Meta-Prompting
The Consultant Who Writes Their Own Brief
7 min read
The Consultant Who Writes Their Own Brief
Instead of writing the brief yourself, you ask the consultant to write the best possible brief for the problem — then hand it back to them.
Most people write prompts by instinct. Meta-prompting flips this: you describe your goal to an AI and ask it to generate the optimal prompt for achieving that goal. Then you use that generated prompt. It sounds circular — but it works remarkably well, because the AI knows its own capabilities and failure modes better than most users do.
In Plain English
Meta-prompting is using an AI to write better prompts for an AI. You describe what you want to achieve, ask the model to generate the ideal prompt, and then use that prompt. It consistently outperforms human-written prompts for complex tasks.
The Technical Picture
Meta-prompting leverages the model's latent knowledge of effective prompting patterns to generate task-specific instruction sets. It's particularly powerful for complex structured tasks where the optimal prompt includes edge cases, constraints, and output specifications the user wouldn't naturally think to include.
Real-World Examples
- 'What's the best prompt to give you to generate a detailed competitor analysis?' — then use that prompt
- Anthropic's prompt generator tool uses meta-prompting internally
- Advanced builders use meta-prompting to generate system prompts for their AI products
Ask the AI to write your prompt before you write it yourself — the results will surprise you.