Allow for specifying default types based on name

Feature
My suggestion is to allow for a default_types.py file in the root of a python project. The file should contain a dictonary with the name, type combinations. Which could look like this:

from typing import List
from database import Task
from datetime import date, timedelta

default_types = {
    "day": date,
    "name": str,
    "duration": timedelta,
    "task": Task,
    "first_task": Task,
    "tasks": List[Task],
    "task_id": int
}

As a result:

def get_task_name(task_id):
    task = Task.get(task_id)
    return task.name

would be interpreted as:

def get_task_name(task_id: int):
    task: Task = Task.get(task_id)
    return task.name

The developer would be able to overwrite the default types explicitly. For example:

def update_task_duration(task, duration: Optional[timedelta]):
    if duration is None:
        task.duration = task.default_duration
    else:
        task.duration = duration

Pitch
For a long time I have been irritated about writing type annotations for variables where it as programmer is obvious which types they are, based on its name. It felt like it was discouraging me to make smaller functions because it would lead to more typing.

As a tech lead I want to make my code base consistent and easy to read. I want to encourage developers to give objects good names.

By doing this I fell I could give a very strong incentive to give consistent names and it would result in less uninformative boilerplate and a better developer experience. And it could be adapted very gradually for existing projects one variable name at a time. Where it makes sense for people.

What do you think?

Review the proposed root-level default_types.py format and the examples showing name-based annotations and explicit overrides. Determine the intended behavior and scope for mypy, then define how completion would be verified for inferred function arguments, local variables, and explicit annotation overrides.