Off your baseline?

• Two established questionnaires, the Primals Inventory (PI-18) and the Mini-IPIP, each scored separately on its own range. No combined score, norms or cut-offs.
• An experimental daily check-in: three questions on 0–10 sliders, each kept as its own series and never combined.
• The patient sees no scores. The therapist sees answers as given, with dates, scoring provenance and missing data, in a summary, a timeline, charts and session preparation.
• Rule-based prompts are labelled unvalidated and link to the exact answers that triggered them.

• Turns daily inputs (a short check-in, daily health summaries, calendar counts) into one state per day.
• Learns a rolling 28-day baseline per feature and flags a shift of two or more standard deviations that lasts seven days, only after 14 days of history.
• The model decides whether something changed. A template, or later a language model, only explains it, and the person marks the explanation accurate or incorrect.
• Runs end to end on synthetic histories. Real text encoding, persistent storage, encryption and deployment are not built.

These settings are working assumptions of the prototype, not validated values.