Megalearn AI
A learning game built around hypotheses and experiments
Megalearn is a browser-based learning game for children and parents, built with Three.js and generative AI. Students explain what they think, test the idea in a 3D world, and reflect on the result. The AI responds as a teacher inside that loop instead of simply giving the answer.
Role & Strategy
I co-founded Megalearn and designed and built the system, including agent-based tools for producing game assets and lesson content. The work combines product design, game design, and front-end development in TypeScript and Three.js.
- Text-to-voxel assets: I built a pipeline that turns text prompts into 3D voxel models, reducing the manual work needed to populate each game world.
- Procedural worlds: I designed a modular terrain system that generates environments from parameters such as humidity and temperature.
- Real-code prototypes: I tested the Hypothesis, Experiment, and Conclusion loop in working prototypes, then refined the MVP with teachers and parents.
- Responsive dialogue: Conversation scenes use AI-generated emotion tags to change character expressions and the atmosphere around each exchange.
