I Tried Google’s New AI Health Coach, and It Left Me Utterly Baffled

We may earn a commission from links on this page.

Google has launched its “Personal Health Agent,” an AI coach available in the Fitbit app. Currently it’s in a “preview” mode, and limited to Android users in the U.S. who have a Fitbit Premium subscription. That group includes me, so I tried it, and it gave me some decent workouts. It also told me that the Pixel Watch 4, which Google makes, and which I reviewed, and which am currently wearing, does not exist. So, par for the course when it comes to AI.

How to enable the Fitbit app’s personal health coach

The “public preview” of this new coaching bot is available starting today for Fitbit Premium users in the U.S., provided they use Android. (Support for iOS is coming soon, Google says.) I’m a little unclear on what to call this bot—an email I got from Google calls it their “Personal Health Agent” and describes it as “Google Health’s AI coach.” A Google blog post calls it Fitbit’s “personal health coach.” In any case, it lives in the Fitbit app. 

When the AI coach became available for me, I received a message at the top of my Today screen asking if I’d like to “try new Fitbit features before they’re available to everyone.” If you missed that prompt, you can go to your profile pic at the top right corner of the app and select Public Preview from the menu that appears. 

Joining the public preview launches you into an entirely different version—dare I call it a beta?—of the Fitbit app. It doesn’t yet include menstrual health, mindfulness, nutrition, or community features, so to access those, you’ll need to switch back to the old version of the app. You can swap between versions at any time from the menu under your profile icon.

Setting up my fitness goals

Google says the new chatbot can answer general questions about health…but so can a web search, so I wasn’t too excited about that. What I did want to see was how well the bot could set up a coherent exercise plan for me—that’s the big feature Google is touting. It did pretty well, at first. 

The coach asked the same kinds of questions I would expect a personal trainer to cover when putting together a plan. It seemed to have a nicely structured approach, and gathered this information: 

  • My main goal (I told it I’d like to get back into a consistent habit after time off)

  • My biggest challenge (I said something about motivation and time)

  • How much exercise I was used to doing, including my running mileage and paces (it pulled this from my exercise data but let me make corrections)

  • What activities I like to do

  • When I like to do them (it noticed that my strength sessions tend to fall on Tuesdays and Thursdays)

  • What equipment I have available (it deduced I have space to run outdoors, and strength training equipment)

  • How many days each week I’d like to exercise

It responded well to my adjustments during the conversation. I told it I’d like to alternate strength training and running (with Saturdays off), starting today with a strength training session. It suggested a lower body focus to support my running, which I declined. I named some of my favorite lifts and asked if it could build the strength program around those. We were agreed—a six-day plan with strength and running was coming right up.

The bot told me it would take up to 10 minutes to generate my plan, but it only took about two. My workouts for the week matched what we’d discussed, with a few discrepancies. For example, I asked for pull-ups and it gave me assisted pull-ups. I also didn’t like the six-rep sets of squat and bench press, since I was hoping for heavier lifts with fewer reps. But there is an “adjust plan” button, and with some more back-and-forth, I was able to get it to tweak the workouts to my liking. 

It has trouble planning for the long term

I was excited to look over my plan—to me, a plan sets out the steps to accomplish a goal. For a training plan, that would involve building toward that goal over a matter of weeks or months. For example, a marathon training plan would increase your mileage over time until you can run a strong 26 miles. In my case, with a goal of consistency, almost anything would fit the bill. This is easy mode for a trainer, AI or otherwise.

But what I got in the app was not what I would call a plan. It was four workouts, taking me from today to my rest day on Saturday. There was no way to view next week, or the week after that, or to see how many weeks were even in this alleged plan. I didn’t even have a way to see the last two days of my six-day plan. 

I asked the bot what was coming up next, and it said it wasn’t able to tell me anything about next week. What about the end of this week? (We agreed on six days, after all.) It told me that the week is Tuesday through Saturday. I began to feel like those bodybuilders arguing over how many days are in a week. After some back-and-forth, it delivered me text descriptions of what Sunday and Monday’s workouts might look like, but they were incomplete, not even naming what lifts I’d be doing on the strength day. When I exited the conversation and looked at the workouts in the app, I only had the original four.

I tried asking another way, and the coach was able to give a broad overview of what the next few weeks might contain. Unfortunately the adjustments we’d previously discussed weren’t factored in, so it described how the second week would build on the originally programmed first week, not how it would build on the workouts that were actually on my calendar. If I were comparing this chatbot’s plan to something from, say, the Reddit fitness wiki, pretty much anything on the wiki would have been more comprehensive.

There’s no good way to follow the workouts

I’ve written before about the Pixel Watch’s barebones fitness tracking. (This applies to Fitbits like the Charge 6 as well.) You can turn on a strength training mode on the watch, but you can’t track rest times or note what exercises you did, though there is some ability to create and follow running workouts. 

