What Should Educators Watch At Google I/O?

A podcast for educators on why Google I/O matters far beyond model benchmarks, focusing on agentic AI, school devices, smart glasses, Workspace changes, and NotebookLM’s growing classroom impact.

Transcript

SPEAKER_00 0:00

If this episode makes you think, please let us know in the comments and support us by subscribing and leaving a review. Thank you. Today we are exploring a timely article written ahead of Google IO, Google's flagship annual event arguing that educators should stop obsessing over whether the next Gemini model beats a rival on benchmarks, and instead pay attention to the announcements that will actually reach classrooms over the next year. The piece says the real education story sits not in model leaderboards, but in agenci features, hardware, browsers, operating systems, and updates to workspace and notebook LM that schools are already using or are about to be sold. And I think that is exactly right. Because benchmark culture is very seductive. Faster, smarter, better reasoning, higher scores, a new version number. It feels like progress because it is measurable and dramatic, but for most schools that is not where the real change shows up. Teachers do not suddenly change their practice because a model has gone from 3.2 to 3.5 or 4.0. Schools change when the tools already embedded in daily life quietly gain new capabilities. A browser button does more than it used to. A student device starts acting like an assistant. A workflow moves from suggestion to action. A notebook tool gets new sharing options and suddenly spreads through a department in a week. That is where the real educational story usually sits. In the ordinary, in the infrastructure, in the thing that looks like a product update but actually changes what students can do, what teachers can trust, and what leaders now have to think about. According to the article, the first thing educators should really watch is the shift from chatbot to agent. Google had already unveiled what it calls Gemini intelligence, described in the piece as a proactive layer handling multi-step tasks across phones, watches, cars, and eventually glasses. The article points to examples like book and parking near a calendar event, building a grocery list from a recipe, and autonomously completing web tasks. It also notes reports of an always-on agent and screenshots of a separate Gemini agent that may share information or make purchases without asking. Now pause there because this is where the conversation for schools changes dramatically. A chatbot is one kind of challenge. A student asks for help, gets text back, maybe copies too much, maybe learns something, maybe doesn't. We have already been living inside that tension for a couple of years, but an agent is different. An agent does not just answer, it acts, it browses, it assembles, it executes. According to the article, this changes the homework conversation entirely. If a student can ask an agent to do my research project on the Industrial Revolution, and the system can browse, draft, and submit, then the question is no longer whether the student used AI. The question becomes whether the student did anything at all. That is a much deeper challenge because many current school policies are still written for a world of assistance. A student may use AI for ideas, a student may use AI to improve grammar, a student may use AI to support research, fine. But what happens when the boundary between help and replacement becomes almost invisible? What happens when the workflow itself can be handed off? That pushes schools into much harder territory, not just academic integrity, but agency itself. What counts as student effort? What evidence do we still accept as proof of learning? What work must remain human if the machine can now complete the sequence? And this is why I think schools need to move beyond a simplistic AI allowed or not allowed conversation. That frame is already too old. The more useful questions are this. Where in learning do we want assistance? Where do we need struggle? Where must we still see the student thinking? And how do we design work that makes the learner's judgment visible, not just the final output? Outsource the doing, not the thinking. But once the tools can do more and more of the doing, schools have to get much better at valuing the thinking. The article's second point is about something called Google Book, a new laptop category running a merged Android and Chrome OS platform with hardware expected from major manufacturers. According to the piece, Chrome OS remains in education for now, while Google Book is aimed initially at consumers and premium productivity users. But the article argues that any successor product running Android apps natively and shipping with Gemini intelligence baked into the operating system will inevitably be marketed into education sooner rather than later. This is one of those moments where a school leader needs to read past the launch language. A company may say a product is not for schools yet, but if schools already use the ecosystem, if the device category overlaps with school budgets, and if the AI layer is embedded into the system rather than added later, then education is not separate from the story. It is simply later in the rollout, and the feature the article flags here is fascinating. Magic pointer, an AI-powered cursor suggesting contextual actions for whatever it hovers over, including single prompt photo editing. Now that may sound like a minor convenience feature, but think about it from an educator's perspective. A capability like that does not just save time, it changes what kinds of manipulation are frictionless. It changes how easily images can be altered, how polished a submission can become, how little technical knowledge is needed to transform media. That matters for assessment. Because one of the biggest mistakes schools can make right now is assuming that because a task looked safe last year, it will remain safe next year. The speed of change is in the interfaces. Not only in the models underneath, but in how effortlessly powerful capabilities are being placed in front of ordinary users. If image editing, rewriting, summarizing, and agentic execution are built into the operating system itself, then the threshold for sophisticated assistance drops even further. And when the threshold drops, assessment design has to respond. This is why we need more product, process, and performance. Not just what the students submitted, but how they got there and what they can do live. Not because we are trying to catch them out, but because the product on its own is becoming weaker evidence. The third area the article says educators should watch is Android XR smart glasses. Google had confirmed a hands-on showcase with live translation, heads-up notifications, and Gemini Live among the expected features. And the article says the real test is a simple one. Would I be comfortable with a student wearing this in my classroom? That is exactly the right test. Because the interesting thing about smart glasses is that they