Interesting data, but I think there's a pretty big trap in treating usage intensity as a proxy for economic impact. The finding that early-career workers send substantially more messages than senior people could mean they're getting huge productivity gains. It could also mean they're the ones doing the work that's easiest to delegate to ChatGPT…