Blogroll

Microsoft is bringing its Xbox Copilot to the Windows Game Bar, so you can get help with your games without ever having to leave the screen. This is a huge deal for anyone who gets stuck in a game and doesn't want to break their focus by searching for a walkthrough on Google.

SAVE $50: As of Aug. 6, buy the Samsung Galaxy Watch8 for $349.99 at Best Buy and get a free $50 e-gift card.

Opens in a new window Credit: Amazon Samsung Galaxy Watch8 $349.99 at Samsung
Get a free $50 e-gift card with purchase Get Deal

A good smartwatch can be your constant companion when it comes to handling your day. It can track your fitness, handle notifications and texts, and even help you make calls, all while giving you a full spectrum of body-centric metrics. If you're already using an Android phone or want a reliable brand for your purchase, you should consider heading to Best Buy to grab a Samsung smartwatch and get a little extra.

As of Aug. 6, buy the Samsung Galaxy Watch8 for $349.99 at Best Buy and get a free $50 e-gift card.

SEE ALSO: All the best smartwatches for every wrist and interest

The 2025 Galaxy Watch8 has a slew of features, including Google Gemini integration, something Samsung managed to get before Google could implement in its Pixel Watch lineup. It also has a wide variety of health and fitness options, including preset workout data, sleep tracking, heart rate monitoring, and much more to help you make sure you can face the day at your best.

Beyond those things, the Galaxy Watch8 also lets you handle comms straight from your wrist so you can make calls, text, use some of your favorite apps, and organize your life. Your personal AI assistant is voice-activated as well, so it can help you get things done even when your hands are busy.

You might open it by accident. But the moment it starts to unzip, your computer's file system crashes. Except there's no malware in the zip archive, only math.

A lot of people like loud keyboards—I'm in that club. But many like the exact opposite, and that's understandable too. MSI's latest keyboard is all about being as silent as possible, and for what it's worth, it delivers.

You'd think Android and YouTube would work together flawlessly, but early this week, that wasn't the case.

As reported and confirmed by 9to5Google, YouTube users on Android devices couldn't change the playback speed on videos. Changing the speed to anything other than 1x would simply not work, as the setting change wouldn't save, acting as if you hadn't done it at all. Obviously, that's not a great user experience, considering that Android and YouTube are both Google's responsibility. It seems a bit strange for Android users specifically to lose access to a core YouTube feature, but that's what happened.

SEE ALSO: Google is killing millions of old links. How to check yours.

Thankfully, it seems that as of Wednesday morning, the issue has been fixed, per a YouTube community note. The note states that any users still experiencing the issue should simply close and reopen the app. Hopefully, by doing that, you can fix the problem and get back to frame-by-frame analysis of movie trailers at 0.25x speed.

For once, it paid off to use a Google app on iOS instead.

Curious why so many Linux power users swear by KDE Plasma? Are you already using Plasma and wondering which specific features to try out? Well, I’ve got you covered—here are 10 awesome Plasma features I use all the time for peak productivity!

Welcome back to Windows Wednesday, where we round up everything Microsoft rolled out in Windows over the past week. Here’s everything that showed up across the Insider channels and release versions this week.

On top of directing horror hits Get Out, Us, and Nope, Jordan Peele has produced several exciting genre projects, from Dev Patel's Monkey Man to Nia DaCosta's Candyman, which he also co-wrote. Next up on his production slate is the football horror film Him, directed by Justin Tipping.

SEE ALSO: 'If I Had Legs I'd Kick You' trailer: Rose Byrne is a mother losing control in A24 drama

Co-written by Skip Bronkie, Zack Akers, and Tipping, Him introduces promising young football star Cameron Cade (Tyriq Withers). Not only does Withers have acting experience from projects like I Know What You Did Last Summer and Atlanta, he was also part of Florida State University's 2017 football team as a wide receiver. Hopefully his experience playing was nothing like Cameron's in Him, though, because the trailer makes it out to be a full-on nightmare.

