I'm looking for deeper-level driving analysis - like telling the driver they brake too hard... As well as connectivity to smartphones to use in place of proximity keys/alarms/remote start/etc.
Some of this, we're already seeing, but I think there is a huge opportunity here.
Automatic (YC, I think) do a good amount of that driving analysis, like telling the driver they brake or accelerate too hard, using only OBD-II and an accelerometer.
I think a built-in system doing this kind of thing would be really cool. The technology is obviously already there, but the software isn't, so having an app platform creates a lot of opportunity. I wonder how deep Google will be able to convince manufacturers to take the integration - modern cars expose a ton of data and functionality over the CAN bus, but it's all undocumented and proprietary and is often obfuscated as well.
I think they also provide the interface for free for interested developers... Could be mistaken.
Basically, the idea is to make it better documented and less proprietary... Then support a developer community for apps that can interface with the broader range of data than OBDII can provide.
OpenXC effectively just moves "proprietary vehicle knowledge" into firmware from client software - it's not a standard interface for cars, but rather a standard interface for car interfaces.
That's a neat idea since it prevents every app from needing to handle CANbus and whatever other proprietary signaling each vehicle uses, but it doesn't really expand vehicle support because the CAN and other messages still need to be documented officially (as Ford have done) or reverse-engineered.
The current implementation of OpenXC actually makes the setup more expensive, as well - an OpenXC module, since it does more, requires more hardware than a simple CAN-to-UART/Bluetooth gateway like an ELM327 clone.
Some of this, we're already seeing, but I think there is a huge opportunity here.