[gmerc]

All I can say is lol. DeepSeek showing 3 order of magnitude efficiency gains over the performative capital furnace that was training and inference absolutely moved the bar here.