Sakana AI announced that a paper on "Smart Cellular Bricks," written with researchers at the IT University of Copenhagen and Autodesk, has been accepted in Nature Communications, with code released at github.com/rmorenoga/cube3D. The system is a set of identical cubic printed-circuit-board modules, each with a microcontroller, six-face connectors and an LED, running the same 3D Neural Cellular Automaton and communicating only with physically attached neighbours; no module knows its position. In simulation the method reached 98.97% classification accuracy across classes such as planes, chairs, cars, tables, houses, guitars and boats; transferred unchanged to hardware, four shapes of 26 to 197 bricks converged correctly in all cases in under 60 update cycles, roughly three minutes. Joint training for damage detection kept 98.9% shape accuracy with 94.8% damage-detection accuracy, and simulations scaled to 64x64x64 grids and assemblies of over 18,000 cubes.