The Today screen ends on a line about what is likely next, and a reminder arrives when something looks overdue. This page is what sits behind both. You do not need any of it to use moppet.
Before moppet knows anything about your child, it uses the feeding and sleep cadence published for their age. Which body it follows depends on where the child is, taken from the time zone on their profile:
That changes the numbers, because the guidance itself differs. The AAP describes a newborn feeding roughly every two to three hours. The NHS and the WHO describe responsive feeding, which in practice runs a little more often early on. So a newborn's starting cadence is about two and a half hours under the AAP and a little over two under the NHS and WHO. The two converge as the child gets older, and both land near four hours by six months.
It is keyed to the child, not to you. A grandparent in London and a parent in New York see the same “Next” for the same child. A shared record that told the two of them different things would be worse than no record at all.
Diapers and sleep are not split by country. No health body specifies a materially different diaper cadence, and sleep works the same way everywhere, so those figures are the same wherever you are.
A feed at 2am is not a feed at 2pm, so moppet keeps the two apart. The overnight gap is longer, and it lengthens faster as your child grows. Your child's own day and night patterns are learned separately too, so a good run of nights does not stretch out your afternoon.
A nap is not counted from the last nap. It is counted from the moment your child last woke up, plus the length of time a child that age can comfortably stay awake. That stretch is the wake window, and it widens as they grow: under an hour for a newborn, several hours by the time they are two. It moves gradually, so the window grows by a few minutes a week and never jumps on a birthday.
The ranges come from pediatric sleep practice, and the total-sleep figures behind them are the American Academy of Sleep Medicine consensus, which the AAP endorses. While a sleep timer is running there is no nap prediction at all, because your child is already asleep. Today says so instead.
The published cadence is only the starting point. moppet looks at the last couple of weeks of your own logs, takes the typical gap for the time of day, and blends it with the published figure. Early on the published figure dominates, because it is all there is. Within a few days of ordinary logging your child's own rhythm carries most of the weight.
It uses the typical gap and not the average, so one four-hour Sunday afternoon does not drag the whole estimate out.
A feed that happens but never gets logged does not leave a hole in the record. It makes two gaps look like one long one, and the error only runs one way: missing logs make your child look slower than they are, never faster. Left alone, an app would read a patchy day as a placid baby and push the next feed hours out.
So moppet weighs its own evidence. A gap far longer than the published cadence for that age is more likely to be a missed log than a real one, and counts for less. The less of the day moppet can see, the more it leans on published guidance, and the window cannot drift far from it either way. If you log normally, none of this changes anything you see. It is a correction for missing information.
Underneath everything there is a floor, so a long unlogged stretch cannot push a feed reminder out of the day. You can set your own in Settings › Reminders. When the floor and the prediction disagree, the earlier one wins.
moppet also stops when a reminder would stop being useful:
The prediction runs on your own logged data, on our servers. Nothing is sent to an outside service to work it out. Notifications are the exception: to reach your phone a reminder passes through Expo and Apple in the United States, carrying your child's first name, and for a medication reminder the medicine and the dose. The alert when another caregiver stops a timer takes that same path, carrying that caregiver's name and the span the timer covered.
Treat the result as a likely window. It is a nudge about your record rather than an instruction about your child.