A tide, never a deadline
every age · Updated 2026-08-23
This app will never tell you your child has to sleep at 14:20. That is its founding rule, and everything else here is a footnote to it. The screen you look at is water — a tide of sleep pressure rising toward a shore — and the times it implies are wide, soft windows. Never appointments.
Why a precise time would be a lie
Here is what a minute-precise predicted time does inside a home. It stops being information and turns into an appointment. The appointment gets missed — children are not trains — and the miss gets read as a verdict on the parent. I have caught myself doing it with my own numbers: I wrote the code that produced them, I know exactly how soft they are, and I still felt the sting when the day disagreed.
The deeper problem is honesty. My model's error, measured on 671 days of my own child's sleep, is tens of minutes — and that is true of every model in every sleep app, whether they say so or not. The science itself (the two-process model this engine implements) describes windows, not points. So printing "13:47" would be precision theater: dishonest and stressful at the same time.
A model whose own test error is measured in tens of minutes has no business printing a minute-precise appointment.
What you get instead
- The whole screen is the instrument: a waterline that is your child's sleep pressure, drawn from their own history, rising while they are awake and draining while they sleep.
- A wave is the honest shape for this. It builds, it crests, and — this matters — if you miss one, there is always a next wave. That is not a consolation I invented; it is how ultradian sleep biology actually works.
- Exact times exist in the details, where detail belongs. I am not hiding numbers; I am ranking them.
- When a window passes, the app answers in calm amber and recoverable language: carry gently, aim a little earlier next time. No alarm red, no countdown, nothing that ticks.
The part I like most is that none of this is a concession. Rigor and mercy usually pull in opposite directions; here they point at the same design. The window is both the kinder thing to show and the honest reading of the evidence.
The rule was first written down as "a gauge, never a deadline" — the dated origin story is in the notebook. The instrument has since become the tide you see today; the rule hasn't moved an inch.
Where this comes from
- My own dataset: 671 days, 2,181 hand-logged sleep sessions of one child, month 1 to month 22, used to fit and validate the engine. The honest scorecard, losses included, is in the notebook.