Google Is Already Behind on AI Coding. Brin Memo Confirms It.

Google has a coding gap to close — and its own cofounder just said so out loud. In a memo sent to DeepMind staff in recent days, Sergey Brin wrote: "To win the final sprint, we must urgently bridge the gap in agentic execution and turn our models into primary developers," Sherwood News reported, citing three people who read the document.

Google's response is a dedicated strike team. Sebastian Borgeaud, a research engineer who previously led pretraining for Google DeepMind, is running it, with both Brin and CTO Koray Kavukcuoglu directly involved, according to Sherwood News and Capital Brief. The goal, described by one person who read the memo: build AI that can improve itself.

The gap that prompted the memo is not abstract. Google CFO Anat Ashkenazi said in February that roughly 50 percent of Google's code is written by coding agents, Capital Brief reported. At Anthropic, between 70 and 90 percent of company-wide code is written by AI — a figure confirmed by an Anthropic spokesperson — and the head of Claude Code, Boris Cherny, had not written a line of code manually in more than two months as of January, Fortune reported. The measurement contexts differ — the CFO figure covers a broader range of AI-assisted coding, while Anthropic's figure skews toward autonomous agent writes — but the directional gap is large enough that the methodological difference does not erase it.

The compounding logic is explicit. Labs that write more of their own code with AI build faster, which builds better models, which write more of their own code. Dario Amodei, Anthropic's chief executive, predicted at the World Economic Forum in January that the industry is six to twelve months away from AI handling most or all of software engineering work from start to finish, Fortune reported. Whether Google can close the gap before the compounding tilts further in Anthropic's direction is the question worth watching.