ok so you say that LLM-generated code "works" and/or is "useful." fine. but AI-generated images and video also "work" and are "useful"—they depict, evoke, convince—broadly, they do many of the things that we expect images and video to do. (right-wing social media is proof of this imo.) indeed I feel like the fact that AI-generated images "work" in this fashion is *part of the reason* that people dislike them. yet people cite the "usefulness" of LLM-generated code to *justify* its use. weird!