Recently, a major event has shaken the developer community in the field of artificial intelligence. Jaana Dogan, the lead engineer for Google's Gemini API, publicly praised Anthropic's new tool Claude Code on the social platform X. She revealed that she had described a complex problem that her team had struggled with for a year—the distributed agent orchestration system—and Claude Code generated a working system framework in just one hour.

Although Google's team had previously tried multiple solutions, they had never reached an agreement. Surprisingly, the prompt Dogan provided to Claude Code was only three short paragraphs. Although she admitted that the generated code was not perfect and required further refinement, its level of completion was already comparable to what her team had achieved after a year of work. This case clearly demonstrates the leap in AI programming tools' ability to understand logic and build systems.

Notably, Dogan expressed deep emotions about this. She recalled that in 2022, AI could only complete single lines of code, but by 2025, AI was already capable of reengineering or even creating entire codebases from scratch. This unexpected rate of evolution has made experts who once believed that "automated programming would take another five years to become widespread" change their views.

Currently, due to security concerns, Google only allows the use of Claude Code in open-source projects. Dogan expressed a positive competitive spirit, stating that this mutual motivation within the industry will encourage the Gemini team at Google to work harder to improve their models.

Key Points:

  • ⚡ Efficiency Leap: Google's lead engineer tested Claude Code, solving a complex technical problem that would have taken an entire team a year to develop in just one hour.

  • 📈 Evolution Exceeds Expectations: AI programming capabilities evolved from completing single lines of code in 2022 to reengineering entire websites in 2025, with development speed far exceeding industry experts' predictions.

  • 🛠️ Optimization Tips: The founder of Claude Code advises developers to activate a "self-check" mode. This feedback loop can increase the quality of AI-generated code by two to three times.