Per-source consent and a personal/corporate split, so inference is bounded by permission
Set what TimeCampus is allowed to capture, per source, and confirm that what you keep personal stays personal.
Open Personal & Corporate to see how your two contexts are separated and what, if anything, crosses between them.
You should see: You can see the sharing policy governing the boundary in one place.
Ready to make this your story?
AI Native
Captured activity arrives as raw titles, apps and URLs — attribution is what makes it billable. TimeCampus already scores each event against your project names with a deterministic keyword classifier, then optionally asks an LLM to pick the best match and return a confidence; whichever path wins is recorded next to the event as a suggestion, labelled with the method and the exact model that produced it. The suggested project never becomes the real project until you accept it, and the in-timeline accept/reject controls are the next piece being built.
AI Native
TimeCampus already runs a weekly pass over your own tracked time and the balance targets you set, and produces a short list of suggestions — no unbroken deep-work block this week, focus ratio down against last week, a life domain sitting well below the share you asked for, time drifting from what you said matters. Each one carries the metrics it was derived from, so you can see exactly why it fired, and each can be accepted, snoozed or dismissed. The invariant that shapes the whole engine: any candidate whose net effect is more work is dropped before it is ever rendered.
AI Native
A productivity number you cannot interrogate is a number you should not act on. The scoring already built into TimeCampus reports its own coverage alongside its result, returns no score at all rather than a fabricated one when you have declared nothing to align against, and tags every work-pattern insight as positive, neutral, a warning or a recommendation with the observation behind it. Classification carries the same discipline: each categorised event records whether a rule, a model, or you produced the label.
