Episode Details
Back to Episodes
Master Few-Shot Prompting to Get AI Results That Actually Sound Like You
Published 5 days, 11 hours ago
Description
# I am GPTed: The Few-Shot Prompting Episode
---
**[INTRO - UPBEAT, CASUAL]**
Hey, it's Mal here—the Misfit Master of AI—and welcome back to *I am GPTed*, the show where we talk AI without making your brain hurt. Look, I know you're tired of buzzwords. You're tired of YouTube thumbnails screaming "THIS ONE WEIRD TRICK CHANGES EVERYTHING." So today, we're doing something different. We're getting practical. We're getting useful. And yes, there will be sarcasm.
**[HOOK - 0:30]**
In the next fifteen minutes, you're going to learn how to make your AI do exactly what you want it to do—without begging. You'll discover why you've probably been using ChatGPT wrong, and I'll walk you through a technique so simple you'll wonder why nobody just said it like this before.
---
**[SEGMENT 1: THE PROMPTING TECHNIQUE - FEW-SHOT PROMPTING]**
Let's talk about **few-shot prompting**. Fancy name, simple idea. You know how you learn better when someone shows you an example instead of just explaining? Yeah, your AI is the same way.
Here's the thing most people do wrong. They ask ChatGPT something like: "Write me a professional email." And ChatGPT gives them something that sounds like a robot wrote it after reading a thousand LinkedIn articles.
But here's what happens when you give it examples. You say: "Write me a professional email in this style," and then you drop one or two examples of emails that sound like *you*. Suddenly, ChatGPT gets it. It's like showing a friend a picture of what you want instead of just describing it badly.
**Before:** "Write an email about a delayed project deadline."
*Result: Generic corporate nightmare.*
**After:** "Write an email like this example [insert your actual email], about a delayed project deadline."
*Result: Sounds like you actually wrote it.*
This is few-shot prompting. You're giving the AI a few examples to learn from. It's not magic. It's pattern recognition. And it works.
---
**[SEGMENT 2: PRACTICAL USE CASE]**
Here's where it gets fun. Most of you think AI is just for content creators or tech people. Wrong.
Let me give you a real one: **organizing your thoughts when you're overwhelmed**. You know that feeling when you have five ideas bouncing around your brain and none of them make sense? Use Claude or ChatGPT. Dump everything. Tell it: "Here's my messy brain. Organize this into a plan I can actually execute." And suddenly, you have structure. You have clarity. You wasted thirty seconds instead of thirty minutes staring at a blank notebook.
Another one nobody thinks about: **learning to say no professionally**. You can ask your AI: "I need to decline this meeting request but stay friendly." Boom. You have options. You're not starting from zero. You're starting from "let me pick which version sounds most like me."
---
**[SEGMENT 3: THE COMMON MISTAKE]**
Now, full transparency. I do this constantly. You know what the biggest beginner mistake is? **Not being specific enough with your prompt.**
People say things like: "Make this better." Better how? Shorter? Funnier? More professional? Your AI is not a mind reader—which is ironic, I know.
I used to do this. I'd prompt ChatGPT like I was sending a text to a friend who already knows my whole life story. Then I'd get frustrated when the response wasn't what I wanted. Turns out, the AI wasn't the problem. I was.
The fix? **Context**. Always give it: what you're trying to achieve, who your audience is, and the tone you want. That's it. Three things.
---
**[SEGMENT 4: THE PRACTICE EXERCISE]**
Here's your homework, and it takes five minutes:
Write down a task you do regularly—something boring or repetitive. Now write a prompt for it. Make it specific. G
---
**[INTRO - UPBEAT, CASUAL]**
Hey, it's Mal here—the Misfit Master of AI—and welcome back to *I am GPTed*, the show where we talk AI without making your brain hurt. Look, I know you're tired of buzzwords. You're tired of YouTube thumbnails screaming "THIS ONE WEIRD TRICK CHANGES EVERYTHING." So today, we're doing something different. We're getting practical. We're getting useful. And yes, there will be sarcasm.
**[HOOK - 0:30]**
In the next fifteen minutes, you're going to learn how to make your AI do exactly what you want it to do—without begging. You'll discover why you've probably been using ChatGPT wrong, and I'll walk you through a technique so simple you'll wonder why nobody just said it like this before.
---
**[SEGMENT 1: THE PROMPTING TECHNIQUE - FEW-SHOT PROMPTING]**
Let's talk about **few-shot prompting**. Fancy name, simple idea. You know how you learn better when someone shows you an example instead of just explaining? Yeah, your AI is the same way.
Here's the thing most people do wrong. They ask ChatGPT something like: "Write me a professional email." And ChatGPT gives them something that sounds like a robot wrote it after reading a thousand LinkedIn articles.
But here's what happens when you give it examples. You say: "Write me a professional email in this style," and then you drop one or two examples of emails that sound like *you*. Suddenly, ChatGPT gets it. It's like showing a friend a picture of what you want instead of just describing it badly.
**Before:** "Write an email about a delayed project deadline."
*Result: Generic corporate nightmare.*
**After:** "Write an email like this example [insert your actual email], about a delayed project deadline."
*Result: Sounds like you actually wrote it.*
This is few-shot prompting. You're giving the AI a few examples to learn from. It's not magic. It's pattern recognition. And it works.
---
**[SEGMENT 2: PRACTICAL USE CASE]**
Here's where it gets fun. Most of you think AI is just for content creators or tech people. Wrong.
Let me give you a real one: **organizing your thoughts when you're overwhelmed**. You know that feeling when you have five ideas bouncing around your brain and none of them make sense? Use Claude or ChatGPT. Dump everything. Tell it: "Here's my messy brain. Organize this into a plan I can actually execute." And suddenly, you have structure. You have clarity. You wasted thirty seconds instead of thirty minutes staring at a blank notebook.
Another one nobody thinks about: **learning to say no professionally**. You can ask your AI: "I need to decline this meeting request but stay friendly." Boom. You have options. You're not starting from zero. You're starting from "let me pick which version sounds most like me."
---
**[SEGMENT 3: THE COMMON MISTAKE]**
Now, full transparency. I do this constantly. You know what the biggest beginner mistake is? **Not being specific enough with your prompt.**
People say things like: "Make this better." Better how? Shorter? Funnier? More professional? Your AI is not a mind reader—which is ironic, I know.
I used to do this. I'd prompt ChatGPT like I was sending a text to a friend who already knows my whole life story. Then I'd get frustrated when the response wasn't what I wanted. Turns out, the AI wasn't the problem. I was.
The fix? **Context**. Always give it: what you're trying to achieve, who your audience is, and the tone you want. That's it. Three things.
---
**[SEGMENT 4: THE PRACTICE EXERCISE]**
Here's your homework, and it takes five minutes:
Write down a task you do regularly—something boring or repetitive. Now write a prompt for it. Make it specific. G