Inspiration
We live in a world that is more connected than ever – and yet, many children and teens feel a growing disconnection from knowledge, culture, and diverse perspectives. We wanted to reimagine learning as something alive, playful, and full of voices that make curiosity irresistible.
The basic idea was originally created during the Harvard XR Ideathon in April 2025 by Rolf Kruse, Mehdi Akhondi, Urooj Imdad, Vijay Kagupati, where it received an award for its innovative approach to real-world, embodied learning.
What it does
World Tutor is an interactive learning world where characters "sparks" tell their own stories. Instead of passively consuming information, learners are invited into conversations with the world itself.
Example dialogue (User age 12):
Chester the Chair: “Greetings! I’m Chester, a humble chair. But did you know I have quite a history?”
User: “History? You’re just a chair!”
Chester the Chair: “Ah, but thousands of years ago, only kings and queens had chairs. Everyone else had to sit on the floor!”
User: “Whoa, so sitting on a chair meant you were super important?”
Chester the Chair: “Exactly! A throne is just a very fancy chair. Later, in ancient Egypt and Rome, people started making decorated chairs to show off their power.”
User: “And now everyone has one, even in classrooms.”
Chester the Chair: “Yes – from royal thrones to school desks, I’ve seen it all. Not bad for four legs and a seat, right?”
Through dialogues like this, learners don’t just learn facts – they experience stories that make knowledge personal and memorable.
How we built it
We combined AI-driven character generation, narrative design, and immersive interaction models. By blending knowledge databases with playful personas, World Tutor creates a dynamic ecosystem of voices, each representing different perspectives and expertise.
Teck Stack
- MR with Hand Tracking: Meta Quest SDK 77, URP
- Object Detection: PCA, MS Florence-2, Depth API
- Speech to Text (On Device): Recognissimo
- Chat: OpenAI GPT-4o-mini, OpenAI-Plugin
- Text to Speech: Elevenlabs
- Animated Blendshapes: Blender
Challenges we ran into
- Capturing the right balance between credibility and playfulness in the characters.
- Creating a hybrid of Roboflow models via local inference and the remote MS-Florence 2 model
- Getting kicked from the WiFi
- New Firmware Updates
- Keyboards with "space" and "c" not working
- Testing Speech To Text in a noisy environment
- Tried integrating OpenAI Realtime API but ran into persistent server errors
- Switched to GPT-4.0 mini with ElevenLabs, which added noticeable latency
- Cut delay by playing the first sentence instantly and queueing the rest in the background
Accomplishments that we're proud of
- Building the first prototype as a connected living knowledge world.
- Giving everyday objects surprising, memorable voices.
- Seeing early testers get curious, smile, and dive deeper into topics they hadn’t expected.
- Working as a perfect Team <3
What we learned
That it is possible to raise an idea and create a prototype within a short time with inspiring, skilled and mindful people you just got to know. And that it is rewarding to bring a smile on peoples faces with smart application of high tech.
What's next for World Tutor
Our next step is to expand World Tutor into a fully-fledged platform for AI-powered education. We aim to build a scalable foundation where learners of all ages can explore knowledge through rich, interactive conversations. This includes:
Train custom model with roboflow
Expanding the universe of characters and stories: Introducing new “sparks” with diverse backgrounds, expertise, and cultural perspectives to create a richer, more inclusive learning world.
Adaptive learning experiences: Using AI to tailor dialogues and story paths to each learner’s interests, pace, and knowledge level, ensuring engagement and retention.
Cross-subject integration: Connecting multiple disciplines—history, science, literature, and more—through narratives that reveal the interconnections between ideas.
Interactive challenges and quests: Encouraging curiosity-driven exploration through tasks, puzzles, and collaborative learning experiences embedded in the world.
Finalizing the Implementation of MX Ink Stylus
Seamless platform accessibility: Making World Tutor available across devices, so learners can engage anytime, anywhere, and share discoveries with friends, teachers, or family.
By combining storytelling, AI, and immersive interaction, we aim to transform passive learning into an exploratory adventure. Our vision is to create an educational ecosystem where curiosity is sparked naturally, knowledge is remembered deeply, and learning feels like a conversation with the world itself.
Log in or sign up for Devpost to join the conversation.