We should add another stage into that routine: research. It's no good looking at problems that other people have solved (or proven that they can't be), when we

We should add another stage into that routine: research. It's no good looking at problems that other people have solved (or proven that they can't be), when we could be looking at new ways to approach problems instead. For example, we'd know that "statically inferred duck typing" is already well researched. It's called structural typing.