Currently what is called camera_pose in a lot of places is not actually the camera_pose, but the camera_matrix, which is essentially the inverse of the camera pose. It is probably best to store and manipulate everything as a camera_matrix because that is what we actually want to supply to the renderer (if we switched to actually storing the pose, we'd have to invert that matrix every time).
Go through and try to make that naming consistent and obvious everywhere.
Currently what is called camera_pose in a lot of places is not actually the camera_pose, but the camera_matrix, which is essentially the inverse of the camera pose. It is probably best to store and manipulate everything as a camera_matrix because that is what we actually want to supply to the renderer (if we switched to actually storing the pose, we'd have to invert that matrix every time).
Go through and try to make that naming consistent and obvious everywhere.