Lock In

Login
Main
Map
Map
Report
Profile

Inspiration

Study timer apps like Forest and Yeolpumta (YPT) track hours but never verify if students are actually focused. A student can run a timer for hours while scrolling Instagram. We wanted to build something that measures real cognitive engagement, not just time spent sitting.

What it does

Lock In is a macOS-native study app that combines AI-powered focus verification with campus building competition.

Core Loop:
Open the app → Start a study session → AI monitors focus via webcam and screen → Earn a Session Score → Contribute to building territory → Become Building King → Weekly reset

Key Features:

AI Focus Verification -- Webcam tracks gaze, posture, and blink patterns. Screen activity is analyzed using Meta’s TRIBEv2 model to estimate cognitive engagement.
Session Score -- Focus (80%) + Duration (20%). No fake hours, only real concentration counts.
Campus Territory -- Buildings are divided into a hex grid where students compete for control in real time.
Building King -- Top scorer per building earns the title. Resets weekly.
Brain Activity Report -- Post-session insights on cognitive engagement and personalized study tips.
World ID Authentication -- Ensures 1 person = 1 account. Prevents bots and multi-account abuse while preserving anonymity.
Admin Dashboard (B2B) -- Universities receive anonymized insights into study behavior, space usage, and focus trends.

How we built it

Frontend: Swift / SwiftUI (macOS), Apple MapKit, Canvas-based hex grid visualization
Backend: FastAPI + SQLite, JWT authentication
AI/ML: TRIBEv2 (brain activation prediction), TensorFlow.js Face Mesh (gaze tracking), ScreenCaptureKit
Auth: World ID (IDKit integration)
Data: OpenStreetMap Overpass API for real campus building data

Challenges we ran into

Mapbox SDK does not support macOS → pivoted to Apple MapKit
macOS App Sandbox blocked localhost connections → required entitlement configuration
Swift Charts conflicts with MapKit → resolved by isolating chart logic
Race conditions in session API calls → fixed with async flow control
Git merge conflicts across frontend/backend development
Integrating World ID into a macOS environment

Accomplishments that we're proud of

Real-time hex grid territory system with dynamic updates
Integration of Meta’s TRIBEv2 model (released just one week ago)
Full pipeline: World ID → session → AI scoring → territory update → analytics
B2B dashboard designed based on real university needs
Authentic campus mapping using OpenStreetMap data
Ensured both real-user verification and data reliability while maintaining user anonymity
Real-time WebSocket updates for territory changes

What we learned

Measuring real focus is significantly harder than tracking time
macOS development has stricter sandbox and networking constraints than expected
Combining multiple AI signals (vision + screen activity) improves reliability
Gamification dramatically increases motivation compared to passive tracking

What's next for Lock In

MiniKit integration for World App deployment
World Chain for verifiable academic achievements
iOS companion app
Pilot program with Purdue University
Introduce competitive challenges and optional reward-based (e.g., stake-based) study modes