The idea is to make libraries preserve as much type information as possible, as a principle. Once type information is erased it can't be restored. For regular application code you don't need to use those features.

But regular application code also contains libraries. Type information is useful even if you're the only user of those APIs.

My point was more related to the level of expressiveness required of a type system in order to allow a programmer to produce reliable code without getting in their way. I think TypeScript leans more towards cumbersome than useful.

For example, I'm more familiar with Go's type system, which is on the other side of that scale. It is certainly much less expressive and powerful than TypeScript, and I have found it frustrating and limiting in many ways, but in most day-to-day scenarios it's reasonably adequate. Are Go programs inherently worse off than TypeScript programs? Does a Go programmer have a worse experience overall? I would say: no.