Whisker-Wise Wisdom: CatGPT’s Ultimate Guide to Prompt Engineering

For April Fools this year, my cheeky prank involved ChatGPT and my name-sake animal! In the name of a good laugh, I unveiled my very own fantastic feline version of everyone’s favorite Chatbot. But the LinkedIn post routed unwary clicking connections to Rick Astley’s iconic music video. The innocent prank got my gears purring as I tried to turn ChatGPT into CatGPT for the joke.

This cheeky blog will walk you through how I created a CatGPT persona via prompt engineering with ChatGPT 4.0 and the hilarious results! Read on if you dare! Be sure to also check out my Ultimate Guide to Prompt Engineering if you want to go deeper.

Cat’s Outta The Bag

For anyone who clicked the link on April Fool’s Day, sorry, not sorry! Hope a little dose of “never gonna give you up” at least broke up the monotony of your day and gave you a smile!

So how exactly did I create CatGPT? Step-on came with creating a believable image to hook the readers. The secret is in the prompt engineering! There is a bit of magic and science when it comes to interacting effectively with Large Language Models (LLMs) like OpenAI’s ChatGPT and DALLE.

Understanding the mechanics of simple prompt engineering will help you get the most out of the AI chatbot taking over the world.

What Exactly is Prompt Engineering?

At the most basic level, a prompt is a sentence that describes the image or written output you want. think of it like a cheat code you are giving the Generative AI tool to help it better guess the output you seek.

ChatGPT Prompt engineering is the art of crafting effective directions for the AI to follow. Prompt engineering is your backstage pass to making generative AI do a standing ovation-worthy performance. Even beginners can get amazing ai-powered results with effective ChatGPT prompts.

It's all about finessing those input prompts with a blend of creativity and strategic thinking. Pushing the AI to not just understand your query but to read between the lines. By refining how you ask, you're essentially programming this AI Helper to deliver responses that are as rich in detail as they are in relevance. Tailored precisely to what you – the maestro of Legal Lingo – are aiming to showcase.

In a more traditional sense, prompt engineering is the strategic formulation of inputs to guide an AI model, like GAI, to generate specific and useful outputs. For Legal Professionals, it means crafting prompts that convey the Legal objectives, target demographics, and the desired tone. These enable the AI to produce relevant content and visuals.

What Are the Key Ingredients for a Good Prompt?

At the most basic a prompt should contain the following to guide the GAI in creating an output. There is no perfect order of operations in prompt writing, but there are key ingredients. With those in the mix, the results will be best.

1.     Task: What do you want ChatGPT to do?

ChatGPT is more powerful than simply googling for an answer; with guidance, it can create a myriad of outputs. Images in a variety of visual styles, blog outlines, Summaries, identifying tone, creating memo bullets, deposition questions, and PowerPoint decks, oh my. The sky is the limit, but you must guide the AI.

CatGPT Image Prompt: To engage the Dalle Image generator within ChatGPT 4, I directed the GAI to draw me a specific image. “Draw a picture of a white cat holding a tablet that says "CatGPT" and looks like the UX of ChatGPT.”

2.     Context: What unique factors do you want to be reflected?

Additional guidance to completing the task, 250 words, 6-section blog, specific factors in a matter you are ideating. The more guidance, the better.

CatGPT Image Prompt: I added constraints on background and color accents to make the image more on-brand for Technocat. “Pink and teal accents, white background.”

3.     Persona/Style: Who is ChatGPT answering the questions as?

This is where you can humanize ChatGPT's output or, in my case, ask it to respond in the voice of a cheeky feline protagonist. Identifying who you want ChatGPT to replicate in the response is how you escape a robotic, clearly AI-generated response.

Be specific when telling GAI who it is. Tell the AI to write like a document review attorney, litigation support director, legal technology influencer, or your favorite author?

CatGPT Image Prompt: To guide the AI on the Visual style I wanted the output to have I used, I provided mood words that can set the tone of the entire image. Also known as artistic styles. Since I prefer images that are modern and have a dash of eighties glam, I requested the output be “Cyberpunk and Synthwave.”

Cyberpunk: Cyberpunk is like a sci-fi movie set in a gritty, chaotic future. It's got super-advanced tech mixed with a society that's kind of falling apart. It's all about rebellion, cool gadgets, and a whole lot of neon lights.

Synthwave: Now, picture the rad music from your favorite 80s movies and games. It's like cruising down a neon-lit highway in a DeLorean.

