Is OpenAI's success because of the easy API they provide?

Question

For almost everything, they have a working API. You want chat completions? There's an API for that. You prefer more? Use GPT-3.5's API. You want embeddings for classification, search, etc.? There's the Ada API for that.On top of that, they have provided good Jupyter Notebooks as examples that you can use right now. You can use Numpy arrays and FAISS for semantic search, but why do it when the OpenAI's API is a few LoCs away?When people bring up LLaMA and other alt-GPTs, my first reaction is: Okay, but can a high-schooler get it up and running for their side project? And usually the answer is "no".

sharemywin · Accepted Answer

It's another one of those 8 year overnight successes.
They're not successful yet, but I would attribute their success to waitlist and the login to the chatbot. plus the brand they built up over the last 8 years.
most of the other chatbots failed because they didn't force people to login which caused people to be ridiculous.
FOMO was a big part. And a really good easy to use product. that demoed the api.
And the people behind it already had big reputations in tech.

qup · Answer

They were first, they have the best product right now, it's cheap enough that you can easily "pay to not have a headache," I think those are what people like about their offering.
As soon as something as good as GPT-4 can be self-hosted, I'll be right in there--as a full-stack solo dev it's kinda right up my alley. But maybe GPT-5 will be out...
It's hard enough to keep up with everything being released, I don't need to fiddle with the details, too.