Google I/O showed off the hands-on Gemini AI, but I have mixed feelings

AI continues to give me mixed feelings about the future, and that’s all the more important after another Google I/O with Gemini last week. On the one hand, Google continues to show some of the most practical and useful ideas for AI, but until I can trust it, I have a hard time getting excited.

There are a million ideas in the AI landscape, but I think it’s safe to say that many of them aren’t worth your time, attention, or all the resources needed to build them. Very often when I hear about new AI products or initiatives, my conclusion is basically “why should I use this?” There are so many ideas that exist as AI for the sake of AI, or to try to find shortcuts that take humanity out of the equation. Those are the products I struggle with: AI absolutely has a place in modern life and these models are powerful, but I firmly believe that everything should be built with the real human using it in mind.

That’s my general thought on why Google’s AI efforts have caught my attention the most.

While most of the AI industry has been doubling down on profitable applications like coding, Google is still trying to figure out where AI can be useful to the average person in places beyond a simple, generic chatbot like ChatGPT. Some great examples of this can be found on Google Home and Fitbit (now Google Health); That’s why I’m really interested in continuing to see more of Google’s AI efforts, unlike many others who I find hard to care about at all.

Announcement: Scroll to see more content

Google I/O 2026 was a combination of both paths. On the one hand, there were some really useful and practical ideas. “Universal Car” As a deal tracker and proactive shopping cart that works on the web it has huge potential. Wearing “Generative UI” in Search to create interactive, use case-specific miniature applications is a great tool for understanding a topic rather than just staring at a wall of text. But, on the other hand, there were some ideas that didn’t convince me so much. The realistic Gemini Omni video It’s undeniably impressive, but the social impact of being able to create such a compelling video is enormous, and the Internet is doing it. already filled with enough AI garbage.

“Live documents” It also worries me. The idea of turning to an AI to help you create outlines is innocent enough, but the example in Google’s scenario of using AI to create a last-minute speech that you forgot to make yourself just seems… bad is the word I’ll use, but the real feeling it brings to mind is “I don’t want to live on this planet anymore.” Personally, I hate these AI demos that basically just show people how to be lazy; It’s human nature to take the path of least resistance, and I need Google and all these other AI companies will stop acting like everyone uses these tools. They are not taking away humanity’s ability to think for itself.

Anywaythe main focus of I/O felt like it was on “agents”, the tools that allow you to pass controls (mostly) to the AI to perform tasks on your behalf. There are mixed feelings here again. On the one hand, I can incorporate the idea of using these tools to speed up my existing workflows. Gemini Spark had a some interesting examplesHow to identify fraudulent subscription fees on your credit card or cardling and summarize deadlines from your inbox for easier viewing.

But as much as I thought about how I could apply this to my life, I kept coming back to the same thought I’ve had about this era of generative AI since the beginning: I just don’t trust it.

Time and time again, Gemini and every other AI tool I’ve tried to integrate into my life, at some point, simply failed to do their job correctly or confidently lied to me about what they were doing. This has improved slightly over time, but is still an inherent problem: Recently, our own Will Sattelberg started testing Google Health Coach and, the first dayhallucinated a run of several kilometers that he simply did not complete. My own experience with Coach has not yielded the same results, but it is not at all uncommon for me to have similar hallucinations or “safe lies” when interacting with the various forms of Gemini. Yet we are constantly told that this is somehow normal.

That’s where my concern for agents comes in. I have no doubt that the tools Google is developing here will be able to do what they promise, I just worry about when they will inevitably fail. Going back to those two examples Google provided, what if my agent completely overlooks one of those hidden fees at a time when my budget is tight? What if I get a random deadline for a project I’m working on? The agent is not going to say When he’s wrong about something, he’ll just say it with 100% confidence, like he always does. With that in mind, I just have to go in and see for myself and at that point, am I saving time?

The first thing we will get out of Gemini Spark It is the “Daily Summary” a feature that intelligently analyzes your data from other Google products to extract some things that you may need to address. I’ve already started watching this and it’s actually really cool. It was able to pull relevant information from my browsing and Gmail, such as an appointment for a phone call (which I never added to my calendar) that even leveraged a previous conversation with Gemini Live about the topic. Clean! Useful! But it also stopped my recent research into a used EV for seemingly no reason. It seemed to go back to “AI for AI’s sake,” because the The summary you provided was completely useless..

Overall, I think Google I/O 2026 was still one of the best displays of practical AI we’ve seen, especially recently, but it still left me torn. Google continues to iterate and test new ideas as it develops these practical applications, which is excellentand what we should demand from all AI companies. But at the same time, there are so many places here where what I’m seeing is something I don’t think I can trust, or basically a glitch in the AI, and that’s not even mentioning the ongoing changes to Search that seem to continue. destroying the web.

That’s where I am: mixed between optimism for the future of Google’s AI efforts and equally concerned about the impact it will have.

What about you? What did you think of Google I/O 2026?

This week’s top stories

More from Google I/O 2026

There’s a lot more to Google I/O 2026, so here’s a quick list of the highlights:

One of the biggest revelations? YouTube Premium Lite is now included in Google AI Pro, which simply makes that storage plan even better.

Launch of Google Health

Out of I/O, the Fitbit app has officially been replaced by Google Health, and the update is now rolling out to Android along with a useful new widget.