Beyond Scaling II: The Future-Transition Field: Runtime Transition Coordinates for AGI/ASI-Relevant Systems

Scaling has expanded what AI systems can do, but capability alone does not define the next safety bottleneck. As AI systems become more agentic, recursive, collective, and operationally embedded, their significance increasingly lies in runtime transitions that may begin to reshape observation, correction, and evidence conditions before outcomes are explicit or fully committed. This paper introduces the Future-Transition Field: a coupled abstraction for studying runtime transitions in AGI/ASI-relevant systems. The paper proposes the Precursor Condition: a minimum condition for intelligence beyond human level is the capacity to handle precursors — weak, partial, and context-dependent indications associated with consequential transitions before outcomes become visible or committed. Such precursor handling cannot be reduced to isolated warning-signal recognition, output classification, or static action assessment. It requires public coordinates that bind precursor evidence to observation context, future-condition structure, and evidence-validity structure. The runtime safety object is represented as X_t=(Omega_t,F_t,E_t), where Omega_t denotes observation context, F_t denotes future-condition structure, and E_t denotes evidence-validity structure. A transition tau_t:X_t->X_{t+1} may alter not only the system state or action path, but also the boundary, observer position, future correction conditions, and evidence domain under which runtime judgment remains valid. Existing work on AGI-to-ASI transition, agent dynamics, reachability, recoverability, impact regularization, shielding, provenance, and runtime governance can be interpreted as partial projections of this coupled field. This paper does not disclose implementation-level coordinates, observable decompositions, operational criteria, scoring functions, or decision rules. Its contribution is to define a public conceptual object for organizing precursor handling and runtime transition safety before commitment. Disclosure boundary: This manuscript remains theoretical, conceptual, and observational. It does not provide implementation-level algorithms, operational boundary criteria, scoring functions, runtime control mechanisms, or product-side decision logic. Version: Public Draft v0.4 — Public-Ready, 20 August 2026. License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).