Jaana Dogan, a principal engineer at Google working on the Gemini API, recently shared a surprising experience while testing Claude Code. She said that after giving the tool a short, three-paragraph problem description, Claude Code produced a solution in about one hour that looked very similar to what her Google team had been developing for almost a year.
The task focused on complex systems that manage and coordinate multiple AI agents, an area where her team had explored many ideas but had not finalised a single design.
Claude Code Shows How Fast AI Coding Assistants Are Evolving
According to Dogan, the problem involved building distributed agent orchestrators, which help different AI agents work together smoothly. She explained that Google has been experimenting with several approaches since last year, and internal teams were still not fully aligned on the final direction.
I’m not joking and this isn’t funny. We have been trying to build distributed agent orchestrators at Google since last year. There are various options, not everyone is aligned… I gave Claude Code a description of the problem, it generated what we built last year in an hour.
— Jaana Dogan ヤナ ドガン (@rakyll) January 2, 2026
To test Claude Code fairly, she avoided using any confidential information and instead created a simplified version using only public ideas.
Even with this limited input, Claude Code generated a design that closely matched Google’s internal work. Dogan was clear that the output was not perfect and still needs refinement.
However, she said the speed and quality of the result were impressive enough to change how people should think about AI coding tools. She encouraged sceptics to test these systems in areas where they already have strong technical knowledge, as that is where the real value becomes clear.
Claude Code Sparks Debate On The Future Of AI Coding Assistants
When asked if Google uses Claude Code internally, Dogan replied that it is allowed only for open-source projects, not for internal development. Responding to questions about when Gemini would reach a similar level, she said her team is working intensely on both the models and the supporting systems.
Dogan also stressed that progress in AI is not a zero-sum game. She said it is important to acknowledge strong work from competitors and added that Claude Code’s performance made her more motivated.
She reflected on how quickly AI coding has advanced: from writing single lines of code in 2022, to full sections in 2023, to multi-file projects in 2024, and now entire codebases in 2025. Her post gained millions of views, with many users noting how AI could reduce organisational friction and accelerate individual creativity.
