Anthropic Engineering Leader Says Her 200-Person Team Writes Zero Lines of Code Itself

Katelyn Lesse, Anthropic’s head of platform engineering, says her roughly 200-person team has shifted entirely to AI-generated code, with every line now written by Claude Code rather than by human engineers directly.

Lesse’s team oversees Anthropic’s core services, developer tools and product infrastructure, making the scale of the shift notable given the team’s central role in keeping the company’s platform running.

Despite the full handoff of actual code writing, Lesse said human engineers remain essential to the process, continuing to design systems and verify the outputs that AI agents produce, while the agents themselves handle implementation, monitoring and other routine engineering tasks.

As an example, Lesse pointed to Anthropic's migration of a major codebase from Python to Rust, saying staff developed validation methods to confirm the final product met requirements, while AI agents carried out the underlying build work and ongoing monitoring.

Lesse made the comments at Grafana Labs' ObservabilityCON, offering a rare, concrete look inside how one of the leading AI labs is restructuring its own engineering workflows around the tools it builds.

The disclosure adds a striking real-world data point to the broader debate over how quickly AI coding tools are reshaping software engineering roles, with Anthropic's own platform team now serving as a live example of a workflow where humans oversee and verify rather than write code directly.