Prompt Patterns That Work
What you will learn
- Apply chain-of-thought prompting to improve AI reasoning on complex problems
- Use few-shot examples to get consistent, predictable output formats
- Understand when each prompt pattern is most effective
- Combine multiple patterns in a single prompt for maximum effectiveness
# Prompt Patterns That Work
Most prompting advice is vague: "be specific," "provide context," "iterate." This lesson teaches concrete patterns — specific prompt structures that reliably improve AI output. These are not tips. They are techniques used by AI researchers.
Pattern 1: Chain-of-Thought (CoT)
Chain-of-thought prompting asks the AI to reason through a problem step by step before giving an answer. This dramatically improves accuracy on complex tasks.
Without CoT:
`
If a store has 23 apples and receives 3 boxes with 12 apples each,
then sells 15 apples, how many remain?
`
AI might jump to a wrong answer.
With CoT:
`
If a store has 23 apples and receives 3 boxes with 12 apples each,
then sells 15 apples, how many remain?
Think through this step by step before giving your answer.
The AI will write: "Starting with 23. Three boxes of 12 = 36 more. 23 + 36 = 59. Sell 15. 59 - 15 = 44 apples remain."
When to use CoT: - Math or logic problems - Multi-factor decisions - Debugging code - Analyzing complex scenarios - Any task where the reasoning matters as much as the answer
Unlock this lesson
Upgrade to Pro to access the full content
What you'll learn:
- Apply chain-of-thought prompting to improve AI reasoning on complex problems
- Use few-shot examples to get consistent, predictable output formats
- Understand when each prompt pattern is most effective