08 JAN 2026
David decided: module planner outputs define lecture candidates
in Kerra, lecture-note candidates should come from AI plans, not deterministic fallback guesses.
a lecture candidate in Kerra is a card a student can click to generate lecture notes or a workspace for a course module. the intended flow is that a module planner agent reads the change report and chooses which plans should become candidates.
David pushed back when the system created fallback candidates without the planner doing the decision work, because that breaks the product model.
one candidate = 1 plan
candidates arent meant to be deterministic
the candidate is supposed to represent an intelligent judgment about the course material, not a mechanical placeholder that happens to have the right shape.