The problem is that you can't verify that hash if you don't have the content of the block. I may give you the following hash:
I see your point. I agree with you that trust would be less self-verifiable and delegated to an extent. The idea is that the hash would be verified against previous transactions up to the previous state, but the problem remains: how to verify the previous state?
Correct me if I'm wrong, but I understand trust here as being statistical, as in, "highly unlikely or computationally too demanding to be otherwise". By the same logic, how likely can the entire network be fooled into accepting a wrong but publicly verifiable state for a month or a year?