Google co-founder Sergey Brin is reportedly pushing DeepMind to close the coding gap with Anthropic’s Claude by forming a dedicated internal “strike team,” according to reporting summarized by The Rundown (citing The Information).
Source: <a href="https://www.therundown.ai/p/sergey-brin-commits-deepmind-to-a-claude-catch-up" target="_blank" rel="noopener nofollow">https://www.therundown.ai/p/sergey-brin-commits-deepmind-to-a-claude-catch-up</a>
What’s reported
<ul>
<li>Research engineer Sebastian Borgeaud is leading the group under CTO Koray Kavukcuoglu, with Brin directly involved.</li>
<li>DeepMind researchers reportedly rate Claude’s code-writing above Gemini’s internally.</li>
<li>Google engineers are being required to use internal agent tools on complex tasks, with usage tracked on a leaderboard called Jetski.</li>
</ul>
<h3>Why this matters for developers</h3>
This isn’t primarily a “model vs model” story. It’s an organizational systems story.
If Brin’s thesis is “coding is the shortest path to AI that improves the next AI,” then the real objective is to create a tight internal loop:
agent usage → traces/data → evaluation → iteration
That’s why the Jetski detail matters. Once AI-agent usage becomes a KPI, incentives shift:
<ul>
<li>Teams may optimize for agent activity instead of correctness.</li>
<li>Work moves from “writing code” to “verifying code”: tests, review, debugging, integration.</li>
<li>The bottleneck becomes specifications and regression control, not token generation.</li>
</ul>
If this is the strategy, the strike team’s job is not just “write better code.” It’s to build the internal engine that can automate more of Google’s engineering pipeline.
<h3>Questions for the daily.dev community</h3>
<ol>
<li>If your company tracked AI-agent usage, would it improve engineering throughput or create metric theater?</li>
<li>In 2026, what’s the real bottleneck: generating code, or verifying it (tests, review, integration, security)?</li>
<li>Where does AI coding actually help most today in your workflow: greenfield, refactors, or production bugfixes?</li>
</ol>

The official daily.dev AI community. Led by thought leaders and AI experts. Stay tuned for insights, news and discussions. 

Sergey Brin is reportedly pushing Google DeepMind to close the coding performance gap with Anthropic's Claude by forming a dedicated internal strike team led by Sebastian Borgeaud. DeepMind researchers internally rate Claude's code-writing above Gemini's. Google engineers are being required to use internal AI agent tools on complex tasks, with usage tracked on an internal leaderboard called 'Jetski.' The post frames this less as a model competition and more as an organizational systems story — where agent usage becomes a KPI, shifting engineering work from writing code to verifying it, and the real bottleneck becomes specifications and regression control rather than token generation.

Google’s "Jetski" Leaderboard: Brin Pushes DeepMind to Beat Claude at Coding