I started my career in the mid 1980s. One of the women in my group - a brilliant software engineer and mathemetician - attended a QA conference and came back saying 'we discuss graphing number of new bugs for thousand lines of new and changed non commentary source code but according to analysis this doesn't even scratch the surface, and we don't even really do bugs per ncncsl well.' That was 1986, and software quality has only declined since then.
Hopefully any new nuclear power stations will not include software, but it probably will.