Google Has a Real AI Problem — And I/O Won't Fix It in Two Days

Google I/O 2026 starts Tuesday, May 19 at 10 a.m. PT. Alphabet CEO Sundar Pichai will take the stage and announce a wave of AI updates. The press will cover it like a triumph.

Don't buy it — at least not entirely.

According to MIT Technology Review, Google enters this conference in third place in the foundation model race. A year ago, Google was riding momentum from its Gemini 2.5 Pro launch. Today, Anthropic's Claude Code and OpenAI's Codex have left Google's coding tools in the dust.

How bad is it? Bad enough that engineers at Google DeepMind — Google's own AI research division — were reportedly fighting over access to Claude, Anthropic's product, according to reporting cited by MIT Technology Review. Google had to let some of its own people use a competitor's tool so they wouldn't fall further behind.

The Coding Comeback Attempt

Google is reportedly taking this seriously. According to MIT Technology Review, there's a new dedicated AI coding team at DeepMind. Nobel Prize-winning scientist John Jumper — who shared the 2024 Nobel Prize in Chemistry with DeepMind CEO Demis Hassabis for work on protein-structure prediction software AlphaFold — has reportedly been pulled into the effort.

Expect a coding-related announcement at I/O. MIT Technology Review points to a possible update to Google's agentic coding platform as the likely vehicle.

Google's own engineers still preferred Claude as recently as last month. Unless DeepMind made significant, undisclosed progress in the last few weeks, Google is unlikely to reclaim the coding frontier at this conference. A press release does not equal a breakthrough.

Where Google Actually Leads: Science

Google DeepMind is not irrelevant. It is the only frontier AI company to have earned a Nobel Prize. In AI for scientific research, Google has a genuine, substantial lead over its competitors.

MIT Technology Review notes Google has released tools like its AI co-scientist, which formulates research hypotheses and plans autonomously. That work gets far less media attention than ChatGPT's latest chatbot trick. It also matters far more in the long run.

Science and health AI announcements will likely get buried under Gemini headlines. They deserve attention.

Gemini 4, Android 17, and the Product Pile

On the consumer and developer side, the product announcements will be substantial. According to PCMag, Gemini is expected to receive a full version bump — potentially Gemini 4 — along with a unified multimodal model that handles text, images, audio, video, and code in a single prompt. Larger context windows are also expected.

PCMag also reports on a new proactive agentic feature, potentially called Remy (yes, after the rat from Ratatouille), designed to autonomously handle tasks like answering emails and making calendar entries without you asking.

Android 17 was previewed at a separate event on May 12 called the Android Show. According to CNET, the final release is expected sometime in June or July, timed alongside new Pixel hardware. Google has been holding Pixel 11 details close — I/O may offer a first look.

According to Engadget and PCMag, Google has also been working on a merged Android and ChromeOS laptop platform. That's a direct shot at Microsoft's Windows dominance in laptops. Whether it lands is another question.

Android XR and the Mixed Reality Push

Multiple sources — Engadget, CNET, and PCMag — all flag Android XR as a topic to watch. Google has been quietly building out its mixed reality platform. I/O is likely the venue to show more of what that actually looks like on smart glasses hardware.

Apple's Vision Pro and Meta's Ray-Bans are already competing in this space. Google has been conspicuously quiet. Expect that to change Tuesday.

The Small Stuff That Signals Something Real

Engadget also reported that Google is rolling out redesigned app icons across its entire product suite — Drive, Calendar, Gmail, Meet and others. The 2021 redesign was widely hated because it made all icons look nearly identical. The new icons restore visual differentiation. A company paying attention to basic user frustration shows attention to detail.

What the Coverage Is Getting Wrong

Most tech media will frame I/O as a Google comeback story. That is the easy narrative. It's also incomplete.

The reality is more complicated. Google has real strengths — science AI, search integration, hardware ecosystem, YouTube, Android's market share. It also has a genuine, documented weakness in the exact capability — coding AI — that enterprise customers and developers care most about right now.

Coverage from Engadget and CNET leans heavily on what Google will announce. MIT Technology Review is the only source that asked the harder question: does the announcement match the actual capability?

What This Means for Regular People

If you use Google products — Gmail, Maps, Android, Search — your experience is about to get significantly more AI-integrated whether you asked for it or not. Some of that will be genuinely useful. Some of it will be intrusive.

If you're a developer or enterprise customer choosing between AI coding tools right now, Google is not your best option based on current evidence. Anthropic and OpenAI have the lead. Google knows it.

And if you're a taxpayer, the AI race between American companies and Chinese competitors like DeepSeek is a national security issue. A Google that's falling behind in core AI capabilities is a weaker American hand in that competition.

Tuesday's keynote will be polished, well-produced, and full of impressive demos.