Custom spaced-repetition scheduler
An SM-2 style engine tuned for real course rhythms: graduated intervals, lapse handling, and wrong-answer priority that resurfaces what you fumbled.
Spaced repetition, active recall, and a little dungeon crawling. Built solo by a kinesiology student who needed to remember an entire degree, and kept polishing until it felt like a game worth opening.
This is the real interaction, running right here on the page. The deck happens to be about why the app works.
Keyboard works too: 1–4 to answer, T/F for true or false, Enter for next.
One scheduler underneath, several ways to face it. Each mode exists because a real study week needed it.
Your weak cards become the boss fight. Wrong answers hit back.
Fact cards flip into fill-the-blank. Retrieval, not recognition.
The runs through the carpal tunnel.
Exactly what the scheduler says is due today. Nothing else.
A timed mock exam sampled from your lectures, scored like the real thing.
Rewards off, sounds down, timer on. Just you and the cards.
A daily-goal streak that only counts days you actually studied.
An SM-2 style engine tuned for real course rhythms: graduated intervals, lapse handling, and wrong-answer priority that resurfaces what you fumbled.
Every sound is synthesized at runtime from one A-major pentatonic palette, so any two overlapping tones stay consonant. The whole soundscape ships as code.
Strategy-per-asset service worker, per-course code splitting, and IndexedDB storage. Airplane mode is a supported study environment.
Optional cloud backup to a personal Cloudflare Worker. A 120-bit code is both identity and AES-256 key; the server only ever sees ciphertext.
A content-integrity test suite validates every card in every deck: gradeability, unique ids, valid lecture keys. Bad content cannot reach the build.
Weekly lecture slides become validated flashcards through a documented authoring loop with automated card counting and drift locks.
The game layer is sugar. The engine underneath follows three findings that keep replicating.
Testing yourself beats re-reading for long-term retention, even when the testing feels harder and less productive.
Roediger & Karpicke, 2006The same hours spread across days beat the same hours crammed into one night, by a wide and reliable margin.
Cepeda et al., 2006Conditions that slow learning down, like mixing topics and delaying review, tend to improve what survives a month later.
Bjork, 1994 · Rohrer & Taylor, 2007There is no account to create, no analytics script watching you, and no server that ever sees your progress in plain text. Everything lives on your device, and the optional sync is encrypted with a key only you hold.