Back to full roadmap
topicfoundation
Zero-shot Prompting
Instructing without examples — enough for 80% of tasks on modern models.
2 hours2 resources1 prereqs
Modern LLMs are instruction-tuned, so most tasks don't need examples. A clear, specific, measurable instruction is enough.
Anatomy:
- Role — "You are a Python expert"
- Task — "Refactor the code below"
- Context — the code block
- Constraints — "Add type hints, write a docstring, keep under 50 lines"
- Format — "Return only the final code, no explanation"
What you'll gain
By using the 5-part scaffold for every task, you'll hit 80% quality on the first try.