Nothing here is proprietary magic. Three well-studied findings — spacing, retrieval, and cognitive load — implemented literally, in the one place they matter: your daily hour.
Ebbinghaus showed that without review, most of what you learn is gone within days. Each well-timed review flattens that curve and stretches the next interval.
Pebble runs SM-2 over whole problem structures rather than flashcards, aiming for desirable difficulty: late enough that recall costs effort, early enough that it still succeeds. You never pick what to review — deciding is the tax that kills most self-study.
Re-reading a solution feels like learning; it's recognition, the shallowest form of processing. Explaining forces retrieval, compression and honesty — gaps you can hide from a compiler are impossible to hide from a plain-language walkthrough.
So most reviews in Pebble are spoken, not typed: state the invariant, name the edge cases, give the complexity. Five minutes at maximum depth instead of forty minutes of boilerplate.
Sweller's cognitive load theory separates effort that builds schemas from effort that is pure overhead. "What should I practise today?" is overhead — and it runs on the same limited working memory the problem needs. Pebble moves planning, sequencing and scheduling out of your head entirely.
{{ s.note }}