This is the most important research-backed technique. ChatGPT is bad at complex reasoning unless you force it to show its work.

Let's apply the theory. Here are three "Hero" prompts for common jobs.