What no single vendor can see alone.
Cross-vendor signal alignment, agreement scoring per field, lifetime baseline carried across hardware churn. The 9pm glucose spike that caused the 2am HRV crash lives here — in the synthesis layer, not in any single ring app or CGM dashboard.
Switch between raw and synthesized. Disagreement becomes visible.
Toggle to Raw · per-source on HRV to see Oura, Whoop, and Apple Health drift apart during travel days — the 3-hour timezone shift physically changes how each sensor reads the same physiology. Synthesis reconciles the disagreement into one weighted-by-confidence value.
The 9pm glucose spike caused the 2am HRV crash.
A late dinner spikes glucose. The body burns through clearance mechanisms for hours. Around 2am, the autonomic nervous system registers the load and HRV collapses. Three sensors, one causal chain — only visible when their streams are aligned in time.
09:45 pm
Late dinner logged · MyFitnessPal flags 2940 kcal day total + 2 drinks
10:00 pm
Glucose peaks at 158 mg/dL · 60 mg/dL above your dinner-window baseline
02:00 am
HRV crashes to 31 ms · 21 ms below 30-day median, deepest dip in 30 days
No single vendor sees this causal chain. Dexcom owns the glucose curve. Oura owns the HRV. MyFitnessPal owns the food log. The spike-→-crash relationship lives in the cross-vendor synthesis — exactly the ground a hardware vendor cedes by definition.
Which vendors corroborate which signals.
When two sources cover the same field, Continuum scores the agreement per day and surfaces it in every insight's provenance. Disagreements are flagged, never hidden.
| source ↓ / validates → | Apple | Oura | Whoop | Dexcom | MyFitnessPal | Strava |
|---|---|---|---|---|---|---|
| Apple Health | · | ✓ | ✓ | — | — | ✓ |
| Oura Ring | ✓ | · | ✓ | — | — | — |
| Whoop 4.0 | ✓ | ✓ | · | — | — | — |
| Dexcom Stelo | — | — | — | · | — | — |
| MyFitnessPal | — | — | — | — | · | — |
| Strava | ✓ | — | ✓ | — | — | · |
✓ = sources cover overlapping fields and produce cross-validatable signals. Dexcom and MyFitnessPal carry no peer overlap — their data is reported with single-source confidence and no cross-vendor agreement multiplier.
One continuous record across hardware churn.
Apple Watch in 2024. Whoop in 2025. Oura in 2026. The user accumulates 5–10 vendor models over a decade. Every hardware switch resets the vendor's personal model — but Continuum's baseline carries through.
HRV baseline
51ms
range 47–54ms over 21 months
Records ingested
665K
across 4 vendors, 21 months
Vendor switches
4
zero gaps in baseline
Hardware vendors can't do this without surrendering lock-in. Their personal models are a moat against churn. Continuum's moat is the opposite shape — it grows with every hardware switch the user makes. Time becomes a structural advantage.
Three commitments the inference layer rests on.
Sensor fusion is the moat
Glucose × HRV × calendar × food. The cross-modal correlations no single hardware vendor can see, because they only own one signal in the chain.
Disagreement is a feature
When sources disagree, the synthesized signal is flagged low-quality and surfaced with reduced confidence. Continuum doesn't hide vendor noise; it scores it.
Time is the moat's compounder
Every month adds another 30 days of personalized model fit. Every hardware switch tests baseline carryover — the longer the user is on Continuum, the more durable the model becomes.
The synthesis layer is one endpoint away.
GET /v1/timeline returns the user's lifetime baseline — synthesized across every connected vendor, with per-field agreement scores and vendor provenance attached. Drop into any wellness app, employer dashboard, or clinical surface.
// pull lifetime baseline
GET api.continuum.health/v1/timeline
?user=usr_8f2a91
&fields=hrv,sleep,glucose
&range=lifetime
// → { fields[], vendors[], agreement[] }