Gemini Spark is the most impressive and terrifying AI experience I’ve had yet
It’s a remarkable piece of technology. But the future sure is creepy.
AI
Tech
Google
Gemini Spark is the most impressive and terrifying AI experience I’ve had yet
It’s a remarkable piece of technology. But the future sure is creepy.
According to every product demo from the last four years, planning a trip is a killer use case for AI. Just tell it where you’re going, they all promise, and your chatbot / agent / other buzzword will exhaustively search travel options, read up on all the fun things to do, check all the local hotspots, and offer you a fully fledged itinerary. So far, I’ve found this to work only in the most generic ways: If you want to do the six most obvious things in any city on planet Earth, AI has you covered, but that’s about as far as it goes.
I had a very different experience using Spark, Google’s new always-on AI agent. Spark is a hugely ambitious thing: Google intends it to be the interface through which you can use external apps, and over time even operate your computer. (“OpenClaw with better internet access” is a not-wrong way to describe it.) Spark is currently rolling out to Google’s $99 / month AI Ultra plan, but Google allowed me to try it early. I tested some simple action-oriented stuff, like having Spark go through my Gmail inbox and suggest a bunch of things I should unsubscribe from and having it comb my Google Docs for old tasks I still haven’t finished. In both cases, it did a fine job, even creating me a nicely organized document with a bunch of links to quickly unsubscribe from various marketing emails.
Then I gave Spark a simple trip-planning job. “I’m going to be in Hershey PA with my wife, two kids, and dog the weekend of July 18th. Can you make a plan for the whole weekend, including places to stay, eat, things to do, and everything else?” I left out a few salient details, like the concert tickets I have for that Saturday night, but figured I’d start with the six most obvious things to do in Hershey and go from there.
Gemini Spark is the most impressive and terrifying AI experience I’ve had yet
It’s a remarkable piece of technology. But the future sure is creepy.
AI
Tech
Google
Gemini Spark is the most impressive and terrifying AI experience I’ve had yet
It’s a remarkable piece of technology. But the future sure is creepy.
According to every product demo from the last four years, planning a trip is a killer use case for AI. Just tell it where you’re going, they all promise, and your chatbot / agent / other buzzword will exhaustively search travel options, read up on all the fun things to do, check all the local hotspots, and offer you a fully fledged itinerary. So far, I’ve found this to work only in the most generic ways: If you want to do the six most obvious things in any city on planet Earth, AI has you covered, but that’s about as far as it goes.
A few minutes later, Spark pinged me back. “I have created a comprehensive, family-friendly, and dog-friendly weekend itinerary for your trip to Hershey, PA, from Friday, July 17 to Sunday, July 19, 2026.” It shared a link to a Google Doc it had made me, and a couple thousand words of shockingly detailed, useful itinerary.
To begin with, it offered driving directions from my house, an address that of course Google knows but I had not offered. It included a few hotel options, including their pet fees, and some dog-friendly activities that Frida might like. I never told Google my dog’s name is Frida; my only guess is that Spark found it through emails from my vet.
Spark also casually noted that my son Lewis will get into Hershey Park for free, because he’s not a year old yet, but that because Arthur is three, he’ll need a ticket. I don’t know if Spark was guessing what time Lewis naps in the afternoon, or if it knew it somehow, but it was right to schedule nap time for 1:30PM.
The whole Spark itinerary was filled with details like this. It included my wife’s name, and took into consideration the fact that she doesn’t like to eat onions or scallions. It included the Thomas Rhett and Niall Horan concert on Saturday night, presumably based on the Ticketmaster confirmation in my email, and noted that parking is included in the tickets we bought. When I got to the part where it mentioned getting a babysitter that night, I remembered to note that my parents are coming along for just that purpose, so I added a note to the conversation.
“That is a wonderful update!” Spark replied, happily calling my parents by their names, and switching its recommendations from a hotel to an Airbnb. When I asked Spark to put all the information in a Google Doc and share it with Anna, it found my wife’s email, attached the document, drafted a note that sounded like we were business colleagues instead of a married couple, and sent it along.
The only time Spark failed me was when I asked it to book an Airbnb. It prompted me to allow Gemini to interact with websites on my behalf, navigated to Airbnb, and appeared to be promptly blocked. “Due to security and authentication policies on Airbnb, I am unable to log in, handle payment, or complete bookings directly on your behalf.” It instead offered up a few relevant places with availability on the right days, and reminded me of the information I’d need to book.
On the one hand, this is one of the most astonishingly impressive AI experiences I have ever had. Google’s AI prowess, combined with the vast quantity of data it has on me through Google’s Personal Intelligence feature, produced a personalized and useful itinerary that was well suited to my needs and my family. It put together the itinerary, and presented it to me, the way an actual human assistant would have — with lots of details specific to our situation, with the names of the people who matter, and with affordances made for all of our specific needs. Every time I read the itinerary I’m blown away by another detail of it; I suspect we’ll follow it almost exactly.
On the other hand, I can’t shake the deeply creepy feeling I get from the whole thing. What Spark did feels sort of magical, and very invasive. It’s weird that Spark is so casually telling me the names and ages of my children, reminding me that it knows where I live, and finding information I know for a fact I’ve never volunteered to Google. Intellectually, I know that Google knows an incredible amount about me — add up my emails, my calendar, my photos, and my search history, and you’ve pretty much got me pegged. But seeing Spark treat all that data not as something to be protected, but as something to be mined, just feels bad.
