"It still blows my mind how companies with billions of dollars to burn get UI wrong. Like the first Windows 10."
It's much harder to build a UI that a billion people like than one that a hundred people like.
Startups have the luxury of being able to pick their customers. They can (and should) make the UI really, really tailored to their initial customers, because that's all they've got. Nail that and the initial customers evangelize the product, and bring new customers in.
Big companies do not have that luxury. Most peoples' desires are contradictory; some people like things one way, other people like it another way, some people want it configurable just the way they want it, most people get confused and leave if you put in too many knobs and configuration options. When you're doing UI design for a billion-dollar company, you have to shoot for the middle-ground that will piss off the fewest people. That's invariably something that doesn't make people all that excited either.
I was involved in 4 UI redesigns for Google Search, one as the first engineer, two as a tech lead, and one as a consultant. Usually the version we launched was a pale shadow of some of the cooler ideas that the UI designers came up with, and made nobody particularly excited, including the people who designed it. But that's the reality of serving over a billion people. Every feature cut was cut because it would make the design unusable for some fraction of the userbase; every misfeature added was added because somebody really wanted it, and in most cases we had the metrics to prove it.