Comment by alganet

> I don't agree that you can pick one cherry example

Benchmarks and evaluations are made of cherry picked examples. What makes my example invalid, and benchmark prompts valid? (it's a rethorical question, you don't need to answer).

> write documentation to make it easy for LLMs to assimilate.

If we ever do that, it means LLMs failed at their job. They are supposed to help and understand us, not the other way around.