The piece draws an intriguing parallel between large language models (LLMs) like ChatGPT and Borges' "Library of Babel," suggesting that navigating vast amounts of information is a key challenge in the age of AI. It argues that while LLMs can produce outputs based on prompts, the quality of those outputs can vary widely. The author emphasizes that even with a hypothetical oracle agent—an LLM that generates flawless code—the fundamental need for control in software development remains. This control is vital for ensuring security, confidentiality, and integrity, as even slight deviations in input can lead to significant problems. The article introduces the "oracle argument," questioning whether a simplified prompt could yield a program that meets stringent specifications and security requirements. It touches on concepts from information theory, highlighting that every program has a minimal complexity representation. The author suggests that while LLMs might streamline the process of generating software, they do not eliminate complexity. Instead, they might shift it into a different domain—prompt engineering. The author argues that programming languages are essential tools for managing complexity, countering the notion that coding is merely a difficult barrier for non-programmers. Serious programming languages aim to simplify complex systems, and as long as there’s a need for control in software operations, complexity will persist in some form. The notion that AI could render technical skills obsolete is challenged, emphasizing that understanding and managing complexity remains crucial in software engineering.