A nutrition app should show its work.

I built Svelio because I did not trust the answers I was getting from the apps already on the market. The problem was not that they were occasionally off. It was that I could not tell when. One app would give a food one number and another would give it a different one, and the same app would give the same food two different answers on two different days. An estimate you cannot check is not a measurement, and a number you cannot repeat is not something to make a health decision on. That is the whole design brief. Read the label wherever a label exists. Grade against a published standard rather than an opinion. Show which food, what portion, and where every figure came from. Say plainly when something is an estimate, and say nothing at all rather than fill a gap with a number that merely looks right. Here is what that costs in practice. A crowded plate hides food behind other food, so Svelio lets you shoot the same meal from up to three angles and counts anything visible in any one of them. What it will not do is tell you that makes the number right. We tested whether a second angle improves the portion weight itself, and on weighed plates it did not. So the feature stays, because seeing a piece you would otherwise have missed is worth something, and the claim does not, because we could not support it. That is the trade this whole product makes. I would rather ship an honest limitation than a confident number.