When a Simple Algorithm will Do
summary
A local-notification generator analyzes the past six weeks of logs, counts only the first log of the day, doubles the two most recent weeks, and picks times by hour; if a pattern appears at least magic number: three times for that weekday it schedules a repeating time-based notification an hour later plus 12 minutes. No CoreML, just heuristics and calendar triggers, a 21-day training mode for early users, and background refreshes to keep reminders accurate.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 1:55The problem with tracking apps
- 2:57Making it easy to use
- 6:07Todays system
- 7:11The dumb way
- 8:37The right way
- 9:41Machine learning
- 11:07Caffeine habits
- 11:52Notifications
- 12:31Data Collection
- 13:24Day of the Week
- 15:20Coffee
- 15:50First Log
- 17:30Data Breakdown
- 19:40The Magic Number
- 20:22Example
- 21:07Triggers
- 21:48Randomness
- 23:27Training Mode
- 24:51Training
- 35:54Coding
- 36:26Gotchas
- 39:34Repeating reminders
- 40:48What I learned
- 41:01Good oldfashioned thinking
- 41:33Apple TV
- 41:56Recap
related talks
25mIntelligent Notification Systems at Scale
31mHow Do Push Notification Recommendations at Reddit Work?
06mIntermittent Learning: On-Device Machine Learning on Intermittently Powered System
11mPersonalizing your app in real-time: Powered by Firebase and ML
40mNext-generation Alerting and Fault Detection
01h 05mWhat the hell am I doing?
21mRise of the Machines: Removing the Human-in-the-Loop
25mBeyond Send-Time Optimization