There's a lot for us to learn about LLMs. Claude 4's prompt guide notes that when you really want a good output it might help to encourage the model, and provides the example

"Don't hold back. Give it your all."

We don't really understand how they work, do we.

https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/claude-4-best-practices#example-formatting-preferences