The software industry is entering a phase of deep change. AI firms are no longer treating artificial intelligence as a helper for engineers. They are letting it do the work. Anthropic is now the clearest example of this shift.

At Cisco’s AI Summit, Anthropic chief product officer Mike Krieger said the company’s AI model, Claude, now writes nearly all of Anthropic’s code. Human developers no longer write features line by line. They guide, review, and approve what the system produces.

“Claude is being written by Claude,” Krieger said during an onstage discussion with Cisco president Jeetu Patel. He added that Claude products and core code are now generated by the model itself.

This marks a sharp move away from traditional software development. Until recently, AI tools helped engineers complete tasks faster. At Anthropic, the balance has flipped. The system now generates full blocks of production code, while humans provide oversight.

Krieger said teams often produce pull requests that run to several thousand lines. These are not drafts or experiments. They are working code used in real products. To support this process, Anthropic has built internal checks that allow teams to trust machine-written output.

How Anthropic Claude is Rewriting the Future of Engineering and IT?

Claude also reviews code. Engineers submit pull requests and ask the model to act as a critical reviewer. It flags security issues, points out weak logic, and suggests cleaner structure. Krieger said this has improved both speed and quality.

This approach builds on a prediction made last year by Anthropic CEO Dario Amodei. Amodei said AI would eventually write 90 percent of all code. Krieger said that threshold has already been crossed inside the company.

The timing matters. Krieger’s comments came days after Anthropic launched a new set of workplace automation tools for its Claude Cowork agent. These tools triggered a sharp sell-off in Indian IT stocks.

Claude Cowork now supports 11 plug-ins that handle tasks across legal, sales, marketing, and data analysis. Unlike earlier AI tools, these agents can act on their own. They no longer rely on software platforms such as Salesforce or ServiceNow to complete tasks.

This change unsettled investors. Shares of Indian IT firms such as Infosys, TCS, and HCLTech fell as markets reassessed the role of traditional IT services in an AI-driven world. Jefferies described the moment as a “SaaSpocalypse,” suggesting that many software platforms could lose relevance.

The concern runs deeper than tools or stock prices. It raises questions about jobs. At the World Economic Forum in Davos last month, Amodei warned that software engineering as a profession could fade within a year. He said AI has moved from speeding up work to replacing it.

Amodei shared feedback from his own teams. Some engineering leads, he said, no longer write code. They let Anthropic’s most advanced model handle the task and focus on edits and decisions.

How AI is Redefining White-Collar Work

Amodei has also warned that AI could remove most white-collar roles within five years. His views carry weight because Anthropic builds some of the most capable coding systems in use today.

Zoho founder Sridhar Vembu echoed this concern in a post on X. He said the industry should pay close attention to Amodei because Anthropic has the strongest coding tools available.

What emerges from these developments is a clear signal. Software creation is changing at its core. Coding is shifting from a craft based on manual input to a process built on supervision and judgment. Companies that adapt may move faster than ever. Those that rely on older models may struggle to keep pace.

For now, Anthropic stands at the front of this shift. By letting AI write its own code, the company has turned a long-term prediction into a present reality.