There is still too much variance in the utility of LLMs. Hopefully, the uncertainty around utility will decrease while some concrete ranked score metric increas

There is still too much variance in the utility of LLMs. Hopefully, the uncertainty around utility will decrease while some concrete ranked score metric increases. Until that point, the variance is preventing the maximization of utility of LLMs, and by extension, agentic AI.