Lumi

Problem

Learning work is often scattered across notes, chat history, quizzes, and revision tools, leaving students to operate the system and reconstruct their progress themselves.

From an agentic-learning idea to Lumi

Lumi is the working expression of the Agentic Learning idea: a learning environment where a student can focus on what they want to learn while an agent coordinates the surrounding workflow.

The goal is not to replace the student’s thinking or hand out answers. It is to help keep the learning loop connected: capture material, explain it, practise actively, record observed results, and return to the right topic for revision.

What is working now

The current local-first build brings notes, visual Canvases, saved academic materials, interactive quizzes and flashcards, and specialist learning agents into one workspace. Useful output is saved to Artifacts, a searchable library instead of a disposable chat scrollback.

The shared learning record preserves scores, misconception patterns, and due reviews locally so teaching, testing, and review can work from the same evidence. Lumi also lets a learner choose a local Ollama connection or an explicit cloud provider, while keeping that boundary visible.

What comes next

The next phase is about validating the complete learning loop with real study sessions: whether the agents create useful practice, whether the evidence-based review is genuinely helpful, and how much autonomy is useful without taking control away from the learner.

For the detailed product update, including the data, privacy, model-choice, and portability boundaries, read Lumi: a learning workspace that stays yours.