Your phone, your bed, your data — that's it.

Step 1 — Before bed, tap what's true tonight

30 seconds of tagging. The first time you do it, it feels like overhead. By night 5 it's a habit, like setting an alarm.

What might be hurting:

🍷 alcohol  ·  ☕ late caffeine  ·  😰 stress  ·  🍽️ large meal  ·  💊 a new medication  ·  🏃 hard workout

What you're trying:

🛌 side-sleep  ·  📐 elevated head  ·  👃 nasal spray  ·  🪶 anti-snore pillow  ·  😷 CPAP  ·  🧘 wind-down routine

Tagging is optional. The app will produce a normal report whether you tag or not. But the Lifestyle Lab™ Factor Impact analysis only works on tagged nights — that's how I know what your variable was that night.

Step 2 — Phone on the nightstand. Plugged in.

Anywhere within arm's reach. 50–90 cm from your head is what I tested in the validation cohort. Microphone facing you. Do Not Disturb on.

Charge it. The app uses about 8–12% battery per night, but the rest of the night your phone keeps running its normal background stuff. Better to wake up with a full battery than a dead one because you forgot to plug in.

One thing worth saying clearly: the audio doesn't leave your phone. The model runs on your device. Your sleep audio is intimate — it includes you, your partner, conversations after you went to bed, the dog. I built this on-device-only because if I were the user, I wouldn't put my own sleep audio in someone else's cloud.

Step 3 — Sleep how you sleep

That's it. There's nothing to wear. Nothing to wake up and check. The app runs as a foreground service with a wake lock; you can lock your phone normally.

The first 3 nights are baseline. You'll see scores those nights, but trend signals only become reliable after that. If you want to test something specific — a side-sleep belt, an anti-snore pillow, no alcohol on weeknights — plan for 14 nights of tagged data before drawing conclusions.

Step 4 — In the morning, the report is there

You wake up, open the app, and the report is there. No "syncing your data" delay because the data never went anywhere — it's already on your phone.

What's in it:

• BRI for the night. Same per-hour event-rate scale clinicians call AHI. The breathing side.
• SRI for the night. Snoring Rate Index — events per hour, plus a Quiet/Light/Moderate/Heavy band. The snoring side. The two indices are independent — you can have a heavy SRI with a normal BRI, or vice versa.
• Q/L/M/H per-minute texture. Four percentages of your night plus a 15-minute color strip across the timeline so you can see when the heavy minutes actually happened, not just how many.
• Every event, time-stamped. A list of every snore, every breathing pause, every reduced breathing event with start time and duration to the second.
• Audio playback per event. Tap any event, hear what actually happened. Free gives you 3 plays per report. Pro is unlimited. The first 5 events on night one usually change your view of what's been going on.
• Today's tagged factors. What you said before bed gets locked in alongside the report.

Most people listen to maybe 5 events from their first night. Then they stop listening every night. The data does its work even without you pressing play. But on night one, before you trust the numbers, listening once usually changes things.

Step 5 — Doctor-Ready Cadence™ shows up Sunday

This is the Pro-only step, and it's the one I kept thinking about while I was using my own prototype: most sleep apps trap your data inside the app. If you want to bring it to a clinic visit, you screenshot a chart and call it a handout. That's not how clinicians read sleep reports.

So: every Sunday at 7 a.m. local time, Pro subscribers get a PDF in their inbox. 7-day BRI trend. SRI trend with Q/L/M/H texture. Lifestyle Lab™ findings (which factors moved BRI, which moved SRI). A Pro-only Bedside Report page with a Loudness-pattern view for the partner-facing side of things. One PDF that reads the way clinicians actually read sleep reports. On the 1st of each month, a longer monthly summary designed to fit a 15-minute clinic visit.

Print it. Forward it. Bring it. Don't bring it. Up to you. The point is the data leaves the app and lives somewhere you can use.

See a sample weekly briefing PDF →

Why on-device AI (the privacy version, plain)

Your sleep audio is more personal than most people realize. It includes you snoring (private), your partner's breathing (theirs to share, not mine to sell), conversations you had right before falling asleep (no one else's business), and the dog scratching at the door at 2 a.m. (yours).

The AI runs on your phone. Audio is stored encrypted in the app's local storage. The only way audio leaves your phone is if you explicitly tap Export or Share on a specific clip — usually to send to a doctor, occasionally to a partner who didn't believe you.

What we send to our servers: anonymous app analytics (crashes, feature usage), and the metrics needed to render your reports — never the audio itself. Full data inventory at /legal/privacy.

Why we'd rather miss than alarm

Our model is precision-first by design. If we can't be confident an event is real, we don't flag it. The trade-off is real: if your true BRI is 12, we might report 10. But if we say BRI 18, you can trust it's at least 18.

I made that choice because false alarms damage trust. The first time the app wakes you up worried about an event that didn't really happen, you'd start ignoring it. Missed borderline events are recoverable — keep tracking, the trend will surface them. Phantom alarms are not — once you stop trusting the app, it doesn't matter how good the next number is.

For the people who'd rather see every borderline event flagged: a future Pro setting may let you switch to a higher-sensitivity mode. Not yet. The default is what we validated against the lab (n=70+ paired nights, 80% sensitivity / 91% precision on breathing irregularities). See /accuracy for the methodology.

What you do with the data after a few weeks

Most of the actual usefulness shows up around night 14, not night 1. Here's what tends to happen:

• Nights 1-3 — you see the numbers. You're surprised. Maybe alarmed. You play back a few events.
• Nights 4-7 — the trend chart starts showing variation between nights. You notice that wine-on-Tuesday looks different from no-wine-on-Wednesday.
• Nights 8-14 — you're tagging consistently. The variation in your data starts to feel like a signal, not noise.
• Day 15 — open Lifestyle Lab™ Factor Impact. See — for you, not the average user — which factors actually moved your number.
• Beyond — bring the data somewhere it can be useful. The Doctor-Ready PDF is for that. So is, sometimes, just showing your spouse what their concern actually looks like as a number.