Michael Deal
TT_01 / 2026

TrimTab Fit.

TT_01

An iOS weight-management app built around food-quality classification, not calorie counting.

ROLE /
Design · Build · Ship
YEAR /
2026
STATUS /
Live
STACK /
React Native · Expo · Supabase · RevenueCat · Gemini 2.5 Flash
PLATFORMS /
iOS
TrimTab Fit — hero

The problem

Calorie counting works for nobody. The people who can actually stick with it don't need an app; everyone else gets a few weeks in, gives up, and feels worse than when they started. The pattern repeats every January.

What I built

TrimTab is a weight-management app with no calorie counter. You snap a photo of what you're eating and the app sorts the meal into a small set of quality tiers — green, yellow, red — backed by a model that understands ingredients and cooking method, not portion size. Stack a week of greens and the app stays quiet. Stack a week of reds and it shows you the pattern.

The design is deliberately calmer than what the App Store rewards. No streaks, no rings, no nags. The home screen is a single chart and a button to log the next meal.

How I built it

React Native + Expo for the app, Supabase for auth and storage, Gemini 2.5 Flash for classification (cheap, fast, and surprisingly precise on dish photos). RevenueCat for subscriptions, with a free tier that's actually usable. The biggest engineering bet was treating the model output as a structured tier — not a paragraph of text — so the UI can render instantly without waiting on streaming.

What I learned

The hardest part wasn't classification; it was the UI restraint to not over-explain. The first version showed the model's reasoning under every meal. Users ignored it. The second version showed only the tier and a one-line note. Engagement doubled.