Still open, and the one item here that needs work. Per-NIGHT translation was tested and abandoned: across the 45 paired Oura/sleep-study nights the correlation is r = 0.195, and every one-parameter model (ratio 1.13, ratio 1.10, offset +0.68h) has NEGATIVE R² — worse than guessing the mean. The correction is smaller than its own error bar. Nights are shown as the device reported them, with the source named. What is NOT solved is doing it per METRIC. Sleep duration, HRV and SpO2 each disagree between devices in their own way and for their own reasons: the Apple Watch reads HRV 1.24x Oura across 526 paired nights, and that is a DEFINITION difference (SDNN vs RMSSD) rather than device error, while sleep duration differs for entirely unrelated reasons. One factor per device was always the wrong shape — it should be one per metric, if at all. Worth deciding together which metrics justify the paired-night cost and which should simply stay labelled by source.

All three routes closed as of 2026-09-02. The Fitbit Web API sunsets this month and dev.fitbit.com has already stopped issuing new app registrations, so there is no key to get. Its replacement, the Google Health API, puts sleep and HRV behind RESTRICTED scopes needing OAuth verification plus an annual third-party security assessment — Google refused the scope outright. The Apple Health bridge, which would have needed no code at all, is permitted but unused: the app has read AND write access to sleep and has never written a record. Not worth reopening. HRV already has three sources and sleep duration three more; a fourth needs ~20 paired nights before it can say anything, and the 12-week HRV plan names Oura as the single source through 2026-11-21 with the Fitbit listed as a confound to exclude. The real constraint is Oura’s coverage — 19 of the last 30 nights — which no extra device fixes.

Set and rep detail is the one thing the adherence integration cannot get today. Nothing to build until the key lands; then strength can show prescribed-vs-completed per set.

John S can import them into the Bridge program from an upload, screenshot or rough recall, so the blocker is getting sessions recorded at all. The last one that reached the database was the manual “Gym with Ben” entry on 2026-08-11.