dsh-edu

Ask AI a question and it usually dumps three definitions, five pages of formulas, and two reference links in one breath—then you stare at the screen wondering which part matters, which symbol was never defined, and whether you actually understood anything. dsh-edu turns DeepSeek Harness into a properly prepared instructor: it first gathers authoritative sources from textbooks, university course pages, and papers, then walks through each topic in a fixed pedagogical order—why it matters, intuition and analogy, precise definition with every symbol defined before use, derivation, a fully worked example, common misconceptions, and how the idea connects to the rest of the subject. After every lesson it writes structured class notes to a persistent file, and at the end it quizzes you interactively, re-teaching anything you missed with fresh examples before marking the module complete.

The teaching loop is powered by three dedicated tools. edu_course creates and tracks the syllabus with a module checklist and progress state. edu_notes writes each lesson's explanation, key points, pitfalls, and sources into course files that survive across sessions. edu_quiz asks questions through the conversation's built-in question UI, auto-grades single- and multi-select answers, hands open-ended answers back to the model for judgement, and logs every result. Because progress lives in workspace files and mode state lives in the session log, you can resume or fork a session and pick up exactly where you left off.

It is well suited to learners who want to work through a subject systematically (linear algebra, quantum mechanics, operating systems, or any course that has a real textbook), to researchers who need AI to act as a tutor rather than a search engine, and to engineers or grad students who want their self-study to leave behind reviewable, persistent notes instead of a scrollable chat history.