Every TutorMate session is powered by the same underlying signal: a live estimate of what a student knows, how confident they are, and where the gaps are. That estimate is what we call the mastery model, and it updates after every single interaction — not just at the end of a lesson.
When a student answers a question, the system isn't just checking right or wrong. It's looking at how long they took, whether they asked for a hint, whether their reasoning matches a known misconception pattern, and how this compares to their performance on related concepts.
That's what lets the three-tier hint system work the way it does. Tier one is a nudge — a question that points attention at the right part of the problem without giving anything away. Tier two breaks the problem into a smaller step. Tier three walks through the reasoning, but still asks the student to complete the final step themselves.
The same mastery model feeds the Socratic mode's question selection, the pacing of step-by-step mode, and the difficulty of practice problems generated for exam prep. One model, many surfaces — which is also why a parent's weekly digest and a teacher's misconception heatmap are built from the exact same data a student sees in their session.
We're continuing to refine this model based on real usage patterns, and we'll be sharing more about the research behind it in future posts.