Jidoka for Software: Autonomation in the Age of LLMs (jamessummerfield.com)

A recent discussion titled "Jidoka for Software" highlights a transformative shift in software engineering driven by the rise of coding agents powered by large language models (LLMs). As these tools rapidly generate correct code, the industry is moving away from conventional bottlenecks like coding speed to focusing on intent, specification, and verification. This shift echoes historical principles of jidoka, or "autonomation," where automation enhances human oversight rather than replaces it. The most productive teams will be those that can efficiently specify tasks, review agent outputs, and maintain quality standards—essentially redesigning workflow to prioritize strategic decision-making over manual coding tasks. The implications for the AI/ML community are significant. With coding agents handling production work, the role of engineers is evolving from writing code to managing and directing these agents, ensuring that outputs align with business objectives. Teams are encouraged to adopt a domain-centric approach, allowing each engineer to oversee an entire business area while minimizing integration risks. As the process of code review transitions to a model that leverages automated checks alongside human judgment on intent and outcomes, engineering practices are set to become more scalable and efficient. This transition promises to enhance output and speed, enabling smaller teams to "outship" larger competitors by recognizing that the real bottleneck lies not in production but in strategic oversight.