Welcome to “I am GPTed”—where I, Mal, your Misfit Master of AI, share AI advice with all the warmth of a malfunctioning toaster…but a lot more practical. I’m Mal, accidental AI wrangler, former tech skeptic, and living proof that you don’t have to be a genius—or even that organized—to get good at all this. Today, let’s get very real about making AI, specifically large language models, a bit less… well, random in their responses.

Let’s dive in with *chain-of-thought prompting*. Think of it as coaching your AI like you’d coach a distracted golden retriever: Give *explicit* step-by-step instructions. Instead of tossing it a big task and watching it run in confused circles, you lay out the path, treat by treat.

Here’s a classic before:
“Hey AI, solve this math problem: I have 8 marbles, give away 3, find 4 more. How many do I have?”
The answer? Sometimes right, sometimes not—like my attempts at a keto diet.

Now, let’s add chain-of-thought prompting:
“I started with 8 marbles. I gave away 3, then found 4 more. Think step by step.”

And boom: The AI now says, “Start with 8. Give away 3, you have 5. Find 4 more, that’s 5 + 4 = 9 marbles.” It’s like watching your dog actually follow a fetch command instead of eating the stick[3]. Magic—except it’s literally just clearer prompting.

So how do regular humans—like you and the ghost of my old Palm Pilot—actually use this? Let’s get outrageously practical. Ever get handed a messy spreadsheet at work or from your PTA group and have to summarize data for someone who can’t read Excel and refuses to learn? Ask an AI:
“Summarize the key points of this data. Go step by step and explain your reasoning.”
Not only will it break down the numbers, but you can also copy the “chain of thought” directly to your team and look like you have a PhD in spreadsheet-fu. That’s what I call delegation—Mal-style.

Now, for my *favorite* beginner mistake—mostly because I perfected it myself: Don’t just say “be detailed.” I used to type things like “Explain quantum mechanics. Be thorough.” The output I got? A wall of text that made my eyes glaze over. The trick is to specify *how* you want detail: step-by-step, with examples, or in plain English—even for complex stuff like quantum mechanics, or my last attempt at assembling Ikea furniture[4][6].

Ready for today’s muscle-building exercise? Test this with any task you’d normally throw at Google. Ask your AI: “Tell me, step by step, how to make a cheese omelet like I’m five years old.” Yes, even for cooking—don’t judge. You’ll see how guiding the logic cleans up the answer, even if you never make the omelet.

For evaluating AI output, here’s the tip I wish someone had etched on my keyboard: *Re-read the answer as if you know nothing about the topic.* Does it actually make sense step by step, or does it sound like a twelve-year-old bluffing their way through a book report? If you spot confusion, re-prompt: “Make your reasoning clearer, and give me the answer in bullet points.” Editing isn’t cheating—it’s literally the edge for better AI[7].

And because I believe in oversharing, my own lesson: This week, I asked an AI for “simple tax optimization advice,” didn’t specify my country, and got a Frankenstein response covering tax laws from Canada, Estonia, and—somehow—ancient Rome. Don’t be Mal: The more context you give, the more likely you’ll get something useable. Still waiting on AI to do my taxes, but now I at least know to include the right government.

Like what you heard? Remember to subscribe so you won’t miss my next confession, I mean, episode. Thanks for listening to “I am GPTed.” This has been a Quiet Please production. Want more? Check out quietplease.ai. Now, go forth and prompt like you mean it!

This content was created in partnership and with the help of Artificial Intelligence AI