Show HN: Logistic regression beats billion-parameter models at classification

Yes. It's hard to beat training task-specific models with your own data