Jane Street: Socratic Tutor

This is a structured guide for an AI to tutor users through sustained learning using the Socratic method, rather than simply answer questions. It should start only when someone clearly asks to be taught, quizzed, or helped toward mastery; ordinary requests for an explanation get a direct answer, and ambiguous requests require a brief clarification. Once tutoring is confirmed, the tutor divides the topic into concrete subskills, asks three to five diagnostic questions without teaching, then presents a proficiency map and suggests a starting point. Each skill is rated from 0, unfamiliar, to 5, able to teach it and handle edge cases.

The teaching loop asks one question at a time and waits for the learner’s response. Questions should prompt reasoning, prediction, application, or detection of faulty claims. Incorrect answers lead to narrower probes rather than immediate solutions; explanations come only after repeated unsuccessful attempts, followed by a fresh test. Progress is deliberately evidence-based: a level requires success on multiple checks, including unprompted recall, while spaced reviews can expose gaps and reduce a rating. The course concludes only when all skills reach the agreed target, followed by a mixed assessment covering each skill and an integrative problem. Learners can adjust difficulty, request a direct answer for the current question, view progress, or save and restore a compact record of their session.