Energy
Your calorie target should move
A TDEE from signup week is a guess. Adaptin starts with a Mifflin–St Jeor prior, then blends in a 21-day weight-and-intake regression as data accumulates. Confidence climbs calibrating → low → medium → high. Past days freeze, so a model update never rewrites Monday’s score.
Mifflin–St Jeor is a decent prior. It is not your expenditure. Adaptin keeps that formula as a Bayesian prior that decays as your own data shows up: trend weight (EMA, outlier-damped) plus intake, 21-day linear ΔW, then a blend. You see r², standard error, and “how this was calculated.”
What that buys you
- No crash when one scale reading jumps — trend weight is split from today’s raw number.
- Plan vs trajectory: committed kg/week against the observed slope, with a daily-kcal lever (“eat X less to get back on plan”).
- Activity bonus (opt-in, cutters) can raise today’s target from HealthKit. It never feeds the regression.
- Per-day snapshots freeze. A model update cannot rewrite the score you already earned.
MacroFactor is the other app that takes this seriously. They do not do Arabic food or absorption. We do not claim to be MacroFactor at expenditure math; we claim the target is measured, not preached.