No one here means quite many Discourse users, for example.
I get the frustration.
As an uneducated noob, I tried to create my first Custom GPT yesterday. Imagine my frustration that the helper chatGPT was not able to convince the custom got to not disclose the source knowledge documents to a user. Go around that a few times and I’ve used up all my Pro prompts, come back in 2.5 hours. The custom GPT is still not finished.
Because this got necroed I’ve re-read the OP and I think a significant point is being missed: the API is designed to be used for many more applications than just creating an inferior “competitor” to ChatGPT - of course you are going to fail in that pursuit and of course the price of the API is going to exceed the cost of the back-end of ChatGPT.
But that’s not what the public API is for - it’s for the myriad of more specialised uses that will be found in the ecosystem outside of a general online Chatbot. Many of those applications are discussed in this very forum.