there's an entire 3d visualization that builders have in their minds that simply cannot be condensed down to a sequence of tokens, we need much stronger ways to represent intent that doesn't just rely on one modality
we discuss this very extensively at @HanomiAI : Current 3D generative AI models are fantastic for virtual content, but hit a wall with real-world manufacturing needs. There is a significant gap between 3D gen-AI content and what is actually produced on the shop floor
(1/6) π§΅