Features · Baku
Features

What Baku measures —
and what it does with it.

Seven screens, one model. Everything below is calibrated to your own history, and everything below is either measured or labelled as an estimate.

Today

One screen, one orbit.

Your whole day on a 24-hour dial: last night’s sleep, every session, elevated-stress windows, and the readiness fingerprint in the centre. The bright arc is the day so far; the marker ahead of it is tonight’s lights-out target.

BED 23:12 WAKE 07:29 DIP HRV SLP RHR LOAD SpO₂ 81 READINESS
The dialSleep, training and stress as arcs on a 24-hour orbit, with bed, wake and afternoon-dip stations.
The fingerprintFive readiness components as a polygon against your own baseline — a pattern, not a single number.
Tonight’s planA lights-out target derived from your debt and your typical wake time, not a fixed bedtime.
Home-screen widgetThe dial and the day on your home screen, without opening the app.
Sleep

What the night actually did.

Duration is the least interesting thing about a night. Baku reads architecture, timing, regularity and the autonomic recovery underneath it.

82 SLEEP SCORE 7H 32M · 91% EFFICIENT DEEP 18%REM 21%LIGHT 61% 81 MIN95 MIN276 MIN
ArchitectureDeep, REM, light and awake — with the caveat that stage percentages are device-specific, so they are never trended across a switch.
EfficiencyTime asleep against time in bed, using awake time that was actually measured rather than assumed.
Timing & regularityYour chronotype, and how far each night drifts from it. Regularity predicts more than duration does.
Naps, separatedAn afternoon nap is not a short night. Baku splits them so neither corrupts the other.
Overnight HRVParasympathetic recovery through the night, read against your own distribution.
Respiratory rateA slow-moving baseline whose deviation is among the earliest signals there is.
Readiness

A pattern, not a verdict.

Five components, each scored against your own distribution and drawn as a shape. You learn to read the shape faster than you would read five numbers.

HRVSLP RHRLOAD SpO₂
Your own baselineEvery component is read against your distribution, so “normal” means normal for you — not normal for a study population.
Nothing is inventedA component whose input was not measured is dropped and the others reweighted. It never contributes a zero.
The weakest vertexThe shape shows which system is holding the day back — something a single number structurally cannot tell you.
Training

Load you can act on.

Sessions, zones, and the balance between what you have just done and what you are conditioned for.

1.04 UNDEROPTIMALHIGHSPIKE 0.81.31.5 ACUTE LOAD AGAINST CHRONIC
Load balanceAcute against chronic, so a hard week reads as a hard week and a taper reads as a taper.
Heart-rate zonesAnchored to your own maximum — measured from what you have actually done, or stated by you.
Session detailEvery session with its heart-rate trace, time in zone, and what it cost you.
MonotonyTraining that never varies is its own risk. Baku flags it.
Longevity

Biological age, with a range.

A composite of VO₂max, HRV, resting heart rate and sleep — reported with the uncertainty it actually carries, and split into drivers so you know which one to work on.

29.4 YOU, 34 26.7 32.1 VO₂MAXHRVRESTING HRSLEEP −6.1−4.8−1.9−4.4
Biological ageA number and a plausible range. The range narrows as data arrives instead of feigning precision.
The four driversEach expressed as an age equivalent, so you can see which is holding you back.
Pace of agingWhether the gap is widening or closing, over months rather than days.
VO₂maxCardiorespiratory fitness estimated from what you actually did, with a validity window.
Zone 2 minutesAgainst the weekly guideline, and against your own target.
Early signalsRespiratory rate, resting HR and HRV moving together — the pattern that precedes feeling unwell.
Trends

Is the line going up?

Daily numbers are noise at this range. Baku combines six pillars — recovery, sleep, fitness, rhythm, activity and calm — into a single vitality line, then shows you which of them moved it.

VITALITY · LAST 90 DAYS 57 81 WHAT MOVED IT ▲RECOVERY▲SLEEP●FITNESS▲RHYTHM▼ACTIVITY▲CALM
The vitality lineOne composite over 30, 90 or 365 days. The same number the Today constellation draws, followed over time.
The six pillarsRecovery, Sleep, Fitness, Rhythm, Activity and Calm — each with its own contribution, so a falling line has an address.
Fitness is held, and says soVO₂max does not change nightly. It stays at your last real reading and is marked as held rather than re-estimated to look busy.
Six raw metricsHRV in ms, resting HR in bpm, sleep duration in hours, deep and REM in minutes, SpO₂ as a percentage. Tap any for its full chart.
Weekly averages & bestFour-week rolling means and your personal best on each metric, so a good night has something to be good against.
Your correlationsWhich of your metrics actually move together — computed from your data, not borrowed from a study of somebody else.
Experiments

n = 1, done properly.

Magnesium, a caffeine cut-off, a cold shower, a screen curfew. Name the change, pick what it should move, and Baku runs the before-and-after — including telling you when the run was too short to have proved anything.

MAGNESIUM BEFORE BED 200MG, 45 MIN BEFORE LIGHTS OUT NIGHT 14 / 21 SIGNAL SO FAR DEEP SLEEP +12 MIN REM SLEEP 7 MORE NIGHTS HRV NO CHANGE NO EFFECT
Describe the change“Magnesium before bed”, and what you actually did: 200 mg, 45 minutes before lights out. Your words, so the record still makes sense in six months.
Choose what it should movePick from your tracked metrics. Each one shows its minimum run length before it can give an answer at all.
Warned before you startChoose 7 days for a metric that needs 21 and Baku says so up front — rather than letting you finish and calling it a null result.
Signal so farLive interim numbers and a night count: 14 of 21. You can see whether it is worth continuing without waiting for the end.
One at a timeRun two overlapping interventions and it warns you they will confound each other, because a result you cannot attribute is not a result.
“Not enough yet” is a resultAn effect that has not cleared its bar is reported as unresolved, with the nights still needed — never rounded up into a verdict.
Integrations

Whatever you already wear.

Baku reads Apple Health, so an Apple Watch, a Garmin, or most other trackers already qualify. Oura connects directly. Wear more than one and each is handled on its own terms.

Apple Watch
Oura
Garmin, via Apple Health
Anything else that writes to Apple Health

Wear two devices and Baku knows they disagree — a watch and a ring measure the same night differently, so scores are corrected per device rather than pooled into one misleading average. One caveat worth stating: Garmin does not publish HRV to Apple Health, so on a Garmin that term is dropped and the rest reweighted rather than filled in with a guess. An Apple Watch does publish it, and Oura connects directly.

Standards

Every number obeys the same rules.

This is the reason to trust the rest of this page.

The rules every number here obeys. A readiness score appears once Baku has seven nights of your sleep behind it. Trends are drawn within a single device’s era, so new hardware never masquerades as a change in you. And every value is computed from what was actually recorded — a night that was not measured is not a bad night.

Each estimate carries the range it genuinely warrants, and that range tightens as your record grows. A number that visibly sharpens is the instrument working.

Privacy

It stays on your phone.

Your health data is never sold or shared, and the daily written summary is produced by a model running on the iPhone itself.

On-device summariesApple’s on-device model writes the daily read. Nothing leaves the phone to produce it.
Export everythingOne tap, the whole record, yours to keep.
Delete everythingAlso one tap, and it means it.