Noob question: At first glance, Flyte and Metaflow seem very similar to me. • Both use decorators to mark an isolated “step/task” in a DAG/flow. • Both use decorators to configure Kubernetes resource allocation parameters for said “steps/tasks.” In the materials I’ve digested, Metaflow augments: • the feature that anything committed to a class variable — i.e. self.model = … — becomes “versioned”/retained. • out-the-box “experiment-tracking”-esque visualizations. Or perhaps you could also say this as a differentiator: • Metaflow’s design and features are intended more for operational ML situations • Flyte’s design and features are intended more for general orchestration situations Okay, sorry for the verbosity, what am I getting right/wrong/missing? Hot takes welcome :)

here's a recent thread touching the topic