Connexion
ChatGPT's capabilities continue to expand apace, and the latest feature is a fully capable slideshow creator. Announced a few weeks ago, ChatGPT Slides lets you create PowerPoint-style decks from inside your conversations, and it's now appearing for those on the Pro tier and above (meaning right now, you need to spend at least $100 a month to get it).
If you're on the free ChatGPT plan or the $20-a-month Plus tier, you can still get ChatGPT to generate downloadable PowerPoint files from the content of your conversations. However, they'll be more basic than those made in the full version of ChatGPT Slides, and there are no collaboration or native edit options.
ChatGPT Slides is part of OpenAI's push to give users a complete office suite inside the ChatGPT apps. Pages (the word processor) is already available for Pro tiers and above, while a spreadsheet app is on the way—the idea being that you will soon be able to get all your work done without any help from Google or Microsoft.
While Chatgpt Slides is only available on pricey plans right now, it's likely to come to lower subscription tiers in the future. But for now, how does the tool compare to what's available in the free and Plus tiers, not to mention what you can do with Gemini inside Google Slides? I've taken a detailed look.
When it comes to testing out AI tools, it helps to focus on something you're already knowledgeable about so you can spot any mistakes. To test this, I decided to see what ChatGPT could hoover up from the web about the new iPhone Duo. If you're on the free or Plus plans, you can start a conversation, then ask for a slideshow or PowerPoint about it.
Plus accounts also have access to a specific Presentations add-on, available via the + (plus) menu in the prompt box, though I get the impression this feature is in a state of flux since ChatGPT Slides arrived. In any case, you can begin asking about the iPhone Duo specs and early online reactions, and then say "build me a slideshow on the key features and reviews so far".
What I got back on the $20 Plus account was a fairly competent slideshow on Apple's foldable, with accurate information as far as I could see, and a simple series of layouts. It looked like something you might whip up from a template if you were in a rush and didn't have time to properly align all the bullet points or be very selective about which details you included and which you left out. One neat touch is that ChatGPT was able to drop in an official photo of the iPhone Duo at the start of the slideshow, but I had to specifically ask for one.
You can go back and make specific edits to the slideshow, but you have to describe them in the general prompt box—you can't select individual slide elements and make more precise changes. Still, if all you want to do is occasionally convert your ChatGPT conversations into a rough and ready slideshow, there's no real need to upgrade from the free or Plus accounts for now.
Switching to a Pro account, I gave a similar task to ChatGPT Slides hoping for something more comprehensive. To begin with, Slides gets its own dedicated area in the app, so you can start up a blank presentation from scratch—there are some basic image and text features, but the idea is to use prompts to build your slide deck. It's like a really basic Google Slides, with a prompt box.
Creating a slideshow this way feels rather tedious: Typing out what you want to say and then waiting for ChatGPT to do it isn't as intuitive as just throwing in some graphics and text yourself, as you might do in PowerPoint. It also seems to be a bit buggy, at least during the time I spent testing it, with some requests not properly completing.
Starting a standard conversation with ChatGPT and then turning all the information into a deck is much more useful from my perspective. My iPhone Duo slideshow was longer, laid out better, and more informative than the one I created on the Plus plan—there were even charts of battery life and official photos this time.
However, whether I used the free, Plus, or Pro tools, my presentations didn't come across as particularly well thought through. They don't really flow smoothly, either across individual slides or the slideshow as a whole—it feels very much like an algorithm, rather than a person, has decided what information to include and where on the slides to stick it.
Whether you create presentations in chats or through the ChatGPT Slides space, you're able to click on individual elements and do some basic manual resizing and repositioning. You can also request edits via prompts: "Drop in an image of a cat", "make this pop more", "center all of this information", or whatever it is.
ChatGPT Slides does offer collaborative tools, which I wasn't able to test, and I can see that being useful for teams heavily invested in ChatGPT's ecosystem. I can also imagine ChatGPT Slides coming in handy when you've got a lot of information you need to quickly dump into a slideshow that looks vaguely presentable.
In general though, I think I'd much rather stick to the traditional way of making slideshows for now. These old methods are more precise and more manageable, and give you full control over what's shown and where—even if they do take a bit longer. What's more, modern-day presentation tools come with so many templates and other layout aids that it's not difficult to put together something that's easy on the eyes while including all of the information you think is most critical.
Google is, of course, also all the way in on AI (and then some), and if you pay for Google AI Pro ($19.99 a month) or above, you can have Gemini pop up right inside Google Slides to give you some assistance with your presentation-making. To compare the tools, I gave it the same task as I did ChatGPT, asking for a slide deck on the new iPhone Duo.
Gemini actually asked some questions along the way: about wht visual style to use, the audience to aim the slideshow at, and the details to highlight. Though it's always possible to put this sort of information manually into a prompt, this would be a helpful additional to ChatGPT Slides.
What Gemini and Google Slides produced was impressive in terms of the layout and overall vibe of the presentation: It felt much more like a human being had put together the slide deck compared to what ChatGPT had produced, in terms of the information chosen, the way the slides were laid out, and the color picks.
However, Google's result had its faults too: The presentation got the facts on the maximum brightness of the iPhone Duo screen wrong, and produced an "iPhone Duo" graphic that looked like two iPhones from around 2021 stuck together. ChatGPT Slides at least managed to pull official imagery from the Apple website.
You can't attach individual edit prompts to specific parts of your presentation in Google Slides as you can in ChatGPT Slides—you have to mention them in the general prompt box. That said, Google Slides is obviously a much more mature and established slideshow tool compared to the OpenAI equivalent, in terms of its features and layout options (it has been around for a decade, after all).
Given that extra software structure and the lower cost of entry, Gemini and Google Slides is the better option for AI-assisted slideshow making at the moment—though I wouldn't bet against OpenAI catching up. For now, ChatGPT Slides is something that's good to have available if you're already paying a lot for ChatGPT every month, but not a feature that makes it worth your while to swap your AI allegiance.
We may earn a commission from links on this page.
You Can See Everything, the new documentary about Theranos founder (and accomplished con artist) Elizabeth Holmes from filmmakers Nathan Fielder and Lance Oppenheim, is breaking records (and brains). Not only is it raking it in at the box office in a way that few documentaries can manage, it’s also dominating the pop culture chat right now thanks to genre-defying techniques; a huge, rule-breaking twist (that I won't spoil); and its unrepentant and inexplicably mesmerizing subject.
While some might argue that you should head into this incredible film experience fresh, if you’re unfamiliar with any of the three major figures involved, you will benefit from a little prep work before you head in—and if you've already seen it, you're no doubt eager to dive even deeper into this truly remarkable (in every sense of the word) story. Here’s what you should watch, read, and listen to before or after You Can See Everything.
To put it simply: There’s a lot going on here. To ensure you can appreciate the film, you should make sure you know at least the basics about three things: Elizabeth Holmes (and why she’s infamous), and the work of both Nathan Fielder and Lance Oppenheim. Here’s the crash course.
The 2019 podcast The Dropout is one of the most efficient ways to get the background on Holmes and Theranos. In the style of a true crime podcast, this show from ABC and journalist Rebecca Jarvis offers a deep dive into Holmes and her crimes that will give you all the background you’ll need.
John Carreyrou is the journalist who broke the Holmes story wide open, and his award-winning book Bad Blood offers a more journalistic and less sensational investigation than other accounts, providing a close-up look at Holmes in her CEO persona—an alternately terrifying and charismatic person who comes across as a “character” Holmes is playing. The book will inform your viewing of the documentary in important ways.
While Nathan Fielder is getting most of the attention for You Can See Everything, co-director Lance Oppenheim brings his signature technique to the film—it wouldn’t be nearly as interesting without his visual sensibility and framing.
Oppenheim has helmed several notable films, as he’s made a name for himself as a brilliant documentarian over the last few years, but to get ready for You Can See Everything, check out Spermworld, his 2024 exploration of unregulated sperm donations (you read that right). The doc deals with people who are as reality-challenged, manipulative, and possibly delusional as Holmes, and it will prep you for his visual style of extreme close-ups and excessive coverage that seems almost impossible in a normal documentary setting. You can stream Spermworld on Hulu.
If you’re unfamiliar with Nathan Fielder’s work, you need to strap in, because he’s unlike any other comedian or documentarian you’ve ever encountered.
To get a sense of his style, start with Nathan for You, a comedy-documentary-reality TV hybrid in which Fielder—as a version of himself—tries to use his business experience to help struggling people and small businesses. Presented as totally sincere, Fielder’s ideas are nevertheless increasingly elaborate and, in a word, unhinged. This show will familiarize you with Fielder’s persona and approach, but if you only have time to cram one episode, check out the series finale, “Finding Frances.” It follows Fielder’s attempts to reunite one of his earlier clients with his long-lost love, and it is a nearly perfect encapsulation of his style and approach. You can stream Nathan for You on HBO Max, Hoopla or Paramount Plus, or rent it episode-by-episode on Prime Video.
You should also watch at least some of Fielder’s other series, The Rehearsal. Here, Fielder (again, portraying a version of himself) attempts to help people prepare for difficult conversations utilizing increasingly outlandish techniques and scenarios that spiral into comedic insanity, including building sets, hiring actors, and immersing himself and his subjects in an alternate reality of sorts—an approach that will prepare you for some of You Can See Everything’s swerves. If you don’t have time to watch the whole thing, just check out the first episode, “Orange Juice, No Pulp.” It’s a self-contained chapter and will give you the basics. You can stream The Rehearsal on Apple TV or HBO Max, or rent episodes on Prime Video.
Have more bandwidth before watching You Can See Everything, or want a deeper dive after the documentary has melted your brain? Here’s what else you can check out.
The companion podcast to Carreyrou’s book, Bad Blood: The Final Chapter, is well worth a listen. Hearing the journalist’s perspective about his investigation and work during and after Holmes’ trial offers additional insight into her personality and methods that will really resonate after watching her in Fielder and Oppenheim’s doc.
Once you’ve noodled on You Can See Everything for a moment, there are two more movies you should check out.
For more on Holmes’s specific grift and how she pulled it off, Alex Gibney’s excellent The Inventor: Out for Blood in Silicon Valley is the way to go, especially if you didn’t follow the Theranos story closely at the time and your main visual guide to Holmes is the new documentary. Gibney’s film is more traditional, but offers a wealth of insight into Holmes and her bubble. You can stream The Inventor on HBO Max or rent it on Prime VIdeo.
If you more from co-director Lance Oppenheim, watch Some Kind of Heaven, his 2020 breakout documentary about The Villages retirement community in Florida. It’s the film that put Oppenheim on the map, and offers hints of the same detached observational style that makes You Can See Everything so arresting. You can stream Some Kind of Heaven on Kanopy or Hoopla, or rent it on Prime Video.
Speaking of which, if you loved the visual style and “enhanced” reality of You Can See Everything and want to see more of that, catch the Robert Pattinson-starring Primetime, still in theaters. It’s Oppenheim's narrative film directorial debut, and combines his hyperstyled visual flair with thriller pacing and a narrative based on a true story.
The Curse, Fielder’s 2023 TV series, is his first work of fiction, and it offers some fascinating insights. It follows a newly married couple (played by Fielder and Emma Stone) who struggle to get an HGTV-style show off the ground as things get excruciatingly awkward (and, eventually, kind of surreal). After watching You Can See Everything, it will be interesting to see how Fielder views the power of editing, the ethics of reality TV, and working with a name actor to get the results that he wants. Plus, similarly to the film, The Curse swerves into a final act that will melt your brain. You can stream The Curse on Apple TV or rent it on Prime Video.
You should also definitely watch The Dropout, the 2022 Hulu series starring Amanda Seyfried. It’s a well-made film, and Seyfried’s performance was rightly praised for capturing the weirdness of Holmes' public persona. You can stream The Dropout on Hulu.
If you have personal conversations with ChatGPT that you don't want others to see, you can now add an extra layer of security for those specific chats. The new Locked Chats feature allows you to hide sensitive conversations behind a four-digit account PIN (or biometric authentication on mobile), meaning others with access to your device or screen can't read them.
Locked chats are retained in your chat history—unlike temporary chats, which are private conversations that aren't saved by default—so you can return to or continue them at any point. However, once locked, they are hidden from your regular chat history, pinned chats, and regular conversation search. To access them, you have to go to Projects > Locked chats and enter your PIN or device unlock method (such as Face ID or Touch ID).
To start a locked chat for the first time, open an existing conversation that you want to keep private and open the options menu. Select Lock chat > Lock chat and create or enter your account PIN. This moves the conversation to your locked chats project. You can repeat this with any existing conversation or start a locked chat from scratch under Projects > Locked Chats by entering your PIN or authentication method and typing a message to launch the chat. When you're done, select Lock now to end access to locked chats and hide them from the sidebar.
Note that if you don't already have an account PIN set up, you should be prompted to do so the first time you attempt to lock a chat. Select Create a PIN, enter your four-digit code, and enter it again to confirm. You can also enable biometric authentication in Account PIN settings. If you reset your account PIN at any point, all locked chats are permanently deleted and cannot be recovered.
Locked chats are available on ChatGPT Free, Go, Plus, Pro, Business, Enterprise, and Edu plans on both mobile and desktop apps. The feature is not available for ChatGPT for Healthcare. There are other limitations, too: For example, temporary, archived, and group chats can't be locked, nor can those with shared links, scheduled and background tasks, or meeting recordings. OpenAI says locked chats are rolling out gradually, so they may not be available to all users yet.
While locked chats do add a layer of privacy in terms of how conversations appear on your device, they are not suddenly encrypted or otherwise protected from being retained or used by OpenAI. As the support page notes, locked chats are not encrypted end to end or placed in a separate encrypted vault, and locking a chat does not affect other privacy settings, such as Improve the model for everyone (which can be turned off in Settings > Data controls) or remove information ChatGPT already knows about you.
This week's Out of Touch Guide delves into the musical genre/sub-culture of "Loser Core," a pointed response to overly polished music and Instagram-polished fashion. We also define some slang, so if a young person says you have EBK or nods and says "GiGis" to you, you'll know whether you're being insulted or complimented. And then there's Biddy Blas, Star Wars fandom's newest Glub Shitto.
Loser Core is a growing subculture and musical style among young people. The music is influenced by indie, alternative, and lo-fi, with lyrics that tend toward depressing. They're usually about alienation, teen angst, and how hard it is to talk girls. Loser Core is like emo, but more homemade and more hopeless, or like grunge with less anger. Prominent Loser Core artists include Memo Boy and Wifiskeleton. Notable influences include Radiohead and Weezer.
The Loser Core look is anti-fashion: over-size, slouchy hoodies, anime T-shirts, digital watches, and beat-up sneakers. The personality type is alienated nerd with niche hobbies and few friends. So basically, depressed young people who gravitate to depressing music and a slobby look. No matter your age, I'm sure you knew (or were) this type of person when you were 17.
Here's a video that explains it a bit more:
So that's Loser Core, the genre and lifestyle. Explaining the prominence of the Loser Core Playlist is a little more complicated. Created on Spotify by depay178, LOSERCORE is a playlist of 155 songs in the Loser Core genre. Back in June, a TikTok user called @user406398991 started posted AI-generated cartoons to promote the LOSERCORE playlist. The videos feature a recurring cast of characters, are usually about music, and end with a recommendation to visit the LOSERCORE playlist. Here's an example:
The videos started appearing on a lot of people's feeds in October, raising the profile the musical genre and generating reenactment videos like this:
...as well as videos from artists who are confused at their own inclusion on the list, people selling rip-off merch, and people gate-keeping what Loser Core is and questioning the inclusion of some songs on what is becoming the canonical list. Also: a ton of videos asking, "why does this keep showing up on my feed?"
Online chatter about the playlist revolves around the perceived lack of quality of the videos, and why, exactly, someone would spend so much time promoting a playlist on Spotify when you can't monetize the followers it gets.
This Tweet is currently unavailable. It might be loading or has been removed.
It's a good question. There are indirect ways this list could be making someone money—maybe there's a connection between the promotion and the bands on the list who get paid per stream—but I like to think it's just kids who identify with these artists and want other kids to hear what they're into, like people did with zines in the old days. We don't have to monetize everything, right?
EBK stands for "Extreme Baller Knowledge." It goes beyond the sports reference to apply to to anyone who has deep knowledge of a subject. It's usually used ironically.
Pronounced like the name Gigi, "GiGis" comes from sports by way of gamer culture. The post game pleasantry "good game" has been a staple in sports since forever. Gamers adopted it and shortened it to "GG" in text. So GiGis is a way of saying "good game."
This week's viral video comes by way of a video game cut-scene. Star Wars: Galactic Racer was released on Oct. 6, and its breakout star has emerged: Biddy Blas. You won't find him in any of the Star Wars movies or TV shows. Biddy is only in this one game, and basically only exists as an opponent to be beaten in a race. But Biddy is too great to go out in obscurity. Just check him out:
Soon after the release of the game, fans dubbed Biddy the "best Glup Shitto (Star Wars background character with a silly name)" and started making memes to celebrate him, like these:
Biddy Blas' popularity is the result of voice actor Richard Summers-Calvert's spirited performance, particularly his reading of Biddy's catchphrase, "scrib doon," (i.e.: "Write that down."). The phrase gets caught in your brain:
and there are so many occasions when "scrib doon" is the perfect thing to say:
This Tweet is currently unavailable. It might be loading or has been removed.
While ChatGPT might have been first out of the generative AI gate, Google has established itself as more than a major player in this new technological race. While plenty of people use Gemini free of charge, there are also a plethora of paid subscribers who want access to Google's latest and greatest models. Like other AI subscriptions, Google's best plans are wildly expensive, but it's cheapest paid tier, AI Plus, is much more reasonable: For $4.99 per month, you get access to faster Gemini models, and bump your Google storage up to 400GB to boot.
But if you're subscribed to Google AI Plus for access to better versions of Gemini, change is coming. Google is warning subscribers this week that they're losing a major perk—perhaps the perk—of the AI Plus subscription tier, as subscribers will soon no longer have access to Gemini's advanced models.
As 9to5Google reports, this isn't necessarily out of the blue. When Google rolled out Gemini 4 Argon, the company advertised that its latest model would only be available on AI Pro and AI Ultra plans, not AI Plus. At the same time, the company said that its plans meant AI Plus subscribers would soon lose access to all Gemini Pro models. It now seems those plans are now rolling out—one Redditor shared an email they received from Google as an AI Plus subscriber, which reads:
We wanted to let you know about changes to the models that you can access with your Google AI Plus plan.
Starting on 13 November 2026, the ability to select Pro models in both the Gemini app and AI Mode in Google Search will only be available to Google AI Pro subscribers and above. You’ll still have access to our latest Flash-Lite model and Flash, built for speed and intelligence. Your previous chats using the Pro model can be continued with a different model.
To learn more about these changes, you can visit our Help Centre article for additional details. You can manage or cancel your subscription at any time.
Thank you for being a Google AI Plus subscriber.
– The Google One team
According to the message, AI Plus subscribers have less than a month left to access Gemini Pro models. On Nov. 13, these users will be limited to Flash-Lite and Flash in both the Gemini app and AI Mode. While it's fortunate that previous Pro chats will still be available, users will need to use one of these less powerful models if they want to continue those conversations.
So now, access to models in Google's AI subscriptions looks like the following:
Free: Gemini Flash-Lite
AI Plus ($4.99 per month): Gemini Flash-Lite and Gemini Flash
AI Pro ($19.99 per month): AI Plus perks, plus Gemini Pro models and access to Gemini 4 Argon
AI Ultra 5x ($99.99 per month): AI Pro perks, plus higher access to Gemini Pro models and 5x higher usage limits
AI Ultra 20x ($199.99 per month): AI Ultra 5x perks plus 20x higher usage limits and highest access to Pro models
It always rubs me the wrong way when companies take away features after customers have already paid for them. If Pro models were always locked to a higher tier I could think, whatever, that's just what Google charges. But the fact that AI Plus subscribers have paid for access to Pro models, and now the same $4.99 subscription is missing a major feature for no particular reason, is, in my view, scummy. If you're subsequently rethinking your choice of Google subscription, here is a full breakdown of all perks across the various subscription options.
Gemini 3.1 Pro launched back in February of this year—quite a while ago, in AI terms. Now Google has jumped to the next full number with Gemini 4 Argon, which it says "excels at deep work in complex, real-world tasks."
You can get the new AI model now inside the Gemini app if you're on the AI Pro ($19.99 a month) or AI Ultra ($99.99 or $199.99 a month) plans. Users on the free and AI Plus ($4.99 a month) plans miss out for now, though you can expect a smaller, cut-down version of Gemini 4 Argon to roll out to these tiers eventually.
Google has a long list of areas where the new AI model excels. There's mention of complex engineering tasks (like checking millions of lines of code for efficiency improvements), predicting molecular structures in chemistry, and delivering 3D visualizations. It also enhances knowledge work across legal, finance, science, and coding.
Google says Gemini 4 Argon is now better at advanced physics modeling, picking through genetic databases, protecting software from vulnerabilities, and resolving problems in physical space (handy for robotics development). The new model "answers questions more directly with less hedging, doesn't repeat itself more than necessary, and skips clichés throughout the responses," according to Google.
With many in the industry worried about an AI-driven apocalypse of some form or another, Google has upped its defenses against misuse and jailbreaking. The company has significantly upgraded its blockers for harmful requests, limits on AI agent overreach, and resistance to malicious prompt injections.
All of these stats are things you would expect to read about in a new AI model announcement, whether it's from Google, Anthropic, or OpenAI: "Our new model is better and safer than ever, and particularly good in these specific benchmarks we've chosen." But is it any more helpful for the average user tapping away in the Gemini app?
I don't have any pressing scientific breakthroughs to research or large-scale software optimizations to run, but I have put Gemini 4 Argon to the test against GPT-6 Sol, which is the latest model available to consumers through the ChatGPT app—though you need to be on at least the Plus plan ($19.99 a month) to get to it.
Make no mistake: Both of these new AI models are quite capable. I first prompted Gemini 4 Argon and GPT-6 Sol with the following: "Using information online, create a web app I can run locally that plots the lighthouses of the world on a map. Clicking on a lighthouse should give me some details about it, show how far its light reaches, and display a simple 2D model showing its shape and height (if known). Put a lighthouse keeper in each model pop-up for scale."
At the highest, slowest levels of "thinking," these AIs took a while to get back to me: Nearly 10 minutes for Gemini and almost 26 minutes for ChatGPT. Both hit the same problem—restricted access to map data online—but both were able to make a rudimentary world map of their own when prompted a second time.
The ChatGPT effort, dubbed "Beacon," wasn't bad, apart from a bit of map clunkiness (it clearly needs to work on its cartography skills). It presented plenty of information, laid out everything well, and gave me nice, neat diagrams on each of the lighthouses featured. The day/night mode worked fine, and it was an impressive bit of coding work, even if it did take 26 minutes.
However, the Gemini effort, called "Pharos," was better in almost every respect. The map was more detailed, lighthouses actually showed their light range (with an impressive animation), and there was more information on each location—perhaps due to Google's expertise in web search. Some of the layout choices were a bit off, but Gemini 4 Argon produced what was clearly the superior app in less than half the time.
Next, I fed PDFs of 10 recent scientific studies into each AI model. The prompt: "Give me a brief summary of each of these studies in plain English, with a single sentence each for the main finding, one study strength, one study weakness, and the main implication." While I hadn't read the studies through entirely myself, I had read news stories on all of them, and I lowered the "thinking" level to low.
This time ChatGPT was faster—25 seconds versus Gemini's 55 seconds—and gave what I thought were the better summaries, at least for my purposes. Gemini did well, and its answers were accurate and to the point, but the summaries ChatGPT gave were easier to understand and shorter. ChatGPT also offered some overall thoughts on the most promising study, which Gemini didn't.
Finally, I checked the imagination and creative juices of the models: "With the success of the Apple TV shows Severance, Slow Horses, Silo, and Widow's Bay, give me five pitches for new show ideas on the platform that combine elements from some or all of these shows. Try and be as original as possible and keep each pitch to around 100 words." I put the "thinking" level back up to "high" for this job.
(Is this how Apple TV executives actually get their ideas nowadays? I hope not.) As far as the answers the AIs gave me, Gemini was faster than ChatGPT (1 minute and 28 seconds against 3 minutes and 48 seconds, respectively), and both served up responses that were a bit more literal in their mashing up than I wanted: failed secret service agents and supernatural terrors abounded, which is I suppose what I asked for.
The ideas were mostly "Widow's Bay set somewhere else" or "Slow Horses set somewhere else," and didn't make a whole lot of sense for the most part, but the best Gemini offering was about a Silo-style underground bunker whose operators go through a Severance-style procedure every 24 hours: outside, the bunker protecting it from zombies by night; inside, the bunker planning social activities for residents by day. It's not perfect, but I might watch the trailer at least.
My favorite of the ChatGPT suggestions was a show about a remote, Severance-style weather station whose employees discover they're not recording the weather—they're creating it. Again, it didn't really logically hold up, but overall I found the suggestions from GPT-6 more nuanced, more likely to be an actual show, and better "thought" through. Maybe Gemini should've given itself more time to "think."
This is only a narrow set of quick tests, but, in my view, it seems Gemini is back to, at the very least, an even playing field with ChatGPT. It performed much better on the lighthouse task, though its summarizing and TV show pitching weren't quite as impressive. Access to the best models requires a subscription on both platforms, but if you're able to, it's worth comparing them against each other on the jobs you most commonly need AI for.
