Longitudinal health infrastructure for people, agents, and care.
QED is building Health Core: open-core infrastructure for a user-owned health record and the modular AI agents that reason over it — with provenance, uncertainty, and consent built in.
In plain terms: one health record you own, that AI can reason over safely — instead of a dozen apps that each hold a fragment.
Sources → Profile → Surfaces
Live modelThe product surface can change. The profile, permissions, and provenance model are the durable layer.
Health software is still built around fragments.
Wearables collect signals. Clinics capture episodes. Apps optimise one domain. No personal health record holds a durable, patient-owned model of you across time.
The Longitudinal Health Profile is the durable object.
The Longitudinal Health Profile (LHP) is one living record of your whole health over time. Health Core turns raw events into a profile that can be inspected, corrected, permissioned, exported, and reused — without losing continuity.
Every data point shows its source.
Each observation keeps its source, timestamp, and chain of interpretation visible.
Nothing gets dropped over time.
Changes, unresolved questions, experiments, and care events stay connected across years.
Guesses stay labelled as guesses.
Inferences carry confidence and can remain provisional instead of being flattened into fact.
You can change or delete any of it.
Inspect, correct, approve, export, restrict, and delete meaningful parts of your profile.
The app is one viewport. The profile is the continuity layer. Agents and clinician-facing summaries inherit the same context, instead of rebuilding it from scratch. Explore Health Core →
What the record actually looks like.
An illustrative view of a Longitudinal Health Profile. Each entry keeps its source and a confidence level — and agents reason across them instead of guessing in isolation.
An illustrative view of a Longitudinal Health Profile. Each entry keeps its source and a confidence level, and a recovery agent connects them across time.
Sleep below baseline
Mar 126 nights of short, fragmented sleep versus your February baseline.
WearableHigh confidenceNew medication started
Mar 14Logged a change to a prescription, with dose and start date.
You loggedConfirmedResting heart rate rising
Mar 18Resting HR up ~9 bpm against your personal baseline.
WearableMedium confidence- Recovery agentProvisional
Resting heart rate began rising about two days after the medication change. Worth raising at your appointment — I can prepare a summary.
Illustrative sample data — not a real profile or medical advice.
Specialist agents should share context, not create more silos.
Health Core is designed so agents can be narrow, accountable, and useful without each maintaining a separate version of the person.
One shared Longitudinal Health Profile. Many specialist agents, each connected to the same context.
Domain agents
Sleep, recovery, symptom, medication, clinical prep, and experiment agents operate over the same profile.
Permissioned writes
Agents declare what they can read, what they can write, and which actions require explicit approval.
Shared context
Contradictions and low-confidence interpretations remain visible because agents do not maintain private versions of the person.
Early agent domains under exploration:
- Sleep
- Recovery
- Symptoms
- Medication
- Experiments
- Clinical prep
A shared foundation for different health surfaces.
The same longitudinal profile can support a personal app, a clinician-ready summary, and developer-built agents without forcing each surface to start from zero.
A clearer longitudinal picture of symptoms, habits, goals, experiments, and care history.
Learn more →Patient-owned context that can make encounters better prepared without replacing clinical judgment.
Learn more →Stable primitives for modular health agents that need provenance, permissions, and continuity.
Learn more →Building something on longitudinal health context, or want it for yourself?
Request beta accessThe non-negotiables we build to.
These are engineering constraints, not slogans — the properties without which longitudinal health infrastructure cannot be trusted.
Recommendations are calibrated to your history — not a population average that bears no resemblance to you.
One consistent model of the person. Agents do not contradict each other without explanation.
Unresolved threads and follow-ups are preserved. A longitudinal agent is not a stateless chatbot.
You stay in control of your own model — more capable and informed, not more dependent on the platform.
What we will not compromise.
- Agents stay narrow and accountable, never generic bots
- Uncertainty is preserved, never flattened into false certainty
- Composition is modular, so the system can be inspected and extended
- The person owns the model, with provenance at every layer
Agency is concrete. Over any meaningful part of your profile, you can:
- Inspect
- Correct
- Approve
- Export
- Restrict
- Delete
Developer alignment matters in health infrastructure.
Health Core is being shaped as open-core infrastructure so the primitives can be inspected, extended, and trusted by serious builders, not hidden behind a closed wellness app.
Developer model →Early, serious, and deliberately scoped.
QED is building the foundations first: schema, profile, permissions, provenance, and the first agent surfaces. The work is intentionally narrow before it becomes expansive.
Health Core schema and ingestion primitives
The Longitudinal Health Profile interface
Consent, export, correction, and provenance flows
Early modular agents for personal context and care preparation
Follow or collaborate on the foundations.
We're speaking with early users, clinicians, researchers, developers, and partners who care about longitudinal context, data continuity, and safer health agents.
- Your health model belongs to you — inspect, correct, export, or delete it.
- Provenance and uncertainty are visible, never buried.
- Not a diagnostic or treatment system.