Origins
A thirty-year project, briefly.
Guardian Angel is not a new project. It is the technical and clinical realization of a research program that has been carried forward since 1994 by Szolovits, Doyle, Long, Kohane, and Pauker, and by the generations of work that followed at MIT, Harvard, Boston Children's, and elsewhere.
The original manifesto
The Guardian Angel technical report, MIT/LCS TR-604, proposed a lifelong patient-centered health system with active assistance, longitudinal memory, and preference-aware reasoning. The thesis has held; the infrastructure has finally caught up.
Read the 1994 manifesto · Original HTML (MIT) · Original PDF (DSpace)
PING and Indivo
Architecture and reference implementations for patient-controlled records. Published in BMJ, JAMIA, and International Journal of Medical Informatics. Renamed Indivo and adopted as the technical basis of the Dossia personally controlled health record initiative.
The first deployment wave
Microsoft HealthVault, Google Health, and Dossia validated the thesis at scale but did not sustain, primarily because EHR data extraction was manual, patient-side AI did not exist, and information-blocking enforcement had not yet arrived.
The substitutable-apps shift
SMART on FHIR established a real standards-based path for patient-facing applications. The strategy shifted from process-heavy interoperability to app-level substitutability on a shared API.
The policy era
21st Century Cures Act, ONC final rules, and information-blocking enforcement established patient access as a legal right rather than an institutional courtesy.
The architecture is operational
The five forces have converged: AI, EHR dissatisfaction, patient access, shared-decision demand, and personal compute. Guardian Angel is being built.
The source document
Read the 1994 technical report.
The intellectual foundation under everything on this site. Preserved verbatim, with the modern operational argument layered on top in Architecture, Evidence, and Governance.