Is Protean safe to bet a real project on?
The cost of a framework that turns out wrong is not the first week, it is the years afterward spent unable to move off it. Here is the honest picture: what Protean is for, where it stops, and what you keep control of.
A real system already runs on it
A 100k-line application with 41 aggregates runs in production on Protean today. It is younger than Axon or Marten and it is pre-1.0, and its public surface changes only under a written compatibility contract.
No lock-in
Your model is ordinary, typed Python that you own, with no base classes to inherit and no generated code you cannot read. If you move off Protean later, the business logic is the part you keep.
It tells you where it stops
Protean is for systems whose hard part is the business rules. It is honest about the systems it is wrong for: raw speed, very high data volume, and table-first applications. The applicability charter says so plainly.