Which is just an implementation of precisely the oracle scheme I talked about. Introducing precisely the problems I mentioned: blockchain bloat, additional trust anchor which can never be 100% validated and thus destroys the idea of requiring zero trust in any particular individual person or system.
Extending the chain of trust down to the source of information isn't exactly going to happen at a snap of a finger. If you trust TLS and trust what the server gives you then today that is a fine option, the server is signing a piece of information and giving it to you, this is no different from you signing a transaction and publishing that transaction to bitcoin the trust ends in just the same way in each case with the caveat of 'value' being created from a web service. An attacker gains access to either the producer of the transaction (client code or web service) or the keys to sign the transaction (tls or bitcoin private key). That said I agree that modern web services could much more easily be made to give an incorrect response.
I don't think your argument applies any differently to other cryptography, it all looks like jumping through hoops for no guaranteed safety. If you believe that the only thing that matters is 100% safety then you are going to be very disappointed with computer security in general.
I don't see it as so black and white. Sure, zero trust is the ideal, but there are potentially benefits even there's some trust involved. For example, I saw a headline indicating that Unicef was going to experiment with block chain technology and asset management. Charity strikes me as particularly suitable domain since benefactors are understandably concerned about how their donations are being used and there's little need or even desire for secrecy. Block chain could provide an unparalleled degree of transparency and accountability even if the humans in control of the private keys ultimately need to be trusted. "Trust but verify" as Reagan said, and it would help a great deal with the latter if not the former.
Once you're willing to trust, you don't need the proof-of-work that makes blockchain blockchain though. A distributed append-only, log-structured, cryptographically verified is a great idea with plenty of practical applications; OTOH "mining" is an immense waste of energy to minimal benefit.
