Software Engineering’s Reliance on Informal Knowledge and Its Limits in the Age of AI

The article examines how modern software development heavily depends on informal, undocumented, or poorly documented knowledge—often passed through conversations, code comments, or short-lived internal wikis. A historian-turned-software engineer argues that this “oral tradition” has become a defining feature of the industry. While early design documents may exist, they are frequently outdated as systems evolve rapidly. Much of the real understanding of a system lives in developers’ heads, making it vulnerable to turnover and organizational change. Agile development practices, which prioritize working software over comprehensive documentation, have further reinforced this tendency toward underdocumentation. The article highlights that although documentation is widely acknowledged as valuable, it is often neglected due to time constraints, developer preferences, and shifting industry culture. As a result, software systems can become difficult to maintain or understand for new engineers who must reconstruct intent from code alone. This creates an “archaeological” process where developers reverse-engineer decisions without reliable written context. The discussion also addresses the growing expectation that generative AI could fill documentation gaps by summarizing codebases or generating explanations automatically. The author is skeptical of this idea, arguing that while large language models can describe what code does, they cannot reliably capture why it was written in a particular way or what trade-offs influenced its design. Documentation, in this view, is not just descriptive but cognitive—it helps developers clarify thinking and decision-making before implementation. Ultimately, the piece suggests that if software engineering continues to undervalue intentional documentation, even advanced AI tools will not fully resolve the loss of institutional knowledge. The gap between written code and human reasoning remains a core challenge that technology alone may not bridge.