With that in mind, I didn’t expect to be able to follow the strength workouts from my watch, but I figured it was worth asking. The bot told me to just track a basic strength workout from the watch (which records heart rate and total time, nothing else) and follow the exercises from my phone. Fair enough. 

But wait! The app just shows each exercise with a checkbox next to it. If you’re supposed to do three sets of six reps, you only get one checkbox, not three. And there’s no way to note how much weight you used so you can build on it next time. The bot told me we’d be doing some progressive overload, but how to progress if we’re not tracking how much weight I’m using? 

OK, maybe strength is hard for a simple bot to track, but running workouts should be straightforward, right? The old version of the Fitbit app (which you can still access if you quit the preview) could recommend personalized running workouts and load them onto your watch, so that the watch coaches you through the different paces and intervals. I tried one out when writing my Pixel Watch 4 review, so I know the device can do it. I was hoping for a similar experience here. 

But when I asked the bot about how to follow the running workouts, things got weird. It gave me step-by-step instructions to find the workouts on my Pixel watch, but the instructions were wrong. For example, it told me to swipe up to access the app list, but that’s not how you access the app list. And it told me that my workout should appear on a certain screen, but there were no workouts on that screen. 

I let the bot talk me through a troubleshooting process, which derailed when I mentioned that my device was a Pixel Watch 4. That watch doesn’t exist, it told me. There is only a Pixel Watch 1 and a Pixel Watch 2. 

What? The Pixel Watch 3 was released more than a year ago. The Pixel Watch 4 is the current model. I am wearing one right now. I asked the bot where it was getting its information about Pixel Watch models, and it responded by admitting to hallucinating the “nonexistent ‘Pixel Watch 4.’” Hmm.

The bottom line: Promising tech, if it ever works

As with many AI products these days, the best conclusion I can offer is that this would be a cool feature if it worked well—but it currently doesn’t. 

Here are a few things that it does handle competently at the moment: The onboarding conversation is well structured and gathers the right information (or at least it did for my fairly simple situation). The bot understood what I meant when I used lingo like “heavy singles with some back-offs.” It was able to pull data from my workout history, like my running mileage and the types of equipment I’m likely to have access to. 

But there’s so much it can’t do, including some really basic, fundamental things. It can’t plan for the long term, which is the whole point of a plan. It also can’t give me a way to follow the workouts it comes up with.

This brings me back to the question of why somebody would want to use this AI coach in the first place. Sure, it can come up with an idea for a workout, but so can anybody who has ever typed a query into a search engine. Finding simple workout ideas on the internet is like searching for grains of sand on a beach. Adding another to the pile isn’t innovative.

But if the AI could convert the workout it generated into a format I could follow with Google’s tech (be it their app or watch), that function would be useful, and it wouldn’t duplicate something I can find in a million other places. The ability to track your progress over time would also be useful, but that means the app would have to record your weights so it can actually program progressive overload, not just talk about progressive overload. Things like that are what a personal fitness coach really needs to provide, and this chatbot just isn’t right now.

Nine New AI Features Coming to Adobe’s Creative Apps

Adobe, the company behind big creative programs like Photoshop and Premiere, just wrapped up its 2025 Adobe Max keynote, and you know what that means. That’s right: more AI. Over the course of the three-hour presentation, the company went big on automating creativity, introducing new generative AI tools for Photoshop, Lightroom, Premiere Pro, and other Creative Cloud apps. Some of these are expansions of tools that already exist, like better generative fill, while other are all new—like Firefly’s new AI audio generation.

Adobe Express can design based on vibes

Adobe AI Express

Credit: Adobe

Before getting into the meatier stuff, let’s start with Adobe’s entry-level apps. While Adobe is known for professional-level programs like Photoshop, the company also has its own free basic web editor (although there’s also a mobile app) to help it compete with alternatives like Canva. Called Adobe Express, the tool’s been getting a steady stream of upgrades since its debut in 2015, and with the introduction of generative AI, has been quick to jump onto the trend to try to make itself easier to use.

Enter today’s “AI Assistant in Adobe Express.” When toggled on through a switch in the app’s top-left corner, the assistant will replace your tools with a chatbox where you can instruct it to either make a new design from scratch or edit an existing one. Should you need your tools again, you can bring them back by toggling the assistant off, although Adobe’s demos for the feature also show the assistant bringing up contextual sliders when needed, like one for resizing.

While this is not Adobe Express’ first venture into generative AI, the idea is to make getting started or quickly editing a piece less intimidating, by having inexperienced users spend most of their time in a chatbox rather than having to click through a toolbar. Adobe says, like its other AI tools, it pulls from a number of “commercially safe” sources, including the company’s font and stock image libraries and its Firefly AI models.

