Bearly

A cognitive-science-backed desktop AI that turns passive studying into a personalised metacognitive loop.

Most students study longer, not smarter. They re-read notes, watch lectures on 2x speed, and convince themselves they understand — until the exam.

Bearly fixes that. It is a persistent desktop AI companion that sits on top of your workspace and forces active engagement with the material: you explain concepts, get diagnosed on your gaps, and drill the exact things you don't know — all without leaving your document. And unlike any other study tool, it remembers you — building a hyper-personalised learning profile that gets smarter with every session.

The result: a metacognitive study loop that builds genuine understanding, not just exam confidence.

Why Bearly?

Problem	Bearly's Solution
Passive re-reading feels productive but isn't	Teach-the-AI mode reveals what you actually understand
Generic AI tools don't know your weaknesses	Learning Memory builds a persistent profile — your strengths, struggles, and study style — that deepens with every session
Flashcards are tedious to make	One click generates AI flashcards from your study session
Pomodoro apps don't stop you opening YouTube	Built-in distraction detection nudges you back in real time
Study tools live in a separate tab you have to switch to	Bearly stays on-screen, above your work, always one glance away
No one helps you plan how to study	Voice AI structures your session around your goals and schedule
Study tools ignore how you're actually feeling	Bearly reads your mood and adapts its tone and approach accordingly

The Science Behind Bearly

Every core feature in Bearly is grounded in peer-reviewed cognitive science:

Technique	Research Backing	Bearly Feature
Retrieval Practice	Testing yourself on material improves long-term retention by up to 50% vs re-reading (Roediger & Karpicke, 2006, Science)	Active Recall, Ninja question pop-ups
The Feynman Technique	Explaining a concept in simple terms is one of the most effective ways to identify gaps in understanding	Reverse Teaching
Spaced Repetition	Distributing practice over time reduces forgetting by following the forgetting curve (Ebbinghaus, 1885; Cepeda et al., 2006)	Exam calendar reminders, flashcard system
Metacognition	Students who track their own knowledge gaps and adapt their study strategies consistently outperform peers (Flavell, 1979; Dunlosky et al., 2013)	Learning Memory profile
Focused Work + Break Cycles	Short, timed work sessions maintain cognitive performance and reduce mental fatigue (Ariga & Lleras, 2011)	Pomodoro Timer

Features

• Pomodoro Timer — Focus sessions with auto-hide sidebar, mini floating mode, and live distraction detection
• Active Recall — Generate quiz questions from your notes and test yourself with spaced repetition reminders
• Reverse Teaching — Explain concepts to the AI; it gives feedback, identifies gaps, and can generate flashcards from your session
• Flashcards — In-house deck management system:
  • Dedicated viewer window with flip animations and Next/Back navigation
  • Create decks and manually add cards via text input
  • AI-generate flashcards from any Reverse Teaching session using MiniMax
  • All cards stored persistently in flashcards.json
• Study Assistant — AI study buddy that:
  • Has conversations to plan and structure your study session around your goals
  • Reads your mood and adapts its tone, pace, and suggestions accordingly
  • Pops up random questions every 10 minutes during study sessions
  • Voice chat powered by ElevenLabs
  • Detects distractions (nudges you when visiting YouTube, social media, etc.)
  • Dock companion that stays visible at all times
• Calendar — Track exams with spaced repetition reminders (14, 7, 3, 1 days before)
• PDF Upload — Load PDF content as study material for quizzes and voice context
• Learning Memory — A metacognitive learning profile that hyper-personalises Bearly over time:
  • After every session, AI extracts and classifies what you studied, where you struggled, and how you learn best
  • Tracks: subjects covered, conceptual strengths, knowledge gaps, preferred study approach, and energy patterns
  • This profile is injected into every future AI interaction — so Bearly already knows you before you say a word
  • Grounded in metacognition research: students aware of their own learning gaps significantly outperform those who aren't (Dunlosky et al., 2013)
  • Similar in concept to ChatGPT Memory, but purpose-built for studying: the more you use it, the more tailored every response becomes

Getting Started

Prerequisites

• Node.js v18 or higher
• ElevenLabs API key — for the voice Study Assistant
• MiniMax API key — for AI feedback, flashcard generation, and learning memory

Setup

1. Clone the repository

git clone https://github.com/gust10/studyyyy.git
cd studyyyy

2. Install dependencies

npm install

3. Add your API keys

Open main.js and replace the placeholder values near the top of the file:

const MINIMAX_API_KEY = 'your-minimax-key-here';
const ELEVENLABS_API_KEY = 'your-elevenlabs-key-here';
const ELEVENLABS_AGENT_ID = 'your-agent-id-here';

4. Launch Bearly

npm start

Tip: Bearly is optimised for macOS. The app will anchor to the left edge of your screen and auto-hide after a few seconds of inactivity — hover over the edge to bring it back.

Project Structure

├── main.js                 # Electron main process, window management, IPC handlers
├── index.html              # Main sidebar UI
├── styles.css              # Global styles
├── quiz.html               # Quiz overlay
├── ninja.html              # Ninja popup (random questions)
├── ninja-dock.html         # Dock companion
├── ninja-chat.html         # Voice chat window
├── pomodoro.html           # Pomodoro timer
├── calendar.html           # Exam calendar
├── blur-overlay.html       # Fullscreen distraction overlay
├── flashcards-viewer.html  # Flashcard deck viewer window
├── flashcards.json         # Flashcard deck and card data
└── tabs/
    ├── active-recall.js
    ├── reverse-teaching.js
    └── reverse-teaching.html

Data Storage

User data (questions, exams, learning memory, flashcards) is stored in the app's user data directory:

• Windows: %APPDATA%/Bearly/Bearly-data/
• macOS: ~/Library/Application Support/Bearly/Bearly-data/

Supported Platforms

• macOS (highly recommended)
• Windows

Distraction detection uses platform-specific APIs (PowerShell on Windows, AppleScript on macOS).

License

ISC

Team Members

• Kei Jonathan McCall-Pohl - Software Development
• Maddalena Di Salvo - Software Development
• Hyunsung Shin - Software Development
• Prajitno Fiona Keira - Design
• Yash Relekar - Design

Recommendations

• Strongly Recommended for macOS
• Strongly Recommended for pdfs ~8000 characters / 400 words