NeuronBook

NeuronBook Logo
Start / Launch Page (Before Signed in)
Info Launch Page
Get Started / Sign up
Home
Content Library
Neural Network

Inspiration

Studying from PDFs is still one of the most common ways students learn, yet the experience is largely passive. Highlighting and rereading often create an illusion of understanding, but retention and true mastery remain low. As students ourselves, we noticed that the moments we learned best were when we were actively questioned and forced to explain concepts in our own words.

NeuronBook was inspired by the Socratic method and learning science principles such as active recall and spaced repetition. We wanted to transform static PDFs into interactive learning environments that challenge users to think, reflect, and build understanding over time instead of passively consuming information.

What it does

NeuronBook turns passive PDF reading into active mastery.

Users upload a PDF and read it in an interactive workspace. As they progress through the document, NeuronBook generates AI-powered Socratic questions tied to the current page to test understanding in real time. Users can answer questions, rate their difficulty, flag them for later review, or skip them.

Each interaction contributes to a Neural Trace — a visual knowledge graph where concepts are represented as nodes and relationships as connections. Concepts strengthen, fade, or get pruned over time based on user performance and difficulty ratings. NeuronBook also enriches learning by pulling in external research and explanations when users struggle, creating a personalized and evolving map of understanding.

How we built it

The frontend is built with Next.js, TypeScript, TailwindCSS, and shadcn/ui, providing a clean and responsive interface. Clerk handles authentication with Google and email sign-in.

We used the Foxit Web SDK to render and interact with PDFs directly in the browser. A Python Flask backend acts as the central API layer, coordinating AI logic, external APIs, and data storage.

AI workflows are orchestrated using LangChain, generating summaries, Socratic questions, and concept extraction. Sanity serves as our database and content layer, storing user data, questions, answers, embeddings, and the neural knowledge graph structure. The Neural Trace is visualized using react-force-graph. When deeper explanations are needed, You.com’s API is used to augment concepts with relevant research and sources.

Challenges we ran into

One major challenge was coordinating real-time interactions between the PDF viewer, AI generation, and the neural graph without overwhelming the user. Designing meaningful Socratic questions that adapt to context and difficulty was also non-trivial.

Another challenge was modeling learning as a graph in a way that was both visually intuitive and technically scalable. We had to carefully balance performance, data structure design, and clarity to ensure the Neural Trace felt insightful rather than cluttered.

Accomplishments that we're proud of

We are first proud of our team effort and teamwork; we were able to come up and create the app effectively. We are also proud of everyone's shared efforts and how everyone played their roles and finished their objectives.

What we learned

We learned how initiatives and teamwork can be shared online whilst everyone is at long distances from each other. We learned everyone's skills, like frontend skills and developing backend, with no experience in either, and requesting API for connections and multiple languages, like advancing our skills with NextJS, Typescript, Figma, etc.

What's next for NeuronBook

We plan to build and make this a startup, as everyone has agreed on how great and easy the idea is to implement to the population and how we can inspire more individuals in reading and engagement with documents, pdfs, books, etc. We plan to put maximum effort and teamwork into implementing this as a startup.