Google’s Secret AI Strike Team Has One Goal: Self-Improving AI
Sergey Brin is back — and he’s not building products. He’s rebuilding Google’s future from the inside out.
The Rundown
Google co-founder Sergey Brin is personally stepping back into the arena — assembling a high-priority “strike team” inside DeepMind to push Gemini ahead of Anthropic’s Claude in one critical domain: coding.
But this isn’t just about better autocomplete or developer tools.
The real mission? Build AI that can improve itself.
And according to Brin, coding is the fastest path to get there.
Why Coding Is the Battleground
Inside Google, there’s a growing realization: the next leap in AI won’t come from bigger models alone.
It will come from recursive improvement — systems that can write, test, and refine their own code.
That’s the unlock.
And right now, Google isn’t leading.
Internal assessments reportedly show that Anthropic’s Claude is outperforming Gemini in code generation — a gap that triggered Brin’s direct involvement.
Instead of waiting for incremental improvements, he’s accelerating things with a focused internal push.
Meet the Strike Team
Leading the effort is Sebastian Borgeaud, a research engineer known for running DeepMind’s pretraining systems — essentially the foundation layer of modern AI models.
He’s operating under DeepMind CTO Koray Kavukcuoglu, with Sergey Brin closely involved.
This isn’t just another research group.
It’s a high-urgency, mission-driven unit with a singular focus:
- Make Gemini better at coding than any competitor
- Turn coding into a feedback loop for self-improvement
- Accelerate Google’s internal AI transformation
Think of it less like a product team — and more like an internal special forces unit.
Google Is Forcing Its Own AI Adoption
One of the most interesting parts of this story isn’t the model — it’s the behavior change inside Google.
Gemini engineers are now required to use Google’s internal AI agent tools when working on complex tasks.
And yes — it’s being tracked.
There’s even an internal leaderboard called “Jetski” ranking usage.
This creates a powerful dynamic:
- Engineers are incentivized to rely on AI
- AI tools get real-world stress testing
- Feedback loops improve faster
In simple terms:
Google is forcing itself to become an AI-first company from within.
This Isn’t About Catching Up — It’s About Rewiring Google
From the outside, it might look like a response to competition.
But this goes deeper.
This isn’t just about beating Anthropic or OpenAI on benchmarks.
It’s about transforming how Google operates at its core.
Because the companies that win this race won’t just have better models —
They’ll have AI deeply embedded into every layer of their organization.
Continue reading for the bigger picture → how this changes the AI race entirely.
The Real Prize: AI That Builds AI
In Brin’s internal memo, one idea stands out above everything else:
The ultimate goal is AI that trains the next generation of AI.
This is the concept of recursive self-improvement — and it’s one of the most important ideas in the entire AI landscape.
Here’s why it matters:
- Today: Humans train AI models
- Tomorrow: AI assists humans in training models
- Endgame: AI trains itself
Once that loop becomes efficient, progress doesn’t just continue — it accelerates.
Exponentially.
Why Google Fell Behind (And Why It Matters)
At the end of 2025, Google dominated the AI conversation.
Gemini was everywhere. Momentum was strong.
But in early 2026, things slowed.
Meanwhile:
- Anthropic doubled down on deeply integrated AI systems
- OpenAI continued embedding AI into workflows and tools
The gap isn’t just model quality anymore.
It’s how deeply AI is woven into the company itself.
And that’s exactly what Brin is trying to fix.
The Hidden Strategy Behind This Move
This “strike team” signals something bigger than a technical upgrade.
It reveals Google’s new strategy:
- Use coding as the gateway capability
- Turn engineers into AI-powered operators
- Build internal systems that improve themselves
Instead of shipping features first, Google is rebuilding its internal engine.
Because once that engine works —
Everything else becomes easier.
What Most People Are Missing
Most coverage will focus on competition:
Google vs Anthropic vs OpenAI.
But the real story is structural.
We’re watching a shift from:
- AI as a tool → to AI as infrastructure
The winners won’t just build better models.
They’ll build companies where:
- AI writes code
- AI improves systems
- AI accelerates decision-making
That’s a completely different level of advantage.
What Happens Next
If this strategy works, expect three major shifts:
- Faster model iterations — because AI helps train AI
- Internal automation — entire workflows handled by agents
- Compounding advantage — improvement speed becomes the moat
And if it doesn’t?
Google risks falling behind companies that already operate this way.
Final Thought
Sergey Brin isn’t trying to win a feature war.
He’s trying to change how Google evolves.
Because in the AI era, the most powerful company won’t be the one with the smartest model —
It’ll be the one that improves the fastest.
If you enjoyed this breakdown, share it with someone building in AI.
The Daily Upgrade
Smarter every day, one insight at a time.