As an alternative approach, I toyed with using theslice.__new__ overloads from typeshed to infer the type arguments. However, this had a big impact on the tests (since all tests involving slice expressions now needed a fixture definingslice), and I'm not sure if there are any benefits from doing it that way.

--- a/mypy/checkexpr.py+++ b/mypy/checkexpr.py@@ -5616,7 +5616,12 @@ class ExpressionChecker(ExpressionVisitor[Type]):             if index:                 t = self.accept(index)                 self.chk.check_subtype(t, expected, index, message_registry.INVALID_SLICE_INDEX)-        return self.named_type("builtins.slice")+        args = [+            TempNode(NoneType()) if arg is None else arg+            for arg in [e.begin_index, e.end_index, e.stride]+        ]+        slice_type = TypeType(self.named_type("builtins.slice"))+        return self.check_call(slice_type, args, [nodes.ARG_POS] * 3, e, [None] * 3)[0]

Sure, done!

(FWIW, that's how Ioriginally wrote it 😉. I thought to try separating this logic from that loop since I figured it may be removed soon, now that generic slices let us do better checking on slice indices than the hardcoded SupportsIndex/None checks in that loop)

Diff frommypy_primer, showing the effect of this PR on open source code:

pandas-stubs (https://github.com/pandas-dev/pandas-stubs)+ tests/test_frame.py:2246: error: Expression is of type "tuple[Index[int], slice[None, None, None]]", not "tuple[Index[int], slice[Any, Any, Any]]"  [assert-type]- tests/test_frame.py:226: error: Expression is of type "Any", not "DataFrame"  [assert-type]- tests/test_frame.py:227: error: Expression is of type "Any", not "DataFrame"  [assert-type]