IMO it's a mistake to get too caught up in the (admittedly self-described) goal of modelling AI's economic impacts. Instead, this is a super rare and valuable

IMO it's a mistake to get too caught up in the (admittedly self-described) goal of modelling AI's economic impacts. Instead, this is a super rare and valuable look into who/what/how folks are doing with Claude across millions of conversations, nicely categorized by function and task. The economic impact data (i.e. wages) that they might overlay onto that usage data is a separate thing that -- of course -- is more subjective and likely to be part of some PR machinery about the public value of AI etc. But as to sharing the raw usage data itself - we should applaud it! What a useful window into how this stuff is being used in the real-world. Will OpenAI release similar data? Why or why not? I hope they will. It elevates the discussion for everyone, and frankly would be 'good business' if it gets people thinking about who/how AI could be used at their organization with more granularity.