Anthropic Platform Team: All Code Generated by Claude, Engineers Shift to Verification
Anthropic's platform engineering lead revealed that all code maintained by the team is now generated by Claude, shifting engineers' roles to system design and verification. AI agents are already authorized to autonomously fix tests, with plans to allow direct codebase modifications, while junior technical hiring has dropped significantly.
Full Report
On October 22, Katelyn Lesse, Head of Platform Engineering at Anthropic, revealed during ObservabilityCON hosted by Grafana Labs on Tuesday (local time) that "all of our code is written by Claude—literally everything," referring to the team’s system development and maintenance work.
The team now relies entirely on Claude to generate the code required for writing and maintaining their systems. Engineers have shifted their focus primarily to system design, validation, and ensuring code quality. According to reports, the team has begun allowing AI agents to autonomously fix tests and is exploring ways to enable these agents to directly modify the codebase.
Public records indicate that Katelyn Lesse leads Anthropic’s platform engineering team, which is responsible for the company’s core services, developer tools, and product infrastructure. As reported by Business Insider, the team comprised approximately 200 employees as of this summer.
Lesse stated that engineers remain responsible for system design and validation before handing off actual coding tasks to Claude. During a large-scale migration of the codebase from Python to Rust, the team designed validation mechanisms to ensure the final output met expectations. Meanwhile, AI agents handled the bulk of specific tasks involved in developing code components, monitoring them, and validating the results.
Lesse also disclosed that Anthropic has already permitted AI agents to autonomously fix tests without human intervention. With the improvement in AI model capabilities over the past year, the company is considering how to gradually allow agents to directly modify the production codebase. She noted that while engineering leadership remains cautious about this shift, the boundaries for allowing autonomous code modifications are evolving.
Anthropic’s adoption of AI programming tools extends beyond its platform engineering team. According to a technical report published by Anthropic in June 2026, as of May that year, more than 80% of the lines of code merged into the company’s production codebase were written by Claude.
Another report released by Anthropic on October 16 indicated that AI agents are reshaping the company’s organizational structure and hiring practices. Between early 2025 and 2026, the proportion of junior staff within the technology department dropped from 21% to 14%, while the share of junior technical roles in tech department hires fell from 27% to 9%. Anthropic stated that as AI tools take on more responsibilities for code writing and bug fixing, recruiters are placing greater emphasis on candidates’ autonomy and judgment.