The head of one of the world’s most talked-about AI coding tools says he has completely stopped writing code himself and says the change has made his work more enjoyable, not less. Boris Cherny, creator and head of Claude Code at Anthropic, said that since November, 100 per cent of his production code has been written by AI. He has not edited a single line by hand. Instead, he now spends his time reviewing the code, checking for safety and correctness, and deciding what should be built next.
Speaking on Lenny’s Podcast, Cherny said he ships between 10 and 30 pull requests every day, often with several AI agents running at the same time. “I have never enjoyed coding as much as I do today,” he said, explaining that AI has removed the repetitive and time-consuming parts of the job.
Coding without typing at Anthropic?
Claude Code began as a small internal project at Anthropic and was not an immediate hit when it was first released. Early users struggled to understand a terminal-based AI coding tool. Over time, as the system improved, usage grew sharply.
Cherny said Claude Code now writes about 4 per cent of all public code commits on GitHub. The share is higher when private repositories are included. Inside Anthropic, the AI reviews every pull request before a human does, acting as an automatic check on code quality.
For Cherny himself, the shift to fully AI-written code happened gradually. Early versions of Claude Code wrote only a small portion of his work. That figure rose through the year and crossed 100 per cent in November. Since then, he said, he has not manually written or edited code.
Claude Code is no longer limited to writing code when asked. Cherny said the system now looks through bug reports, user feedback and internal data, and suggests fixes or new features on its own. This has changed how teams work.
In some cases, AI has solved problems faster than experienced engineers using traditional methods. Cherny described a situation where a newer team member let Claude Code find and fix a memory issue on its own, while he was still analysing the problem manually.
The tool is also being used for tasks beyond engineering. Cherny said AI now helps with project management, syncing information across documents, and handling routine work such as emails and internal coordination.
Cherny says jobs and titles may change
Cherny believes these changes will affect how people think about tech jobs. As AI handles more of the actual coding, he expects the role of “software engineer” to change or even disappear over time. Instead, people may focus more on deciding what to build and why.
He warned that this transition will not be easy. “It’s going to be painful for a lot of people,” he said, especially as AI moves into areas such as product planning, which were once considered human-only roles.
For now, humans are still part of the process. Engineers continue to review AI-written code, especially where safety and accuracy matter. But the direction is clear. AI is taking over the act of writing software, while people increasingly guide, supervise and make final decisions.
Cherny said the speed of change has been hard even for those building the technology. “It’s important to remember how unusual this is,” he said. “Things are changing faster than anyone expected.”
