Google Engineer: Claude Code Solved in One Hour What Her Team Worked on for a Year

Summary

Jaana Dogan, Principal Engineer at Google, reports that Anthropic's Claude Code generated a distributed agent orchestration system in just one hour—a problem Google's team has been working on for a year. While the result isn't perfect, it demonstrates the rapid advancement of AI-assisted coding tools. Claude Code creator Boris Cherny recommends enabling the tool with self-verification capabilities to double or triple output quality.

People

Topics

Detailed Summary

The Experiment: One Hour Against a Year of Development

Jaana Dogan, Principal Engineer at Google and responsible for the Gemini API, shared her experience with Claude Code from Anthropic in a post on X. She gave the system a problem description for a distributed agent orchestration system—a technical problem Google's team had already been working on for a year—and received a working result back within one hour.

The task involved developing systems that can coordinate multiple AI agents. Dogan explained that Google had explored various approaches to this problem without reaching consensus. She emphasized that the input prompt was not particularly detailed—just three paragraphs—and that she created a simplified version based on existing ideas to test Claude Code, since she couldn't use internal company data.

Quality and Realistic Deployment

Dogan admits that the result is not perfect and requires refinement. However, it is comparable to what Google's team had previously developed. She recommends that skeptics of coding agents test them in areas where they have deep expertise.

When asked whether Google uses Claude Code, Dogan answered that it is only allowed for open-source projects, not for internal work. She also emphasizes that the industry is not a zero-sum game and that it makes sense to give competitors credit for their achievements: "Claude Code is impressive work, I'm excited and more motivated to push us all forward."

Rapid Evolution of AI Coding Tools

Dogan outlined the rapid development of AI-assisted programming:

• 2022: Systems could complete individual lines of code
• 2023: They handled entire sections
• 2024: Work across multiple files and build simple applications
• 2025: Can create and restructure entire codebases

She admitted that in 2022 she didn't believe the 2024 milestone could be practically feasible to scale as a global developer product. In 2023, today's level seemed five years away. "Quality and efficiency gains in this domain are beyond what anyone could have imagined so far," she wrote.

Workflow Tips from Claude Code's Creator

Boris Cherny, the creator of Claude Code, published tips for optimal use of the tool at the same time:

Self-verification as key: His top recommendation is to give Claude the ability to verify its own work. This feedback loop doubles or triples the quality of the final output.

Planning mode: Cherny recommends starting most sessions in planning mode and iterating with Claude until the plan is solid. After that, Claude can usually complete the task in one pass.

Automation through slash commands: For recurring workflows, he uses slash commands and sub-agents that automate specific tasks such as code simplification or app testing.

Background agents: For longer tasks, Cherny runs background agents that review Claude's work when it's complete.

Parallel instances: He runs multiple Claude instances in parallel to tackle different tasks simultaneously. His default model is Opus 4.5.

Integration and collaboration: During code reviews, Cherny's team tags Claude directly in colleagues' pull requests to add documentation. Claude Code also integrates with external tools like Slack, BigQuery for data analysis, and Sentry for error logs.

Key Takeaways

• Claude Code solved in one hour a problem Google worked on for a year, demonstrating exponential progress in AI-assisted coding tools
• The performance is impressive but not perfect—improvements and refinements are still necessary
• AI coding tools have evolved in four years from line completion (2022) to full codebase generation (2025)
• Self-verification feedback loops can improve Claude Code quality by two to three times
• The technology industry benefits from fair competition, not zero-sum thinking
• Practical best practices like planning mode, automation, and parallel processing maximize Claude Code's effectiveness

Metadata