Instructor Notes
Timing and Schedule
This lesson is designed for a single 60-minute slot. It pairs naturally with the Issue Tracking lesson and continues its StarSort scenario.
| Time | Episode | Notes |
|---|---|---|
| ~8 min | Introduction | Benefits table, StarSort callback, and the spack scavenger hunt. |
| ~16 min | Basic Pull Requests | The good-PR checklist (moved here, up front) + opening the StarSort fix PR. |
| ~12 min | Labels and Templates | Labels recap + building a PR template. |
| ~9 min | Spotting ‘Good’ (and Bad) PRs | The PR Critique game — formerly a lecture, now hands-on. |
| ~15 min | Code Reviews | Partner reviews — the collaborative high point. Protect this time. |
Total ~60 min.
The StarSort continuation
This lesson continues the StarSort story from Issue
Tracking: learners fix the bug they filed last time. It works fine
standalone too — if they skipped Issue Tracking, “fix the StarSort bug”
just means “make a small change to your README.md.” Play
the maintainer when they @-mention you or request a
review.
Running the PR Critique game (good-prs episode)
Learners diagnose three deliberately flawed PRs (too many changes / no description / oversized). Run it fast and out loud — call on the room, then reveal the solution block. It reinforces the better practices through spotting violations, which is the reviewer skill.
Running the partner reviews (reviews episode)
The two partner-review exercises are the best part of the lesson — protect this time. Learners need to add each other as collaborators (or share a repo) to formally request changes.
- Encourage kind, specific feedback (suggest, don’t just criticize) — see Conventional Comments in the learner references.
GenAI thread
GenAI is woven throughout, never a standalone block:
- Basic PRs — draft a PR description from a diff (then supply the why).
- Labels — generate a PR template, then trim.
- Good/Bad PRs — AI as a first-pass reviewer / diff-summarizer, with limits.
- Reviews — AI review assistants before human review; data/IP caveat for research code.
Through-line: AI drafts and assists; the human owns the merge decision.