Built in Minutes, Checked by Script: Instructor-Made Interactive Chemistry Study Guides with Generative AI

Interactive study guides, with practice problems that change every time, help on request and small simulations, are valuable and slow to build; few instructors have time to make one for every unit. I describe a workflow that lets one instructor, a chemist with no programming background, build such guides with a generative AI coding assistant in about ten minutes each, plus about half an hour of polishing, and deliver them through GitHub Pages to Canvas or D2L Brightspace. Twenty pages for General Chemistry I, a course for non-science majors, biochemistry and Organic Chemistry I illustrate it; seventeen were built with this workflow over about a month. Each guide is a single web page that runs in any current browser, needs no account and sends nothing to a server. The guides share a design: a roadmap of the topic, sections in teaching order, problems whose numbers change each time, feedback on wrong answers, and small virtual instruments such as a titration, a calorimeter and a thin-layer plate; nine offer three levels of help. The biochemistry and organic guides take their values from named textbooks, and a check of every page, which ran about 1.6 million generated problems and recomputed each answer independently, found no wrong answer key but did find faults that reading had missed: named mistakes accepted as correct, correct answers rejected, an answer given away, and physically impossible setups; all were fixed and the check run again. No learning outcomes are claimed; evaluation is future work.