Vaulted masonry structures are a fundamental class of load-bearing systems in architectural heritage whose structural behavior is governed by a strong interdependence between form and force [1]. Their stability relies predominantly on axial compression, enabling efficient load transfer. In many historical constructions, vaults were conceived to remain stable not only in their final configuration but also throughout the assembly process, without extensive temporary support [2]. In this context, self-supporting behavior is the ability to maintain static equilibrium at each assembly step without the need for auxiliary supporting forces. While the structural optimization of vaulted structures has been addressed in prior research[3, 4], most existing approaches focus on final equilibrium and do not account for the evolving conditions during assembly. Methods that consider stability during assembly either decouple this assessment from the form-finding process [5] or enforce self-supporting behavior through typology-specific constructive rules [6], thereby conservatively restricting the design space. As a result, a generalized form-finding methodology that directly incorporates intermediate equilibrium conditions into the design process remains an open challenge. This paper introduces a computational framework in which stability during assembly acts as an explicit driver of form-finding for vaulted masonry structures. For a prescribed assembly sequence, successive intermediate configurations are generated and evaluated, and the auxiliary supporting forces required to maintain static equilibrium are quantified and accumulated across all assembly steps. This accumulated measure is embedded in a gradient-based optimization that steers the geometry toward being self-supporting under the specified assembly sequence. In this formulation, gradient-based optimization serves as a unifying interface that can integrate additional constraints and criteria within the same computational framework.