>100 trillion token analysis of reasoning model usage over time Full piece from @MaikaThoughts, @AnjneyMidha, @xanderatallah, and @cclark: