It's more about knowing the tricks to get llms to give you the output you want. However, there's no reason to think any trick would be relevant even in a year.

It's more about knowing the tricks to get llms to give you the output you want. However, there's no reason to think any trick would be relevant even in a year. As llms get better, why wouldn't we just have them auto rewrite prompts using appropriate prompt engineering tricks?