When objects become part of a collection, they collectively take on additional meaning. These clusters of objects could serve to group results, or be used to store a users collected finds.
In the video you can see a small cluster of objects. Introduced here is the idea of using a spinning motion to highlight certain objects. The faster the spin could relate to a certain data metric. Personally I think this is preferable to using a flashing on off state to hi-light data points.
The user can rotate this cluster, exploring the visual summary of this dataset or set of results. They may also filter further, indicated here by the the spinning objects exploding outwards from it’s original position. This uses the idea of a 3D space to place objects in certain spaces to indicate relevance. It’s also important to note the objects transition to these new positions over time, rather than instantly moving there in an instant. Transitions allow people to see that a change has taken place, take a moment to take that in and then react accordingly.