There are people who don’t use the Plan function when they write Claude Codex, but that’s more important than thinking.
Imagine a kick-off meeting with another team or external company. No team starts right away without a kick-off. If each team starts off on its own without a direction, then the cost of subsequent shifts will be much higher.
If you don’t communicate enough with AI from the beginning and create too many things at once, then subsequent corrections tend to take place only within the framework of the results that have already been created.
So until the desired results are realised to be visible, it is better to continue the conversation in the Plan phase than in the actual work.
There are many advantages to writing a plan.First, the loss is small when you get the wrong direction.When you create a code or document, and then find out that the direction is wrong, it costs a lot to revert, but in the Plan phase you can correct the direction with a few words.
AI often interprets difficult requirements on its own, and if you first look into the interpretation at the Plan stage, you can correct the contradictory assumptions before the actual work begins.
Finally, Plan also acts as a natural control over the scope of work.When you start right away without a plan, AI often creates more than it needs, but if you clearly define the scope at the Plan stage, the results are more definite.
After all, Plan is not about a slowing down process, but about an investment that reduces the time to switch later.
Loading comments.