why I built MNEMO
A neuroscience-inspired learning platform I built so studying would actually stick — quizzes, an AI assistant, and a dashboard that reacts to how you learn.
Most study apps I tried felt like a content dump with a quiz bolted on. You watch something, you click through a few questions, you forget it by Thursday. I wanted something that treated memory as the actual product, not a side effect.
That's why I built MNEMO.
It's a full-stack learning platform — Next.js App Router, TypeScript, Tailwind, Supabase for auth and real-time data, deployed on Vercel. Live at mnemo-brain.vercel.app. The idea is simple: you sign in, you take quizzes, an AI assistant sits next to the material, and the dashboard tracks what you're actually retaining instead of what you "completed."
What it does
You get a dashboard with progress and recommended topics. You start a quiz, get immediate feedback, retry the weak spots. When a concept doesn't land, the assistant explains it in context instead of sending you to a generic FAQ. Progress syncs in real time, so the UI doesn't feel like a static course site.
Under the hood the quiz state has to stay clean — scoring, retries, and "where did I leave off" without the component tree turning into soup. That was the first hard part. The second was wiring Supabase realtime without the dashboard flickering every time a row updated. Optimistic UI plus reconnection handling sounds boring until you skip it and the whole page feels cheap.
What was hard
State in a quiz is deceptively messy. One wrong model and "retry this question" fights with "submit the whole set." I ended up treating the quiz as a small state machine instead of a pile of useState flags.
Auth looks easy until you care about session restore on mobile. Supabase Auth got me there faster than rolling my own, which was the point — I wanted to spend time on the learning loop, not OAuth edge cases.
What's next
Spaced repetition is the obvious missing piece. If MNEMO is about memory, the schedule should be the engine, not a chart. After that: quizzes generated from uploaded notes, and a path that works offline. The SaaS shell is there. The science still has room to go deeper.