Simon Willison in conversation with Cat Wu & Thariq Shihipar, Anthropic

The integration of AI agents like Claude Code and Claude Tag is fundamentally transforming software engineering by shifting the focus from manual implementation to high-level product strategy and business sense. These tools allow engineers to delegate menial tasks, enabling faster iteration cycles and more ambitious project goals. Claude Tag, a proactive, multiplayer agent, now manages over 65% of product engineering pull requests at Anthropic by operating within Slack channels and utilizing team memory. Reliability and safety are maintained through rigorous internal dogfooding, extensive evaluation sets, and automated sandboxing, which mitigate risks like prompt injection and data exfiltration. As these models evolve, the role of the engineer is increasingly defined by the ability to orchestrate complex agentic workflows and maintain high-quality standards, rather than writing individual lines of code.