The Finetuner’s Fallacy: When to Pretrain with Your Finetuning Data - Best AI papers explained | Wave AI Podcast Notes