Google Co-Founder Sergey Brin Calls for Urgent Focus on AI Agent Development at DeepMind
In a significant internal directive, Google co-founder Sergey Brin has reportedly urged employees at DeepMind, the Alphabet-owned artificial intelligence research unit, to prioritize the enhancement of AI agents. This move comes as the company intensifies its efforts to advance automated coding and research capabilities, positioning itself in a competitive race within the frontier AI landscape.
Internal Memo Highlights Urgency in Bridging AI Agent Gaps
According to a report from The Information, Brin emphasized the need for urgency in a recent internal memo, specifically targeting the development of agent-based AI systems designed to handle complex, multi-step tasks. He wrote, “To win the final sprint, we must urgently bridge the gap in agentic execution and turn our models into primary developers” of code. This directive underscores Google's strategic shift toward leveraging AI for internal development processes, with the broader goal of automating parts of its own software engineering workflows.
Competitive Pressure from Anthropic Drives Google's Push
The report cites sources familiar with the development, indicating that Google's latest push is partly motivated by recent AI model releases from Anthropic. Researchers at Google DeepMind reportedly view Anthropic's coding tools as superior to the code-writing capabilities of Google's own Gemini models. This competitive pressure has prompted Google to assemble a dedicated strike team of researchers and engineers, led by Sebastian Borgeaud, a research engineer at DeepMind who previously oversaw pretraining efforts. The team is focused on improving performance for long-term coding tasks, such as building new software, which often require models to read multiple files and understand user intent.
Leadership Involvement and Long-Term AI Goals
Both Sergey Brin and Google DeepMind Chief Technology Officer Koray Kavukcuoglu have been directly involved with this strike team, highlighting the initiative's importance to Google's leadership. The end goal, as noted in the report, is achieving AI takeoff—where AI systems can improve themselves autonomously. Brin has told staffers that enhancing Google AI's coding abilities is a critical step toward this eventual objective. An advanced coding agent, combined with AI that can solve math problems and run experiments, could potentially automate the work of AI researchers and engineers at scale, mirroring efforts by competitors like OpenAI, which already uses internal tools to boost researcher efficiency.
Strategic Shift in Google's AI Focus
Coding has emerged as a key focus area this year across leading AI labs, with OpenAI and Google working to catch up with Anthropic's advancements in code generation. In response, Google is now placing greater emphasis on models that write code for internal use, moving away from its earlier focus on tools for external customers. This approach involves training models on Google's proprietary codebase, which differs from the public datasets typically used for general-purpose coding agents. While models trained on internal Google code cannot be released publicly, they could facilitate the development of improved models that may later be made available to the broader market.
Google's Internal Adoption and Future Outlook
A Google spokesperson confirmed to The Information that the company has seen “tremendous adoption” of internal coding tools, stating, “Their use has been turbocharging our model and AI tooling development—we’re really focused here.” This internal focus aligns with broader industry trends, as AI labs increasingly prioritize coding automation to drive innovation and efficiency. As Google DeepMind accelerates its efforts under Brin's guidance, the race to develop superior AI agents is set to intensify, with potential implications for the future of software development and artificial intelligence research worldwide.
