I personally choose to work with typed languages most of the time, and I’m thoroughly convinced of their value, but acting like it is literally impossible to write a large project without types is just inaccurate.
But it would be significantly more expensive.
So I should probably preface and say that it is not possible within the budget I have (We are ~2 full time developing and maintaining everything on a 140k LOC platform spanning from infrastructure, backend, multiple frontends).
However, we do have types and don't do type coercion. Types are sound full stack db -> backend -> frontends. We don't do any funky things. We don't use SSR (as the business does not require it). We test well, but not too much. We keep a sane architecture.
We can also skip out on a lot of project shenanigans and mostly focus on our work.
We don't (yet) use AI as a part of our workflows, and frankly, I don't see we can keep this way of working with the current gen AI.
I've worked on one of those, and I would not recommend it.
There are, but they are certainly a pain in the ass to work with.
I once had to work on a semi-large Python codebase and got so frustrated that I wanted to port the code to some other language. The best part? I wrote the majority of that codebase myself, just months earlier.
I ranted about it here: https://avi.im/blag/2023/refactoring-python/
Yikes, this framework will never fare well in any decent sized or above project.
Even Typescript is problematic sometimes, as it has several forms of coercion.
I manage 2 large scale production apps using Typescript (Along with the rest of the infrastructure) with a small team.
This simply would not be possible, had I not been guaranteed that things at least type check when reading diffs from PRs.