Quizzes & Hint Chats

Teachers author quizzes by hand or generate a draft from course content with AI, then curate it. Learners attempt, get server-side graded results, and can generate an AI review that explains each wrong answer.

Stuck mid-attempt? Open the hint chat: it is configured hints-only, so it nudges and points instead of answering.

The testing effect: answering questions is not just measurement, it IS studying — each retrieval strengthens the memory. But that only works if the answer cannot leak into the attempt; a hint that states the correct option converts practice into copying.

Separately, trust: if correct answers ever reached the browser, a curious learner could read them from the network traffic. Integrity has to live server-side.

Grading runs entirely on the server and compares against stored correct answers — the client only ever receives your score and per-question correctness. The hint mode’s never-reveal rules live in the tutor’s system prompt (streamed text cannot be filtered afterwards).

Your attempts feed the twin with full evidence context: which concepts, whether a hint chat existed, whether an answer review from a previous attempt was available — that is how the twin can tell supported success from independent success.

Any quiz lesson in a course. The AI review appears after submission, and both the review text and the misses are searchable in Learning memory.