Apple Watch recovery, strain, and sleep coach app demo

[Note: the creator shared a partial prompt. Wording below is theirs.] GOAL: Build "Pulsr": a recovery, strain and sleep coach for the Apple Watch I already own. MOTION: iPhone app + Apple Watch app. Free, open-source, no subscription, no server. I attached 14 screenshots of WHOOP as a quality bar. Match the depth and polish, not the brand: original name, icon, copy and score names. Nothing that looks like WHOOP's logo. ## How to work 1. Map first. Go through every screenshot and list each feature next to the Apple Health data that can power it (HRV SDNN, resting HR, respiratory rate, sleep stages, wrist temperature, SpO2, VO2 max, workouts, steps). Mark what Apple Watch cannot measure and skip it. 2. Engine before UI. Write all scoring as pure Swift functions with no SwiftUI. Run them on 60 days of realistic synthetic data and print the results. Check that the numbers look like a real person's before building a single screen. 3. Design first, code later. Make 4 different visual directions as one interactive HTML page (iPhone screens + an Apple Watch face each). Let me pick, iterate on my feedback, then build the final direction with every screen in light AND dark. 4. Only then write the app, and port the design 1:1. 5. Verify for real: build for the Simulator, run every screen in light and dark with demo data, fix what looks wrong, then install on my iPhone and Apple Watch. Show me screenshots. ## Stack - SwiftUI, iOS 26 + watchOS 26, Liquid Glass (glassEffect, GlassEffectContainer). - XcodeGen project.yml, generated .xcodeproj committed so nobody needs XcodeGen. - Shared/ folder compiled into both targets: models, scoring engine, HealthKit source, demo source, theme. - HealthKit read-only, async query descriptors. Never write to Health. - Signing in a git-ignored Config/Local.xcconfig (team ID + bundle prefix) with a committed .example file. - A `-demo` launch argument so the Simulator (which has no Health data) shows demo data. ## Scores (document every formula in docs/ALGORITHMS.md) - Recovery 0–100%: ln(HRV), RHR, respiratory rate and sleep performance as z-scores against MY 30-day baseline, shrunk toward population priors for the first 7 nights ("calibrating"). Green ≥ 67, yellow 34–66, red < 34. - Strain 0–21: Banister TRIMP from minute heart rate, interpolating watch gaps ≤ 10 min. Daily strain target from recovery. - Sleep: need = goal + yesterday's strain + sleep debt measured against the goal (so it can't compound), performance, consistency, efficiency, stages, recommended bedtime. - Stress 0–3 from heart rate above resting, workouts excluded. - Health monitor: each overnight vital vs. my own ±2 SD range. - Body age: VO2 max, RHR, HRV, sleep, steps, zone 2 time vs. age norms. Show what moves it. - Journal: yes/no habits, next-day recovery impact once there are ≥ 3 yes and ≥ 3 no days. - Coach: Apple's on-device Foundation Models with my numbers as context, offline fallback. ## The orb (the signature) - Replace rings with a living orb: ribbons of light inside a glowing volume,