Let's Do Math
Release readyA study app that counts only the time actually spent focused
iOS app (student · parent · admin) · 2026
Overview
A study management app designed around two kinds of user: a middle-school student and their parent. The student works through formula and concept cards by unit, runs a timer that measures focus, sits a 20-question maths exam assembled automatically each day, and reviews what they got wrong. The parent sees weekly focus minutes, score trends, and streaks on a dashboard.
Curriculum material for grades 1–3 across 31 units was seeded into the database as the content base, and AI-generated questions only reach students after an admin has reviewed them.
Engineering highlights
Making the measurement trustworthy
A three-state model — focused, idle, away — is derived from app state and the last interaction timestamp, with periodic heartbeats and an offline queue. Crucially, the server recomputes session time from those heartbeats, so the client cannot inflate the number.
An AI pipeline split by cost and risk
Expensive question generation runs as a local batch with structured output, answer verification, and de-duplication, landing in a review queue. The daily exam is then assembled in plain SQL from unseen approved items by difficulty. No per-student token spend each day, and an exam still appears even if generation fails.
Scripts that hold the data quality
Around thirty seeding and QA scripts — content ingest, classification, de-duplication, review-report export, production sync — keep the question bank clean.
Motivation mechanics
Streaks, points, and badges; a wrong-answer notebook for re-drilling; custom exams where you pick scope and question count; and a unit-relation graph that links related formulas.