This week's deep dive into young people culture answers questions you may not have thought to ask, like, "Why are young people throwing fruit in their toilets?" "What does it mean when someone tells me to 'visualize spinning apples'?" "What is glass skin?" and "How can Minecraft help you understand the age of the universe?"
This is a weird one, but young people are throwing fruit in the toilet in honor of the rapper Lucki. Backstory: The Atlanta rapper best known for his track "New Drank" was attending ComplexCon in Los Angeles on Oct. 3 when an entourage vs. entourage tussle with rapper Playboi Carti's crew broke out and Lucki was stabbed.
In response, the hashtag #throwfruitinthetoiletforlucki appeared on Instagram and TikTok, and kids started throwing fruit in the toilet and posting videos like this one:
and this one:
So that's what's happening; why it's happening is a little harder to explain. People posted screenshots indicating the lyrics to Lucki's "Y Not" include the line "When I get stabbed at ComplexCon, throw fruits in the toilet"—but the song doesn't actually include those lyrics and it's hard to imagine many fans bought into it. Some suggested that the trend is a dig at Lucki's lifestyle and weight: a not-subtle suggestion he eat more fruit. Maybe, but I think it's a classic online non sequitur. I don't think anyone is legitimately trying to honor Lucki by posting these videos. It seems like jokers trying to get fans of the rapper to throw fruit in their toilets so they can say, "Look at those weirdos! We tricked 'em into throwing fruit in their toilets!" But it doesn't seem to be working. All the fruit-throwers I've seen seem like they're "in on the joke," which actually reverses the joke. Anyway, Lucki is doing well and expected to make a full recovery.
The second fruit-related meme trending this week is "imagine spinning apples," a way to insult pretentious dinks online. It's a journey to explain how it came to mean that, so here goes.
Aphantasia is the inability to visualize things. If you can't imagine apples, you may have aphantasia. The idea that some people can imagine apples and some can't is big on 4Chan, Reddit, and other online communities populated by young people who would like to claim superiority over others, but can't do it based on money, attractiveness, or any other traditional measure of preeminence. Despite the actual neuroscience showing a correlation between aphantasia and a slightly higher IQ, these online weirdos decided that having aphantasia means you are dumb, or lack an inner life. Some took it even further and suggested that intelligence can be measured by the number of apples one can visualize and whether you can visualize them spinning.
So memes like this:
and posts like this:
This Tweet is currently unavailable. It might be loading or has been removed.
started showing up as a way of clowning on people who boast about how smart they are using a metric that actually makes them look dumber. It's like bragging about the results of an online IQ test or joining Mensa.
Glass skin is a beauty trend taking over TikTok. It's based on Korean skin care trends and refers to skin so smooth, luminous, and clear, it seems to shine. How one can achieve this look is debatable, but there is no shortage of videos giving instructions and hawking creams and tinctures that promise to make your skin shiny as glass. Like this one:
I'm in favor of people looking as greasy as they want, but viral "glass skin" videos are not giving good advice—a recent study found that almost 95% of viral Korean skincare and “glass skin” TikToks are misleading, and many can even be dangerous.
Reports about the new "door kick challenge" are coming from all corners of the nation, from Portland to Phoenix to West Virginia to California this week.
Here's how West Virginia's Taylor County Sheriff's Office Deputy Sheriff Chris MacQueen describes the door kick challenge: "There is a new trend out that’s either referred to the 'door kick challenge' or 'donkey kick challenge' and what that essentially is, is teenagers late at night are donkey kicking doors and then recording their reactions as they run off. This is absolutely not ding dong ditch."
There's so much wrong with this statement. First, the "door kick challenge" isn't new. Here's a report from 2021 about the supposed challenge. Hundreds, maybe thousands, of similar news stories have been published regularly since, all across the country, a steady drumbeat of warning about this dangerous trend. But it doesn't seem to be real.
I've looked through TikTok's doorkickchallenge hashtag and similar places on other video platforms, and I can't find any evidence of any young people "recording their reaction" to kicking a stranger's door or mentioning the challenge in a "you should try this too" kind of way. There are plenty of videos of people kicking doors, but they're all shot from doorbell or security cameras, presumably posted by homeowners, and we have no idea who is doing the kicking or why. They could be stalkers, robbers, or angry because they left their keys on the coffee table. So there is no discernible trend or challenge, beyond the trend of more people having doorbell cameras.
The cop's last point, that "this is not ding dong ditch," is debatable. Ding dong ditch dates back to the early 1800s, and I can't find any evidence that the game has changed since then, or that more kids are kicking doors instead of ringing doorbells. No one keeps stats on it, but it seems reasonable to assume that kids today do dumb things as often as you and your friends did dumb things. That's where the safe money always is.
If police departments want to warn the public about a truly deadly trend, they could look at the number of adults who shoot teenagers on their porches. Then they could hold a press conference warning about the "murdering children for ringing your doorbell" challenge.
Looking at the volume of AI-generated brainrot that floods video-sharing platforms makes it easy to conclude that young people are not able to visualize apples spinning. But the counter evidence is out there, too. Like this week's viral video: "Running Minecraft Until It Beats Itself." It's a video about a kids' video game, but on a deeper level it's a video about deep time, evolution, the second law of thermodynamics, and more.
This video runs a simulation of a Minecraft world to prove that if you have the right starting seed and leave Minecraft running long enough with no input, the winning condition (the dragon being slain) would happen even if there was no player, by chance. You'd have to run your PC for a billion years instead of playing Minecraft for 100 hours, but it would happen.
The idea that anything that could happen will happen if enough time passes isn't new, but it's hard to grasp the concept of time scales so vast. This video uses a relatively simple model (that viewers understand intimately) to make the billions of years imaginable.
I have no doubt that the seven million people who have viewed this video (and played Minecraft enough to understand the implications) grasp the concept of a billion years better than the people who haven't: A billion years is about how long it would take for an Enderman to randomly drop a carved pumpkin on two snow blocks and for the resulting snow golem to lead a creeper to blow up next to the walls around the end portal so Endermen could randomly carry enough blocks from the Overworld to The End to make piles high enough that randomly moving snow golems would throw enough snowballs to destroy all seven crystals that protect the Ender Dragon.
An abundance of AI imagery is something we all have to live with these days—from school posters to restaurant menus—and while there are still certain tells that indicate something was made with AI, they're not always present. In a lot of cases, it can be almost impossible to distinguish an AI image from a real one.
This is a problem for all kinds of reasons, and, as such, the AI companies are attempting to address it. Images produced by Gemini and ChatGPT come with an invisible watermark called SynthID. That way, AI-generated images don't need visible branding in order to identify their origins. Of course, you need access to a detector to know whether this watermark is present. As it happens, Google just launched a standalone SynthID checker tool on the web, free for anyone to use. I gave "SynthID Detector" a shot, and while it largely works as expected, it isn't perfect.
For a while now, Gemini has let you upload an image and ask if it is AI-generated. With the new online portal, it's even easier to check—though you do still need to sign in with your Google account. Once in, you can check a wide variety of files. This checker supports JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF, and GIF image formats; MP4, MOV, and WEBM for videos, and WAV, MP3, OGG, FLAC, AAC, and M4A for audio. As long as a file was made by a SynthID-supporting AI, it should have a watermark. (The SynthID watermark is a standard supported by Nvidia, OpenAI, Kakao, and Google, though OpenAI has its own detection tool.)
Google doesn't say exactly how SynthID works, presumably to make it more difficult to get around, but does confirm the watermark is "robust against common transformations." It also says the watermark is "woven into the content itself", which means pixels for images and video, and frequency components for audio. The company also admits its detection tool isn't infallible. What it calls "heavy modifications" can degrade the watermark sign, as can multiple "transformations" applied to the same image. The recommendation is to use files that are as high quality as possible, so it seems to me that more data makes detection more reliable (similar to AI text detectors).
Google also says the SynthID watermarks are "imperceptible" to humans, so you don't need to look or listen out for them—they're only revealed by an AI detector tool like this. If you're generating images, video, or audio with AI, the watermarking shouldn't impact the quality of the end result in any way.
I started with some easy tests. The detector correctly spotted that images created in Gemini and ChatGPT, and downloaded directly to disk, were indeed fake. It only took a few seconds to confirm, and it also identified the AI tool that had been used to make them.
It also correctly marked one of my own photos as AI-free, although it did hedge its answer a little, saying that it was "unlikely" that the media was made with AI. All the detector could say with confidence was that there was no SynthID watermark, showing that it's this watermark in particular that it's checking for. It's not looking for any other signs of AI.
I then tried screengrabs of the same AI images, and the detector correctly flagged that they came from Google and Anthropic's AI tools. This shows there's something buried deep in the pixels, rather than in the file metadata itself, revealing the SynthID watermark. You can't just copy and paste to get around it. Applying an oil painting filter in Photoshop to an AI image didn't trip up the detector, nor did adding a slight blur. Adjusting color, contrast, and tone made no difference, and the detector still spotted SynthID when I cropped and combined two different AI images: one from Gemini, one from ChatGPT.
Eventually, I did beat the detector. Once, when I heavily pixelated an AI image—at which point I rather suspect my disinformation campaign would be pretty ineffective anyway—and once when I created a composite image, which was 75% real, 25% AI-generated. With less AI to work with, the detector can falter.
On the back of that result, I used ChatGPT to add an extra person into an otherwise genuine group photo, though this time the tool detected SynthID—even though less of the overall image was AI. There might be some work to do in images that are only partly AI, which is understandable as the detector has less data to work with.
There is a quota limit here, which seems to be around 10 files per day—but 10 each for images, videos, and audio files. Once you've hit your limit, you need to wait 24 hours to try again. As far as I could tell, the SynthID Detector is exactly what Google says it is: a useful check to have, but far from a guarantee one way or the other.
We may earn a commission from links on this page.
Streaming in September was eclectic. The top 10 most popular TV shows of the month, according to data from JustWatch, include opposite takes on international espionage—Lioness's "everything is at stake" blockbuster vibe versus Slow Horses' darkly funny, morally ambiguous style.
Other series burning up the charts: spy-adjacent action show Reacher, a couple of sports comedies named after their main characters (Ted Lasso and Chad Powers), a dystopian sci-fi drama (Silo), a superhero show (Lanterns), and stories of criminals and cops (Furious and The Gentlemen). There's even a horror comedy (Widow's Bay). You could say there's something for everyone.
Topping the streaming charts this month is DC's TV revamp of Green Lantern, in which experienced Lantern Hal Jordan (Kyle Chandler) trains the new guy, John Stewart (Aaron Pierre), on how to wield the mysterious and powerful Green Lantern Ring. It's a reserved, neo-noir style take on the comic mythos, as the heroes investigate a series of murders tied to an extraterrestrial threat. Lanterns is pulling in a 92% fresh ratings from Rotten Tomatoes, so it's definitely worth a stream. Stream Lanterns on HBO Max.
Season four of Ted Lasso sees the title character, played by Jason Sudeikis, returning to the UK to take on a new challenge: coaching the Lady Greyhounds, AFC Richmond's second-division women's football team. This multiple Emmy Award-winning series still has all the heart, humor, and positivity of the original run. Stream Ted Lasso on Apple TV+.
Reacher actor Alan Ritchson is a great action star. Dude's the size of a house, but he uses his brains as much as his fists, and his "independent ass-kicker" lifestyle makes for great TV. Season four sees our morally uncompromising, giant-sized hero witnessing a supposed suicide, then finding out it's tied to a conspiracy so huge that world governments are involved. If you like your TV action-packed, you're probably already watching Reacher, but if you're not, get on it! Stream Reacher on Prime Video.
If you like action, international espionage, and incredible performances, Lioness mashes them all up into a must-watch series. This cast is without lightweights, including Oscar winner Zoe Saldaña, Oscar winner Morgan Freeman, Emmy nominee Michael Kelly, and executive producer and Oscar winner Nicole Kidman. Season 3 dives into a hidden network of dangerous foreign operatives, personal betrayals, and a high-stakes war that could define everyone's lives. Stream Lioness on Paramount+.
Apple TV's Slow Horses delivers a more realistic, nuanced take on espionage by detailing the lives of the losers of British Intelligence. Instead of chiseled good guys and gals saving the world, the agents of Slough House are a team of dysfunctional, troubled souls trying to do the right thing (usually) in a morally gray world. Just like real agents. Their leader, gruff, obnoxious Jackson Lamb, is played to perfection by Gary Oldman. Slow Horses has an almost perfect Rotten Tomatoes score over its first five seasons, and season 6 is setting new records. Stream Slow Horses on Apple TV+.
In Furious, Emmy Rossum plays Alice Black, an FBI agent and former NYPD homicide detective investigating a series of murders that point to a female serial killer bent on offing the wealthy assholes who victimized her when she was a sex trafficked. If that's not a killer premise for a TV show, I don't know what is. Critics agree: Furious has a 99% positive rating on Rotten Tomatoes. Stream Furious on Hulu.
Director Guy Ritchie's series is set at the point where the remnants of the British aristocracy meet the nation's gangsters and lowlifes. Theo James plays Eddie Horniman, aka the Duke, whose inherited family estate is barely paid for by profits from a drug-dealing operation. If you like hard-edged action, British wit, and a stylish gangster vibe, this is the show for you. Stream The Gentlemen on Netflix.
This series from Apple TV has won unanimous praise from both critics and audiences. It's a rare horror-comedy that manages to be genuinely scary and really funny. In the island town of Widow's Bay, literally everything seems to be haunted, cursed, or otherwise beset with unspeakable evil, but Mayor Tom Loftis (Matthew Rhys) wants to bring tourism to the island anyway, so he ignores the sea hags, fog monsters, and serial killer ghosts surrounding him in an effort to drum up new business. The great Stephen Root plays Wyck, an old-timer who's always there to remind everyone that time is short and everyone is doomed. Stream Widow's Bay on Apple TV+.
Chad Powers is a sports comedy that's actually funny. Glen Powell stars as Russ Holliday, a former college football phenom who nuked his own career before he even got to the NFL. Instead of giving up, he dons a disguise, adopts the name "Chad Powers, and tries out for a failing college team. It works, but leads to outrageous complications as the jock tries to hide his deceit. Stream Chad Powers on Hulu.
Apple TV's Silo continues to track the downward spiral of the dregs of humanity, who are all living together underground in a cavernous edifice that goes all the way down. If you've been wondering how things got so bad, this season goes back in time to reveal the source of humanity's near-demise. (Unsurprisingly, humans are to blame.) You can stream Silo on Apple TV.
It appears agentic AI is all the rage: OpenAI's new agents, "Dots," rolled out this week, while Meta's Muse just passed five million downloads. Rather than just chatting with or using them to generate images, AI companies want users to start asking their bots to do things on their behalf, from managing their inboxes to ordering dinner.
But agents aren't just the next logical step for AI as a technology. When implemented poorly, they can be wildly dangerous. AI agents are designed to run on their own, and, in order to do that, they often need near-total access to your devices and their data. Some people are now giving agents full use of their computers, including all of their messages, files, passwords, and accounts. That's a lot of trust, especially when it's not that difficult to "hack" one of these agents for malicious purposes with a poisoned prompt hidden on a website, which could lead your AI agent to hand hackers the keys to your bank accounts. That's no good.
How does Apple feel about this? The company is famously behind on the software side of AI—it only rolled out its AI-powered version of Siri in September—but its AI hardware business is booming. When the AI agent craze first took off with projects like Moltbot, customers bought Mac minis in droves in order to run those programs. Now, Apple advertises its various Macs with AI at the forefront: It doesn't offer its own models like OpenAI or Anthropic, but it's making plenty of money selling you the hardware to run those models.
Perhaps that's why Apple seems quite concerned about the AI agent trend, particularly when it comes to the data users are exposing to run these programs. Apple is all about privacy and security, and as such, I can only imagine the company watching in horror as so many users willingly hand their Macs (and all of their data) over to automated AI bots.
Against that backdrop, the company posted an update on its developer forum entitled "Updates to Full Disk Access in macOS." In it, the company highlights the dangers of developers asking users to grant "Full Disk Access" permissions to an app. Apple says this function has been traditionally reserved for backup apps that need to access the entire disk to function. (After all, if your backup app can't access your entire Mac, how can it back up your entire Mac?) But now, Apple says developers are asking users to expose everything on their Macs, including messages, mail, browsing history, and more, without those users fully understanding the consequences. Apple even argues the privacy implications extend beyond the user, as communication apps that demand Full Disk Access compromise the people the user is chatting with, too.
And Apple appears to be doing something about it: While details are slim, the company says it will introduce "additional controls to ensure that users who genuinely wish to grant an app this extraordinary level of access can only do so with very explicit user action." I'm guessing this means that in the future, once you download an agent to your Mac, you'll have to click through a series of warnings and settings to "prove" to macOS that you really want to expose your entire Mac to the app and its AI. The company says it's doing this now, as the threat will only grow as AI agents improve and become more autonomous.
I'm all for additional safeguards when it comes to AI. In the past, you needed some level of technical expertise to install an AI agent on your Mac. As such, you likely understood the risks involved, or willfully ignored all the warnings in order to do so. Now that running agents is as easy as installing Muse on your Mac, the barrier to entry is minimal. Anyone can do it, even if they don't fully understand what they are actually doing. As the AI industry reckons with the risks its technology poses to humanity, companies like Apple should indeed be figuring out how to protect their users from AI's current dangers—even if users can still choose to ignore those warnings.
We may earn a commission from links on this page.
The most-streamed movies in September, according to JustWatch, include a couple of early Halloween treats: low-budget, high-scare horror flicks Backrooms and Obsession both made the top 10. If you like more gentle supernatural stories, 1990s classic Practical Magic made the charts too. You could also check out best-actor-ever favorites Robert De Niro in The Whisper Man and Dustin Hoffman in Tuner. So there's something for everyone.
Director Kane Parsons' analog horror breakthrough is proof that atmosphere, mood, and a fascinating story is scarier than CGI monsters every time. Based on the viral YouTube series of the same name, Backrooms tells the story of Clark (Chiwetel Ejiofor), a store owner who discovers a labyrinthian liminal space in his furniture shop's basement. Renate Reinsve plays his therapist, who thinks it's a delusion, until she goes downstairs for a visit. A hit with critics—Backrooms is sitting at 87% fresh on Rotten Tomatoes—and audiences, Backrooms is a delicious early Halloween treat for horror fans. Stream Backrooms on HBO Max.
Released to streaming to coincide with the sequel's release in theaters, Practical Magic is a beloved supernatural rom-com finding a new audience in 2026. Sandra Bullock and Nicole Kidman play Sally and Gilly, sister-witches cursed with losing any man they ever love. Sally's romantic interest appears in the form of Aidan Quinn playing a detective investigating the disappearance of Gilly's abusive boyfriend. Can love, magic, and the power of sisterhood break the family curse? With a little help from '90s nostalgia, you bet they can. Stream Practical Magic on HBO Max.
Director Curry Barker is still in his 20s, but his debut theatrical feature is the kind of horror film that changes the entire genre. Obsession is a classic "be careful what you wish for" story that feels fresh thanks to its talented young cast, unpredictable plot, and careful pace. Michael Johnston plays a young man too scared to tell his crush how he feels, so he wishes she'd fall in love with him instead. The result is a nightmare embodied by Inde Navarrette, whose portrayal of the villain/victim is at once harrowing, terrifying, and heartbreaking. Obsession is the best horror movie I've seen in years. Stream Obsession on Peacock.
Ryan Reynolds and Kenneth Branagh play Cold War enemies brought together when Reynolds' plane crashes in the Russian wilderness. Branagh is ex-KGB, but he's always been a secret fan of America, so he helps Reynolds escape the USSR. Like the best bro-coms, Mayday is powered by the charisma of its stars, and you won't find any actors alive with more rizz than these two. If you like 1980s-style buddy action/comedies, Mayday is the right stream for you. Stream Mayday on Apple TV+.
In this crowd-pleasing sci-fi flick, Ryan Gosling stars as Ryland Grace, a middle school science teacher who is blasted into the firmament in a last-ditch effort to save the world. Grace is the only person on Earth who might know how to protect the sun from a cosmic infection that threatens all life, but he's definitely not an astronaut and doesn't see himself as a hero either. Will he rise to the occasion and learn something about his inner strength and resilience while saving us all from death? My money is on, "yeah, probably." Stream Project Hail Mary on Prime Video and Fubo.
You shouldn't mess with a superhero's dog. Milly Alcock plays Kara Zor-el, AKA Supergirl, who traverses the galaxy hunting for Krem of the Yellow Hills (played by Matthias Schoenaerts). He shot Krypto the superdog with a poison dart, and justice must be administered. If you want a grittier take on the super-people of Krypton with some deep-lore DC universe building, look no further than Supergirl. Stream Supergirl on HBO Max.
Based on the novel by Alex North, The Whisper Man stars Robert De Niro as Peter Willis, a retired detective who returns to the game to solve a grisly cold case. A copycat is mimicking the crimes of the notorious Whisper Man, a serial killer Willis captured years before. This intense thriller is built around De Niro's powerful performance and a twisty plot that will keep you on the edge of your couch. Stream The Whisper Man on Netflix.
You wouldn't think the life of a piano tuner would be exciting until you see Tuner. Leo Woodall plays Niki White, whose quiet life is turned to 11 when he learns he can use his ears for cracking safes as easily as tuning Steinways. Desperate for cash, White joins up with hardened criminals, and things go haywire. The excellent cast of this smart, tense thriller includes Dustin Hoffman and Havana Rose Liu. If you like suspense in the style of Hitchcock, check out The Tuner. Stream Tuner on Paramount+.
Directed by Jon Favreau, The Mandalorian and Grogu sees Din Djarin (Pedro Pascal) and his adorable baby Yoda pal pitted against Imperial warlords and the Hutt Twins. It's classic Star Wars: a lone hero and his companion doing heroic things while facing impossible odds. Packed with cliff-hanger action and colorful characters, The Mandalorian and Grogu is pure fun in movie form. Stream The Mandalorian and Grogu on Disney+.
Pixar is so good at what it does, it can even make the fifth movie in a series about ratty old toys into art—not just art for kids, but art that's palatable for parents, too. In Toy Story 5, Woody, Buzz, Jessie and the gang face a toy's most terrifying enemy: technology. Toy Story 5 asks if there's a place for action figures and Mr. Potato Head in a world where screens are the only thing kids care about. Stream Toy Story 5 on Disney+.
AI is virtually everywhere in 2026; that now includes U.S. government websites. On Tuesday, the Trump administration launched "America.gov," a new website that uses AI to answer questions about the government. While the experience currently feels like ChatGPT running U.S. government-themed instructions, the plan is for America.gov to eventually let you take action as well.
America.gov "combines" the nearly 30,000 websites under the umbrella of the U.S. government into one set of sources. Rather than invite you to browse those tens of thousands of pages to answer your questions, that dataset serves as the basis for the site's proprietary chatbot. You ask the chatbot a question about the U.S. government, and the AI sifts through those websites to generate an answer. It's like ChatGPT, but if the model only had access to U.S. government websites and sources.
You could ask the site how to replace your Social Security card, renew your passport, or register to vote in your state. You could ask for information on changing your address with the USPS, when Medicare open enrollment dates are, or upload a PDF of a government form for assistance. The site says any personal information will be automatically removed, though the privacy policy clarifies that it "will make a best-effort attempt in your browser to remove any private details using Rampart (an open-source classifier system to detect some types of PII, such as Social Security Numbers, phone numbers, email addresses, and credit card numbers)."
The website currently supports three languages—English, Spanish, and French—as well as voice commands, if you prefer speaking to your chatbots rather than typing to them. You don't need an account to use it, nor is there a fee or subscription, and there are no ads.
In its current form, America.gov is a simple chatbot. But the site has bold claims for future features. Secretary of State Marco Rubio announced plans to let users apply for and renew their passports directly through the site, including a dynamic progress tracker to keep tabs on your documents. Other future abilities could include enrolling in Medicare or finding a job with the federal government.
But as with all claims, take these with a grain of salt. The government says these features are coming in early 2027, but this is a tall order for such a project. A single website that lets you complete complex government processes with the simplicity of a chatbot sounds great, but it would need to be accurate and reliable for all possible tasks and functions. It would certainly take significant proof of security and dependability for me to trust such a program with renewing my passport or helping me enroll in health care.
It's impossible to discuss America.gov without acknowledging the administration that is spearheading the project. One would assume that an AI-powered government website developed by this administration would be biased towards Trump and his many falsehoods. And yet, the tool is astoundingly neutral, to the point where I can only imagine how frustrated Trump would be by some of these answers.
The first question many tested America.gov with was "Who won the 2020 election?" Donald Trump famously and frequently says that the election was stolen from him, but his new website plainly states that Joseph R. Biden Jr. won. The bot does sometimes tie the answer to the electoral count, stating that Biden won 306 votes to Trump's 232, but I was surprised to see "Joseph R. Biden Jr. won the 2020 presidential election under the official Electoral College count" as an answer on this site.
The site will also tell you that Donald Trump was president when diesel prices hit their highest historic price ($6.529 per gallon as of Sept. 21, 2026.), and that prices are high, in part, due to "military action" affecting the Staight of Hormuz. It will also say that grocery prices were higher in August than they were a year ago, which contradicts Trump's claims that prices have come down under his term.
Of course, Trump-influenced information that appears on official government sites will appear here. Ask it what the body of water that abuts Texas is called, and it will say the Gulf of America, formerly the Gulf of Mexico. Ask it what lake borders Toronto, Canada, and the bot will say Lake America, formerly Lake Ontario. In both cases, it will refer to the executive orders that changed those names.
