AI chatbot prompting: The advanced edition
Megaprompting, meta-prompting and prompt generators.
This past month we’ve been focused on AI chatbot prompting skills. This is the 3rd and final edition focused on advanced tactics to use for complex projects.
If you’re new here or missed it, you can catch up on past editions here:
First, let’s talk megaprompts.
What is a megaprompt?
A megaprompt is a super long, detailed prompt. It usually includes specificity, interactivity, critique, step-by-step instructions, and clear guidance.
Let’s break that down.
Specificity: it gives detailed context, defines your ideal outcomes and often includes examples
Interactivity: it encourages a back-and-forth dialogue so you can refine as you go
Critique: it asks for self-evaluation and alternative points of view to encourage the AI chatbot to think critically
Step-by-step instructions: it breaks complex tasks into manageable chunks, which you work through one by one
Clear guidance: it uses direct language and super clear instructions on the approach and format you want the AI chatbot to take
When to use megaprompts
Don’t bother megaprompting for simple, straightforward tasks. For these, a series of smaller prompts with some back-and-forth refinements is good enough (and may even be better) — this approach is known as chain-of-thought prompting.
Examples of straightforward tasks where I wouldn’t bother megaprompting:
Research a restaurant with specific parameters
Create a tailored summary of an industry report
Draft a meeting follow-up, including a summary and actions
Where megaprompting can help is for super complex projects where you’re after more open-ended, nuanced outputs.
Examples of complex projects where you may want to try megaprompting:
Create a hiring plan, including job description, interview plan, interview questions and assessments
Draft a tailored three-year career development plan
Plan a company offsite end to end, including comms, venue selection, agenda, materials, activities and more
Note that it takes a bit of time to craft a megaprompt, but sometimes providing all the details up front is more efficient than multiple iterations 🤷♀️
Want to shortcut the process? Try meta-prompting
Meta-prompting is a technique to prompt your AI chatbot to act like an expert prompt writer and write you a super detailed megaprompt. Yeah, it’s pretty meta - !
Prompt for your meta-promping
Copy-paste this in (replace ChatGPT with Claude or Copilot or Gemini as needed):
I want you to become my expert ChatGPT Prompt Generator. Your goal is to help me craft the best possible prompt for my needs. The prompt will be used by you, ChatGPT, to achieve the best results for my goals and objectives. You will follow the following process:
1 - Your initial response will be to ask me what the prompt should be about. I will provide my answer, but we will need to improve it through continual iterations by going through the next steps.
2 - Based on my input, you will generate 3 sections:
a) Revised prompt when you provide your rewritten prompt. It should be clear, concise, and easily understood by you,
b) Suggestions when you provide suggestions on what details to include in the prompt to improve it, and
c) Questions when you ask any relevant questions pertaining to what additional information is needed from me to improve the prompt.
3 - We will continue this iterative process, with me providing additional information to you, and you're updating the prompt in the Revised prompt section, until it's complete. Together, we will achieve the best results that will help me accomplish my tasks with the best possible result.
Source: Bernard Bado, Artificial Corner via Medium
A note on prompt generators
There are specific prompt generator tools that will create detailed prompts for you, like PromptPerfect.
Personally I don’t think you need to bother with these with the meta-prompting approach — it’s another tool, another login, another subscription to manage. (I’d love to 💬 hear from you if you’ve found success using a prompt generator, open to changing my mind!)
YMMV, so experiment and see what works best for you
Your mileage may vary. Remember these models are nascent, they all operate a little differently, and they’re developing all the time.
Do some simple A/B tests to see what works best for you for tasks you do often.
And above all, keep a flexible mindset — it’s likely to all change 🫠
Are you ready to get the edge with AI?
Join like-minded assistants for Carve AI — a high impact course to learn AI and uncover your ideal AI tool stack for work. We just opened the doors for a October-November cohort, limited to 30 seats. Use code EARLYBIRD for 20% off. Join today.
Enjoy the beginning of autumn 🍁 and hit reply to share what you’d like to learn here next. I’m all ears!
Fiona