WhatDo?!

Every cooking decision is the same broken loop: open the fridge, stare, close the fridge, open the app, get told to make something that needs three things you don’t have, go to the store, come home, find the thing you thought you were out of, repeat.

Recipe apps assume a fantasy pantry. Inventory apps assume you’re tracking boxes, not food that rots. Meal planners assume you’ll cook what you scheduled even when the kids are melting down and the chicken is still frozen. None of them meet you where you actually live — in a real kitchen, on a weeknight, with the clock running.

WhatDo?! was born from that gap. A kitchen inventory that knows how food ages, paired with a recipe engine that only suggests things you can actually make right now.

WhatDo is a local-first app with a cloud backbone. Every device keeps its own SwiftData store so the app works offline and launches instantly. A custom delta-log sync engine reconciles changes across devices and households without requiring CloudKit’s assumptions about ownership.

The AI layer runs on Cloudflare Workers, which gives us low-latency inference, tight quota control per user, and the ability to adapt suggestions over time without shipping a new app build. Recipe generation is grounded in the user’s real pantry so the model can’t hallucinate ingredients that aren’t there.