EVALUATIONS ARE ALL WE NEED

Summary from the TLDR newsletter

While large language models seem intelligent and highly capable, we still don't know how to evaluate them and spot talent within them.