Anthropic platform engineering chief Katelyn Lesse said Claude writes all the code her team uses to build and maintain its systems. Her statement concerned that team, which handles Anthropic’s core services, developer tools and product infrastructure. Lesse also said people continue to design systems and verification processes.
What Lesse said about the platform team
Lesse made the claim at Grafana Labs’ ObservabilityCON. The “all” applied to her platform engineering team, not to every Anthropic engineering group or all of the company’s code.
A separate statement attributed to Anthropic in May 2026 put Claude’s share at more than 80% of the company’s production code. That earlier figure covered Anthropic company-wide; Lesse’s later statement concerned her team.
People still design systems and verification
Lesse described agents as handling implementation and monitoring work, while people design systems and decide how to verify them. During a major migration of a codebase from Python to Rust, employees developed a way to verify the finished product.
She said agents already repaired tests autonomously, without human intervention. Anthropic was considering whether agents should make changes directly to the codebase, a separate step from repairing tests.
What Claude Code does
Anthropic describes Claude Code as a coding agent that can read a codebase, edit files and run commands. Its product description says it asks permission before changing files or running commands. Those product details provide context for AI-assisted coding, while Lesse’s account describes her team’s workflow.
