"It’s become abundantly clear over the course of 2024 that writing good automated evals for LLM-powered systems is the skill that’s most needed to build useful applications on top of these models."

Comments