Developers report productivity gains with Claude

The software, Claude Code from startup Anthropic, has stunned programmers and tech executives who tested it over the holidays, with many reporting productivity gains they didn’t think were possible.

Lots of developers spent their holiday break testing the newest version, Claude Opus 4.5, which runs on their computers. Tech firms have been using code-writing AI for years, and the older ones worked about as well as a new hire. What people are seeing now feels different.

Malte Ubl runs technology at Vercel, which helps build and host websites for people using Claude Code. He finished a tough project in one week that he says would have taken him a year on his own. Ubl worked 10 hours every day during his vacation, making new software. He compared the good feeling from each win to pulling a slot machine lever in Vegas.

Google has been getting most of the attention lately after showing off a really good AI model and tools. But people who follow this stuff closely are still talking about what Anthropic just put out. Claude’s website traffic more than doubled in December compared to last year, and daily visitors on computers went up 12 percent around the world versus last month. That’s based on numbers from Similarweb and Sensor Tower.

The AI agent revolution finally arrives

Tech experts have been saying for a while that AI “agents” would be able to do almost anything for us, but that hasn’t really happened yet. For lots of users, Claude Code was the first time they saw this kind of AI work. It gave them an idea of what might be coming. People are using it to look at government money data, fix broken wedding photos, build websites from nothing, answer tons of emails, or get food delivered.

Shopify’s CEO Tobias Lütke wrote on X that he used it to make software that looks at his MRI scan. Boris Cherny, who’s in charge of Claude Code, said one person hooked up a camera and watched their tomato plants grow with it. “It’s just so different than the AI that came before,” Cherny said.

Since so many regular people wanted to try it, Cherny and his team made a version called Cowork. Instead of the old-school command line screen that the main version has, Cowork looks friendlier and easier to use. They built it in about 10 days, using Claude Code.

Anthropic’s market position strengthens

Anthropic, which people think will go public this year, has always focused on making AI that’s really good at coding first, then at “tooling”, which means the AI can use different software without much help from people. Most tests show it’s the best at coding, and it’s also one of the best at tooling, based on research from UC Berkeley.

Now people want to know what happens next. “The bigger story here is going to be when this goes beyond software engineering,” said David Hsu, who runs Retool, a business-AI startup. Not many Americans work as software engineers. “How far does it go?”

If you're reading this, you’re already ahead. Stay there with our newsletter.