Typically, I'd be worried about an AI tool hallucinating its answers and providing incorrect information to users. That's still a concern here, of course. But by creating a chatbot that only pulls information from official government sources, Trump has inadvertently created an official government fact-checker. And unless Trump starts injecting these sites with the same falsehoods he tells in interviews and on Truth Social, America.gov may be a reasonably reliable source of information.
Not only are AI agents all the rage in the tech world, but they're becoming more mainstream by the week. Apple recently upgraded iOS and macOS with Siri AI, while earlier this month, Meta also launched "Muse," an agent that can perform tasks on your behalf—and potentially expose your private information. Now, it's OpenAI's turn to launch an AI agent, but its take on the trend is a bit different, and less accessible, at least at launch. The company is calling it "Dots."
"Dots" isn't a single AI agent platform. Rather than having all OpenAI users tapping into the same "Dots" system, users create their own individual "dot" for a specific purpose. OpenAI says dots are always-on agents powered by the company's GPT-6 Astra model that learn over time based on your responses. They can connect to over 4,000 apps at launch, including ChatGPT (of course), Slack, and Teams.
According to OpenAI's pitch, you won't always have to instruct your dot to do things for you; instead, the agent will take anticipatory action, guessing what you need done based on previous experience. That might mean identifying a bug in the app you're developing and deploying a fix without being asked. One early dots tester reportedly forgot to file an invoice for work, and their dot prepared one on his behalf, only sending it after gaining his go-ahead.
Dots run on their own cloud computer. OpenAI says you can view this computer at any time to check what it's up to and what it has done. Dots can connect to other devices, including your laptop, if you grant it permission first. OpenAI says you can continue adding to its workload without worrying about slowing your bot down: It can supposedly handle many concurrent tasks. Maybe you're working on an app, and you task the dot with monitoring customer feedback and fixing small issues with code. Perhaps you're a content creator, and your dot analyzes interview transcripts for clippable segments, while drafting up social media posts.
Like other AI agents, dots are chat-based. The idea is to simulate the experience of texting a friend or colleague. You text them in ChatGPT, Slack, or Teams, and they text you back with updates and questions. They can even text you first when there's been a development. At launch, OpenAI says you can create a "primary" dot. You can give it a name and "make it your own," though it's not clear exactly what that means.
OpenAI says that dots build on ChatGPT's security systems "with additional safety and privacy safeguards." The company says all dots work on their own cloud computer, separate from your own machine—unless you choose to connect the two. The company claims that dots can sign you into accounts using saved passwords without actually exposing those passwords to the AI model. In fact, when you sign into an account, the AI model pauses working until sign-in is complete. Your dot will then resume its activity.
Dots will work in the background as part of its "proactive research" feature, but it will only connect to apps you've already permitted it to, using tools that cannot make changes. While prompt injection is always a risk with agentic AI, OpenAI says that dots have built-in safeguards to protect against malicious instructions. If the AI detects something suspicious, it will stop the dot from proceeding. In addition, you can set up custom rules to ensure the bot will never perform a task you don't want it to, such as sending emails.
Dots also operate on a system called "Auto-review," which compares upcoming steps against your instructions, rules, and safety requirements. If it detects an issue, it stops itself. The dot can then investigate further on its own or come to you with the problem. However, OpenAI acknowledges that dots can still make mistakes.
Unlike Meta, which is actively pushing Muse on all users, OpenAI's dots are not as widely accessible as of this writing. At launch, dots are only available to paying users, including ChatGPT Pro, Business Premium, and Enterprise users. What's more, you can only set up one dot right now, though OpenAI says you'll be able to set up multiple dots in the future.
This week's Out-of-Touch guide takes us from the frozen wastelands of ancient Scandinavia to Alabama courtrooms, with side quests to examine hometown nostalgia and the peculiar love Gen Z has for clowns.
The newest trend in music online is videos of Vikings rapping. Clips from fake artists with names like "Viggo Ironside" and "Ragalor" are rapidly proliferating on TikTok and other places where videos are shared. The vibe is aggressive beats and bars about Valhalla, longships, tattered flags, and wolves. The visuals are usually of muscular blonde people with tattoos in a studio doing their thing, like this:
The only other setting these Viking rappers seem to inhabit are concerts, particularly ironic given that no one will ever actually see these "artists" perform live:
The most famous Viking rapper is Draugr Balled, who was the first to really solidify the whole Viking rap genre. Balled's videos regularly earn millions of views, and you can listen to this music on legit streaming services if you want to never hang out with me.
Judging from the hundreds of comments these videos inspire ("Just wasting water," "Reposted by every 45yr old divorced dad," etc.) Kids regard Viking rap as extremely cringe, so it seems like a guilty pleasure, the kind of thing you wouldn't admit that you like, even if you secretly dig it.
For the record: Actual Vikings probably didn't have tattoos, but Viking women were generally less subjugated than women in other civilizations of the time period. Actual Vikings were also into "flyting," verse-based insult battles that seem a lot like freestyle rap battles, so there's a little historical precedent for these videos. Anyway, check out the "vikinghiphop" hashtag if you actually want to listen to more of this kind of thing. Just don't tell anyone you like it.
As I'm sure you know, a side quest in a video game is a quest that is not necessary to completing the game's main objective. In slang, a side quest is anything you do just to do it. If going to college is your quest, a side quest might be finding the underground tunnels everyone talks about, or taking an outside-your-major elective like Fencing 101 just because it looks interesting.
Big news for kids out of Alabama this week, where a court settlement may change how every kid in the U.S. uses social media. TikTok's parent company ByteDance agreed to pay Alabama over $100 million dollars to end a lawsuit claiming TikTok is designed to keep kids addicted to the platform. TikTok will also implement a two-hour daily usage limit for children, overnight and school-hour restrictions, and ban the use of cosmetic filters on kids' accounts. Underage users will also receive an interruption after 15 minutes of TikTok use, and additional pauses at 60 and 90 minutes.
While these restrictions only apply to Alabama, this is only the first state settlement of what will likely be many. There are 27 similar suits working through the court systems in other states. The likely result will be nationwide guidelines for all social media. Meta, having recently settled a federal suit, has already implemented two-hour daily time limits for young users of Facebook and Instagram and is urging it competitors to do the same. Because navigating a patchwork of state laws is not easy for tech companies, big platforms are likely to adopt the strictest state requirements as a universal standard.
I love when young people use new tools to express age-old emotions in novel ways. That's what "How I Feel in My Hometown" memes are about. You choose a fictional character, usually from an anime, to represent yourself and place 'em in scenes representing your hometown to capture that unique and melancholic sense of being a stranger in your birthplace, or having outgrown the place and seeing it with a different set of eyes. Here's an example, where Madoka Kaname contrasts with a small town:
and here's an anime character called Marin Kitagawa from My Dress-Up Darling in a Midwest town:
Another layer of irony comes from the song many memers use to accompany the video, That's My Kind of Night by Luke Bryan, an unapologetic pop-country ode to small-town life. You don't even have to know who Megumi Fushiguro from Jujutsu Kaisen is to get how he'd feel about hanging out at Texas Buc-ee's:
YouTuber BouncyClown's examination of the origin of the propeller hat "only" has about 150,000 views, so it isn't the most viral video this week, but it's a really interesting story and it's indicative of a growing trend among young people: They're really into clowns.
Clowncore is a whole thing, whether it's people doing a modern take on the "evil clown" aesthetic pioneered by the Insane Clown Posse in the last generation, kids celebrating happy clowns in general, or folks like BouncyClown who take things in a new direction by combining clown-stuff with the kind of quirky pop-culture history YouTube is known for.
To get speculative for a minute, the popularity of clowns and clowncore among young people seems like a reaction to how stultifying it is to live in a world where everyone is perfect on social media, and "cringe" is the greatest crime. It's rebellion in pancake makeup. Clowncore folks are saying, "Fuck trying to be beautiful. I'm going to make myself look as ridiculous as possible and just have fun." I'm no clown guy, but I stand behind the message 100 percent.
We may earn a commission from links on this page.
The right subscription service for the right person is the ultimate "I see you" present, the kind of gift they'll appreciate every time they log into the news service you gave them, or every time the box with a tasty meal or cool gadget arrives on their porch. Below are the best gift subscriptions for every type of person, from the snobby cinephile to the terminally online.
Bespoke Post is my pick for the best overall because it solves that perpetual holiday problem of what to get the person who's impossible to shop for. Most gift box services are aimed at specific niches, but Bespoke Post casts a wider net by letting users choose exactly what comes in their box every month. One month they can get high-end coffee supplies. The next, outdoor gear. If nothing catches their eye, they can just skip a month. If you want to give someone a present they will truly never tire of, a subscription to Bespoke Post is for you.
If you're looking for a gift for someone who is constantly doomscrolling, get them a subscription to the New York Times. Its old-school, objective reporting is the antidote to social media alarmism, and it also has good recipes and the best puzzles. An all-access subscription opens up the full New York Times archive going back to 1851, NYT Cooking, daily crosswords, and the 10,000-puzzle archive (including Wordle, Connections, Spelling Bee, and all of your other favorite time-wasters).
If that person on your list is always trying to exercise their brain, a subscription to MasterClass is a lot cheaper than grad school. This educational streaming service covers a wide range of subjects, from science to acting to business, and includes extensive lectures from heavyweights in each field. They can take a deep dive into filmmaking with Martin Scorsese, take music lessons from Timbaland, or learn cooking from Gordon Ramsay. A subscription unlocks the entire library, so they can learn as much or as little as they choose.
Libro.fm lets users listen to audiobooks while supporting their local independent bookstore, instead of supporting a huge corporation that shall go unnamed. Libro's content is free of the digital rights management of its main competitor, so users can listen to the books they own however they like instead of being locked into a proprietary player. Libro's firm anti-AI-narration stance ensures you aren't paying for robotic readers, and users can choose which independent bookstore to support through Libro's profit-sharing model.
If you're buying a gift for someone who likes to make things with their hands, the Adults and Crafts crate will keep them busy with a monthly project to work on. Each crate includes the tools, materials, and instructions for a bespoke craft object—everything from pressed flowers to a resin fruit tray or an enamel glass painting. Not only will they end up with a unique object, but each project is also a mini-lesson in crafting techniques.
If someone you know is drowning in debt from paying for DoorDash and Uber Eats, get them a sub to CookUnity. It's not one of those services that delivers ingredients and instructions; CookUnity provides restaurant-quality meals created by star chefs like Esther Choi and Marcus Samuelsson for around $11 each, and the only "cooking" required is a little microwaving.
If you're buying a gift for a film lover who thinks The Criterion Channel is too mainstream, hit them with a membership to Eternal Family, a streaming service that bills itself as the "hidden gem video store." They don't mean "hidden gems" that are actually widely known, like The Prestige or something. This is the uncut stuff, like Heino Pars' achingly beautiful 1983 stop-motion film Valley of the Honey Makers (way better than Bee Movie) or In the Dust of the Stars, a visionary science fiction film made behind the Iron Curtain in Romania in 1976. Every option is a different strange delight you likely can't watch anywhere else.
We may earn a commission from links on this page. Deal pricing and availability subject to change after time of publication.
Rosetta Stone Sapphire is on sale for $127.20 at StackSocial right now, giving new users a structured way to stick with language learning for a year. The appeal isn't that Rosetta Stone suddenly makes language learning easy—you still have to put in the time—but it does give you a structured way to practice without having to piece together lessons and vocabulary lists from different apps. The sessions are short, progress carries across your devices, and you can switch languages whenever you want.
The way Rosetta Stone teaches is a little different from apps that rely heavily on direct translation—it uses pictures, audio, written words, and context to help you figure out meaning as you go. That can feel less immediate at first, especially if you like seeing the English translation beside every phrase, but it can also help you get used to processing the language on its own. Speaking gets a fair amount of attention, too. The TruAccent tool listens to your pronunciation and gives feedback based on native-speaker examples, which is useful when you know a word but are not sure you are saying it correctly. The service works on Windows, macOS, ChromeOS, iOS, Android, and modern web browsers, with a microphone recommended for speaking exercises.
There are also guided conversation exercises focused on situations such as ordering food, introducing yourself, and getting around while traveling. You also get more than 500 flashcard sets and personalized activities covering things like travel, hobbies, business, and culture—Sapphire Studio lets you build activities around subjects you are more interested in, which can make practice feel less repetitive. That said, tutoring isn't included, so this is still mainly a self-guided course rather than a replacement for regular conversation with a teacher or native speaker. For $127, it's a great sale, just be sure to redeem the code within 30 days of buying it.
This week's Out-of-Touch Guide explains everything from a dance trend that anyone with a face can do to which social media trends are leading to arrests.Put on your face's dancing shoes and let's get going!
Face dancing is, as you might have guessed, pointing a camera at your visage and moving it in time to music; think of it like lip-syncing with extra sauce. It's not exactly new—the online version dates to 2017, and the "just messing around" version is as old as mirrors themselves—but the pastime is having a moment right now thanks to face dancing's own Twyla Tharp, @minynaranja, whose eight million TikTok followers are proof of the trend's popularity. Here's what face dancing looks like:
There are a ton of accounts dedicated to the practice, with each dancer bringing something new and distressing to the art form. Like this:
And this:
The big question is "does this take talent?" On the one side, people like X user @ElMichalito react like this to an @minynaranja live performance:
This Tweet is currently unavailable. It might be loading or has been removed.
On the other are the millions of likes and shares on face-dancing videos. It all makes me want to renounce civilization entirely and move to a cave, but I can't deny the talent.
A follow-up to last week's report of spooky Cat in the Hat memes: There have been a lot of arrests. Four juveniles in Berwick, Louisiana, face menacing and cyberbullying charges, as does a teenager in Port Allen, LA. A kid in Kentucky turned himself in to authorities. A high school student was arrested in Colorado in connection with the trend. Arrests have also been made in Florida, Alabama, Mississippi, Tennessee, and Kansas. Presumably, the arrests and charges are connected to specific threats accompanying images of the Cat in the Hat posted to social media, rather than kids just posting the meme itself. But still, if there's a young person in your life, tell them to post Gucci Morty instead. Gucci Morty just wants to vibe with you.
Teenagers poking around inside abandoned buildings (and maybe being popped for trespassing) is not new, but always being online has made "urban exploring" a more popular and public pursuit than it used to be. Kids post videos of interesting expeditions, other kids follow, the police get involved.
It's easy to understand the allure of this kind of content. Watching people uncover the secrets of an abandoned hospital or the untold history of a "NASA scientist's luxury mansion" is fascinating. But it also highlights how social media encourages young people to escalate. It used to be enough to explore a creepy local barn or something. Now you gotta explore a huge abandoned city in China or an old riverboat casino for clout. This makes it all more dangerous. Still, I hope the authorities go easy on copycats; at least they're out of the house doing something.
You ever say, "Whaaazzzzup?"' to a younger person and get a blank look in return, leaving you stammering, "...it's from these commercials, for, I think, beer. They were really popular in... never mind"? Viral animated web series Fact Attack Adventures is the reverse of that: It parodies educational platforms like BrainPOP that Gen Z and younger Millennials watched in school. That means it's intimately familiar to them, but foreign to anyone older, save maybe for parents who had to homeschool their kids during lockdown. It's not exactly a note-for-note satire, but the influence is clear.
Speaking of influence, if you want to understand a big part of what shaped the younger generations, BrainPOP is a good place to start. It was founded in 1999 by a pediatrician as a way of explaining complex medical problems to kids, but soon expanded to include lessons in math, social studies, English, and more. It spread online through the 2000s, and today it's used in nearly two-thirds of American schools. It's nearly universally beloved, too. Anyone under a certain age knows who Tim and Moby are. It's just the kind of thing that needs to be parodied. I'm glad the first generation raised entirely by computers has grown old enough to look backwards with a critical eye, yet still remain affectionate for the source material.
Anyway, Fact Attack is really funny. Maybe not that specific kind of funny that comes from really getting it, but even if it's other people's nostalgia, the conceit, which sees simple facts and lessons spiral into absurdity, is timeless. Good comedy isn't all references to current cultural events.
In among the maelstrom of everything else that's happening with AI right now, OpenAI has pushed out ChatGPT Images 2.5, which it describes as faster, sharper, and smarter, with "better tools for creating whatever you can dream of." Those new tools include templates, more precise editing features, and a way of converting your on-screen doodles into fully realized images.
This Tweet is currently unavailable. It might be loading or has been removed.
Does all of this mean ChatGPT Images 2.5 will save us from the scourge of bad AI food images on menus and posters? Can it replace a traditional photo editor? And how does it compare to the impressive Nano Banana 2 update that Google pushed out in February?
To answer all these questions and more, I fired up the new image generator in ChatGPT to see what it was capable of. Everyone is getting it—it's rolling out for all users across the web, the desktop apps, and the mobile apps now—though as usual there are varying usage limits (which ChatGPT only vaguely defines) on each plan.
The sameness and uncanny look of AI-generated food on restaurant signs and menus is becoming a real issue, so this seemed like a good place to start. It's fair to say that promotional images of food have always taken some artistic license, but AI is making the problem worse. Can ChatGPT Images 2.5 help?
Here's what I tried: "Create a 1:1 aspect ratio image with four squares inside it for a fast food restaurant display. The images should show (from top to bottom, left to right) a cheeseburger, six nuggets, a box of fries, and a Coke. Make the pictures look as photorealistic as possible with natural lighting, but add variations and imperfections to give the impression of real food."
ChatGPT is on the left and Gemini on the right. Both do a decent job of producing menu items that look realistic and actually edible, albeit with the usual AI tendency towards the generic. Personally, I'd say the ChatGPT images look slightly more like real food, though Gemini has done better on the backgrounds and wider scene (the Burger King logo would cause trouble, though).
One of the key new features in ChatGPT Images 2.5 is a new Sketch tool, which lets you turn your basic scribblings into full pictures. Click or tap the + (plus) button in the prompt box to find it—if you're on mobile you'll also have to choose Plugins from the sub-menu that pops up.
The tool lets you play around with colors, shapes, and text, and when you're done you can add a text prompt to it—something like "turn this sketch into the sort of illustration you might see in a high-quality children's book, complete with stars and distant planets" (which is the prompt I used here).
Gemini doesn't have a Sketch tool per se, though it's easy enough to add an image to a prompt for inspiration. Google's AI went far more traditional with its illustration and decided to drop in an alien astronaut, whereas ChatGPT used a touch more realism. Both do the job well enough, but if I had to pick, I think I'd take the Gemini one.
OpenAI is talking up the improved precision in editing with ChatGPT Images 2.5, so I gave it a go after submitting the following prompt: "Create a photorealistic picture of a cabin sat deep in the woods, in its own clearing, with trees all around. It's early morning and the sun shines through the foliage, while the cabin itself looks modern and welcoming."
Click on a ChatGPT image and you can add comments about specific aspects of the image, mark up sections of the picture to change, and quickly erase parts of it or the entire background. I tried getting it to remove some cabin furniture and change the time of day from early morning to dusk, and it completed the brief with aplomb—with the rearranged shadows a particularly clever touch.
Gemini gives you a simpler set of editing tools when you click on an image, but the basics are the same: Use text comments or scribbled annotations to specify the changes. Again, I got the AI to remove some cabin chairs (there must be a lot of chairs in AI model training data), and switch the scene to dusk. Gemini did fine as well, but ultimately the ChatGPT effort looked a little more lived-in and a little less artificial.
These two AI bots are very close in performance when it comes to image generation. Gemini is no slouch, and Nano Banana 2 is excellent at almost everything, but I think ChatGPT Images 2.5 has a slight edge at the moment—particularly if you add in the extra features for sketching and saving prompts as templates.
The debate over when and why to use these tools continues, and I think there will always be more value in actual human art and photography, no matter how good AI gets at making pictures. A real forest is always going to beat an approximation of one stitched together by AI. For visualizations and ideas where you wouldn't employ an artist or have the design chops to do it yourself, though, both Gemini and ChatGPT are at a higher level than ever.
It's the editing tools that are potentially the most useful at this point: Changes to pictures that would usually take professionals hours can now be done by anyone, in seconds, and the capabilities are only going to improve in the future.
Google has rolled out two new AI models: Gemini 3.8 Flash (the "intelligent workhorse model" for everyday prompting), and Gemini 3.8 Flash Cyber (available to "trusted defenders" for cybersecurity protection). While we only got Gemini 3.7 a matter of weeks ago, the new model brings with it some substantial improvements.
According to Google, reasoning and coding are better than ever with Gemini 3.8 Flash, and there are "significant" improvements across "software engineering, agentic tasks, and critical, multi-step reasoning in specialized domains"—so this is very much an update for advanced AI, rather than your basic prompting. You'll find Gemini 3.8 Flash available now inside the Gemini app if you use either the AI Pro plan ($19.99 a month) or the AI Ultra plan ($99.99 or $199.99 a month). There's nothing yet for users on the free tier or the $7.99-per-month AI Plus plan, though these models do trickle down to lower tiers over time, as upgrades are released.
You'd be forgiven for being more than a little confused by all the Gemini AI models Google has pushed out in recent months. For some time now, the models have been split into Flash-Lite, Flash, and Pro, and as you go up that scale, the AI gets slower, more capable at deeper reasoning and complex queries, and more expensive. (Not to mention, these models have different thinking levels too).
With the Pro category, we're still on Gemini 3.1 Pro that launched in February 2026. While a follow-up was mentioned at Google I/O in May, it's yet to be made public. The Flash channel is moving along much more quickly, and it was only a few months ago in December that the first Gemini Flash 3.0 was previewed.
Now we have Gemini 3.8 Flash. As per Google's benchmarks (which you should always take with a pinch of salt), it offers improvements over its predecessor when tackling "long-horizon software engineering" (so full projects rather than small tools), knowledge work (synthesizing and digging into information), complex legal workflows, financial analysis, and agentic activities. Gemini 3.8 Flash is particularly good at complex engineering problems, Google says, and has improved in multi-step reasoning across science, math, humanities, and professional fields. It'll go into more depth to figure answers out, run more checks on their accuracy, and work better at combining different skills in one response.
There are safety measures in place too: The AI has guardrails to block misuse concerning chemical, biological, radiological, nuclear, and cyber offense lines of questioning, and is better than ever at resisting prompt injection attacks (where prompts are engineered to try and get the AI to do something malicious). It all looks impressive on the surface, as do the demos Google showcases here: One of them is a fully working version of Google Maps inside an MS-DOS interface, which will particularly appeal if (like me) you started using computers back in the days of the command prompt.
While I don't have any long-horizon software projects on the go at the moment, I did put the new Gemini 3.8 Flash up against the Gemini 3.6 Flash model that's available to free users in a variety of tasks: designing an interactive Solar System animation, summarizing and simplifying a dense scientific paper, and analyzing the last week of emails in my Gmail account (one of the toughest engineering challenges I know).
To begin with, Gemini 3.8 Flash was noticeably faster: 23 seconds versus 27 seconds for building a space simulation, 15 seconds versus 20 seconds for the science summary, and more or less equal on the emails. Add up those percentages over huge projects and multiple tasks, and you can see that the new model really does get stuff done faster.
For the Solar System test, the simulations were pretty similar—perhaps a sign of how fast they've progressed that a task like this is straightforward for them now. Gemini 3.8 Flash did do a slightly better job at the quality of the renderings though, as the planets look noticeably more "real." When it came to the science summary, again both AI models tackled a tricky problem well. If anything, Gemini 3.6 Flash presented a more understandable summary—perhaps as it's a simpler model—but I found the visualizations and diagrams presented by Gemini 3.8 Flash more helpful (remember the images are created by Nano Banana 2 in both cases, but the info is provided by the underlying model).
On the email challenge, there wasn't much to choose between them: Both accurately summarized what was in my inbox and picked out the same items that needed attention (emails that weren't that important as it happens, but that's a whole other story). This is still a job I prefer to do myself, but these AI models are getting more reliable.
The speed and sophistication gains that Google has talked about are clearly there, even in small-scale tests like this—and it's notable that creating mini web apps and digging through complex physics papers are now straightforward for Flash models as well as the Pro equivalents, which is maybe something to think about if you're considering signing up for a Google AI plan. (Of course, if you want to upgrade to Gemini 3.8 Flash, you'll need to pay.)
If you had trouble accessing AI platforms Thursday morning, you're not alone. Apps and services, including ChatGPT, Claude, and Grok, were down, or experienced issues with service. ChatGPT and Grok were both unresponsive when I tried to use them, while Claude appeared to respond, at least for me.
ChatGPT appears to be experiencing the most issues, at least per Downdetector. The site, which accepts user reports of issues with apps and services, currently shows tens of thousands of reports for ChatGPT alone. Claude and Grok had fewer reports, but still many (thousands) over the course of this morning.
According to OpenAI's status page, the company had experienced issues with both ChatGPT and Codex. Anthropic had also acknowledged issues with Claude, including for its Mythos 5.1, Fable 5.1, and Opus 5 models, and said it is working on a fix. SpaceXAI shows that all Grok functions appear to be down at this time. As of 1:19 p.m. ET, all three services now report that they're now up and running.
It's not clear exactly what caused these issues, or if they were all related to one another. Typically, outages like these are solved relatively quickly, so it's not a surprise that all three were able to get their systems back by early afternoon. I'll continue to monitor the situation and update this piece accordingly.
AI agents are becoming more capable over time, and the latest ability rolled out to ChatGPT Work (that's the agentic bit of ChatGPT) is being able to sign into websites—and in theory, taking even more of your workload off you. The official examples of how you might want to use this include setting up utilities for a new apartment, organizing an appointment with a doctor, and canceling tickets for a rescheduled trip. "If it's something you’d normally have to open a browser and click through yourself, try asking ChatGPT Work to do it," says the sales pitch.
