first time here
Three things and you are set. Every card regenerates, so you learn the method instead of the answer. You type your answers and they are checked as math or text, with typos forgiven. Then you come back tomorrow, and the schedule does the rest.
hover anything to see what it does. press ? for the keys.
Ready when you are.
0 due · 0 new
how this works
Every card here is a template, not a fact. When a card comes up, its numbers are drawn fresh, so the integral you set up last week comes back with different coefficients and the only thing worth remembering is the method. Type your answer and it gets checked properly: 1/2, 0.5 and sqrt(2)/2 are all read as math, and for indefinite integrals any answer off by a constant counts.
Scheduling is spaced repetition (FSRS, the algorithm behind modern Anki). The app grades you from what actually happened: wrong is again, right but slow or with a hint is hard, right and quick is easy. You can override the grade before moving on. Cards you miss come back a few problems later until they stick, and those practice reps never touch the schedule.
Keys: enter submits and continues, h shows a hint, 1–4 override the grade, esc ends the session early, and ? shows the whole map. Everything is saved in your browser, and nothing leaves it unless you turn on sync under stats.
enter checks · h hints · u undoes · esc ends · ? shows every key
Done for today.
These are the two built-in decks. Your own cards come in through add from notes, about thirty seconds each, and show up here tagged edited.
add from notes
select the words worth remembering, mark them, then make cards. the app hides a different mark each review. AI drafting reads the whole paste and proposes cards for you to approve; it needs your API key in stats, and your notes go only to Anthropic.
new card
how templates work
Params draw fresh values each review. Types: int (min, max, optional ne list to exclude values), float (min, max, dp), pick (one record from a list, reach into it with {{q.field}}), and computed (an expr over earlier params, for clean displays). {{a:+}} writes the value with its sign, for building polynomials.
Answer types: number (expr evaluated with the drawn params), expression (expr plus vars; checked by sampling, so any equivalent form counts; set upToConstant for indefinite integrals and domain to keep samples away from singularities), choice (options plus a correct index), and self (you grade yourself, for derivations). Optional display is the pretty form shown after answering. Saving draws the card 100 times and makes the reference answer pass its own checker, so a broken card cannot reach a session.
| card | skills | state | due |
|---|
brain dump
You get three minutes and a blank page, notes closed. Write everything you can summon about one topic. Free recall like this beat concept mapping in Karpicke and Blunt's experiments, and it shows you what you actually have instead of what feels familiar.
skill strength, weakest first
Strength is the chance you would get the skill's cards right now, from the scheduler's memory model. It decays between reviews, which is the point: this list is what to worry about before an exam.
Nothing studied yet. One session and this page fills in.
Charts, records, and badges appear after your first review.
badges
the year so far
every review, one square per day. keep the wall green.
reviews per day, last 30
coming due, next 14 days
sync between devices
Optional. Your reviews and cards sync through a free Supabase project that you own; this app has no server of its own. The readme has the three-step setup, then paste the project's URL and anon key here.
exam dates
Give a deck an exam date and nothing in it gets scheduled past that day: reviews bunch up as the exam approaches instead of drifting beyond it. The today page counts down.
settings
everything lives in this browser's local storage.