Honestly, I think it's nothing special to say that certain technologies have an end point. We had lots of advancements in single core CPUs but eventually more

Honestly, I think it's nothing special to say that certain technologies have an end point. We had lots of advancements in single core CPUs but eventually more than that was necessary, now the same is happening with monolithic chips vs chiplet designs. Same for something like HTTP/1.1 and HTTP/2 and now HTTP/3. Same for traditional rendering vs something like raytracing and other approaches. I assume it's the same for typical spell checking and writing assistants vs LLM based ones. That it's the same for typical autocomplete solutions vs LLM based ones. It does seem that there weren't former technological solutions for images/animations/models etc. (maybe the likes of Mixamo and animation retargeting, but not much for replacing a concept artist for shops that can't afford one). Each technology, including the various forms of AI have their limitations, with the exception of how much money has been spent on training the likes of models behind ChatGPT etc. Nothing wrong with that, I'll use LLMs what they're good for and look for something else once new technologies become available.