Hook gps-based augmented reality apps into a database of user-created 3d models (paired with orientations and locations), so that sets of models can be configurably layered on top of the real world. Get around opacity problems by having the opacity inversely proportional to distance from the object.
Each user creates a station/profile that streams real-time audio and/or stores pre-recorded material. Can link up with multiple other users' stations & all incoming signals are mixable & tweakable at each user endpoint. Like facebook except instead of images & text & whatnot, communication is primarily audio-based. Could include video streaming to facilitate realtime collaboration. Bandwidth permitting, naturally.