I think I've seen an old article by a Microsoft employee saying "if you didn't change the UI, you changed nothing".
They used the calculator as an example, saying that it saw several improvements but since it always looked the same, people thought nothing had changed. Improvements included things like better arithmetics, so that (10/3-3)*3=1.
Developing a calculator that one could be proud of would be a pretty sweet job. By that I mean producing a product that consistently produced correct results, offered useful features, and had a user interface that is more pleasant to use than a four function calculator.
Did that at an earlier point in my career, and it absolutely was, but calculators pretty quickly reach a point where "less is more", and at that point it's just better to leave it alone.
I also use Python or bc as a desktop, but that really isn't the point.
People sometimes criticize the Windows calculator for perceived errors in the results even though it is a superficially simple program. What they fail to account for are the edge cases, those quirks of computation that manifest themselves for various reasons (e.g. the precision of floating point, simplifications that appear simple to a person yet are non-trivial to implement in a program). A shipping product is not the same thing as a simple calculator tutorial a novice programmer may follow.
They used the calculator as an example, saying that it saw several improvements but since it always looked the same, people thought nothing had changed. Improvements included things like better arithmetics, so that (10/3-3)*3=1.