What is duck typing in Python?

Question presented to candidate: "People say Python uses duck typing. What does that mean in practice, and what are the upsides and the risks?"

What a strong answer should cover:

• Duck typing: code relies on what an object can do (its methods and attributes), not on its declared class; "if it walks like a duck and quacks like a duck, treat it as a duck".
• Python checks at the moment of use: a missing method raises AttributeError (or TypeError for operators) at run time, not at compile time.
• The built-ins work the same way: len(), for, in, + and with call special methods such as __len__, __iter__, __contains__, __add__, so any class that defines them plugs into the language; file-like objects such as io.StringIO are the classic example.
• The usual style is EAFP (try it, catch the exception) rather than LBYL (hasattr/isinstance checks first), and isinstance against a concrete class is the opposite of duck typing.
• Trade-offs: flexible and easy to test with fakes, but errors surface late and a name can match while the meaning does not; typing.Protocol lets a type checker verify the shape statically.

Code / implementation expected: Optional — a function that accepts any object with the right method, and a class that works with len()/for because it defines the special methods.