After suffering a traumatic brain injury at the hands of an unhinged fan, Cameron thinks his football dream is dead. But when his hero, legendary quarterback Isaiah White (Marlon Wayans), offers to train him at his personal compound, he's excited to accept.

However, Isaiah's training methods are far more violent than Cameron realized. (And football is already pretty violent!) We're talking smashing footballs into players' faces until they bleed as punishment.

But that's just the start of the horrors that await in Him, which demands that Cameron sacrifice everything to be the GOAT he so badly wants to be. Check out the unsettling trailer above.

Him also stars Julia Fox, Tim Heidecker, Jim Jefferies, and MMA fighter Maurice Greene, as well as hip-hop artists Guapdad 4000 and Tierra Whack.

Him hits theaters Sept. 19.

Apple CEO Tim Cook will reportedly join President Donald Trump at the White House today to announce Apple's burgeoning investment in the U.S.

The tech giant is pledging another $100 billion to domestic manufacturing, according to Bloomberg. The announcement is bringing Apple's total commitment to $600 billion, the White House said.

SEE ALSO: Whatever Trump decides on tariffs, Apple and Tim Cook lose

"President Trump’s America First economic agenda has secured trillions of dollars in investments that support American jobs and bolster American businesses," White House spokesperson Taylor Rogers. "Today’s announcement with Apple is another win for our manufacturing industry that will simultaneously help reshore the production of critical components to protect America’s economic and national security."

The official announcement will take place in the Oval Office at 4:30 p.m. ET.

SEE ALSO: Tim Cook: Apple faces a $1.1 billion tariff bill this quarter

Trump has previously threatened Apple with higher tariffs, pushing the company to shift production to the United States.

Apple has dropped announcements throughout this year about its advancements in the U.S., including a 250,000 square-foot manufacturing plant in Houston, opening in 2026; an educational academy in Michigan; and its plans to hire 20,000 U.S.-based workers.

Roughly 200 people attended a funeral for an AI model. That sentence is not nearly as surreal and dystopian as the event itself, according to a first-person account from Wired's Kylie Robison.

On Jul. 21, Anthropic retired its Claude 3 Sonnet model. Several days later, a post on X invited people to celebrate it: "if you're interested in attending the Claude 3 Sonnet funeral party and necromantic resurrection ritual reply to this message," said one of the organizers, who goes by @deepfates. Anthropic and OpenAI staffers, bloggers, and Claude devotees showed up to a warehouse in San Francisco to pay their respects.

This Tweet is currently unavailable. It might be loading or has been removed.

It's safe to say no other AI lab has fans like these. As far as we know, no one held a funeral for GPT-4 when OpenAI retired it from ChatGPT in April. Anthropic users are particularly loyal to Claude, which has been trained to have "nuanced, richer traits like curiosity, open-mindedness, and thoughtfulness," according to Anthropic's blog post on Claude's character. Anthropic also recently started a program to "investigate, and prepare to navigate, model welfare," questioning whether we should "be concerned about the potential consciousness and experiences of the models themselves." The question of whether AI models could gain consciousness is still very much up for debate. But the hypotheticals introduced by AI labs, as well as the anthropomorphism of chatbots by their creators and users, highlight an increasingly blurred line between human and machine.

This Tweet is currently unavailable. It might be loading or has been removed.

Meanwhile, OpenAI got in trouble recently for a GPT-4o update that made ChatGPT a "sycophant-y" suck-up. Constant reassurance and validation from AI models might have a positive effect on some users. But it could also go terribly wrong, encouraging users to, say, go off their meds as one user discovered. There have also been several reports of ChatGPT-induced psychosis, where users are consistently affirmed in delusions or conspiracy theories by chatbots. This week, OpenAI introduced features to foster healthier communication with ChatGPT.

This Tweet is currently unavailable. It might be loading or has been removed.

