Zach Lloyd predicts a shift from interactive coding agentsAn AI coding agent is a tool-using AI system that can inspect, modify, and validate software within a repository. toward software factories that coordinate the development life cycleAgent orchestration coordinates AI agents, tools, people, tasks, state, and control flow so a larger workflow reaches a verified outcome.. Ideas enter a loop of triage, specification, implementation, review, verification, shipping and monitoringWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort.. Human checkpoints remain part of the process, including reviewing specifications, code and the resulting product. In this account, engineers increasingly design and manage the system that builds software rather than writing every change themselves.
Zach Lloyd uses open-source development to explain why this approach matters. If software becomes cheaper to reproduce, he argues that a strong product also needs advantages such as distribution, community, brand or data. Building in the open can help establish an ecosystem, while agent automations can reduce the burden of noisy issues, poor contributions, code review and change verification. He presents this as experience and reasoning from an evolving workflow, rather than a claim that every part already works perfectly.
Zach Lloyd outlines a practical factory architecture: intake channels, a control plane, cloud execution environments and a data layer that lets agents retain context and learn from previous work. Straightforward issues can move directly to implementation, while harder changes receive product and technical specifications. Agent review, risk-based human review, computer-use checks, screenshots, videos and CI/CD provide different forms of verificationAgent completion verification checks evidence that an agent achieved the requested outcome rather than merely producing intermediate work.. Monitoring feeds new problems and observations back into the next development cycle.
Zach Lloyd recommends measuring shipped output alongside human effort and token cost, then using feedback to improve the workflow. His example is an observer agent that notices how senior engineers correct a review agent's comments and uses that evidence to improve its skillsAn AI agent skill is a reusable package of instructions, resources, and tool guidance for performing a bounded kind of work. for later runs. The engineering challenge becomes choosing appropriate context, skills and intervention points for a particular product, not merely running more agents.
Zach Lloyd's closing discussion emphasizes adaptability, critical thinking, learning speed and the ability to understand systems, architecture, code and specifications. Teams still need to tune their process to their domain while concentrating on their core product. Human taste and product judgment remain essential because an automated factory has little value if it repeatedly produces things people do not need.
Watch on YouTube




