Gemini tech chief reveals Claude Code solved year-long problem in one hour

Jaana Dogan, a senior engineer at Google, ignited a debate in the tech industry with a viral post in which she revealed that Anthropic’s Claude Code was able to replicate a complex system architecture in just one hour. In contrast, her team at Google Gemini API had spent nearly a year planning for the same task.

The experiment began when Dogan provided the AI a basic description of a problem her team had been struggling to solve, which was to build “distributed agent orchestrators.” She described this challenge as designing a traffic control system for multiple AI agents working in a team.

Despite using a prompt of only three paragraphs and strictly avoiding proprietary In an internal discussion, an AI generated a working prototype that closely resembled a solution developed by Google over a substantial period.

“I’m not joking, and this isn’t funny. We have been trying to build distributed agent orchestrators at Google since last year,” Dogan wrote on X (formerly Twitter) on January 2, 2026. “I provided Claude Code with a description of the problem, and it generated what we developed last year in just an hour.”

Dogan emphasized that the AI’s output was not a finished product and would still require human refinement. However, this incident underscored a growing sentiment in the tech world: AI tools are circumventing the corporate bureaucracy of meetings and consensus-building that often slows down human teams.

When questioned if Google uses the rival tool internally, Dogan responded that it is allowed only for open-source projects. Regarding Google’s own progress, she confirmed to followers that her team is “working hard right now” on the models and infrastructure to compete.

Dogan’s observation, which quickly amassed over 4 million views, highlights a significant shift in AI coding capabilities: from generating single lines of code in 2022 to overhauling entire codebases by 2025. Her message serves as a stark warning: AI may not be replacing human engineers outright, but it is already taking over the extensive time previously dedicated to the planning stage of their work.