MasteryMint

Inspiration

As a passionate self-learner, I’ve always been driven by one question: "How can I learn deeply, faster, and retain it longer?"

In exploring this, I discovered research-backed learning techniques like Spaced Repetition, Active Recall, Dual Coding, and Elaborative Interrogation. But no app brought all these together in one seamless experience — especially one powered by AI. So I built MasteryMint to fill that gap.

What it does

MasteryMint is an AI-powered learning assistant that helps users understand, retain, and master any concept or document using cognitive science.

Users can:

Ask any question or upload a text document
Get simplified explanations, summaries, visual diagrams, and audio lessons
Use flashcards and quizzes generated automatically
Track their progress via a Knowledge Tracker
Learn using proven methods like spaced repetition, chunking, and dual coding

How I built it

Frontend: Designed with Subframe, implemented with Bolt.new
Backend: AI responses via DeepSeek AI on OpenRouter, Supabase for database.
Voice Narration: Integrated with ElevenLabs
Diagrams: Generated using Mermaid.js, dynamically rendered and scaled
Progress Tracking: Visualized user learning stats with graphs
Deployment: Used Netlify for deploying, Entri for domain name

I built everything from scratch, alone, using bolt.new infrastructure.

Challenges I ran into

Prompt engineering for pedagogically structured AI responses
Rendering and scaling Mermaid diagrams properly across devices
Hiding error icons (Mermaid bomb) during failed renders
Handling irrelevant or malformed user input gracefully
Managing token limitations and errors from AI providers

Accomplishments that I'm proud of

Built a full-stack SPA learning app solo in under a week
Seamlessly combined multiple learning strategies into one flow
Designed a clean, UX-focused interface with Subframe
Integrated multiple AI APIs and tools efficiently
Delivered a unique learning product that truly empowers understanding

📚 What I learned

Educational UX requires more than good design — it needs brain-aware structure
Working with AI models is as much art as science, especially prompt tuning
Tools like Bolt.new can dramatically accelerate solo development and vibe coding
Cognitive science is a treasure trove for improving human learning