This seems poorly reasoned. The premise that there's a hard divide between trustworthy and untrustworthy systems is, at best, incomplete. The authors definitio

This seems poorly reasoned. The premise that there's a hard divide between trustworthy and untrustworthy systems is, at best, incomplete. The authors definition would consider 1-password an incoherent system, yet it's exceptionally effective at mitigating a whole class of attacks. You could argue that any system where you don't control at least 1) the update mechanism, or 2 the keys isn't sufficiently secure. That's one valid approach, but I would recommend anyone who thinks this read Reflections on Trusting Trust: https://www.cs.cmu.edu/~rdriley/487/papers/Thompson_1984_ReflectionsonTrustingTrust.pdf https://www.cs.cmu.edu/~rdriley/487/papers/Thompson_1984_Ref.... The more interesting and important problem isn't preventing all possible attacks. It's knowing what attack vectors are worth preventing, and what the consequences for a successful hack are. It's definitely not black and white