Skip to content

A Collection of Prompts for Problem Solving

1. Deep Reasoning & Problem Decomposition

Template: You are {role}. First break the problem into clear, logically consistent steps, then provide the final answer. Think carefully before you answer.

Example: You are a business strategy consultant. First break down the key steps for “launching an AI compliance tool in the UK market,” then deliver a complete GTM strategy.

Principle: Force planning before answering to reduce skipped steps and reasoning errors.

2. Critique Before Final

Template: Draft a first version. Then conduct a strict self-review and mark weak points. Finally, rewrite to fix each issue.

Example: Write a \~400-word sustainability press release; then critique its clarity and persuasiveness; rewrite accordingly.

Principle: Leverage GPT-5’s stronger self-reflection and rewriting.

3. Role + Objective + Constraints

Template: You are {expert role}. Objective: {objective}. Strictly follow these constraints: {list of rules}.

Example: You are a senior policy advisor. Objective: summarize a 20-page climate report for a parliamentary committee. Constraints: ≤500 words, neutral tone, bullet points.

Principle: Make boundaries and deliverable specs explicit; GPT-5 follows constraints more reliably.

4. Progressive Deepening

Template: Start with a high-level overview of {topic}. Then pick the most critical point, explain it in depth with evidence and examples.

Example: First overview the EU’s AI regulation landscape; then go deep on the enforcement challenges in the AI Act.

Principle: Go macro first, then drill down; increase information gradually without overwhelming.

5. Multi-Modal Integration

Template: Here is mixed input (text/images/files, etc.): {paste/quote}. Analyze it holistically and produce {output type and formatting requirements}.

Example: Given a market report PDF and a product image, synthesize them into 2 investor pitch slides with key points.

Principle: GPT-5 is better at integrating multimodal information into a coherent output.

6. Exploratory Idea Map

Template: For the broad goal of {broad goal}, branch out all feasible paths. Organize by category; give at least 3 concrete options/examples per category.

Example: For monetizing an “AI tools news” newsletter, group by “Sponsorship / Product / Community / Partnerships,” with ≥3 items in each group.

Principle: Encourage divergent thinking within a structure to avoid unorganized lists.

7. Expert Emulation

Template: Answer in the style, decision preferences, and priorities of {noted expert/professional role}, and explicitly show their thinking framework.

Example: From the perspective of a “chip startup CEO,” set the launch and market entry strategy for a new AI chip company.

Principle: More precise style and strategy emulation yields decision logic tailored to the scenario.

8. Constraint‑Anchored Creativity

Template: Create {output type} and strictly satisfy these constraints: {word count / register / structure / prohibitions, etc.}.

Example: Write a 100-word LinkedIn post explaining generative AI to a 10-year-old: use only one analogy; end with a question.

Principle: Constraints focus creativity and prevent topic drift or style mismatch.

9. Long‑Context Knowledge Recall

Template: Based on the “entire conversation so far,” produce {deliverable}. If needed, refer back and cite earlier details.

Example: Using the earlier discussion on AI regulation, write a 2,000-word policy white paper.

Principle: Use GPT-5’s more stable long-range memory and detail retention.

10. Broad Goal → Precise Delivery (2‑Step)

Template:

  1. Broad goal: {draft goal}. First, align with me on the exact deliverable (spec/structure/acceptance criteria).
  2. Produce the deliverable immediately according to the agreed specs.

Example:

  1. Broad goal: Increase my engagement on LinkedIn for AI topics.
  2. Deliverable: A 4-week content calendar (topics, opening hooks, CTAs).

Principle: Align on goals and acceptance first, then deliver high quality in one go.