Learn by doing,
without the ceiling.
Interactive lessons where the visualization is the explanation. Predict what a simulation will do, watch it prove you wrong, then implement the thing yourself and be graded on whether it actually works.
Every topic runs from foundation to frontier on one page. Nothing is taught twice — lessons link through a concept graph. Progress, XP and spaced repetition live in your browser and never touch a server.
How it works
Predict, then confront
You commit to an answer before the simulation runs. Being wrong on purpose is the mechanism — reading that RK4 loses energy teaches nothing next to predicting it will not, and watching it.
Implement, and be measured
Code challenges are graded by measuring the convergence order your implementation actually achieves. Any correct method passes; a plausible one that is secretly second-order does not.
Foundation to frontier
One page carries the first-pass explanation and the graduate treatment of the same idea. You choose the depth; the platform does not choose it for you.
Recall, scheduled
Cards are authored inline where the idea appears and scheduled with FSRS-6. A review item can be an interactive prediction, not just a front and a back.
One concept, one owner
A concept is taught by exactly one lesson, enforced at build time. Everything else links to it — so the graph stays a graph and the material never repeats itself.
Yours, locally
XP, streaks and review state live in SQLite in your browser. No account, no server, no subscription. Export the file whenever you want.