Comment by Klonoar

> why are none of these gains being used to get back towards native app development?

One argument might be that, like with any LLM output, you still do need to know it well enough to know if it's good or not implementation-wise. You still need that knowledge to understand if your performance for rendering in some scenarios is going to fall off a cliff.

Web (via browsers or Electron/etc) are mostly one train of thought. When you're doing native application development using host OS frameworks, you have to actually know the framework. LLMs don't really save you from that; i.e, I could have an LLM spit out whatever flavor of Windows-specific UI I need. I have zero way of knowing whether it's correct or not.