Hey guys! I'm late to the party, but I want to say that I admire what you're trying to do here. I've thought about this myself, as I have wanted a simple way to send serial data without needing to connect any processor, write code, etc. Even UART's need a processor to set up baud rates, etc. A simple, propritary system using digital logic parts is ideal for alot of applications. A few people on here have tried to do this, with varying success. Despite what you might hear, simple serial communication is most often done this way.
Photon, you're trying to use manchester encoding in your scheme and you're connecting the clock signal from your transmitter to your receiver. By connecting the clock, the encoding becomes unnessary. You typically use manchester encoding when the clock is not connected to the receiver. So, you can forget about needing to code and decode it. I think your best bet is just to connect the clock.
Ccurtis: I've studied your diagram, and I'm convinced that you would get errors in a real system. The fact that your clock doesn't have a function until a start pulse is received is inconsequential to the operation of the circuit. There need to either be a synchronous clock at the receiver, or else a clock signal that is periodically synchronized. I've always been interested in the idea of using a bistable multivibrator which is synchronized with the arriving data. With careful design, I think it could be made to work. That's where manchester encoding would come in handy.