GPTs need a way to manage versions, some form of a limited repository would be super helpful for developers. I know for simple agents it doesn’t matter much and you can copy and paste instructions, but things can go wild with increasing complexity.
With support for commit messages and diff is more than enough.
I agree.
The simplest and most effective way is just to use GitHub. As a developer, you simply manage all your GPTs as regular GitHub repositories and the GPT framework would pick up any versions automatically for “deployment” for the runtime.
That would ease the burden on OpenAI and simplify the development process for developers.
The idea has been fully demonstrated by the cloud hosting company render.com, so it is a proven solution. You do a single git push when you are ready, and your production site goes live right away.