The middle of a project is rarely tidy. That is often where the strongest connections appear.
Make before deciding
Small prototypes expose problems and possibilities more clearly than long debates.
Keep the system visible
Exploration works best when every experiment can still be judged against the same goal.