I'm Pavel, co-founder of Recombee and an associate professor at CTU Prague. I built p-book because I kept running into a problem: most explainability work on recommender systems is aimed at researchers or regulators, not at the people actually affected by these systems every day — including kids.
The technical experiment here is whether personalizing how you explore a topic actually improves comprehension. Recombee drives the recommendations inside the book itself, so the book is eating its own dog food.
The thing I'm most unsure about: the four navigation modes (Missions, Browse, Read, Map) feel genuinely different to me as the builder, but I don't know if they feel different to a kid or a parent sitting with a kid. That's the feedback I most need.
Happy to talk about the technical architecture if anyone's curious — the spaced repetition layer on top of a live recommendation API is an unusual combination.