XPRIZE submission

Evaluate this in two minutes

A short, honest path through the product: what works, what it is built on, and what is represented rather than rendered.

01
Estimate
Four inputs produce a VO₂max estimate, fitness age, percentile and live zone bands. No account required.
Open
02
Prescribe
Generate a session. Note the six phases, the per-phase BPM anchors derived from heart-rate targets, and the Gemini coaching cue returned with the plan.
Open
03
Verify
Open the public audit explorer. Records are append-only and hash-chained; the header reports live chain verification from a database function, not client code.
Open
04
Trace rights
The eleven-step Content ID workflow shows disclosure, watermarking, hashing and reference delivery for every render.
Open
05
Integrate
Four public JSON endpoints with a live health check. Everything the app does is reachable over HTTP.
Open

Google stack usage

Coaching cues generated per prescription through the Gemini API; the model receives physiology and mood tokens only, never identity.

Postgres with row-level security, security-definer role checks and an append-only trigger enforcing immutability of sealed rows.

Installable PWA today with home-screen delivery; the same public API is the contract for the Android Health Connect and Apple HealthKit shells.

Each render reserves an asset id at generation time and ships fingerprint plus ownership metadata for reference delivery.

Stated limits

VO₂max from heart-rate ratio carries ±10–15% error against lab testing; the UI states this wherever an estimate is shown.
Audio rendering and inaudible watermark embedding are represented by their record fields in this build; the ledger schema and hashes are real and verifiable.
Device connections model scope grants and revocation; live vendor OAuth requires per-partner credentials outside this environment.
Not a medical device. No diagnostic claims are made anywhere in the product.

Estimate