Retrieval-augmented CAD 🔥🧱 This is a fantastic project by @PolySpectra that actually proposes a novel way to think about 3D generative AI. If you're familiar with the space, a lot of work in AI + 3D has dealt with how to directly model 3D representations with neural nets, from NeRFs to voxels to point clouds to polygons. But this is a cool approach where we use LLMs to generate a domain-specific language code CAD - code that can convert into 3D CAD models. So we reduce the problem to generating code instead of trying to tackle the visual space directly. And because this code is domain-specific, using RAG (with @llama_index) to index guides about the DSL will help the LLM better write the DSL to generate coherent models. Check out the blog post here! llamaindex.ai/blog/unlocking…