Google Formed a Strike Team to Catch Claude Code. Sergey Brin Is Running It.

Google co-founder Sergey Brin sent an internal memo to AI engineers and researchers with an unambiguous message: “To win the final sprint, we must urgently bridge the gap in agentic execution and turn our models into primary developers of final code.”

The Information reported on April 21 that Google DeepMind has assembled a dedicated group with a single mandate: improve Gemini’s coding capabilities fast enough to close the gap with Anthropic’s Claude. Brin, who had been operating in a semi-retired advisory role for years, is now directly involved. So is DeepMind CTO Koray Kavukcuoglu. The team is led by Sebastian Borgeaud, who previously ran pretraining for Gemini.

The Gap Brin Is Worried About

The specific area isn’t code completion or autocomplete benchmarks. It’s long-horizon agentic coding: reading through large codebases, maintaining context over hours-long sessions, and executing plans that span dozens of files and thousands of lines.

This is precisely where Claude Code built its reputation after Anthropic launched it for general availability in mid-2025. Developers running complex, multi-step tasks on Claude Code consistently report it outperforming alternatives on sustained, coherent execution. Benchmarks tell part of the story, but the developer community’s workflow adoption tells a clearer one.

The numbers back up Brin’s concern. According to Google’s own leadership, Anthropic writes close to 100 percent of its code with AI assistance. Google is at roughly 50 percent. That’s not a rounding error — it’s a flywheel problem. An AI company that uses AI more effectively in its own development can ship improvements faster, which widens the gap further. Brin’s memo is essentially an acknowledgment that Google is on the wrong side of that dynamic.

The Internal Tension

The story gets more pointed from there. According to reporting by The Decoder, some Google DeepMind engineers have been granted access to Anthropic’s Claude for coding purposes, while the rest of Google’s engineering staff is restricted to internal Gemini models. This two-tier access created real friction.

When the question of equalizing access came up internally, one proposed resolution was to remove Claude access from everyone. DeepMind engineers objected strongly enough that several reportedly threatened to leave if that happened. The result: DeepMind kept its Claude access, the divide remained, and the internal tension became visible enough to leak.

This puts Google in an awkward position. The company is simultaneously building what it insists is a world-class AI coding tool and quietly giving its best AI researchers access to a competitor’s product for their own work. The message that sends internally is hard to spin.

What Google Is Doing About It

The strike team’s mandate, per Brin’s memo, isn’t just catching up on a benchmark. The stated goal is building systems that can improve themselves, which is the compounding advantage Anthropic currently holds. If your AI tools help you build better AI tools, you get faster with time. Google is behind on that loop.

BusinessToday reported that Google employees from the AI unit cited concerns the company is losing its edge specifically on AI-assisted development tooling. That’s a significant admission from the company that created Gemini.

On the product side, Google Antigravity and Firebase Studio are Google’s main competitive responses for developers. Gemini CLI has been shipping regular updates, including dynamic sandbox expansion and worktree support in v0.37.0. But these efforts are scattered across teams, which is part of what the strike team is meant to fix: putting dedicated, focused resources on a single problem rather than distributing it across organizations.

Why This Matters

Google is not a company that usually lets competitive gaps go public this way. The leaked memo, the reporting on internal access policies, and the TechRadar coverage all happening within a few days of each other signals that the situation is significant enough to generate pressure that’s hard to contain.

For developers, the immediate read is that competition in AI coding tools is about to intensify from Google’s direction. The strike team has Brin’s direct involvement and DeepMind leadership behind it. Whether that translates into a better Gemini for coding in three months or three years is the open question — but the direction of travel is now unambiguous.

Sources: The Information, The Decoder, TechRadar, BusinessToday