Most prep tools tell you which questions you got wrong. Crux figures out why — and the next question you see is the one that actually moves you forward.
Crux currently covers nine major math curricula. More are coming. Pick the closest match.
Every wrong attempt is classified — conceptual gap, procedural slip, careless mistake, time pressure, or motivational dip. The next intervention is chosen against the diagnosis, not just the symptom.
Crux picks the next topic to maximise expected score per hour of effort. Your weakest topic isn't always your highest-priority topic. The engine knows the difference and explains it.
Real-world hooks, contexts drawn from your actual interests, and a careful balance of new and review — so the topics you hate today aren't the ones you avoid tomorrow.
Every other week, a short no-AI probe — practice with no hints, no chat, no help — keeps the mastery estimate real, not inflated by scaffolding.
This is a starting prior. The diagnostic will refine it shortly.
Crux uses this to draw real-world contexts that feel less like math homework. Pick anything that's true.
~10 questions. The engine learns your starting mastery. No grade is shown to you — this is calibration, not a test.
—
Bands show 80% confidence interval — narrower = more evidence.
Things the engine has flagged. Resolved ones drop off this list.
Crux uses a cookie on this browser to remember you. To continue on a different device or browser, export your progress here and import it on the other side. There's no account, no email.