Math CoTeacher

The first AI teacher that models how math understanding actually develops and uses that to build solid math foundations.

Three out of four 8th graders in America are below proficient in math. Math CoTeacher will fix that, running one-on-one diagnostic conversations, pinpointing misconceptions on a cognitive learning trajectory, and teaching like an expert human, at scale.

Built by a Math Teacher of the Year and two-time founder who created a proven cognitive map from arithmetic to algebra.

Are you an AI/ML engineer or an investor who wants to do what's never been done before and end 60 years of education stagnation?

Email Me Today → mark@mathcoteacher.com

Generational Failing

Why Me

I built and proved the human version of this system in the analog world.

I have the framework, the data playbook, and the hard won understanding of classrooms, businesses, and state-level politics needed to build this responsibly and at scale. That rare mix is what this moment in education demands.

From Math CoTeacher to Universal Teacher

How Math CoTeacher Works

  1. Elicit: Student solves a problem. AI elicits the student’s reasoning, not just their answers.
  2. Log: the entire interaction as training data, utterances, inferred state, chosen teaching move, and outcome as training data.
  3. Diagnose: AI maps the reasoning onto a detailed cognitive learning trajectory: Ex: “Accurate visual fraction arithmetic, accurate multiplicative justification, fragile on formal symbolic identity, level 5”
  4. Learn: Every interaction trains the system. Framework gets sharper. Instruction gets tighter.
  5. Teach: Meet the learner exactly where they are developmentally with the best lesson, explanation, worked example, and practice problems to move them up the stack.

See what the data from a sample interaction in JSON or JSONL format looks like.

Experiment Power at Scale: Building the Factory That Takes Us From Now to the Future

  1. Cognitive scientists know how to make maps called cognitive learning trajectories (CLTs). Teachers learn how to use them.
  2. CoTeacher can learn to do both. This is experiment power AND diffusion of that knowledge at scale, which is unimaginable today.
  3. Learning the meta skill of map making as we sharpen math and move into other K-12 subjects provides the training ground for learning a domain-general algorithm for building cognitive learning trajectories (CLTs).
  4. Math CoTeacher will be the first instance of a larger thing, an engine for designing, testing, and scaling how humans learn anything.
  5. This is building the first factory AND the assembly line for creating CLTs.
  6. The power of compounding: each new CLT compounds making CoTeacher more valuable. Everyone wants a data moat. CoTeacher will have an ever widening one compounded by flywheel effects that it can ride to becoming the world's universal teacher.

Next Steps

Build the founding team. Raise a seed round. Deploy an MVP with design partners by December.

Email Me → mark@mathcoteacher.com

The Curse of Knowledge

  1. If you're here, you are likely good at math.
  2. Most of America and the world is not.
  3. You likely can't relate to the daily frustration that turns to apathy for students or to the anxiety elementary school teachers feel about teaching the math they never really learned themselves.
  4. That's the Curse of Knowledge, not any moral failing.
  5. Try this math exercise. See if it can help you feel a little bit of what math class is like for the 3 out of 4 8th graders in America who are below proficient and why Math CoTeacher could be such a boon for mathematics education.
  6. You probably *did* feel the frustration of being ahead of everyone else, being bored, not challenged, and maybe even picked on for being good at math.
  7. When I say math education in America serves no one well, that's part of what I mean. Join me and make math education effective and satisfying for every kind of student.