Teach It to Learn It.
Students coach an AI protégé that starts with zero problem-solving ability. It only learns what they teach it — to succeed, students have to organize their thinking and explain their reasoning clearly. The result is deeper conceptual understanding and the ability to communicate mathematically.
Content based on AMC 8 and AMC 10 competition problems, chosen for their rigor and range of mathematical reasoning.
Based on the Protégé Effect — the research-backed finding that teaching deepens understanding.
1.56×
learning gains
ALTER-Math / Univ. of Utah, 2024
This isn't a hunch. A 2024 University of Utah study with 6,000 middle schoolers found that students who taught an AI agent learned 1.56× as much as those using a standard online math platform. Two decades of research on the Protégé Effect confirm it: teaching is one of the most effective ways to learn. Protegz is built on that finding.
How It Works
Three steps. One powerful learning loop.
Coach
Explain problem-solving strategies to your AI protégé step by step. Guide it through competition-level math — the why, not just the what.
Challenge
My answer: (C) 15
Your protégé attempts the problem using only the strategies you taught. No shortcuts, no outside help — just your explanations put to the test.
Track
Watch your protégé's brain grow. See which concepts stuck, track mastery across topics, and identify where your teaching can improve.
A student teaches their protégé to find the area of a triangle.
After each session, see what clicked — and watch mastery grow.
Who It's For
For Educators
Teachers, tutors, math circle leaders, coaches
Moving students past correct answers and into real mathematical thinking is one of the hardest challenges in math education. Protegz gives every student a partner who can only learn what they're taught — which forces students to organize their thinking, make connections explicit, and confront what they don't actually know. You see the reasoning itself, not just the result.
- Review full coaching transcripts
- Track understanding across topics and techniques, student by student
- Spot classroom-wide gaps at a glance
- Share an invite code — students join in seconds
The Student Experience
Here’s what your students actually do: they pick a problem, coach their protégé through it step by step, then watch it attempt the problem using only what they taught. They see immediately whether their explanation landed — and where it fell short. Over time, they build a visible map of what their protégé has learned across topics and techniques.
- Problems drawn from AMC 8 and AMC 10 — rigorous, varied, well-structured
- Progress tracking across 50+ math topics and techniques
- A Playbook of concept refreshers for review before teaching
- Works independently — no teacher facilitation required during sessions
See where your whole classroom stands — technique by technique.
Free During Beta
Full access, no credit card required.
Pro
Everything unlocked while we're in beta.
All problems
Full protege progress tracking
Unlimited coaching sessions
Concepts playbook
Generous AI usage limits
See It in Action
Try coaching the protégé through one problem — no sign-up, about 90 seconds.
Protegz was created by Venkat Narayanan (Ph.D. Robotics, CMU · former Principal Tech Lead, Aurora) from a simple observation: over a decade of teaching and mentoring — from university classrooms to engineering teams — the deepest learning always came from being the one teaching. Protegz brings that experience to every student.