The relevance of programming languages in the era of AI agents raises important questions about our choices in coding. With AI tools like GitHub Copilot effectively generating Python scripts, the feedback loop of a programming language’s popularity becomes evident. Python’s extensive training data allows AI to produce reliable code, which in turn increases its adoption. However, while AI can automate aspects of coding, it doesn't replace the need for skilled programmers who understand the nuances of their projects. The article highlights two main points regarding the interplay between AI and programming languages. First, languages with strong static type systems, like Scala and Rust, provide quicker feedback for AI agents. This is crucial because a compiler’s ability to catch errors reduces reliance on unit tests, which can be time-consuming. The iterative process of fixing compiler errors leads to more dependable code. Second, the need for human oversight remains vital. Programmers must be able to review and reason about the code generated by AI. The risk of “comprehension debt” arises when team members lack understanding of the original code, making future modifications challenging. Peter Naur's insights on software aging emphasize that changes made without an understanding of the original design can degrade the code's structure. This degradation complicates future updates, increasing costs and introducing bugs. Clear, well-structured source code is essential for preserving knowledge over time, as it allows teams to maintain the intent behind the software. In this context, functional programming and its emphasis on reasoning remain relevant, as they provide tools for understanding and validating AI-generated outputs, ensuring that programmers can still engage meaningfully with the code.