Using a tie-and-strut model makes it easier to explain the difference here. A tie-and-strut model is a simplification of the structure which describes where the structure is under tension, and where it is under compression. The model gets effectively simplified into a truss. Below we have two examples of the axial load on members of two different trusses with identical loading (positive values are tension, negative are compression):
Looking at the values, you'll see that the results are mirror images of each other: same values, but different signs. Importantly, notice that the first truss has a negative (compressive) force of 634.9 between the supports and the upper chord, while the second one has a positive (tensile) force of 634.7 between the supports and the lower chord.
Now, looking at the video, we basically have two possible layouts, which are converted into the overlaid trusses:
Comparing with the truss results shown above, it is clear that the first diagonal emanating from the support will suffer compression in the first format and tension in the second layout.
Now, what will resist the forces represented by that diagonal? The combined plywood layers, of course. However, that force needs to be transmitted from layer to layer. In the case of compression, that transfer is done by contact. However, in tension, that transfer occurs via the glue. If the glue were to fail, then the plywood layers will separate and the structure will progressively collapse:
Meanwhile, with the other orientation, that force is compressive, so such a failure is impossible (other failure states will certainly happen before the "compressive strut" collapses).