Claude-Yap

Learning that yaps back 😎

Comment

Inspiration

We’ve observed the rise of short attention spans and passiveness amongst students, and at the same time the traditional education system has not evolved to effectively address these needs. While the cycle of lectures, sections, and exams might work for some, the rise of AI both contests this model and presents an opportunity for new educational solutions. We were also inspired by gaps in the current edtech space. While tools might be engaging, gamification doesn’t always translate to true understanding - We wanted to ask: in this new landscape, how can education still be effective, and can it still be engaging and fun?

What it does

This question led to Claude-Yap, a tool that transforms academic materials, like research papers, lecture slides, and textbooks, into personalized podcasts. But it’s more than just a summary. Users can choose their preferred format: a debate mode, where speakers challenge each other's views, or a student-teacher mode, where the podcast speakers roleplay being a learner vs. a teacher where learners ask questions and the teacher guides the conversation. This structure makes learning feel more like eavesdropping on a really smart, engaging conversation, rather than just passively consuming a text. One of our final favorite features allows students to speak directly with an AI-powered avatar about the topic they just heard, encouraging them to explain concepts in their own words. That kind of active recall is proven to improve retention, and it's a powerful antidote to the cut-and-paste culture that AI tools can sometimes enable.

How we built it

  • Frontend: React with React Router, Tailwind CSS for styling, Framer Motion for animations
  • Backend: Python with Flask, Claude API for AI content generation, Cartesia API for high-quality text-to-speech, PDF processing libraries, Tavus for replica.

Challenges we ran into

  • Improving how long it takes to generate the podcast
  • Prompt Engineering

Accomplishments that we're proud of

  • Debate mode and student-teacher mode for the podcast
  • Active recall session with replica

What we learned

  • We learned how to use Claude, prompting tool
  • Integrating AI in products and navigating the challenges that come with this like prompt engineering.
  • Feature prioritization
  • Teamwork

What's next for Claude-Yap

Looking ahead, we’re excited to make Claude-Yap even more powerful:

  • User profiles: Users create a profile detailing their interests that Claude-Yap analyzes. Imagine a podcast that explains CRISPR gene editing using metaphors from birdwatching because that’s a student’s hobby. That’s the kind of learning that sticks.
  • Spaced Repetition: Claude-Yap will prompt students to revisit and discuss past topics on a personalized schedule. By combining active discussion with time-based reinforcement, we're building a system that doesn't just help students learn content, but retain it long-term.
  • Deep Research: Currently, users upload PDFs and get a podcast in return. But with deep research, users can simply ask Claude-Yap about a topic—“Explain the water crisis in the American Southwest,” or “Give me a breakdown of neural lace technology”—and we’ll automatically gather relevant sources and build an episode from scratch. It's not just a tool for reviewing what you already have—it's a launchpad for discovering what you don’t know yet.

Built With

Share this project:

Updates