This article critically explores the ways by which the Web could become a more learning-oriented medium in the age of, but also in spite of, the newly bred algorithmic cultures. The social dimension of algorithms is reported in literature as being a socio-technological entanglement that has a powerful influence on users' practices and their lived world. They do not only govern what is visible (and inherently, what is obscured), what is valued and noteworthy, but also have the power to enable and assign meaningfulness in managing how information is perceived by users. This incurs a certain knowledge logic which is pervasive in algorithmic culture. This article posits that inquiry about the relation between algorithms and learning needs definitions as well as the stance of not extensively relying on them. When asking what an algorithm is, or how to define the process of learning and knowledge acquisition, one must also keep in mind that a definition is mostly blind to the ambiguity and slipperiness of contexts, hiding the gaps that hinder the objective circumscription of a concept. This article proposes to mind these "gaps" through discussing controversies that may (or rather actually do) happen regarding contextual or theoretical differences in the interpretation of key concepts such as learning, knowledge and culture. To extend the discussion, I will expose alternative material which allows a wider consideration of the concept of learning and emphasises a dimension of learning seldom taken into account: contextual dependence. The chief characteristic of data processed by algorithms being their decontextualisation, I will discuss the agonistic relationship that is emerging from learning in the age of algorithmic cultures, to explore the possibilities of bridging the gaps and exploit the valuable resources the Web has to provide to enrich another dimension of learning in our lived world: its contextual relatedness.