Thesis 01 · Consumer health

Biomarker Expansion

Consumer devices measure a dozen signals. The human body emits hundreds of thousands. The next medical-grade markets will be built in that gap.

The leading commercial wearables track somewhere between 10 and 15 biomarkers: heart rate, heart-rate variability, blood oxygen, skin temperature, sleep staging, steps and a few others. Dozens more are already being measured in clinical trials and next-generation lab prototypes.

Widen the lens and the gap is enormous. Health data is still in its infancy, and the teams that obsess over new ways to measure the body, and new ways to connect what they measure, will reshape the industry.

01How much of the body do we actually measure?

Very little. There are roughly 34,500 validated molecular biomarkers, but only about 1,400 digital measures for them. The metabolome and proteome hold more than 200,000 candidate markers, some not yet discovered. For nearly all of them, we don't know how they relate to disease, longevity or performance.

The measurement gap · biomarkers, log scale
Tracked by leading consumer wearables~12
Digital measures that exist~1,400
Validated molecular biomarkers~34,500
Candidates in the metabolome + proteome200,000+

Each gridline is ten times more of the body. Consumer health lives in the first sliver.

02Why now?

Three curves are crossing.

  • Sensing is leaving the wrist. Contactless radio, optical, breath and sweat chemistry, and continuous molecular sensing.
  • Foundation models learn physiology from raw signals instead of hand-built features.
  • Regulators have paths for software that keeps learning, such as the FDA's Predetermined Change Control Plans.

03What does one signal, read differently, unlock?

I spent five months as the first go-to-market hire at Emerald Innovations, an MIT CSAIL spin-out led by MacArthur Fellow and MIT professor Dina Katabi. Emerald builds a Wi-Fi-like device that tracks a person's breathing and movement from radio signals, with nothing worn on the body.

By licensing those devices into clinical trials at biotech and pharma companies, Emerald built a physiological dataset across young and old, healthy and sick, and trained a foundation model on it. From breathing alone, during a single night's sleep, that model can detect Parkinson's disease. Add movement and gait, and the same signal stream speaks to pain, mobility and a range of other conditions.

My job was to turn a licensing business into a subscription product for care facilities. I sourced 47 senior-living facilities, brought 16 into discovery, and landed a verbal pilot with a Boston skilled-nursing facility.

Plenty of devices track breath rate. Almost none track how you breathe. That's one existing biomarker, read differently. There are hundreds of thousands left.

What I look for

  • Modality: is this a new way to measure the body, or a new reading of a signal we already capture?
  • Flywheel: does every trial, partner or user make the model better in a way a competitor can't buy?
  • Validation: is there a credible path to clinical evidence and reimbursement?
  • Buyer: does the team understand who pays as deeply as they understand the science?

What would make me wrong?

  • Validation and reimbursement timelines outlast venture timelines.
  • Incumbent device makers fold new sensors in faster than startups can build distribution.

Diligence questions

What I ask founders building against this thesis. Open any question to see why it matters.

ISignal

Is the measurement real, new and capturable?

  1. 1.What are you measuring that no consumer device measures today?

    The thesis only holds if the signal expands what's measured, not just how conveniently.

  2. 2.How was the signal validated, against what reference, in how many people?

    Small, homogeneous cohorts are where most digital biomarkers quietly fail.

  3. 3.What does it take to capture it every day, passively?

    Adherence decides whether a biomarker becomes a dataset.

IIData

Does the data compound?

  1. 4.How does each new user, trial or partner make the model better?

    Emerald's edge came from bootstrapping data through trials. I look for the same loop.

  2. 5.Who owns the data, and what consent covers future uses?

    The most valuable use of a dataset is usually one nobody planned for.

  3. 6.How deep is the longitudinal record: days, months or years?

    Progression and prediction need time series, not snapshots.

IIIClinic

What's the path to medical grade?

  1. 7.Which regulatory path are you on, and what's the timeline?

    Wellness, 510(k) and De Novo are different companies with different burn.

  2. 8.What clinical evidence exists today, and what will the next study prove?

    Each study should unlock a buyer, not just a paper.

IVBuyer

Who pays, and why now?

  1. 9.Who signs the first contract: consumer, employer, payer, provider or pharma?

    At Emerald, the hardest work was turning a licensing model into one a facility would buy.

  2. 10.What does pilot-to-contract look like, and how long does it take?

    Pilots are cheap to start and expensive to convert.

  3. 11.Is there a reimbursement code, or does the ROI stand on its own?

    Without one, the product has to pay for itself inside the buyer's budget.

If you're building at this edge, I want to hear from you.

Tell me what you're measuring and who it's for. I read every submission myself and connect the strongest founders with the investors and operators who should know them.

Next: Thesis 02 · Inverse Design for Matter →
Raising now?

140 characters max