I cant wait for models to get smaller enough that they can run on commodity devices.

Hope we can build an app like Whispr Flow using this with the model running completely on device.