<p>Additive manufacturing (AM) can be considered as a “material-writing” process in which geometry, microstructure, and behavior are assembled locally, voxel by voxel, into complex architectures. Inspired by natural construction processes, this article reframes material heterogeneity as a mechanism for achieving performance, robustness, and adaptability rather than as a defect to be eliminated. Recent advances in site-specific microstructure control, compositional grading, and solid-state joining demonstrate that AM already enables spatial architecturing of material behavior across multiple length scales, opening the pathway to designing materials with completely novel functionalities. When coupled with <i>in&#xa0;situ</i> sensing, data-driven models, and autonomous experimentation, AM has the potential to function as a programmable material machine. At the same time, fully voxel-resolved design introduces challenges related to complexity, interpretability, qualification, and reproducibility, and may be unnecessary for many applications. A pragmatic path forward lies in introducing material heterogeneities selectively, along dominant loading paths or other functional regions where it may provide decisive advantages. While this parallels familiar strategies such as the use of coatings, graded materials, and layered architectures, it is ultimately shaped by the limits of current AM systems. As these systems grow more capable, it is worth thinking what design space could unfold beyond what can be achieved today and what new applications that may&#xa0;offer.</p> Graphical abstract <p></p>