Harmonica is building the first multi-context health model, a single system that reads a person's genome, their day to day physiology, and the metabolic signals in between, then turns all of it into one continuous picture of their health. Most of medicine still runs on episodes. A fifteen minute visit twice a year, a lab drawn once something already feels wrong, a specialist working from notes that are weeks old. Everything that happens in the spaces between those moments stays invisible to the people who are meant to act on it. Harmonica exists to close that gap, and it starts at the very beginning of life, where the need is most acute and the data is richest.
The mission is easy to state and hard to earn. American healthcare spends close to five trillion dollars a year and still produces outcomes that rank far down the global list, and a large share of that spending pays for conditions that earlier information could have prevented. We do not think the answer is another dashboard or another isolated point solution. The answer is context. When a parent, a pediatrician, or a care team can watch how a child's biology, vitals, and behavior move together over time, the events that would have arrived as emergencies become patterns that can be read in advance. That is the change Harmonica is built to make, for one family first and eventually for many.
The technology rests on three data layers that have never been gathered together from birth. The first is the genome. We sequence at clinical grade and screen for the high confidence variants that change what a clinician should do today, alongside polygenic scores that we treat as one input among many rather than a verdict. A single sequence can quietly answer questions that otherwise take a twelve to eighteen month diagnostic journey to reach, and for conditions like Dravet syndrome or spinal muscular atrophy, reaching them before symptoms set in is the entire difference between outcomes.
The second layer is continuous physiology, measured against each person's own baseline rather than a population average, so a heart rate or oxygen trend that drifts in the wrong direction is caught while it is still a trend. The third is serial metabolic monitoring, simple blood spot panels that show when a pathway is under strain well before a crisis. On their own, each layer is partial and easy to misread. Fused into one model, they reinforce each other. Genomics describes what is plausible, metabolism shows what is happening, and physiology reveals where it is surfacing in real time. The intelligence that emerges is delivered where decisions actually get made, in the parent's app, in the report a clinician reads before an appointment, and over time in the systems that hospitals and researchers already rely on.
We are building it the way durable health companies have to be built, from the data outward. The first product is a kit for new parents, a wearable and a whole genome paired with a subscription that keeps the picture current. That earns the trust and the longitudinal record no competitor can assemble after the fact, because it can only be collected by a company that owns the full path from sensor to sequence to the family's own experience. From there the same model powers provider tools, population insight, and licensable intelligence for the partners who shape care at scale. We are early and honest about it. Clinical validation is underway, the regulatory work is in progress, and every claim here is one we intend to prove. What we are already certain of is the shape of the problem, and the shape of the answer.
