Beacon

Insights into attention and activity.

Comment

Beacon — Project Story (Updated)

Inspiration

Beacon was inspired by a simple observation: most monitoring and accountability tools either overwhelm users with excessive surveillance or provide shallow insights without meaningful context. We wanted to rethink monitoring from the ground up — creating a system that feels intentional, secure, and transparent, closer to a mission-control interface than a traditional productivity tracker.

Our goal was to build a platform that balances awareness and privacy, clearly defining when tracking happens and when it intentionally does not. Beacon introduces structure to monitoring through session-based design, ensuring data collection always has context.

What it does

Beacon is a session-based monitoring and analytics platform designed around two distinct session types:

1. Block Sessions

  • Category-based sessions with no behavioral tracking
  • Records:
    • Participants
    • Session duration
    • Session metadata
  • No flags are generated
  • Excluded entirely from dashboards and analytics
  • Designed for privacy-preserving collaboration periods

2. Track Sessions

  • Fully monitored sessions within a defined timeframe
  • Behavioral signals are analyzed and flagged
  • Tracks:
    • Who participated
    • Which flags occurred
    • When events happened
    • User-specific activity summaries

Behavioral Signal Detection

Beacon currently monitors two independent signals:

Eye Flag (Attention Detection)

  • Triggered when attention deviation is detected through tracking eye movement
  • Automatically records 15 seconds of video surrounding the event
  • Captures contextual footage (before/during detection)
  • Video stored securely for review
  • Only short contextual clips are saved — not continuous recording

Cursor Flag

  • Triggered when anomalous cursor behavior occurs:
    • Cursor idle for 60+ seconds, or
    • Periodic erratic movement patterns detected

This helps identify disengagement or abnormal interaction behavior without constant monitoring.

Note: Keyboard tracking functionality is planned for the future. The backend, database structure, and Gemini API integration are already in place to support it, but it is not currently active. When implemented, it will track typing within a 15-second window, classify activity as work/non-work, and store only the flagged results — never raw text.

User & Session Features

  • Add User functionality for session participation management
  • Session dashboard displaying:
    • Category
    • Time duration
    • Status
    • Flag counts (track sessions only)
  • Individual user views showing:
    • Sessions participated in
    • Flags received per tracking session
  • Block sessions show participation only — no behavioral metrics
  • Planned future updates will include per-user keyboard flag visibility within sessions

How we built it

Beacon was built using a modern full-stack architecture emphasizing security and scalability.

Backend & Infrastructure

  • Supabase
    • Authentication and user management
    • Relational database storage
    • Secure object storage buckets
    • Instance-based video storage
  • Dedicated video buckets per session instance
  • Event-driven flag recording pipeline

AI & Planned Keyboard Integration

  • Gemini API
    • Already integrated into the backend
    • Will be used to classify keyboard activity as work-related or not
    • Privacy-first design ensures raw text is never stored
  • Current focus is on eye and cursor flags; keyboard tracking will be activated in future updates

System Architecture

We unified sessions under a single abstraction:

[ Session = (type,\ participants,\ time,\ events) ]

where:

  • Block sessions → ( events = \varnothing )
  • Track sessions → contain structured flag events
  • Future keyboard flags → integrated as additional structured events

Development Workflow

We transitioned development from browser-based prototyping to Cursor, enabling:

  • AI-assisted engineering workflows
  • Faster iteration cycles
  • Integrated refactoring and system reasoning
  • Cleaner architectural evolution

Challenges we ran into

Balancing monitoring with privacy
Designing meaningful signals without storing sensitive user data required careful architectural constraints.

Event-based video capture
Recording only relevant 15-second clips instead of continuous footage required precise timing pipelines and storage handling.

Signal reliability
Defining cursor anomalies and attention signals without excessive false positives was an iterative process.

Planning future AI integration
Keyboard functionality is non-trivial because of privacy and real-time classification requirements, but the system has been structured to handle it securely once implemented.

Accomplishments that we're proud of

  • Built a dual-session system separating privacy-safe and monitored environments
  • Implemented contextual video capture instead of constant surveillance
  • Designed privacy-first AI classification (no raw text storage)
  • Created multi-signal behavioral awareness (eye and cursor)
  • Prepared backend and API for future keyboard functionality
  • Integrated secure cloud storage and authentication via Supabase
  • Developed a structured, security-forward interface

Beacon demonstrates that monitoring tools can be powerful without being invasive.

What we learned

  • Privacy constraints drive better architecture.
  • Context matters more than volume in behavioral analytics.
  • Users trust systems that clearly define tracking boundaries.
  • AI works best as a classifier, not a recorder.
  • Security perception is shaped as much by UI design as backend implementation.

We learned that innovation often comes from deciding what not to collect.

What's next for Beacon

  • Implement keyboard tracking and flagging functionality using Gemini API
  • Advanced session timelines and playback visualization
  • Role-based permissions and access control
  • Behavioral trend insights across tracking sessions
  • Smart summaries generated from session signals
  • Adaptive UI modes based on session category
  • Real-time alert intelligence
  • Expanded anomaly detection models

Long-term, Beacon aims to evolve into a secure situational awareness platform — combining intelligent signals, strong privacy guarantees, and immersive operational design.

Built With

Share this project:

Updates