Anthropic’s Claude Code creator hasn’t written code in 8 months

Boris Cherny, the creator of Anthropic’s Claude Code tool, has not written a single line of code by hand in eight months. Instead, the software engineer spends his days managing anywhere from a few hundred to tens of thousands of AI agents — autonomous programs that write, debug, and experiment with code on his behalf.

From one tab to thousands of agents

Cherny spoke about this dramatic shift at Fortune’s 25th annual Brainstorm Tech conference in Aspen. He noted how much the workflow has evolved: just a year and a half ago, developers were using a single Claude Code tab in one window. Today, each Claude Code instance spawns subagents, and rather than a human prompting Claude directly, another Claude instance now generates those prompts — adding an entirely new layer of AI involvement.

His daily routine now resembles running a product and engineering team more than writing software. He drafts detailed briefs for agents, specifying target outcomes, relevant sections of the codebase, and constraints to follow. Overnight, agents work through bugs and test new features while he sleeps, leaving him summaries to review each morning.

A productivity shift comparable to the printing press

Cherny frames this transformation in sweeping historicalerms, comparing it to Gutenberg’s printing press, which dramatically cut the cost of producing books in the 1440s and accelerated the spread of ideas. He believes AI coding tools are doing something similar — drastically lowering the barrier to building software and putting that power in far more hands. “The role of a builder is just totally changing,” he said at the conference.

However, experts caution that overseeing large numbers of AI agents carries its own cognitive costs. A Harvard Business Review study found that 14% of surveyed workers reported mental fatigue from managing AI tools — a condition researchers have termed “AI brain fry,” linked to brain fog, slower decision-making, and increased errors.