Google recently made a remarkable announcement: 75% of all new code within the company is now generated by AI and then approved by human engineers.
This rapid acceleration didn't happen in a vacuum. A key driver was intense competitive pressure. Reports of co-founder Sergey Brin urging teams to catch up with rivals like Anthropic's Claude Code, combined with a leak of Claude's own agent-focused roadmap, created a sense of urgency inside Google to push its AI coding capabilities forward, fast.
Of course, this was only possible because several foundational pieces were already in place. First, Google committed to a massive increase in capital expenditure, nearly doubling its 2025 spending to expand the computing power needed to run these AI agents at scale. Second, it built the necessary tools for its engineers. This includes the 'Antigravity' IDE, an environment designed from the ground up for managing AI agents, and the Agent Development Kit (ADK), which standardizes how these agents are built and communicate with each other.
This shift fundamentally changes the role of a software engineer. Instead of spending most of their time writing code line-by-line, they are becoming orchestrators and reviewers. Google's own data shows the power of this new model, citing a recent internal migration project that was completed six times faster with this human-plus-agent approach. However, this speed introduces a new challenge: ensuring the quality and security of the AI-generated code. This is why Google strongly emphasizes that all code is 'approved by engineers' and has developed tools like the DeepMind CodeMender agent, which automatically finds and fixes security vulnerabilities.
Ultimately, Google's milestone is more than just an internal metric. It's a clear signal of where the entire software industry is headed. The era of agentic workflows, where humans and AI collaborate to create software, is no longer a future concept—it's happening now.
- Agentic Workflows: A process where tasks are completed by autonomous AI programs (agents) that can reason, plan, and execute steps, often with human oversight.
- Capital Expenditure (CapEx): Funds used by a company to acquire or upgrade physical assets such as property, buildings, or equipment—in this case, powerful computer servers for AI.
- IDE (Integrated Development Environment): A software application that provides comprehensive tools for programmers to write, test, and debug code efficiently.