The tool will start rolling out in public beta today, so you should be able to try it out shortly.

Adobe Premiere is getting built into YouTube Shorts

Premiere Pro for YouTube Shorts

Credit: Adobe

Shorts are the next big thing over on YouTube, and to encourage more people to make them, YouTube is teaming up with Adobe. As an update to both the Premiere iPhone app and YouTube itself, Adobe’s new Create for YouTube Shorts feature allows you to upload your footage and instantly make it publish-ready with Adobe’s font overlays and a number of “exclusive” effects, transitions, and stickers. Or you can directly plug your footage into templates that already have transitions and effects included.

The feature is currently listed as “coming soon,” so it’ll be a bit before you can try it. But once it’s live, Adobe and YouTube say you’ll be able to access it either through the Premiere iPhone app or directly though YouTube, via an “Edit in Adobe Premiere” icon in YouTube shorts.

There is no word yet on an Android or desktop release.

Adobe will add sound to your videos for you

Firefly Audio AI

Credit: Adobe

Sound is easy to overlook when making a new video, and I’ve had to scramble to find a decent soundtrack to add to my videos at the last second more than once. Adobe’s new Firefly AI audio features are looking to save you from that fate, by making it easy to add music and even narration to an otherwise silent video.

Rolling out in public beta today, Firefly’s new “Generate Soundtrack” and “Generate Speech” buttons use AI and a Mad Libs style prompting system to help you quickly score your content from a number of options.

For “Generate Soundtrack,” you’ll upload your video, press the appropriate button, and the app will suggest a prompt for you and give you a palette of adjectives, genre types, and and content types to refine it with. Drag your chosen terms into the prompt box, hit generate, and you’ll get four options, each cutting out at a maximum of five minutes.

It’s a bit odd that you can’t just enter your own terms into the prompt box, although Adobe generative AI head Alexandru Costin told The Verge that’s because AI audio is “a new muscle we need to develop” and that the current approach is “easier and more accessible.”

Like other Firefly generations, audio will be generated using Adobe’s own licensed content, so users won’t have to worry about copyright strikes on videos made using the feature.

“Generate Speech,” meanwhile, gives users access to 50+ text-to-speech voices, either from Adobe Firefly or licensed via ElevenLabs. There’s no Mad Libs prompting here, with Adobe instead allowing for fine-tune control over factors like speed, pitch, tone, and even pronunciation. Currently, over 20 languages are supported.

Taken together, the updates seem to me like an attempt to keep up with platforms like Instagram and TikTok, which have licensed music libraries and text-to-speech built in. Whether a purely AI-powered version can keep up remains to be seen, although putting it into the editor rather than the platform does give creators more choice about where to upload.

Updates inside Photoshop, Lightroom, and Premiere

Adobe Creative Cloud promo image

Credit: Adobe

Finally, for the more hardcore Adobe users, updates are coming to the company’s core apps as well.

First, Photoshop is also getting its own AI assistant, which will be able to use prompts to edit for you. However, unlike Adobe Express, it’s currently in a private beta, so it’ll be some time before most users see it. It’s also limited to the web version of the app for now.

However, not in beta is the ability to choose which AI models the app works with. Previously, Generative Fill, which uses AI to fill in blank spots in backgrounds (or just generate whole canvases from nothing), were limited to Adobe’s Firefly models. Now, users will also be able to use them with Google’s Gemini 2.5 Flash model, or Black Forest’s Flux.1 Kontext model. Given how popular 2.5 Flash has gotten on social media under the name “nano banana,” that’s a big get for Adobe.

Still, Firefly isn’t getting left behind. Adobe says it’s upgraded the model with the ability to generate in a native four-megapixel resolution, and to better render people. It’s also integrating it into a new “Layered Image Editing” tool that can make contextual changes for you across layers, like futzing with shadows after you move an image.

Outside of Photoshop, Lightroom has its own private beta feature called “Assisted Culling.” I’ll admit Lightroom is probably where I have the least experience when it comes to Adobe, but the company says it’ll be able to filter through uploaded photos for you and find the most edit-friendly ones.

Finally, Premiere Pro has its own beta feature, but one that’s graciously public. Called “AI Object Mask,” it’ll automatically detect and track people and objects in your video’s background, so you can more easily add effects like blurs or color grading. It could be useful if, say, you’re shooting in a crowded area where you need to blur a lot of faces.

A little something for everyone

Overall, it was a fairly balanced Max, with a number of features for both pros and beginners. That said, I can’t ignore the focus on AI and automatic generation. On one hand, I get that photoshop’s a bit intimidating. On the other, the more Adobe handles your edits for you, the more it runs the risk of competing with existing easy-edit apps and platforms. I’m curious to see how the industry giant will compete as platforms like TikTok and Instagram continue to offer their own built-in editing tools.

