I dug into some internal metrics and what I found is pretty striking. Their engineering teams are actually running Claude for 60% of their workload—we're talking about the bulk of daily operations here, not just occasional debugging sprints.

The numbers back it up: they're seeing 50% productivity bumps across the board. Compared to where things stood a year ago, they're hitting 2-3x efficiency gains. At some point the language changes, right? We've been calling it "assistance" for a while now, but when AI is handling the majority of your actual work output, maybe we need a better word for what's really happening.