Skip to content
QEDHealth Systems
Health Core

The Longitudinal Health Profile

The LHP is the canonical, user-owned context layer that turns raw events into a coherent model of the whole person. It is not a static profile — it is a living model that stays accurate as you change.

A living model

What the profile holds

Every entry is interpreted in relation to everything before it. The profile preserves meaning over time, not just data points.

What happened, and when it happened
Where the information came from
How confident the system is
What changed over time
What remains unresolved
What the user cares about
What agents have inferred
What actions or follow-ups are pending
Goals, preferences, and experiments
What it looks like

From raw events to a living model

A few entries from different sources become a coherent trajectory — each event keeps its source, confidence, and unresolved threads, so agents can reason over it without losing the thread.

Illustrative · synthetic data

Timeline

Apr 2026
  1. 2 AprLabsUnresolved

    Ferritin 28 ng/mL — below reference range

  2. 9 AprWearable

    Resting heart rate drifted 58 → 64 bpm over a week

  3. 14 AprNote

    “Afternoon fatigue, hard to focus after lunch.”

  4. 16 AprMedication

    Started iron supplement, 65 mg daily

  5. 30 AprWearableChanged

    Sleep efficiency 78% — up from 71%

Event

Unresolved

Ferritin 28 ng/mL

Below reference range

Source
Lab import · MedLab
Recorded
2 Apr 2026, 09:14
Confidence
High
Linked goal
Resolve persistent fatigue
Follow-up
Recheck in 8 weeks

Nutrition agent

Suggest · level 3

Low ferritin on 2 Apr alongside the afternoon-fatigue note from 14 Apr is consistent with low iron. Consider rechecking ferritin in 8 weeks and taking the supplement with vitamin C to aid absorption.

Reading: 2 events · 1 note · 1 goal

Confidence: moderate · awaiting your approval
Core surfaces

Four surfaces, one core loop

Different views into the same profile. Each serves a distinct job — and all strengthen the same underlying model.

Input surface

Capture

A chat-first, voice-capable, quick-log interface — a translation layer from messy human experience into structured longitudinal context. Not just chat: a way to record events, symptoms, behaviours, goals, and reflections.

Profile surface

Understand

A human-readable view of your current health model: timeline, goals, changes, patterns, uncertainty, and agent interpretations. It helps you see what the system believes — and why.

Agent surface

Reason

A modular interface where specialist agents explain, suggest, plan, follow up, and coordinate around your profile. It feels unified, but stays inspectable.

Developer surface

Inspect

A technical view into schema behaviour, ingestion quality, provenance, confidence, and agent outputs — so the system never becomes a black box.

Chat is only one interface

The conversation is a doorway. The profile is the house.

A chatbot answers in the moment and forgets. Health Core maintains a coherent model of your health trajectory, so every interaction builds on the last.

You can talk, log, or import — but the value isn't the message. It's the durable, inspectable model underneath that agents and clinicians can rely on over time.

Design principles

What we won't compromise

These commitments decide what gets built. They set expectations for early partners — and keep the product honest as it grows.

Coherence over features

New data is interpreted against everything before it. We protect one coherent model rather than shipping disconnected features.

Uncertainty is preserved, never buried

Confidence, provenance, ambiguity, and change are first-class. The system never presents inference as fact when it isn't justified.

You stay in control

You can inspect, correct, approve, restrict, export, and delete what the model holds. The profile makes you more capable, not more dependent.

One shared context, no separate truths

Every agent reads and writes the same profile. We won't let separate agents create separate versions of you.

Chat is an interface, not the product

Conversation is one way in. The durable value is the inspectable longitudinal model underneath.

Infrastructure, not lock-in

Decisions favour portability and open-core compatibility over locking your health context into a single app or device.

These flow from the principles behind Health Core. Read the manifesto → or see how we handle trust and control →

Agent autonomy ladder

Autonomy earned through checkpoints

Agents climb an eight-rung ladder, but never jump to higher autonomy without the safety, consent, auditability, and confidence to justify it.

1Levels 1–3

Explain · Interpret · Suggest

Make the profile legible — summarise what it holds, surface patterns and gaps, and recommend next steps for you to consider.

2Levels 4–6

Plan · Follow up · Coordinate

Build cross-domain plans, remember unresolved threads, and hand context between specialist agents while preserving coherence.

3Levels 7–8

Act with consent · Support care

Trigger reminders, update records, or prepare exports — only within approved permissions, bridging daily life to clinical care.

See the full eight-rung ladder in the developer model →

Engineering philosophy

Built on stable primitives

Health Core is longitudinal infrastructure, not a bag of features. Every capability maps back to a small set of durable primitives — so each event carries its own source, confidence, and links.

Longitudinal by default, provenance-first, timeline-aware, uncertainty-aware, user-correctable, and open-core compatible. See the full primitive vocabulary and developer model →

event.json
{
  "id": "evt_8f3a…",
  "type": "observation",
  "observation": { "metric": "ferritin", "value": 28, "unit": "ng/mL" },
  "timestamp": "2026-04-02T09:14:00+10:00",
  "source": { "kind": "lab_import", "name": "MedLab" },
  "provenance": "imported",
  "confidence": 0.94,
  "links": { "goal": "goal_resolve_fatigue", "follow_up": "fu_recheck_8w" }
}
Private beta

Start building your longitudinal profile.

Tell us how you'd use Health Core and we'll be in touch as we open more access — whether you want to use it yourself, support patients, or build agents on the LHP.

Which best describes you?

Non-binding — export your data and leave the beta at any time.