Our honesty commitments
Almost every nutrition app promises "accurate AI" without showing numbers. We publish exactly what we know, what we don't, and what we still need to improve.
We publish our real accuracy
Instead of saying "AI-powered" we show the numbers: MAPE, MAE and RMSE of our scanner against an open dataset, refreshed weekly. If a model iteration regresses, you'll see it before we do.
See accuracy →Every medical claim is sourced
We don't say "studies show" without naming the study. Every calorie figure, every macro recommendation, every formula comes from USDA, BEDCA, EFSA, peer-reviewed journals, or WHO. The sources page lists them all with direct links.
See sources →Your data is yours
We don't sell to third parties. We don't use your data to train our model unless you tell us to (explicit opt-in, not pre-checked). You can export everything as JSON or delete your account in one click — and deletion is real, not soft-delete.
See privacy →When we don't know, we say so
Every analysis has a visible confidence level (verified / estimated / approximate) and an honest calorie range ("320–450 kcal" instead of fake precision "385 kcal"). If a plate is ambiguous or lacks a scale reference, we tell you to add a fork to the next photo.
We don't hallucinate food
The scanner applies hard anti-hallucination rules: if the model sees artichokes and anchovies, it does NOT add ham "because they usually go together". Every plate passes through a sceptical reviewer, USDA validation, and a physical container cap. We catch it before we show you.
No engagement bait
We don't game notifications to force your open. No streaks that artificially break, no fake countdowns, no lying "best price now" tags. Health isn't sold with FOMO.
Notice any gap between what we say and what we do? Email us at hola@renzy.app.