I fail to see how this helps.

The reason why bugs crop up all the time in software isn't because syntax is confusing. It's also not because semantics is confusing--plenty of bugs have perfectly clear, well-understood bugs. The problem is that we as programmers don't think about how our software could fail. We don't ask ourselves "could this multiplication overflow?" frequently enough. We don't look at code and ask "who made sure this pointer points to valid data?" We believe that there's no way the price of an affiliated token could ever reach exactly $0, so we assert that it can't happen.

The way you avoid these bugs is to just simply make it impossible for the system to get into certain states. It's already the case with statically-typed languages that it's impossible to pass a string to a function that expected an int. If you design your API right, you can make it impossible to get an index into an array that is out-of-range (although this is way too rarely done). But, even then, you will still find people who will confidently use the escape hatch to say "this string is clearly UTF-8, I know it is from outside experience, so don't bother checking."