For decades, software engineering has depended less on formal documentation and more on an informal “oral tradition” where experienced developers pass down knowledge to newer engineers through conversations, mentorship, and shared problem-solving. This fragile but deeply important culture may now be threatened by artificial intelligence. As AI increasingly writes, explains, and maintains code, fewer engineers may fully understand the systems they work on or learn the historical reasoning behind technical decisions.
The article describes software development as surprisingly undocumented in practice. Many companies rely on institutional memory stored inside the minds of veteran engineers rather than detailed written records. New developers often learn by asking senior coworkers why systems were built a certain way, what hidden risks exist, or which parts of the codebase should never be touched. This human transfer of context has historically compensated for incomplete documentation and rapidly evolving software environments.
AI-assisted coding tools may weaken this mentorship culture by accelerating development while reducing direct collaboration between engineers. If junior developers increasingly rely on AI-generated answers instead of experienced coworkers, they may become effective at producing code without fully understanding architecture, trade-offs, or long-term system behavior. Critics worry this could create a generation of engineers skilled at prompting AI but less capable of debugging deep infrastructure problems or reconstructing institutional knowledge once experienced developers leave. Similar concerns are already appearing across industries where AI is reshaping training, apprenticeship, and workplace learning.
At the same time, many engineers argue that AI will not eliminate the need for human expertise but instead change how expertise is developed and shared. Software systems remain highly complex, unpredictable, and deeply dependent on human judgment, especially in large legacy environments. Researchers and technologists increasingly believe the future of software engineering may depend on balancing AI-driven productivity with deliberate preservation of mentorship, documentation, and collaborative learning practices that help humans understand not just what code does, but why it exists in the first place.
