Engineering management after the cost of code collapsed (karimjedda.com)

In a recent reflection from an engineering director, the introduction of large language models (LLMs) has dramatically altered traditional management practices in software development. The cost of producing code has significantly decreased, prompting a reevaluation of longstanding principles such as the necessity of extensive coding experience for leadership, the effectiveness of consensus-driven decision-making, and the evaluation of team productivity based on outdated metrics. The director emphasizes that many conventional rules are now based on assumptions that are no longer valid and warns against replacing them without careful consideration. This shift highlights a crucial inflection point for the AI/ML community, where the dynamics of engineering management must pivot from traditional heuristics to a more nuanced understanding of human and AI collaboration. It calls for organizations to audit their assumptions, focusing on outcome-based measurements rather than outdated proxies like pull request counts. The potential for AI to handle tasks such as code checking may increase efficiency but also raises questions about verification and accountability, emphasizing the need for robust specifications to guide AI-assisted processes. Ultimately, this transformation offers both an opportunity to streamline development and a challenge to effectively groom the next generation of engineers in a significantly changed landscape.