Trust me, it's better this way
Definitely a huge opportunity right now is to take an existing growth channel and build a small UI layer on top of GPT-3 and market it.
In terms of defensibility - there is none. And you're still dependent on the API layer - but people will move to build more models to compete if the API layer is charging too much, or build more competitive and similar apps until it gets saturated.
Who will be the Pets.com of AI?