Working in a three-dimensional variant of Winfree's abstract Tile Assembly Model, we show that, for an arbitrary finite, connected shape X ⊂ Z 2 , there is a tile set that uniquely self-assembles into a 3D representation of X at temperature 1 in 3D with optimal program-size complexity (the program-size complexity, also known as tile complexity, of a shape is the minimum number of tile types required to uniquely self-assemble it). Moreover, our construction is "just barely" 3D in the sense that it only places tiles in the z = 0 and z = 1 planes. Our result is essentially a just-barely 3D temperature 1 simulation of a similar 2D temperature 2 result by Soloveichik and Winfree (SICOMP 2007).cooperative self-assembly leads to highly non-trivial theoretical behavior, e.g., Turing universality [25] and the efficient self-assembly of N × N squares [1,18] and other algorithmically specified shapes [23].Despite its theoretical algorithmic capabilities, when cooperative self-assembly is implemented using DNA tiles in the laboratory [2,14,20,21,26], tiles may (and do) erroneously bind in a non-cooperative fashion, which usually results in the production of undesired final structures. In order to completely avoid the erroneous effects of tiles unexpectedly binding in a non-cooperative fashion, the experimenter should only build nanoscale structures using constructions that are guaranteed to work correctly in non-cooperative self-assembly. Thus, characterizing the theoretical power of non-cooperative self-assembly has significant practical implications.Although no characterization of the power of non-cooperative self-assembly exists at the time of this writing, Doty, Patitz and Summers conjecture [7] that 2D non-cooperative self-assembly is weaker than 2D cooperative self-assembly because a certain technical condition, known as "pumpability", is true for any 2D non-cooperative tile set. If the pumpability conjecture is true, then non-cooperative 2D self-assembly can only produce simple, highly-regular shapes and patterns, which are too simple and regular to be the result of complex computation.In addition to the pumpability conjecture, there are a number of results that study the suspected weakness of non-cooperative self-assembly. For example, Rothemund and Winfree [18] proved that, if the final assembly must be fully connected, then the minimum number of unique tile types required to self-assemble an N × N square (i.e., its tile complexity) is exactly 2N − 1. Manuch, Stacho and Stoll [13] showed that the previous tile complexity is also true when the final assembly cannot contain even any glue mismatches. Moreover, at the time of this writing, the only way in which non-cooperative self-assembly has been shown to be unconditionally weaker than cooperative self-assembly is in the sense of intrinsic universality [5,6]. First, Doty, Lutz, Patitz, Schweller, Summers and Woods [5] proved the existence of a universal cooperative tile set that can be programmed to simulate the behavior of any tile set (i.e., the aTAM is intrinsical...