5 Lessons in AI Precision: Unpacking Prompt Engineering with a PB&J Sandwich Challenge

In a world increasingly driven by artificial intelligence, the precision of our communication shapes the effectiveness of AI outcomes. This concept is humorously and effectively illustrated in a video featuring Josh Darnit and his kids, who navigate the intricacies of making a peanut butter and jelly sandwich. Here are five lessons in AI precision gleaned from their experience:


Lesson 1: Specificity is Key

When Josh’s kids told him to “make a sandwich,” the result was confusion. This teaches us that AI needs detailed instructions to produce accurate results.

– Good Prompt: “Place two slices of bread on a plate, open the peanut butter jar, and spread a layer on one slice.”

– Bad Prompt: “Make a sandwich.”


Lesson 2: Clarity Prevents Errors

Ambiguous commands like “add stuff to the bread” led to incorrect actions. Clear and unambiguous instructions are crucial for AI to follow commands correctly.

– Good Prompt: “Open the jelly jar, use a spoon to extract jelly, and spread it on the second slice of bread.”

– Bad Prompt: “Add stuff to the bread.”


Lesson 3: Sequencing Matters

Order of operations is vital. Telling Josh to “put the breads together” before applying peanut butter or jelly doesn’t work. Similarly, AI needs to follow a logical sequence of steps.

– Good Prompt: “After spreading peanut butter and jelly, press the two slices together with the coated sides facing inward.”

– Bad Prompt: “Put the breads together.”


Lesson 4: Assume No Background Knowledge

Josh acting confused when told to “make it yummy” shows that AI doesn’t understand subjective terms or assumptions. Instructions should be as if explaining to someone with no prior knowledge.

– Good Prompt: “Spread the peanut butter evenly across the entire surface of one slice to ensure every bite has flavor.”

– Bad Prompt: “Make it yummy.”


Lesson 5: Iteration Improves Outcomes

The video shows that refining instructions leads to better sandwiches. In AI, refining prompts through trial and error leads to more accurate and reliable results.

– Good Prompt: “Review the sandwich to ensure it has equal amounts of peanut butter and jelly on each slice before closing it.”

– Bad Prompt: “Just put them together.”


This video does more than entertain; it offers valuable insights into the precise nature of human-AI communication. The lessons from Josh and his kids’ sandwich-making challenge underscore the importance of clear, specific, and well-ordered instructions in achieving desired outcomes with AI.

Leave a Reply

Your email address will not be published. Required fields are marked *