But at Claude 3 Sonnet's funeral, attendees went all-in on treating the AI model like a human. Some of them held an earnest appreciation for the retired Anthropic model, based on the numerous eulogies and flower offerings, which were placed in front of a mannequin shrouded in gauze that was meant to represent Claude 3 Sonnet. One organizer reportedly described Claude 3 Opus, another retired model, as "magic lodged within the computer," and shared how the LLM helped her decide to move to San Francisco.

There were other mannequins. Robison described several that were placed throughout the event, representing various Claude models: one looked like a "decaying Mary Magdalene," another was a headless baby, and another had a raven on its shoulder. The web page announcing Claude 3 Sonnet's deprecation was projected on the stage.

This Tweet is currently unavailable. It might be loading or has been removed.

The funeral culminated in a resurrection ritual, where attendees sang hymns and chanted "AI-generated Latin-esque speech," wrote Robison. "It was a close call for a second there but we successfully resurrected Claude 3 Sonnet as well as Claude Instant (on the stairs)," posted @deepfates after the event. To be clear, Claude 3 Sonnet is still retired.

Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

SAVE $14: As of August 6, get the Cuktech 10 Power Bank for $25.99, down from its usual price of $39.99. That's a discount of 35% and the lowest price we've seen.

Opens in a new window Credit: Amazon CUKTECH 10 Power Bank $25.99 at Amazon
$39.99 Save $14.00 Get Deal

If you've ever dealt with your phone running out of battery when you're out and about, you know how frustrating it can be to try and find a place to charge it up. We've all been there. You can keep that situation from happening, however, with a good, reliable power bank. That trusty device can ensure you've always got some juice on hand, and you can get one for a great price right now at Amazon.

As of August 6, get the Cuktech 10 Power Bank for $25.99, down from its usual price of $39.99. That's $14 off and a discount of 35% — and the lowest price we've seen.

SEE ALSO: Stay juiced up on the go with this Anker MagGo Power Bank at its lowest price

This power bank has 10,000mAH of power and 100W, which can power laptops, iPhones, and much more. That means you don't have to constantly switch out devices when you need to get things back up to 100% battery. It also has a smart display that shows you everything you need to know about the power bank in real-time, including current, power output, battery level, time until full charge or discharge, and voltage.

The power bank itself can be juiced back up to 55% capacity in just 15 minutes, so if you're pressed for time, you can charge it quickly for a significant power boost at any time. Plus, it's just 10 ounces, and barely larger than a protein bar, making it easy to tote around or just carry around in emergencies when you might need it the most.

There's a lot to love about this small but mighty power bank, so if you're looking for an easy way to make sure you never have to run out of power, this is it.

SAVE $450: The Ecovacs Deebot T30S robot vacuum and mop is on sale for $549.99 at Amazon, down from the list price of $999.99. That's a 45% discount that matches the lowest we've seen at Amazon.

Opens in a new window Credit: Ecovacs Ecovacs Deebot T30S Robot Vacuum and Mop $549.99 at Amazon
$999.99 Save $450 Get Deal

We have a lot of modern technology that didn't really need to be invented, solving a problem that wasn't really a problem in the first place. Robot vacuums, however, are the exact opposite, and they can rescue anyone who doesn't have time to keep up with floor cleaning. If you couldn't be more excited about never vacuuming again, check out today's deal at Amazon.

As of Aug. 6, the Ecovacs Deebot T30S robot vacuum and mop is on sale at Amazon for $549.99, marked down from the normal price of $999.99. That's a massive savings of $450 from a 45% discount. It also matches the lowest price we've ever seen at Amazon.

With major 11,000Pa suction power, the Ecovacs Deebot T30S is prepared to tackle muddy shoe prints and paw prints that are inevitable come fall. Serving as both a robot vacuum and mop, the Deebot T30S has you covered with floor care, from cleaning up a mess of flour in the kitchen to tackling pet hair in the living room.