These Budget-Friendly Beats Flex Earbuds Are 57% Off Right Now

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

While wireless earbuds can be convenient and allow for greater freedom of movement, Bluetooth earbuds connected by a wire are harder to misplace and are significantly more affordable. The Beats Flex earbuds are Apple’s budget-friendly wired neckband earphones, and right now, they’re just $30.01 (originally $69.95) in black at Walmart—about 57% off the usual price.

If you’re prone to losing your wireless earbuds, the Flex’s earpieces magnetically snap together when you’re not wearing them, allowing you to keep them securely around your neck. When they snap together, audio automatically pauses, saving your battery life. The tangle-free neckband design houses a charging port, microphones, and volume controls.

According to PCMag’s review, which rates them as “Excellent,” they feature a secure in-ear fit with a design that eliminates cable tug and provide balanced sound with powerful bass with 8.2mm drivers. Battery life is around 12 hours on a full charge, and 10 minutes of “Fast Fuel” charging will give you about 90 minutes of listening time.

The earbuds use Class 1 Bluetooth for extended range and Apple’s W1 chip for seamless pairing. Two pairs of Flex can share the same sound source; when a second pair is nearby, you’ll get an onscreen prompt to connect. They’re compatible with Bluetooth 5.0 and support AAC and SBC Bluetooth codecs, but not AptX. Charging is via USB-C. They do lack an official IP rating, so you’ll want to be mindful when wearing them in heavy rain. 

If you don’t need active noise cancellation, don’t need water resistance, and are fine with USB-C charging, the Beats Flex earbuds are an affordable option with a bass-heavy sound and Apple’s W1 chip, making them a particularly good option for budget-conscious iOS users.

Why Now Is the Best Time to Buy a Peloton With a Referral Code

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

If you’ve been considering getting one of Peloton’s new Bikes, Treads, or Rows, this might be the time—as long as you know someone who is already an All-Access Member. Until Nov. 11, referral codes from existing members get you more money off your hardware purchases.

The discount details

Last month, Peloton rolled out upgraded versions of its fleet, including the Bike, Bike+, Tread+, Tread+, and Row. The new devices are known as the Cross Training series and, as such, are named the Cross Training Bike, Crosstraining Tread+, etc. If you want to know how they differ from the old models, I wrote an explainer here, but the basic gist is that the proprietary touchscreens on which you follow guided classes now swivel around, making it easier to hop off your machine and do, say, a strength or yoga class. The Plus line also comes with a movement-tracking camera that uses AI to correct your form, suggest when it’s time to add more weights or reps, and monitor your workouts.

As you can assume, these are pricier than their predecessors, which is why the current referral promotion might be of interest to you. If you get a referral code from an existing All-Access Member, here’s what you can get off at checkout:

  • $300 (instead of the usual $250) off your cart when you purchase a new Peloton Cross Training Bike

  • $700 off your cart (instead of $500) when you purchase a new Peloton Cross Training Bike+

  • $500 off (instead of $300) when you get a Peloton Cross Training Tread

  • $1,200 off (instead of $600) a purchase of a new Peloton Cross Training Tread+

  • $200 off a new Peloton Cross Training Row+

All the discounts will return to normal after Nov. 11. (Note: This probably isn’t the best time to get a Cross Training Row+, as the typical referral discount on that is $300.)

Your referrer has to be an All Access Member, not someone who uses one of the less-expensive, app-only versions of Peloton’s digital offerings. The referrer can’t be renting their machine, their membership can’t be paused, and they can’t be a Peloton employee.

Notice, too, that this specifically applies to new devices. That means you can’t use the referral code to get money off a purchase on Peloton’s resale platform, Repowered, although that’s always a solid option if you’re looking for equipment on a budget. You don’t have to have the Cross Training series to get a good workout (although I enjoyed my demos of those devices); I have been using the same Peloton Bike for nearly five years and love it—I ride it every day.

You can redeem your code on the Peloton website or at Peloton stores. Bear in mind, however, you have to be a new customer.

How to find a referral code

If you want to be the referrer, not the referee, here’s how to generate the code that will get you some brownie points with your fitness-focused friends. The easiest way is to go to this link and log into your Peloton account.

Otherwise, head to members.onepeloton.com and select the hamburger menu in the top right corner, then hit Refer friends. If you’re not seeing that or the link above didn’t work, you’re not eligible to share a code. You can also do this on your mobile app by tapping the Community tab, then Add Friends on the upper right, then Refer Friends.

