Understanding code is essential for engineers, yet modern codebases are complex and overwhelming, making it difficult to grasp what matters. New engineers can take between 3 to 9 months to fully ramp up, while senior engineers often spend over 5 hours weekly onboarding others. These productivity losses highlight a significant challenge in engineering work. AI coding tools have yet to effectively address focused onboarding and navigation for debugging and feature development. Cognition’s Codemaps aims to improve this process. It allows engineers to generate a visual map of their codebase, providing a snapshot that helps clarify functionality and dependencies. Users can choose between Fast and Smart models to create their Codemap and navigate easily through related code sections. This tool improves upon previous solutions by linking answers to specific lines of code, making it easier to trace issues in client-server interactions or data pipelines. The article emphasizes the shift in the engineer's role from simply writing code to being accountable for understanding it, especially as AI takes over routine tasks. Codemaps enhances this accountability by providing both engineers and AI a shared understanding of the code structure and data flow. This collaborative approach contrasts with the trend of replacing engineers for low-value tasks, advocating instead for a synergy between human and machine capabilities to tackle the more challenging aspects of coding efficiently.