Retention is a clinical problem before it is a business problem. Solve the first, and the second follows.
Our Thesis
Why TeraHealth,
Why Now
The prescription is a commodity.The protocol is the business.
01The Shift
The market split in two.
Healthcare is dividing into two economies. One is insurance-billed and script-driven — and it's being commoditized fast, by whoever can write the cheapest, fastest prescription. The other is cash-pay and protocol-driven: GLP-1 therapy, longevity medicine, functional medicine. These practices don't compete on access to a drug. They compete on whether the patient is still enrolled in month twelve.
That's a retention problem, and retention is decided by what happens between visits — whether the pillars below are working as one coordinated plan, or as disconnected point solutions the patient quietly abandons.
NutritionSupplementsFitnessSleepStressRecoveryToxin avoidance
02The Gap
Everyone built the edges. Nobody built the center.
The last few years consolidated everything around the clinical decision: practice management, lab ordering, supplement fulfillment, ambient scribes. Data platforms narrate labs. Fulfillment platforms ship supplements. Operations platforms schedule visits. None of them generate a clinical hypothesis a care team can act on.
That center — a system that turns labs, symptoms, wearable trends, and patient history into a structured hypothesis for the care team to review — is the piece nobody has built.
It's also the hardest piece, which is the honest reason it's still open: it requires reasoning that's grounded in versioned clinical rules, not a model's best guess, and a workflow built for a multidisciplinary team rather than a single sign-off. We built Tera to meet that requirement from day one. The architecture behind it lives on our Safe AI page.
03Why Now
Two curves crossed in the same window.
Capability
The model got useful — inside constraints.
Language models finally got capable enough to be useful inside a constrained clinical system — fluent enough to parse messy patient data, structured enough to be governed by deterministic rules rather than left to their own judgment. That gap between capable and trustworthy is exactly why Safe AI was the unlock, not a side benefit of it: a model good enough to draft but not yet good enough to decide is precisely the tool a rules-anchored system was built for.
Market
The category hit scale.
Cash-pay protocol medicine crossed from niche to scale, fast enough that the practices running it are actively looking for infrastructure, not just tools — not point solutions they have to stitch together themselves.
Neither curve was there three years ago. Neither will still be open in three more — outcome data compounds, and whoever has been running supervised, evidence-cited protocols the longest will have the best model of what actually keeps a patient enrolled. This gets built once, in this window.
04Our Beliefs
What we believe.
The first two weeks of a protocol predict the next year of adherence.
Tera generates structured clinical hypotheses, grounded in clinical rules and evidence, with a citation behind every recommendation. The care team reviews and decides, every time, no exceptions.
The care team should build one coordinated plan across nutrition, supplements, fitness, sleep, stress, recovery, and toxin avoidance — not seven disconnected point solutions — because the patient only has one calendar.
Get the protocol right and retention follows. Retention is the economics of this business — not a metric next to it.