The gap between a mediocre and a great LLM response is often just how you asked. Prompting is a skill worth developing.

**Basic principles:**

1. **Be specific**: 'Explain transformers to a 10-year-old using a library analogy' > 'Explain transformers'

2. **Give context**: Include relevant background the model needs

3. **Define format**: 'Respond as a bullet list with max 5 items'

4. **Few-shot examples**: Show 2-3 examples of the output format you want

5. **Chain-of-thought**: 'Think step by step' improves reasoning tasks significantly

6. **Assign a role**: 'You are a senior ML engineer reviewing this code'

**Advanced:** ReAct prompting, tool use, structured output (JSON mode).

Prompt Engineering: Getting the Most from LLMs