Every mode runs the same core — the ledger, the optimisation map, the optimisation path, and the agent. What changes is who holds keys and how deep the data runs. Same map, sparser constellation; modes are configuration, not products.
The clinic runs vasusOS as its health-optimisation copilot — sessions, intake and retest, protocol boards, review packs. Members get an optional lite login: their path and results, read-mostly. No tracking between visits unless the member connects something — the ledger deepens only when they turn a key, and the agent's retention move is to suggest what they could add to track their journey with this clinic.
A solo journey. Manual entry is first-class — typed numbers carry self_reported provenance beside device data. Working with a clinic that isn't on vasusOS? Hand them a generated, cited report — or grant a discreet lite view. Everything shared is scoped and revocable.
The full duo, and the destination every mode ramps toward: verified sessions × lived weeks × research, two keys on every change. Clinic-led members connect a source and arrive; member-led reports carry an invitation for the clinic to claim their view.
| data rails | Apple Health · Google Health / Health Connect (read, and sessions written back) · Oura, Eight Sleep, WHOOP, Garmin via an aggregation rail · Strava, own activities only |
| systems of record | All main booking and CRM systems: Pabau, Phorest, Zenoti, Cliniko, Fresha and others. Completed appointments in; only the shared review summary out. |
| agent layers | ChatGPT Health first, others as partner paths open. Any agent calls the same governed tools; answers arrive with their tool receipts. Writes are drafts until a human approves. |
Live: the vasus.ai evidence platform — 600,000+ papers, ranked, cited, zero personal data.
Design-partner build: the ledger, the console, the member app, the booking-system connector suite, the expectation model.
Later: treatment device feeds, the cohort layer at n ≥ 50, the assistant beyond scripted admin.