Tutor Platform — one engine, many faces

A tutor engine that teaches people to properly master a subject, starting with the people who have been told to "adopt AI" and do not know where to begin. It uses diagnosis, active recall, and deliberate-practice drills rather than the passive explain-at-you register most AI tutoring defaults to, and it gates progress on demonstrated understanding rather than time spent.

One engine, two faces

The architecture is one engine behind two faces. A generic personal tutor that can teach me any topic I want, which is the test bed that proves the engine can teach anyone anything, and named niche tutors that are curated, mastery-gated, and commercial. The first committed niche is AI-skills development, where the depth of taste and the content moat from my own research corpus are strongest. Maths at GCSE level is next.

Why the split matters

The longer-term thesis is a niche factory. Once the engine, the content schema, and the freshness pipeline exist, a new subject is authored as content, treated as data, rather than built as new code. That is what turns a single tutor into a platform: the expensive engineering happens once, and each new niche is a content exercise on top of it rather than another build.

Me using the generic face daily is what proves that authoring path actually works, by use rather than just by architecture. If I can point the same engine at an unfamiliar topic and be taken to real mastery, the claim that a new niche is just content holds. If I cannot, no amount of clean architecture rescues it. Dogfooding is the validation, which is a pattern that runs through everything I build: the personal use case is user zero, and the thing only earns a wider audience once it has proven itself on me first.