I do agree with you that it isn't required to improve readability, it definitely isn't. But in some I really do think it helps a lot if done well - in the case of the transformers library a mention in the article, the parts of the library they've begun adding type annotations too make it a lot clearer what is going into each function, and what is coming back out.

As you said, clean consistent code, and well written documentation will always be primary. I think type hints can make a nice bonus for *some* libraries.

Either way, we will see how it goes - I don't see a wide adoption of them as the standard (certainly not anytime soon). But, I think it's nice option to have.

Freelance ML engineer learning and writing about everything.