The starting point for the open source project plan tree recommended by this group of friends is to condense the solutions discussed with AI, decisions made, current progress, and where to continue next time into a Markdown planning tree. It doesn't make decisions for you, it makes your decisions clearer and more complete Solved two of the most important things in AI programming. Firstly, align costs. Every time a new session is opened, simply throw the planning tree into it without having to explain the background again. Secondly, it forces you to really think it through. When you want to write decisions into a planning tree, you will find that many 'almost understood' are actually 'unclear'. Forcing you to turn vague cognition into structured text is itself clarifying your thoughts. The 'big loop mode' workflow used in plan tree refers to the process of thoroughly discussing, clarifying, and documenting feasible solutions, followed by batch execution by AI. After completion, progress, evidence, and remaining issues are written back to the plan tree. First, prepare the tree species, and then let the machine chop wood. I think this is the posture that AI programming should have. Think about it, the most scarce thing in the AI era is not execution power. Execution power is given by AI, but a sense of direction, the kind of sense of direction that says' I know what I'm doing, why I'm doing it, and where to go next '. Plan tree is a tool that helps you preserve this clarity. It can reduce the deviation of AI implementation functions. If you also have the same problem, I recommend you try it