Your experience pretty much mirrors my own. I hate to be the 'they're holding it wrong' guy but there's certainly a lot of people out there that have no real idea how to effectively leverage AI.
That’s a problem with the tool not the people. AI is marketed literally as writing one sentence and having some app perfectly output. Just check any of the landing pages for Claude code or codex or GitHub copilot…
No, it literally isn't. I just looked at the landing pages for Claude Code, Codex, Cursor, and Copilot, and literally none of them have anything about "writing one sentence and having some app perfectly output", or anything remotely like that. In fact, just the opposite: they all make clear that they're built for ongoing collaboration with AI, and have detailed descriptions of what that looks like. No one advertising the idea that you can one-shot perfect apps with these tools.