A mobile training partner that combines real handwriting, computer-vision grading, and habit-forming gamification — already in classrooms.
Built with researchers and real classroom teachers — not bolted on.
Apps got faster. Students got worse.
More screen-time. Flat outcomes. The medium is the problem.
Three independent findings, the same conclusion: the act of working out by hand is the learning.
Handwriting activates motor cortex and working-memory regions in tandem. Tapping doesn't.
van der Meer & van der Weel, Frontiers in Psychology, 2017Writing forces synthesis. Students who write notes outperform those who type — even when typed notes are longer.
Mueller & Oppenheimer, Psychological Science, 2014The chain of steps — not the final answer — predicts long-term retention.
Rittle-Johnson et al., J. Ed. Psychology, 2019The loop is built around the pen.
Read the full research →
The adaptive engine picks one at the edge of your ability — not too easy, not too hard.
Pen, paper, full working. Phone goes face-down. Real cognitive load, no shortcuts.
The camera reads every line of your handwriting — every digit, every step.
Encouragement first, then exactly where the logic broke. Try again — no shame, no streak penalty.
When the chain of reasoning checks out — confetti, XP, streak intact. The dopamine lands on paper, where it belongs.
While students work on paper, teachers get a live view of what's actually happening — not just who finished, but where each student's reasoning broke and why.
Mastery, streaks, and time-on-task — sorted by who needs you next, not by surname.
"7 of 24 students stuck on logarithm rules, step 2" — so you know what to re-teach tomorrow.
Weekly reports show which examples worked, what to repeat, and where to push the bar higher.
The engine adapts to where each student actually is — so the kid who hates math gets unstuck, the kid who's coasting gets pushed, and the kid who's already ahead doesn't get bored.
Slows down. Re-introduces the concept. Walks through worked examples. Lets them practice as many times as they need — no raised hands, no embarrassment.
"I used to skip math homework. Now it's the first thing I open."
Spots when answers are right but reasoning is shaky. Picks problems at the edge of ability — not too easy, not too hard. Builds the habit before the test does.
"Now I actually want to know why I got it wrong."
Skips the easy ones. Surfaces stretch problems and proofs. Tracks elegance, not just correctness — so the strong students stay engaged instead of disengaging.
"It finally gives me problems that feel hard."
Engagement isn't a feature. It's what happens when the difficulty curve fits.
Student data is the most sensitive thing a school handles. MathXP is built to the strictest European data-protection standards — privacy and security designed in from the ground up.
The system and its models never know who a student is. Names and identities stay with the teacher alone — MathXP grades the maths, not the person.
Student work is read and analysed by MathXP's own models on EU servers — never handed off to external US LLM providers.
EU-hosted, end-to-end. Handwriting images stay inside the EU. No transit, no replication abroad.
Schools own everything. DPA per institution. Right to deletion within 30 days. Full audit log per account.
No training on student work. Foundation models never see identifiable student data. Nothing sold or shared.
Curriculum-bounded outputs. No open-ended chat. No off-topic generation. No minors-targeted ads. Ever.
MathXP is already in classrooms. If you'd like to explore a pilot for your school, leave your details and we'll be in touch.