Remember how I said you’ll get brownie points with the people you refer? You get something else, too: You can give out up to $1,200 in referrals, but you also get a free month of your All Access Membership for every person you refer. Once someone makes a purchase with your code, you’ll get en email containing a code of your own, plus instructions on how to redeem it, after three to five business days. The voucher codes are valid for three months after their issue date.

This 2-in-1 Microsoft Surface Pro 7 Is on Sale for Just $660 Right Now

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

This brand-new Microsoft Surface Pro 7 offers portability, flexibility, and power that still holds up for most people who want a dependable Windows machine without overpaying, and it’s on sale for $659.99 at StackSocial right now. The 12.3-inch PixelSense touchscreen remains sharp and responsive, with a 2736×1824 resolution that makes documents, spreadsheets, and streaming look crisp. It’s lightweight at 1.7 pounds, and the built-in kickstand turns it from tablet to laptop on demand—something the Surface line has always nailed.

The Surface Pro 7 is powered by a 10th Gen Intel Core i5 processor, paired with 8GB of RAM and a 256GB SSD. That combination is no longer cutting-edge, but it’s still sufficient for everyday tasks like writing, browsing, video calls, and light creative projects. The battery is rated for up to 10.5 hours, which means you’ll likely get through a workday before needing to recharge. You also get a mix of ports that newer devices sometimes skip: both USB-A and USB-C, a headphone jack, and even a microSDXC card reader for quick storage expansion. Wifi 6 and Bluetooth 5.0 ensure modern connections, and for those invested in the Surface ecosystem, it still works with accessories like the Type Cover keyboard and Surface Pen, though they’re sold separately.

This model won’t compete with the latest Surface Pro 11 or high-end ultrabooks if you’re looking for raw speed, and the cameras top out at 1080p video when many laptops now ship with 1080p or even 4K webcams. It also doesn’t come with a warranty, so the price reflects a final-sale deal. Still, for someone who needs a 2-in-1 that balances tablet portability with laptop functionality, the Surface Pro 7 remains a solid choice. It’s a way to get a brand-new Microsoft device with decent specs at a reasonable cost, if you’re willing to accept that it’s no longer the newest kid on the block.

Here’s Why I Won’t Install ChatGPT Atlas on My Mac

OpenAI has officially entered the browser wars. Last week, the company announced ChatGPT Atlas, a new web browser with ChatGPT integration. At the moment, it’s Mac-only, but I wouldn’t recommend even my Apple friends jumping on board immediately—at least not without understanding the underlying risks.

Atlas’ AI web browsing

If you’ve already used other AI browsers, like Perplexity’s Comet, Atlas is going to feel familiar. In fact, that’s also likely true if you’ve used any web browser before: Atlas is built on Chromium, the engine that powers browsers like Google Chrome, Microsoft Edge, and Opera. That means the core mechanics of Atlas are fairly standard; there’s nothing particularly revolutionary happening here when it comes to sorting tabs or the browsing experience itself.

The same is true when it comes to some ChatGPT interactions. As with other AI browsers, ChatGPT is assigned to the sidebar of the browser window. You can call it up by clicking on the “Ask ChatGPT” button, where you can ask it questions about the content you’re currently browsing. You can also ask ChatGPT for writing assistance any time you enter an open text field in the browser.

Like Comet, Atlas has an agent mode, but the latter’s is built off the existing ChatGPT agent. The idea is that you can task Atlas with performing functions on your behalf. So, rather than pulling up DoorDash’s website and ordering yourself dinner, you could ask Atlas to order dinner for you. You can even watch Atlas get to work, and see its thinking behind each decision. OpenAI has other ideas for how to use Atlas’ agent mode, including giving the browser a recipe to shop for, or asking the bot to run through team documents at work to generate a brief.

Deeper ChatGPT integration is what might set Atlas apart from the competition. If you’re a regular ChatGPT user, you’ll probably appreciate it having that contextual awareness of your past conversations—if you’ve already asked ChatGPT about a topic, and you’re currently researching it in a browser window, you can pick up the conversation and assume ChatGPT will remember what you’ve already talked about.

chatgpt atlas memory

Credit: OpenAi

Similarly, Atlas will track your browsing and activity history and call upon it in future sessions. Perhaps you’ll open your browser to find personalized suggestions on which sites and topics to explore next. Does that sound creepy? Absolutely. But if you’re someone who doesn’t mind the privacy trade-off, there could be some benefits there. In OpenAI’s announcement, it suggested asking Atlas to pull up all the job postings you looked at last week, and produce a summary of industry trends to use in interview prep. If you find that these memories are a bit too much for you, you can disable them from the browser settings. (OpenAI says deleting your browsing history also deletes associated browser memories, and the browser’s incognito window logs you out of ChatGPT.)

