Steering in the Volumetric Time Model

This preprint develops a formal account of agency within the Volumetric Time Model (VTM), which treats spacetime as a compact Lorentzian manifold whose governing equations and boundary constraints define a fixed ensemble of admissible histories. Within this framework, uncertainty arises from Bayesian conditioning on partial information rather than temporal indeterminacy. Agency is defined operationally as steering: the capacity of an embedded agent to bias posterior probability mass over pre-existing admissible histories through low-latency feedback, rather than the creation of new futures. Steering is quantified using delayed mutual information between actions and observations, making agency explicitly dependent on latency, noise, and energetic cost. The paper introduces the concept of an agency horizon, beyond which feedback delays render intentional control indistinguishable from noise, and shows how sustained deviations from statistical geodesics in history space require continuous thermodynamic expenditure. Multi-agent consistency is addressed by intersecting admissible history sets under shared global constraints. This work provides an information-theoretic and physically conservative account of agency compatible with relativistic block-universe models, predictive processing, and thermodynamic limits, without invoking retrocausality or violations of global consistency.