Lesson 2: Core Prompting Techniques

The Four Core Prompting Techniques

There are four fundamental techniques that form the foundation of prompt engineering. Master these and you'll handle 90% of prompting scenarios.

1. Zero-Shot Prompting

The simplest technique — give the model a task with no examples.

When to use: Simple, straightforward tasks where the model already has sufficient knowledge.

Example:

Classify this review as positive, negative, or neutral:
"The product arrived on time and works as expected."

Result: Positive

Pros: Fast, simple Cons: Less reliable for complex or nuanced tasks

2. Few-Shot Prompting

Provide examples of the desired input-output pattern before asking for new output.

When to use: Complex tasks, specific formatting requirements, or when you need consistent output.

Example:

Classify these reviews as positive, negative, or neutral:

Review: "Amazing quality, will buy again!" → Positive
Review: "Terrible, broke after one day." → Negative
Review: "It's okay, nothing special." → Neutral

Review: "The customer service was fantastic!" →

Result: Positive

Pros: More accurate, consistent output Cons: Uses more tokens, requires example selection

3. Chain-of-Thought (CoT) Prompting

Ask the model to show its reasoning step by step before giving the final answer.

When to use: Math, logic, complex analysis, or multi-step problems.

Example:

Solve this step by step:

A store sells shirts for $25 each. If you buy 3 or more, you get 20% off.
How much would 5 shirts cost?

Show your reasoning.

Step-by-step output:

1. Price per shirt: $25
2. Number of shirts: 5
3. Since 5 ≥ 3, discount applies
4. Discount: 20% of $25 = $5
5. Discounted price per shirt: $25 - $5 = $20
6. Total: 5 × $20 = $100

Pros: Improves accuracy for complex tasks, shows reasoning Cons: Longer outputs, more tokens

4. Self-Consistency

Run the same prompt multiple times and take the most common answer.

When to use: When accuracy is critical and you can afford multiple API calls.

How to implement:

1. Generate 3-5 responses to the same prompt
2. Compare the answers
3. Use the most frequently occurring answer

Pros: Higher accuracy Cons: More expensive, slower

Advanced Techniques

Tree of Thought (ToT)

Explore multiple reasoning paths simultaneously and evaluate each.

When to use: Creative tasks, complex problem-solving, strategic planning.

Example:

I need to choose between 3 marketing strategies. Evaluate each:

Strategy A: Influencer marketing
Strategy B: Content marketing
Strategy C: Paid advertising

For each strategy, consider:
1. Cost
2. Time to results
3. Long-term value
4. Risk level

Then recommend the best option for a startup with limited budget.

Persona Prompting

Assign a specific role or expertise to the model.

When to use: When you need specialized knowledge or a specific perspective.

Example:

You are a senior financial advisor with 20 years of experience.
A 30-year-old client asks: "Should I invest in index funds or individual stocks?"

Provide advice in a conversational tone, explaining the reasoning
for your recommendation.

Prompt Chaining

Break complex tasks into a series of simpler prompts, where each output feeds into the next.

When to use: Multi-step tasks, research projects, content creation pipelines.

Example Chain:

1. "List 10 trending topics in AI for 2026"
2. "For each topic, write a one-paragraph summary"
3. "Expand topic #3 into a 500-word blog post outline"

Choosing the Right Technique

Common Mistakes to Avoid

• Mistake: Using few-shot when zero-shot works — Fix: Start simple, add examples only if needed.
• Mistake: Not enough examples in few-shot — Fix: 3-5 examples usually work best.
• Mistake: Forcing CoT on simple tasks — Fix: Use CoT only for multi-step reasoning.
• Mistake: Copying examples that are too similar — Fix: Use diverse examples to show the pattern.

Professional Tips & Tricks

• Start with zero-shot, upgrade to few-shot only if the output isn't good enough.
• For chain-of-thought, explicitly ask for "step by step" — don't assume the model will do it.
• Mix techniques: few-shot + CoT often produces the best results.
• Test your prompts with edge cases, not just happy paths.

Key Takeaways

• Zero-shot is fastest; few-shot is more reliable; CoT improves reasoning.
• Choose the technique based on task complexity and accuracy needs.
• Few-shot works best with 3-5 diverse examples.
• Chain-of-thought significantly improves performance on reasoning tasks.
• Combine techniques for optimal results.

Next up: Advanced prompting patterns — structured outputs, chain-of-thought variants, and prompt templates.