Compound Leverage

Do you want to get more done with less effort?

The cheat code is crafting prompts that make ChatGPT and all the other AI tools work smarter.

In the last few weeks, we’ve seen new models and products from OpenAI, such as Canvas and the 01 preview model. SearchGPT is on its way.

With all the AI tools out there, the key is getting the most out of AI is writing better prompts.

Which comes down to two things:

• Applying critical thinking and

• Writing quality

Crafting good prompts is the difference between ChatGPT being helpful or frustrating.

The formula is being specific and direct and providing context like you're talking to an expert colleague.

Instead of spending time building prompt templates, refocus your energy around prompt frameworks and experimentation.

Your knowledge of a topic is the secret sauce combined with your ability to apply non-linear thinking to prompt writing.

Prompts fall into four main categories.

1. Understand

2. Create

3. Learn and

4. Code

Here's how to use these prompt categories and get the most out of them.

1. Understand Prompts These prompts help you gather and analyze information. Use them to have the AI summarize documents, extract key insights, or provide background on a topic.

Example: "Summarize the attached RFP, focusing on the key requirements and compliance criteria."

2. Create Prompts: Do you happen to need the AI to generate something from scratch? Create prompts come in handy for writing tasks like drafting emails, creating content, or brainstorming ideas.

Example: "Generate a series of social media posts promoting our new product launch, focusing on the key benefits for our target audience."

3. Learn Prompts Got a new skill to master? Learn prompts turn the AI into your tutor. Use them to get explanations, walkthroughs, and feedback as you learn.

Example: "Explain the basic principles of responsive web design and provide examples of best practices."

4. Code Prompts For developing code, Code prompts are helpful for wanna-be coders like me. Use them to have the AI generate code snippets, troubleshoot bugs, or explain complex concepts.

Example: "Write a Python function that takes a list of numbers and returns the sum of the squares of each number."

The more advanced prompts combine categories.

Take this example of an Understand/Create hybrid prompt for evaluating an RFP:

"You are acting as a proposal writing expert who has won billions in proposals.

Your task is to help me assess an RFP to decide whether to bid. Please ask me a series of questions to gather the information needed to make your recommendation, including:

1. The RFP details and compliance requirements

2. My business capabilities and bidding strategy

3. My knowledge of and interactions with the offeror

4. Any additional relevant documents

5. Data on past awardees

Based on my responses, please provide a recommendation on whether to bid, considering that our philosophy is to bid only when we have a 50% or greater chance of winning."

This prompt guides the AI through a complex analysis task by chaining together specific questions, yielding a clear recommendation grounded in multiple factors.

The possibilities are limitless. The better you get at writing prompts, the more tools like ChatGPT and Claude can do for you.

Experiment with the four prompt types.

Before you write a prompt, could you ask yourself what type of prompt I am writing?

Is it one of the four or a combination?

As you become more skilled, you will find new and interesting things to do using prompts, and that's when the real magic happens.

Google did a study and found the best prompts were an average of 21 words.

It’s about quality, not length.

If you need help on what to write, please just change your prompting approach to asking ChatGPT to write the prompt.

As you do this, you will see how to structure your prompts better.

Download the Google Prompting Guide for pointers here.

Happy prompting.

Marvin