Hello, this is TobiraAI — thank you for reading as always.
Today’s goal: A Complete Guide to “Advanced Prompts” for Beginners — Using Chain of Thought, Premortem, ReAct, and 8 Extra Tricks.
If you haven’t read the previous chapter (Part 7), I recommend doing that first. But even trying one technique from this article will help you get better results with AI.
As I mentioned before, starting can be difficult — that’s why I suggest joining Mr. Shinichi Shichiri’s free ChatPro seminar (by Generative AI Institute, ex-pa.jp).
1. Introduction
In the previous part, we learned that AI conversations have a “structure.” The first step was REX: Role, Example, Expectation.
Now, we’ll add a bit of spice — to make AI think, test, and reflect.
These small changes turn AI into a thinking partner, not just a tool.
2. Chain of Thought — Let AI “Show Its Work”
Asking AI to explain how it thinks improves accuracy — just like telling students to “show your steps.”
Especially useful for comparisons, priorities, or step-by-step reasoning.
Example prompt:
“Explain your reasoning step by step (premise → comparison → decision). Summarize in a table and give 3 next steps.”
🟢 Tip: Add phrases like “explain step by step” — AI will automatically organize its logic.
3. Premortem — Predict Failures in Advance
Instead of reacting to failure later, imagine it first and plan prevention.
Ask AI to list possible risks, early warning signs, and backup plans.
🟢 Tip: Have AI identify “early signs” (e.g., “less than 10 signups in 3 days = risk”).
4. ReAct — Think → Act → Adjust
ReAct alternates reasoning and action. Perfect for travel planning, study plans, or writing improvement.
🟢 Tip: Break “Act” into small steps so adjustments are easy.
5. +8 Bonus Techniques for Beginners
- Chain of Density — Summarize stepwise for precision.
- Self-Critique — Let AI review its own output.
- Delimiter Use — Prevent confusion with clear sections.
- Clarify First — Make AI ask questions before answering.
- Checklist Creation — Have AI build its own review list.
- Role Switching — Move from planner → critic → editor.
- Step Limit — Prevent overlong answers.
- Rubber Duck Method — Use AI as a “thinking listener.”
6. Example: Combining REX + Advanced Prompts
“You are an editor. Summarize quarterly sales in 3 key points (CoT), then self-critique for clarity.”
This two-pass method gets you to a “90% perfect” draft quickly.
7. Human-in-the-Loop Final Check
Always verify key details (names, dates, data). AI assists — you finalize.
8. Summary
With REX to structure, CoT to reason, Premortem to prevent, and ReAct to iterate, AI becomes your second self.
Try “REX once a day, advanced prompt once a week.”
You’ll feel the difference in thinking quality and time efficiency.
AI evolves not linearly, but exponentially — staying curious is key.
And if this still feels difficult, join Mr. Shichiri’s free seminar — seeing is believing.
TobiraAI_ A specialist in Education, AI, and History 