Yeah from an overall perspective codegen still has to happen somewhere. However the benefit here is that the responsibility can be moved to the owner of the API vs. the consumer (like is standard practice today with operation-based codegen).
Consumers can then update a pre-generated client at-will (assuming there are no breaking schema changes) like we do with REST API clients.
Consumers can then update a pre-generated client at-will (assuming there are no breaking schema changes) like we do with REST API clients.