AI that can code is not replacing engineers at Anthropic just yet.
In a conversation on Wednesday with Salesforce CEO Marc Benioff at the annual Dreamforce conference, Anthropic CEO Dario Amodei told Benioff that even though Claude AI is now writing 90% of code for most teams at the company, humans are still essential.
"I made this prediction that, you know, in six months, 90% of code would be written by AI models," Amodei said. "Some people think that prediction is wrong, but within Anthropic and within a number of companies that we work with, that is absolutely true now."
Benioff followed up by asking when that percentage will rise, and if that means Anthropic will now need fewer engineers.
Amodei said people shouldn't "misinterpret" Claude's ability to write features and help solve long-running bugs.
"If Claude is writing 90% of the code, what that means, usually, is, you need just as many software engineers. You might need more, because they can then be more leverage," said Amodei.
"They can focus on the 10% that's editing the code or writing the 10% that's the hardest, or supervising a group of AI models. And so what happens is, you know, you just end up being 10 times more productive," Amodei added.
Amodei said it was about "rebalancing" rather than replacing.
Anthropic teams are not alone in coding with AI. In March, Garry Tan, the president and CEO of the startup incubator Y Combinator, said in an X post that about a quarter of the founders in the company's 2025 winter batch are generating up to 95% of their code with AI.
A recent study from Stanford found that the rise of AI coding tools is already impacting entry-level software engineering jobs, which could deter young job seekers from the field and result in a broken talent pipeline.
The Stanford researchers found that as of July 2025, employment for developers between 22 and 25 years old had fallen by nearly 20% compared to its peak in late 2022, which coincided with the launch of ChatGPT in November 2022.
Workers with more experience, however, are much less susceptible to the impacts of AI coding tools.
