Twin Logo
Chrome Extention
Sign up
Texting Twin
Twin Video Scraping

Twin

Inspiration

As a group of students, we understand the struggle to stay motivated and focused when learning alone. We were inspired by new ways of interfacing with large-context language models, and we saw an opportunity to go beyond traditional AI assistants like Siri. Our vision is to build Twin—an AI companion that lives in something you carry with you every day, your phone! Twin understands your habits through context awareness and checks in with you like a supportive study buddy. Whether you're learning a new subject or need someone to help keep you on track, Twin has covered you. By blending accountability, conversation, and progress tracking, Twin makes learning more engaging, less isolating, and easier to stick with.

What it does

Twin has two core parts. First, an intelligent browser extension that continuously tracks your browsing activity—using web scraping and video transcription—to build rich, meaningful context for AI interactions.

Second, and what makes Twin truly unique, is its presence right on your phone. Instead of opening another app, Twin lives in your text messages, chatting with you like a friend. It checks in when you’re focused, helps you reflect, and keeps you accountable—making the experience feel natural, personal, and always within reach.

How we built it

Twin is built as a modern web ecosystem combining several key technologies:

Frontend Components

Chrome Extension: Built with Manifest V3, featuring a custom popup interface styled with a clean, minimal design to toggle tracking and change user data.
Activity Tracking: JavaScript-based content scripts that monitor user interactions across web pages
Real-time Updates: Dynamic status updates showing the latest tracked activity with human-readable time formatting

Backend Infrastructure

Flask API: Python-based REST API handling authentication, data storage, activity processing, and Twilio messaging
Supabase Integration: Leveraging Supabase for secure user authentication and scalable database storage
Real-time Sync: Seamless synchronization between browser activity and cloud storage
Firebase Node.js Cloud Function: Tackle all visual contexts to process images and videos into text serverlessly.
Cohere AI Agents: Intelligent pipelines for tool use and using Cohere's language models to analyze browsing patterns, generate learning summaries, and provide contextual insights from user activity data

Challenges we ran into

Building Twin presented several significant technical challenges:

Real-time Data Sync: Ensuring seamless sync between browser activity and cloud storage while maintaining performance was complex, especially handling network interruptions and offline scenarios.
Multi-step Tool Use with Cohere Agents: Designing multi-step reasoning pipelines with Cohere agents proved challenging but rewarding. We had to ensure the agent could reliably chain tools together for smooth user interactions.

Accomplishments that we're proud of

Seamless Behavior Tracking: Created an extension that tracks activity so smoothly users barely notice it's running
Real-time Insights: Successfully created a system that can provide immediate, contextual information about user activity patterns

What we learned

This project taught us valuable lessons about:

Browser Extension Architecture: Deep understanding of Chrome extension APIs, service workers, and content script limitations
Real-time Data Systems: Implementing efficient synchronization between client-side tracking and cloud-based storage
AI Context Integration: How to structure and format activity data to be maximally useful for large language model interactions
Cohere Agent Development: Building intelligent agents that can analyze user behavior patterns, generate contextual learning summaries, and provide personalized insights through Cohere's advanced language models

Twin represents just the beginning of a new era of AI-human collaboration, where technology truly understands and adapts to our digital lives in a truly human-like manner.