lol. Yup. It's even easier than that under Python ≥ 3.12, where you no longer need to explicitly declare your type variables (i.e., TypeVar() objects). Instead, just subscript your types by your type variables and away you go: e.g.,

from beartype.typing import Generic

class Something[_T](Generic[_T]):  # <-- look, ma! no "TypeVar"
    data: _T
    
    def __init__(self, data: _T = None) -> None:
        self.data = data
        
s = Something[int](5)
s.data # Type checker shows int! Even when I obscure my int value!

The runtime type-checking issue, of course, is that no runtime type-checker (including @beartype) propagates generic subscriptions. That is, @beartype currently fails to "copy" the int type hint subscripting Something[int] into either the data: int field or the def __init__(self, data: int = None) -> None: constructor. The int is entirely ignored at runtime.

In fact, the core typing.Generic superclass goes out of its way to destroy the int type hint subscripting Something[int] at runtime. Any runtime type-checker (including @beartype) trying to type-check generic subscriptions thus needs to do CrazyStuff™ like monkey-patching the core typing.Generic superclass (e.g., to record generic subscriptions in some sort of instance variable like __beartype_child_hints__). Probably not a great idea... but what you gonna do, huh? 🤔

@beartype: We didn't make crazy. We just monkey-patch the crazy away.