An AI examiner that actually listens.
The IB Spanish oral is 15 minutes of preparation followed by a 12-minute conversation with an examiner. Students rarely get to practice in that exact format — teachers don't have the hours, and ChatGPT doesn't time you, doesn't catch tense slips, doesn't notice when you're stalling for a vocabulary word.
Six small algorithms wired into a single session: a state machine for the 15+12 prep/oral structure, Spanish tense detection, a response-depth analyzer, anti-cheat visibility detection, vocabulary-complexity scoring, and a speaking-pace analyzer. The teacher gets a dashboard; the student gets an examiner.
Built for one teacher, used by her IB Spanish students. Not public — the use case is intimate and the rubric assumptions are class-specific. The product works because it matches one teacher's actual rubric, not a universal one.
Stay private. Generalizing it would require turning a teacher's judgment into a config, and the value is in the judgment.