What powers the model
EstudyLog runs on a topic-aware spaced repetition stack layered on top of Firestore. Every course outline you provide becomes an evolving map of concepts, weighted by recency, difficulty, ratings, and completion status.
Topic graph
Each outline block becomes a node. We track total minutes, average session spacing, and confidence ratings so the model knows what’s solid and what needs reinforcement.
Decay curves
Sessions age along adaptive decay curves tuned for student-paced learning. We never expose the raw coefficients, but you’ll see the results in the priority list surfaced in Courses, Dashboard, and Calendar views.
Coverage balance
Minutes without a topic assignment land in an “unmatched” bucket. The model boosts those topics until you fill in the gaps, ensuring outlines stay aligned with real activity.
What you see in the app
- Dashboard quick stats: your streak, pace, and a curated prompt for the next high-impact course.
- Course coverage panel: study intensity, untouched sections, and within-topic spacing suggestions.
- Calendar highlights: heat map intensity derived from the same spaced repetition scores.
Why we keep the details private
The exact weighting, decay functions, and reinforcement rules are part of our competitive advantage. We’re transparent about the signals we use - session length, ratings, recency, and topic coverage - while keeping the implementation sealed so students get consistent results and so we can iterate safely.
Coming soon
- Adaptive exams that sample weak topics based on the current spacing score.
- Instructor dashboards that surface aggregated coverage (opt-in and anonymized).
- Optional Apple Intelligence enhancements for on-device outline parsing on iOS 18.
Questions about the model? Email model@estudylog.com and we’ll loop you into the research updates.