[mitchtbaum]

That seems to have worked because the king had an unmanageable level of overconfidence, whereas this worked because they already had mutual trust[0]. Advice from a friend passes easily through the "harm test" heuristic filter which takes place immediately after hearing any untrusted (doubted) person advising one to change course (and potentially other places if someone learns they need to apply it there too).

By mixing in advanced machinery, our innate heuristics like harm measurement need many more dimensions of analysis. Hackers, in tune with modern machines, recognize this as a blunder since we have seen trust misused with secrets in machines before; still how can a "[s]cientist and security researcher" and "farmer and shoe-repair-man with a handheld" alike learn to recognize wider effects of their machine-enabled actions?

0: https://twitter.com/search?q=from%3ASc00bzT%20to%3ADefuseSec...