The company includes a setting called “ChatGPT page visibility,” which lets you control whether ChatGPT can actually see the webpage you’re visiting. If you choose “Not Allowed,” you can block the bot from seeing what you’re doing, which is good. But then again, doing that defeats the purpose of Atlas a bit. If you don’t want ChatGPT seeing what you’re doing, you might as well use a browser that doesn’t have ChatGPT built right into it. (The company does promise it won’t train ChatGPT on your browsing data unless you opt into it, but why would you do that?)

Is Atlas safe to use?

chatgpt visibility

Credit: OpenAI

I’m of the opinion that if the safety of a browser is in question, it’s best not to bother with. That’s the case with Atlas, as well as other AI browsers, which is why I’m making it a bit of a personal rule from here on out to not download or use them on my personal computers.

The main issue with browsers that incorporate AI agents is that they are susceptible to indirect prompt injection attacks. Brave has done a lot of research on this subject, particularly with Comet. In short, bad actors can potentially hide malicious instructions on websites that the AI agents see as no different from a typical user request. Because the browser is designed to act on your behalf, these malicious instructions can command the AI to do things you definitely don’t want it to do. You might ask Atlas to summarize a webpage, but because a bad actor hid a command to do something involving your email, bank account, or corporate intranet on the site, it does that instead.

To OpenAI’s credit, the company has compiled a list of safeguards to mitigate risks with Atlas. Atlas cannot run code directly in the browser, nor can it download files or install extensions. The browser has no way to access other apps on your Mac, nor its file system. If agent mode needs to access sensitive sites, like your bank, it’ll pause to make sure “you’re watching.” To that point, you can use Atlas’ agent in logged out mode, which limits its ability to access sensitive data or take actions “as you” on websites. But even OpenAI admits that after thousands of hours of testing, their safeguards “will not stop every attack that emerges as AI agents grow in popularity.” The company says it’ll patch new vulnerabilities as it finds them, but if bad actors find them first, they might trick these AI agents into doing some terrible things.

That assertion has already been proven right. Within Atlas’ first few days available to the public, one researcher discovered the browser is susceptible to clipboard injection attacks, in which bad actors can lace elements of their website with code that would add malicious links to your clipboard when clicked. Since Atlas can act on your behalf, you might not even realize if it clicks on such a button, and, as such, you may accidentally paste the malicious link into your browser yourself.

To me, the risks currently far outweigh the benefits. I don’t yet see much reason to have a bot in my browser do things on my behalf, but even if I did, I wouldn’t use it just yet. The risk that someone injects a website with a malicious command and derails my AI agent—and my digital life—is too great, especially when I can book those flights or order that delivery on my own without issue.

This Nest Thermostat and Temperature Sensor Bundle Is $70 Off

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

Just in time for winter, a bundle including the fourth-generation Google Nest Learning Thermostat and a second-generation Nest temperature sensor is $70 off at Walmart. That brings the price down to $209.55 (originally $279.99). The smart thermostat allows you to customize your home’s heating and cooling schedule with AI, conserving energy and ensuring consistent temperatures throughout your entire home.

This fourth-generation Nest thermostat offers several updates from its predecessor, including support for Matter, which allows compatibility with a wide range of smart home ecosystems (though it doesn’t support Thread). According to this PCMag review, it’s simple to install, comes with a remote room sensor, and uses AI to learn preferences and create a customized heating and cooling schedule that adjusts as your routine changes—though you can also manually make changes and reject suggested tweaks.

It has a larger display than the previous generation (2.7 inches with a 600×600 pixel resolution, compared to 2.1 inches with a 480×480 pixel resolution) with a thin stainless steel ring that allows you to scroll through menu options and adjust temperature settings or switch between modes like Comfort, Eco, and Sleep. The thermostat works with most 24V heating and cooling systems and includes ambient light, humidity, motion, and temperature sensors. You can control it via the Google Home app or Google Assistant voice commands.

This model also has a new version of Nest’s Farsight technology, which uses a motion sensor to turn on the display when you enter the room and adapts what’s shown based on how far away you are from the thermostat, showing more details as you get closer. Unlike more feature-rich options like the Ecobee Smart Thermostat Premium, it can’t detect air quality or double as a home security hub when paired with sensors and cameras. 

Ultimately, this bundle—currently around 25% off—is worth it if you’re seeking a smart thermostat with a large display, a stylish aesthetic, and remote room sensing. If you value design and simplicity over maximum features, it’s a better choice than the Ecobee Smart Thermostat Premium.

Threads Now Has Self-Deleting ‘Ghost Posts’

If you’ve watched enough Mission Impossible (or Inspector Gadget, if your parents wouldn’t let you watch Mission Impossible), you’re familiar with the self-destructing message. Now, Meta’s X competitor, Threads, is making them a core feature.

