The BittsAI Story: Rethinking Learning Tools for the AI Era
•
5 min read
The Core Problem
While every learner is different, most online learning platforms employ the same approach (One Size Fits All). These issues ranges from limited question banks, superficial analytics to distracting user interfaces, and outdated information.
The Assembly Line Approach
Traditional platforms have treated learning like an assembly line for years. Same questions for everyone, random/preset order and identical explanations. They track what you get wrong but never why or how you're thinking through the problems. When they do collect data, there's no feedback loop closing the gap between insight and action. Worse, the constant navigation between pages for analytics, testing, searching, or finding alternative explanations breaks your focus and slows learning.
My NCLEX Wake Up Call
Back in May 2021 during my NCLEX preparation I subscribed to an online learning platform like everyone else. The question banks were solid but fundamentally limited. 3,000 questions sounds impressive until you realize entire topics might have only five questions or none at all. If you need to study something new after passing, you most definitely need another subscription and probably another interface to learn.
The analytics were laughable: percentage correct, bar graphs, maybe category breakdowns. Nothing about how I was thinking through problems, where my reasoning broke down, or what patterns existed in my mistakes.
The Missing Feedback Loop
AI tools like Claude and ChatGPT could explain most concepts brilliantly, but they couldn't track my progress, render interactive question/flashcard UIs, adapt to my weaknesses, or tell me when I was ready to move on. Traditional question banks tracked progress but couldn't generate fresh content or provide deep explanations. I needed both systems linked to each other.
Feedback loop image adapted from Jones, Nelson, & Gerzon (2021)
The Spark
The breaking point came when colleagues asked for help with their certification exams. "You're good at explaining things and you love solving problems. Why not build something that actually works the way we need it to?"
Why not indeed? I already knew how to build web applications. I'd been teaching myself TypeScript and Next.js precisely to create tools that didn't exist yet. In an era where AI could generate almost infinite personalized content, why are we still choosing the traditional learning approach?
The Vision
What if a platform could do both? Generate unlimited questions, track real learning behaviors, adapt in real time, and cost less than a single traditional subscription? That conversation didn't just spark an idea. It became BittsAI.
The Solution! What BittsAI Does Differently
Unlimited, Adaptive Content Generation
BittsAI generates questions on demand, as many as you need, on any topic you're studying. There is no more cycling through the same questions hoping to hit your weak areas. The system creates 8+ question types (multiple choice, multiple response, multiple response select N, numerical free text, highlight, dropdown, true/false) that adapt based on your performance equivalent to Computer Adaptive Testing (CAT) principles.
When you struggle with a concept, the next question builds on that weakness. When you demonstrate mastery, it moves you forward. Every question is unique, contextualized to what you're learning and tailored to how you're performing in real time.
Personalized Analytics
BittsAI tracks what traditional platforms ignore: how you learn, not just if you got it right. Every interaction captures granular data: clicks to correct answer, time spent per question, hint usage, confidence levels, and reasoning patterns. When you hit a personalized threshold (5-25 questions, your choice), the system generates detailed analytics with specific recommendations on what to work on next.
The analytics render in real time within your chat, keeping you in flow instead of breaking focus to navigate separate dashboards.
BittsAI personalized analytics showing detailed performance breakdown with specific concept weaknesses identified (Medication administration, Initial nursing interventions, Managing adverse reactions), time spent metrics, confidence level tracking, and actionable recommendations for targeted review, all rendered inline within the chat interface
Learning in One Place, Without Context Switching
Everything happens in a single conversational interface. Generate questions, get immediate feedback, review explanations, create flashcards, access analytics, search the web for current information without leaving the chat.
If you need to reference a concept mid question, use pin documents for quick access during study sessions. Want to explore an alternative explanation? Branch the conversation and link back to the original thread. The platform eliminates the cognitive overhead of juggling multiple tabs, tools, and interfaces.
Study Mode even takes this further It's a dedicated environment for structured learning with specific, measurable objectives. When you're ready for deep focus, everything you need is already there.
BittsAI unified chat interface showing real time question rendering, conversation messages, calculator and pinned reference document on sidebar demonstrating seamless workflow without tab switching
FSRS Spaced Repetition
Spaced repetition is an evidence-based learning technique that is usually performed with flashcards. Flashcards generated during study sessions automatically sync to a FSRS (Free Spaced Repetition Scheduler) system. Any created flashcard is available for review immediately and at the optimal moment. No manual imports, no separate apps, no friction between learning and retention.
BittsAI flashcard interface with FSRS spaced repetition scheduling showing automatically synced cards from study sessions
Building BittsAI
The Tech Stack Decision
I built BittsAI with tools that solved specific problems, not just what was trendy.
Core Infrastructure:
- Next.js + TypeScript - Server side rendering for speed & type safety
- Supabase - Real time database subscriptions, authentication and storage without managing servers
- Vercel - Deployment, analytics and edge functions for low latency
- Stripe - Clean payment integration and subscription management
AI & Data Layer:
- Vercel AI SDK - SFree open-source library with amazing tools needed to build AI-powered products.
- TanStack Query - Data fetching, caching, and optimistic updates
- Tavily - Web search integration
- Zod - Runtime validation for AI tools and responses and user inputs
User Experience:
- shadcn/ui + Tailwind - Consistent, accessible component library with utility-first styling
- Motion - Smooth animations without performance hits
- Lucide - Lightweight, consistent iconography
- Recharts - Interactive analytics visualizations
Learning Features:
What's Coming Next
Referencing Past Chats
Building conversation memory using RAG (Retrieval Augmented Generation) to enhance learning continuity. When you reference "that study session on renal physiology last week," or "what did I learn yesterday" the system will find it and pull relevant context automatically. No more manually searching through chat history.
Overview of RAG based conversation with memory. Generated using BittsAI platform
Dynamic Memory System
Moving beyond static custom instructions to adaptive memory that learns your preferences, common mistakes, and learning patterns over time. The platform will remember you struggled with a specific concept and proactively adjust future content.
3D Anatomy Integration
I am trying to leverage my clinical background to build interactive 3D models for anatomy learning. Instead of static diagrams, manipulate organs, visualize spatial relationships, and answer questions while exploring structures in 3D. Currently mapping out how to integrate this feature using BioDigital which Powers the World's Understanding of the Human Body.
Sample 3D anatomy sourced from Biodigital showing Temporomandibular joint and paranasal sinuses
Conclusion
BittsAI started with a simple frustration, it has evolved to a full stack platform that generates unlimited adaptive content, tracks actual learning behaviors, and costs less than traditional question banks.
Building it while working as an RN has taught me that the best solutions come from living the problem. I believe that building EdTech platforms need people who understand both learning workflows and the technical implementation
For learners and professionals, education in the AI era should be unlimited, personalized, and affordable explore BittsAI.