ebb

What’s real in this build, and what we assumed

This is an evaluation build: a working front end with every journey clickable, running entirely in your browser. Nothing is sent anywhere. Here is exactly where the edges are.

Real in this build

  • Every screen, control and state on phone, web, Mac, clinic and admin.
  • One shared record across all of them. Open two tabs and they update each other live.
  • Insights computed from the patient’s logs, with n-counts and evidence thresholds.
  • The companion’s guardrails and grounding: it only cites published articles.
  • Safety detection for crisis language and red-flag symptoms, on logs, messages and questions.
  • RTM data-day and minute counting against the 2026 code rules.
  • Print-ready report, CSV and JSON exports, offline queue, audit trail.

Scripted or simulated

  • Denise’s 90 days of logs are generated from a fixed seed, counted back from today, with planted sleep, activity and weather effects.
  • Voice logging types a scripted transcript; production uses on-device speech recognition.
  • The AI is deterministic lexicons and statistics; production uses an LLM with retrieval over reviewed content and the same guardrail contract.
  • Weather and Apple Health data are simulated. Other panel patients have weekly summaries only.
  • Sign-in, Epic hand-off, e-prescribing and OARRS checks raise a message instead of calling out.

Assumptions we made

What we need from you

How this becomes the product