Sequential thinking

Sequential thinking is the style where an AI agent processes information and makes decisions one step at a time, each step depending on the previous one. It shows up in Chain-of-Thought prompting, ReAct think-act-observe loops, and linear RAG pipelines. The upside is predictability and easy debugging; the downside is tunnel vision — once the chain commits to a path, it doesn't naturally reconsider alternatives. The contrast is system thinking, which treats the task as a network of dependencies instead of a line.