OpenAI API is pretty easy to use. Where is the value add? Is this for developers? If so it's really not that difficult to use OpenAI and its increasingly getting easier as the models get better. I feel like you may be overstating the problem here.
Agreed that the OpenAI API is pretty easy to use. But to have a model perform a specific task involves more than just calling the API - you have to manage prompts, finite context windows, and integrations with other APIs.
We use GPT under the hood for our website summarizer for example, which involved connecting the OpenAI API to a web scraper and then handling cases where the website had more text than GPT can handle. Finally we also experimented with a bunch of prompts to make sure that it responds based on the user's needs.
> you have to manage prompts, finite context windows, and integrations with other APIs.
For many non-straightforward tasks, a dev doubts somebody else handles these problems with care. Who can guarantee that a general-purpose handling of these could be better than the specific team that aims to solve the problem?
Finite context windows are, for many complex tasks, are still a problem unsolved, only heuristics applied. You basically manage injection of cues when you hit that finite context limit. However it's not guaranteed that you'll select good cues.
Handling prompts with a vector db is still not the same thing for serving long contexts. Details of how that vector db is built in the first place can also be critical to the solution.
In other words, the product must be targeted at people whose first thought is not "I can do it better than you", but rather "I want these outcomes but I don't know how to achieve them and I'm glad you have a solution".
Yeah super fair, agreed that general-purpose handling will rarely (if ever) beat a specifically dedicated team - anything specific and critical will still require custom work. That being said, we think that there are some straightforward applications that can be annoying to solve and are reasonably generalizable.
I can see the nocode angle though.