I love love love seeing how reflections are handled by making essentially a virtual geometry under the floor.

This implies pretty obvious & severe limitations that it's not modelling where lights are or how things bounce, but it is really neat to see & works shockingly well for this use case of moving forward through a scene.

If I recall correctly gaussian splats are direction-dependent, so they can lean reflections "correctly" but since there's no picture in the training set that forces it to do the reflection correctly there's no reason for it to.