🌟 Inspiration
I’ve always been fascinated (and frustrated) by how inaccessible cutting-edge research can be. Most of it is locked behind paywalls or buried in dense, jargon-heavy papers. I wanted to build something that explains research, not just finds it—a tool that could genuinely help anyone understand frontier science, no matter their background.
I also wanted to challenge static tools like NotebookLM by creating something more dynamic and immersive—my own personal “nemesis” to it.
💡 What it does
Feynman turns advanced research papers into narrated podcasts with transcripts and interactive, 3Blue1Brown-style visual explainers.
Unlike NotebookLM, which mainly summarizes static text, Feynman actively searches, analyzes, and contextualizes relevant papers to produce rich, understandable, and interactive content. Users can listen to podcasts while following animations that break down complex ideas step by step.
🛠️ How I built it
I combined:
- Advanced NLP pipelines to summarize and rewrite papers in natural, conversational audio scripts
- Semantic search over research archives to find the most relevant papers
- A custom animation engine to generate step-by-step interactive visual explanations
I also built a narrated podcast player with transcripts so users can learn through audio while engaging with the visuals.
🧗♀️ Challenges I ran into
- Translating dense, highly technical research into natural, accessible language without losing meaning
- Designing visualizations that simplify concepts without oversimplifying or distorting them
- Ensuring the AI surfaces relevant, high-quality papers for the topic at hand
🏆 Accomplishments that I'm proud of
- Successfully transforming static, academic papers into engaging, understandable podcasts with visuals
- Creating interactive visual explainers that mimic the clarity of 3Blue1Brown
- Building a truly individual project that can help learners, educators, and researchers anywhere in the world
📚 What I learned
I learned that research accessibility is about far more than just finding the right paper—it’s about making it understandable.
Balancing accuracy, clarity, and engagement in AI-generated explanations is incredibly challenging but also rewarding. I also learned how hard it is to design good educational visuals that truly aid learning.
🚀 What's next for Feynman: The Ultimate Nemesis to NotebookLM
I want to keep pushing this vision forward by:
- Integrating Image and video models for more content rich learning.
- Improving the animation system to support even richer, more interactive visual explanations
- Building a collaborative platform where researchers and educators can share and remix explainers
- Making advanced knowledge as easy to learn as listening to your favorite podcast
- Adding stripe and payment tiers.
My ultimate goal is to help billions of minds access and understand frontier research—no matter where they are in the world.
Built With
- arvix
- bolt
- supabase
- typescript
Log in or sign up for Devpost to join the conversation.