Knovine

Knovine connects everyday AI learning to a knowledge map and the next classroom lesson. I study Mathematics Education at Seoul National University. In my own AI-assisted learning, while teaching students, and in conversations with teachers I know, I kept encountering the same gap: a useful question gets answered, but the learner's reasoning rarely informs their next explanation or their teacher's next lesson. Knovine makes learning state persistent, inspectable and learner-controlled. An observation preserves the learner's original words, its concept, a provisional interpretation and suggested connections. Personal learners can start immediately with an empty map or choose a subject guide. School and academy learners can join a teacher-defined course. The same evidence layer supports each setting. The demo follows a sample class through the complete connected workflow. A student asks about turgor pressure in ChatGPT. With permission, the Knovine MCP tools read existing class context and save the question. A new concept appears in the learner's map, with the original words attached. The student can continue inside Knovine, where Meta's Muse Spark 1.3 uses the same earlier observations to adapt its explanation. When class sharing is enabled, the teacher sees those original questions. A short outline becomes a complete Meta-generated lesson with differentiated prompts and an exit question. The teacher reviews, edits and accepts it; the original plan and revision history remain available. Turning sharing off removes the observations from the teacher's view while retaining the personal record. This working implementation has real ChatGPT tool calls, real Meta tutoring and teacher proposals, durable SQLite storage, learner/class isolation, and 37 passing tests plus a production build. Questions and AI answers are never treated as verified mastery. The external Muse agent connector is planned, separate from the working in-app Muse Spark model. The learner-state idea predates the event. The submitted repository contains this hackathon implementation and documents reused tools and its development boundaries. My goal is a learner-owned context layer across AI tools that helps people explore what comes next and gives teachers a practical basis for adapting a lesson.