GOOGLE’S 9 HOUR AI PROMPT ENGINEERING COURSE – IN 1 CARTOON!
Summary:
I recently completed Google’s Coursera course on prompt engineering to continue sharpening my use of AI—for both work and play. Like many of you, I want to get better, faster, and more reliable outputs from these powerful tools. To help me remember the core principles, I’ve adopted and modified a fun and memorable mnemonic (more on that in a second). But first, here’s the big idea:
To write effective prompts, you need to include seven core sections:
Teeny Tiny Crabs Riding Fat Enormous Iguanas
T = Task T = Tone C = Context R = References F = Format E = Engage I = Iterate
Include each of these elements in your prompt, and you’ll dramatically improve your results.
Where This Came From
Much of this is based on Tina Huang’s excellent YouTube summary, “Google’s 9-Hour Prompt Engineering Course in 20 Minutes.” (As someone once said: “If you steal from one person, it’s plagiarism. If you steal from many, it’s research.” I do a lot of research!)
The Google course itself is built around five inputs for effective prompt design:
Task, Context, References, Engage, Iterate
Their trainers introduced the mnemonic:
Thoughtfully Create Really Excellent Inputs
Tina found it hard to remember—and I agreed. She came up with a more vivid version:
Tiny Crabs Riding Really Enormous Iguanas
To that, I added two key concepts from my own experience—Tone and Format—to make it more complete.
The final version?
Teeny Tiny Crabs Riding Fat Enormous Iguanas
Why Add “Tone” and “Format”?
I’ve written hundreds of prompts across multiple platforms while using AI in professional consulting and nonprofit fundraising. Here’s why those two extra sections matter:
- Tone is critical for anything involving writing. For example, when I want a professional yet conversational article (like this one), I say so in the prompt. Tone is often influenced by the Context section, especially when I open with something like:
“I am the head of a boutique consulting firm that specializes in…”
This helps the AI match voice, audience, and style.
- Format becomes essential for large-scale research tasks. I’m currently using AI to research the charitable giving practices of 200+ companies to prioritize potential donors for several nonprofit clients. In this kind of work, it’s not enough to just get raw information—I need it delivered in tables, bulleted lists, or other clean formats that are easy to review and analyze.
Gemini (Google’s AI) is particularly strong at formatting outputs into tables, and with one click, those can be exported into Google Sheets—saving tons of time.
A Quick Walkthrough of Each Section
- #Task#: Be as specific as possible. Clearly define what you want the AI to do. I often assign the AI a role or persona, like:
“You are an expert content strategist writing a blog post on how to use AI in marketing research.”
This helps focus the AI’s responses.
- #Tone#: Set the mood—professional, playful, technical, academic, or conversational. If your writing has a target voice, state it upfront.
- #Context#: Describe who you are, why you’re asking the question, and who the audience is. The better the AI understands your intent, the better its output.
- #References#: Upload or include everything relevant: job descriptions, resumes, bios, organizational goals, website URLs—whatever provides grounding for the AI’s output. The more context it has, the smarter it behaves.
- #Format#: Be clear about how you want the information delivered. Want a comparison table? A checklist? A three-paragraph summary? Tell it.
- #Engage#: Don’t copy and paste the output blindly. Engage with it. Review it. Challenge it. AI tools still make plenty of hallucinations (mistakes), so check citations, dates, and factual claims before you use them. Ask for sources. Fact-check relentlessly.
- #Iterate#: Rarely do you get the perfect response on the first try. In my experience, it takes 2–3 iterations to get what I really want.
Here’s a fun example:
I used ChatGPT to create a cartoon for a party invitation featuring my American Bully, Izzy. I uploaded a photo and asked for a cartoon of her in a chef’s hat next to a grill.
- Version 1: Looked like Izzy, but she was wearing her leather biker jacket from the photo.
- Version 2: I asked for a chef’s jacket—got that—but the dog didn’t look like Izzy anymore.
- Version 3: I asked to combine the head from version 1 and the jacket from version 2—and nailed it!
Iterating means providing clear, specific feedback. When the AI gets something wrong, tell it what and why. That’s how you teach it—and improve your results over time.
One Final Tip: Use Section Headers in Your Prompts
I have read several books and articles about AI and cannot remember where I learned this, but some recommend separating your prompts into the TTCRFEI sections and placing a “#” before and after the section title, e.g. #Task#. Experts suggest this will improve the quality of your outputs by adding clarity and structure, focus, and reducing ambiguity. By segmenting the prompt, you can make sure the AI knows what it is reading and how to use the information. For example, after reading # Task #, the AI knows the following text defines what it needs to do. After # Reference #, it understands that the text is supplementary information that it should consider. It also avoids or at least reduces misinterpretation.
Let’s Keep Learning Together
I hope this is helpful. If you’ve developed your own strategies for writing better prompts—or if you’ve tested this framework—I’d love to hear your thoughts. The more we share, the better we all get.
And Yes, I still recommend investing in the Coursera Google course.
One of my mantras as CEO of The Loyalty Group was “training without testing is little more than orientation.” My (and Tina’s) shortened version/cartoons notwithstanding, if you are relatively new to AI, I would recommend taking the Coursera course. It’s free, and if you have “fast ears” and can listen to the video portions at 1.5–2.0X speed, it takes far less than 9 hours. The two things I like best about it are:
- Completing the course requires you to show actual prompts you have written, along with improvements you have made.
- The course has an exam at the end of each of the four major sections. If you score at least 80% on the exams, you receive this certificate:
