learn-shit

name	learn-shit
description	Generate Jupyter notebooks that teach a user-chosen topic by interleaving compact theory, runnable experiments, visualizations, and a small culminating build. Use when the user asks to learn a topic through a notebook, wants a theory-plus-practice tutorial, mentions the spinning donut tutorial style, or asks for an .ipynb lesson in this format.

Learn Shit

Quick Start

When the user gives a learning topic, create a real .ipynb tutorial that feels like the reference notebook in references/spinning_donut.ipynb: small conceptual steps, immediate code checks, formulas where useful, and a concrete mini-project that combines the ideas.

Default to Chinese prose if the user writes in Chinese; otherwise use the user's language.

Workflow

1. Read the user's topic and infer the target level. If unclear, assume a curious beginner who can run Python.
2. Read or inspect references/spinning_donut.ipynb directly as the style and structure reference. Do not rely on a separate extracted summary when the notebook itself is available.
3. Choose one tangible final artifact for the lesson: simulation, visualization, parser, toy model, solver, animation, or terminal/HTML display.
4. Build the notebook as a sequence:
   • Title cell: # <topic>; add a subtitle or short intro only if it helps. Do not force 101 into the title.
   • Motivation and environment setup.
   • "What are the steps?" roadmap with 3-6 bullets.
   • Concept blocks: each concept gets a short markdown explanation, optional math, then a small code cell that proves or explores it.
   • Composition blocks: combine earlier pieces into larger functions.
   • Final build: render, animate, simulate, or evaluate something visible.
   • Reflection/exercises: 3-5 small modifications the learner can try.
5. Save as a valid notebook with executable Python code and minimal dependencies.

Notebook Style Rules

• Prefer short markdown cells over long lectures.
• Every abstract idea should be followed quickly by runnable code.
• Use numpy, matplotlib, IPython.display, and standard library tools when enough; add heavier libraries only when the topic truly needs them.
• Introduce names by using them, then explain just enough to keep momentum.
• Show intermediate arrays, tables, plots, or printed output so the learner can inspect the machinery.
• Keep code cells focused. One cell should teach one idea or build one reusable piece.
• Let the final artifact be playful or inspectable, not just a numeric answer.
• Avoid hidden magic: define helper functions in the notebook before using them.
• Include comments only where they orient the learner.

Bundled Reference

Use the complete bundled notebook at references/spinning_donut.ipynb as the example. Match its overall rhythm and notebook-native teaching style while adapting the content to the requested subject.