I'm going to throw down a gauntlet: encryption design that doesn't start with designing for the human users is already fundamentally broken. Why? Because it cannot represent a solid system. This isn't machines in the Matrix talking to one another: the needs of the human users are the only thing that matters.
Combine that with the past few decades of experience in software encryption systems, and we realize that this is hard to achieve. The non-adoption and misuse of GPG that Moxie speaks of, the train-wreck of passwords, and so on. It really feels like we've hit a wall, where there are no significant gains to be had unless these design problems are taken seriously. "Perfect" cryptosystems that have major real-world usability flaws are in fact worthless. Implementing new systems of this kind is a waste of effort and a disservice to everyone who vitally needs the services of a robust encryption infrastructure.
I view the current state of security in applications a a result of two forces: 1) The human dynamics of the roles involved: The disciplines of security and user interface design overlap very infrequently, and very rarely done alongside one another, collaboratively; one always tends to drive the other. For user-friendly and secure communication software to be written in an open source environment, too many stars have to align. 2) The market value of data: Software which can be monitored, and data of its users aggregated, analyzed, acted upon, and sold is more lucrative than making something which is truly secure, even from the authors.
I believe usable encryption is possible, but I don't expect it to come from grassroots open source efforts, nor from the big market players. If anyone, I expect to see it come a collaboration between large non-profits (such as Mozilla, EFF), and universities. In the current environment, they're one of the few with sufficient resources, organization, and motivation to get it done.
Sure, every once in a while you have good security luck out and land a great designer (TextSecure), but that's extremely rare. The vast majority of security teams don't value design as much as security because they don't have the incentive to.
So how do we fix the incentives? Can we? Should we?