Have you added a big feature after launch that required changes to multiple parts of the codebase?
I doubt that a codebase almost purely built with AI will survive in the long term.
If there is no developer that understands what is happening, after a while the AI will start running in circles due to a lack of guidance and new features or patches become almost impossible to push out reliably.
AI is good at writing small-ish tools from scratch. Most software development isn't like that, and people claiming it is are disingenuous. That being said, it is also true that if you are doing that kind of software development, AI will be very useful to you in the same way someone writing the code for you would be. It works, yay, you don't fully understand it, boo, but it still works. Maintaining it will be more difficult and error-prone, but it'll actually get you to build something, which is great!
That's a very tall claim, and one that I personally doubt. When it comes to refactoring old code with unclear requirements and inconsistent style, I have not seen AI do well. Of course, these things are bad, but that's the reality of most code out there - maybe you're lucky enough that you don't have to wade through that stuff.
Hardly wrote a single line of code myself.
Just crossed 372 sign ups and 73 paying users.
AI has been extremely valuable to me. Objectively speaking I have already gotten thousands of USD in value out of it.
Just my anecdote.