Musk: diffusion wins on any bitstream; CoBit applies one diffusion recipe to text, speech, images, proteins
LucaAmb · x · 2026-11-03
Elon Musk argues diffusion "will obviously work on any bitstream" — for text the only question is whether first-sentence latency is worth it, and since most AI workload will be video understanding/generation, diffusion is likely the biggest winner overall, raising the compute-to-memory-bandwidth ratio.
GBatz97's team took the idea literally and built CoBit: text, speech, images and proteins are all converted to bits and processed with the same diffusion recipe.
More from Models
- DeepSeek V4.1 Flash 1017 doubles ARC-AGI-3 score to 5.81%, leads open models — teortaxesTex · 2026-11-03
- Trajectory launches post-training platform to let anyone train, eval and deploy trillion-parameter models — yenkel · 2026-11-03
- Grok tests free usage resets in exchange for training on your data — koltregaskes · 2026-11-03
- Sol 6.1 still best cost-per-intelligence worker; Fable 5.5 more well-rounded — haider1 · 2026-11-03
- Anthropic halves its Max 20x plan usage and renames it "Ultra" — EntityRemovalGuy · 2026-11-03
- Critics tear apart "Weight-Space Programming": Atari test spoon-feeds answers, Qwen case fails on Gemma — anshulkundaje · 2026-11-03