The consequences of task switching in supervisory programming (martinfowler.com)

Recent discussions at The Pragmatic Summit highlighted the evolving role of senior developers as large language models (LLMs) become integral to software development. Attendees noted that while some senior developers are initially resistant to using LLMs, hands-on engagement has often led to a change in their perceptions, with a significant portion becoming advocates for the technology. This shift suggests that practical experience is key in understanding LLMs' potential, particularly as advancements in the models continue to emerge. Interestingly, concerns about LLMs displacing junior developers were addressed, with the consensus that these newcomers will thrive due to their adaptability and familiarity with AI tools. Mid-level developers, however, may face challenges in navigating this new landscape. Furthermore, the concept of "cognitive debt" was introduced, emphasizing the mental overload that can result from frequent task switching and the management of multiple AI agents. As programming increasingly resembles supervision of these agents, the industry must explore effective workflows that minimize context-switching while harnessing the efficiency of LLMs. The implications of these dynamics extend to organizational structures, potentially influencing team sizes and collaboration styles in programming, laying the groundwork for a significant transformation in software development practices.