Apple Watch (Series 12 / Ultra 4)

Apple's 2026 generation — Series 12 ($399) and Ultra 4 ($799) — built on the S11 chip with a redesigned optical + electrical HR sensor (repositioned photodiodes, larger green LEDs, new ECG electrodes). The generational headline is sampling frequency, not raw accuracy: background HR goes from ~every 5 min to every 5 s ("60x more frequent"), HRV to as often as every 5 min (24x), reversing Apple's historical "accuracy over availability" stance. On top: new ML step counting (claimed "most accurate of any smartwatch," 1,000-step video-validated), a dynamic readiness-score, health-age, camera-based at-home movement/VO2max evaluations (on-device, AirPods Pro 3 compatible), a $119 Quest Diagnostics 50-biomarker panel bookable in the Health app (US-first), and audio intelligence (Life Rewind, Siri recap, auto-Shazam — later-2026 beta, EU excluded at launch). Battery: Series 12 24h/10h-workout; Ultra 4 50h normal, 84h low-power, up to 45h max-extended workout with 1 s GPS retained.

Context from prior videos: Apple Watch was already first-tier in rob-ter-horst's sleep-tracking data and among the best wrist HR trackers before this sensor overhaul. All headline claims are pending his reference-device-validation testing (both models ordered).

dc-rainmaker's same-day Apple Park hands-on added: sensor specifics (ECG electrode area +24% in a "petal" layout, eight windows with double photodiodes replacing four dots, all-new LED array), sdnn-vs-rmssd both now exposed per-feature, HRV baseline needs 7 days of wear (≥4h sleep/night), readiness score is new-watch-only, the health app redesign (later 2026) needs an Apple Intelligence iPhone but not a new watch, max-extended workout mode records full-fidelity 1s GPS + HR while displaying a 5-min rolling average, audio runs in the S11's secure Exclave (ambient-audio-privacy), display is 60% stronger, and both ship September 18.

His Ultra 4 hands-on supplied the missing mechanism: the LED array is regrouped into clusters of four green, four red, and four infrared diodes, which moves 24/7 background sensing off infrared and onto the green LEDs — see green-led-background-sensing. Also Ultra-specific: 50h normal battery (audio intelligence included in the claim), Extended mode taking GPS workouts 20h→25h, Max Extended reaching ~48h on single-band GNSS while keeping 1-second GPS (full-fidelity-recording-reduced-display), both modes moved to the sport profile page; daytime HRV first appears 2 hours after wake; RMSSD and SDNN are both exposed to developers; new translucent ocean bands; orders open the Saturday before the September 18 ship date.

Unresolved: the max-extended workout figure reads 45h in Rob's analysis and 48h here.

In-depth review verdict (DC Rainmaker, 2026-09-17, vs Polar H10 across six sports): the sensor is "very accurate thus far" — genuinely strong plots even for road-cycling vibration and rowing — but with quality-gated recording gaps of 11-40s that Apple's study never scored (quality-gated-data-omission); cold-weather testing still pending. Workout display refreshes HR only every 5s regardless of sampling. Readiness is realistic but launch-buggy (scores missing for hours, two watches out of sync). Daytime HRV needs 7 clean days per hourly time slice — never completed on an irregular schedule. GPS: GLONASS removed (no impact), Ultra 4 multiband > Series 12, Garmin Fenix 9 still better. The 50h battery claim held in testing. Upgrade advice: strong from Ultra 1 (excluded from watchOS 27), skip from Ultra 3 unless readiness/HRV/audio matter.

Sources: Quantified Scientist analysis, DC Rainmaker Series 12 hands-on, DC Rainmaker Ultra 4 hands-on