Instructor Notes

Timing and Schedule


This lesson is designed for a single 90-minute slot. It is taught after the collaborative-git section (issues, branches, pull requests), so learners already know how to branch, commit, and push — you can move briskly through anything git-related.

Time Episode Notes
~10 min Introduction Benefits/challenges tables + Spack browsing challenge. Seeds the genAI theme.
~12 min Types of Software Documentation Dev/user split + the new Diátaxis four-mode lens.
~28 min Documentation Better Practices The two sandwich exercises are the heart of the lesson — protect this time.
~22 min Documentation in Practice README, docstrings (“why not what”), CITATION/licensing, and the Euler’s Method practice (with genAI twist).
~18 min Documentation Tools Style guides, Sphinx quickstart, and publishing.

Total ~90 min.

The Optional Tools exercise

The final Tools exercise is intentionally optional / take-home — learners almost certainly won’t finish it live, and that’s by design. If you’re short on time, point them to it and move on; do not cut the sandwich exercises to make room.

It now offers two paths:

  • Option A — a personal website (YOURUSERNAME.github.io from an academic/personal template). This is the recommended long-term takeaway: most learners will get more lasting value from a CV/portfolio site than from publishing the practice docs. The instructor having their own such site to show off is a great motivator.
  • Option B — publish the Sphinx docs via Read the Docs or GitHub Pages Actions, for those who want to close the loop on the project-docs workflow.

GenAI

GenAI is woven through the whole lesson rather than tacked on at the end:

  • Introduction — sets up “AI makes drafts cheap; the bottleneck moves to review.”
  • Better Practices — ties LLM hallucination to the sandwich “benefit of context” point.
  • Practice — the Euler’s Method exercise has a step to draft a docstring with an LLM and then critique/correct it.
  • Tools — AI assistants (Copilot, IDE docstring generators) with verify-it and data-policy caveats.

Reinforce the through-line: generate with AI, but always review.

Exercise Notes


  • Sandwich (x2): The point is discovering one’s own bias / benefit of context. The second pass should be visibly more detailed than the first — that contrast is the lesson.
  • Euler’s Method: Used in both Practice (free-form docs) and Tools (Google-style docstring). Learners can build on their first attempt.
  • Sphinx quickstart: Have learners use a virtual environment; default quickstart options are fine. This is the input to the optional publishing step.