What really blows my mind about this is that they are using off-the-shelf T4 Ligase to ligate the junctions. I figured this was going to be some tour-de-force of enzyme engineering, but nope, all the reagents are pretty much commercially available.

It is super clever and exciting. Note that people have been able to assemble short (<100 bases) DNA oligomer fragments of synthetic DNA into longer fragments using "splint" oligos since forever. But in this case, each splint has to be custom engineered to only bind to the junction of interest (in practice it is pretty tricky and expensive to do this.) These guys figured out a way to use engineered sequences to make the match, and used a clever (but also more or less standard) way to chew up the engineered stuff, leaving behind only the desired long assembly with no scars at the end of the process.