Tare

The problem

Every tracker hands you Mifflin-St Jeor times an activity multiplier and never revisits it. That formula is routinely wrong by 300–400 kcal a day, which is the entire margin someone is trying to manage. The number people actually need — what they burn — is treated as an input nobody can measure.

What I built

• It is measurable, and the arithmetic runs both ways: intake minus expenditure equals the change in body mass. With logged intake and a smoothed weight trend, expenditure is the unknown you solve for. A filter runs over both and adjusts the target weekly.
• The trend, not the scale. Water, salt and glycogen move a reading by a kilo overnight — more than a week of real change. The app acts on the smoothed line and shows the raw readings as faint dots behind it.
• Safety limits that are not adjustable: no target below 1,200 kcal for women or 1,500 for men, no rate faster than ~1% of bodyweight a week, no goal weight below a healthy BMI. When a limit binds, it says so rather than quietly changing the number.
• Honest about its own uncertainty. It says "still learning" for the fortnight before it has enough data, instead of projecting confidence from a formula on day one.
• One outbound request in the entire app — a barcode to Open Food Facts. The 13,348-food database is bundled, so search and logging work in a supermarket basement.