I think simple solutions work great when you are not at scale. But to achieve greater scale or quality (or in the case of computing, reliability), you need to d

I think simple solutions work great when you are not at scale. But to achieve greater scale or quality (or in the case of computing, reliability), you need to delve deeper and deeper into the nuances and make things more complex. It's kind of like when you know more about something, the bigger picture isn't as important anymore when you can focus deeply on the specific problem. In this case, he's arguing that we are just trying to create a better mouse trap, whereas in reality we've hit that 80/20 rule and to get that last bit requires making some big changes, disproportional it seems to what it took to get the first 80.