The rise of agentic LLMs has fundamentally shifted the definition of 'code.' Where code once served as the final artifact of the software development lifecycle, LLMs can now generate it on demand, reducing code to a regenerable means rather than an end. This evolution echoes sentiments expressed by Elon Musk, who suggested that source code is becoming akin to assembly language, with AI capable of producing efficient binaries directly.

However, this shift raises questions about the role of traditional programming languages. While LLMs may interpret prompts and context as the new 'code,' programming languages still serve a critical function: reducing ambiguity in specifications that natural language like English inherently lacks. Without precise, unambiguous languages, interpreting specs over time becomes error-prone.

In essence, code is not becoming irrelevant but is instead evolving. As long as deterministic execution matters, programming languages remain essential—even if their role shifts toward clearer specification rather than direct machine instruction.