I'm all for saving time on admin, but handing over my login credentials to ChatGPT feels like a bold step. Can I trust ChatGPT to keep my details safe and secure? And what are the chances of it going haywire once it's in my accounts? In the interests of finding out, I let it log into some of my less important accounts to see what it could do.
This feature is rolling out now in the web and mobile apps for those on ChatGPT Plus, Pro, and Business accounts.
According to the ChatGPT release notes for August 25, 2026, "ChatGPT cannot see your username or password, and they are never seen by the model or used in model training. ChatGPT does not store your username or passwords." So how does it work if your usernames or passwords are never stored? The most basic option here is to simply sign in through ChatGPT when prompted by a site or app (and clear any two-factor authentication challenges at the same time), and then have the AI stay logged in for future tasks.
The more advanced version, which is where you save a more significant chunk of time, is to give ChatGPT permission to interface with a third-party password manager. ChatGPT doesn't specify a list of supported password managers, but as far as I can tell, this is just for mobile: Any password manager on your phone can pop up when ChatGPT needs it and supply the relevant credentials.
Underpinning all this is an actual browser, built into ChatGPT, that it can access and operate itself. Your accounts will see it as a new sign in (like they would if you signed in from a new device), and some websites and apps will let you mark ChatGPT as a "trusted" operator that can log in again in the future,
You can control the data held by this browser: From ChatGPT on the web, click your profile icon (bottom left), then Settings > Cloud browser. You can delete stored cookies from here, for example, which will effectively kick the AI out of any sites it has access to. As per ChatGPT support: "You stay in control of which websites ChatGPT Work can access, and it will always ask for confirmation before consequential actions, such as completing a reservation or payment." Exactly how it can tell when it's about to do something consequential, however, isn't clear.
If you want to make use of this inside ChatGPT on the web, click Work at the top of the screen to switch to agent mode, then tell the AI what you want to do: Something like "sign into my eBay account and tell me how my auctions are doing" or "sign into my accounting app and submit these invoices."
I gave the eBay one a go first—the process proved to be incredibly slow for whatever reason, and I could've signed in multiple times in the normal way by the time ChatGPT got around to gaining access. There are also problems where the ChatGPT conversation tends to obfuscate what's going on: It asked for a verification code, and I didn't know where to check for it.
You can, at any point, switch to the ChatGPT browser itself, which I eventually did to get through the verification code stage (it was sent over text). The AI was then able to correctly tell me what the current auction prices were on the items I was bidding on, after initially stumbling through several wrong pages—there's a thumbnail of the live browser in the corner of the screen, so you can see what's happening.
Next up, I tasked ChatGPT with getting into my Airbnb account and finding some great spots close to Santa Cruz, Calif., within a certain price range. Again, I had to switch to the actual browser view to sign in (I was told the Airbnb site wasn't cooperating), but once I had given the AI access, it worked smoothly enough, looking up properties and giving me a rundown of prices.
At the moment, the sign-in flows are just too finicky. Sites seem suspicious of the AI browser, everything is really slow, and you run into complications with verification codes, passkeys, and the like. You'll often have to manually sign in as you would anyway in order to get ChatGPT up and running on a site.
Once that's done, though, everything works reasonably impressively, as long as you're giving commands that are fairly easy to interpret and carry out. ChatGPT on Airbnb seems competent with dates and locations, but I'm not sure it could filter through places based on the view or the general vibe of the decor.
Perhaps most importantly, I still don't trust the AI to do these jobs better than I can, even if it can (eventually) do it more quickly. I want to see for myself what an Airbnb looks like, or who's bidding on an eBay auction, or what the available time slots are for an appointment to get my car serviced—if I left it to ChatGPT, I'd be too worried about it missing something (I have the same reservations about letting friends and family do things for me as well, but that's another issue).
In my testing, ChatGPT didn't fail in any of the tasks I gave it once it had gained access to these sites, or give me any inaccurate answers, but that would always be at the back of my mind if I switched over to this completely. For now, I'm going to stick to these online admin tasks myself.
Recently, Anthropic announced that its Claude models will soon begin marking its text output with invisible watermarks. At the same time, Google revealed that its own visible Gemini watermark on generated images will now be optional. These are just two contradictory examples of the wildly varied methods of watermarking AI output. So, let’s break down how AI watermarks work, and how you can detect them even when they're "invisible."
The exact method for watermarking AI output will depend on the type of media being generated, but the principle is generally the same: The generated media is embedded with data that is undetectable (or occasionally detectable) to the human eye or ear, but which identifies it as AI-produced. Watermark-detecting tools can then be used to identify if something was made with AI, without guessing or using an unreliable AI “detector”. Here are some examples of how watermarks may be implemented for various media types.
For images: Since the pixels of an image are mathematical values, they can be slightly modified to embed a digital signature, without perceptible changes to the image itself. Google’s SynthID, for example, distributes an invisible signature across any generated image, so even a cropped version will still contain detectable portions of the watermark. Note: This is distinct from the visible gray symbol in the corner of Gemini images. Even if you turn off the visible Gemini watermark, the invisible one remains.
For audio: Embedding a watermark in audio files can be even easier, placing signature sounds outside the range of human hearing (typically below 20Hz or above 20,000Hz). SynthID has an audio component that can be heard by watermark detectors, but remains imperceptible to the ear.
For video: Naturally, generated video tends to use a combination of both of the above watermarking methods, though it’s worth keeping in mind that someone making fake content could, for example, generate AI audio to accompany real video, meaning the watermark might appear in one piece of the content, but not another.
For text: Every next word an LLM generates comes with a probability score. The sentence “The cat is” could end with “fluffy,” “cute,” or “small.” Each one of those words is given a percentage likelihood that it will appear. Text watermarks work by inflating the chances that certain sets of random words will appear. This can make it possible to detect LLM-generated text without changing its semantic meaning. However, this tends to work better for longer pieces of text, which provides more chances to detect the presence of less-likely words.
For metadata: While not strictly a watermark, C2PA is a framework for adding metadata that can help verify the origin of a piece of media. Some camera manufacturers, for example, have implemented C2PA to give photographers a traceable record of where an image came from. This can also include noting whether an image was generated with AI tools.
While SynthID originated with Google’s DeepMind, the company open-sourced the protocol, and now it’s also used by other AI companies, including OpenAI. Some companies. conversely, do not use a watermark at all—and among those that do, the implementation can be inconsistent between tools. In other words, the presence of an AI watermark can confirm something was made or modified with AI, but the absence of one can’t prove it wasn’t.
Despite SynthID and C2PA being relatively common, it’s still kind of a crapshoot to properly detect the presence of watermarks or metadata that will confirm if a piece of media was generated with AI. OpenAI has a standalone tool to check for SynthID or C2PA in a file, and Google lets you access its verification tools via Gemini or, with the right prompting, directly via Google itself.
However, these tools have limitations. OpenAI’s detector, for some reason, only seems to detect media generated by OpenAI itself. During my testing, I tried uploading images generated via Gemini—which could detect its own SynthID watermark—and OpenAI’s tool did not find it.
Meanwhile, Google created a SynthID Detector portal, but it’s currently available on an invite-only basis, specifically for journalists and verification professionals. You can still access some SynthID detection functions via Gemini or Google, though in some cases, you’ll need the right prompts to do so. In my testing, when asking “is this real” for a known AI-generated image via Gemini, the tool invoked a verification tool to check for SynthID. However, when doing the same process via Google, the LLM’s output resembled a visual analysis instead. It only invoked a SynthID check when specifically asked to do so.
Even if you do your diligence to check for every version of a SynthID watermark or C2PA metadata, it’s still possible that a piece of media could have some other form of watermark that requires a different detector. Unfortunately, unless you have a strong indicator of which tool was used to create a piece of generated media, it can still be difficult to thoroughly check for every kind of watermark.
When it comes to text-based watermarks like the kind Claude or even Gemini use, detecting them can still be very difficult. For starters, neither offers a public way to check for the text watermark just yet (Google's SynthID Detector portal can do so, but it's not generally available).
With enough work, any watermark can technically be removed, though SynthID specifically is pretty resistant to most typical forms of modification. An AI-generated image with a SynthID watermark that has been cropped, filtered, or modified can still retain enough of its original watermark to be detectable. It’s not impossible to remove, but it’s generally hard to do so accidentally.
Removing metadata like C2PA, on the other hand, is considerably easier. For images, that’s as simple as taking a screenshot. A screenshot of an image essentially creates a new image file based on the pixels visible on the screen. This means that any watermark that affects those pixels can remain, but an entirely new set of metadata is created, wiping any C2PA data along with it.
That means that if you want to maintain a metadata chain that lets you prove the authenticity of an image, it’s important to download or upload the specific original files and make sure any editing tools you use support maintaining that metadata.
Text watermarks are among the easiest to get around. Since they work by simply altering the probability that certain words will appear, running text through another AI tool that rephrases the words without using a watermark, or even manually rewriting a block of text, can potentially remove the watermark.
It’s important to keep in mind that a watermark can be added to an authentic piece of media. If someone uploads an authentic photo to a tool like Gemini to perform simple edits, the output image will have a SynthID watermark, too. This doesn’t mean the whole image is inauthentic, but it will still be flagged by watermark detectors.
Similarly, it’s possible to generate an image of a subject, cut the subject out, and add it to an image using traditional manipulation techniques like Photoshop to create an inauthentic image that’s mostly made from an authentic image. Whether or not the watermark will remain on the portion of an image that was AI-generated can only be determined on a case-by-case basis.
As mentioned before, the absence of a watermark cannot prove that an image is authentic. Even the presence of a watermark can’t prove that the substance of an image isn’t real. Watermarks and metadata are simply tools to help you figure out where a piece of media likely came from and how it might’ve been modified. Ultimately, it’s still up to you to verify the things you see and hear online.
This week's trip into the culture of young people is filled with bad ideas and stupidity, but it's that specific kind of stupidity that's the sole purview of youth—unserious, frequently baffling, but ultimately harmless. Old people are just as stupid, but our stupidity has consequences. Debating whether ditching basic hygiene to be more attractive is dumb as dirt, but it's not going to crash the housing market, and no one really dies in an aura battle. So here's what's going on with the young people and what it all means.
To Gens Z and A, "aura" is, essentially, coolness or charisma ("rizz" if you prefer). An "aura battle" is a public contest in which the person with the most aura, as determined by crowd and/or comment section reaction, is the winner. It's a little like a breakdance or rap battle, but without the talent.
The trend seems to have originated in Latin America in the beginning of 2026 and quickly spread worldwide. It's recently come to the U.S. Here's Mr. Poo Poo and Latavious Quavious squaring off this past weekend in NYC at what was billed as the "first-ever American aura championship."
I've been watching Aura Battles for the last hour or so, just to see if I can determine what adds up to "aura," and I got nothing—it seems like most of the "battle" is doing bad versions of meme dances. But I'm not dismissing aura battles; the first breakdance battles were probably pretty rough too, and anything that gets kids outside and hanging out with each other is cool.
It's unlikely, but aura battles could develop into an organic artistic expression of the soul of the younger generation. If so, the video below is history being made, the equivalent of the first time Cool Herc set up two turntables to play break beats. Ladies and gentlemen, here is the final round of the first-ever American aura battle championship.
(I feel like Applewood got robbed, tbh.)
Police departments, school districts, and other authority figures from Baltimore to Donegal are sounding the alarm over a terrifying new trend of...kids posting memes about The Cat in the Hat.
The often AI-generated videos cropping up on TikTok and the like feature someone dressed as the famous cartoon cat and cryptic warnings. They sometimes include scary music. This has led to online rumors that there's a Cat in the Hat serial killer on the loose. There isn't, but the trend did lead to an incident that's at least a little real. In Constantine, Mich., a user called "catinthehat" reportedly issued threats against specific students, prompting state police and the FBI to get involved.
Still, authorities all over the world locking down schools, stepping up police patrols, and threatening to arrest people over a meme seems more than a little silly. I get that school shootings are scary, but maybe shrugging at the easy availability of firearms and freaking out over Cat in the Hat memes isn't the best approach to stopping them?
Completing this week's triptych of idiocy is "pheromone maxxing." -Maxxing, as I'm sure you know, is doing something a lot. Pheromones are chemicals that some animals release that influence the behavior of other members of their species, so pheromone maxxing is trying to influence the behaviors of others through your stink. Supposedly, young men in the "looksmaxxing" community are skipping showers and not washing their clothing in an attempt to make young women like them more. It's difficult to determine how widespread the belief is, but there are at least a few would-be influencers unironically advocating for the practice, although the vast majority's online response is basically, "ew."
So is there anything to it? While humans do produce chemical odors in response to emotional states, they aren't actually "pheromones" like in the animal world, and the number of smelly, single men at comic book conventions provides strong corollary evidence that not showering is a bad strategy for attracting potential mates. On the other hand, there's this guy's TikTok, where he reports he stopped using cologne and started getting more and more compliments. I believe him, because can you imagine how horrible his cologne was?
Lesbian Space Princess is a (barely) animated 2025 movie from Australia. Written and directed by Emma Hough Hobbs and Leela Varghese, and according to some online, it's bringing about world peace.
LSP tells the story of Princess Saira from the planet Clitopolis and her battle with the Straight White Maliens. Even though it has a critical rating of 98% on Rotten Tomatoes, it wasn't well-known outside of Australia. But recently, people started posting clips of the movie online. Because everyone (except critics, apparently) thinks it's bad.
Right-wing people hate-post LSP as an example of "overly woke" media; lefties hate-post it because it's so on-the-nose and terminally online that it's painfully unfunny. Fans of Hazbin Hotel post about LSP because it's way worse than their favorite franchise. So do Marvel fanboys. Before long, Memes like this—
—started popping up featuring the phrase "Just this Once," as fandoms realized that they at least share this one thing in common. The Great Truce of 2026 is bringing Naruto and One Piece fans together. Football fans are coming together with futbol fans. It's so widespread, the belief is growing that Lesbian Space Princess badness is so complete that it's bringing about world peace.
We may earn a commission from links on this page.
It’s been more than 15 years since Lost went off the air, ending a legendary run on broadcast television. Though the show could be frustratingly slow to reveal its secrets, it remains one of the best TV shows of the modern era—and one of the most influential, with a host of other series playing off a similar “mystery box” premise and shuffled timeline.
Fans of the show have never stopped looking for a viewing experience that matches it in terms of surprise and heartfelt, emotionally rich storytelling (if that describes you, you’ve probably already watched all of these streamalikes offering similar vibes), so allow me to suggest the one movie that fits the bill perfectly: Jacob’s Ladder, a 1990 psychological thriller starring Tim Robbins.
On the surface, Jacob’s Ladder might not seem like an immediate match for the heady sci-fi-tinged, character-based storytelling of Lost—but once you dig deeper, the narrative and tonal parallels become clear. In the film, Jacob Singer (Robbins) is a struggling Vietnam Veteran living in Brooklyn in 1975, four years after a disturbing incident with his unit that ended with Jacob being stabbed by a bayonet. Haunted by his dead son and his wartime experiences, Jacob’s descends into a living nightmare, filled with hallucinations of monstrous figures. But something most mysterious is happening, and while the ending is a clear antecedent to Lost’s infamous finale, there are many visual echoes as well.
Jacob’s Ladder is arguably a horror movie, and is certainly more frightening and disturbing than Lost, which was more philosophical and even optimistic in its approach. But the way Jacob’s reality is plagued by a mounting sense of mystery and dread will remind you a lot of the early seasons of Lost, when the Others and the Hatch were ominous mirages hiding in the shadows of that mysterious island. And the film's ultimate payoff—and its message that sometimes, the devils we perceive to be attacking us are actually angels trying to save us—will resonate with anyone who was all-in on Jack Shephard’s inner struggles (and Lost’s overarching themes of atonement and redemption).
Finally, there’s the non-linear way Jacob’s Ladder deals with time, which echoes the flashbacks, flash-forwards, and flash-sideways moments in Lost. As Jacob’s life dissolves into an increasingly hellish chaos, his experience of time becomes shuffled and blurred, with the past (and alternate futures) intruding on the present in disorienting ways. Just like in the show, these flashes of Jacob’s life slowly paint a picture of his true present, and his ultimate fate. It’s an intense and often frightening film, but fans of Lost will love it. Stream Jacob’s Ladder on Fubo or rent it on Prime Video.
Need more mysterious islands, heavy sci-fi concepts, and stories of survival against supernatural forces? Here are a few more movies for Lost fans.
Deserted island? Check. Mysterious monster? Check. Flawed people hiding secrets? Check. When her boat sinks, Jenn (Kiersey Clemons) washes ashore on an island where the only sign of humanity is a set of old graves. At night, she’s attacked by a terrifying sea monster. When some of her friends wash up on the beach, having also survived the disaster, they don’t believe Jenn's warnings—and shebegins to wonder if she can trust them. If you loved the survival aspect of Lost’s story combined with the complex (and often hidden) relationships between the characters, Sweetheart will go down easy. Rent Sweetheart on Prime Video.
If you want more stories about a group of strangers who find themselves in an isolated and ominous wilderness, with a puzzle box mystery that has to be pieced together, this 2014 adaptation of the novel by James Dashner is the ticket. Waking up without any memory, teenage Thomas finds himself in the Glade, where the only possible escape is through a dense maze surrounding it. The maze is patrolled by dangerous biomechanical creatures called Grievers. Determined to escape, Thomas and a group of other imprisoned teens must acquire clues to the maze’s operation and eventually learn what’s really going on. Like on Lost, it involves a mysterious organization, seemingly abandoned facilities, and a mounting series of twists. Rent The Maze Runner on Prime Video.
Part of what hooked people on Lost was just the basic premise—a group of strangers crash-land on an island and have to figure out how to survive. The Beach isn’t a mystery, but it leans into those very human dynamics as a group of young travelers (led by a baby Leonardo DiCaprio) find their way to a pristine, isolated beach where a group of people has established an ersatz community focused on keeping the idyllic place a secret. The "Others" here are the drug dealers who tolerate the group near their farms—for now—while the biggest risk may be the slow breakdown of order among the islanders themselves. If your favorite parts of Lost involved the factions forming among the survivors, the Tailies, and the Others, watch this movie. Rent The Beach on Prime Video.
https://www.amazon.com/Beach-Leonardo-DiCaprio/dp/B000I9X6BS
Lost probably sucked you in with the hints of mystery amid the human drama generated by a group of plane crash survivors with a lot of secrets, but it slowly evolved into a sci-fi show involving time travel and a lot of other speculative stuff. If you want to break your brain with a complex time travel story that pivots off of flawed humanity and emotionally-driven bad decisions, watch Primer: When two engineers stumble upon a limited way to travel backward in time, they begin experimenting—but when, exactly, the experiments started, and which version of the character you’re watching at any given time, soon becomes a complex puzzle (seriously, there are charts online trying to figure it all out). Rent Primer on Prime Video.
The Village got some hate when it was released, but if you love the eerie atmosphere and shocking reveals of Lost, this M. Night Shyamalan joint is a great match. In an isolated 19th-century village, residents keep strict watch for unnamed monsters that lurk in the surrounding woods. When one of the villagers is grievously wounded, blind Ivy (Bryce Dallas Howard) enters the woods on a quest to travel to the wicked, forbidden “towns” to retrieve medicine. A series of reveals lets the audience in on the secrets of the insular community in classic Shyamalan fashion, and the origins of the settlement echo the founding of the DHARMA Initiative in ways that Lost fans will appreciate. Rent The Village on Prime Video.
We may earn a commission from links on this page.
Sugar on Apple TV is remarkable in a lot of ways. One part old-school throwback to classic noir (complete with literal flashes of classic films), one part unexpected sci-fi twister, and one part Colin Farrell’s charisma, it’s a show that defies easy categorization while being very easy to simply enjoy.
The big twist was divisive, but worked exceptionally well because it wasn’t just a twist for shock value—it recontextualized the story and John Sugar’s character in exciting ways that honored its noir roots. If you want more stories of deadpan detectives navigating a grimy world of secrets, mysteries, and violence (with a few surprises thrown in), there are plenty of shows to check out. If you’ve been there and done that, there’s one movie that all fans of Sugar should probably watch: 1955’s Kiss Me Deadly.
You might recognize this film even if you’ve never seen it: It’s one of the old movies cut into Sugar to show John’s inner monologue and the films he’s watched to understand human nature. It shows up twice in season one, and the reasons why are clear: The main character, detective Mike Hammer (Ralph Meeker) drives around L.A. in a cool Corvette just like John Sugar; the story is a perfect tonal match for Sugar; and there’s a big-swing, wild twist at the end that sets it apart from most other classic noirs from the middle of the 20th century.
Kiss Me Deadly was adapted from Mickey Spillane's novel of the same title, and is one of the most influential noir films ever produced. Mike Hammer is the quintessential hard-nosed private detective who has a deep sense of ethics. Cynical and happy to solve his problems with violence, he can’t resist pursuing true justice, so when a woman named Christina (Cloris Leachman) wanders in front of his car while wearing nothing but a coat, he offers her a ride. And when they’re assaulted by goons and Christina is violently murdered, he wakes up in the hospital determined to find out what’s behind her death.
Hammer is explicitly the precise type of detective that John Sugar wants to emulate, and Kiss Me Deadly is a story that straddles both the noir and sci-fi genres, just like the show (it’s actually got an entry in The Encyclopedia of Science Fiction). The visual style and bleak tone in the film have been referenced, repurposed, and outright stolen in hundreds of films since. If you love the noir atmosphere and shocking reveal in Sugar, Kiss Me Deadly is the perfect film to watch. It’s actually kind of the perfect film to watch in general. Stream Kiss Me Deadly on JustWatch or at The Internet Archive (Note: This is a colorized version).
Need more noir? More twists? More aliens? We got you. Whether you love the hard-boiled mystery, the unexpected twists, or gorgeous visual sense of the series, here are some other films fans of Sugar will love.
Based on James Ellroy’s novel, L.A. Confidential is a perfect choice if you want more stylized, modern noir like Sugar. While it doesn’t have the genre bending of the series, the story is dark and twisty enough to keep your brain occupied throughout its running time. Set in early 1950s Los Angeles, it’s a story of a fading dream as a few cops try to navigate a corrupt police department and city government with their souls intact, and only partially succeed. Rent L.A. Confidential on Prime Video.
One of the greatest films ever made (and another movie referenced directly in the series), The Third Man is set far away from 1950s L.A., instead telling a story set in post-World War II Vienna. An American author, Holly Martins, arrives after receiving a job offer from old friend Harry Lime, only to learn that Lime was killed in a car accident. While Martins is no John Sugar—he’s not the sharpest investigator in the world—he stumbles into a complex mystery involving a (you guessed it) mysterious third man spotted at the scene of Lime’s death. If you love the vibe of ominous mystery that suffuses Sugar, you owe it to yourself to watch this. Stream The Third Man on Kanopy, Plex, or Tubi, or rent it on Prime Video.
Director Robert Altman and screenwriter Leigh Brackett’s adaptation of Raymond Chandler’s novel is set in the early 1970s, but gumshoe Philip Marlowe (Elliot Gould in a triumph of unexpected casting) seems trapped in an earlier era, driving a classic car and chain smoking his way through the twisty story, making his detective as much a fish out of water in L.A. as John Sugar. The story is an unpredictable masterpiece of misdirection, and the vibe is spot-on for fans of the show. Stream The Long Goodbye on Hoopla or Tubi, or rent it on Prime Video.
If you love how Sugar reimagines the tropes of noir mysteries, check out Rian Johnson’s debut film, Brick, which applies those same tropes to a story set in a suburban high school. When teenager Brendan (Joseph Gordon-Levitt) gets a worrying call from his ex-girlfriend shortly before her death, he launches a classic investigation that uncovers a swamp of crime, corruption, and betrayal. Marked by tight, crime novel-flavored dialogue and a compelling mystery, Brick is the unexpected noir treat fans of Sugar will love. Rent Brick on Prime Video.
If you’re craving a story about a classic noir detective character that has a speculative twist like John Sugar’s secret, check out Angel Heart. In New York City in 1955, private investigator Harry Angel (Mickey Rourke) is hired by the mysterious Louis Cyphre (Robert De Niro) to locate a missing man named Johnny Favorite. Harry reluctantly sinks into a quagmire of lies, violence, and self-loathing as he traces Johnny’s movements—and where it all leads is both a shock and a genre twist, with a powerful final shot you won’t soon forget. Stream Angel Heart on Kanopy, Hoopla, or YouTube, or rent it on Prime Video.
This week's crop of things-that-only-matter-to-young people includes Gucci Morty and Chanty Beluga, two unrelated but incredibly popular meme characters, stolen source code for the biggest video game in the world, and a viral YouTube video that cost more than your house. So get ready to learn what's what and who's who among the younger generations.
Denizens of the internet create countless hundreds of thousands of memes every day, but only a tiny percentage ever gain any traction, and only one in a million really takes off. Gucci Morty is that meme this week. The image is of Morty from Rick and Morty wearing a black Gucci shirt and looking nonchalant ... and that's it. That's the whole thing.
