Context is king

The director at my client asked if I knew anyone familiar with recent HIPAA electronic data interface changes. I asked around and got a bunch of "No clue" responses. So I said, "Let's see what ChatGPT says." 

You might wonder why I picked GPT over Claude. It's not necessarily a model preference, but a token-optimization preference. I use Claude to maintain my health conference directory, so I don't normally waste non-related queries and want to preserve my 5-hr usage as much as possible. 

I gave GPT a simple prompt, and it quickly returned all the info that the director wanted. I had it put it into a nicely formatted PDF, and she loved it. She also asked me what prompt I used, which I gave her. This is where the topic of the day fits in. 

Context is king. She has the free GPT plan and barely uses it. If she entered the same prompt, she would get a response quite different from the one I got. I've been using GPT for a while, and I've trained it well enough to get the kind of response I want. So, even though people say, "Oh well, anyone can ask GPT or Claude." Sure, that's true, but context is king.