The PyKEEN project began in 2018 coincident with a boom in the development, evaluation, and application of knowledge graph embedding models (KGEMs) to a variety of real-world tasks. Along with the goal of reimplementing and reproducing landmark models and experiments, PyKEEN endeavored to both philosophically and technically define a unified, extensible, reusable framework for KGEMs and their components. We will begin by briefly describing the intertwined journeys of our project and our team of young researchers and software engineers. After reflecting on the academic and industrial adoption of PyKEEN, we will finish by presenting some exciting current and future directions for PyKEEN to align with related software projects, to embrace maturing paradigms like inductive learning, and to diversify to new technologies like graph neural networks and language models.