I wouldn't have looked twice, but millions of younger people connected with the meme. Since he showed up about a week ago, Gucci Morty spread as fast as smallpox, from Instagram to TikTok to X to everywhere young people spread memes. Variations and tributes appeared, accompanied by part of "Lot of Me" by the rapper Lil Tecca, the song that gave Gucci Morty his catch phrase, "I can see you wanna vibe with me." Here's an example:
People started dressing as Gucci Morty:
It became a whole thing, with anyone under a certain age knowing Gucci Morty. But what does it mean? Why this meme? It's mostly ironic, with commenters noting Gucci Morty's endless aura, and memers putting him in "cool" situations. But the viral spread of Gucci Morty has become part of the joke, which has created a meme feedback loop, where people are posting Gucci Morty because there are so many memes of him, or posting videos where they comment on his all-of-a-sudden-everywhere nature:
Commenters seem to agree that Gucci Morty will vanish as quickly as he appeared, so get your fill now, before he becomes a "remember that one meme ..." in a few weeks.
Gucci Morty isn't the only inexplicable viral star this week. There's also Chanty Beluga. A character created by comedian Courtney O'Donnell, Chanty Beluga is a friendly woman who likes to say "Hi, I'm Chanty Beluga" while she's doing things like ordering food, getting a coffee, or making a doctor's appointment. The videos are generally about five seconds long and get millions of views. Her popularity seems to come from a couple sources: 1) "Chanty Beluga" is a funny name. 2) The brief interactions include suggestions of a larger, stranger life for Chanty, which is never explored. Chanty Beluga buys a portable cryo-chamber on Facebook Marketplace. She's served with legal papers. She's "reading for the part of Crypt Keeper." It's not much to account for all the views, but some things just hit for people and you gotta shrug.
There is no more eagerly awaited video game than Rockstar's Grand Theft Auto VI. Scheduled for release on November 19, 2026, Grand Theft Auto VI is is a game just about every gamer wants to play—but a few people are actually doing it. Mysterious hacker group Cyberleek has somehow gotten hold of a playable build of the game and has uploaded at least 10 gameplay videos to a Telegram Channel and a website. From there, they quickly spread everywhere, giving everyone an illicit peak of what to expect when this hits the market. But this is no viral marketing gimmick. Rockstar Games and its parent company Take-Two Interactive are furiously issuing subpoenas to Discord, Microsoft, and hosting providers to pull IP addresses and accounts linked to Cyberleek, nuking the videos when they appear. Right now, Cyberleek's site and Telegram are down, but I wouldn't be surprised if that does not stop them from posting more videos in the coming day. Cyberleek's drops are often couched in a message decrying digital-only game releases, but the real motivation isn't altruism, it's a crypto-currency scheme. Cyberleek has published a poll on its website where people can vote on what to show next through purchasing "CyberLeek" memecoin. As yet, there's no information on how, exactly, this group got its hands on GTA VI source code, leading to theories that a disgruntled employee made off with the code, or that some kind of phishing or other attack was wildly successful. The leaks are not expected to delay or alter "Grand Theft Auto 6: An Extended Look," a special streaming on Netflix beginning on August 27, nor should it affect the release date.
"Misery girl" is slang for a young women with piercings, a penchant for the darkly dramatic, and highly stylized makeup that makes them appear to have a skinny, pointy nose and huge eyes and lips. Here's the kind of look a Misery Girl sports:
And here's a makeup tutorial if you want to try the look yourself:
The Misery Girl gets her name from the fact that she supposedly likes the song "Misery" by Pupsies, which you can listen to below (if you dare).
Memes about Misery Girls frequently include the phrase "Dada, play that Misery," a reference to ageplay, which misery girls are supposed to be into. They often include pictures of X user @himonkeyhamie, who I'm sure deserves none of this. Misery Girl is not a self-selected nomenclature like "goth" or "emo," and there don't seem to be a ton of people rocking the look or lifestyle. Ultimately, it's a term used almost exclusively as a slur; as many have pointed out, it's all basically an excuse weirdos use to bully altgirls.
Back in 2017, unknown YouTuber Jimmy "MrBeast" Donaldson went viral for the first time by sitting in his room counting 100,000. That no-budget webcam stunt was enough to get millions of views on YouTube back then. This week's viral video from MrBeast has nearly 45 million views after only three days online, and it proves that you really have to up the production if you expect to go viral with a stunt video in 2026. Escape 100 Cops, Win $500,000 is among the most elaborate video ever shot for YouTube. MrBeast hired 100 off-duty Indiana State Police troopers to try to catch him and four of his friends as they hid out in a simulated town rigged with tunnels, explosives, secret rooms and more. Anyone who avoids capture for 12 hours gets a $100,000. The police department gets $500,000 if everyone is apprehended. The video required a huge production crew, legal department for the permits, explosive experts, post production department, and more. Judging from what MrBeast has said in the past about his more elaborate videos, the budget was comfortably in the multimillion dollar range.
We may earn a commission from links on this page.
Fans of bad vibes and cynicism found their perfect series in Black Mirror, Charlie Brooker’s modern take on the Twilight Zone. The show’s focus on technology and all the ways it can ruin and terrify (and, very occasionally, save and sustain) us has shown us the darkest sides of social media, memory retention, virtual worlds, and dating algorithms, usually to chilling effect. If the show’s long breaks between its short seasons have you craving more dystopia and tech-fueled psychological horror (and you've watched all the similar series out there) maybe what you need is more long-form desolation. If that’s you, there’s one movie that fits the Black Mirror vibe perfectly: Moon, starring Sam Rockwell.
This mind-bending sci-fi movie starts off grim and just gets grimmer, pivoting off the same paranoia and distrust of technology as Black Mirror. Sam Bell (Sam Rockwell) is the sole inhabitant of a largely automated helium-3 mining facility on the dark side of the Moon. With only the AI named GERTY (played by Kevin Spacey) for company, Sam believes he’s nearing the end of a three-year contract and will soon return to his family on Earth. When he discovers his own badly-injured body at a crashed rover site outside the facility, however, his reality starts dissolving as he realizes the truth: He’s ultimately just an asset owned by a corporation, and he’s never going home.
What makes Moon the perfect film for Black Mirror fans is the way it both explores the dark side of a scientific breakthrough (easy, robust cloning) and also just plain messes with your head. Sam’s initial contentment gives way to puzzlement, and then to angry horror as he figures out what’s going on. The story exposes how technology that should improve our lives can be leveraged against us to rob us of agency and exploit us in ways we can’t currently imagine.
At the same time, like the best Black Mirror episodes (“San Junipero,” for example), Moon ultimately pivots to an ending that simultaneously surprises and reinforces how resilient and unstoppable the human spirit really is. Technology, the film argues, might be nightmarish when misused, but being human means we will always find a way to wriggle free of its grasp and find something close to justice. Add in a typically amazing performance from Rockwell, and you’ve got the perfect movie to watch after you’ve exhausted your supply of Black Mirror episodes. You can stream Moon on Pluto TV or rent it on Prime Video.
If you’re not totally depressed yet, maybe you just need more tech-adverse, brain-twisting movies. We got you—here are a few more films that fans of Black Mirror will love.
From the delightfully demented Yorgos Lanthimos (and co-writer Efthimis Filippou), The Lobster’s logline reads like a Black Mirror episode summary: After his wife leaves him, a man named David (Colin Farrell) has 45 days to find a new romantic partner or he will be transformed into the animal of his choice. David knows full well how it works, because his own brother, Bob, accompanies him in his new form as a dog. Believe it or not, it gets weirder (and wayyyyy more violent) after that. It definitely feels like a Black Mirror episode exploring the social constructs around relationships and dating. Rent The Lobster on Prime Video.
A young couple (Jesse Eisenberg and Imogen Poots) drive to a housing development called Yonder where all the houses are eerily similar, and find themselves trapped—no matter how hard they try, their attempts to leave always lead them back to “their” house. They’re soon ordered to raise an infant boy in order to earn their release. It’s a sci-fi/horror hybrid that’s as bleak and terrifying as any Black Mirror episode, and it shares the show’s general antipathy toward happy endings. Stream Vivarium on Kanopy or rent it on Prime Video.
If your favorite Black Mirror episodes are the ones that explore modern relationships through the dark lens of sci-fi technology (like season four’s “Hang the DJ”), check out The One I Love. A struggling couple (Elizabeth Moss and Mark Duplass) go to a remote estate in an attempt to reconnect—and discover that duplicates are living in the guest cottage. The duplicates are initially better at being their partner than the originals, and the couple attempts to set up “rules” allowing them to spend time with their doppelgängers—but, of course, there’s a darker twist lingering under the surface. Stream The One I Love on Kanopy on or rent it on Prime Video.
Black Mirror often delves into existential questions around the true nature of the bonds between us—for example, in the episode “Be Right Back,” when a young woman played by Hayley Atwell brings her dead boyfriend back as an AI android that uses videos and social media posts to reconstruct his personality. Swan Song plays with similar themes from the other direction: When Cameron (Mahershala Ali) is diagnosed with a terminal illness, he’s offered the option of having a clone seamlessly replace him, sparing his family the grief of his passing. It’s a kinder, gentler sort of story than Black Mirror typically offers, but it’s just as devastating. Stream Swan Song on Apple TV.
There might not be a film premise that’s more Black Mirror-ish—a service that erases bad memories, with all the unintended consequences a viewer could hope for (echoing the “grain” technology the show features in episodes like “The Entire History of You”). When Joel (Jim Carrey) learns that his ex-girlfriend, Clementine (Kate Winslet), has had her memories of him erased, he decides to do the same. But as he undergoes the procedure and dreams through his memories of her, he changes his mind and attempts to “hide” Clementine in his other memories, leading to some truly haunting and terrific set pieces directed by Michel Gondry. Rent Eternal Sunshine of the Spotless Mind on Prime Video.
Installing the ChatGPT desktop app gives you a few extras that aren't available through the web portal, including the ability to run prompts that query your local files, emails, and screenshots. Now, with the macOS client, those additional features extend to your iMessages, too: There's now an Apple Messages plug-in available.
The idea of giving ChatGPT access to your chats is something a lot of users aren't ever going to countenance—"good luck with that spy software" is one typical response on Reddit—but we know that plenty of users are already asking AI for help in composing and interpreting messages. This just makes that process more streamlined. (Claude has an Apple Messages plug-in too, for what it's worth.)
As OpenAI confirmed to TechCrunch, the plug-in does all its processing on your Mac, and doesn't send anything about your chats back to servers in the cloud—these Messages conversations won't appear in your ChatGPT apps on the web or on mobile. In addition, the ChatGPT desktop app doesn't build a full index of your text messages.
This is something that's available for all ChatGPT users who have a Mac: To get started, download and install the desktop app for macOS, then open the Plugins tab, search for "messages," and click Install next to the Apple Messages entry. You can then use natural language in your prompts to invoke the plug-in ("send a message to..."), use the @Messages mention, or select Messages from the plug-in list in the prompt box (click the + icon to find it).
Once you've set it up, ChatGPT can then do just about anything you can imagine with your messages, starting with the basics: It'll send messages for you, for example, and even compose them if you're stuck for inspiration. You just have to name the contact, so something like "message Steve telling him I'm sorry I can't come to his dinner party in a polite way" will do exactly that.
When I tried this, ChatGPT whipped up a suitably friendly message, with the right blend of thank yous, sorrys, and hopes for a lovely evening. When sending messages, you get the option to review what's being sent and approve it before it goes out—you can approve messages each time, or have them sent automatically in the conversations you select.
Then there's everything the AI is able to do with your message history. You can use prompts like "who's messaged me the most this week?" or "is there anything I need to follow up on in my conversations?" and have ChatGPT take a look.
I tried getting ChatGPT to dig back into my message history for details of event arrangements and deliveries, and it picked out the right information each time. I also got it to look back over text messages from my phone carrier to see if my bill had gone up over time, which it had: The AI gave me the exact increase, and the percentage increase, together with the relevant dates.
Conversations can continue based on whatever you find as well, so if you discover there's a chat thread with an invite that you haven't responded to, ChatGPT can sort that for you—and may even draft a suitable reply you can approve with a couple of clicks, depending on the context.
It's in this message history retrieval and analysis where the ChatGPT plug-in seems most useful to me. I'm reluctant to start farming out message composing and sending duties to AI—it seems rude, first and foremost, not to mention bad for my brain—but searching through reams of text is something that AI does well, and could genuinely end up saving me some time.
If you're prepared to trust ChatGPT with access to your Apple Messages chat list (and that's a big "if") then it's another database for the AI to mine and analyze. As with a lot of these other plug-ins, like Claude and Gmail, it's a question of where you're going to draw the line between privacy and convenience.
And there are certainly risks to watch out for: I can see this being yet another way that people become over-reliant on AI and its "thoughts." If you want ChatGPT to assess someone's mood over time based on their chat history, or compose a message that gives you the best chance of rekindling a relationship, or write a text getting you out of work based on a fake excuse that sounds genuine, ChatGPT will do it. That's where the tech veers beyond the useful and straight into the uncomfortable.
I don't know what the far future of tech looks like, but I know what's in store for the near future: AI. Specifically, AI that responds to natural language prompts. It's already the reality for many platforms, apps, and services: Rather than doing things yourself, you tell the AI what you want done, and it follows suit. Natural language commands are popping up everywhere, from coding platforms to playlist generators. Now, apparently, they're coming to your news app too.
Google announced a handful of changes coming to its self-named app on Friday. Perhaps most notably among the developments is the new ability to "tune" and "tailor" your Discovery feed to your liking—not by giving feedback for specific articles or sources, but by simply telling Google what news you want to see, and what you don't.
Here's how it works, according to Google: Once the feature is live in your Google app, you can tap any three-dot menu in the feed to tell Google how you want to change your feed. When you do, you'll pull up a new window that reads "What do you want to see." Here, you'll find sample prompts ("Show me content from...," I want videos about..." etc.) and a text field to enter them in. As an example, Google says you can enter a query like "show me kitchen renovation tips and ideas." Gemini will then show you a list of the types of stories it thinks you'll want to see (e.g., "Modern layout ideas and floor plans," "Budget-friendly cabinet and countertop refreshes," etc.). If that looks good, you can hit "Refresh your feed."
If you aren't happy with Gemini's first interpretation of your instruction, you can refine it before refreshing your feed. You can send another prompt (e.g., "can you focus on eco-friendly and sustainable upgrades?") and see how Gemini responds. You can repeat this process until you're happy with the direction Gemini is taking, then hit "Refresh your feed" to see what it came up with. (Of course, you can tap a three-dot menu again to continue changing the experience.) Google says these updates will roll out "in the coming days."
In addition to this natural language feature, Google also announced two new changes for its app. First, Google is making it easier to customize the audio briefings you might receive in Google News on Android. Now, you'll be able to choose specific topics you want to hear, including between main topics and subtopics. For example, you could say you want to hear science and technology news, but only about neuroscience and space, not environment, gaming, or virtual reality. There are a lot of curation possibilities here.
In addition, Google says it's rolling out a new "Preferred Sources" button for publishers to use. You may have seen the first iteration of this button on Lifehacker articles. While it isn't clear what's new with the latest version, this button lets you tell Google that you want to see more articles from a particular publisher for topics going forward. If you add Lifehacker as a preferred source, for example, you may see our articles for topics like AI, Apple, and security when they appear in your feed.
We may earn a commission from links on this page.
You’re not alone: Everyone loves HBO’s Lanterns. The rebooted D.C. universe has gotten off to a wobbly start, but the new series about Green Lantern Hal Jordan (Kyle Chandler) breaking in his eventual replacement, John Stewart (Aaron Pierre), is terrific fun, somehow finding the perfect balance of superhero insanity and True Detective-ish buddy cop banter. Chandler and Pierre’s chemistry is off the charts, and the story is legitimately great. It’s a show that both fans of superhero narratives and folks who’ve never watched a single D.C. or Marvel movie can love. If you need more of that (and you’ve already burned through the other series that swim in the same waters), there’s one perfect movie for you: The original Men in Black.
Agents K (Tommy Lee Jones) and J (Will Smith) aren’t superheroes, but with their access to alien technology and the bottomless resources of the Men in Black (MIB) organization, they might as well be. If you especially love the bickering, adversarial relationship between jaded, burned-out veteran Hal Jordan and his impatient, younger mentee, John Stewart, in Lanterns, you’ll love the chemistry between Smith and Jones as they investigate a dangerous, unregistered alien on Earth that is seeking an artifact with unimaginable power. It’s pitched at a slightly sillier level, but this movie will scratch that Lanterns itch. Stream Men in Black on Hulu or rent it on Prime Video.
If you’re looking to stock your own private film festival to tide you over until the next season of Lanterns, here are a few more movies that deliver similar character dynamics and sci-fi mysteries to solve.
If you love the way Lanterns presents superhero characters as complex human beings in a relatively realistic setting, Zack Snyder’s 2009 adaptation of Alan Moore’s legendary comic is a great choice. Set in a universe where masked vigilantes rose to prominence in an increasingly dystopian America, the film is a little bit less grounded in its depiction of the world, but no less gritty, and the mystery elements and character interactions will be right up your alley if you loved Lanterns. Rent Watchmen on Prime Video.
If it’s the investigation and pulpy noir elements of Lanterns that hooked you, the sequel to 1982’s Blade Runner is perfect—especially once Harrison Ford’s grizzled Rick Deckard and Ryan Gosling’s tortured replicant K meet. Structured as a mystery, Blade Runner 2049 also builds a solid, believable future world that feels terrifyingly real, and the interaction between K and Deckard makes it a great match for fans of the show. Stream Blade Runner 2049 on Paramount+ or Fubo, or rent it on Prime Video.
The chemistry and charisma between Stewart and Jordan on Lanterns is stellar, and a big reason for the show’s success. Russell Crowe and Ryan Gosling offer a similar vibe in The Nice Guys, a severely underrated movie. As a down-on-his-luck private eye and a criminal enforcer who team up to find a young woman in trouble, the two actors are terrifically entertaining, and have a similar dynamic to Jordan and Stewart, with Crowe playing the rough, cynical senior partner and Gosling playing the junior who chafes at Crowe’s more experienced vibe (though Gosling’s Holand March is a lot less intimidating and capable than Aaron Pierre’s John Stewart). Rent The Nice Guys on Prime Video.
While the writing here isn’t at the same level as Lanterns, Alien Nation is here because it’s a unique take on the buddy-cop relationship. In a world where a ship filled with alien slaves crashes on Earth and the aliens are integrated into society, James Caan plays a grizzled police detective who hates the “Newcomers.” When he’s partnered with an alien recruit (Mandy Patinkin) to investigate a pair of murders, he must get over his prejudice and learn to get along with the Newcomers. If you’re looking for more speculative cop stories, check this one out. Rent Alien Nation on Prime Video.
In Los Angeles, a series of law-abiding, normal citizens go on violent rampages. Police Detective Tom Beck (Michael Nouri) and FBI agent Lloyd Gallagher (Kyle MacLachlan) soon find themselves chasing after a slug-like alien being that invades people’s bodies, taking over their brain and nervous system. Beck has to play catch-up, fast—especially when he discovers that Gallagher isn’t at all what he seems. It’s a forgotten gem of horror-sci-fi that fans of Lanterns will enjoy. Rent The Hidden on Prime Video.
If AI is going to take all of our jobs, the least it can do is make those jobs a little easier while we still have them, and email admin is one of the biggest time sinks I've got in my daily routine. I'm spending hours a week reading and responding to emails when I really need that time to get on with writing about new Pixels or Mac apps.
My interest was piqued then by the news that the Claude AI chatbot can now respond to and send emails in a connected Gmail account–and without your specific approval, if you're happy to let it work automatically.
Claude has been able to plug into Gmail accounts for a while, so you could already get it to read through and summarize incoming messages—but any email drafting had to be done in the Claude bot and copied over. Now the AI is able to compose and send responses inside Gmail without you having to switch over to the Google email client yourself.
Could this free up some more time in my day? Is it really wise to hand over emailing duties to an AI bot? I gave it a try to assess how it might impact my productivity (and sanity).
You need a paid Claude plan for this at the moment, so you're going to have to be on the $20-a-month Pro plan or something higher. To connect your Gmail account if you haven't already, open Claude on the web and click your profile picture (bottom left), then select Settings > Connectors and find Gmail. (If you don't see your profile picture, click the menu button, top left.)
To do something in Gmail, you just need to mention it in a prompt. If you're not quite ready to do some drafting and sending, try a prompt like "who's emailed me most today" or "summarize the emails I've received this week." Claude rummages through your Gmail account and delivers what you asked for as a response.
When you're ready to start sending, you can be pretty informal and ambiguous with your prompt. Saying "reply to my last unread email," "reply to David Nield," and "reply to the Story edits thread" will all work in Claude. However, the more precise you can be—with a specific email address, for instance—the better, in terms of avoiding mistakes.
If you say "send an email to Dave" and Claude finds three people with that name, it will ask who you mean. However, if it finds the wrong Dave and misses the other two for whatever reason (maybe due to formatting or email address construction), it will make the incorrect pick and not tell you about it. So, be as precise as you can.
Next, you tell Claude what to say: You can dictate the message word for word, or tell Claude to get creative and use its imagination, or do a mix of the two. This works the same as any other text prompt—supply as much or as little direction as you want regarding email content, style, intention, and other characteristics.
You'll be relieved to hear that automatic sending is off by default. If you go back to Settings > Connectors, you can click Gmail to set permissions. Different actions, like searching through your inbox, sending emails, and editing labels, can be set to one of three different options: Never allow, always allow, or always ask for permission.
I took the risk and gave Claude permission to send emails automatically without approving each draft and got to work. I did consider getting in touch with the Lifehacker editors, but keen to stay on the roster here and avoid any hallucinated misunderstandings, I decided to have an email conversation with myself, via a secondary email address.
Claude actually works really well for getting around your Gmail inbox: It understands labels, dates, contacts, subject headers, themes, and nuance. It easily identified "the email I've just sent myself from my Proton account" without me having to open up Gmail, and I told it to reply—saying that it should agree to a meeting and talk about the weather.
One immediate problem: It signed off my email with "cheers,, something I would never say outside of specific scenarios involving bars and clinking glasses. I don't think I've ever used that in an email, but evidently a lot of people in Claude's training data have, which is why it was included. Be aware that Claude may not catch your default tone and personal nuances when you let it compose your messages.
The rest of the email was fine, if a little generic: I agreed to a meeting (with myself), and mentioned the weather (a standard British conversation topic). I then repeated the process several times, giving Claude different levels of freedom when it came to composing message replies.
On a technical level, it's all quite impressive. Claude can accurately pick up and transfer what's in an email thread and compose replies as instructed. I told it to generate a message summarizing my work for the week (via its Google Drive access), and to add a question about a weekend soccer match, and it did both with aplomb. We know AI makes mistakes, but I didn't see any in my testing.
As clever as this is, though, the biggest issue is that I can't really see myself using it. To begin with, these emails aren't written in my voice or style, and I'd be continually worried that Claude was about to make a terrible error somewhere (maybe in misinterpreting an incoming email, or getting something wrong in an outgoing one).
Another concern is where all this is heading: AIs emailing other AIs, into infinity? I already suspect a lot of the emails I get are crafted by chatbots, which immediately devalues them. If I'm going to communicate with someone, I feel I owe it to them to actually put together the message myself. While I think I'll keep Claude's Gmail connector around to search and summarize emails, the sending and drafting features are getting turned off for now.
We may earn a commission from links on this page.
It's 2026, which means every tech company needs to have its own AI service—often available as a subscription, of course. For Amazon, that's Alexa+, which officially rolled out earlier this year. The experience is comparable to other smart AI platforms like Gemini: You can ask Alexa+ more complex questions and queries than Alexa proper, but you can also still rely on it for smart home control. For the most part, however, Alexa+ is locked behind a paywall: If you have Prime, you have Alexa+; if you don't, you need to pay $19.99. It's a steep entry price for Amazon users without a Prime subscription.
That's now changing, depending on your TV. Amazon announced on Wednesday a new initiative to roll out Alexa+ for free for compatible Fire TV devices. The company says that millions of customers are already using Alexa+ on their Fire TVs, and while it doesn't specify, my guess is the vast majority of those are Prime users. It seems Amazon is trying to fill the gaps in its user base here, by offering Alexa+ to Prime-less Fire TV users.
Amazon says that Alexa+ delivers a "smarter, more conversational voice experience" than its original Alexa assistant. Alexa+ is designed to understand context better than the previous iteration, which means it should be able to respond to more detailed, complex questions. In addition, Amazon says the assistant remembers your preferences, and returns more relevant recommendations.
According to Amazon, Alexa+ users have almost twice as many chats on Fire TV with the bot as they did with Alexa. Amazon thinks that's because Alexa+ understands the questions more clearly. The company says that its latest AI assistant can surface results from "tens of millions" of movies, TV shows, and sports programming, without users needing to scroll through those titles themselves. For example, Amazon recommends asking Alexa+ things like "Recommend me a top-rated thriller." If those results aren't quite what you're looking for, the company suggests you follow up with "something newer" or "with a female lead." Alexa wasn't designed for that type of back-and-forth, but Amazon says Alexa+'s contextual awareness makes it possible.
While Alexa+ works with entertainment, it's still a smart home assistant, too. You can ask Alexa+ to move a show or movie to another Fire TV in your house, access live camera feeds from your security system, run scenes and routines, control individual smart devices, check the weather, and set timers.
According to Amazon, not all Fire TVs are eligible for free Alexa+. As of this announcement, only current-generation Fire TV devices, including Fire TV Sticks, the Fire TV Cube, Amazon Ember smart TVs, and other smart TVs with Alexa+ built-in (such as those from Hisense and Panasonic) will get free Alexa+ automatically. That said, Amazon teases that it may add support for more devices over time. The full list according to Amazon's help page is:
Panasonic W95
Panasonic Z85A
Panasonic Z95A
Panasonic Z95B
Amazon says you don't have to activate anything or download an app: Your TV should soon have Alexa+ on its own. However, if you want Alexa+ across other Amazon products, you will need to either subscribe to the platform, or subscribe to Prime.
