In a classical curriculum, the calendar rules. Chapter three arrives in week five, whether or not it has been understood. Gaps accumulate without ever being closed, and surface at the final exam — too late to act.
Mastery-based teaching inverts the constraint: what is fixed is the required standard; what varies is the time taken to reach it. A student who already has command of a module validates it and moves on. A struggling student takes it again, differently, until validation.
This approach has been known for a long time but remains rare, for a simple reason: it is expensive. It requires continuously diagnosing each learner's level and producing different exercises for each profile. At thirty students per teacher, that is materially impossible.
This is precisely what artificial intelligence makes practical. Diagnosis is continuous, exercises are generated from actual mistakes, and the teacher recovers time for what no machine does: supporting, correcting a line of reasoning, restoring motivation.
We accept the consequence: our programme durations are indicative. A programme announced over two years may be completed sooner by a fast-moving student, or later by someone working alongside. The qualification always certifies the same standard.