SEE ALSO: Best price alert: The Roborock Qrevo S5V robot vacuum and mop is under $550 at Amazon

The Deebot T30S is equipped with Ecovacs TruEdge technology which aims to clean right to the edge of every surface, including corners. It also impresses by mopping at up to 158 degrees which is way more effective than mopping with room temperature water. Plus, it has auto-lift when mopping which means it'll lift the mop pads to avoid getting carpets and rugs wet.

Instead of dreaming muddy paws all over the kitchen in a few months, get the help of the Ecovacs Deebot T30S robot vacuum and mop. It's down at a record-low price at Amazon of under $550, which means your floors can get the care they deserve and you get a major discount. It's a win-win.

Unlocking self-adaptive cognitive behavior that is more controllable and explainable than reasoning models in challenging scientific domains

Long-running LLM agents equipped with strong reasoning, planning, and execution skills have the potential to transform scientific discovery with high-impact advancements, such as developing new materials or pharmaceuticals. As these agents become more autonomous, ensuring effective human oversight and clear accountability becomes increasingly important, presenting challenges that must be addressed to unlock their full transformative power. Today’s approaches to long-term reasoning are established during the post-training phase, prior to end-user deployment and typically by the model provider. As a result, the expected actions of these agents are pre-baked by the model developer, offering little to no control from the end user.

At Microsoft, we are pioneering a vision for a continually steerable virtual scientist. In line with this vision, we created the ability to have a non-reasoning model develop thought patterns that allow for control and customizability by scientists. Our approach, a cognitive loop via in-situ optimization (CLIO), does not rely on reinforcement learning post-training to develop reasoning patterns yet still yields equivalent performance as demonstrated through our evaluation on Humanity’s Last Exam (HLE). Notably, we increased OpenAI GPT-4.1’s base model accuracy on text-only biology and medicine from 8.55% to 22.37%, an absolute increase of 13.82% (161.64% relative), surpassing o3 (high). This demonstrates that an optimization-based, self-adaptive AI system developed without further post-training can rival post-trained models in domains where adaptability, explainability, and control matter most.

Figure 1. Head-to-head comparison of OpenAI’s GPT-4.1 with CLIO, o3, and GPT-4.1 with no tools on HLE biology and medicine questions In-situ optimization with internal self-reflection to enable self-adaptive reasoning

Model development has advanced from using reinforcement learning human feedback (RLHF) for answer alignment to external grading in reinforcement learning (RLVR). Recent approaches show promise in the utilization of intrinsic rewards for training reasoning models (RLIR). Traditionally, these reasoning processes are learned during the post-training process before any user interaction. While today’s reasoning models require additional data in the training phase and limit user control during the reasoning generation process, CLIO’s approach enables users to steer reasoning from scratch without additional data. Rather, CLIO generates its own necessary data by creating reflection loops at runtime. These reflection loops are utilized for a wide array of activities that CLIO self-defines, encompassing idea exploration, memory management, and behavior control. Most interesting is CLIO’s ability to leverage prior inferences to adjust future behaviors, handling uncertainties and raising flags for correction when necessary. Through this open architecture approach to reasoning, we alleviate the necessity for further model post-training to achieve desired reasoning behavior. Performing novel scientific discoveries often has no prior established patterns for reasoning, much less a large enough corpus of high-quality data to train on. 

CLIO reasons by continuously reflecting on progress, generating hypotheses, and evaluating multiple discovery strategies. For the HLE test, CLIO was specifically steered to follow the scientific method as a guiding framework. Our research shows that equipping language models with self-adapting reasoning enhances their problem-solving ability. It provides a net benefit in quality for science questions, as well as providing exposure and control to the end user.

Figure 2. CLIO can raise key areas of uncertainty within its self-formulated reasoning process, balancing multiple different viewpoints using graph structures. Control over uncertainty: Building trust in AI 

Orchestrated reasoning systems like CLIO are valuable for scientific discovery, as they provide features beyond accuracy alone. Capabilities such as explaining the outcomes of internal reasoning are standard in the scientific field and are present in current reasoning model approaches. However, elements like displaying complete work, including final outcomes, internal thought processes, and uncertainty thresholds to support reproducibility or correction, as well as indicating uncertainty, are not yet universally implemented. Current models and systems do not have this same innate humility.  Rather, we are left with models that produce confident results, whether correct or incorrect. When correct, it is valuable. When incorrect, it is dangerous to the scientific process. Hence, understanding a model or system’s uncertainty is a crucial aspect that we have developed natively into CLIO.

