Xeno-AI — Operational Shapes Across Configurations, Ignition, Reachability, and World Effects

This paper develops Xeno-AI as a framework for analyzing artificial intelligence without assuming in advance a fixed definition of intelligence, agency, representation, or artificiality. Rather than treating AI as a stable technological category, it identifies configuration-dependent operational shapes that emerge through particular combinations of architecture, memory, embodiment, sensing, temporal horizon, resource access, external tools, institutional coupling, and feedback. The concept of ignition marks the conditions under which capacities such as discrimination, prediction, inference, planning, self-revision, or autonomous intervention become non-redundant within a given configuration. Reachability refers to the set of states, tasks, interventions, and transformations that become practically accessible under those conditions, while world effects track how locally generated AI operations alter subsequent physical, informational, economic, institutional, or scientific arrangements. The framework therefore distinguishes nominal AI classification from operationally demonstrated capacity and allows the same system to instantiate different forms of intelligence across changing configurations. Xeno-AI also rejects the assumption that human cognition provides the universal benchmark for artificial intelligence, while avoiding the opposite assumption that every adaptive machine should count as intelligent. The broader claim is that AI should be studied through the configuration-specific emergence, persistence, transformation, and real-world consequences of operational capacities rather than through a single fixed essence or architecture.