Google I/O Preview: A Pressing Bid to Regain AI Leadership

As Google opens its annual developer conference, I/O, this week, the tech giant steps onto the stage in an unfamiliar position: third place in the foundation model race. Just a year ago, after the launch of Gemini 2.5 Pro, Google was neck-and-neck with rivals; today, the landscape has shifted dramatically. The company's coding tools have fallen behind those from Anthropic and OpenAI, prompting an urgent internal response. Yet in scientific AI, Google remains a clear frontrunner. Here's what to watch as the conference unfolds.

A Pressing Coding Comeback

Google's AI coding capabilities have been overshadowed by Anthropic's Claude Code and OpenAI's Codex for months. The gap is so wide that some DeepMind engineers have reportedly been allowed to use Claude for their work—a striking concession. According to The Information, a new AI coding team has been formed at DeepMind to address this crisis. Even more telling, the Los Angeles Times reports that Nobel laureate John Jumper—co-creator of AlphaFold—is now lending his expertise to the coding effort.

At I/O, we can expect a major coding release, likely an update to the Antigravity agentic coding platform. However, expectations should be tempered. Internal access to advanced models hasn't closed the gap with Claude Code, and unless astonishing progress has been made in the past month, Google probably won't reclaim the coding frontier overnight.

Science and Health: Google's Unshaken Stronghold

While coding may be a weakness, AI for science is where Google DeepMind shines. It remains the only frontier AI lab with a Nobel Prize to its name—awarded to Demis Hassabis and John Jumper for AlphaFold. As large language models reshape scientific discovery, Google has doubled down with tools like the AI co-scientist, described by a Stanford researcher as an "oracle" for generating hypotheses and research plans. Another recent release, AlphaEvolve, extends this capability into evolutionary biology.

These scientific contributions may receive less media buzz than chat bots or coding assistants, but their potential impact is immense. At I/O, expect Google to showcase how its AI is accelerating drug discovery, protein folding, and materials science—areas where it still leads the pack.

What to Watch for Overall

Google I/O 2025 will be a tale of two strengths: an urgent push to catch up in coding and a confident stride forward in science. The company's ability to balance these priorities will determine its trajectory in the AI race. For the next two days, all eyes will be on Mountain View—and on whether Google can turn its weaknesses into strengths without sacrificing its core innovations.

This article is based on reporting from The Algorithm newsletter and other sources.