Skip to content

AGENTS.md

Latest commit

 

History

History
42 lines (35 loc) · 2.38 KB

AGENTS.md

File metadata and controls

42 lines (35 loc) · 2.38 KB

AGENTS.md — Small Models Demo Class

Project Purpose

This repository contains Jupyter notebooks for a freshman-level college course on small language models. All notebooks are teaching materials, not production code.

Primary Rule

Always optimize for student comprehension, never for code elegance or efficiency. These two goals are often in direct conflict. Always choose comprehension. A student should be able to read the notebook and understand each chunk of code.

Mandatory Code Constraints

  • All code must be written in the global scope, executed sequentially cell by cell.
  • Prefer lines of code to functions unless the notebook lesson is specifically teaching functions.
  • Never use list comprehensions. Always use explicit for-loops.
  • Never use lambda expressions.
  • Never abstract repeated code into helpers. Repetition is acceptable and often pedagogically valuable — students benefit from seeing the same pattern twice.
  • Imports go in the first cell only, one per line, with a comment explaining what each library does.
  • When necessary use try;except blocks to import libraries that may not be installed, but do not add any error handling beyond printing an error message.

Notebook Structure the Agent Must Follow

  1. Start with a markdown cell stating the learning objective in one sentence.
  2. Alternate markdown and code cells throughout. Never put two code cells in a row without a markdown explanation between them.
  3. End with a markdown cell summarizing what the student just did.

Language and Explanation Level

  • Explain as if the reader has never seen a concept before.
  • Define every technical term the first time it appears.
  • Never assume context. Never say "as you know" or "recall that."
  • Preferred sentence length: short. Preferred vocabulary: common English words.

What the Agent Should Never Do

  • Never wrap the main demo logic in a main() function or any equivalent.
  • Never add error handling (try/except) unless the lesson is about error handling.
  • Never generate a README or extra files unless explicitly asked.
  • Never refactor working code to be "cleaner" — clarity of the original steps matters more than clean code.
  • Don't name dataframes df. Use descriptive names like sales_data or customer_info.
  • Don't use generic variable names like x or temp. Use descriptive names that indicate their purpose, such as total_sales or average_age.