A DNA file sat unused on Pietro Schirano’s computer for years. Earlier this month, he sent it to Claude Code—Anthropic’s “agentic coding tool”—for analysis. “I’m attaching my raw Ancestry DNA file,” he told the tool.

The AI created copies of itself on Schirano’s device, each mimicking a specialist in a different genomic area: one focused on cardiovascular disease, another on aging, a third on autoimmune disorders. “A lot of the findings aligned with my life,” said Schirano, who worked as an engineer at Anthropic before founding MagicPath, an AI product design startup. “I always thought I handled caffeine better than all my friends. It was an inside joke: I can drink seven espressos just because I’m Italian.” Claude Code’s analysis revealed Schirano indeed has a gene that lets him metabolize caffeine faster than average, that he’s genetically predisposed to Alzheimer’s, and recommended supplements based on his DNA profile.

Launched in February 2025, Claude Code was Anthropic’s first successful AI agent—a system that acts on the user’s behalf instead of just conversing in a chat interface. It can access files and programs on a user’s computer, and even run “sub-agents” for specific tasks, like the ones that analyzed different parts of Schirano’s genome. It has gained a loyal following of hobbyists using it to do their taxes, design patterns, and even autonomously care for a tomato plant.

Yet most people have never heard of Claude Code. That’s because its primary access method is a command line interface—the old-school computer terminal that fell out of use with the general public sometime in the last millennium. That obscurity may soon change. On Monday, Anthropic introduced Claude Cowork, which the company calls “Claude Code for the rest of your work.”

“It’s going to blow the minds of many non-coders,” says Martin DeVido, the developer behind the experiment using Claude Code to grow a tomato plant.

Claude Cowork aims to bring Claude Code’s agentic capabilities to a broader audience by offering a friendlier user interface and hiding some of the complexity that made Claude Code daunting for beginners. The tool, initially available as a research preview for Max plan subscribers paying $200 monthly, has “rough edges,” according to Felix Reiseberg, its lead engineer. One user reported the app gave her “scary error messages” and wouldn’t connect to her calendar.

This shouldn’t be surprising: the app was built in under two weeks, mostly by Claude Code itself. It’s a sign of the times that AI tools are building themselves—and sometimes break as a result. Still, Claude Code started as a demo; six months later, it hit $1 billion in annualized revenue—roughly 15% of Anthropic’s total at the time.

Claude Code and Cowork are part of a wave of efforts to turn AI chatbots into agents. Other agentic coding tools like Cursor and OpenAI’s Codex have also found success with programmers; AI browsers such as ChatGPT Atlas and Perplexity’s Comet browser aim to let AI models act on users’ behalf online. “If Claude Code didn’t exist, I’d expect Codex or Cursor to have similar popularity,” says Jean-Stanislas Denain, a researcher at the AI institute Epoch AI—it represents other tools giving AI models agentic abilities.

It remains unclear if Opus 4.5—the model powering Claude Cowork—is as good at general knowledge work as it is at coding via Claude Code. “Opus 4.5 is definitely a big deal for coding,” Mantas Mazeika, a research scientist at the Center for AI Safety, told TIME. But it only completed nine out of 240 human freelancer projects in the center’s tests—projects including creating architectural plans and developing video games. “The main bottlenecks are specific cognitive limitations that Claude Code doesn’t address.”

“We won’t know from this initial rollout if a wide range of non-coding users would actually get value from it,” Denain said, noting the tool is only available to a small number of customers.

Still, for some, Claude Code, Cowork, and similar tools mark the shift from AI chatbots to agents—a transition predicted even before ChatGPT’s official launch. “Claude Code is just another data point in a broader trend that’s been ongoing for years and will continue,” says Daniel Kokotajlo, executive director of the AI Futures Project. Kokotajlo predicted in 2021 that the “age of the AI assistant” would arrive in 2026. (His team recently forecast that artificial superintelligence—AI substantially smarter than any human—would arrive in 2034.)

“I actually think Claude Cowork will be a far bigger disruption to the economic landscape than anything we’ve seen so far, because it will really impact white-collar jobs,” Schirano said.