arrive carrying two truths at once. According to the article, they could be genuinely useful for inclusive classrooms, language learning and accessibility. Live translation alone could be enormously helpful for some students. The ability to receive context-sensitive support in real time could transform some learning experiences. But the article also points to a very exposed legal and regulatory backdrop, including a UK inquiry into Meta's smart glasses and a US class action complaint over footage reviewed by contractors. It then raises unanswered questions about data retention, visual input, and whether footage will be used to train Gemini. This is where school leadership gets difficult, because the right response is not panic, but it also cannot be naivety. A wearable AI device is not just another gadget, it changes the privacy conditions of a room. It changes what might be recorded, inferred, translated, prompted, or retained. It changes what classmates and teachers may feel about being observed, and if these products become fashionable as well as functional, then schools will meet them sooner than many policies are ready for. So a really practical leadership question becomes this. Does your acceptable use policy currently say anything meaningful about wearable recording devices with on device AI? Many do not, or they say something written for an earlier era, when smart glasses were fringe products rather than potentially mainstream devices. The article is right that this conversation is coming whether schools are ready or not. And again, there is an education principle underneath the device story. We should not reduce new technology to either threat or opportunity. It is usually both. Live translation could be empowering. Hidden capture could be corrosive. The leadership task is to hold both truths at once and design accordingly. Then the article turns to what it calls the quiet changes that reach every classroom tomorrow. Gemini inside Chrome and Google Workspace. According to the piece, Chrome for Android is gaining built-in image generation and one click page summarization through the Gemini icon. The article also points to an ongoing legal dispute over claims that Google enabled Gemini across Gmail, chat, and Meet accounts without explicit user consent, which Google disputes. For schools using Google Workspace for education, the practical issue it argues is whether new announcements change what data Gemini can access in student and staff accounts, and what the opt-out path looks like. Honestly, this may be the most important section of all, because schools love to focus on the dramatic announcement, the headset, the new device, the keynote moment. But the things that really reshape teacher practice are often much quieter. A browser gains summarization. A workspace tool gains deeper access. An icon appears in a familiar place. A default setting shifts, and suddenly the everyday environment in which staff and students work is different. Not next year, tomorrow. That means leaders cannot treat AI strategy as a special project floating above core infrastructure. It has to be part of ordinary digital governance. Who has access to what? What permissions are turned on? What data is being touched? What features are enabled by default? What can students do inside the tools they already use? The article is absolutely right that if a school has not audited Gemini's current access across workspace, this is the moment to do it. And that is a good example of something I think schools need much more of right now. Boring AI leadership. Not performative AI leadership, not just keynote and big vision language, boring leadership, settings, permissions, procurement questions, device management, policy review, audit trails. These are not glamorous, but this is where culture and safety are actually built. If you ignore the boring part, the shiny part will run straight over you. Finally, the article highlights Notebook LM, and I think this is especially significant because it points to a tool educators are already using at skill. According to the piece, teachers use it for lesson planning, students use it for revision, and university instructors use it for course design, largely because it constrains Gemini to a user-selected source set rather than the open web. The article notes reports that notebooks can now sync directly and generate outputs such as video overviews and charts, arguing that any update to source types, output formats, or sharing controls will land in classrooms within days, not months, because the user base is already there. This is such an important observation, because when a tool already has trust, even partial trust, its updates matter more than a brand new product launch. Notebook LM has gained traction in education precisely because it feels more bounded, more source aware, more useful for working with a defined body of material, rather than just wander in the open web. And that makes it educationally interesting, because schools do not only need raw generative power, they need structured thinking support, they need tools that help students work with texts, compare ideas, organize knowledge, revise from sources and build understanding with some guardrails in place. But even here, the same question returns. Are we using it to support thought or replace it? That is always the hinge. A source-grounded notebook can be brilliant for helping a student explore a complex reading set. It can also become a way of skipping the reading if the student simply consumes the generated summary and mistakes that for understanding. So the tool itself is not the answer. The design of the learning around it is the answer. What do we ask students to do with the notebook? How do we make their response visible? How do we ensure the summary is a starting point, not the final act? And maybe that is the deepest thread running through this whole article. The big educational changes are increasingly arriving sideways, through browsers, through devices, through operating systems, through assistants that become agents, through wearables that blur support and surveillance, through notebook tools that change what revision feels like. If schools wait for one giant headline moment to tell them the future has arrived, they will miss how it is already embedding itself in ordinary practice. So for me the challenge this article poses is quite simple. Stop watching AI like a horse race and start watching it like an infrastructure shift. Ask not only which model is winning, but where agency is moving, where control is slipping, where convenience is rising, where evidence of learning is weakening, and where the human role needs to become more visible and more deliberate. Because the benchmark story may excite analysts, but the story that matters for schools is much closer to the desk, the browser tab, the school issued device, the homework task, and the question every educator now has to keep asking what exactly are we helping students become in a world where the tools can increasingly act on their behalf? That's all for today, thanks for listening.