incredible talk by @DrJimFan on pre-training and post-training in humanoids! they scale data just like we do for LLMs, creating scaling laws and increasingly finding ways to make data collection easier