Rather than just watching, I turned to the keyboard and had a crack at implementing the same Roman numeral conversion in Go.
What I hadn’t appreciated is how poorly defined Roman numerals are. Try implementing an `isValid(roman string) bool` method if you want to see what I mean. There seems to be very little agreement on what constitutes a valid Roman numeral. The `toRoman()` and `fromRoman()` methods seem unevenly distributed in terms of complexity.