Anthropic engineer alerts: AI to radically change all US computer-based jobs

An experienced Anthropic engineer said a freshly developed generation of AI agents capable of managing computers will revolutionise nearly every internet-driven job in the US.

Additionally, he indicated that the transition is approaching rapidly.

Boris Cherny—the creator of Claude Code at Anthropic, the organization recognised for its Claude chatbot—recently joined "Lenny's Podcast," presented by Lenny Rachitsky.

He stated that AI systems capable of integrating with workplace computer programs—such as those Anthropic provides access to—are swiftly advancing and could soon transform duties for software developers, project managers, designers, and other professionals.

"It will expand to nearly any task manageable on a computer," Cherny explained. "Initially, it will be highly disruptive and challenging for many people."

Claude Code is Anthropic's AI programming agent built upon its Claude models. The company rolled out its recent updates, named Opus 4.6, in early February.

Unlike typical chatbots that produce text or images, an AI agent can utilize digital tools—executing commands, evaluating documents, messaging peers, completing tasks throughout apps, and even designing websites.

In essence, Claude Code can increasingly operate a computer similarly to a human—though the company recently acknowledged it hasn't yet reached the expertise of a seasoned human.

"It's something that I believe introduces agentic AI to those unfamiliar with it, and users are beginning to understand it for the first time," he mentioned.

Cherny reports that his team already leverages AI for speedier work. Efficiency per engineer has soared since the launch of Claude Code, he noted. 

He anticipated ongoing advancements in the models.

In a recent chat with Y Combinator's "Lightcone" podcast, Cherny suggested the job title software engineer might "disappear" by 2026.

The broader implications remain unclear, he cautioned.

"As a community, this is a dialogue we need to resolve collectively," he expressed to Rachitsky. "Anybody can develop software whenever they choose."

For those adapting to the shift, his guidance is straightforward: engage with AI tools and understand their operations.