The logic is deterministic. The AI only narrates.
Rules written from peer-reviewed literature decide every clinical call — never a language model. The AI rewrites them into plain English and answers your questions, but it can't invent a threshold. That's our safety posture, and our trust story.
Rules, not vibes
Every threshold, flag and escalation is a rule written by our clinical team from the literature. Given the same inputs, you get the same output, every time — auditable and versioned.
A doctor signs off
A registered clinician reviews every panel before release. Critical values are routed to them for action, never handled by the coach or the model.
Read against you
Population reference ranges are wide and blunt. We track each marker against your own history and flag change using Reference Change Value, so ‘normal’ doesn’t hide a real trend.
The model only explains
The AI turns the rule output into plain English and answers your questions, grounded strictly in your numbers and the rules. It cannot set a threshold or overrule the clinician.
Was it a real change, or was it just noise?
Every test has analytical variation (the lab) and biological variation (you, day to day). A number that ticks from 56 to 58 hasn't necessarily moved. Reference Change Value is the threshold a marker must cross before we'll call the difference real — so we never celebrate noise or scare you with it.
CVi — your within-person biological variation
Z — confidence (1.96 for 95%)
Why these markers
We prioritise markers with the strongest outcome evidence over the familiar ones. A few examples of the reasoning behind the panel:
Counts every atherogenic particle, not just the cholesterol they carry. A better predictor of cardiovascular risk than standard LDL-C, and the marker we lead with for heart health.
Largely genetic and measured once in a lifetime. A high Lp(a) meaningfully changes your risk picture and is invisible to a standard cholesterol panel.
Together they catch metabolic drift years before glucose alone would. We watch the trend against your activity and sleep, not a single reading.
A sensitive marker of low-grade inflammation. Fused with your HRV and recovery data, it separates a training load from a genuine inflammatory signal.
Wellness & optimisation — not diagnosis
Arcaevo is built for healthy people who want to stay that way. When a result crosses a clinical threshold, we don't coach it — we flag it to our reviewing clinician and tell you, plainly, to see your GP. Read our clinical safety statement for the full escalation policy.
Clinical safety statement →