I take all the documentation from the platform that I’m working with, and I either create a custom GPT with it, or if you don’t have a paid account, just upload it as you start the session. Do something like this:

Here’s the documentation for Simple Practice. You are a Simple Practice consultant, I’ll ask you how to accomplish something using Simple practice, and you’ll provide concise instructions on how to accomplish that based on the documentation. Ask clarifying questions to improve your answers. Include code snippets where appropriate.


Then, here’s the kicker, expect that 2/3rds of the answers are going to be wrong. When that happens, tell GPT what is happening, and usually you’ll get closer and closer.

I also like to use a versioning system (git) that way I can easily go back to when things were working better in case I break things, it also allows me to easily see what GPT is changing, and I can decide if I actually want those changes.

ChatGPT is like having a very junior developer on your team, great for simple copy pastas and ceremonial coding structure, but lousy for actually understanding the why behind a requirement, or the implications for getting code working to the overall system.