Compiling LLMs into a MegaKernel: A path to low-latency inference

LLMs pose an interesting problem for DSL designers

How Do Olympiad Medalists Judge LLMs in Competitive Programming?

Clinical knowledge in LLMs does not translate to human interactions