what we need is a llm that compiles straight from twitter to cuda

→ View original post on X — @plinz