Chain of Thought
A prompting technique where the model reasons step by step before answering, instead of jumping to a conclusion. Adding 'think through this step by step' — or using models with built-in reasoning — measurably improves accuracy on complex, multi-part problems. For creators, it shines in planning: ask an LLM to reason through a storyboard scene by scene, or to break a complex composite image into ordered generation steps, before producing final prompts. The visible reasoning also lets you catch wrong assumptions early and correct course before wasting generations.