Google co-founder Sergey Brin has told DeepMind employees in a recent internal memo that the company must "urgently bridge the gap in agentic execution" with Anthropic's Claude, according to a report from The Information published Sunday. The response: a dedicated "strike team" of researchers and engineers assembled inside DeepMind with a single mission: make Gemini competitive in AI coding.

"To win the final sprint, we must urgently bridge the gap in agentic execution and turn our models into primary developers," Brin wrote, according to three people with direct knowledge of the memo. The Verge confirmed that Brin specifically demanded "every Gemini engineer must be forced to use internal agents for complex, multistep tasks."

The admission is remarkable. Google, the company that invented the transformer architecture that powers every modern AI model, is now openly acknowledging it has fallen behind a startup that didn't exist four years ago. Not in research. Not in scale. In the thing that actually makes money right now: writing code.

Claude Code Forced Google's Hand

Anthropic's Claude Code has been dominating the AI coding market since its expanded release earlier this year. The tool's ability to handle complex, multi-file coding tasks with minimal supervision has made it the default choice for developers at startups and enterprises alike. Cursor, which relies heavily on Claude models, just raised $2 billion at a $50 billion valuation. The coding use case alone is generating billions in annual revenue across the ecosystem.

Google's Gemini models, despite massive investment, have struggled to match Claude's performance on real-world software engineering tasks. On SWE-bench Pro, the industry standard for coding evaluation, Anthropic's models have consistently outperformed Gemini variants. Developers have voted with their wallets: Ramp data from Q1 2026 shows Anthropic now commands 37% of enterprise AI spending, overtaking OpenAI's 33%. Google's share of that market remains in single digits.

The Self-Improving AI Angle

What makes Brin's memo especially notable is the stated end goal. This isn't just about shipping better developer tools. The strike team's objective, according to The Information, is to create an AI system that can improve itself. In other words, the race to build the best AI coder is a proxy war for the race to build AI that can do its own AI research.

If an AI model can reliably write, test, debug, and ship complex software, the next logical step is pointing it at its own codebase. Google, Anthropic, and OpenAI all understand this. Whoever builds the best coding agent gets a compounding advantage: their AI gets better at making AI better.

Brin's return to hands-on involvement at DeepMind signals how seriously Google is treating this gap. The co-founder has been increasingly active inside the AI division over the past year, but a direct memo to engineering staff demanding they adopt internal agents is a new level of urgency.

The Competitive Landscape

Google isn't the only company scrambling. OpenAI declared "Code Red" on Anthropic earlier this month, restructuring its entire product strategy around enterprise and coding use cases. It killed Sora (shutting down the app April 26) and is consolidating everything into ChatGPT as a super-app. Microsoft, Google's other major AI competitor, is simultaneously building its own models to reduce dependence on OpenAI.

The AI coding wars have become the first real revenue battleground for frontier AI labs. Cursor at $50 billion. Claude Code driving Anthropic's enterprise growth. GitHub Copilot underpinning Microsoft's AI strategy. And now Google assembling a strike team to avoid being left behind entirely.

For Google, the stakes extend beyond revenue. The company uses AI internally to write a significant and growing portion of its own code. If Gemini can't match Claude's coding capabilities, Google faces a choice: use an inferior internal tool or pay a competitor to help build Google's own products. Neither option is acceptable for a company that views AI as an existential priority.

Google, Anthropic, and OpenAI did not respond to requests for comment on the strike team or Brin's memo.

First reported by The Information. Additional reporting by The Verge, Sherwood News, and Heise.