ankarhem.dev / learn
This repository has no description
Star
Fork 0
Watch
TypeScript 100%
3 1 0
assets
extensions
skills
.gitignore
README.md
README.md

learn #

This fork is adapted for omo.dev. The original pi version lives upstream at amosblomqvist/learn.

video

My AI learning system from this video: How I Use AI to Learn Things.

This is a personal system I built for myself, shared as-is. Built as a pi configuration: the teaching philosophy encoded in a skill, a few small extensions, and subagent briefs.

What's in it #

  • skills/teach/ — the philosophy and the process; researcher.md is the brief for the researcher subagent
  • skills/visualize/ — adds a correct, minimal diagram to a lesson when an idea is clearer as a picture; mermaid-maker.md and svg-maker.md are the briefs for the maker subagents
  • extensions/ask-user-question/ — the agent asks you questions through a UI popup
  • extensions/quiz/ — graded questions with instant feedback (✓/✗, correct answer, explanation)
  • extensions/md-log/ — link a markdown file to the session

Install #

This repo is a .omo directory. From your learning project's root:

git clone <this repo> .omo

Then open omo in that directory and trust the project when asked (project skills and extensions only load in trusted projects).

Requirements #

  • omo. Subagents run through its built-in task tool, which spawns a worker that reads the brief file and follows it.
  • The visual makers must run on a model that can see images; the skill dispatches them with category="unspecified-high". Point that category at a vision-capable model in omo.jsonc.
  • Rendering: npx (for @mermaid-js/mermaid-cli, which uses a puppeteer-managed Chrome) and rsvg-convert (or nix shell nixpkgs#librsvg, or ImageMagick as a fallback).
  • ask-user-question — use the copy bundled here. If your setup already has an ask-user-question extension, use this one in its place. Popups from different extensions serialize through a shared UI lock, which only works when it's the same implementation.

Notes #

You can run the system without subagents. The main session does the teaching. You just lose the researcher (truth verification) and the generated visuals.

The teaching skill is written for one learner (me). Edit the skill to fit how you learn best.