Tomorrow, the tech world turns its gaze to Mountain View, California, as Google throws open the doors for its annual developer conference, I/O. This year, the atmosphere feels a little different; Google is currently navigating a challenging landscape, finding itself a clear third in the fiercely competitive foundation model race.
It’s a stark contrast to just a year ago, when Google I/O 2025 saw the company riding high on the success of Gemini 2.5 Pro. Back then, distinguishing between the top-tier large language models often felt like splitting hairs, a testament to Google’s formidable position.
However, the reputation of a foundation model today increasingly hinges on its coding prowess. For many months, Google’s internal coding tools have been significantly outmaneuvered by offerings like Anthropic’s Claude Code and OpenAI’s Codex, which have set new industry benchmarks.
Reports even suggest that some engineers at DeepMind, Google’s premier AI division, have been permitted to use Claude for their work. This underscores the critical need for Google to catch up, lest they fall further behind in this crucial domain.
Google’s Quest for AI Coding Dominance
As I arrive at the conference, my keenest attention will be fixed on any aggressive moves Google makes to reclaim its frontrunner status. The company is reportedly taking its AI coding crisis very seriously, with whispers of a dedicated new AI coding team forming within DeepMind.
Adding significant weight to these efforts, John Jumper, who shared a Nobel Prize for his groundbreaking work on the protein structure prediction software AlphaFold, is lending his expertise. It would be surprising if we don’t witness a major new coding release at I/O, possibly a substantial update to their agentic coding platform, Antigravity.
Yet, it’s crucial to temper expectations for immediate, transformative breakthroughs. Even internally, Googlers with access to advanced proprietary models were reportedly scrambling for Claude Code access last month. Unless Google has achieved astonishing, rapid progress since then, a return to the absolute coding frontier might not happen overnight.
Pioneering AI in Scientific Discovery
While coding might represent a current vulnerability for Google DeepMind, its prowess in scientific AI remains a conspicuous strength. This is the only frontier AI company boasting a Nobel Prize, earned through the revolutionary AlphaFold project.
As Large Language Models (LLMs) continue to reshape the landscape of AI for science, Google has only solidified its leadership. Last year, the company unveiled remarkable scientific AI tools, including the “AI co-scientist,” which can formulate hypotheses and research plans in response to complex user queries, lauded as an “oracle” by some researchers.
Another impressive system, AlphaEvolve, iteratively discovers novel solutions for intricate mathematical and computational problems. Any new scientific tools announced at this I/O will undoubtedly be worth deep scrutiny, further cementing Google’s role at the forefront of AI-powered research.
Navigating the Future of Health AI
Another critical area to watch closely is Google’s advancements in health and medicine. Google has consistently conducted some of the most advanced research in LLM-based health tools, demonstrating significant potential for real-world impact.
However, OpenAI arguably captured the public’s imagination and defined the health AI conversation with the release of ChatGPT Health earlier this year. Tomorrow, Google is set to make its AI-powered Health Coach publicly available, but promotional materials suggest its primary focus is on fitness and diet advice rather than direct medical concerns.
This raises an intriguing question: Is Google genuinely lagging behind in the direct medical application space, or is the company demonstrating commendable caution in such a high-stakes, sensitive domain? The answer will be key to understanding Google’s long-term strategy in healthcare AI.
Beyond the Demos: The Underlying Currents
Adding another layer to the I/O buzz, the high-profile Elon Musk v. Sam Altman trial will be concluding just north of Mountain View in Oakland. The AI world has seen its fair share of executive drama lately, with animosity between figures like Altman and Anthropic CEO Dario Amodei often making headlines.
In contrast, DeepMind’s Demis Hassabis has largely steered clear of such public controversies, presenting himself as a Nobel Prize-winning scientist focused on innovation. His relatively drama-free public persona offers a refreshing counterpoint in an often tumultuous industry.
Despite this, Google itself isn’t entirely immune to controversy. Just last month, a significant group of 600 employees, many from DeepMind, penned a letter protesting an impending Department of Defense deal, which Google proceeded to sign the very next day. While executives like Hassabis and Sundar Pichai will surely attempt to sidestep these sensitive subjects on stage, such undercurrents are likely to influence perceptions and discussions.
It will be fascinating to observe whether Google can successfully maintain its desired veneer of neutrality amidst these internal and external pressures. The I/O conference is not just about product announcements; it’s also a crucial moment for Google to project its vision and address its challenges in the rapidly evolving world of artificial intelligence.
Source: MIT Tech Review – AI
