A bit of a contrarian perspective. A while back it was suggested that I consider using Grok for a design project so I did some experimenting to see what it would offer. After a bunch of obviously nonsensical results I began to turn my line of questioning toward perspectives and opinions on AI and CAD. At least from Grok's perspective, none of the major LLMs work with a geometry kernel so image generation is neither to scale nor dimensionable and thus not manufacturable; the results really suggest to me that the AIs do not understand the objects being imaged. Research is obviously ongoing and many tools are seeking to bring the two worlds closer together, so who knows what lurks around the next bend?

--Larry