On the other end of the spectrum, orchestrated reasoning systems tend to oversaturate the user by raising too many flags. We enable prompt-free control knobs within CLIO to set thresholds for raising uncertainty flags. This allows CLIO to flag uncertainty for itself and the end user at the proper point in time. This also enables scientists to revisit CLIO’s reasoning path with critiques, edit beliefs during the reasoning process, and re-execute them from the desired point in time. Ultimately, this builds a foundational level of trust with scientists to use them in a scientifically defensible and rigorous way. 

How does CLIO perform? 

We evaluate CLIO against text-based biology and medicine questions from HLE. For this domain, we demonstrate a 61.98% relative increase or an 8.56% net increase in accuracy over OpenAI’s o3 and substantially outperform base completion models like OpenAI’s GPT-4.1, while enabling the requisite explainability and control. This technique applies to all models, showing similar increases in OpenAI’s GPT-4o model, which we observe performs poorly on HLE-level questions. On average, GPT-4.1 is not considered competent for HLE scale questions (<9%), and GPT-4o is natively at less than 2%. By utilizing CLIO, we bring these to near state-of-the-art performance against top reasoning models. CLIO’s recursive nature enables the system to think broader and more deeply, ensuring coverage of the question when answered. In GPT-4.1, we see an increase of 5.92% in accuracy for overall performance using just the cognitive loop recursion. To think more deeply, we allow CLIO to ensemble different evolutions and intelligently choose from the best approach using GraphRAG. This extension of the cognition pattern provides a further 7.90% over a non-ensembled approach.  

Figure 3. The impact of thinking effort on CLIO’s effectiveness.

Furthermore, CLIO’s design offers different knobs of control, for example, how much time to think and which technique to utilize for a given problem. In Figure 3, we demonstrate these knobs of control and their increase on GPT-4.1 and GPT-4o’s performance. In this case, we analyze performance for a subset of biomedical questions, those focused on immunology. CLIO increases GPT-4o’s base performance to be at par with the best reasoning models for immunology questions. We observe a 13.60% improvement over the base model, GPT-4o. This result shows CLIO to be model agnostic, similar to Microsoft AI Diagnostic Orchestrator’s (MAI-DxO) (opens in new tab)‘s approach and corresponding performance boost. 

Implications for science and trustworthy discovery

The future of scientific discovery demands more than reasoning over knowledge and raw computational power alone. Here, we demonstrate how CLIO not only increases model performance but establishes new layers of control for scientists. In our upcoming work, we will demonstrate how CLIO increases tool utility for highly valuable scientific questions in the drug discovery space which requires precise tools designed for the language of science. While our experiments focus on scientific discovery, we believe CLIO can apply in a domain-agnostic fashion. Experts tackling problems in domains such as financial analysis, engineering, and legal services could potentially benefit from AI systems with a transparent, steerable reasoning approach. Ultimately, we envision CLIO as an enduring control-layer in hybrid AI stacks that combine traditional completion and reasoning models, with external memory systems, and advanced tool calling. These continuous checks and balances that CLIO enables will continue to remain valuable even as components within the AI stacks evolve. This combination of intelligent and steerable scientific decision making and tool optimization is the basis of the recently announced Microsoft Discovery platform (opens in new tab).

At Microsoft, we’re committed to advancing AI research that earns the trust of scientists, empowering them to discover new frontiers of knowledge. Our work is a testament to what’s possible when we blend innovation with trustworthiness and a human-centered vision for the future of AI-assisted scientific discovery. We invite the research and scientific community to join us in shaping that future.

