EduRealm AI

Script Introduction
Learning Page
AI Summary Report
AI Script Creation
AI Script Creation

Inspiration

In many K–12 humanities classrooms, learning is still dominated by memorization and passive listening. Students rarely engage in reasoning, discussion, or interpretation, even though these abilities are central to modern curriculum goals.

We were inspired by scripted role-playing games, where participants actively reason, collaborate, and view problems from different perspectives. We believed this structure could transform learning into a context-driven, reasoning-centered experience, especially when combined with AI support.

This led us to build EduRealm: an AI-assisted educational script-based learning platform.

What it does

EduRealm turns a single lesson into a script-based learning experience.

Each script corresponds to one class session and is aligned with subject, grade, and textbook.
Students learn through role-play, discussion, and task-driven reasoning, rather than lectures.
An AI learning companion supports the process by guiding discussion, organizing ideas, and recording reasoning paths—without giving direct answers.
After completion, students receive an AI-generated learning report combining task performance, discussion participation, and assessment results.

EduRealm also enables teacher and user co-creation, allowing educators to design and share curriculum-aligned scripts.

How we built it

Core technologies

Google AI Studio as the main development and prototyping environment
Gemini 3 APIs for interaction, reasoning support, and content generation
Supabase for backend data storage and session management

Gemini 3 usage (key focus)

gemini-3-flash-preview
- Powers the AI learning companion
- Provides fast, real-time responses during student discussion
- Summarizes conversations, organizes reasoning, and generates learning reports
gemini-3-pro-preview and other Gemini models (user-selectable)
- Used in AI-assisted script creation
- Generates teaching objectives, outlines, roles, scenes, and assessments
- Supports step-by-step generation rather than one-click output

Design decisions

AI is positioned as a learning companion, not a teacher
Students remain the main agents of reasoning and decision-making
Teachers retain full control over educational goals and content

Using Google AI Studio’s natural-language-driven workflow, we rapidly implemented frontend logic, AI behavior, and backend integration despite limited prior engineering experience.

Challenges we ran into

Limited full-stack development experience
Difficulty expressing complex system requirements to AI
Coordinating AI logic with frontend and database design

We addressed these challenges by:

Letting Gemini act as a professional engineer to self-check architecture and logic
Iteratively refining prompts with references and examples
Using AI collaboratively across Google AI Studio and Supabase to validate system design

Accomplishments that we're proud of

Preserved student and teacher autonomy with clear AI boundaries
Designed a novel learning model combining scripted gameplay and AI guidance
Built a complete, functional platform using Gemini 3 APIs
Enabled co-creation, where users are both learners and content creators

What we learned

How to design AI systems that support thinking instead of replacing it
How different Gemini models serve distinct roles:
- fast interaction vs. structured content generation
How Google AI Studio significantly lowers the barrier to building complex AI-driven applications
How AI-assisted debugging accelerates learning for new developers