Good comment, I like arguments backed by stats! I do recognise the validity of this line of thought.
Food for thought for you, though: is solving this problem (a "miracle") not quite similar to the problem of making a safe FAI, possibly with hard takeoff?
That is to say, shouldn't we be concerned that we can't (at least gesture towards) solving this problem of most people having crappy governments, when we are also supposed to be solving the problem of AI alignment? I feel as though these two problems are intimately related beyond just the issue of bad governance interfering with or actively hindering safe AI research.
I would have liked to have seen more examples of what different languages shepherd you to do.
Python, in my experience, shepherds you to import a lot of modules rather than writing your own. There's so many helpful ones available and that's kind of a reality shock if you switch to some other languages.