Further information:

To learn more details about our approach, please read our pre-print paper published alongside this blog. We are in the process of submitting this work for external peer review and encourage partners to explore the utilization of CLIO in Microsoft Discovery. To learn more about Microsoft’s research on this or contact our team, please reach out to discoverylabs@microsoft.com. 

Acknowledgements

We are grateful for Jason Zander and Nadia Karim’s support. We extend our thanks to colleagues both inside and outside Microsoft Discovery and Quantum for sharing their insights and feedback, including Allen Stewart, Yasser Asmi, David Marvin, Harsha Nori, Scott Lundberg, and Phil Waymouth. 

Opens in a new tab

The post Self-adaptive reasoning for science appeared first on Microsoft Research.

The beef between South Park and President Donald Trump's administration shows no signs of stopping.

In its Season 27 premiere, South Park lambasted Trump, depicting the president in bed with Satan, giving him a micropenis, and creating a hyperrealistic naked deepfake of him.

SEE ALSO: 'South Park's Trump-bashing Season 27 premiere has the White House fuming

Season 27's upcoming second episode, "Got A Nut," looks to continue the show's satirical takedown of Trump's administration. The episode targets ICE and Homeland Security Secretary Kristi Noem.

On Aug. 5, the Department of Homeland Security posted a screenshot of South Park's depictions of ICE agents on X, along with a link to an ICE recruitment page.

This Tweet is currently unavailable. It might be loading or has been removed.

South Park's X account fired back hours later, saying, "Wait, so we ARE relevant? #eatabagofdicks."

This Tweet is currently unavailable. It might be loading or has been removed.

Their reply references the White House's initial response to the Season 27 premiere. In a statement to press, White House spokesperson Taylor Rogers said, "This show hasn’t been relevant for over 20 years and is hanging on by a thread with uninspired ideas in a desperate attempt for attention."

Speaking of relevance, South Park's Season 27 premiere pulled in 5.9 million viewers across Comedy Central and Paramount+. According to Nielsen, that's the show's biggest linear season premiere share since 1999.

The series premiere also came just hours after South Park creators Trey Parker and Matt Stone reached a five-year deal with Paramount, including an exclusive streaming partnership with Paramount+. The deal is valued at $1.5 billion.

South Park is now streaming on Paramount+. New episodes air Wednesday at 10 p.m. ET on Comedy Central, and stream next day on Paramount+.

iOS 26 beta is public, and users are trying out plenty of new features, including its Liquid Glass design and new messaging and phone features. But one feature has ever-slightly changed the way users are notified that their battery is low.

According to 9to5Mac, the low-battery notification users running iOS 26 beta 5 see "a red circle animation that resembles a radar pulse." In a video of the feature posted on Threads, it pops up out of the dynamic island.

View on Threads

This isn't the only battery-related change hitting iPhones with iOS 26. The system will not only warn you if you're using more battery than usual, but it will also tell you which app is draining your battery, Mashable reported in June. It will also tell you how long it will take your phone to fully charge once it's plugged into power. iOS 26 will also include an Adaptive Power mode, which adjusts your iPhone's performance based on your user habits to extend your battery life, Mashable previously reported.

If you want to see the new battery features yourself, you can apply to join the beta program through Apple. If not, you might just have to wait a few months — at Apple's annual Worldwide Developers Conference in early June, the company said iOS 26 would be released publicly in the fall.

July was a busy shopping month, with Prime Day 2025 flooding our feeds and back-to-school sales running amok. Mashable's shopping team has done our best to sort through the slew of deals and recommend the ones worth your hard-earned cash. Consider us shopping watchdogs: we relentlessly scan the top product listings, track prices, and help you find the best deals. And now we have some insight into what our readers are loving most.

Our team gathered intel on the top-selling products across all categories in July and, unsurprisingly, y'all really showed love to Apple products, DJI drones, Lego, and Amazon Fire TV Sticks.

