A senior Google engineer has sparked debate online after revealing that an AI coding assistant produced, in just one hour, a solution similar to what her team had been working on for nearly a year.
Jaana Dogan, a principal engineer at Google involved with the Gemini API, shared that she tested Claude Code by providing it with a short, three-paragraph description of a technical problem. To her surprise, the AI quickly generated a design that closely resembled Google’s own internal work.
The project involved building systems that coordinate and manage multiple AI agents — an area Google teams have been researching extensively but had not finalised. Dogan noted that several design ideas were still being debated internally.
To ensure fairness, she deliberately avoided using confidential information and instead relied only on public concepts when describing the problem to the AI. Even with that limitation, Claude Code produced a structured proposal that mirrored much of Google’s thinking.
Dogan acknowledged that the output was not flawless and still required human refinement. However, she emphasised that the speed, quality, and direction of the result were startling — and should reshape how engineers think about AI tools.
She encouraged critics to test coding assistants like Claude Code in areas where they already have strong expertise, saying that’s where the tools reveal their true strengths.
When asked whether Claude Code is used inside Google, Dogan clarified that it is permitted only for open-source work, not for internal proprietary projects. On questions about when Gemini might match such capabilities, she said teams are working aggressively on both the models and the infrastructure around them.
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
Importantly, Dogan also rejected the idea that AI progress is a competition where only one company wins. She praised Claude Code’s development and said it actually made her feel more motivated.
Reflecting on the rapid pace of change, she noted how AI coding has evolved from generating small snippets in 2022, to larger sections in 2023, full projects in 2024 — and now potentially entire codebases. Her comments quickly went viral, highlighting growing conversations about productivity, creativity, and the future role of developers.