Peanut butter & jelly sandwich

With AI tools like ChatGPT and image generators making headlines daily, it’s easy to think the machines are doing all the heavy lifting now. But AI is powerful—not magical. It only works as well as the instructions it’s given. That’s where prompt engineers come in.

Let me explain.

🥪 Prompt Engineering Is Like Making a PB&J in Second Grade
Remember learning to write instructions in elementary by explaining how to make a peanut butter and jelly sandwich? The teacher would follow your steps literally. If you skipped “take the bread out of the bag,” you’d end up with peanut butter smeared on a plastic-wrapped loaf. Funny? Yes. Functional? No.

In this scenario, the teacher is the AI—and you’re the prompt engineer.

AI doesn’t assume. It doesn’t infer. It just does what it’s told. And if your prompts are vague, out-of-order, or incomplete, you’ll get results that are confusing, broken, or just plain weird.

🍽️ Now Try Explaining Beef Wellington
Now imagine asking that same second grader to describe how to make Beef Wellington. You might end up with a soggy meatball in a crescent roll.

That’s AI when it’s given a vague prompt for something complex—like “build me an app that connects my CRM to my phone.” Sure, it might produce something. But without technically precise instructions, the output will likely be a half-baked mess. That’s why prompt engineering isn’t just about creativity—it’s about clarity, structure, and technical awareness.


🧠 You Don’t Know What You Don’t Know
Here’s the tricky part: if you’re not trained in how things work behind the scenes, you won’t even know what your prompt is missing.

Ask an AI to “sync my CRM with my email and phone” and it might give you some starting code. But it won’t warn you about:

  • API authentication
  • Encryption
  • Device compatibility
  • VoIP setup
  • Error handling
  • Calendar syncing

A good prompt engineer knows to include those things—or how to guide the AI step-by-step to get there.


🖼️ The Nose With No Septum
Real story: my husband used an image generator to create a toy animation game for our toddler. Simple prompts but not the expected result. Here’s a couple examples.

Prompt: A cartoonish nose against a bright blue background. The nostrils flare as the nose breathes in, and air gets sucked in.

Prompt: Dynamic rotation: An orange sports car with sleek lines and shiny wheels. The wheels start to spin as the car rears back, then tears off, leaving skid marks.

Why? Because AI doesn’t “understand.” It just guesses based on patterns in its training data. Without precise prompts, it generates chaos—or creepy noses.


🛠️ Breaking Down Problems Is a Prompt Engineer’s Superpower
AI shines when given small, specific tasks. Prompt engineers break down big goals—like “design a mobile app”—into clear, granular instructions.

They think like architects:

  • What are the components?
  • What are the dependencies?
  • What does success look like?

AI can help build faster—but only when it’s told exactly what to do. That precision? That’s not AI. That’s prompt engineering.


👨‍🏫 Even Harvard Gets in on the PB&J Analogy
In Harvard’s famous CS50 course, Professor David J. Malan demonstrates how literal computers are by asking students to explain how to make a peanut butter and jelly sandwich. He follows their instructions word-for-word—often ending in disaster (and jelly everywhere).

It’s hilarious, but it makes the point: machines don’t “get it.” They follow steps. And if your steps are off, the outcome is, too. Sound familiar?


🤖 AI Is a Tool, Not a Mind Reader
AI is a powerful assistant—but it needs a skilled guide. Prompt engineers know how to frame problems, structure tasks, and steer AI toward useful outcomes. It’s not about knowing everything—it’s about knowing how to ask.


🎯 Final Takeaway
AI won’t replace your job. But prompt engineers? They’re the translators, the architects, the directors behind the scenes. The future belongs to those who can speak fluently to machines—and that’s what the career is all about.

So next time someone says AI is doing all the work, remind them of the peanut butter and jelly sandwich. AI won’t work or scale unless someone knows how to give it the right instructions.

Here are a few more of the weird animations that were generated for my son’s game. Enjoy!

https://www.canva.com/design/DAG0FxxNHxI/8InfcRPb5S3vdzoni3mtAw/view?utm_content=DAG0FxxNHxI&utm_campaign=designshare&utm_medium=link2&utm_source=uniquelinks&utlId=hd8500be1fa