The claim that “adaptivity is a prerequisite for cognition” has been central to enactivism (Di Paolo 2005, Thompson 2007), a framework aimed at explaining how cognition arises from the dynamics of life. However, there is a substantiated worry that adaptivity functions as a "Trojan horse," subtly reintroducing representational assumptions into an ostensibly anti-representational framework (Villalobos 2013). Adaptivity in its current conceptualisation undermines enactivism’s promise as a revolutionary alternative to representational cognitive science. Enactivists describe adaptivity as involving self-monitoring, anticipatory regulation, and meta-adaptation—processes traditionally aligned with representationalist cognitive science. This tension is particularly evident in anticipatory regulation, where embodied prediction, though framed as emergent and situated, risks collapsing into a form of representationalism. Unfortunately, enactivists fail to provide a non-representational explanation for how these adaptive mechanisms operate. Thus, if enactivism is to uphold its non-representationalist commitment, it must offer a clearer account of how adaptive capacities emerge solely through dynamic, embodied interactions. As a potential solution, I propose reconceptualising anticipation as temporally extended, implicitly predictive interaction patterns. On this account, anticipation that seemingly trades in internal representations is explained as mechanisms of learned lagged-response mechanisms. This explanation preserves the anticipatory and yet non-representational character of adaptivity.