Robots and AI technologies are often compared to humans. But, such systems have their own capability distributions which anthropomorphic projection and comparing against human benchmarks fail to capture. The idiosyncratic combinations of what these systems can do and cannot do calls for non-anthropomorphic descriptions and concepts. Easy tasks for humans are typically hard for such computational systems, while tasks humans find harder can be much easier to engineer. Thus, depending on what they are to be used for, what is technologically feasible and economically viable, the design of these systems mixes various such tasks, without any evolutionary logic binding the parts together. The result is systems with a genuinely novel profile with action and interaction possibilities—with their own affordances—entering our open environments. This paper introduces two interconnected concepts to capture this: affordance mixtures and affordance jumps. For robots and AI, affordance mixtures name the specific combinations of human-like and non-human affordances, which further combines designed-object characteristics with an active-agent-like behavior—at radically discontinuous levels and in configurations that are not encountered in nature. Affordance jumps are the specific discontinuities within that mixed profile: absent co-occurrences of affordances, surface affordances implying depths that are not there, or affordances exceeding anything the surface suggests, and familiar outputs through unfamiliar mechanisms with uninferrable failure modes. Together these concepts have implications for how these technologies and interaction with them should be ontologically understood, morally and empirically evaluated, and how they should be designed.