Jaana Dogan, a principal engineer at Google working on the Gemini API, has shared a striking account that highlights how rapidly AI-powered coding tools are advancing. In a recent post on X, Dogan revealed that after testing Anthropic’s AI coding assistant, Claude Code, she received a solution within an hour that closely resembled what her Google team had been developing for nearly a year.
According to Dogan, the challenge she presented to Claude Code revolved around distributed agent orchestrators—complex systems designed to manage and coordinate multiple AI agents working together. She explained that Google engineers have been exploring various strategies for this problem since last year, but aligning on a final architecture has remained difficult.
Expressing her surprise, Dogan wrote that she was not exaggerating when she said Claude Code generated a design similar to Google’s year-long effort in roughly an hour. She clarified that the prompt she provided was intentionally limited to just three short paragraphs and did not include any confidential or internal Google information. Instead, she framed the problem using publicly known concepts to test how the AI would perform independently.
While the output was not flawless and still requires refinement, Dogan said the experience demonstrates how powerful AI coding agents have become. She encouraged sceptics to experiment with such tools in domains where they already possess deep technical expertise, as that is where the value becomes most apparent.
Responding to questions online, Dogan confirmed that Google permits the use of Claude Code only for open-source projects, not internal development. When asked when Google’s Gemini would reach similar capabilities, she responded that teams are actively working on improving both the models and the supporting infrastructure.
Dogan also stressed that progress in artificial intelligence should not be viewed as a winner-takes-all race. She said recognising strong work from competitors is both practical and motivating. Praising Anthropic’s effort, she described Claude Code as impressive and said it has increased her motivation to push innovation forward across the industry.
Reflecting on the pace of change, Dogan outlined how AI-assisted programming has evolved dramatically in just a few years. In 2022, such systems were capable of completing individual lines of code. By 2023, they could manage entire code blocks. In 2024, they began working across multiple files and assembling basic applications. By 2025, AI tools are now capable of creating, refactoring, and restructuring entire codebases.
She added that only a few years ago, she believed today’s capabilities were still far in the future. “The gains in quality and efficiency are far beyond what anyone could have realistically predicted,” she wrote.
The post quickly went viral, gathering over four million views and sparking widespread discussion. Many users highlighted how AI tools can bypass organisational slowdowns and empower individual developers. Others noted that AI-driven code generation is set to dramatically accelerate software development and feature creation across the tech industry.