The arguments for typed languages seems to often take a motte-and-bailey form: The easily-defendable “motte” form being similar to “You always need to know if the thing you got was a string or a complex object.” which, when agreed to, becomes the “bailey” form: “Now that we have agreed that types are always good, please declare for every single numeric variable what width of integer they should use, and if they are signed or not. It’s the only way. You don’t want to return to the ways of chaos, do you?”

I am not entirely convinced of the validity of the latter argument.