For now, there is a lot of novelty at work, AI and being able to make computers realize ideas for the first time.

Let's see if it holds up in the long run. Building a prototype that makes feel-good hormones run is one thing. Slogging, or slopping, through the 20 remaining pareto points to a more polished experience is another thing. I suspect people will go back to full featured apps they don't have to constantly maintain and update.

I'm sure someone will chime in AI will solve all this, but I'd pre-counter by saying that the inherent limits of LLMs and agents based on transformers preclude this reality from materializing