Just wanted to comment that I love your take on this. Especially that "the Collatz conjecture is essentially the simplest hard problem, by the defensible (albeit not unique) standard of "what it takes to understand the behavior of a Turing machine"".