At Anthropic, AI now writes most of the code and staff steer it

Anthropic, the company behind the AI tool Claude, says Claude wrote more than 90% of the contributions to its code in September. The detail comes from the company's IPO filing, a long document released Monday. Its engineers now spend more time directing and checking the AI's work than typing code line by line.

What the filing says

Anthropic grew from about 940 employees at the end of 2024 to about 5,300 as of late September. That is more than fivefold growth in under two years. The company says its plan is to use Claude to grow without adding staff at the same pace, which it says leads to "extraordinary contributions per employee."

• Claude wrote over 90% of code contributions in September.
• Average daily code contributions per engineer in the third quarter were more than 45 times the 2024 level.
• Engineers increasingly direct and review Claude's output.

What outsiders make of it

Recruiter Martha Heller noted that an AI company will naturally praise its own product. She also said the company still needs people to oversee what the AI does. Aaron Cleavinger, an executive recruiter, said head count is becoming a less meaningful measure of what a company can do. He said it is not yet clear what replaces it, but firms may focus more on worker productivity and business impact.

Fewer entry-level roles

Anthropic said in a recent internal report that it is hiring a smaller share of junior technical workers. AI agents, which are programs that carry out tasks on their own, now write code, fix bugs and handle other jobs across the firm. That could leave a thinner pipeline of newer workers to promote later.

Keeping talent

The company says it depends heavily on senior leaders and key technical staff, and that losing them could slow its work. Competitors are courting the few people with the right skills. Anthropic reported only about 1% "regrettable attrition" in the 12 months ending September 30. Safety researcher Jacob Coxon resigned last month, saying AI companies were moving too fast without a sufficient plan to control the technology.

What it means for work

Anthropic also describes a "low-ego" culture, where the best argument wins over the loudest voice or most senior title. Experts quoted say that is harder to keep as a company grows. For readers, the larger point is that if AI handles more routine work, oversight and results may count for more, and entry-level openings may be harder to find.