This is the trade we’re all being asked to make right now. There is a direct correlation between how much of yourself you’re willing to share with an AI system and how useful that system can be. Google is in such a strong position precisely because it already has all that information, while OpenAI, Anthropic, and the rest are desperately trying to figure out how to accumulate it. The AI tools we’re being promised are the ones that know us intimately, that can take action on our behalf, that can make decisions without even needing us around. None of that works unless we open ourselves up completely to the machine. So that’s what we’re being asked, even compelled, to do.
You know the phrase, “If you’re not paying for it, you’re the product”? AI takes that one step further. We actually are paying for it. And we — our correspondence, our photos, our very lives — are both the raw material and the end product, everything constantly mined and sorted and fed back to us in new ways. Some of them might be incredible; all of them will require this trade. I suspect I’m going to have a fabulous weekend in Hershey this summer, but I’ll never shake the feeling that I’m being watched. Supposedly for my own benefit.
I had a very different experience using Spark, Google’s new always-on AI agent. Spark is a hugely ambitious thing: Google intends it to be the interface through which you can use external apps, and over time even operate your computer. (“OpenClaw with better internet access” is a not-wrong way to describe it.) Spark is currently rolling out to Google’s $99 / month AI Ultra plan, but Google allowed me to try it early. I tested some simple action-oriented stuff, like having Spark go through my Gmail inbox and suggest a bunch of things I should unsubscribe from and having it comb my Google Docs for old tasks I still haven’t finished. In both cases, it did a fine job, even creating me a nicely organized document with a bunch of links to quickly unsubscribe from various marketing emails.
Then I gave Spark a simple trip-planning job. “I’m going to be in Hershey PA with my wife, two kids, and dog the weekend of July 18th. Can you make a plan for the whole weekend, including places to stay, eat, things to do, and everything else?” I left out a few salient details, like the concert tickets I have for that Saturday night, but figured I’d start with the six most obvious things to do in Hershey and go from there.
A few minutes later, Spark pinged me back. “I have created a comprehensive, family-friendly, and dog-friendly weekend itinerary for your trip to Hershey, PA, from Friday, July 17 to Sunday, July 19, 2026.” It shared a link to a Google Doc it had made me, and a couple thousand words of shockingly detailed, useful itinerary.
To begin with, it offered driving directions from my house, an address that of course Google knows but I had not offered. It included a few hotel options, including their pet fees, and some dog-friendly activities that Frida might like. I never told Google my dog’s name is Frida; my only guess is that Spark found it through emails from my vet.
Spark also casually noted that my son Lewis will get into Hershey Park for free, because he’s not a year old yet, but that because Arthur is three, he’ll need a ticket. I don’t know if Spark was guessing what time Lewis naps in the afternoon, or if it knew it somehow, but it was right to schedule nap time for 1:30PM.
The whole Spark itinerary was filled with details like this. It included my wife’s name, and took into consideration the fact that she doesn’t like to eat onions or scallions. It included the Thomas Rhett and Niall Horan concert on Saturday night, presumably based on the Ticketmaster confirmation in my email, and noted that parking is included in the tickets we bought. When I got to the part where it mentioned getting a babysitter that night, I remembered to note that my parents are coming along for just that purpose, so I added a note to the conversation.
“That is a wonderful update!” Spark replied, happily calling my parents by their names, and switching its recommendations from a hotel to an Airbnb. When I asked Spark to put all the information in a Google Doc and share it with Anna, it found my wife’s email, attached the document, drafted a note that sounded like we were business colleagues instead of a married couple, and sent it along.
The only time Spark failed me was when I asked it to book an Airbnb. It prompted me to allow Gemini to interact with websites on my behalf, navigated to Airbnb, and appeared to be promptly blocked. “Due to security and authentication policies on Airbnb, I am unable to log in, handle payment, or complete bookings directly on your behalf.” It instead offered up a few relevant places with availability on the right days, and reminded me of the information I’d need to book.
On the one hand, this is one of the most astonishingly impressive AI experiences I have ever had. Google’s AI prowess, combined with the vast quantity of data it has on me through Google’s Personal Intelligence feature, produced a personalized and useful itinerary that was well suited to my needs and my family. It put together the itinerary, and presented it to me, the way an actual human assistant would have — with lots of details specific to our situation, with the names of the people who matter, and with affordances made for all of our specific needs. Every time I read the itinerary I’m blown away by another detail of it; I suspect we’ll follow it almost exactly.
On the other hand, I can’t shake the deeply creepy feeling I get from the whole thing. What Spark did feels sort of magical, and very invasive. It’s weird that Spark is so casually telling me the names and ages of my children, reminding me that it knows where I live, and finding information I know for a fact I’ve never volunteered to Google. Intellectually, I know that Google knows an incredible amount about me — add up my emails, my calendar, my photos, and my search history, and you’ve pretty much got me pegged. But seeing Spark treat all that data not as something to be protected, but as something to be mined, just feels bad.
This is the trade we’re all being asked to make right now. There is a direct correlation between how much of yourself you’re willing to share with an AI system and how useful that system can be. Google is in such a strong position precisely because it already has all that information, while OpenAI, Anthropic, and the rest are desperately trying to figure out how to accumulate it. The AI tools we’re being promised are the ones that know us intimately, that can take action on our behalf, that can make decisions without even needing us around. None of that works unless we open ourselves up completely to the machine. So that’s what we’re being asked, even compelled, to do.
You know the phrase, “If you’re not paying for it, you’re the product”? AI takes that one step further. We actually are paying for it. And we — our correspondence, our photos, our very lives — are both the raw material and the end product, everything constantly mined and sorted and fed back to us in new ways. Some of them might be incredible; all of them will require this trade. I suspect I’m going to have a fabulous weekend in Hershey this summer, but I’ll never shake the feeling that I’m being watched. Supposedly for my own benefit.