# Geistr Tutor – AI Learning System

> A desktop-first AI learning product that turns real sources and meaningful goals into short, active lessons with feedback, evidence, and a fitting next step.

## Overview

Geistr Tutor grows from a personal learning pattern. I learned programming and AI engineering mainly through YouTube, research, building my own products, and problems that appeared while building. That learning was strongest when it did not stop at passive consumption: I wanted to make something, encountered a gap, looked up the foundations I needed, compared options, and kept experimenting until I had a useful result. Tutor is meant to preserve that curiosity and connection to real goals while giving it clearer, more helpful structure.

The starting point is therefore not a course catalog or an empty chat window. A learner brings a source, a question, or a concrete goal: perhaps a documentary, article, paper, tutorial, or problem from their own project. Tutor clarifies why the topic matters, what the learner wants to be able to do afterward, and how closely the original source should shape the journey. The source can stay central, important claims can be checked deliberately, or the learner can choose a broader research foundation. The result is a persistent topic workspace rather than one forgettable conversation.

Inside that workspace, a dedicated tutor continues the learning relationship. The first orientation, Lesson 0, is neither an exam nor a generic beginner course. It connects the learner's mission to the high-level shape of the topic, discovers prior knowledge, observations, uncertainties, and preferred explanations. Only then does the learning path adapt to what the person can actually explain, apply, or question. This is the essential difference from a fixed content sequence: experience in programming does not automatically mean knowledge of neuroscience, music, or design, and the same source should create different learning journeys for different goals.

Lessons are designed as rich, calm working surfaces rather than long walls of generated text. They can combine concise explanations, visual material, examples, activities, tasks, source segments, and contextual questions. Keeping the original material connected is especially important: Video Focus Mode holds a relevant source interval, one clear point of attention, and questions about that exact moment together. When someone selects text in a lesson or gets stuck on a source, they can ask directly beside that context instead of having to reconstruct the whole situation in a separate chat.

Tutor deliberately does not treat progress as a completion percentage alone. Opening a lesson or clicking Continue does not prove understanding. Stronger signals come from an explanation in the learner's own words, application, correction, transfer to a new situation, an artifact, or later retrieval. Answers and submissions remain attached to their activities, feedback appears where the work happened, and a weak result can invite revision without preventing progress. The next step should respond to visible work, open questions, and the learner's purpose—not to an artificial mastery score.

Technically, Geistr Tutor is a desktop-first product built on the reusable Geistr Core foundation and Pi SDK. Geistr Core provides reusable desktop, agent, session, tool, and persistence systems; Tutor focuses on turning those foundations into a genuine learning relationship with topic workspaces, sources, lessons, feedback, and evidence. Topics, answers, observations, transcripts, and artifacts stay local or in learner-controlled storage by default. The public website documents the current product journey with real recordings. The desktop product remains in private development; public installers for Windows, macOS, and Linux are not available yet.

## Challenges

- Turning passive reading or watching into active learning with one concrete next step
- Preserving a personal mission and a strong original source instead of immediately layering on a generic course
- Deriving difficulty, examples, and teaching form from demonstrated, topic-specific prior knowledge
- Making questions possible exactly where a term, paragraph, or video moment matters
- Connecting feedback and revision to real work without reducing people to points, grades, or completion pressure
- Organizing sources, answers, transcripts, and learning observations in a local-first, traceable way

## Solutions

- Mission, success criteria, and source policy are clarified explicitly before topic preparation begins
- A persistent topic workspace connects sources, purpose, learning path, lessons, activities, feedback, and progress
- Lesson 0 calibrates the starting point before later lessons are pre-generated from assumptions
- Structured lessons combine explanations, visual material, activities, source segments, and contextual help
- Video Focus Mode and selected lesson text give questions the appropriate source and time context automatically
- Durable activity submissions, inline feedback, and separate coverage/evidence views make development visible without presenting completion as mastery

## Outcomes

- A real desktop-first learning journey from source intake through Lesson 0, active lessons, contextual questions, feedback, and evidence is implemented
- Video Focus Mode brings a meaningful video interval, a learning note, and timestamp-aware questions together in one dedicated learning view
- Submissions and feedback stay with the relevant activity; revision is possible without blocking the learner's path forward
- Coverage and learning evidence are presented separately so completed work is not represented as proven understanding
- The public project website documents the journey with real recordings, screenshots, and detailed product explanation
- The current status remains explicit: a private desktop product in development, with public installers and the Tutor repository to follow later

## Details

- Category: AI Learning Product
- Role: Product vision, learning model, UX, design & implementation
- Timeline: July 2026
- Status: in-progress
- Technologies: Electron, React, TypeScript, Pi SDK, SQLite, Geistr Core

## Links

- live: https://tutor.m-reschreiter.at
- github: https://github.com/Asm3r96/geistr-core