Human intent
Bring fragmented measurements into one understandable history without allowing derived or AI-generated interpretation to overwrite the underlying record.
← All projects08 / HEALTH / PROVENANCE / 2026
KehaYour body, over time.
Working personal-data product and architecture experiment
A provenance-first personal health record that keeps raw observations, deterministic derived metrics and interpretation as visibly different layers.
System / agent action
Normalize measurements from Apple Health, wearables and records into observations, derive metrics separately and expose interpretation with traceable source relationships.
Hard boundary
Raw observations remain owner-only and are not mixed with derived metrics or AI interpretation. A federation experiment further separates source observations, deterministic derivation and interpretation into independently owned subgraphs.
INTENT PRESERVATION / CORE
Raw observations remain owner-only and are not mixed with derived metrics or AI interpretation. A federation experiment further separates source observations, deterministic derivation and interpretation into independently owned subgraphs.
Evidence in the system
- Apple Health remains an iPhone-only source; the web app does not pretend to read HealthKit directly.
- The production observation table uses owner-only row-level security.
- A federated demo can trace a derived claim back to each source observation used as evidence.