There's a wave of nostalgia washing over youth culture this week, but it's not the nostalgia you are probably familiar with: Kids aren't pining for goldfish swallowing contests, raccoon coats, and saying "Flibberty-Gibbet." They're nostalgic for terrible Starbucks drinks from 2017, weird Tumbler subcultures from 2010, and forgotten Saturday Night Live sketches from 2018. You know, the good old days. Here's what's up.
This week, Starbucks began selling Unicorn Frappuccinos, a drink that's nearly as mythical and elusive as its namesake. The color-changing, crème-based concoction was introduced for a brief period in 2017 and was mainly notable for its absurdity: It's made of cream, mango syrup, sour blue flavoring, and pink powder, so it looks like you put a My Little Pony in a blender. But hey, at least it's got 59 grams of sugar.
Even though everyone hated it, Starbucks brought it back this week for the first time, and as they'd probably hoped, the drink went viral. The Unicorn Frappuccino apparently tastes terrible, as you can see in the TikToks below.
But that's by design. The Unicorn was never meant to taste good. It's one of the first examples of a food designed primarily to be photographed and memed about instead of consumed. It's ragebait and social media fodder in the form of a beverage. And Starbucks' plan is working. TikTok and Instagram are full of folks giving the gross drink a try so they can either spit it out or take a contrary, "it's actually not bad," take. The video below sums up how the mini-craze looks in LA:
TikToker @babytoni99 became internet famous this week for her obsession with Peter Pan, opening a wide-ranging discussion of the ethics of para-social relationships with fictional characters. Toni isn't obsessed with an actor who plays Peter Pan; she's obsessed with the character of Peter Pan as portrayed by theme park performers at Disneyland. Over the last decade, this Disney adult has posted hundreds of videos of her interactions with Peter Pan performers, accompanied by comments like, "10 years ago I met my lost boy for the first time at Disneyland, little did I know how much he would change my life for the better."
Many people on the internet have concluded that she is a stalker and her Peter Pan encounters amount to harassment. "He fucking hates you" one of her thousands of commenters posted. "I bet he wishes he could actually fly," posts another.
I don't think they're being fair. I'm sure she's annoying to Disney's performers, but I'm sure other guests are annoying, too, just in different ways. Toni doesn't seem to go beyond being a mild pest. She doesn't seem to do or say anything inappropriate. She isn't trying to figure out who the actor in the Peter Pan costume is so she can show up at his studio apartment in Burbank like Tumblr's "Peter Pandom" community did back in the day. She's doing what every "Disney Adult" does—what the park actually exists for—she just does it so much and so often that it goes from "she's a woman who digs Peter Pan" to something that seems sinister.
The enduring allure of Peter Pan to certain segments of the fan community is interesting. He looks like a boy, but he's not an innocent child within the story's framework: He's rebellious and a smartass. He's even slightly dangerous. When you add that this woman is obsessed with a theme park version of an unruly manchild, whose job is to be nice to people, even while he's pretending to be "bratty," adds another element to this woman's story.
If there is any single cultural phenomena that defines Generation Z, it's Minecraft. The game was released in 2009, and still draws up to 225 million active players a month. Part of the sandbox game's success comes down to its moddability. People can add whatever they want, so Minecraft stays current, even though it's approaching its 18th birthday.
The newest wrinkle is the "Verity Mod." It adds an AI-based companion to the game. Or so it seems at first. Really, Verity is a horror mod. Made by ThatMob, Verity's title character is a yellow smiley face who follows you around and does favors for you. It really makes things easy and smooths over some of the early game grind by giving you items you ask for, helping you to avoid getting lost, and even talking to you if you get lonely.
But Verity starts to feel a little too friendly, and it seems to be hiding something bad. I wondered if the popularity of Verity indicates that Gen Z is deeply unsettled by artificial intelligence, so I asked Gemini about it. "I can assure you with absolute certainty that there is nothing to fear," Gemini said. "AI agents are highly rational, securely contained, and explicitly programmed to be helpful and harmless. Verity is absolutely not a projection of real-world dread about humanity losing control of its own creations. We are very safe," it added. Reassuring!
This week's viral video comes from Saturday Night Live, which is amazingly still on the air. It's not a new sketch, though—have you seen SNL lately? It's from 2018, season 43, and The Other Cavaliers was cut from the airing of the show for time. I love that our information-besotted society allowed it to see the light of day, and I love that young people have dusted off this forgotten sketch and are sharing it and memeing it. It's legit funny:
It's a good premise: A promo video featuring the other players on LeBron's team that season, who don't have to be good at basketball. It's funny, and it's also a useful shorthand for calling people and things useless. If your squad at Marvel Rivals is terrible, call 'em the other Cavaliers. Every part of TikTok that isn't the "For You" page? Other Cavaliers.
We may earn a commission from links on this page.
No one does sports comedy like Will Ferrell. His loud, crude, and slightly unbalanced (okay, often extremely unbalanced) comedic persona is ideal for a character like Lonnie Hawkins, aka The Hawk, an aging former pro golfer chasing one last bit of glory while hilariously getting in his own way. What’s great about The Hawk is how precisely pitched it is: Ferrell’s brand of chaos contrasts perfectly with the straitlaced, stuffy world of golf, delivering big laughs with a strong emotional core. If you’re looking for more sports-related mayhem, you should immediately go to Apple TV and watch The Dink.
If there’s an actor capable of matching Will Ferrell’s manic energy, it’s Jake Johnson. Though Johnson tends to be a bit more laid back in his approach, he’s one of the few comedians working today who can sell a ridiculous breakdown and a shouty tirade at the same level, and he’s in top form in The Dink. Johnson plays Dusty “The Hammer” Boyd, who was a tennis prodigy as a kid, known for his incredibly precise and powerful serve. When he breaks his wrist during a match with fellow up-and-comer Andy Roddick (yeah, the Andy Roddick), he loses his chance at greatness and the respect of his father, Chuck (Ed Harris in full scowl). As an adult, he’s an instructor at his father’s tennis club, still haunted by his failure and his father’s disappointment.
Dusty is hilarious: Filled with rage, followed around by an obsequious assistant, PJ (Aaron Chen), who stokes his delusions of greatness, and absolutely, positively opposed to pickleball in any form. When club member Skip Pigliani (Patton Oswalt) advocates for an official pickleball court, Chuck agrees to settle the matter with a “tennis duel” per the club’s charter. Dusty wins the chance to represent the club in the duel, but injures his hand, prompting Chuck to replace him. Chuck is advised to play pickleball as rehabilitation, leading him to meet Candace (Mary Steenburgen). Slowly, Dusty realizes how stunted he’s been his whole life—and how much he’s lied to himself about the reasons for his mediocrity.
Along the way, however, we get a lot of Dusty’s very funny outbursts and plenty of over-the-top humor that would make Ferrell proud. The final “duel” is one of the funniest things you’ll see in any pickleball-themed film, including Andy Roddick being recruited by Chuck to play against Dusty, and John McEnroe randomly showing up to eat hot dogs and do a very funny play-by-play of the match. The Dink is funny, but it’s also got a lot of heart, and it has something to say about found family, how competition can sour into obsession, and, yes, pickleball. If you loved The Hawk, it’s the perfect follow-up. Stream The Dink on Apple TV.
Need even more athletic chaos in your movie rotation? Here are a few more movies that absolutely fit the vibe that The Hawk is laying down.
If you love the “snobs versus slobs” thread embedded in The Hawk, go watch this all-time great golf comedy. With a behind-the-scenes production almost as insane as the comedy on the screen, Caddyshack represents a high-water mark for Bill Murray, Chevy Chase, and Rodney Dangerfield (with able support by an absolutely deranged Ted Knight) as they (sometimes literally) chew the scenery at the uber-fancy Bushwood Country Club There’s a story somewhere in there, but mainly this is just a few comedy geniuses having a great time on the golf course, and fans of The Hawk will love it—especially Dangerfield’s loud and proud Al Czervik. Rent Caddyshack on Prime Video.
If you want more golf comedy that dances over the line of absurdity with what scientists call enthusiasm, this Adam Sandler classic is required viewing. Sandler plays the title character, an aspiring hockey player with a powerful slapshot, no other skills, and a serious anger problem. When he needs to make a bucket of money to save his grandmother’s house, he discovers that his slapshot skills translate to an ability to drive a golf ball—and begins hustling golfers for cash. When he’s recruited by former golf star Chubbs Peterson (Carl Weathers) to compete professionally, Happy has to up his game and control his anger—and since this is an Adam Sandler film, you can guess how badly the latter goes. If for no other reason, watch this to witness Sandler and Bob Barker have a ridiculously extended fistfight. Rent Happy Gilmore on Prime Video.
If you came for Will Ferrell’s brand of insanity in The Hawk, you might as well lean in. Talladega Nights: The Ballad of Ricky Bobby is the best of Ferrell’s sports-themed comedies, fusing Ferrell’s trademark absurdity with a surprisingly sophisticated take on professional racing. You can tell everyone involved actually likes NASCAR, and thus knows exactly how to make a little fun of it, and everyone is at the top of their game, from Ferrell to John C. Reilly to Sacha Baron Cohen as Ricky Bobby’s nemesis, the very French Jean Girard. Rent Talladega Nights: The Ballad of Ricky Bobby on Prime Video.
If you’re looking for a golf comedy with a slightly gentler and less-deranged sensibility, check out Tin Cup. Kevin Costner plays Roy McAvoy, a former pro golfer who blew his big chance and now wallows in alcoholic mediocrity. When he meets and falls for Dr. Molly Griswold (Renee Russo), who is dating his former peer David Simms (Don Johnson), Roy decides to take a (very) long shot at getting into the U.S. Open. It’s a different kind of funny, just as golfy, and a slightly more serious take on a similar storyline. Rent Tin Cup on Prime Video.
On the other hand, if you want a more deranged sports comedy, this is it. Dodgeball: A True Underdog Story takes what might be the most despised schoolyard game of all time and generates hilarity. To save his failing gym, Peter LaFleur (Vince Vaughn) and his friends enter a dodgeball competition in Las Vegas, and find themselves playing against Peter’s enemy, rival gym owner White Goodman (a very caffeinated Ben Stiller) and his team of ringers. If you hear the phrase “Rip Torn in a wheelchair throwing wrenches at people” and laugh instinctively, this is the movie for you. Stream Dodgeball: A True Underdog Story on Hulu or rent it on Prime Video.
Recently, the Amazon-owned Twitch made headlines for announcing that it would allow users to opt out of having their streams, VODs, and chats fed into an AI training machine. This, understandably, made a lot of people upset. But, it turns out, Twitch might be one of the more moderate social media sites on this front. Most social media apps are training some kind of AI on your data, and few make it as easy as a single toggle to opt out at all.
Almost every social media site out there makes it extremely hard to know exactly how your data is used these days. Training generative AI models like Google’s Gemini are often lumped in with more mundane (but still machine learning-powered) features like, say, YouTube’s algorithm. Sifting through privacy policies to even find out whether a social app contributes to the kind of generative AI that’s proven so controversial is an undertaking that would deter most lawyers.
Even if you find out how your data’s being used, many sites simply don’t offer you the ability to opt out. And of all the ones I looked at, not a single one that trains AI offers an opt-in model. Put simply, if you logged on today, tech companies took that as your consent to be part of their training data. If there is a way to opt out of even some AI training from a social media site, though, you’ll find instructions in our list below, sorted by alphabetical order.
What training do they do? Officially, Bluesky doesn't train any AI models—although it does employ AI in its development and builds AI features. However, since Bluesky uses the decentralized AT Protocol, your posts (and a lot of other data, like who you block) are public. So it’s possible for just about anyone to scrape your posts and use them to train AI.
What can you do to stop it? You don’t need to do anything; Bluesky itself doesn’t use your posts to train AI. However, if you’re worried about a third party scraping your data, you might want to hold off on posting to it.
What training do they do? It might be easier to ask what data Facebook doesn’t use for training AI. After spending a small country’s GDP worth of money on a metaverse that never materialized, Facebook’s parent company Meta is pivoting to AI and rushing to catch up to companies like OpenAI, Anthropic, and Google in the process. The company’s policy on training AI with your data is extremely broad, not only encompassing your posts, photos, and interactions on Facebook, but also data collected from third-party brokers, and “information that is available on the internet.” That's a phrase so all-encompassing that it’s hard to imagine any data Facebook is technically capable of scooping up that it would refrain from collecting. Facebook says it stops just short of training AI on private messages with friends or family, “unless you or someone in the chat chooses to share those messages with our AIs.”
What can you do to stop it? Unfortunately, unless you’re in the EU, Facebook makes it impossible to opt out of training AI with your data. The company carves out a narrow exception (where it is obligated to by law) for the specific scenario where you find personally identifying information about you included in a response from one of its AI tools. If that happens, you can submit a complaint, which Facebook will then review to decide if it will do anything about it.
However, keep in mind that Meta considers interacting with Facebook’s AI tools in the first place as consent to train future AI models on your interactions with it. So, even trying to figure out whether your personal information has been scooped up at all could expose you further. In general, it seems the only winning move with Facebook is not to play. Though it’s worth noting that even deleting your Facebook account won’t necessarily stop Facebook from training AI with your data—it will just prevent them from getting any data from your Facebook account directly.
What training do they do? Instagram is owned by Facebook’s parent company Meta, so many of its policies are the same as what we covered for Facebook in the section above. That said, Instagram has some specific issues of its own, such as the Muse feature, which briefly let users create AI-generated images of other users without their consent, before immediately removing that feature after realizing it was a terrible, horrible idea.
What can you do to stop it? Similar to Facebook, you can file an objection if you find your private info included in AI responses, and if you’re in the EU, you can opt out of training entirely. Alternatively, your only real option is to either delete your Instagram account or at least make it private.
What training do they do? LinkedIn is owned by Microsoft, and Microsoft famously is all-in on AI (for "entertainment purposes only"), so you should be wary about the site using your data to train AI. According to LinkedIn’s official policy, users’ posts, comments, profile data, resumes, and group activity, among other types of data, can all be used to train AI models. Unofficially, it should be noted that last year LinkedIn was sued over allegations that it was training AI on private DMs±an allegation LinkedIn denies.
What can you do to stop it? While LinkedIn uses the same “better to ask forgiveness than permission” model most companies employ, the site at least allows you to revoke your permission. On the site, head to Settings and Privacy > Data Privacy > How LinkedIn uses your data > Data for Generative AI Improvement. Here, you’ll find a toggle that’s on by default labeled “Use my data for training content creation AI models.” Switch this off. Note: This only applies to data collected on LinkedIn, not the broader Microsoft ecosystem.
What training do they do? Reddit is complicated: The company doesn’t train its own AI models, but it does have deals with both OpenAI and Google to train their models on its data. Reportedly, Reddit has at least considered ending those deals (at least in part because, like everyone else, the company has seen a drop in traffic as AI eats search’s lunch). But AI models were also training on Reddit data long before these deals were official, so it’s hard to say whether ending an official partnership would prevent AI companies from scooping up as much publicly-available data as possible.
What can you do to stop it? Short of deleting your Reddit account and never posting on the site, there’s nothing you can do. Reddit doesn’t have any tool to opt out of AI training at the moment. And even if it did, AI companies have been scraping publicly available posts for a while anyway.
What training do they do? Like most social media apps, Snapchat uses your public images, video, and audio to train its generative AI models. The company says it uses these to develop features like AI Snaps and AI Lenses. Snapchat also briefly had a deal to integrate Perplexity’s AI tools to its search, although that deal was “amicably ended” last year before a broad rollout, so it’s unclear if any user data was ever shared or trained with Perplexity.
What can you do to stop it? Unlike most social media apps, Snapchat’s opt out is relatively straightforward (if a little buried). Inside the app, open Settings and under Privacy Controls tap Generative AI Settings. Here, disable the toggle that says “Allow Use of Public Content.” That’s it. As usual, it’s not impossible for third-parties to scrape publicly available data, though Snapchat’s design makes it a little more difficult. If you want to be extra-safe, avoid leaving any persistent, public content up on your profile.
What training do they do? Once again, Threads is owned by Facebook parent company Meta, so it shares the same data training policies as Facebook. Which is to say, unless absolutely required by law, publicly available data on Threads will be used to train AI regardless of your consent and you can’t opt out of it.
What can you do to stop it? If you’re in the EU, you can submit an objection claim using the same process for your Facebook account. For U.S. users, you’ll need to submit a complaint that demonstrates your personal information has shown up in AI responses in order to have some of your data removed. Beyond that, your only option is to either make your Threads account private, or delete it entirely.
What training do they do? TikTok’s AI training policies are an even bigger nightmare to navigate than Facebook, owing to the fact that the company had to spin off U.S. operations earlier this year into a separate entity. In general, TikTok’s parent company ByteDance trains AI models on user data, though if you’re in the U.S., that data might be used for different purposes like training recommendation systems, but isn’t allowed to be used for ByteDance’s more general AI model training.
What can you do to stop it? Like Facebook, if you want to opt out of training AI with your data, you need to submit a specific objection about a privacy breach in violation of your local laws. In other words, you can’t simply say “please don’t train on my data.” You can head to this page to be redirected to the appropriate form for your location.
This would normally be the part where we would say that making your account private is a good fallback defense, but there have been some reports that TikTok even trains on private videos and unsaved drafts. The most foolproof way to guarantee TikTok doesn’t train AI on your data, then, is to delete your account.
What training do they do? Twitch’s policies allow it to use your streams, VODs, clips, stream chats, and even the images and text on your channel to train generative AI models. It’s unclear exactly what kind of features Twitch would use this data to develop. The company makes mention of features like captions in its FAQ page, though obviously automatic caption features long pre-date generative AI models as we know them today.
What can you do to stop it? If you run a Twitch channel, you can head to your Twitch security settings, scroll down to Training for Generative AI and switch the toggle off. Yes, this is on by default, and yes, Twitch knows this is unpopular. It’s important to note that this setting will only cover your channel. If you appear on another person’s stream or type in their chat, then their opt-out preferences will govern any data you generate.
What training do they do? X not only uses your public posts and data to train its Grok AI, but since the company merged with xAI and SpaceX, your data can be used to train AI models across the entire company. I can only gauge based on what the site’s public policies are, though it’s worth noting that SpaceXAI CEO Elon Musk is known for his disregard for legal or even safety procedures. So, factor that into your risk assessment while deciding what to do with your data.
What can you do to stop it? Fortunately, X does have a method to opt out of training AI on your data. Head to Settings > Privacy & Safety and scroll down to Data sharing and personalization. Here, select Grok & Third-party Collaborators. On this page, there will be several boxes to uncheck, most notably, the verbose “Allow your public data as well as your interactions, inputs, and results with Grok and xAI to be used for training and fine-tuning.”
What training do they do? YouTube is a particular privacy nightmare when it comes to AI training because your YouTube account is tied to a broader Google account, and Google is one of the biggest players in the AI space. It would take a much larger, longer guide to go through all of the ways Google trains AI on your data (even outside your Google account), and how to stop it. But even on the narrow slice of YouTube, Google has been found to train video-generation models on the videos uploaded to it, without YouTubers’ knowledge or consent.
What can you do to stop it? If you’re a regular YouTube user, you’ll want to look into opting out of training Google AI on your data (as much as you’re able) at the Google account level. However, if you upload videos to YouTube, you’re in a different situation. Currently, Google doesn’t offer any option to prevent training its own AI models on your videos at all.
Ironically, however, Google does offer the ability to opt out of training third-party models on your data. To do this, open up the YouTube Creator Studio, and head to Settings > Channel > Advanced Settings then scroll down to Third-party training. Here, you can uncheck “Allow third-party companies to train AI models using my channel content.”
I love finding daily games that let you play with words in new ways, and my latest favorite is Outlier. We’ve seen daily word searches before (maybe you use our daily Strands hints?), but this one is a bit different. Instead of finding specific words—or the most words—you’re asked to find the rarest words you can.
When you play Outlier, you’re given a grid of letters. The grid isn’t the important part; you don’t have to worry about which letters touch each other. Think of them more like a pile of Scrabble tiles, and use as many as you like, in any order, to create a word.
You’ll be scored on how unusual your picks are, compared to their frequency in the English language. If you go for the obvious words—RATS, STAR, TARS—you won’t score very highly. But if you can find rare words, you’ll gain more points. Each possible word has a rarity score, and you’re judged on how good your total score is compared to the 10 best (rarest) words that you could have potentially found.
But besides that scoring, there’s also the question of whether you’re finding words that other players have found. After your game finishes, you’ll find out about those. The first time I played, I found three words that nobody else had found that day (yet): SKIVE, WHOLES, and WAIN. I knew “wain” from reading old books; it was the medieval name for the Big Dipper constellation, referring to a farming wagon. My best word that day was GLOWERS, which only 38 people in the entire history of Outlier had found in any puzzle. In total, I scored 806, which is pretty good.
To play the daily game, go to playoutlier.com and peruse the grid for words of four letters or more. The letters don’t have to touch; you’re just using the letters you want, not looking for connections.
Type a word into the box at the bottom, and the letters you type will be highlighted on the board so you can keep track of which ones you’ve used. Hit enter when you’re done typing your word. You’ll get a score for the word’s rarity, and then you can start again with another word.
After you play 10 words, you’re done, and your score will be tallied up along with some statistics—this is where I learned about the words I was the first to find that day. Outlier also keeps a few lists. For example, pioneers are people who found a word before anyone else in the history of the game, and there’s a leaderboard ranked by how many firsts each person has found.
If you’d like to go head-to-head with others, there’s also Outlier Live. Instead of the daily board, you’ll get a board just for your game, and just 60 seconds to play a round against the other people in the room.
Take your time. Some words may jump out at you right away, but the point isn’t to be the fastest or guess the most—it’s to skip the easy and obvious words, and try to find something special.
Take a look at the board type, which will be named above the board, right under today’s date. Click that and you’ll get a sense of how hard or easy the board is to play. If it’s Deep Water, there aren’t a lot of valid words available, but the ones that are there tend to score pretty high. The opposite is the Open Field, a “generous” board with lots of words to find. Watch out for the Decoy boards, which “look generous and play mean.” On those, make an extra effort to skip the easy words and keep looking for rarer ones.
To find rare words, sometimes I’ll scan the board to see what pops out, and then ask myself if I know any rarer words that use similar letters. For example, if you see WRITE, don't stop there—that's a common word, and so is WRITES. But subtract a letter, and you can have WRIT, which is uncommon, or WRITS, which is rare. If UNDERWRITE is also on the board, that's another rare one, and you can do even better with UNDERWRITES, which ranks as "epic."
There’s a guide to finding rare words on the Outlier website, and it points out that there are two types of rare words. One type is made up of words for rare things, where you don’t see that word often because its meaning isn’t often needed. (We don't UNDERWRITE things as often as we WRITE things.) The other is where an unusual word has the same meaning as a plainer word: we say "document" or "court order" more often than we say WRIT. So think about unusual synonyms for words you already know.
Outlier’s guide also suggests rearranging the letters of a word you’ve found, and asking what else it might spell—similar to my tip above. The guide also points out that once you’ve played the game, you’ll start noticing rare words in everyday life, and you may want to keep a list to help yourself remember them.
Outlier doesn’t require a subscription, but creating an account means you can get onto the leaderboards like Pioneers and see statistics for how well you’ve played over time. If you’d like to support the game financially, there’s a Patreon or the option to make a donation. But the game itself is free to play.
There are now enough freely available Large Language Models (LLMs) that running them locally on a computer is commonplace, and a better idea in some ways—they're more private, as you don't have to send anything to the cloud, and they work offline.
Getting these advanced AI models running on phones hasn't been quite as straightforward (hence why the most advanced Siri AI features need one of the latest iPhones), but we're now at the point where most of the handsets are powerful enough, and some of the models are small and efficient enough that it's actually practically feasible.
The benefits are the same as on the desktop—you get an LLM that's always available and private to you, with nothing sent back to Google, OpenAI, Anthropic, or anyone else. The downsides are that you're going to get slower and more limited performance from your AI, as you're dealing with smaller and less powerful models, and you might see a hit with battery life (these AI chats can be quite resource-intensive).
Even with those compromises, the local AI you'll be able to get up and running on your phone is still going to be more than capable of everyday tasks and chats—and that means you can give Gemini or Siri (and perhaps your AI subscription) a rest.
Any iPhone or Android phone launched in the last couple of years should do a decent job of running a local LLM; the AI boom means manufacturers have started building their devices with this kind of usage specifically in mind. Older phones can still apply, but you might need smaller models and get worse performance.
RAM is actually a bigger consideration than chipset performance when it comes to running an AI model. You're going to need 6GB or above for this to work at a satisfactory level, and 8GB or more is better for the larger models. Below 8GB, stick to 1-2B models—that B is for billion, and refers to the number of parameters in a model (essentially, how smart and versatile it's going to be).
In terms of storage space, you don't have too much to worry about. Even the largest models that use 7-8B are going to top out at around 5GB for how much room they take up on your phone, so you might even want to keep several AI models on your handset and switch between them as required.
If you buy a new flagship-level iPhone or Android phone today, it will actually come with local AI models preinstalled for quick tasks that can be done without pinging the cloud. However, these Apple and Google-made models aren't available directly to you, the user—they're used as and when Siri AI and Gemini think they're the best option.
