AI Tutors

Each item in a course — a video, an article, a quiz — can carry its own AI tutor with a teacher-chosen personality (Socratic by default). You chat in plain language; the tutor answers with the course material behind it and shows you which lesson each part of the answer came from.

On quizzes, the tutor switches to a hints-only mode: it will nudge, reframe, and point at the right chapter, but it will not hand you the correct answer.

Answers from a general-purpose chatbot drift: they are not about YOUR course, and they arrive without provenance. Grounding the tutor in the course material (retrieval-augmented generation) keeps help relevant and lets you verify it — every reply cites the lesson it drew from.

The Socratic default and hints-only mode exist because retrieval practice — struggling to produce an answer yourself — is one of the most robust findings in learning science. A tutor that simply states answers feels helpful in the moment and quietly removes the work that builds memory.

When a course is published, its content is split into chunks and embedded into a vector index. Your question retrieves the most relevant chunks, which are placed in the tutor’s context alongside its personality prompt and your teacher’s own guidance.

Citations are extracted from the exact chunks used, so the "Ch. 7 · Oxidative phosphorylation" chip you see is not decoration — it is the source. Hint modes are enforced in the tutor’s system prompt, because streamed answers cannot be filtered after the fact.

Inside any lesson, open the tutor sidebar. On a quiz, use the hint chat — and notice it coaches instead of answering.