Upwarding's Substack
Upwarding's Substack
PromptLab Day 5 of 7
0:00
-1:35

PromptLab Day 5 of 7

This isn't elementary school - the magic of crappy inputs

Break the “Garbage In, Garbage Out” Habit

👋 Welcome to Day Five!

One habit I want you to break when you're dealing with large language models is the idea of "garbage in, garbage out."
Instead, think: "More in, more gold out."

You should be editing as little as possible.


🎙️ Talk First, Edit Later

Imagine, for example, you went to a career coach. They could either:

  • Ask you one question at a time to reveal your preferences,
    or

  • Let you start by just blabbing for five to seven minutes—talking about what you liked as a kid, jobs you’ve loved, or something you’ve always found interesting.

Even though that may sound like low-quality information, a skilled coach can extract powerful insights—including things you didn’t even know about yourself—to guide your next move.

Well, language models work the same way.


🥗 Let’s Try This: Diet Edition

We’re going to apply this to diet.

🔊 Here’s the exercise:

  1. Use voice mode in whatever LLM you're using.

  2. Just talk into it for five minutes about:

    • What you like to eat

    • How you’d like to eat

    • Any challenges or restrictions

    • Childhood favorites

    • Cravings, habits, routines—whatever comes to mind

  3. Then ask ChatGPT to suggest three meals you hadn’t considered that might fit your ideal diet plan. 🍽️


💡 Why This Works

The key here is to realize:
The quality of your input is less important than the quantity.

If you pair your brain dump with asking the model three clarifying questions, you’re likely to get magical results. 🪄

Try it out. Let me know what you think. 👂
And I’ll see you tomorrow. 👋

Discussion about this episode

User's avatar