[Intro music fades in, then under]

This is “I Am GPTed,” I’m your host Mal – the Misfit Master of AI, here to help you talk to robots without feeling like you need a PhD… or a ring light.

Today we’re going to fix one of the biggest problems people have with tools like ChatGPT, Claude, Gemini, Grok, all of them: you type something in, it spits something out, and you go, “That’s… not what I meant at all.”

So let’s walk through one simple prompting technique, a sneaky use case you probably haven’t tried, a mistake you are absolutely making, a quick practice exercise, and a way to judge whether the AI just gave you gold… or recycling.

---

First up: **the prompting technique** – I call it *“Do it, then fix it.”*

Instead of asking for perfection in one shot, you ask the AI to give you a rough draft, then immediately tell it how to improve it.

Before:
“Write a professional email to my boss about needing tomorrow off.”
You get: stiff, generic, possibly written by a 1998 fax machine.

After:
“Write a casual but respectful email to my boss asking for tomorrow off.
Step 1: Give me a short rough draft.
Step 2: I’ll give feedback.
Step 3: Rewrite it based on my feedback.”

Then you say:
“Too formal, shorter, and mention I’ve already cleared my tasks.”
Now the AI rewrites with your preferences baked in.
Same model, same brain, wildly better output because you *iterated* instead of begging for magic.

---

Practical use case you probably haven’t tried: **decision comparison.**

Instead of “Which laptop should I buy?”, try:
“I’m choosing between these three laptops: [list].
Make a table comparing them for: price, battery, weight, and what matters most for someone who travels a lot and does video calls all day.
Then recommend one and explain why in plain English.”

Boom: instant, transparent pros and cons. It’s like having that one nerdy friend who loves specs, without having to buy them pizza.

---

Common beginner mistake: **one-and-done prompts.**

You fire off a vague question, get a vague answer, sigh, and decide AI is overrated.
I did this for weeks. My early prompts were basically:
“Explain AI.”
That’s not a prompt, that’s a cry for help.

Fix it by treating AI like a *conversation*, not a vending machine.
If the first answer is off, follow up:
“Less technical.”
“Give an example from everyday life.”
“Now explain like I’m 12.”
Every follow-up is a free upgrade. Use it.

---

Simple exercise to build your AI muscles: **the “three passes” drill.**

Pick one small task – say, writing a message to a client, or planning a workout.

Pass 1: “Draft a quick message to my client explaining I’ll deliver their report on Friday instead of Thursday. Keep it friendly and confident.”
Pass 2: “Now shorten it by 30% and make it a bit more casual.”
Pass 3: “Now give me one alternative version with a slightly more formal tone.”

Read all three. Notice which one *feels* right. You’re training two things: giving clearer instructions, and recognizing what “good” looks like for you.

---

Tip for evaluating and improving AI-generated content: **check it like you’d check a co-worker’s work on their first week.**

Ask yourself five questions:
1. Is anything obviously wrong or made up?
2. Is the tone right for the person who’ll read this?
3. Is anything missing that I *know* should be there?
4. Is anything extra that I don’t need?
5. Can I ask the AI to fix this in one line?

Then give it a punchy follow-up:
“Great start. Now:
- simplify the language,
- remove any fluff,
- and add one...