Levels, CGMs, and the case for real-time metabolic data

Executive overview

88% of Americans are metabolically unhealthy, yet CGMs — devices that give real-time blood glucose data — have only ever been prescribed to diabetics. Josh Clemente, former SpaceX life support engineer, used one on himself and discovered his "healthy" diet was spiking his glucose into diabetic ranges. The insight: the data exists, but no one has built the software layer to make it actionable for the non-diabetic majority.

The metabolic crisis is a feedback problem — real-time glucose data closes the loop diet advice never could.

From SpaceX to metabolic health

• Clemente joined SpaceX as employee ~678, starting as a manufacturing engineer building and launching Falcon 9 hardware.
• Led the pressurized life support systems team for the Dragon Crew capsule — oxygen systems, fire suppression, docking pressurization.
• Hit severe burnout: low energy, mood crashes, symptomatic episodes of shakiness and cold sweats with no diagnosis.
• A research paper by Dominic DiAgostino showed ketogenic rodents surviving five times longer in high-oxygen toxic environments — shattered his "a calorie is a calorie" worldview.
• Began finger-pricking 60 times a day, plotting glucose in Excel; the gaps in the data made the signal useless.
• A friend brought a CGM back from Australia (available OTC there); first use immediately revealed glucose traces "more like a heart rate than a blood sugar" — massive spikes and crashes.
• Foods he considered healthy — sweet potatoes, brown rice, quinoa, oatmeal — pushed him into diabetic-range responses.
• Correlated the crashes directly with years of symptomatic episodes; tuned his diet within two weeks using trial and error.

Why CGMs had never reached consumers

• CGMs were developed exclusively for diabetes management — type 1 (autoimmune, no insulin production) and type 2 (insulin resistance built over time).
• Dexcom and Abbott's Libre dominated a therapeutic market; pre-2017 devices cost $1,000+/month and required prescriptions.
• When Clemente asked his doctor for one — showing his finger-prick spreadsheet — he was told: "You'd be ashamed to ask if you saw the people who actually need it."
• The insurance coding system forces physicians into diagnostic buckets; without a diagnosis code, no prescription.
• 90 million Americans have pre-diabetes; 84% don't know it; 70% will convert to type 2 in their lifetime.

The "why now" for direct-to-consumer digital health

• Companies like Hims, Hers, and Roman proved a new model: combine telehealth with low-risk medical products into a consumer checkout experience.
• Physicians in an independent network review asynchronous intake forms; mail-order pharmacy handles fulfillment — regulatory compliance hidden behind a seamless UX.
• COVID accelerated telehealth reform: multi-state licensing, platform flexibility (FaceTime, Google Meet), and broader asynchronous consultation acceptance.
• The consumer path unlocks traditional market forces — competition, economies of scale, price compression — that the insurance-reimbursement path suppresses.
• Historically, digital health split into two lanes: generic wellness advice (too vague) or post-diagnosis managed care (high friction, captured audience). Levels is carving a third lane.

What Levels is building

• An insights layer on top of raw CGM data — analogous to what Oura and Whoop did for heart rate, or what Apple did with step counts and rings.
• The app converts raw milligrams-per-deciliter readings into meal scores (1–10), correlating nutrition, exercise, sleep, and stress in one view.
• Users learn that a 10-minute walk after an indulgent meal can nearly eliminate a glucose spike — a lesson that sticks because it's grounded in their own data.
• Levels holds what it claims is the largest non-diabetic CGM dataset by an order of magnitude, even while in invite-only beta.
• 7,000 beta users through the program; 105,000 on the waitlist; no paid marketing — growth driven entirely by content and podcasts.

Business model and trajectory

• One-month program at $399: user gets CGM hardware (Abbott Libre), software access, and a physician consultation via telehealth network.
• Margins on the one-month program: ~50–60%. Subscription tier exists but currently near break-even, constrained by sensor hardware costs.
• Clemente frames the roadmap using Tesla's arc: current product is the Roadster (expensive, limited scale); Model 3 and eventually a mass-market option follow as hardware costs fall.
• The real target market — lower socioeconomic groups with highest metabolic risk — requires hardware commoditization, software intelligence, and regulatory change to reach.
• Deliberate strategy: build premium direct-to-consumer first to force quality, then use that data and experience to earn payer/employer reimbursement from a position of strength.
• Subscription data moat: longitudinal personal glucose history makes switching costly and enables progressively better phenotyping and personalized recommendations.

The metabolic health crisis at scale

• Glucose variability — number and amplitude of spikes — is directly correlated with cardiovascular disease; it's inflammatory.
• Stress alone, without any food intake, can push glucose above pre-diabetic thresholds (Clemente demonstrated this live via CGM in a stressful meeting).
• 70% of oatmeal-eating Levels users exceeded the pre-diabetes postprandial threshold — a food universally described as "heart healthy."
• ~30% of the global population is estimated to be pre-diabetic; billions are on a trajectory toward metabolic disease without knowing it.
• Policy and population-level nutrition advice fails because it targets the average person; real-time individual data enables each person to solve for themselves.

Hardware future and the Apple Watch question

• Levels currently uses Abbott Freestyle Libre; Clemente has worn a CGM continuously for three years.
• The core value gap is not in the hardware — it's in the software intelligence between raw data and behavior change.
• Apple faces a harder problem than Levels: non-invasive measurement of a colorless, small, water-soluble molecule (glucose) in a primarily-water fluid, at tight concentrations, without breaking the skin.
• Suspected technique: Raman spectroscopy (light scattering). Clemente is skeptical it arrives imminently but would welcome it.
• Historically Apple ships hardware before the software ecosystem — Levels is positioned as the insight layer that would sit on top of any new sensor Apple or others bring to market, just as Instagram sat on top of iPhone cameras.
• Levels is open to joint hardware development or creating market pressure that drives sensor innovation if necessary.