tutorkit
tutorkit
Teach a topic over many sessions and remember what stuck. tutorkit keeps one learning repo with a folder per topic, pitches each lesson at the edge of what the user already knows, writes the lesson as a printable HTML artifact, and schedules retrieval cues that come due later. A tutor is someone who remembers you and adapts; that persistence is the whole design.
Four postures, not four presentations: explain answers and writes nothing, lesson teaches and is the only mode that opens a track, drill tests recall across topics, exam measures and refuses to teach.
status is a fifth mode and not a fifth posture. It teaches nothing, asks nothing, and grades nothing. It reads the two router files, prints where every track stands, and crowns the next move. Run it to decide which of the four postures to run.
What tutorkit is not
- Not a code writer. It never implements a feature in the user's project. It reads their code to build examples and stops there.
- Not a decision report. "Which queue library should we use" is a research question, even though both read primary sources.
- Not project documentation. It documents nothing about the user's repo.
- Not a spaced-repetition engine. A fixed five-step interval ladder, no SM-2 or FSRS tuning, no Anki export, no retention modelling. The schedule exists to make review happen, not to be optimal.
- Not general-purpose by design. It works for any topic and it is tuned for software engineering, where repo grounding, runnable exercises, and source-code citations are all sharper.
- It never commits the learning repo on its own.