Not a chat window that hands you the answer. Not a quiz that hands out badges. A structured lesson: short concept, examiner-graded past-paper drills, and homework that auto-targets the skills you missed.
No credit card · 50 free credits to try every feature
Why it's different
Generic AI chatbots and gamified quizzes both fall short for exam-track maths. We rebuilt the loop around the markscheme.
A chatbot will give you the answer if you ask, and skip the working if you don't. Our lessons follow a fixed workflow — concept → understanding check → past-paper drill — with the agent visibly held to it.
Pretty progress bars don't help on exam day. Every past-paper question we ship has the official IB markscheme attached; the grader checks method marks, answer marks, and reasoning marks the same way an examiner does.
When you reach a lesson milestone, the platform reads your per-skill score on the concept and generates a homework set that biases toward the sub-skills you missed. No two students get the same paper.
Inside a lesson
The flow is deliberate: diagnose where you are, teach the missing piece, drill the past-paper questions, then close the loop. If you stall, the engine drops you into a foundation-rebuilder loop before pushing you back up.
Three quick-pick cards or free text — "I want to revise partial fractions" biases the whole lesson from move one.
A teacher-designed chunk plus a 4-6 minute video — not a 25-minute lecture you'll skim.
Real IB questions, mid-to-hard. The agent grades against the markscheme and shows the working when you miss.
If you slip on a hard question, you don't loop on it. The engine drops to 2-3 AI-generated easier questions on the same skill, then climbs back up.
Cross the understanding threshold and a homework set is auto-assigned, tagged to your weakness. The lesson stays open — keep asking, keep practising.
Three-dimensional mastery
Every graded attempt updates a structured profile across three independent axes, mapped to Bloom's taxonomy. You see exactly which dimension is weakest, and the engine picks questions accordingly.
Can you explain the why, not just the what? Bumps up on every correct answer that demonstrates you know the rule, not just the recipe.
Can you carry the same idea to an unfamiliar shape? Word problems and parametric variants verify you're not pattern-matching the question stem.
Can you reason about edge cases and connect concepts? Cross-topic past papers and 'what happens if…' prompts feed this score.
Every lesson tracks your skill-grain proficiency — not just "calculus" but "power rule", "chain rule", "set-derivative-to-zero", etc. When the agent picks the five questions for your homework set, it pulls from the pool ordered by skill-overlap with your weakest areas first.
The result: roughly 60% of your homework questions deliberately drill the sub-skills you actually missed during the lesson, with the rest filling out general practice in the same difficulty band.
The platform isn't a black box. Every step the agent walks — the order, the question difficulty, the prompts it reads — is exposed as a node graph and a prompt panel. Customise per class, see exactly what changed, roll back if you don't like it.
Educator overview →Sign in with Google or email, pick a concept from the IB Math syllabus, and the first lesson is yours in under a minute. 50 free credits — no card required.