Google Principal Engineer Jaana Dogan revealed on January 2, 2026, that Anthropic’s Claude Code AI coding tool recreated in one hour what her team spent a year developing. The distributed agent orchestration system software that coordinates multiple AI agents matched Google’s year-long effort from a simple three-paragraph prompt, highlighting how rapidly AI-powered development tools are advancing even at the world’s leading tech companies.
What Happened
Dogan, who leads work on Google’s Gemini API, posted on X that Claude Code generated a working system comparable to what her team had been building since last year. The prompt wasn’t detailed, just three paragraphs describing the problem. She built a simplified version based on existing ideas to test Claude Code, since she couldn’t use internal company details.
“I’m not joking and this isn’t funny,” Dogan wrote, explaining that various approaches had been explored at Google without reaching consensus. The output wasn’t perfect and required refinement, but the result demonstrated where AI coding capabilities stand today.
She recommended skeptics of coding agents try them in domains where they already have expertise to properly judge the results. Claude Code is only permitted at Google for open-source projects, not internal work.
Why It Matters
Dogan’s assessment gained attention partly because she was praising a competitor’s product while working at Google. When asked when Google’s Gemini would reach similar capabilities, Dogan responded that her team is “working hard right now” on the models and infrastructure.
The incident underscores how AI coding tools are reshaping software development across the industry. Google disclosed in July 2025 that AI now writes approximately 50% of new code at the company, up from 25% reported in late 2024. Anthropic CEO Dario Amodei said in October 2025 that most teams at his company now use Claude to write 90% of their code.
Boris Cherny, creator of Claude Code at Anthropic, shared in late December that 100% of his contributions to the project over the previous 30 days 259 pull requests and roughly 40,000 lines of code were written entirely by Claude Code using the Opus 4.5 model. He didn’t open an integrated development environment (IDE) even once during that month.
How Claude Code Works
Claude Code is an agentic coding tool that operates directly in the terminal, integrated into local development environments. Unlike traditional AI coding assistants requiring web-based interfaces, Claude Code understands codebases, automates routine tasks, and provides intelligent suggestions through natural language commands.
Key capabilities include:
- Refactoring code and debugging errors across multiple files
- Managing Git workflows without leaving the terminal
- Generating entire codebases from high-level descriptions
- Self-checking work through feedback loops that can double or triple output quality
- Maintaining long context windows up to 200,000 tokens
The tool launched in early 2025 and operates with a focus on terminal-based interaction, eliminating the need for additional servers or browser-based interfaces.
Evolution of AI-Assisted Coding
Dogan outlined the swift progression of AI programming capabilities:
- 2022: Systems could complete individual lines of code
- 2023: Entire sections could be generated
- 2024: Tools worked across multiple files and built simple apps
- 2025: Tools can create and restructure entire codebases
“In 2022, I didn’t believe the 2024 milestone could be practically feasible as a global developer product,” Dogan wrote. “Quality and efficiency gains in this domain are beyond what anyone could have imagined so far”.
The shift is transforming how software gets built, with developers increasingly serving as reviewers and orchestrators of AI-generated code rather than writing every line themselves.
Google-Anthropic Partnership
Google is actually a major investor in Anthropic, having committed approximately $3 billion to the AI startup. The companies deepened their partnership in October 2025 when Google agreed to supply Anthropic with up to one million specialized tensor processing units in a deal worth tens of billions of dollars.
The partnership is expected to bring over a gigawatt of capacity online in 2026, representing one of the largest hardware commitments in AI industry history. Anthropic cited better cost efficiency, strong performance, and Google’s prior experience with the processors as reasons for using Google’s TPUs for training and running Claude AI models.
“This industry has never been a zero-sum game, so it’s easy to give credit where it’s due even when it’s a competitor,” Dogan wrote. “Claude Code is impressive work, I’m excited and more motivated to push us all forward”.
What’s Next
Both Google and Anthropic continue racing to advance AI coding capabilities. Dogan confirmed her team is actively working on improving Gemini’s models and infrastructure to reach competitive levels.
The Google-Anthropic TPU agreement will dramatically expand computing resources throughout 2026, likely accelerating development of next-generation AI coding tools. As AI-written code becomes the norm rather than exception at major tech companies, the role of human developers will continue evolving toward higher-level oversight, architecture decisions, and quality control.
The broader implication extends beyond coding efficiency it signals that even the most advanced engineering teams at companies like Google can benefit from external AI tools for complex technical challenges previously thought to require months or years of specialized work.
Featured Snippet Boxes
What is Claude Code and how does it work?
Claude Code is an AI coding tool developed by Anthropic that operates directly in the terminal. It generates, refactors, and debugs code through natural language commands, understanding entire codebases with context windows up to 200,000 tokens. The tool can create complete systems from simple descriptions and self-check its work for improved quality.
How did Claude Code recreate a year of Google work in one hour?
Google Principal Engineer Jaana Dogan gave Claude Code a three-paragraph description of a distributed agent orchestration problem her team had been solving since last year. Claude Code generated a working system in one hour that matched what Google built, though it required refinement. The output demonstrated AI coding capabilities can now handle complex architectural challenges quickly.
How much code does AI write at Google and Anthropic?
Google disclosed in July 2025 that AI writes approximately 50% of new code at the company, up from 25% in late 2024. At Anthropic, CEO Dario Amodei said most teams use Claude to write 90% of their code. Claude Code creator Boris Cherny confirmed 100% of his contributions in December 2024 were AI-generated.
What is the relationship between Google and Anthropic?
Google is a major investor in Anthropic, having committed approximately $3 billion to the AI startup. In October 2025, they expanded their partnership with Google agreeing to supply up to one million tensor processing units worth tens of billions of dollars. The deal will bring over a gigawatt of computing capacity online in 2026.