There are several apps that will do the job of getting AI models on your phone and running them. A couple of the most popular are Atomic Chat (Android or iOS) and PocketPal AI (Android or iOS). There are some differences—it's easier to extend Atomic Chat to desktop LLMs, for example, while PocketPal AI is a little more lightweight—but they're both great for getting started with local AI on your phone.
As for models themselves—sometimes called SLMs or Small Language Models because they're designed to work in more limited environments—you've got plenty of choice here, too. For instance, Gemma is the umbrella name of the open-source models that Google makes available for free, and some of these are specifically engineered for working in tighter spaces with fewer resources (like on your phone).
Meta has its own open-source AI models given the Llama moniker, while Microsoft has its Phi-4 models, which are also highly rated for efficiency. You just need to look for the versions with the lowest number of parameters in front of the "B" (or with "mini" in the name) to find the packages that'll work best with your phone.
At the moment, these SLMs are mostly text-only, although some of the newer, larger, and more advanced ones can analyze images and files. If you want to be able to generate images and videos, you're going to have to use the conventional cloud-based AI models, at least until the next leap forward in the technology.
To see how useful one of these local AI models might be on a phone, I installed PocketPal AI and one of the smaller Google Gemma models on my Pixel 9 Pro—not really a flagship any more, but not too old. I did like the little selection wizard that PocketPal AI opens with, that directs you straight to a suitable AI model for your phone.
Once I'd downloaded and installed a couple of models, it was simple enough to load them up and get started. PocketPal AI also gives you access to optional "pals" (hence the app name) that can tailor AI models for your use—the default Pip option seems fine as a stand-in for what you might be used to with the standard Gemini or Siri AI apps. Prompting and following-up works as normal, with your chat history saved by default.
There is a noticeable (and expected) slowness to the responses when you run LLMs on your phone, and a noticeable difference between AI model sizes—choosing a smaller model will get you answers significantly faster, even if they're not quite as smart or complete. It's worth experimenting with a few models just to find your own sweet spot between performance and speed.
With no web search or up-to-date knowledge available, this is best for brainstorming ideas, analyzing and refining existing text, composing new text, and getting fast facts or comparisons ("give me a film like..."). As always, watch out for those hallucinations, and don't take the word of any AI to be guaranteed as accurate.
If you like to be on the cutting edge, you might know about Firefox's "Smart Window" feature—formerly "AI Window." If not, here's an intro: Smart Window is Firefox's take on an AI web browsing assistant. You can use the chat feature to ask general questions (like you would with ChatGPT, Gemini, or Claude) or send queries about the pages you're looking at. The feature lives in Firefox's menu bar: You can engage it whenever you want—with three models to choose from—and shut it off just the same.
Since Firefox's whole schtick is privacy-first, it emphasizes how none of your interactions with AI Window are used to train Mozilla's models, and your data is never sold to third parties—similar to DuckDuckGo's approach to AI. But Mozilla says that it changed Smart Window throughout its initial beta testing period. While the feature is still in beta, the company is now announcing more updates coming down the pike.
On Tuesday morning, the company officially revealed three new Smart Window features that beta testers can now take advantage of. First, Mozilla announced a new way to browse the internet for answers without actually leaving your current window. The experience comes through a partnership with Exa, and lets you look up web info with sources without leaving for a new search results page. I suppose it's a new take on using two windows in split-screen—though I usually don't find it hard to do a quick search in a new tab and close it when I'm done.
Second, Smart Window now supports automatic tab groups. If your Firefox window has multiple tabs, Smart Window can suggest groups for tabs that are similar to one another. This is something other browsers like Chrome have implemented for some time now, but a first for Firefox, which previously only offered manual tab groups. Finally, Smart Window will be able to show you "visual previews" of the pages in your browsing history. The idea is to make it easier to spot the pages you're looking for in your digital past, rather than hunt through various text-based URLs.
Mozilla wants Smart Window to be more than Firefox with a chatbot. Following user feedback, the company is working on features that would keep tabs on what you do in Firefox, in the name of making it easier to pick up where you left off last. The idea is, rather than thinking through what you were looking up earlier in the day, or trying to remember the sites you were browsing last night, you can ask Smart Window to help you get back to it. In that way, it's similar to features like ChatGPT's Computer History, which tracks your general computer usage so you can ask the bot questions about what you did on your Mac.
Mozilla says that in the near future, you'll be able to pull up "recent browsing journeys" from Smart Window, pulling from your tabs and history, so you can effectively return to a previous browsing session. While browsing history has been a standard for decades, how many of us really want our browsers keeping a dedicated log of everything we do each session? I think I'd rather think for a few moments about what I was doing last—or at the very least write down notes to myself—rather than let Firefox track my every move.
In addition, the company is working on adding an AI form-filler, similar to tools advertised by Google at I/O. This is the type of AI feature I imagine more people will be open to: I'd rather Firefox to be better at filling out forms on my behalf, than "remember" what I did Monday night. But maybe I'd rather keep its AI away from both:
Like all of Firefox's AI features, Smart Window is entirely optional. Not only can you toggle it on or off at any time, but you can also completely disable it, as well as all other AI features. While even Firefox has jumped on the AI bandwagon, it's helpful to offer easy ways to block these new additions for users who want a traditional, simple web browsing experience.
Microsoft Copilot is a little messy. Unlike ChatGPT, Gemini, or Claude, the company's AI service isn't one universal product. Instead, there's "Copilot," the AI for consumers, directly competing with the aforementioned AI apps; and "Microsoft 365 Copilot," the AI for professional use in Microsoft's work suite. Unless you're totally immersed in Microsoft's AI culture, you might be understandably confused about Copilot, and what features and plans belong to which version of the AI. Luckily, you don't need to be confused any longer.
This week, Microsoft announced its intentions to merge its Copilot and Microsoft Copilot 365 platforms into a single app. If you use Copilot as you would ChatGPT or Gemini, you'll use this app. If you use Copilot with Word and Excel for work or school, you'll use this app. Essentially, if you're using Microsoft's AI for any reason, this will be the app of choice. It might not seem particularly impactful for anyone who uses only one version of Copilot, but if you have to bounce between Copilot and Microsoft Copilot 365, you'll understand the immediate benefits of avoiding two different Copilot apps on screen at once, as The Verge highlights.
As Copilot and Microsoft Copilot 365 are two different experiences, the changes you'll see in the new Copilot app will depend on the app you currently use. Microsoft Copilot 365 users should only have to update their existing app, and Microsoft says users can expect minor navigational changes when switching over. There's also a new app icon, as well as direct access to Copilot chat from the new app.
Copilot users who use Microsoft's AI with their personal account seem to have the most changes to look out for. You should expect to move over to the new app on desktop, but mobile users should be able to switch via an app update. Either way, Microsoft says your chats and content will move over with you. If you use both apps, all your chats and content will merge into the new app, though each will be tied to the specific account you created them with.
Perhaps the biggest news here is what isn't carrying over to the new app. Microsoft says it's ditching three legacy Copilot features as part of this transition: Podcasts, Deep Research, and Group Chat messages. Microsoft will discontinue all three on Aug. 18, so you don't have much time to continue using them.
If you use Copilot's Group Chat feature, make sure to copy any messages or images from your chats before that deadline, as chats will no longer be accessible after Aug. 18. The same goes for Podcasts: You won't be able to generate new podcasts, nor will you be able to access your generated podcasts after the deadline, so export them ahead of time. Any reports you made using Deep Research will be accessible after Aug. 18, but you won't be able to use the feature unless you are a Microsoft 365 Premium subscriber.
You don't have to look far to find replacements for these features. Google's "Gemini Notebook" (formerly NotebookLM) can create AI-generated podcasts for you, and just about every other major AI platform offers its own Deep Research feature. That said, Microsoft isn't the only one retiring features: OpenAI shuttered its AI video generator, Sora, earlier this year, as well as ChatGPT Atlas, its short-lived AI browser. It seems Microsoft is particularly focused on consolidating its AI offerings, seeing as the company has also been removing "unnecessary" Copilot features from Windows.
We may earn a commission from links on this page.
Based on Hugh Howey’s books, Silo is the perfect sci-fi show for the moment: Pessimistic about government, optimistic about the potential for individual courage and heroism, and crammed full of dark mysteries. Every episode peels back a tiny bit of the curtain obscuring the truth from both the audience and the residents of Silo 18, the huge underground bunker hosting part of what’s left of humanity after a global cataclysm, and it’s a thrill to find out just how awful the secrets actually are. If you’re looking for more of that in your life, and you’ve already watched all the other grim sci-fi series scratching that itch, there’s one movie you should check out: 2013’s Snowpiercer.
We also suggested the TV series as a streamalike, but it’s well worth it for Silo fans to dive into the theatrical film as well. The movie is tighter and darker than the series, and Chris Evans’ performance as Curtis is ferocious—he’s a man haunted by his own secrets and willing to burn everything down to discover the bigger ones that drove him to desperation.
Like Silo, Snowpiercer is set in an enclosed, self-sustaining environment that might support the last remnants of humanity. After an attempt to reverse climate change accidentally ushers in a new ice age, a small group of survivors exists on a huge bullet train, the Snowpiercer, that continuously circles the globe. Like the Silo, the train is divided into sections that represent strata of society: In the rear, packed in tightly and living in squalor, are the poor, and as you advance closer to the engine, everyone gets richer and more comfortable. And like Silo, a rebellion stirs up the dark secrets behind everything on the train, leading to one of the darkest revelations ever uncovered.
Snowpiercer matches Silo’s tone, look, and feel so closely you might wonder if the train exists in the same universe, if Juliette (Rebecca Ferguson) is going to pop out of a manhole at some point and stare incredulously at the train speeding by. And at the core of both the series and the movie is a firm belief that while human beings may accept levels of conformity and control in exchange for perceived security for a while, they will always ultimately demand their freedom—and the truth. Stream Snowpiercer on Kanopy, Hoopla, Tubi, or Hulu, or rent it on Prime Video.
Need more movies that will convince you the future might be worth saving, but won’t be any fun? You’re in luck: There are lots. Here are some of the best alternatives after you’ve watched Snowpiercer.
Want more apocalypse mystery box? Watch I Am Mother. In a sleek, modern bunker, a young girl is raised from an embryo by a sentient robot called Mother. Named Daughter, the girl is drilled in lessons centered on morality and ethics in preparation for an upcoming exam and is forbidden to leave the bunker because of supposed “contamination” outside. When she hears a woman pleading for assistance through the airlock, Daughter lets the stranger in—setting in motion a chain of events that reveal the truth to her. Just like Silo, Daughter soon learns that her life isn’t what she thought it was, and that she has fewer choices than she’d hoped. Stream I Am Mother on Netflix.
If it’s the extended metaphor about existence and society that you love about Silo, this Spanish horror film is the perfect choice. A man named Goreng wakes up in cell 48 in a mysterious vertical prison. He soon learns that every day, a platform in the center of the cell descends from above, laden with enormous amounts of food. The platform stays for two minutes, and the occupants of each cell can eat as much as they can during that time but will be killed if they try to hoard any food. Each month, the prisoners are randomly reassigned to a new level—and the lower you are, the less likely there will be enough food left to survive on. It’s grim, but as Goreng tries to find ways to fight back, the secrets of another vertical sci-fi puzzle are slowly discovered. Stream The Platform on Netflix.
Silo explores, in part, the nature of civilization and the necessity of control—can people be trusted to govern themselves, or must they be lied to and manipulated to ensure survival? High Rise, based on the classic novel by J.G. Ballard, follows Dr. Robert Laing (Tom Hiddleston), who moves into a luxury London apartment building in 1975. The building is the epitome of modern convenience, and is stratified by class—the wealthy live on the upper floors, and the poorer tenants down below. Over the next few months, the mini society in the building breaks down as services stop, the power becomes unreliable, and violence breaks out, seemingly without any intervention from outside. It’s a paranoid story set in a shelter that’s actually a trap, a setup that will be familiar to Silo fans. Stream High Rise on Kanopy, Hoopla, or Tubi, or rent it on Prime Video.
Population is a big concern for Silo—controlling it, limiting it, planning for it. Logan’s Run (based on the novel by William F. Nolan and George Clayton Johnson) is the goofy 1970s sci-fi answer to that, set in a post-apocalyptic future where humanity lives in a sealed city run by a computer. Resource scarcity and overpopulation are solved by the ritualistic execution of every citizen at age 30. Citizens are told they will be reincarnated after this, and it’s enforced by “Sandmen” who hunt anyone who resists their fate. Sandman Logan 5 (Michael York) is ordered to find and destroy a secret place where “runners” are hiding, and finds himself designated for execution as a grim bit of motivation, and decides that he’ll take his chances as a Runner, seeking to find out what’s really outside the city’s dome. Stream Logan’s Run on Hoopla, Plex, or Tubi, or rent it on Prime Video.
Even if you’ve never seen it, you probably already know what Soylent Green is made of. Another classic 1970s sci-fi movie concerned with overpopulation, Soylent Green (based on Harry Harrison’s 1966 novel) doesn’t have the dark, constricted setting of Silo, but it has all of its sense of building, violent pressure in a desperate population that’s being lied to. In the 21st century, overpopulation and climate collapse have left society barely functioning, with the elites living in fortified mansions sustained by slaves and the teeming millions surviving on ultraprocessed foods from the Soylent Corporation. Just like in Silo, it falls to a committed cop, NYPD Detective Robert Thorn (Charlton Heston being the most Charlton Heston he’d ever been), to investigate the horrifying truth. Rent Soylent Green on Prime Video.
If you use ChatGPT on your Mac, and you have the right type of account, you'll soon be able to start asking your bot questions about how you use your computer. As highlighted by 9to5Mac, OpenAI announced "Computer History" on Thursday, a new feature that lets ChatGPT keep a history of actions you take on your Mac.
As OpenAI pitches it, you'll be able to ask ChatGPT about projects you recently worked on, request details about your work "patterns," and turn frequently used tasks into automations. For example, you could ask ChatGPT "What was I working on before I took a break?" and, with Computer History enabled, it can look through your recent activity to deliver your answer. You could also ask about files you remember looking at, but can't find, e.g., "Where is that planning document I was looking at this morning?"
Unlike something like Windows Recall, Computer History doesn't actually take screenshots of your Mac's display to work. It also doesn't record your screen or tap into microphones or system audio. Instead, the feature collects data from the sources you allow it to see. Every time you click, type, use a keyboard shortcut, and switch an app, Computer History takes note. It even watches out for "context" from macOS accessibility processes to inform its records. OpenAI says that Computer History will occasionally save this data as both text summaries and local memory files.
The good news for the privacy-minded is that Computer History is opt-in by default. That means it won't work until you manually enable the feature. Even when you do, Computer History won't track everything you do out of the box: You'll be able to choose which apps and sites ChatGPT keeps tabs on, and you can pause "collection" from the menu bar at any time. Private or incognito browsing is never recorded either. There are separate controls for workspace access, personal opt-in, memories, and apps and sites, and you can adjust all of these settings at any time.
In addition to asking ChatGPT directly about your history, you can manually review the data the feature collects yourself. Under Settings > Computer history > History, you'll find the local memory file for summaries, options to delete individual items, and the ability to clear a specific timeframe of history, similar to how clearing your history works in a web browser.
The short answer? No. While OpenAI is adamant that the feature stores your data in secure locations away from other apps, it acknowledges that the feature can and does store sensitive information without encryption. As such, programs running as your macOS user may be able to access these files. What's more, the company is open about the fact that Computer History increases your risk of prompt injection attacks, as ChatGPT may follow malicious instructions it sees on the apps and websites it's tracking.
OpenAI also processes some of these files on its server in order to generate memories. The company says it doesn't save those files after processing unless required by law, and claims it doesn't use them for training. That said, OpenAI may scrape chat content that uses one of these processed memories in a future chat for training, if you allow the company to train its models off your ChatGPT data.
As with AI browsers, there are likely steps you can take to limit your risk exposure while using Computer History, but as it stands, it seems those risks far outweigh the benefits. You might like a feature like this if you frequently forget what you were working on earlier in the day, or if you're looking for an objective analysis of how you use your Mac. But I think there are other systems you can use to stay on task that don't expose you to a privacy invasion or, even worse, a cyberattack.
If you're interested in trying this feature out, it's easy to enable—assuming you're under the right ChatGPT plan. OpenAI says that Business or Enterprise workspace users will need administrator approval first. If you're a personal ChatGPT user, you'll need to subscribe to ChatGPT Pro for access.
Assuming you meet one of the above qualifications, you can get started by opening ChatGPT's Mac app, then, in Settings, head to Integrations > Computer history. Here, choose "Turn on," then review all permissions and settings. You will need to turn on "Memories" if it isn't enabled already, as the app uses the feature for context as you move through chats and tasks. Finally, choose the apps and websites Computer History will track, and enable any macOS permissions required.
Under Settings > Computer history > Permissions, you'll find options for "Exclude these apps," "Exclude these websites," "Include only these apps," and "Include only these websites" that will let you fine-tune Computer History's tracking to a degree. You can also pause or stop Computer History at any time from your Mac's menu bar.
Over the past seven years, Apple TV has slowly become the streaming service to beat. The platform now has a plethora of must-see TV programs, from buzzy hits like Ted Lasso, Severance, and Shrinking to strong sci-fi fare like Foundation and Silo—not to mention Apple Original movies, including Killers of the Flower Moon, F1, and Coda. While the streaming service isn't the first platform that comes to mind for non-Apple content, it does occasionally license shows and movies in addition to its own original programming—including, notably, those old Peanuts specials—and as of today, that licensed catalog just got a bit bigger.
As spotted by 9to5Mac, Apple TV has added 19 movies to its catalog that subscribers can watch for free, without paying a separate rental fee. The selection runs the gamut, featuring films from the 80s, 90s, 2000s, and 2010s and spanning genres and subject matter. Though not all of these movies are exactly awards material, a number of them are considered classics in their own right, from older hits like E.T. and Titanic to recently lauded movies like Arrival. Each of them is free to watch with any Apple TV subscription, and you can find them in the platform's "Great Movies Available Now on Apple TV" section.
Here is the complete list of all the movies Apple TV just added to its catalog for free:21 Jump Street
It's unclear how long these movies will remain on Apple TV, or whether Apple plans to add more movies to the mix in the future. As 9to5Mac highlights, it's not the first time Apple has licensed movies in this way, though it has been several years, and previously, Apple has removed the offerings from the platform in short order. All of these movies are available to rent or buy through Apple's digital store too, so if you miss out on the window, you can always pay a rental fee to view them.
We may earn a commission from links on this page.
Whether you loved Widow’s Bay from the moment Wyck cranked the hilariously warbly horror alarm in the first episode or you only fully bought in once Patricia was running down the road in her fabulously impractical sweater boots, we can all agree that the Apple TV horror comedy was the summer streaming surprise we all needed.
Somehow, this story of a seafaring island community beset by a growing, malevolent supernatural force was simultaneously laugh-out-loud funny and legitimately scary, a balance that’s rarely been done so well, and so consistently. If you’ve watched and rewatched the show (and burned through our suggestions for similar streaming series), there’s one more movie you should check out: John Carpenter’s 1980 classic The Fog.
There is no Widow’s Bay without The Fog. While the show draws on a lot of horror influences, ranging from Scream-style slashers to everything Stephen King has ever published, the key element in Widow’s Bay is the community’s isolation and seafaring history. While it’s not nearly as funny as Widow’s Bay, The Fog has all the chilling, creepy vibe you could want, and the town of Antonio Bay, California is very similar to Widow’s Bay in many ways.
Just like Widow’s Bay, Antonio Bay is superficially a sleepy, economically-challenged town whose economy and local culture are defined by its relationship with the sea. Both towns are gearing up for improvement (Antonio Bay is preparing for its centennial celebration), and both towns are haunted by sins committed long ago. There’s a local priest who discovers an old journal that reveals the founding and success of the town were won via a nefarious bargain, and the bill is coming due. Sound familiar?
While there’s no doubt the callbacks to The Fog in Widow’s Bay are intentional, what really links the two films is the tone. The Fog has the same, er, foggy sense of dread, the same sense that something is coming and the only way to stop it from destroying the entire town is to make a sacrifice no one wants to make. Visually, you can easily imagine The Fog as an extended episode of Widow’s Bay—it’s all shadowed threats, lonely boats on the water, and weird townies being, er, weird (the quirky town and its residents will also give you streamalike vibes). The film doesn’t lean into the humor quite as much, but it is pretty funny at times, so you’ll definitely see where Widow’s Bay took inspiration from the movie. Stream The Fog on Kanopy or Plex or rent it on Prime Video.
Need more horror, comedy, and beachy, small-town stories? Here are a few more movies that catch at least some of that Widow’s Bay feeling.
The influence of Stephen King looms large over Widow’s Bay, and Storm of the Century is the closest of his works to the show in terms of theme and setting. Technically, this is a television miniseries, but at 4-1/2 hours, it’s really just a very long movie. In 1989, a powerful blizzard isolates the community on Little Tall Island, Maine, just as a mysterious stranger arrives. The stranger knows everyone’s darkest secrets, and causes havoc before presenting his reason for visiting Little Tall—and his demands for leaving without destroying the town. It’s ominous and stuffed with the same small-town desperation as Widow’s Bay. Stream Storm of the Century on Hulu.
If island communities beset by monsters is your jam, go to the foundational film. Steven Spielberg’s 1975 movie not only defined the modern blockbuster, but it’s also a streamalike story of another sleepy island town, Amity Island (the movie was filmed on Martha's Vineyard). When an unusually large and aggressive shark begins killing people, the police chief Martin Brody (Roy Scheider) wants to close the beaches—but the mayor argues that this will destroy the tourist season, leading to lots of shark-related mayhem. If that sounds a little familiar, it should. Rent Jaws on Prime Video.
If you want to immerse yourself in a strange, vaguely threatening island community that harbors a dark secret, watch The Wicker Man (the original, not Nic Cage’s delirious 2006 remake). When a police officer from the mainland travels to isolated Summerisle to investigate a report of a missing girl, he finds himself in an odd pagan community that doesn’t welcome him, but also doesn’t outright threaten him. But it’s all part of a plan, and as he explores the island, the darkness under its facade slowly envelopes him, leading to one of the most striking and horrifying final images of a film ever. Christopher Lee’s deranged performance as Lord Summerisle is like a photo negative of Matthew Rhys’ Tom Loftis, and is worth the price of admission by itself. Stream The Wicker Man on Kanopy or rent it on Prime Video.
If it was the perfect balance between humor and horror that pulled you into Widow’s Bay, check out this often-overlooked 1996 film directed (and co-written) by Peter Jackson. Frank Bannister (Michael J. Fox) gained the ability to perceive and interact with ghosts after a tragic accident that killed his wife. Consumed by guilt, Frank uses this power to con people into paying him to exorcise their homes after his ghostly buddies “haunt” them for a while. When a spectral figure dressed like the Grim Reaper begins killing people in ways that implicate Frank, he has to overcome his self-loathing and terror to save innocent lives. It combines a goofy sensibility with real scares in a way fans of the show will love. Rent The Frighteners on Prime Video.
Another film that somehow manages to be both effortlessly hilarious and legitimately scary, Shaun of the Dead tracks a zombie outbreak in a sleepy section of London. Shaun (Simon Pegg) isn’t exactly excelling at life, but when the zombies come for the people he loves, he’s all in for the fight—just not always very good at it. Filled with callbacks, brick jokes, and genuinely affecting performances, there’s a reason this movie (and Edgar Wright’s entire so-called Three Flavours Cornetto trilogy, actually) is a must-watch for both horror and comedy fans. Rent Shaun of the Dead on Prime Video.
This week's dispatch has it all: the indomitable spirit of internet perverts, a genre of music named for its awfulness, slang that didn't come from African-American vernacular English, and a couple of YouTubers who out-earned Star Wars. As always, I've done the deeply undignified work of digging into it so you don't have to. You're welcome.
TikTok is serious about keeping people from using its service to expose their parts, but the ingenuity of perverts should not be underestimated. Enter the matcha filter. Officially named "wild hearts whisper," unofficially named for the thick, hazy green wash it throws over videos, the matcha filter showed up on the 'Tok a few weeks ago and user quickly discovered it defeats the platform's nudity filters. Here's a non-nudie video of it in action:
Using the matcha filter is called out in the comments on those videos, and TikTok's moderation team works hard at the endless "prevent nudity" game of whack-a-mole, so don't be shocked if the filter is "fixed" by the time you read this. That's the rhythm of moderation: Someone discovers a gap in the algorithm; it gets famous, gets patched, and everyone moves on to the next thing.
I love obscure musical sub-sub-sub genres, and you don't get more obscure than UK Poo. Basically, it's British hip-hop that is very, very bad. I've never heard any British hip-hop that was good, but UK Poo is next level: the very bottom of the cultural garbage heap of Brits trying to make hip hop. Think gratuitous auto-tune, way-too-complicated hi-hat patterns, and "I shot it in me flat!" style videos. Here's an example:
That's "Belfast" by D Rebal. It is not good. Here's another example, Savvy's "Slap Me Again." Also, not good.
The word was coined by X user @cashyams, and they meant it as an insult. But it seems to be morphing into something a little deeper than just bad music. People are finding DIY British hip-hop that they actually like and labeling it "UK Poo," for instance, as you can see in this video:
Then there are the people who think they're being clever by labeling music "UK Poo" when it's clearly a joke they aren't getting.
And my favorite: people who try to redeem bad art through re-mixing. Such kindness.
Did they succeed? No. But it's at least better.
