Reproducible data science over data lakes: replayable data pipelines with Bauplan and Nessie.

💩 Repeatable processes are undiscoverable. Since there are no required (or even desired) attributes and relations, there is nothing to nudge folks into a predictable workflow. You have to explain each newcomer that “we decompose Epic into Stories, estimate these Stories, and plan into Sprints” — a flat network cannot guide them.

it's hard, as Salesforce, to re-architect your entire platform to be built on top of this notion of a cloud data warehouse, especially when all these other vendors have already built tools on top of your own platform. I'm not sure how you go from point A to point B there. 

To bring projects like the one described above to fruition, new infrastructure is needed that borrows from open-source software communities and beyond. You can imagine a range of tools, including recruitment and application management, process mining to identify new discrete stages in the creative process, and sophisticated forking tools.