I built InfoDiet because I noticed that screen time tools measure duration but not content quality. Spending an hour on Coursera feels completely different from an hour on Twitter, but every existing tool treats them identically.

InfoDiet runs as a Chrome MV3 service worker that categorizes every site you visit into one of 8 content categories and computes a weekly Information Quality Score — your ratio of educational vs entertainment content.

The technically interesting part was building accurate background time tracking under MV3 constraints. Chrome's service workers are killed after ~30s of inactivity, wiping any module-level state. I solved this using Chrome's session storage API to persist active tab state across worker restarts, with a 1-minute alarm to resync against the actual active tab when the worker wakes up.

Stack: React + TypeScript extension, Next.js backend on Vercel, MongoDB Atlas, Stripe billing.

Happy to answer technical questions about the MV3 architecture or the category classification approach.

https://www.getinfodiet.app/ - 7 day free trial, no credit card required.