Dario Amodei predicted last year that AI would eliminate 50% of entry-level white-collar jobs within years. Unemployment could hit 10-20%. He's since moved closer to the Jevons Paradox, the idea that automation actually creates more demand and more work. Altman said last week he was "pretty wrong" about displacement (see Axios image down below). Anthropic co-founder Olah, in turn, repeated Dario Amodei's warning to the Pope a few days ago. Meanwhile Yale's Budget Lab has been tracking the actual US labor market monthly since ChatGPT launched. Every single update: no meaningful shift in occupational mix. No acceleration in job losses for AI-exposed roles (Image 2 below). Deutsche Bank coined a term for it in January, "AI redundancy washing." Companies blaming AI for layoffs they'd make regardless. So where does that leave us? Amodei could still be right. Exponentials look flat until they don't - the steam engine existed for decades before it restructured entire economies. AI capabilities are compounding fast. The labor data just hasn't caught up yet. Or maybe it won't, at least not in the way anyone predicted. We genuinely don't know! And this is precisely my point here. What we do know is that right now the gap between AI capability curves and actual employment data is wider than it's ever been. And that gap is the only honest starting point for this conversation. However, it was important to me to take a look at the status quo and see where we stand and how the different perspectives and assumptions are developing.