It's an engineering problem humans and AI will solve together. The same discussion is happening in math. It's a question of meaning. In this case it's even more obvious. Fairly soon, your weather apps and your ski games are not going to be interesting. I predict within a year.
Everyone will be able to ride their own version of whatever app. They will increasingly all be good enough, so why does anyone care about your "first"? Who cares about a first when cloning is cheap?
The last 10% is always the differentiator and talking is not going to be enough. Even if it doesn't make it better for someone else, for you to care we will see people return to something like programming.
The rest of the vibe coding people will slowly drift back to being consumers, because they don't care enough and because when everything can be good enough the only thing that matters is what you want, not anyone else. If you care about something talking about it isn't good enough. You will talk to your weather app because really you don't care that much. It's rather transparently for most people a mere question of utility (do I need an umbrella today vs meteorological phenomena are so fascinating).
The optimistic part of your story isn't what you've accomplished it's that your "accomplishment" will be duplicated to the point of such eventual economy shattering irrelevancy (post bubble, post UBI) that no one will have to care.
What this synthesis of recentering human intent for activities a human wants to pursue for no good other than the intrinsic, "thing in itself" with LLM capabilities ends up looking like I'm not sure. But I am certain it will not simply be talking. Likely it will be something multi-modal, but one where users also invent their own programming languages from the axiomatic principles that most agree with their interests/intuitions.