4.     Goal/Audience: Who is ChatGPT talking to, and what is the goal?

You can also tell the Gen AI who you want them to craft their output for. Who will be consuming the output? Legal Technology lawyers, civil litigation judges, law students, ELI5. This helps refine the tone and complexity of the output.

CatGPT Image Prompt: Skipped this because was not relevant for the image.

5.     Format: What format is the output?

Do you want Gen AI to create an outline, a bulleted list, a table, a PowerPoint presentation, an image in a certain style, a video, or a photorealistic image?

CatGPT Image Prompt: And I wrapped the prompt up with a direction to create an image in 16:9 aspect ratio.

6.     Bringing it all together

The result “Draw a picture of a white cat holding a tablet that says "CatGPT" and looks like the UX of ChatGPT, pink and teal accents, white background, cyber and synth. 16:9”

Prompting Purfection in Action

When it came to converting ChatGPT into CatGPT, I focused extra attention on refining the Persona, context, and Audience. My Aim was to train the GAI to respond to any prompt, regardless of task or format, as if it were Chat GPT.

Nailing Down the Who and What

The first step was to determine what CatGPT’’s persona would entail. I started by guiding the GAI to respond with Cat puns, 12% cheeky responses in a voice like legal tech influencer Cat Casey. I also limited the task to responding to legal questions for an audience of lawyers and legal technologists.

Try, Try Again

The first iteration was still a bit stuffy and lengthy, so I refined the prompt further. Be “Funnier and cheekier, perhaps pop culture references.” When that was still a bit too professorial, I added, “Much cheekier, like 40%.”

The magic with interacting with GAI is the ability fine tune with iterations. The AI system can learn from prompting techniques to better create the output you are looking for. The key is to get comfortable with failing your way to success.

Off to the Races

Once I had the persona, the audience, and the context nailed down, I could then bring specific tasks and output formats to the Generative AI model. I directed the GenAI to continue using the framework provided in the prior prompts to guide it in responding.

Bringing it all Together

The final prompt perfection: “Generate legal answers that could be used and generated by ChatGPT a version of ChatGPT but with feline puns, 40% cheeky and in the voice of legal influencer Cat Casey, maintain this voice as you answer new prompts.”

Hello CatGPT

What was the result? Well, check out the CatGPT Natural Language Processing (NLP) and LLM-powered decision-making in action with these cheeky little prompts.

Prompt: What Is Your Advice For A Lawyer Facing Their First Doc Review With Large-Scale Discovery Keep It Short And Sweet

CatGPT

Buckle up, kitten, because you’re about to dive into the biggest yarn ball you’ve ever seen. Here’s the cat's meow of advice:

• Scratch Behind the Ears: Find the sweet spots first. Prioritize documents that are most likely to hold the juicy bits. Use technology-assisted review (TAR) to make this mountain look more like a molehill.
• Don't Get Your Whiskers in a Twist: Stay organized. Keep your documents sorted as meticulously as a cat grooming its fur. This way, you won’t end up chasing your own tail.
• Remember, Curiosity Didn't Actually Kill the Cat: Be curious and thorough, but don't get lost chasing digital mice down rabbit holes. Stay focused on what's relevant to your case.

Now, go show that document mountain who’s the top cat!

Prompt: Where Should A Curious Cat Look For Evidence In. Modern Discovery Cases?

CatGPT

In the vast, sprawling jungle of modern discovery cases, a curious cat must know the best hunting grounds to stalk the elusive prey known as evidence. Here’s where to pounce:

• Digital Trees: Emails, texts, and instant messages are the low-hanging fruit, ripe for the picking. They often hide in plain sight, full of juicy tidbits and secrets.
• The Social Media Bush: It's a wild and untamed garden where the most colorful birds’ tweet. Profiles, posts, comments, and seemingly innocent likes can reveal more than intended.
• The Cloud Burrows: Like digging for hidden treasures, cloud storage spaces hold vast nests of documents, files, and data waiting to be unearthed.
• The Device Dens: Smartphones, laptops, tablets, and even those old-school desktops are like lairs where digital footprints hibernate.
• The App Jungle: From productivity tools to communication platforms, every app potentially harbors unique behaviors, conversations, and data trails.

Remember, each of these territories requires the right tools, a keen eye, and the stealth of a night prowler. Approach with caution, precision, and the right legal warrants, and you may just snag the king of the jungle: undeniable evidence. Happy hunting, you clever cat.

If you want to learn more about the magic and mystery of prompt engineering check out my Comprehensive Guide to Prompt Engineering.