Can't run kimi K3 locally? doesn't matter that much imo. this model is going to thrive in the local AI community because you'll be able to train your own distills for cheap:
→ one trained on your own traces, so your workflow fits in a small model
→ one for 3D generation that runs on a single GPU
→ one distilled on its math reasoning, so olympiad-level math fits on a 12B
→ whatever you want...
software is already becoming personalized and generated on the fly. models might be next 🚀
"Most of this is mature nodes"
Actually, the big bump in Chinese Chip exports is from SK Hynix's Dalian NAND fab & Wuxi DRAM fab, and Samsung's Xi'an NAND fab
@RereIgnuts@fbermingham From war to steel, shipbuilding, AI, and chips, he's been repeating Mao's strategy of encircling the cities from the countryside.