Back to the lessonPreparing the printout…

Prompt Engineering · Module 3 — Execute: Apply, Compare and Automate

Period 8 · AI-Assisted Prompting

[video] Lecture video — Period 8: AI-Assisted Prompting — /api/lesson-media/videos/4d3653ff-a3ca-4f63-bf8e-bb93c553e599-1794842707302-1729a8042000fbac.mp4

Flashcards · Period 8

Meta-prompt — A prompt that asks AI to help plan or write prompts, for example to propose stages and draft a prompt for each.

"Ask what is missing, draft the prompts, but do not run the analysis yet."

A / B / C comparison — A: you write the prompt. B: AI drafts the plan and prompts. C: you review and edit B. The three outputs are compared on one task.

Human-reviewed prompt — AI-generated instructions that you have checked and edited before testing. Review is not optional.

Fair comparison — Same CSV, brief, tool, model settings and output format, run in separate clean chats, scored with the same rubric.

Rubric (0/1/2) — Explicit criteria scored 0, 1 or 2. Five criteria give at most 10 points: Task, Figures, Findings, Coverage, Presentation.

Invented context — Facts a draft adds that the file or brief never supplied: dates, a profit target, a company history. Remove them in review.

Changed scope — A draft silently alters what is asked, such as turning a comparison of regions into a ranking of stores.

Ghi chú: The file has regions, not store identifiers.

Review gate — A point between stages where you check assumptions and outputs before the next stage runs.

Effort measures — Time spent and number of corrections. Record them to judge cost, but do not count them as part of the briefing score.

Task-specific tie — If two versions score the same, report comparable performance on this task in these runs, not a universal winner.

Prompt Engineering · Period 8 · AI-Assisted Prompting · printed on November 20, 2026