grid's worksheet names and column headers, and a few key figures;
- a form's response tallies, with the same minimum number of responses an anonymous form needs before answers show;
- a table's columns and row count;
- a note's headings;
- a deck's slide titles.
Whole contents are never sent. Only outputs the person asking can open are summarised, and the summaries fit within a fixed size.
Autopilot may propose linking an output it created to its own project. It never links anything outside that project.
Your Notebook is never read by AI unless you attach a note yourself, as before.
The Toolbox appears only where it is switched on for your workspace. Where it is off, every tool stays where it was: project tabs, /app/forms, /app/signatures, /app/wiki and /app/documents work exactly as before.