Simulated Students

Two faces of the same idea. In a dedicated AI-student lesson, the student is configured by your teacher: it arrives knowing it is here to learn specific outcomes, at a stated level — "beginner", "exam prep" — and opens the conversation itself. On article and video lessons, teach-back lives in a side panel: the same student persona, scoped to that one lesson.

You explain in plain language. The student pushes back the way a real one would — "backs up? what exactly is flowing along the chain?" — and stays stuck until your explanation actually lands.

The Feynman technique: the fastest way to expose a gap in your understanding is to try to teach the material to someone else. Explaining forces retrieval, organization, and elaboration — and a student that keeps asking exposes exactly where your "understanding" was really just recognition.

The evaluation is not a grade to fear. It measures how completely, clearly, and accurately you taught — and every session feeds your Learning Twin as evidence, the same way a quiz does.

The student runs on your course’s indexed materials and the lesson’s own content — it asks about what the lesson actually covers, never invents a syllabus of its own, and never pretends to understand a vague explanation. "Evaluate my teaching" scores the thread against the lesson’s extracted concepts: coverage (did you teach them all), clarity, and accuracy — with feedback you can act on.

Like everything in your twin, the evidence is visible to you: teach-back sessions show up as concept evidence on your Twin page.

In article and video lessons, open the Teach-back panel (the speech-bubble button beside the tutor). Where your teacher added a dedicated AI-student lesson, it opens ready to teach — the outcomes it needs are listed right above the chat.