I think this is generally a good point if you're using an AI to come up with a project idea and elaborate it.
However, I've spent years sometimes thinking through interesting software architectures and technical approaches and designs for various things, including window managers, editors, game engines, programming languages, and so on, reading relevant books and guides and technical manuals, sketching out architecture diagrams in my notebooks and writing long handwritten design documents in markdown files or in messages to friends. I've even, in some cases, gotten as far as 10,000 lines or so of code sketching out some of the architectural approaches or things I want to try to get a better feel for the problem and the underlying technologies. But I've never had the energy to do the raw code shoveling and debug looping necessary to get out a prototype of my ideas — AI now makes that possible.
Once that prototype is out, I can look at it, inspect it from all angles, tweak it and understand the pros and cons, the limitations and blind spots of my idea, and iterate again. Also, through pair programming with the AI, I can learn about the technologies I'm using through demonstration and see what their limitations and affordances are by seeing what things are easy and concise for the AI to implement and what requires brute forcing it with hacks and huge reams of code and what's performant and what isn't, what leads to confusing architectures and what leads to clean architectures, and all of those things.
I'm still spending my time reading things like Game Engine Architecture, Computer Systems, A Philosophy of Software Design, Designing Data-Intensive Applications, Thinking in Systems, Data-Oriented Design, articles in CSP, fibers, compilers, type systems, ECS, writing down notes and ideas.
So really it seems more to me like boring people who aren't really deeply interested in a subject use AI to do all of the design and ideation for them. And so, of course, it ends up boring and you're just seeing more of it because it lowered the barrier to entry. I think if you're an interesting person with strong opinions about what you want to build and how you want to build it, that is actually interested in exploring the literature with or with out AI help and then pair programming with it in order to explore the problem space, it still ends up interesting.
Most of my recent AI projects have just been small tools for my own usage, but that's because I was kicking the tires. I have some bigger things planned, executing on ideas I have pages and pages, dozens of them, in my notebooks about.