I don't think anyone was suggesting to go for the 'parameterized model' from the start. They could just hide the eyes while typing, as a good starting point.