1. I don't have the mental capacity to keep every single function's argument/return shape in my head, or to manually check it every time I make a change. Unit tests can't deliver 100% code coverage in practice.

2. Nor do I want to perform refactorings with stone age tools like s/setFoo/setBar/g. Setting up type information lets my IDE understand which calls to .push() deal with a native Array and which ones deal with my own class, so it can rename the latter ones when I ask. I can also use tools like "Find References" and avoid false positives.

3. I'm not a one-man-band. My coworkers need to deal with this project too, and new developers need to be introduced to it from time to time, and types serve as documentation and guard rails for them much better than jsdoc or regular comments. (This also serves as a significant barrier against using anything more esoteric like Elm, because nobody around would be familiar with it. TypeScript adds just enough syntax on top of regular JS to keep JS users comfortable.)

---

I do use `any` (and `unknown`), I have no doubts that an edge case can crash my app because I didn't validate something, and I never expect third-party code to work flawlessly whether it has types or not. Rejecting TS completely because "but run time loopholes" is throwing the baby out with the bathwater (or to put it in a more hyperbolic way, being an anti-vaxxer because vaccines are not 100% safe from side-effects). TS and types are an additional safety net/force multiplier†, not a silver bullet. (That said, what is a silver bullet? Because I'd sure like one.)

---

† Only applies to a project that has passed its initial rapid prototyping phase. During wild prototyping rides, types can indeed slow you down. But that's really the same debate as RDBMS vs NoSQL.