Why it's still worth trying: LLMs handle tree-like data and VLM spatial skills are rising

keenanisalive · x · 2026-10-10

Despite the difficulties, the author argues the attempt is worthwhile: LLMs model statistics of more tree-like data, and LLMs (really VLMs) are getting much better at 3D spatial intelligence. He frames it as "Blender primitives" but for smooth organic shapes rather than hard-surface models.

Related event: CMU Professor Builds 3D Dragon from 27KB of LLM-Generated GLSL Implicit Surface Code(12 posts)→

Original post →

More from AGI Musings

AGI Musings channel →