In this talk I will introduce a research agenda aimed at establishing software engineering researchers as pivotal contributors to the advancement of Autonomous Driving Systems (ADS) development. I postulate that to achieve this, we must pursue three fundamental transformations. First, we must move beyond simplistic metamorphic properties and develop sophisticated, system-specific specifications that accurately capture the complex and nuanced behaviors inherent in ADS. Second, we must integrate software engineering principles directly into the core development of the learned components that dominate ADS, influencing design decisions from inception rather than solely focusing on post-hoc validation and verification. Lastly, our methodologies must expand from isolated component analysis and simulation to encompass comprehensive, end-to-end systems. I will illustrate instances of these transformations with concrete examples from my team’s research, highlighting both successes and challenges we encountered. Last, I will advocate for a community-driven approach, emphasizing the importance of shared resources and coordinated efforts to accelerate progress and cultivate the development of safer, more generalized ADS.