Some questions about the implementation process

[yunyaopan]

Hi Kenneth,

I am one of the participant of the AI meetup on Wed. Thank you for sharing your work.

I have some followup questions and appreciate your insights:

1. The PRP document is very long. Do you include the entire PRP document in the context everytime you develop a new feature, or is it just for the one time experiment to rebuild the software from scratch? How do you utilize the PRP document?
2. When you develop new features, do you update both the feature spec (/frontend/specs) as well as the PRP?
3. I remembered you mentioned that for every new feature development you create a new branch. But if we commit often, we can always discard the changes and revert back to a certain commit. What is the purpose of using branches here?