Adaptive Practice

What Adaptive Practice does

Adaptive Practice runs continuously behind every Koydo study session. After each item the learner answers, the engine updates a per-topic mastery estimate, decays the estimate over time to model forgetting, and selects the next item from the candidate pool that is most likely to move the learner forward without producing a discouraging miss.

The engine is subject-aware: math practice uses different difficulty signals than reading-comprehension practice, and the testing verticals (NCLEX, CDL, Matura, JEE, IELTS) each have their own item-difficulty calibration. The user-facing behavior is uniform — the learner sees a smooth study queue — but the underlying scheduling is specialized.

What Adaptive Practice is not

Adaptive Practice is not the same as 'item response theory' (IRT) by itself, though it borrows the productive-challenge concept from the IRT literature. Koydo blends mastery estimates with a forgetting model, a confidence band, and editorial overrides that prevent the engine from drilling a learner endlessly on a category that does not need it.

Adaptive Practice does not skip the curriculum. If a topic is required by the learner's program, it appears in the queue at the correct prerequisite point even if the learner could nominally avoid it via the mastery estimates.

Where Adaptive Practice meets the AI tutor

Adaptive Practice decides which item the learner sees. The Everyday Tutor decides how to explain it when the learner asks for help. The two methods are deliberately separate: the practice engine never depends on tutor sessions, so the queue continues to work even if the learner declines tutoring entirely.