Coding-Tool Arms Race
- Google is assembling an elite internal team to close a perceived coding-model gap with Anthropic. - The effort reportedly involves Sergey Brin and focuses on models that can progressively improve software coding capabilities. - The competition over developer tooling is becoming strategic for big tech, intensifying pressure on third‑party code-assist startups. (the-decoder.com)
Google has put together a small internal team to improve its artificial-intelligence coding tools after executives concluded Anthropic is ahead in software programming, according to reports from The Information and follow-up coverage by The Verge and The Decoder. (theinformation.com) (theverge.com) (the-decoder.com) The group sits inside Google DeepMind and is led by Sebastian Borgeaud, a senior researcher who previously worked on model pretraining, with Google co-founder Sergey Brin and DeepMind chief technology officer Koray Kavukcuoglu directly involved, The Decoder reported, citing The Information. (the-decoder.com) (theinformation.com) The target is not just autocomplete. The team is working on “long-horizon” coding, meaning models that can read a codebase, plan several steps, edit multiple files, run tests, and finish a task with less hand-holding from a human developer. (anthropic.com) (github.com) (the-decoder.com) That has turned into a live contest in enterprise software. Anthropic says Claude Code can change files, run commands, and work from the terminal or an integrated development environment, while Google sells Gemini Code Assist in free, Standard, and Enterprise tiers through Google Cloud and its developer tools. (anthropic.com) (claude.com) (developers.google.com) (docs.cloud.google.com) Google has been expanding its own coding product line. At Google I/O on May 20, 2025, the company said Gemini Code Assist for individuals and Gemini Code Assist for GitHub were generally available and powered by Gemini 2.5. (blog.google) Anthropic has kept shipping on the other side of the market. In April 2026 alone, the company posted updates on redesigned Claude Code desktop tools, “routines,” subagents, and managed agents for enterprise customers. (claude.com) (anthropic.com) Brin has tied the coding push to a bigger goal inside Google. The Decoder reported that he told employees stronger coding agents are a step toward systems that can improve their own software, and The Verge reported he pushed Gemini engineers to use internal agents for complex multi-step work. (the-decoder.com) (theverge.com) The pressure lands on smaller code-assistant companies too. When Google, Anthropic, Microsoft-owned GitHub, and OpenAI all treat programming tools as a route into enterprise budgets, independent startups have to compete against companies that already control the cloud, the model, or the developer workflow. (developers.google.com) (anthropic.com) (theinformation.com) For now, the clearest signal is internal: Google is no longer treating coding help as a side feature. It is treating code generation, debugging, and multi-step software work as a frontline contest with Anthropic. (theinformation.com) (theverge.com)