Claude is now writing Claude. Anthropic chief product officer Mike Krieger confirmed in a recent interview that the company’s artificial intelligence (AI) model has become the primary author of its own development.
According to Krieger, the engineering process at Anthropic has moved away from traditional manual coding faster than expected. He said staff engineers now ship pull requests ranging from 2,000 to 3,000 lines that are generated entirely by the AI. Under this system, human developers have moved toward high-level oversight and the verification of machine-generated code rather than line-by-line creation.
This comes less than a year after CEO Dario Amodei predicted that 90% of all code would eventually be written by AI, a claim that triggered much chatter of concern and skepticism by industry observers and engineers at the time.
"Dario predicted 90%... and today it’s effectively 100%," Krieger said.
Amodei has repeatedly spoken about AI’s growing ability to automate “white-collar” and professional work roles once considered safe from automation.
The Anthropic CEO has warned that significant job cuts could occur within five years, urging consumers and US lawmakers to prepare. He also criticised the government and other AI companies for "sugar-coating" the coming reality: the potential for mass job eliminations across various white-collar professions, particularly at the entry level, including technology, finance, law, and consulting.
Automation fears
This also comes days after a tool released by Anthropic led to a massive sell-off in the IT industry on fears of automation. Anthropic had released a "Legal Plugin" for its Claude Cowork platform. This open-source plugin customises Claude AI for legal tasks such as document review, contract analysis, NDA triage, compliance tracking, and generating template responses or basic briefs.
Shares of Indian IT majors, including Infosys, TCS and HCLTech, fell as investors worried about the long-term impact of such AI tools on traditional tech-services businesses.
According to Krieger, the engineering process at Anthropic has moved away from traditional manual coding faster than expected. He said staff engineers now ship pull requests ranging from 2,000 to 3,000 lines that are generated entirely by the AI. Under this system, human developers have moved toward high-level oversight and the verification of machine-generated code rather than line-by-line creation.
This comes less than a year after CEO Dario Amodei predicted that 90% of all code would eventually be written by AI, a claim that triggered much chatter of concern and skepticism by industry observers and engineers at the time.
"Dario predicted 90%... and today it’s effectively 100%," Krieger said.
Amodei has repeatedly spoken about AI’s growing ability to automate “white-collar” and professional work roles once considered safe from automation.
The Anthropic CEO has warned that significant job cuts could occur within five years, urging consumers and US lawmakers to prepare. He also criticised the government and other AI companies for "sugar-coating" the coming reality: the potential for mass job eliminations across various white-collar professions, particularly at the entry level, including technology, finance, law, and consulting.
Automation fears
This also comes days after a tool released by Anthropic led to a massive sell-off in the IT industry on fears of automation. Anthropic had released a "Legal Plugin" for its Claude Cowork platform. This open-source plugin customises Claude AI for legal tasks such as document review, contract analysis, NDA triage, compliance tracking, and generating template responses or basic briefs.
Shares of Indian IT majors, including Infosys, TCS and HCLTech, fell as investors worried about the long-term impact of such AI tools on traditional tech-services businesses.