And certainly many people who try this approach like it. It works well for a certain kind of team. But this approach over-optimizes for the present, and discounts the future. It's going to get leapfrogged by technologies that are already starting to land.
As the web platform matures as an application runtime, it's starting to become a true compilation target.
Q: Sure, almost everybody transpiles something these days, how is this news?
A: There is a critical tipping point that we haven't hit quite yet. A true compilation target is one that you don't need to understand to use.
We have seen this all before. There was a time when you couldn't be a professional programmer without knowing C. Your higher-level language was implemented in C, and it would break a lot, and you would need to debug the C layer. Further back, you couldn't be a programmer if you didn't know assembly, because C compilers were too leaky an abstraction, and debuggers were not good enough to debug-while-thinking-in-C, so you had to debug-while-thinking-in-assembly.
Every time this process has repeated in the past, the old guard who were already experts resisted the change, and denigrated the newcomers who had the gall to call themselves programmers even though they didn't know assembly! But the newcomers win in the end, because there are so many more of them, and the nexus of activity moves up the stack.
It's the nature of software to make last year's cutting-edge problem into next year's moderately-hard problem into something a precocious child can do a few years after that. Realize there are many of us out there working toward the point where a kid in a weekend can reproduce your $X million app by plugging together reusable components. It's not a question of if, but when.
And we're going to succeed because we control all the language semantics and choose how they map into underlying Javscript (or webassembly) semantics. That ability to choose gives us immense leverage over the future.
The asteroid is coming.