Food tracking dies on data entry. Every traditional app wants taps, searches, and portion menus for a meal that took four minutes to eat, so most people quit inside a week. I was most people.
The apps that actually fixed it did it with a camera, and they price it like a product: a few dollars a month, forever, to send a photo to a model. That is a fair business and a bad deal for one person who already has an API key. So the build-versus-buy math was short.
A camera-first web app. A snap of a meal, a nutrition label, or a grocery receipt gets parsed by a vision model into calories and macros, logged against daily goals, and rolled into a dashboard of patterns over time.
It runs on Vercel behind an app secret, so the link is public and the data is mine alone. Running cost is whatever the photos cost, which is less than a month of the subscription it replaced.