🧠 How to make custom GPT that actually works?

Hello from the heights!

It's Luke Skyward here from Smoothing the Learning Curve!

This time, I've got something really cool to share.

I've been experimenting with custom GPTs, and guess what? I stumbled upon something big.

Jetpack strapped on, visor down, let's s-s-s-s-skyrocket! 🚀

Sponsored Post (by me 😂)

I've been on an adventure this past weekend.

And I cooked up something nice!

Introducing HubermanGPT – your guide through Huberman Lab Podcast!

You can ask it anything about health and performance.

Ready to give it a try? ✨

Click HERE.

In the rest of this post I’ll be providing more details on how I created it, so keep reading!

GPTs hype

Everybody talks about custom GPTs these days. There was a big initial hype, but then people realized it’s not that easy to create a good custom GPT.

• Hallucinations

• Security breach

• AI not following the instructions

• Only 20 files that can be used in the knowledge base

It doesn’t feel like a finished product.

However, I see its potential and decided to test if I can make it actually work.

Custom Actions to the rescue

I think that with the current state of GPTs and limitations regarding built-in knowledge base you can gain advantage by creating a custom vector search.

It might currently give you have an advantage over the rest!

Actions is the feature that lets you get ahead!

This is why I decided to create a custom search API for Huberman Lab Podcast and see if it improves the quality of the custom GPT.

Thanks to this approach I’ve been able to create a database of 3000 Huberman Lab Podcast sections.

Definitely better than 20 files cap provided by OpenAI!

Why is this approach better?

𝗜𝘁 𝘂𝘀𝗲𝘀 𝗮 𝘁𝗼𝗼𝗹 𝘁𝗼 𝗳𝗶𝗻𝗱 𝘁𝗵𝗲 𝗿𝗲𝗹𝗲𝘃𝗮𝗻𝘁 𝗿𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀

It uses a tool to find the most relevant sections of Huberman Lab Podcast.

It ensures:

• 🙈 No hallucinations

• ✅ Higher chance of following the instructions

• 🔖 Every response has links to podcast sections

How to use 𝗛𝘂𝗯𝗲𝗿𝗺𝗮𝗻𝗚𝗣𝗧?

1. Ask it specific health-related question, e.g. "How to sleep better?"

2. Let the AI guide you, e.g. "I'm a 25-year-old male who has a sitting job. What are the questions that could be most relevant to me?"

HubermanGPT will give you specific advice based on the content from Huberman Lab Podcast!

You will also receive links to the most relevant podcast sections!

Look at an example advice on how exercises for someone with sedentary lifestyle 👇

If you'd like to learn from Huberman Lab Podcast faster

Make sure to check out the HubermanGPT!

Thanks for joining this episode of Smoothing the Learning Curve. I'll see you in the next issue. Until then, keep reaching for the stars! 🌌

Clear skies,

Luke Skyward