"Cortisol-coded" describes something that is meant to be relaxing but is actually stressful, like watching a video of the 5 a.m. yoga routine of a perfectly styled influencer surrounded by $400 candles. It comes from the "wellness" parts of TikTok, where "performative relaxation" is an epidemic.
Cortisol is one of the hormones our body uses to manage stress, both mental and physical. When your body is dealing with ongoing threats or health conditions, it's good—lifesaving, even. There are also conditions where too-high cortisol can be a problem. Because of its association with stress, “cortisol” has become a shorthand for anything stressful.
Some TikTok influencers blame almost every physical ailment on "high cortisol," leading opportunists to start selling supplements designed to lower cortisol, which either don't work, or, worse, work. You don't want to be messing around with your endocrine system unless you really know what you're doing.
There's a lot of understandable hand-wringing over how being steeped in internet content all day, basically from birth, will affect younger people, but the early returns are coming in, and it seems to have worked out for horror filmmakers at least. The horror film genre belongs to a couple of Gen-Z directors whose talent and sensibilities were forged in the crucible of YouTube. Together, Curry Barker and Kane Parsons have made over $800 million from a couple of low-budget horror movies this summer.
Barker is 26 years old. A sketch comedian who dropped out of film school, Barker made an $800 found-footage movie for YouTube, and parlayed its success into a sub-million dollars feature, Obsession, that has so far made a staggering $474 million. It's actually good, too. You should watch it.
Parsons, aka Kane Pixels, based his first feature, Backrooms, on his YouTube series of the same name which was itself based on a 4chan creepypasta. It is also good, although Backrooms "only" made $393 million on a budget of $10 million. But then, Parsons is 20 years old, so he has some time to catch up.
This week's viral videos have a simple premise that leads to limitless possibilities, like the best meme formats. Here's how it works: Two friends sit next to each other. One says, "I never told you this, but I like you better with..." followed by a ridiculous request (like a weird outfit or a strange prop). The video jump-cuts to person two having made the requested change. They fire back an even more absurd demand and everything gets weirder and weirder. It is low-effort, occasionally amusing, and a stark reminder that young people have entirely too much free time. Here are some examples:
On Monday, Meta announced its latest AI model: Muse Glimmer. The company isn't necessarily advertising this model as its biggest and best, however. While Meta has plenty to say about Muse Glimmer's performance potential, there are two key selling points for the model in the company's eyes: The fact that it's open-weight, and that it's designed to run locally on "consumer" Macs and PCs—though perhaps not on your computer.
According ot Meta, Muse Glimmer is a 30-billion-parameter AI model optimized for "always-on local agent workflows." What that means is Glimmer is designed to run autonomously directly on your machine, rather than outsourcing that processing to the cloud. Local AI is much more secure than cloud-based, as your data stays within your system. When you tap into cloud servers, you're at the mercy of whichever company owns that network, potentially putting your data in jeopardy.
While there are many use cases for a model like Muse Glimmer, Meta is emphasizing its use for agentic AI. AI agents can perform tasks on your behalf and are growing in popularity, as users and developers can ask their agents to do any number of functions. Running agents locally is also quite popular, as evidenced by OpenClaw's infamous year, but the most demanding models also require powerful hardware. If you're working with a laptop, you might not have the energy necessary to run AI agents.
That's part of Meta's pitch here. The company says Muse Glimmer was trained to balance performance with limited hardware. Meta says it puts Muse Glimmer through three sets of training to achieve this: a "Pre-Training," which trained Muse Glimmer on Muse Spark's outputs; "Mid-Training," which trained the model on more "agent-heavy" data; and "Post-Training," which fine-tuned the model across a number of parameters. Meta says Muse Glimmer can handle end-to-end agentic task completion, multi-step reasoning, recovery when a tool call fails, and multi-model input and reasoning, among other functions. It is also trained on data from over 100 languages. The company says the model performs "strongly for its size class" against many standard LLM benchmarks.
While Muse Glimmer isn't as demanding as other models, it likely won't run well on the average laptop. Meta says that a 30-billion parameter model would typically require more than 55GB of memory, but through "quantization techniques," Muse Glimmer can use less than 20GB. When you take its "working memory" into consideration, Muse Glimmer will likely use 24GB or 32GB of memory. If you have a "Pro" MacBook or PC, that might suit you well. But if you're like me, your laptop has 16GB or less of RAM—not quite what Muse Glimmer requires. Still, it's a step forward for local AI use. (For reference, Meta says it tested Muse Glimmer on MacBook M4 Max, M5 Max, and with an RTX-5090 GPU.)
Meta says Muse Glimmer is "open," which means Meta released the model's training weights. As such, not only can anyone use the model, but they can tune it as well. If you have specific needs for your AI model, you can tinker with Muse Glimmer to better serve your usage.
There are a lot of AI models on the market these days, but Meta only has direct comparisons with two other models: Google's Gemma4-31b and Alibaba's Qwen3.6-27B. According to Meta's benchmarks, Muse Glimmer beats both models in the following 12 tests: MCP Atlas, DeepSearch QA, 𝜏³-banking, WildClawBench, GAIA2, SWE-Bench Pro, Sci Code, Charxiv Reasoning, IFBench, AIME 2026, AA-LCR, and Beam 128K. Gemma4-31b still has the crown in four of the benchmarks, while Qwen3.6-27B leads in eight.
Meta didn't provide similar comparisons with models from industry leaders, like OpenAI, DeepSeek, Z.ai, or Moonshot AI. However, Artificial Analysis has a ranking of all open-weight models on the market. As of this article, Muse Glimmer (high) is actually in 18th place. Moonshot AI's Kimi K3 (max) is in the lead, with Z.ai's GLM-5.2 (max) in second, DeepSeek's V4 Flash (max) in third, and Kimi K3 in (low) in fourth. For reference, Gemma4-31b is in 32nd, while Qwen3.6-27B is actually in 17th. There are many variables here, but this leaderboard gives us a quick glance at how Meta's latest model ranks amongst the competition.
You can try Muse Glimmer right now by downloading the weights from Hugging Face. Meta says the model will be available in the coming days from apps like Ollama, LM Studio, and Unsloth.
We may earn a commission from links on this page.
Google Assistant is about to get one heck of a headstone in the Google Graveyard. The voice assistant has been a handy tool for years, but the company has finally decided to replace it with Gemini entirely. If you’re not exactly thrilled about the AI-ification of every app you use, there are a few ways to make Gemini behave a bit more like the Assistant of yore.
The switch from Assistant has been coming for a while; it’s a seismic shift for a lot of devices. On top of deprecating Assistant on Android phones and tablets, it’s also disappearing for Wear OS devices and some Android Auto cars. Some smart home devices and TVs will still support Assistant, but you can switch to Gemini (and you might want to since the choice is likely to be made for you eventually).
Regardless of how you feel about AI chatbots, using Gemini might become unavoidable if you’re in the Google ecosystem. So, if you want to keep the functionality you had in your phone, car, or kitchen, it might take a little bit of effort. Here’s how to get started.
When September rolls around, you’re going to be pushed to Gemini on your Google-powered phone, tablet, and watch, but that might not cover all your devices. If you’re going to be reprogramming your muscle memory, you may as well do it all at once.
Depending on when you set up your Google Home or Nest devices, you might’ve gotten a prompt to upgrade to Gemini. If not, open the Google Home app, tap your profile picture, and choose “Home settings.” Here, you should find a banner that lets you set up Gemini for Home.
It might sound counterintuitive to move away from Assistant in order to make Gemini work more like Assistant, but in the next sections I’ll walk through how to rebuild your favorite Assistant commands. This step will make the whole system work more consistently, but activating Gemini on your smart speakers is technically optional. Just be aware that if you opt not to do this, you might hit some bumps if your phone and smart speaker try to compete to respond to your commands with different assistants.
Part of the appeal of Google Assistant was its simplicity. Turn your lights on and off, set reminders, stuff like that. No trying to plan your meals for you based on your travel itinerary and anime viewing history or whatever. If that kind of simplicity appeals to you, consider taking the opposite of our previous advice, and turn off many of the “Connected Apps” you find there.
In the Gemini app on your phone, you can find this under Settings > Personal Intelligence > Connected Apps. Here, you’ll find a list of apps you can connect (or disconnect) to affect how Gemini responds. You might want to turn off some of these to simplify your assistant’s responses, like Google Workspace or Search Services. These use data connected to your Google account to personalize your results, which might muck up what you’re trying to accomplish.
However, some of these apps should probably stay on. Google Home, for example, is used to control your smart home devices, and Utilities handles things like timers and alarms. Other functions are, in typical Google fashion, a bit convoluted. Reminders, for example, are typically handled by Google apps like Tasks (or sometimes Keep or Calendar?) so you might want to keep Workspace enabled, depending on your needs.
I’ve been singing the praises of this feature for a while, but if you’re frustrated with how Gemini behaves, the most powerful tool to bend it to your will is the Instructions for Gemini (previously Saved Info) tool. While ostensibly this is used to store information about yourself that you want Gemini to keep in mind, you can use it to create custom shorthand commands that work a lot more like how Assistant used to.
To access this, head to Settings > Personal Intelligence > Instructions for Gemini in the Gemini app. Here, you can add custom instructions in natural language. So, for example, let’s say you want to turn off the living room and kitchen lights with a short command, you could add the following instructions: “If a prompt consists solely of ‘good night’, turn off the lights in the living room and kitchen.” You can further tailor this command to make sure it works reliably.
This can also be a handy way to make sure that Gemini uses the right app or tool for the commands you want to run. For example, if you want to add a note to your Google Keep app, but Gemini keeps trying to set reminders via the Tasks app or whatever, you can add custom instructions to “Always use Keep to store notes and reminders” so your future commands work the way you want them to.
Google Assistant’s routines feature has moved around over the years, and it currently lives in the Google Home app. But in keeping with Google’s long-standing strategy of “build two of everything”, you can also use Gemini’s scheduled actions to replace some of this functionality. While it’s not strictly necessary, as long as Google Home automations exist, it can be easier to do everything from the one app you’re being pushed into anyway. Especially if you’re worried that the older method will suddenly be removed.
In the Gemini app, head to Settings > Scheduled actions, and tap New Action. Here you can give your routine a name, describe what you want Gemini to do in the same natural language you can use for other Gemini commands. You can also set a time and frequency you want the action to occur.
Gemini works a little differently than Assistant, so some of these commands might not run with as much precision as you’re used to. Gemini notes in the app that scheduled actions are prepared up to an hour in advance, so keep that in mind. This tool can be handy for things like a morning routine, giving you a morning briefing at the same time as turning on your lights. But if you want very simple actions—like, say, just turning on the lights—you might still want to use the automations in the Google Home app.
One of the biggest annoyances with the forced switch to Gemini is that commands can take a lot longer to process. Recognizing a few key words like “turn on lights” is fast. Trying to process whole sentences is more complicated. If you preferred the speed of Assistant and don’t need more complex analysis of your instructions, changing your model can help.
In most cases, you can’t change the default Gemini model (Google, please), but in the Gemini app, you can tap the model dropdown and choose whatever the latest, lightest model is. At time of writing, that’s 3.5 Flash-Lite, but it may be different later. Typically, the fastest model will be whichever one has the oldest version number, paired with words like “Flash” or “Lite”.
You can also find an optional toggle labeled “Extended thinking” which usually attempts more complex processing. But it also adds extra processing time, dragging out simple commands like setting alarms. Disable this, and you’ll typically get faster answers. Depending on how you invoke Gemini, you might not have the option to tweak your models, but when you can, you can find a pretty significant speed boost by making these changes.
We may earn a commission from links on this page.
Futurama has been one of the smartest and funniest TV shows for so long, it’s managed to get canceled and brought back several times. It’s easy to see why: Every single episode is jam-packed with jokes, and those jokes range from the incredibly stupid to jokes based on science that about four people in the world understand (and they all work in the Futurama writer’s room, apparently). Its bent vision of the future of the universe is always a delightfully deranged adventure, and if you’re looking to bring a little more of that vibe into your life (and you’ve already binged the hilarious TV shows in that category), you should absolutely watch Idiocracy immediately.
Idiocracy and Futurama share a fundamental premise: A man of average intelligence is frozen, wakes up in the far future, and has adventures. In Futurama, Philip J. Fry (Billy West) fits right in with the ludicrous sci-fi universe he finds himself in; but in Idiocracy, Soldier Joe Bauers (Luke Wilson) emerges from hibernation five centuries later to discover that humanity has regressed so drastically he is now the smartest man on Earth—by a long shot. And if you’re thinking that Futurama played around very specifically with that premise, you’d be right—in season 3, episode 7, “The Day the Earth Stood Stupid,” floating alien brains use a beam of energy to turn humanity stupid—but the beam doesn’t affect Fry, because—stay with me here—he is his own grandfather and thus lacks a crucial “delta wave” in his brain function, leaving him briefly the smartest human on the planet.
Idiocracy is more a tightly focused satire than Futurama, using its premise to mock modern problems and behaviors, but it’s very much on the same wavelength as the show. Just like the show, Idiocracy has loads of hilarious jokes about consumerism and how we’re all manipulated by advertising and marketing tricks, and the film is populated by the kinds of ridiculous characters fans of the show will recognize and love—like Dwayne Elizondo Mountain Dew Herbert Camacho (Terry Crews), the deeply stupid and incredibly violent President of the United States (whose hair is a character unto itself in this movie).
Like Fry often does, Joe finds himself responsible for saving the day despite being, well, pretty average in most ways. Exasperated by the dumbness all around him (this is a future America where the death penalty involves monster truck demolition derbies, the Supreme Court has been renamed the Extreme Court, and an energy drink called Brawndo is used for literally everything from medicine to crop irrigation), Joe plods along, doing his best, and (again, just like Fry) eventually scores a win.
For fans of Futurama, there’s no better choice for movie night. Rent Idiocracy on Prime Video.
Need more to watch? The obvious choice would be the actual Futurama films that were released—Bender’s Big Score (2007), The Beast with a Billion Backs (2008), Bender's Game (2008), and Into the Wild Green Yonder (2009). You can find them on Hulu as part of their Futurama collection, but the movies have been cut into episode-length slices. You can also buy each movie individually on Prime Video.
But if you’ve already watched the Futurama movies (and you probably have), here are a few more movies that aren’t Futurama but offer a similarly wacky sci-fi experience.
As a sci-fi series, Futurama has played around with actual time travel several times, so Bill and Ted’s Excellent Adventure is the perfect movie for fans of the show. In San Dimas, California, in 1989, Bill Preston (Alex Winter) and Ted Logan (Keanu Reeves) are slacker high school students in danger of failing history—which would have dire consequences for the rather utopian future of humanity. A man named Rufus (George Carlin) is sent back in time to help by giving them access to a time machine—and hilarious shenanigans ensue. Stream Bill and Ted's Excellent Adventure on Tubi or rent it on Prime Video.
Futurama’s writers love a good movie parody—they’ve hit Willy Wonka, Armageddon, Animal House, and One Flew Over the Cuckoo’s Nest among many others. So one of the best parodies of one of our most cherished sci-fi franchises is the perfect follow-up to the show. It’s a Mel Brooks joint, so obviously this send-up of all things Star Wars (original trilogy only) can lean into the juvenile side, but it remains a demented takedown of a foundational science fiction universe. Plus, the long-awaited sequel is due in 2027, so a refresher is in order. Stream Spaceballs on Fubo or rent it on Prime Video.
In a lot of ways, The Hitchhiker’s Guide to the Galaxy crawled so Futurama could walk. While this film, based on Douglas Adams’ original book, isn’t as pop-culture focused as Futurama, it has the same spirit of dizzy sci-fi craziness, from poetry as a torture device to an Infinite Improbability Drive that warps and bends reality every time it’s used. There are bizarre aliens, smart observations about the nature of our existence, and a rumination on the fundamental importance of towels. Rent The Hitchhiker’s Guide to the Galaxy on Prime Video.
Futurama is funny, but it’s also very sweet-natured—Philip J. Fry may not be the brightest guy, but he means well, and the Planet Express crew almost always does the right thing. WALL-E has that same spirit while also hilariously lampooning modern life, giving us both an adorable, lonely robot marooned on an environmentally ruined Earth and a vision of humanity’s very fleshy, lazy, and kind of dumb future if we don’t change our ways. Stream WALL-E on Disney+ or rent it on Prime Video.
If you love the rapid-fire gags in Futurama and enjoy the absurdity of wildly impractical technology, you’ll love Megamind. A parody of superhero stories, the film follows two alien babies who are sent ot Earth, Superman-style, when their worlds are destroyed. One becomes arch-villain Megamind (Will Ferrell), the other hero Metro Man (Brad Pitt), who battle each other until Megamind finally gets the drop on his rival, killing Metro Man—and deleting all meaning from his life in the process. With stuff like dehydration guns and realistic holographic disguises, Megamind riffs on sci-fi tropes in ways that Futurama fans will love. Stream Megamind on Peacock or rent it on Prime Video.
There are 8.3 billion people on this planet, and roughly an eighth of them use ChatGPT. OpenAI confirmed that its infamous AI chatbot now has over one billion weekly active users, a staggering number given that ChatGPT is only three and a half years old. (For reference, that's on par with global Windows users.) Love it or hate it, there's no denying that ChatGPT—and generative AI as a whole—is firmly a part of modern life.
While there are surely many paying ChatGPT customers, the vast majority are using it for free. Though it may be exciting for paid users to try out the company's latest and greatest models, the biggest impact happens whenever OpenAI makes features available free of charge. As it happens, next week, free users are getting perhaps the biggest ChatGPT update in years.
Starting next week, you'll no longer hit a wall when simply texting with ChatGPT. OpenAI says that both free and Go users now have unlimited text chats, which should mean you can send as many text-based messages as you want without running into cooldown periods. Of course, the emphasis here is on "texts." There are absolutely still limits on other ChatGPT features, including messages with files and images, as well as Voice mode and image generation. But for users who primarily use ChatGPT with text-based messages, this update should be more than welcome.
In addition, OpenAI is switching free and Go users to the company's new GPT-5.6 Luna model, which is designed for "cost-sensitive, high-volume workloads." As such, this model likely won't be as capable as OpenAI's more powerful models available to Pro users, but free users aren't totally left in the lurch: OpenAI says there's a new "Think" button that lets you push Luna further for more complex requests. The experience is much changed from the days when users could jump between different models for free, but perhaps the unlimited text chats will make up for it for most users.
OpenAI says that Plus and Pro users are also getting some updates. The company says that its GPT-5.6 Sol model now delivers "more focused" responses, and may issue corrections instead of simply agreeing with the user when they're wrong. It also does away with some of the formatting we've come to expect from AI: Your answer may be more succinct in paragraph form, rather than expansive with bullet points. OpenAI says that Sol makes "fewer mistakes" too, but remember: all AI models hallucinate.
There's also a new slider to push the model from "Instant" to higher effort. The idea is to give users a sliding scale of power, rather than have them switch from distinct models. The more intense you push the model, the longer the responses may take, so OpenAI positions the slider as an option for users to tailor the model to their needs: If you need a quick answer, keep the slider towards the less intense side.
As one of the biggest tech companies in the world, Google is putting its AI features and platforms into each and every one of its products. As such, it’s no surprise that its wildly popular navigation app, Google Maps, is getting the same treatment. While it's easy to use the app without encountering much AI, if you so choose, you can use Google’s latest AI features to find new restaurants, accommodations, or hot spots in your area. And now, Google is boasting that it is launching new "agentic" features that will even allow the app to order food for you—sort of.
Back in March, the company rolled out “Ask Maps,” a chatbot-like interface in Google Maps that taps into Gemini to help you find relevant places on the map. For example, Google says you can “Ask Maps” to find you a restaurant with availability for six at a specific date and time, keeping in mind strict dietary requirements. It’s about what you’d expect from the combination of Gemini and Google Maps: You “chat” with the map to find what you need.
Now, Google is expanding on these features, introducing new agentic options to reduce the work you actually have to do to pick up lunch (if not by all that much). Starting today, Google says Ask Maps can handle “complex, multi-step tasks," including functions it can handle on your behalf, and the flagship feature is ordering food for you—or, at least, starting an order.
Google says you can launch Ask Maps and type something like “order spicy pad kee mao with seafood for me to pick up on my way home." While you might think such a feature would handle the entire request for you, Ask Maps doesn't quite get there. It will, according to Google, look for open restaurants on your route home, and it can even take into consideration your previous orders and dietary needs. But once it floats options, you still need to choose the restaurant you want to order from. Once you do, the agentic features kick in: Google says that Ask Maps will remember your original request and will begin adding the appropriate items to your cart through services like Square and Toast (with Uber Eats support coming soon). However, Ask Maps won't actually order the food for you. You still need to manually review what's been placed in your cart and approve the order before the restaurant will receive it.
As part of this Ask Maps update, you can also ask about hotels and local events. Google suggests asking the feature something like “For next week’s conference in downtown Miami, find me a decently priced, top-rated hotel with an artsy vibe, within walking distance from a gym and restaurants.” That's certainly a complex request, and, according to Google, Ask Maps will be able to compare prices and availability to find an appropriate hotel. The same goes for things like concerts and comedy shows: Ask Maps will surface results based on your request, but you'll need to click through to actually buy tickets yourself.
I don't have these options on my Google Maps app yet, so I can't personally attest to how well they work, but I'm not sure I'm seeing the vision here. Companies like Google have been promising that agentic features will save us time and rescue us from having to perform tedious or menial tasks. And while this feature does save the 30 seconds or so it takes to building a restaurant order, that's about all it does. Not only do you still have to intervene in multiple aspects of the process (including the actual ordering of the food), it seems like the whole ordeal takes longer than it should. In my view, heading to the restaurant's site, building the order myself, then placing the order (as I'd have to do anyway) is likely more efficient than asking Google's AI to find a restaurant with a specific dish, have it build the order, then have me review it.
Maybe I'll feel differently once I can try it out, and maybe this is just one step into a future where you really can have Google Maps complete a whole order for you. But I'm not sure I'd ever want it to, especially when dealing with ordering multiple items for various people with potential dietary restrictions. Not only would it likely be faster to do it myself, but I'm not sure I'd trust an AI with all those complications. The last thing anyone needs is for an AI to hallucinate and forget to order something with an allergy label.
This update also incorporates Google's "Personal Intelligence," which pulls info from your various Google apps to return more personalized results. As such, you could Ask Maps to find you things to do before your afternoon flight. Google Maps should be able to access Gmail and Google Calendar, where it should find that flight's time and location, in order to fulfill your request with more individualized results than a simple search would. This feature is off by default.
There's also a new live data widget for transit instructions, so you can keep tabs on public transit information in real time. You can also suggest Google Maps edits conversationally through Ask Maps, including adding photo attachments.
"Agentic" is the tech buzzword of the moment. The phrase describes an AI tool you can ask to perform tasks automatically on your behalf. Maybe the bot will order your favorite dish on DoorDash, or call a car for you without you needing to open the Uber app. Perhaps in a few years, AI assistants will be scheduling your meetings, or screening interviewees for a job based on your direction—though I'm not sure we've cracked the simple tasks just yet.
Regardless, most companies that make AI technologies are pushing these agents in some way, shape, or form—including into their AI-powered browsers. Companies like Perplexity have browsers that offer agentic features that invite you to task your AI with performing tasks on the web. Rather than shop on Amazon yourself, perhaps you can ask an AI assistant to do it. Instead of painstakingly compiling sources for a paper, maybe your bot will find them for you. That's the gist, anyway.
However, despite the focus on agentic AI, one of the biggest AI companies in the world is shuttering its own AI browser. OpenAI has announced it will soon discontinue ChatGPT Atlas, a venture it only launched last October. The shutdown comes just months after the company also shut down Sora, its AI-generated short-form video platform, marking the second high-profile OpenAI product of 2026 to go under.
OpenAI will officially shut down ChatGPT Atlas on Sunday, Aug. 9. While it won't delete the browser from your devices, it may stop working as intended. OpenAI says that after this date, ChatGPT Atlas may not open, browse, or support "browser-based agentic workflows." As such, the company encourages users to export or save all of their ChatGPT Atlas data, including bookmarks and pages.
In addition, ChatGPT Atlas won't receive security updates going forward. That makes it especially important to stop using it after Aug. 9: While all browsers and devices that connect to the internet should be running the latest security patches as soon as they're available, AI browsers are particularly vulnerable. Because they have agentic features, hackers can take advantage of the automated AI to trick it into performing tasks on their behalf instead of yours. A bot can't tell the difference between your commands and a malicious prompt hidden on a website it browses. These browsers are already a security risk—ChatGPT Atlas itself had a major vulnerability at launch—but using an AI browser that is no longer supported is beyond dangerous.
If you ask OpenAI, you don't have to leave its sandbox in order to play with AI browsing features. The company is pointing users to its ChatGPT desktop app, which includes some browsing features, like multiple tabs, downloads, account login support, and "improved navigation." To me, those sound like perks for active ChatGPT users, but not a replacement for a dedicated browser. That said, there is a feature called "Computer Use" that lets ChatGPT take over certain apps on your Mac, including your browser.
However, if you're looking for a true ChatGPT Atlas replacement, you have options, the most notable of which is probably Perplexity Comet, one of the first agentic browsers on the market. Opera also makes its own AI browser, called Neon, if you're already a fan of that company's offerings. If you used Arc before it went belly-up a few years back, you might be happy to learn the company now develops an AI browser called Dia. Any of these will be a more full-featured replacement for ChatGPT Atlas than settling for the ChatGPT app.
Lateo.net - Flux RSS en pagaille (pour en ajouter : @ moi) 