Does code quality still matter? (blog.ploeh.dk)

In a thought-provoking article, Mark Seemann explores the future of coding as large language models (LLMs) take on more programming tasks. He raises critical questions about code quality, emphasizing that while LLMs can generate code faster than human programmers, the importance of code quality remains essential as long as humans supervise it. However, he speculates on a future where LLMs operate independently, potentially leading to a new programming language tailored for machine consumption, dubbed "LLaMe." In such a scenario, traditional concerns regarding code quality—like readability and structure—may become irrelevant, as LLMs wouldn't face the cognitive limitations that humans do. Seemann warns that even if programming shifts entirely to LLMs, technical debt could still accumulate, albeit in forms we may not yet understand. As LLMs would need to create their own coding conventions and learn from their experiences, it raises the question of whether poor coding practices could hinder their efficiency or lead to issues that humans may struggle to address. He concludes that the implications of allowing LLM-driven coding without human oversight are complex and could pose unforeseen challenges. This debate is particularly relevant to the AI/ML community, as it underscores the ongoing need to balance human oversight with the potential capabilities of intelligent systems.