OpenAI recently announced, “You can now create custom versions of ChatGPT that combine instructions, extra knowledge, and any combination of skills.” Read about it on their blog here.

I finally had a chance to jog over there and get my hands dirty. I created a tarot card bot, Mystical Oracle, that will give hopeful, positive readings. It will also generate an image of the drawn card–it’s painfully slow, so I may nix that. There will be a marketplace that will allow you to make some money from your creation. I’ll follow this story closely and will update here.

But BEWARE! The author I follow on Forbes, Dr. Lance B. Eliot, raises some flags and advises caution. He indicates that a bad actor could access your personal information and/or figure out the “secret sauce” that makes your GPT so great. However, he does offer some solutions but will get into more technical detail in an upcoming article.

First and foremost: DO NOT ENTER ANY PERSONAL INFORMATION into the foundation of your bot. Underscore, underscore. To secure your secret sauce, he gives examples, but inevitably points to the art of prompt engineering! Dear reader, if you came from Promptmaru.com you know how to do this. If you are new to this blog, then go to Learnprompting.org. It’s a great place to start your prompt engineering journey and it’s free! Thinking of creating your own GPT? Then, I highly suggest you read Dr. Eliot’s post here.

I’m going back to configure my Oracle to perform better readings and implement some of the guardrails suggested by Dr. Eliot. I’ll tweak more once he publishes the technical part 2 of his article. I’ll give you folks the first chance to play with the Mystical Oracle to see what the cards hold for you. 😉