I endorse this: build a monolith first, then decompose into microservices.
Another argument that +Martin Fowler
doesn't make, but I would, is a kind of reverse Conway's Law: the architecture of the software will be reflected in the communication structures of the team. When starting a project you don't want
to let people or groups go off and work in isolation.
In some sense the peeling off of microservices is a natural maturing of a project, where you think about new aspects of the project in relation to how they are different than how the project is currently. But when starting a project there isn't the shared understanding of the existing product, and you need to actively build that shared understanding.