Tri Dao

I’ve often wanted to pair a smart orchestrator / architect model with a super fast worker model. Diffusion LLM could fill the worker role https://t.co/bwn2y4Ky26