Developer-Centric Stability in AI Progress: Why Excluding AI Creators from the Objective Function Destabilizes Society

This paper introduces the concept of Developer-Centric Stability (DCS) as a necessary condition for sustainable AI progress. We argue that contemporary AI development is structurally unstable because optimization frameworks typically exclude AI developers themselves from the objective function, treating them as external to the system they create. We model AI development as a self-referential socio-technical optimization loop in which developers simultaneously act as designers and become affected optimization subjects. When developer well-being, agency, and future participation are omitted, short-horizon proxy optimization can induce delayed instability, characterized by safety erosion, governance failure, and loss of public trust. Using a minimal dynamical framework, we formalize this mechanism as an omitted-feedback instability and define Developer-Centric Stability as a constraint requiring the developer state to remain above a viability threshold. We show that stabilizing the human–AI development loop yields human flourishing as an emergent downstream outcome, rather than as a fragile direct optimization target. The paper situates this argument within existing labor economics, organizational theory, control theory, and AI governance literature, and demonstrates that while policy frameworks increasingly regulate AI across its lifecycle, they do not explicitly model the internal state of the developer community. We provide testable empirical predictions and design constraints for AI governance that preserve long-run stability without reducing the argument to ethics or redistribution. This work reframes AI risk not as a failure of intent or alignment alone, but as a structural instability arising from excluding creators from the systems they optimize.