Key Takeaways
- Boris Cherny is the creator of Anthropic’s Claude Code tool, which writes code on behalf of developers based on a text prompt.
- Cherny hasn’t handwritten code in eight months; instead, he manages anywhere from a few hundred to tens of thousands of AI agents that do the software work for him.
- Cherny compares the development of AI coding agents to the impact of Gutenberg’s printing press.
The creator of Anthropic’s Claude Code tool, Boris Cherny, says he has not written a line of code by hand for about eight months. Instead, he spends his time overseeing AI agents, or AI programs that can act autonomously to perform tasks. He defines problems and reviews their work, overseeing “thousands” or “tens of thousands” of AI agents at a time.
“This morning I was managing maybe a few hundred,” he said at the 25th annual Fortune Brainstorm Tech conference in Aspen earlier this week. This is a noticeable change from even a year and a half ago, he claimed, when developers were using one tab of Claude Code in one window.
“Fast forward to today, it looks very different,” Cherny said. “You have a Claude Code, but it has subagents.” He explained that the human isn’t prompting Claude anymore — another Claude instance now generates the prompts instead, adding another layer of AI involvement.
What does Boris Cherny’s work entail at Anthropic?
At Anthropic, Cherny leads Claude Code, a coding environment where AI agents generate and edit code across large codebases based on text prompts. Since late 2025, AI has written 100% of his own code. On a typical day, he manages AI agents that create code, fix bugs and experiment with new features, often running overnight while he sleeps.
His work now looks less like typing lines of code into an editor and more like running a product and engineering team. He writes detailed briefs for agents, telling them what outcomes they should achieve, which parts of the code base to work on and constraints they must respect. He then monitors progress and reviews summaries of changes.
Cherny compares coding advances to the printing press
At the conference, Cherny said that this leap in coding productivity is comparable to the invention of the printing press. Johannes Gutenberg’s press, developed in Europe around the 1440s, radically cut the cost and time of producing books, making printed material widely available for the first time. That shift helped improve literacy and accelerated the circulation of new ideas.
Cherny believes that analogously, AI coding assistants are cutting down the effort required to create software and opening that capability to far more people.
He framed coding tools as a democratizing force. “It’s quite exciting, because to me it means I can do much more, and the role of a builder is just totally changing,” he said at the conference.
The dangers of AI agents
Despite Cherny’s optimistic outlook, there are unmistakable cognitive risks associated with overusing AI agents.
A Harvard Business Review study released earlier this year found that using AI tools too much can lead to a new problem called “AI brain fry,” or mental fatigue from overseeing AI tools beyond a user’s cognitive limits. The condition can cause brain fog, headaches, slower decision-making and more errors.
In Harvard Business Review’s survey of 1,488 U.S. workers, 14% said they had experienced this kind of mental fatigue.
Key Takeaways
- Boris Cherny is the creator of Anthropic’s Claude Code tool, which writes code on behalf of developers based on a text prompt.
- Cherny hasn’t handwritten code in eight months; instead, he manages anywhere from a few hundred to tens of thousands of AI agents that do the software work for him.
- Cherny compares the development of AI coding agents to the impact of Gutenberg’s printing press.
The creator of Anthropic’s Claude Code tool, Boris Cherny, says he has not written a line of code by hand for about eight months. Instead, he spends his time overseeing AI agents, or AI programs that can act autonomously to perform tasks. He defines problems and reviews their work, overseeing “thousands” or “tens of thousands” of AI agents at a time.
“This morning I was managing maybe a few hundred,” he said at the 25th annual Fortune Brainstorm Tech conference in Aspen earlier this week. This is a noticeable change from even a year and a half ago, he claimed, when developers were using one tab of Claude Code in one window.
“Fast forward to today, it looks very different,” Cherny said. “You have a Claude Code, but it has subagents.” He explained that the human isn’t prompting Claude anymore — another Claude instance now generates the prompts instead, adding another layer of AI involvement.