Called “Ghost posts,” the new type of message launches today, and it lets you post to your timeline without the pressure of leaving a permanent trace. Ghost posts will appear on your timeline while active, but after 24 hours, they’ll automatically be archived.

“Here today and gone tomorrow,” as Threads puts it, the idea is to encourage posters to share “spontaneous thoughts” and “fresh takes.”

It’s not the first time a social media platform has experimented with ephemeral content—Snapchat is probably the most famous example—but it makes sense. People change, and the person I was 10 years ago may not look like who I am today. Even if you don’t have anything cancel-able on your timeline, it can be embarrassing to see old posts. And if you still stand by everything you’ve ever posted, it’s still a good idea to clear your timeline before leaving a platform behind.

That’s why sites like X already have a number of third-party programs that will delete your old posts for you, but Threads is the first platform of its type to offer the feature officially. Here’s how it works:

How to use Ghost Posts on Threads

Ghost Posts on Threads

Credit: Michelle Ehrhardt

Threads’ Ghost posts are already live, so you can try them out right away. The only catch is they don’t yet seem to be available for desktop (at least based on my testing), so you’ll need to use the Threads app to get started, at least for now.

To write a ghost post, open the Threads app and find the “What’s new?” box under your username (it’s usually right up at the top of the page). Hit the icon that looks like a smiley face inside a dotted-line chat bubble to the right of it.

From there, write your post as you normally would. When you send it, it’ll appear in your followers’ feeds in a chat bubble surrounded by dotted-lines, and on your profile in a special Ghost posts subsection.

Aside from being temporary, ghost posts also hide their metrics and replies. Only you will be able to see who’s interacted with your post, and any replies are sent to your messaging inbox, rather than directly below the post. Everyone else will just see smiley faces next to the post’s metrics, indicating that at least one person has interacted with it.

And that’s it. You’ve successfully haunted your friends’ feeds.

What happens to a Ghost Post after 24 hours?

While Ghost Posts are mostly set-it and forget it, you can technically still interact with them after 24 hours. They’ll be archived, which means others won’t be able to see them, but if you ever want to scroll through your old thoughts, it’s easy enough to look at your archived posts. Just go to your profile page, hit the two-lined button in the top right corner, and tap Archive. Also note that while normal archived posts can be made live again, that’s not the case for Ghost posts.

Similarly, to ensure that Ghost posts do not remain live on other servers once they’re archived, they are unable to be shared to the fediverse, Meta told Lifehacker via email.

Update 10/27 at 5:57 PM E.T.: Added clarification from Meta that Ghost posts cannot be shared to the fediverse.

This LG Gram 14 Laptop Features an Ultra-Light Design and Cloud OS Readiness for Under $400 Right Now

We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.

The LG Gram line has built a reputation for being thin, light, and travel-friendly, and this 2022 14-inch model is currently on sale for just $389.99 on StackSocial. It’s an 11th Gen Intel Core i3 laptop with 8GB of RAM and a 256GB NVMe SSD. On paper, that’s a respectable setup for browsing, multitasking, or office apps, and the 14-inch WUXGA display adds extra vertical space compared to standard 1080p panels. What makes this version different is that it ships without a traditional operating system. It’s designed as a “thin client,” ready to run Cloud OS platforms like IGEL or Windows 10 IoT Enterprise, making it more appealing to business users who rely on virtual desktops than someone looking for a plug-and-play Windows laptop.

The laptop weighs just 2.2 pounds and measures 0.66 inches thick, so tossing it into a backpack for commuting or travel won’t feel like a burden. Additionally, its battery life is rated for up to 12 hours, which means a full day of work can be completed before needing to recharge. As for connectivity, Thunderbolt 4 support opens the door to high-speed connections and external displays, while HDMI, USB-A, and a microSD slot cover most everyday accessories. Rounding it out are the backlit keyboard and fingerprint sensor on the power button, which add a touch of convenience.

That said, there are some important trade-offs to weigh. Because it doesn’t include Windows or macOS, setting it up requires some technical expertise and a clear understanding of the environment you plan to use. The Core i3 processor is suitable for everyday tasks, but it’s not intended for heavy creative work or gaming. And while the unit is brand new in original packaging, it doesn’t come with a product warranty. Still, if you’re in the market for a lightweight, cloud-ready laptop built for business use, the LG Gram 14 delivers solid specs, a sharp 16:10 display, and excellent portability at a reasonable price.

The Out-of-Touch Adults’ Guide to Kid Culture: What is Group 7?

We may earn a commission from links on this page.

It’s so-called spooky season, and there’s nothing more frightening than realizing how out of touch you’ve become, so allow me to throw you a lifeline with an explainer on the week’s trending topics with the youths. This week, everyone is talking about their membership in a completely arbitrary online group, planning Halloween costumes that few people over 30 will understand, and trying on “historically accurate” clothing from 2005.

