Surprise coding breakthrough that made Anthropic into an AI juggernaut

Claude Code, released publicly a year ago, quickly took off with software developers around the world, cementing Anthropic as a leader in lucrative, emerging market for so-called vibe coding products

By Shirin Ghaffary

 

Even Anthropic Chief Executive Officer Dario Amodei was surprised by the initial level of enthusiasm for what became the startup’s next breakthrough product. For months, engineers across the company flocked to an internal AI coding tool called Claude Code that was originally developed as a side project by Boris Cherny, then an employee in an experimental division of Anthropic that he compares to Bell Labs. 

 

“I remember Dario asking, like, ‘Hey, are you forcing engineers to use this? Why is everyone using it?’” Cherny recalled in a recent interview. Actually, Cherny explained, all he had to do was give his co-workers access, and everyone voted with their feet.” 

 

 

It was a sign of things to come. Claude Code, released publicly a year ago this month, quickly took off with software developers around the world, cementing Anthropic as a leader in a lucrative, emerging market for so-called vibe coding products. Other applications like Microsoft Copilot and Cursor were already popular with this cohort thanks to their approachable designs, but Claude Code promised to write and debug code more autonomously. Suddenly, rivals like OpenAI had to race to catch up to Anthropic, rather than the other way around. 

Claude Code hit $1 billion in annualised run-rate revenue in the first six months after its release and has since grown to $2.5 billion, the company said. Once used primarily by AI-forward startups, Claude Code has gained traction with engineering teams at Fortune 500 companies and even among hobbyists lacking technical skills who are interested in building their own apps. It’s been used for everything from growing a tomato plant to helping plan the route of a Nasa Mars rover. On social media, users describe themselves as “Claude-pilled,” or Claude-obsessed.

 

If ChatGPT’s release three-plus years ago showed the potential for generative AI to spit out clever chunks of text, Claude Code’s launch demonstrated how AI can actually perform a portion of a person’s job with limited intervention. Anthropic said some users are now letting Claude Code work autonomously on tasks for more than 45 minutes at a time before stopping it. On average, Claude Code users spend 20 hours a week working with the product.

 

The success of Claude Code is both a testament to AI’s utility in the workplace and a reminder of how hard it can be to predict the ways AI will be used in the real world. Its rapid adoption has revived concerns about job loss, including from Amodei, as well as fears about what happens when autonomous AI tools go awry. (Anthropic says it has staffers working to understand and address both issues.)

 

More imminently, however, the challenge for Anthropic and its peers is proving they can find traction with AI agents for other professions as they have with coders, a group that’s quicker to adopt cutting-edge technology.

 

“There are a few areas where these models are actually really getting to the point where they can really change how work gets done,” said Kate Jensen, head of Americas at Anthropic. “Code happened to be the first one.” 

 

Following Claude Code’s runaway success, Anthropic has been pushing to expand its AI offerings for health care, finance and legal services. Those efforts have helped trigger a market meltdown in recent weeks as investors worry that certain legacy software providers may be rendered obsolete by newer AI advances. On Friday, cybersecurity stocks were the latest to slip after Anthropic unveiled new features in Claude Code to help companies spot security vulnerabilities in their software.

 

“You’re seeing a reaction in large part to just how fast the industry is moving and how quickly the technology is getting better and better,” Jensen said before the security announcement. Though some of the market swing news can be “all-consuming” at times, she said it’s important to remember that legacy software companies can benefit from building on top of Anthropic’s technology. 

For now, much of Anthropic’s traction remains in the realm of coding. At Spotify, roughly two-thirds of the staff have opted in to using Claude Code, outpacing the adoption of any similar product, according to Niklas Gustavsson, the company’s chief architect and vice president of engineering. Spotify has also built its own internal tool called Honk that lets staff talk to Claude Code on Slack to make changes to the music streaming app’s code base. 

 

“For many of us, including myself, it was a bit of a mental jump,” said Gustavsson, who had at least four different Claude Code agents running in the background on his laptop during the interview. “I’ve been in this industry for 30 years now, and the code has always been very front and centre when I’m working with it. Claude Code has completely inverted that.”

 

Gustavsson said it’s hard to capture the gains from AI coding tools in any one metric, but several of Spotify’s internal productivity measurements have gone up since using Claude Code. Anthropic said its engineers’ productivity has grown 200%, by one measure, as Claude Code use increased over the past six months. Its sales teams, data analysts and product managers are using it, too.

 

Cherny, who is now head of Claude Code, feels this change in his day-to-day work. He focuses less on the boring tasks like manually debugging code and more on strategy and talking to customers to get feedback.

 

“We’re starting to see this world where a customer complains about something, and it’s fixed in a matter of minutes,” he said. “It doesn’t take weeks anymore. This is just very exciting for me.”