Point. Speak. Done.

Yesterday I wrote about Thinking Machines and Anthropic shipping two halves of the same machine inside nine days. The race has moved from the model to the assembly.

Today I’ll write about how Google reimagined the mouse pointer.

Seems like every day there is something new. Not sorta kinda new. Something that could have a profound effect on more than you can imagine. So, you may ask, it’s a pointer, just a simple pointer, what could it possibly change.

What Google created

DeepMind published a research blog on what they call an AI-enabled pointer, powered by Gemini. Four principles. Maintain the flow. Show and tell. Embrace “this” and “that.” Turn pixels into actionable entities.

Point at a paragraph in a PDF and say, “summarize this.” Point at a recipe and say, “double everything.” Point at a couch in an ad and say, “show it in my living room.” Point at a building in a photo and say “directions.” The AI sees what you are pointing at, hears the few words you say, and does the thing.

No window switch. No prompt engineering. No copy and paste. No “open ChatGPT, formulate a question, paste the answer back.”

The cursor has been the same shape on every screen for half a century. Google just gave it the ability to understand intent. Rolling out now in preview to Gemini in Chrome for Google AI Pro and Ultra subscribers in the US. Coming to their new Googlebook laptop this fall as “Magic Pointer,” a deeper version that works at the operating system level across the desktop, files, and native Android apps. Eventually across every surface they own. Android. Cars. Wearables. Workspace.

Some implications

The friction tax around AI just collapsed again. Not for early adopters. For everyone with a browser, and quickly for every application.

Point. Say a few words. Watch the machine do the thing. Typing a prompt becomes what you do when pointing is not enough, the way the command line is what you fall back to when the menu cannot do what you need.

Distribution is the variable most AI strategies are missing right now. OpenAI’s ChatGPT has 800 million weekly users. Chrome has roughly 3.5 billion. Anthropic and Apple are both moving slower at the cursor surface than Google is. Google has Chrome plus Android plus Workspace plus Maps plus YouTube plus a forthcoming laptop. They just put the first piece of an intent-to-action layer into the most universal interface in computing, and they will roll it out whether the user asked for it or not.

More implications IMHO

One. Software UI design starts shrinking as a discipline. Most enterprise software companies invest heavily in menus, navigation, settings panels, and discoverability widgets. Every PM has had the meeting about where to put the new setting. Every designer has spent weeks on the third level of a navigation hierarchy. When the user can point and say, “do this,” much of that work can quickly become redundant. My guess is that product organizations will feel this inside eighteen months if not sooner. Organizations should already be asking which of their software vendors get cheaper to build and which might get displaced entirely.

Two. Web analytics breaks. Click tracking, heatmaps, session replay, conversion funnels, A/B testing of UI elements. All of it assumes the user navigates by clicking buttons. The global web analytics market was $6.26 billion in 2025. None of those tools were designed to track a user pointing at a product and saying, “compare these three.” Every CMO who depends on funnel data is going to need to rebuild how attention is measured when interaction stops being click-based.

Three. The SEO collapse accelerates. Google’s AI Overviews already cut publisher traffic by 33 percent globally in 2025, with major outlets reporting drops between 25 and 90 percent. Pew Research found that users click traditional results 8 percent of the time when an AI Overview is present, versus 15 percent without one. The pointer extends that compression to every webpage. If the AI answers “directions” or “compare these three products” without loading a results page, the content economy is downstream of whether Google’s pointer chose to surface you.

Four. Accessibility opens up. Screen readers, voice control, and assistive devices have been around for decades. They work well for some users and badly for others, depending on the disability and the context. A pointer that understands intent and combines pointing with voice gives users with motor, visual, cognitive, and literacy limitations a different kind of access, one that does not depend on either fluent typing or full screen reading. Google did not market this as accessibility. The accessibility implications may be among the most important things.

Now Imagine with me….

A physician points at a section of a patient chart and says, “summarize this admission, flag interactions, check imaging.” Today that is three tools and twenty clicks.

An analyst points at a paragraph in a 10-K filing and says “compare to last quarter, flag changes in tone.” The read-highlight-copy-paste-prompt loop that consumes hundreds of analyst hours a week collapses to a gesture.

A lawyer points at a clause and says, “show me where this differs from our standard.” Contract review compresses by an order of magnitude.

A field technician points a phone at a piece of equipment and says, “maintenance history, part number.” The work order opens itself.

A student points at a problem and says, “explain this step.” Khan Academy is doing this inside their own app. Google just put it in every webpage.

The architecture and the model capability are here.

The Google question

I have been around long enough to remember that Google invented the transformer architecture in 2017 that everything else in this field runs on. They published BERT. They did the early work on large language models. Their work on speech recognition reshaped the field. Their research output is the foundation of the modern AI industry.

They are also famously bad at bringing things to market.

OpenAI commercialized transformers. Anthropic built a competitor on the same architecture. Nvidia built the chip empire that made all of it run. The pattern is consistent. Google invents many things. Someone else turns it into a product people use.

So, the right question on the Magic Pointer is not whether it is real. It is real. The question is whether Google ships it, polishes it, and makes it work well enough that a billion people use it every day, or whether they announce it, demo it, and let the next company figure out how to deliver it.

Watch what happens next. Either Google turns a research breakthrough into a category-defining product, or someone else takes the idea, the model architecture, and the distribution lessons, and ships a cleaner version twelve months from now. Both outcomes are live.

As I said in my previous piece – find the person setting strategy in these organization and you can start to understand what will get to market and change the next phase of computing.

What the pointer takes from you

In my book I argue that whoever holds the memory holds the customer. Cognitive rent is the tax you pay because someone else holds the context you rely on. The Magic Pointer is an extremely efficient cognitive rent collection tool.

When you click a button, Google knows you clicked. When you point at a couch and say “show me this in my living room,” Google knows what you were looking at, what you wanted, what your living room contains, whether you bought a couch this month, and how you describe the things you care about in your own words while you point at them.

Every gesture, every word, every “this” and every “that” feeds the same machine that will eventually decide what to surface to you, what to recommend, and what to act on without asking.

That is the cognitive rent meter, and it just got an upgrade.

What this means for organizations

The benchmark race is over and it was never the race. The model race moved to reliability in real workflows. Assembly is what determines who wins adoption. Distribution sits underneath all of it. Google just played a great card on distribution. The year is not over. Let’s see what others do.

Ask your management team which of your software vendors get cheaper to build and which get displaced when the pointer understands intent. Ask where in your customer journey the user says “this” and “that” out loud, and who owns that signal. Ask what your business model looks like when the user never loads your website, because the pointer answered the question on the page they were already on.

You cannot sit passively on this. Just because you decided on a strategy yesterday does not mean the field is not shifting underneath you today. The game is being played live. It is unnerving, trust me, I always feel like I am behind. But the implications for your organization are happening whether you feel them today or not.

Gemini in Chrome is rolling out in preview now. Magic Pointer hits Googlebook in fall 2026. The runway between a research blog and a feature on every Chrome user’s screen is shorter than any planning cycle most boards run.

Own the default, own the data. Own the data, own the decade.

Harry Glorikian is the author of The Invisible Interface: How AI Turns Intentions Into Actions, And Who Wins (Ideapress Publishing, distributed by Simon & Schuster, June 2026).