Challenges I Ran Into
🔥 Major Technical Hurdles

1. Real-Time Typing Speed Tracking Without Killing Battery
   The Problem:
   Initially tried monitoring every single keystroke system-wide to calculate typing speed and error rates. This absolutely destroyed battery life—phone would drain 40% faster. Users would uninstall immediately.
   The Solution:

Switched to sampling-based approach: capture typing metrics every 30 seconds instead of every keystroke
Used Android's AccessibilityService more efficiently with debouncing
Implemented smart pause detection—stop tracking when phone is idle
Result: Battery impact reduced from 40% to just 8-12%

2. App Usage Categorization Was a Nightmare
   The Problem:
   How do you know if YouTube is "productive" or "distracting"? A student watching physics lectures vs. cat videos—same app, totally different context. Simple app blocking would break legitimate study.
   The Solution:

Built a context-aware categorization system:

Users manually categorize apps per goal (YouTube = productive for "Video Editing Goal", distracting for "Math Study Goal")
Time-of-day intelligence: Instagram during scheduled breaks = OK, during focus sessions = distraction
Machine learning model trained on patterns: if user switches apps every 2 mins = distraction, 20+ mins = likely productive

Added manual override: "This WAS productive" button for edge cases

3. Fatigue Calculation: Way Harder Than Expected
   The Problem:
   Initial algorithm was just total_hours_today / 24 * 100. Completely useless. It didn't account for:

Break quality (15-min phone scroll ≠ 15-min walk)
Task difficulty (calculus ≠ reading notes)
Individual differences (some people can focus 10 hours, others max at 6)

The Solution:
Multi-factor fatigue model:
Fatigue Score = weighted_average(

• Continuous focus time (longer = higher fatigue)
• Typing error rate (errors spike when tired)
• App switching frequency (increases with fatigue)
• Break deficit (hours since last real break)
• Time of day (natural circadian rhythm)
• Historical baseline (personalized to user)
  )
  Took 3 weeks of testing with real users to tune the weights correctly.

4. Firebase Realtime Sync Was Eating Data
   The Problem:
   Syncing app usage data every minute to Firebase for leaderboards consumed 500MB+ monthly data. Unacceptable for students with limited data plans.
   The Solution:

Implemented smart sync strategy:

Batch updates: collect 15 minutes of data, compress, send once
Sync only on WiFi by default (user preference)
Delta sync: only send changes, not full state
Offline-first architecture with Room database

Data usage dropped to ~50MB/month

5. Leaderboard Cheating Was Rampant in Testing
   The Problem:
   Beta testers gamed the system:

Left study apps open overnight
Used auto-clickers to fake activity
Manually inflated session times

The Solution:

Activity validation algorithms:

Typing activity must be detected (idle apps don't count)
Movement patterns analyzed (humans don't type at perfect intervals)
Max session length caps (8 hours without break = suspicious)
Random integrity checks: "Are you still studying?" popup

Implemented soft penalties: suspicious sessions marked with lower quality scores rather than hard bans (benefit of doubt)

6. Accessibility Permissions Scared Users Away
   The Problem:
   Android's AccessibilityService requires scary permission prompts: "This app can read EVERYTHING on your screen, including passwords."
   45% of beta users abandoned signup at this step. Trust issue, not technical issue.
   The Solution:

Transparency campaign:

Added explainer video showing exactly what we track (app names, typing speed) and what we DON'T (passwords, messages, browsing history)
Open-sourced the tracking code on GitHub
Added "Privacy Dashboard" showing all collected data
Gave users granular control: "Track typing" toggle, "Track apps" toggle

Abandonment dropped to 18%

7. Pomodoro Timer Broke During Phone Sleep
   The Problem:
   Android Doze mode would kill the timer service. Users would start a 25-min focus session, phone sleeps, timer stops at 12 minutes. Session data corrupted.
   The Solution:

Used AlarmManager with setExactAndAllowWhileIdle() for critical timer events
Foreground service with persistent notification (can't be killed)
Added recovery logic: if app reopens mid-session, recalculate time from session start timestamp
Fallback: if timer is off by >2 mins, show dialog asking user to confirm actual focus time

🎓 Key Lessons Learned

Battery life is sacred—users will tolerate slow features but not dead batteries
Permissions need education—be transparent about data usage
Context matters more than raw data—YouTube isn't inherently good or bad
Users will cheat—design for adversarial users from day one

The hardest part wasn't the code—it was understanding human behavior and designing systems that work with it, not against it.