Code-writing work in Anthropic's core platform engineering organisation has effectively shifted almost entirely to Claude, Business Insider reported on Oct. 22.

Caitlin Lese (케이틀린 레세), who leads Anthropic's platform engineering group, said all of the team's code is being written by Claude as the company builds and maintains internal systems.

The platform engineering group led by Lese is responsible for core services, developer tools and product infrastructure. As a department in charge of foundational internal systems, her remarks are seen as meaning Anthropic has elevated AI coding tools from a supporting role to the centre of its development system.

On stage at an event, Lese said, "If you look at how our team actually builds and maintains systems, all the code is written by Claude." She added, "Literally all of it." As of this summer, the organisation was tallied at about 200 people.

People have not been removed entirely. Employees still handle system design and verification, while handing off implementation to Claude. Lese explained, "Even in the process of changing the main codebase programming language from Python to Rust, we designed the way employees verify the final output."

She said, "In actual development settings, a large portion of simple implementation and operational checks is already being driven by agents." Lese said, "From making the detailed pieces to monitoring them and getting confidence in the results, agents took on all that hard work."

Anthropic is taking it a step further. Lese said, "Some test fixes are currently handled autonomously by agents without human intervention, and as AI model performance has improved over the past year, we are reviewing ways to entrust even real codebase changes autonomously." She said, "From the standpoint of an engineering leader, it is scary and requires extreme caution," but added, "That boundary is gradually being pushed back."

This trend is not limited to a specific team. Anthropic said in a report last week that as internal use of AI agents increases, the share of junior technical hiring has shrunk further. It has also disclosed that as of May this year, Claude was writing more than 80 percent of the company's production code.

It also ties into the company's scale strategy. Anthropic has maintained a relatively small workforce by using Claude-generated code. The media outlet said Anthropic is preparing for an initial public offering, and its valuation could reach nearly $2 trillion. Revenue growth in the third quarter this year has slowed.

That leaves a key point to watch as how much further Anthropic expands an AI-first development approach for corporate clients based on internal cases. The near disappearance of human coding in the internal infrastructure organisation shows that competition in AI coding tools has moved beyond a productivity-assistance stage and entered a phase of changing how organisations are run.