iterative-lesson-refinementlisted

# Iterative Lesson Refinement ## Overview A QA loop that turns a draft lesson into one that *actually teaches* and that the learner can *defend live*, by exposing it to **isolated personas** instead of grading your own homework. Core principle: a lesson's author cannot see its holes, and a persona you role-play inline just echoes what you already know; isolation is the whole point. ## When to Use - A teaching artifact must survive a skeptical follow-up ("you said cosine, why not Euclidean?"), not just read well. - The learner is time-starved: the artifact must be as SHORT as possible while still letting them do the thing and survive the second question. Brevity is a graded criterion, not an afterthought. - **When NOT to use:** content nobody must defend (internal notes, logs); or a first draft that hasn't been written yet (write the tight draft first, then run the loop). ## The Loop (per topic) 1. **Build/revise** a tight draft. Ground every claim to the learner's real work honestly (HAVE / PARTIAL / GAP). If a companion HTML-craft skill exists, use it for the page itself. 2. **Dispatch 4 ISOLATED subagents**, each reading ONLY the artifact file: no answer key, no authoring context. Locked roles: - **NOOB**: technical generalist who has never met the topic; no outside lookups. Reports confusion, drag, and density/overwhelm flags; takes the cold quiz. Your density detector. - **BASICS**: surface familiarity; mandate = depth gaps, half-explained ter