Spine scenes everyone sees, detours the learner chooses, remediation the system chooses after a wrong answer, and checkpoints everyone must pass.
Every move is recorded. This is the data a future AI tutor learns from: which questions get asked, which metaphors work, where learners get stuck.