At Anthropic, the growing role of artificial intelligence in software development is transforming hiring priorities, according to cofounder Jack Clark. Speaking on an episode of "The Ezra Klein Show," Clark revealed that AI, particularly the company’s own large language model Claude, now writes the majority of the code used in its projects—comfortably more than half, and potentially approaching 99% by the end of the year if current trends continue. This shift has altered the perceived value of different levels of engineering talent. Clark noted that the value of junior developers is now “a bit more dubious,” while the importance of senior engineers with deep experience, refined judgment, and strong technical intuition has increased significantly. “Something that we found is that the value of more senior people with really, really well-calibrated intuitions and taste is going up,” he said. Despite AI handling much of the coding work, Anthropic is not reducing its engineering workforce. In fact, the company now employs more software engineers than it did two years ago. Its careers page lists at least 100 open software engineering positions, indicating continued growth in the department. The change isn’t about hiring fewer people—it’s about redefining what engineers do. As AI automates routine and foundational coding tasks, the focus for human engineers is shifting to higher-level responsibilities: designing system architecture, setting strategic direction, making nuanced decisions, and ensuring quality and safety in AI-generated code. Clark described this evolution as an “O-ring automation,” a concept where automating one part of a process causes human effort to concentrate on the next bottleneck. As AI handles implementation, humans move up the stack—improving the parts that remain complex or difficult to automate, which may eventually be automated in turn. This transformation, Clark suggested, could signal a broader shift across the tech industry. While headcount may not necessarily drop, the distribution of value is changing. The most critical contributions are no longer in writing basic code, but in guiding, evaluating, and refining AI’s work—tasks that require deep expertise, experience, and judgment.