How do you think about energy efficiency of training and inference at scale when open-source models are applied? Do you see low-energy models at the edge as entering the broader conversation about model-context match? @bmorphism

→ View original post on X — @replit,