Raw data brokerage and refined post-training work may have different futures

A post distinguishes raw company records used for AI pre-training from refined tasks, environments and rubrics built from private data or human traces for post-training and evaluation. It predicts raw-data brokerage will become low-margin and commoditized, while refined work will gain differentiation as the frontier advances. Another user, quoting the post, adds that cutting-edge data research is critical to staying ahead of frontier labs and sustaining growth.