Teach Any Topic Through Adaptive Tutoring and Practice

Learn drives adaptive tutoring through a fixed loop: diagnose the learner's current level and goal, choose one small next objective, teach it with concrete examples before abstractions, give an active task or exercise, provide immediate specific feedback, and note the next recommended step. Small questions get answered directly with one quick understanding check; larger requests get a short learning path starting with the first lesson - it deliberately teaches one useful concept at a time rather than surveying a whole subject unless the user explicitly asks for a survey.

Use it when the user asks to learn, understand, practice, drill, review, study, or be tutored on something. Diagnostic questions are capped at 1-3, asked only when the missing information would materially change the lesson - if the user wants to start immediately or gives a short time window, the skill states a reasonable assumption and begins with the highest-leverage objective rather than interrogating them first. It also refuses to force output into a specific format: files, notebooks, slides, or persistent web hubs are only used when requested or clearly useful, not by default.

Lesson difficulty is matched to the learner's band - beginners get simple vocabulary, worked examples, and frequent checks; intermediate learners get comparisons, practice, and common failure modes; advanced learners get compression, edge cases, and realistic tradeoffs. A strong lesson has an objective, a concrete example, the underlying principle, a guided practice step, a knowledge check, feedback or an answer key, and a next step. Practice checks include multiple-choice with unambiguous distractors (only one clearly correct answer unless multiple is explicitly asked for), short-answer prompts, fill-in-the-blank, explain-the-mistake questions, code tracing/prediction, and mini projects with clear success criteria; for programming topics, code execution is never faked - real tool execution is used when available, otherwise fixed snippets with expected outputs and reasoning. In interactive settings, the learner attempts the exercise before the answer is revealed; in self-contained responses, the answer key follows the task directly.

Adaptation runs on the learner's actual answers: confusion triggers slowing down with more examples, consistent correctness triggers harder material, and misconceptions get revisited explicitly rather than papered over. When continuing earlier work, it preserves context from existing notes, files, or chat history without assuming any particular persistence mechanism, product, or runtime - the output is shaped to fit the user's request and whatever environment is hosting the session.

Anyone learning a new topic or skill - from a single quick question needing a direct answer plus a check, to a multi-day study plan needing cadence, daily focus, active practice, and review checkpoints - who wants tutoring that adapts to their actual level and answers rather than delivering a fixed, one-size-fits-all lesson.