Overview
Formerly “Adaptive Classroom,” renamed with the Casework design system. Students work in small groups to solve a curriculum mystery — e.g. Who Started the Great Fire of London? — alternating between device time (private investigation with an AI mentor) and face-to-face group discussion.
How It Works
Every student gets a private AI channel that adapts its support level (struggling / on-track / advanced) invisibly — nobody knows what level of help anyone else is getting, which protects the room’s social dynamics. Evidence is jigsaw-distributed within each group, so the mystery can only be solved by actually talking to each other. The architecture is adapted from the Murder Party dinner-game engine.
Current Status
Phases 1–4 are built: project scaffold, a Postgres schema with row-level security, a secure Claude proxy via edge functions, the full session loop (lobby → rounds of intro → device time → discussion), a teacher dashboard, student and projector screens, the tiered AI mentor, the Great Fire of London pilot lesson, and an engagement layer (activity signals, teacher chips for Active / Idle / Nudge / Extension, and an opt-in auto-nudge loop with graduated gentle → specific → direct interventions). Phase 5 adds a teacher “Hold” (eyes-up-front) and more.