2.10 — Prompting Quiz 📊

1. What are the 4 ingredients of a well-structured prompt?

2. What are tags like <text>…</text> used for in a prompt?

3. You want the AI to answer always in the same tone and the same format. Zero-shot or Few-shot?

4. True or false: with a recent reasoning model, you should always add “think step by step.”

5. Rephrase positively: “Don’t make the answer too long.”

6. A client asks you for a “really original” slogan. Which technique pushes the AI to explore several paths before choosing?

7. True or false: switching from ChatGPT to Claude forces you to entirely rewrite your prompt.

8. You need to analyze a 300-page document in one go. Which assistant is known for its very large context?

9. Fast mode (“Instant”/Flash) or reasoning mode: on which does the explicit Chain-of-Thought help the most?

👉 Show answers

1. Role, Context, Task, Format.
2. To separate instructions from the content to process, so the AI doesn’t confuse them.
3. Few-shot: 2-3 “input → output” examples fix the tone and the format.
4. False. These models already reason internally; the instruction is often redundant. Explicit CoT remains useful on standard models.
5. For example: “Write a concise answer, 3 sentences maximum.”
6. The Tree-of-Thoughts (generate several ideas, critique them, then decide).
7. False. A good prompt is portable: the 4 ingredients work everywhere. You may adjust at the margins, you don’t rewrite.
8. Gemini (known for very large context) — the others also offer “large-context” options. To double-check: it evolves fast.
9. On the fast mode (standard model). In reasoning mode, the model already “thinks” on its own: explicit CoT is often redundant.