Apple Intelligence already gives every recent iPhone an on-device model, a privacy rail, and the motion sensors on your body — for free. What it does not give you is a clinical gait method. And in February 2026, Apple proved it won't: it shelved Project Mulberry, its own "AI doctor," rather than ship it. The horizontal platform is Apple's. The vertical method is ours — and that division is the whole strategy.
For over a year Apple built Project Mulberry — an AI health coach meant to "replicate your doctor," trained with cardiologists, physical therapists, and nutritionists, slated for the Health app. In early 2026, after a leadership change put Eddy Cue over the health division, Apple wound it down — Cue reportedly judged it wasn't good enough and that Oura and Whoop already did it better. Apple retreated to rolling out narrow Health features and educational video.
The most valuable company on earth, with the sensors, the model, and the distribution, could not make a horizontal AI health agent compelling — and said so out loud by shelving it.
That is not bad news for a company like this — it is the thesis, confirmed. The hard, defensible, reimbursable value was never the horizontal agent. It is the vertical method: one body signal, measured rigorously, with cited thresholds, physician attestation, and a billing rail. Gait is precisely that vertical — and it is wide open.
This isn't competition — it's a clean interface. Everything on the left, Apple gives away and improves every year. Everything on the right, Apple structurally will not build. We take their platform and add the clinical layer.
The on-device model turns the method's output into plain language — "your gait dipped 0.08 m/s over three weeks; worth a look" — entirely on the phone. No API cost, no PHI leaves the device. The model phrases the finding; the cited method makes it. Exactly Apple's stated use for the model (summarization, not diagnosis).
"Hey Siri, what's my gait number?" Expose the method as an App Intent and it surfaces in Siri, Spotlight, Shortcuts, a Lock Screen widget, and Apple Intelligence's personal context — the way people already know their resting heart rate.
When a read needs more than a 3B model — multi-metric fusion, longitudinal synthesis — Apple's PCC runs server-class models with cryptographic privacy guarantees, so the trust story never breaks between phone and cloud.
With Mulberry shelved, Apple is rolling out narrow Health features one at a time. A rigorous, cited gait/fall-risk module is exactly the kind of vertical depth a horizontal Health app lacks — a drop-in, not a competitor.
Sensors, the on-device model, and the privacy rail improve every OS release at Apple's expense. Building on them means the capture and interpretation layers get better without our capital. Betting against them — a custom device, a cloud model, a privacy story to prove — is betting against the tide.
Cited clinical thresholds, physician attestation, CPT/RPM billing, an FDA file. Mulberry is the proof Apple retreats from exactly this. The durable value isn't the AI — it's the method + the accountability + the reimbursement wrapped around it. Apple hands us the horizontal for free and leaves the vertical entirely to us.
For a device maker, a health platform, or a payer: MotionSole is the gait/fall-risk vertical that rides on Apple Intelligence — the on-device sensors and model do the capture and phrasing, and our published method, physician-attestation rail, and RPM/RTM billing do the clinical and reimbursable part Apple shelved Mulberry rather than ship. Ride the platform; we bring the vertical.
Start a conversation → See the capture layer