Your learning, observed.

TheLearningCompany is an AI learning companion that watches how you learn, not just what you consume. By observing your study sessions in real-time, it extracts key concepts, identifies knowledge gaps, and helps you retain information through quizzes and flashcards.

What makes this novel: Unlike browser extensions or screen recorders, we provide a fully isolated Docker environment where your study session runs in a private container with zero access to your local files, apps, or notifications. This is not just distraction-free browsing, it's a sandboxed AI-native learning environment where an autonomous agent can observe your screen, take actions on your behalf (clicking, typing, navigating), and generate personalized learning materials. This architecture, combining privacy-first containerization with real-time AI observation and autonomous browser control, has not been implemented before.

System architecture: User connects via NoVNC to an isolated Docker browser, observed by an AI agent powered by Overshoot, OpenRouter, and Browser-Use

We extended the Overshoot JavaScript SDK to support real-time screen sharing via MediaStream, a capability not available in the original implementation. This enables our AI agent to observe your browser session live and extract learning concepts as you study.

Our fork adds native MediaStream support for screen capture, allowing seamless integration with the Chrome DevTools Protocol for real-time vision analysis.

1. Set up environment variables

cp .env.example .env

Add your API keys to .env:

• OPENROUTER_API_KEY: Get from OpenRouter (required for AI assistant)
• VITE_OVERSHOOT_API_KEY: Get from Overshoot (required for AI observation)

2. Start the Docker browser container

cd backend
docker compose -f docker-compose.browser.yaml up

3. Start the backend

cd backend
uv sync
source .venv/bin/activate
python server.py

4. Start the frontend

cd frontend
npm install
npm run dev

Navigate to http://localhost:5173

When you end your private browsing session, the AI presents four learning modes based on the concepts it observed:

Choose how to reinforce what you learned

Flashcards Dynamically generated from your session, exportable to Anki	Quiz Dynamically generated questions to test your understanding
Knowledge Graph Interactive visualization of concept connections	Reflect Chat with AI about what you studied thus far

Built at NexHacks 2026 @ Carnegie Mellon University

• Matheus Kunzler Maldaner - University of Florida
• Aamogh Sawant - Georgia Tech
• Pranav Bhargava - University of Florida
• Autumn Qiu - Carnegie Mellon University