For Python 3.7, you can do:

    from __future__ import annotations

This will defer the parsing of annotations until after the file has been parsed, allowing you to use the actual symbol instead of a string.

TIL!

OK, that works, but it does seem dumb, or even deliberately obtuse.

The language could just as well recognize Node as it recognizes 'Node', but instead it given as a burden for the human programmer to handle.

Is it maybe because mypy is not really part of the Python language, and can't really make changes to it?

This is not mypy, it's "python-the-language": the annotations are values, and, in this case, the method annotations are evaluated when encountered - which is before the end of the class. So you are trying to use a value still not defined. It's like doing something like

    def foo() -> Bar:
       ...

    class Bar:
       ...

obviously the python interpreter cannot find a "Bar" definition during the parsing of the foo function

You're saying this is how the language works.

I'm saying the language could work differently.

The case of referring to a class inside its definition is quite different from your "Bar" case. The thing referred to is already declared, it's just not fully defined yet.

There is no logical reason I can see making it impossible for Python to honor the obvious intention of the programmer here. It seems it has just chosen not to do so. Though I'll admit I haven't thought through every weird corner case :)

PEP 563 is intended to resolve it: https://www.python.org/dev/peps/pep-0563/

As rcfox said, you can already enable this behavior, and it will be included in Python 4.0.

The language will recognise it in a future version by default, or a from future import can be used now.

mypy isn't behind (here), it's ahead.

Great!

The second issue is a bit of a judgment call too, I think. If the types were inferred you could end up with errors that you're not able to see because inferred types end up compatible. Personally, I think I'd rather write them manually so that I'm validating whether the method actually matches what it was intended to return.

I can't find a more authoritative source right now, but I also remember seeing the same thing that this SO answer says - that it was a deliberate choice not to do this so that people could incrementally add type annotations: https://stackoverflow.com/a/38775381

There are also some separate tools that can infer and generate the annotations for you. I haven't tried them personally, but the mypy docs recommends MonkeyType: https://mypy.readthedocs.io/en/stable/existing_code.html#aut...