What is “Group 7” on TikTok?

Everyone on TikTok this week is posting about “Group 7.” Here’s what’s up: On October 17, 26-year-old singer Sophia James did a little social experiment to promote her new single, “So Unfair.” She posted seven videos featuring the song to see which would gain the most traction and reach. The seventh went viral, to the tune of over 75 million plays. And then it took on a life of its own, with over 600,000 videos ultimately being posted under the Group 7 tag.

In the original G7 video, James says, “I don’t know what that says about you. But, you’re in Group 7.” People started defining what membership in Group 7 means to them—even though it’s random—and listing the qualities that Group 7 members supposedly possess. According to this video, members of Group 7 are loving, “we laugh, we share, we watch videos all the way through.”

While many TikTokers are taking their membership in Group 7 as a joke, a growing (and sort of sad) part of Group 7 seems convinced that it’s important. That they’ve been chosen. The longing for a sense of belonging is palpable. There’s also the idea that if all the members of the group follow each other, it would mean “never having to get a normal job.

But it won’t. “Follow-chains” are not new on the internet, and they don’t work. What’s really going to happen is that the algorithm will eventually decide people are sick of their feeds being full of Group 7 videos and curb their spread. In a few weeks, everyone will say, “Remember that weird ‘Group 7’ thing?” This is the inevitable life-cycle of internet memes.

HBO Max adds “six seven” category

Speaking of arbitrary number-based memes, brain-rot slang word of the year “6-7” continues to grow and spread in unexpected ways, including on streaming platform HBO Max, which has added a “Six Seveeeeen” movie category on its service.

The phrase doesn’t mean anything, making a “6-7” film difficult to define, so Max decided to take the easy route by including movies with “six” or “seven” in their titles. Films in the collection include expected entries like The Sixth Sense, Seven, and the sixth and seventh Harry Potter movies (Six Days Seven Nights unfortunately isn’t available on HBO Max, and also, you probably forgot it existed anyway). Low-hanging fruit aside, there are also some inspired choices, like Akira Kurosawa’s The Seven Samurai and Ingmar Bergman’s The Seventh Seal. Like all “how do you do, fellow kids” attempts at co-opting youth slang, HBO’s stunt is annoying and tone-deaf, and maybe a signal that the whole six-seven thing is ending at last, but I do love the idea that some brain-rotted kid out there is going to watch The Seventh Seal and really relate to the soldier’s chess match against Death.

Meme-based costumes to watch for on Halloween

Even before the internet, Halloween was a meme holiday, with costumes of pop culture figures topping the sales charts at Spencer’s Gifts every year. But it’s gotten ridiculous in 2025. Look for every other kid ringing your doorbell this year to be wearing an outfit inspired by a popular online trend and for ghosts and Draculas to be as rare as Loch Ness monster sightings.

Here are some of costumes-based-on-memes that are likely to be popular this Halloween:

Labubu

Dressing up as one of these trendy toys is a no-brainer, although the effort involved in doing it right is considerable. Luckily, there are tons of tutorials online like the one linked above.

Lorax girl

If you’re wondering why so many young women are dressing up as the main character from 2012 movie The Lorax, it’s a whole thing. The costume has grown so popular over the last few years that there’s a legit backlash against it.

KPop Demon Hunters

The flashy outfits from Netflix’s amazingly successful animated movie are a perfect fit for Halloween, but are likely to result in the most “what are you supposed to be” responses from older people.

Chicken Jockey

Like the game that inspired it, The Minecraft Movie remains very popular with young people, so look for representations of the iconic “Chicken Jockey!” scene in costume form on Oct. 31. If you want to do it right, you have to make it yourself, but you can buy a Minecraft Chicken Jockey outfit from Walmart if you want to save time.

Inflatable protest frog

Portland’s Ice protests have given rise to an underground, viral hero: protest frog, so look for imitators rocking the iconic costume on Halloween. It’s a good choice for the lazy in that it makes a statement without any effort. You can just buy an inflatable frog costume on Amazon.

Viral video of the week: “trying on low rise jeans”

Warning: this week’s viral video will probably make you feel ancient. In it, young TikTok user @vintageprincessxoxo tries on “historically accurate low-rise jeans.” She’s clearly surprised by how low low-rise jeans actually are, but as many older folks have pointed out in the comments, the “historically accurate” pair she’s trying on aren’t even all that low. “Darling those are high rise with the double buttons,” a user points out. The comment section is filled with similar cross-generational educational outreach, with millennial mentors cluing the youngster in about whale-tails and the importance of wearing a belt. (“Historically accurate,” though? Ouch.)