Working Mode

I Built a Chrome Extension That Catches Me Slacking Off (Automatically tracks and categorizes your web time) Go from feeling busy to being effective~

Comment

🚀 Working Mode: The Illusion of Productivity

💭 Inspiration — The Illusion of Productivity

During the whirlwind of preparing my master’s applications, my browser turned into a black hole of “productive” chaos. I thought I was working — researching programs, drafting essays, studying. Yet every night ended with the same sinking feeling:

“Wait… what did I actually accomplish today?”

Traditional tools didn’t help.
Manual timers? I lied to them.
To-do lists? I cheated them.
I’d proudly mark a Pomodoro as “productive,” even though half of it was me scrolling YouTube.

So I decided to build something that wouldn’t let me lie —
👉 an objective auditor of my digital time.

⚙️ What It Does

Working Mode is your intelligent, privacy-first productivity analyst — silently running in the background while revealing the truth of your digital habits.

✨ AI-Powered Categorization
Uses Google’s Gemini models to smartly classify every site you visit:
→ Work • Study • Utility • Social • Entertainment (could be customized) Each one rolls up neatly under umbrellas like Work / Rest / Other.

📊 Time Analytics Across Scales
Zoom out to your yearly trend, or zoom in to see how you spent your last hour.

⏱️ Smart Focus Protection
Gentle nudges evolve into stronger reminders when distractions drag on — customizable “soft” and “hard” limits keep you accountable.

🧠 Adaptive Learning
Every correction you make trains your personal rule engine — your extension literally learns how you define focus.

🔒 Privacy-First by Design
No servers. No cloud. No leaks.
Everything happens locally — your data never leaves your machine.

🧩 How We Built It

A mix of clean engineering and subtle intelligence.

🧠 Core Stack

Built on the WebExtensions API using vanilla JavaScript, with Chrome’s storage API for persistence and service workers for background logic.

🤖 Intelligent Classification Engine

  • Multi-layer fallback: user-specified Gemini model → auto-discovery fallback
  • Real-time classification from URLs, titles, and AI insights
  • TTL-based caching to minimize API calls without losing accuracy

🗂️ Data Architecture

  • Dual-layer tracking: fine-grained 1-hour logs + efficient daily aggregates
  • Timezone-aware normalization for global users
  • Optimized data structures for long-term performance

🎨 User Experience

  • Live popup dashboard with animated pie charts and time counters
  • Collapsible domain-category lists for granular insights
  • Fully customizable fine categories with umbrella mappings

🧠 Challenges We Faced

😅 Ambiguous Sites — “youtube.com” can teach you math or waste your evening. We built a confidence-based suggestion system that learns from your feedback.

⚡ Performance vs. Precision — Real-time logging without lag meant deep optimization:

  • Event debouncing
  • Smart caching
  • Lightweight data structures

🌐 API Reliability — Gemini models vary across regions. We built auto-discovery and graceful fallback to stay stable everywhere.

🕒 Timezone Chaos — Global users, different clocks. Solution: IANA timezone normalization with fallback aliasing.

🏆 What We’re Proud Of

💡 Technical Elegance: Blending rule-based caching with AI classification — fast and intelligent.
🧘 User-Centric Design: Supportive reminders that coach rather than scold.
📚 Data Architecture: Dual tracking system = instant feedback + historical analytics.
🛡️ Privacy Achievement: 100% local processing, 0% data leakage. A rare combo in 2025.

💬 What We Learned

  • 📈 Psychology of Focus: People respond better to insight than restriction — data transparency beats blocking.
  • 🤖 Reality of AI Integration: LLMs are powerful but unpredictable; robust fallback systems are essential.
  • 🧩 Extension Optimization: Even tiny inefficiencies can slow the browser — minimalism is king.
  • ⚙️ User Complexity: Fewer options, smarter defaults = happier users.

🔮 What’s Next

🧭 Personalized ML:
Truly adaptive learning — context-aware models that adjust based on your time of day and habits.

💪 Smarter Focus Features:

  • Deep Work scheduling
  • Calendar integration for auto-adjusted classification
  • Context-aware blocking

📊 Enhanced Analytics:

  • Productivity scores
  • Trend comparisons
  • Export to Notion / Obsidian

📱 Cross-Platform Future:
Browser → Desktop → Mobile → Complete digital wellness.

🤝 Collaborative Insights:
Team dashboards (privacy-respecting!) to improve collective focus across remote teams.

✨ Working Mode isn’t just another productivity app.
It’s a mirror — one that finally shows how you truly spend your most precious resource: time.

Built With

Share this project:

Updates