That’s science :). We have now established many strategies for context-independent effective prompting via typical experimental research - see https://help.openai.com/en/articles/6654000-best-practices-f...