MIT Technology Review is reporting that Google enters its annual I/O developer conference as a clear third-place contender in the foundation model race, particularly regarding coding capabilities. The publication noted that a year ago, Google was strong with Gemini 2.5 Pro, but its coding tools have since been outmatched by Anthropic’s Claude Code and OpenAI’s Codex. It said Google has reportedly allowed some DeepMind engineers to use Claude for their work to avoid falling further behind.
The publication expects Google to attempt a coding comeback at I/O. It cited reports from The Information about a new AI coding team at DeepMind and from the Los Angeles Times that Nobel laureate John Jumper is contributing to these efforts. While a major new coding release, possibly an update to the Antigravity agentic coding platform, is anticipated, MIT Technology Review cautioned against expecting transformative progress.
MIT Technology Review highlighted AI for science and health as Google DeepMind’s strengths. It noted Google is the only frontier AI company to have earned a Nobel Prize and has solidified its lead in AI for science with tools like the AI co-scientist and AlphaEvolve. The publication will also monitor Google’s moves in health and medicine, mentioning the upcoming public release of its AI-powered Health Coach, which focuses on fitness and diet advice rather than medical concerns.
Finally, the publication touched on broader AI industry drama, including the ongoing Elon Musk v. Sam Altman trial. It noted that DeepMind CEO Demis Hassabis has largely avoided such public controversies, though Google itself faced internal dissent last month when 600 employees, many from DeepMind, protested a Department of Defense deal that was subsequently signed.
Full Article: What to expect from Google this week