ML Pipeline ETL

When you're building data applications, the kind of like these dependency graphs is such a fundamental primitive that you should be doing it from square one effect. You know, we view diacs for something that scales all the way from single player mode to full-blown ML pipeline ETL. We think that actually should be more of a primitive abstraction rather than to be something where only very late in the game, you decide like, oh, now we need to do orchestration.