I have no idea how he does this. What about you brainiacs? [video=youtube;h1cpw57Tg28]http://www.youtube.com/watch?v=h1cpw57Tg28[/video]
Just by watching, it relies on the camera's fixed position (and it's single point perspective). Even when they move it around they're careful to keep it lined up correctly oriented to the camera's perspective and line of sight. At the end of the video you can see that the wooden frame is not a rectangle with 90 degree corners (note how the proximal end of the frame, relative to the man, is not parallel to the distal end of the frame). This would lead one to suspect a spacial illusion involving a forced (false) perspective. At this point, though not solved, you know an illusion is involved, at which point I begin to suspect all factors they don't rule out ahead of time, i.e., the actual plane of the table top and even the orientation of the actor (one assumes he's sitting in a normal chair but he could be "standing in a hole" at the optimum height for the illusion). If it relied on an illusion of one point perspective then all the action should take place at the center. If you note in the video, all the tiles are slid to the center each time, and then the build begins from that "new" center.
See how there is a break in the footage from 30 secs to 31 secs then in the second row to the left about 4 rows up you can see that it is not the same configuration as it was when it was in the frame. it is configured differently with portions of a tile included. Conclusion is that he altered the configuration and added the 3 pieces during the break in footage.
Well...I dunno. I suspect that the 3 pieces are legit but the 3 tiles may be more like cubes than flat tiles. And they appear to have interlocking edges that probably correspond to grooved edges in other tiles. Note how each of the three is flipped in the same orientation when presented on top of the other tiles -- so we don't see the actual face that each tile had when "in configuration." The actual top face of each of these three tiles, when locked into the grid, may be just a fraction of what appears to be the average sized piece (a piece of the grid). Also, just past the middle of video, the wood frame is placed beside the assembled tiles. At this point it can clearly be seen that the tile assembly edge is much shorter than the inside of the frame (they should appear of equal length) it is supposed to fit in. This verifies the skewed frame claim made from observing the top and bottom of frame not being parallel. Some of the large multi-tile pieces, instead of being flat, have definite, thicker and thinner areas and sides. Typically this is seen when the tile pieces are in motion off camera axes.
