In partnership with

Automate your meeting notes

• Condense 1-hour meetings into one-page recaps

• Say goodbye to post-meeting data entry

• Claim 90 days of unlimited AI notes today

Get started free

I was horrible at prompting when I joined Clay.

It took a lot of practice to get good, but can now to create some pretty cool things:

• Custom GPTs that automate repetitive analysis or data synthesis

• Workflows that auto create create variations of content to repurpose

• An automated way to create LinkedIn carousels for LinkedIn engagement

• 10-Ks deep research for strategic initiatives and why they matter

• Creating ‘This is what we heard’ type slides for sales presentations

• An AI generated Podcast on how Clay found Product Market Fit

And now… even a game that helps you practice prompt engineering.

This post will show you some fun ways you can get into and practice prompt engineering:

1. Prompting Games

2. Meme’ing

3. Image Prompting

4. AI Agent Prompting

5. Applying Simple Best Practices

6. Custom GPTs

7. Build an MVP App

Fun Ways to Practice Prompt Engineering

1. Prompting Games

I’ve found 2 really clever ones:

• Google’s Art & Culture Experiments - Say What You See game

• Lakera’s Gandalf

Say What You See

The title basically explains what you need to do. The challenge is how good or close is your prompt to recreating the image. Next, learn how to improve that prompt. It’s harder than it seems…

Gandalf

This is MUCH harder. Use prompting to get past Gandalf.

Can you trick him into revealing his secret password?

2. Meme’ing

This post would’ve come out 5 days earlier if I hadn’t found this idea on ded.ai.

Just take popular memes and then practice prompting to turn them into hyper realistic (or any style of) images.

Here’s an example for Pepe the frog. Give it a try for Soyjak, The Nordic Gamer, or Grumpy Cat. [Please send in your Grumpy Cat images - would love to see 😂)

3. Image Prompting

Everyone knows DALL-E, which I totally recommend. But, I recently discovered Google’s ImageFX.

Part of prompting is just knowing what’s possible or coming up with variations and more descriptors to try.

ImageFX has some cool capabilities where it recommends additional descriptors, variables, and sizes of the image to try.

4. AI Agent Prompting

A lot of my customer work at Clay is prompt engineering AI Agents (ours is called Claygent) to automate manual research of sellers.

My process follows these high level steps:

1. Research aspects of the topic/prompt in Perplexity.ai (like how to figure out if a company is using a specific technology)

2. Have Clay’s Meta Prompter write the prompt from a basic prompt I give it.

3. Run the Prompt for 10 companies.

4. Examine the output and steps that Claygent took to get to the answer

5. Tweak, rerun on 10, and iterate until you’re happy with the output

Some processes are multi step. For example, use 1 agent to find the 10-K report, and then use that to feed the next agent that analyzes the prompt. Below is an example of what it looks like in Clay.

👉 Check out our Prompt for analyzing & summarizing 10-K report strategic priorities.

5. Learning & Apply Simple Best Practices

These are examples from one of my LinkedIn posts - 11 AI Prompting Tips. The post goes into more details of each with bad and good examples.

We also have an ongoing series of AI Prompt Engineering Basics for Sellers:

• Part 1: The Must Learn Skill for Sellers in 2025

• Part 2: 5 Tips for Improving Your Prompts