Reinforcia

Persistent student modelling and teacher orchestration for secondary education. An AI-native learning system built around the learner, not the lesson.

The thesis

Most edtech optimises the lesson, not the learner. Content gets adapted, sequenced, and gamified, but the underlying model of what a given student actually knows stays shallow, usually just a percentage score on a quiz. Reinforcia starts from the opposite end: build a persistent, evolving model of each student's mastery, and let everything else (content sequencing, teacher intervention, reporting) be generated from that model rather than bolted on top of it.

The thesis is that secondary education is bottlenecked less by content availability and more by the absence of a real time, per student cognitive model that teachers can act on. If you can track mastery probabilistically at the skill level, not just at the subject level, you can tell a teacher exactly which student is about to fall behind on exactly which concept, before it shows up in a test result.

How it works

Reinforcia is built around Bayesian Knowledge Tracing (BKT): each skill a student is learning has a probability of mastery, P(L), that updates with every interaction, correct or incorrect, based on parameters for prior knowledge, learning rate, guess, and slip. Instead of a static gradebook, the student model is a live, continuously updated belief state.

On top of that sits a teacher orchestration layer. Rather than manually tracking 30 students across dozens of skills, teachers get a dashboard that surfaces where mastery is breaking down and generates recommended next actions such as regrouping a topic, flagging a student for intervention, or adjusting pacing. The system is being designed with institutional governance in mind from the start, including data handling and reporting standards that fit how NZ secondary schools actually operate, rather than retrofitted later.

Currently in build, with the first live testing phase gated on infrastructure setup (email and notification delivery via AWS SES), followed by internal testing and an onboarding cohort of real test users.