Path to MVP
The research changed the product — for the better. Here's what we can honestly ship, what must be proven on a real wrist first, and what we will never claim.
The promise, after the evidence
What Smart Wake claims
- Times your nap and night from your detected sleep onset
- Guarantees you're up by your deadline — three-layer alarm stack
- Biases every timing error early, never late
- Does its best to avoid waking you from deep sleep
- Shows you what it detected, every session — trust through visibility
What it will never claim
- "Wakes you in light sleep" — no wrist device can promise stage-perfect waking; current tech misses deep sleep a third of the time or more
- "Never groggy again" — grogginess has more causes than stage
- Sleep scores, insights, coaching — it's an alarm, not a tracker
The reframe that survived adversarial review: the provable, differentiating feature is nap and wake length anchored to real onset. Deep-sleep avoidance is behavior we engineer and measure — it becomes a claim only when our own on-wrist validation earns it.
Five gates before we commit code to the nap architecture
Each is a cheap on-device experiment. Any failure changes the product rather than degrading it silently.
- Can a wrist haptic actually wake a sleeper? Wrist down, watch locked, Sleep Focus on, silent mode on. If haptics can't wake reliably, nap mode is cut — not shipped weak.
- Does live detection survive the battery budget? The OS silently suspends apps that overspend background CPU. Our onset model must run a full session without tripping it.
- How much heart rate does the night window really get? Design assumes the pessimistic answer (motion-first); the test tells us how much better we can do.
- How accurate is our onset detection, measured on real wrists? Nobody — no lab, no competitor — has published this number for a live system. Producing it is the MVP's core deliverable and the moat.
- Reproduce the single-source platform behaviors (battery figures, background alarm quirks) before trusting them.
The MVP cut
| In | Out (for now) |
|---|---|
|
Power nap — onset-anchored, deadline-capped Night — recurring "up by" + 30-min wake hunt Deadline stack — watch window + nap haptics + phone backstop Learning — your averages in 3–10 nights, honest about its limits Session receipt — "fell asleep 13:42 · slept 21 min · woke on target" |
Cycle nap (needs validated onset first — it layers cleanly on top) Auto-start without a tap (roadmap) Any dashboard, score, or trend Android/Wear OS (designed-for, not built — see below) |
What we're honest about
- Cycles resist prediction. Even a perfectly learned personal average still misses tonight's cycle boundaries by ~18 minutes on average — which is why the product hunts in windows instead of promising cycle-perfect math.
- The market is unproven. Every app circling this niche has a handful of ratings. We build it free because the owner wants it to exist — success is it working, honestly, for real people.
- Apple is closing in. Nap detection, sleep scores, better staging — every year. Our moat is the one thing platform marketing never ships: measured, published accuracy.
Android is designed-in, not bolted-on
The brain — onset detection, wake optimizer, learning — is one portable core that consumes timestamped sensor streams and outputs wake decisions. watchOS is shell #1. Wear OS is verified feasible (an existing app proves the full loop works there), with a lower sensor floor the core is built to tolerate: batched data, missing signals, per-device capability tiers. When the Watch version earns its accuracy numbers, the port is an engineering task, not a rethink.
The bar, in one sentence: every feature either works fully — measured, on a real wrist, with its failure modes handled loudly — or it isn't offered.