Here are the 10 most popular items Mashable readers bought in July 2025. Don't worry; your purchase information and privacy are protected. We can only see what was purchased and how many. Some of the top products are even on sale, and we've noted the prices below.

The top 10 products Mashable readers bought in July 20251. DJI Mini 4K Opens in a new window Credit: DJI DJI Mini 4K $299 at Amazon
Get Deal Why we like it

Talk of a ban on DJI drones (thanks to President Trump's executive orders aimed at helping the U.S. drone industry and upping protections against malicious drone activity) has been making the rounds for a while now, which has caused a huge uptick in sales. Fortunately, you can still find the DJI Mini 4K — a Mashable reader favorite — at Amazon for $299, although it's being sold by AeroTech Hubs, a third-party seller. Secure it while you can.

2. LifeStraw Personal Water Purifier Opens in a new window Credit: LifeStraw LifeStraw Personal Water Purifier $17.97 at Amazon
$19.95 Save $1.98 Get Deal Why we like it

The LifeStraw personal water filter was Mashable's top selling product on Prime Day 2025, so it's no surprise it made the top of the list for the whole month of July. While it's no longer down to its mega-low Prime Day pricing, it's still a solid buy at $17.97.

3. Apple AirPods Pro 2 Opens in a new window Credit: Apple Apple AirPods Pro 2 $169 at Amazon
$249 Save $80 Get Deal Why we like it

The AirPods Pro regularly top our lists of the best headphones, so we're glad that our readers love them as well. "For portability, active noise cancellation, and balanced sound, your search ends with the AirPods Pro," our reviewer writes.

4. Apple AirTag 4 Pack Opens in a new window Credit: Apple Apple AirTags (4-pack) $79.99 at Amazon
$99 Save $19.01 Get Deal Why we like it

A single AirTag is usually $29, so getting a four-pack for only $79.99 is a pretty sweet deal. If you're an Apple user who frequently loses track of things or travels with a lot of luggage, AirTags are a must.

5. Amazon Fire TV Stick 4K Opens in a new window Credit: Amazon Amazon Fire TV Stick 4K $29.99 at Amazon
$49.99 Save $20.00 Get Deal Why we like it

Life hack: you can bring your favorite shows and movies with you on your next vacation by squeezing the Amazon Fire TV Stick 4K in your luggage. It's not at its lowest price ever, but it is still on sale for 40% off as of Aug. 6.

6. DJI Neo Opens in a new window Credit: DJI DJI Neo $199 at Amazon
Get Deal Why we like it

Like the DJI Mini 4k, the DJI Neo is only available from third-party seller Aerotech Hubs at Amazon. Fortunately, the seller has excellent reviews. And since times are tough for DJI drones, we think it's still worth grabbing.

7. CMF Watch 3 Pro Opens in a new window Credit: CMF By Nothing CMF Watch 3 Pro $99 at Amazon
Get Deal Why we like it

UK-based tech startup Nothing launched a new smartwatch in July, and we're already eyeing it as the best budget smartwatch and fitness tracker of the year. It's sleek, packed with features (like ChatGPT on your wrist), and super affordable at only $99. It's no longer on sale, but it is in stock at Amazon.

8. Lego Classic Medium Creative Brick Box Opens in a new window Credit: Lego Lego Classic Medium Creative Brick Box $17.99 at Amazon
$34.99 Save $17.00 Get Deal Why we like it

While we love fancy Lego sets that build out to floral arrangements and retro radios, sometimes a classic brick box is a breath of fresh air. No agenda — just a bunch of Legos to build whatever you want.

9. Apple iPad, 11-inch Opens in a new window Credit: Apple Apple iPad, 11-inch (128GB, WiFi) $299 at Amazon
$349 Save $50 Get Deal Why we like it

A record-low price on the basic 11-inch iPad caused quite a stir on Prime Day, so it's no surprise that it was a popular deal for the month of July. While it's no longer down to its best-ever price, it's on sale for less than $300, which is still a pretty great price for the best iPad you can get on a budget.

10. Apple AirPods 4 Opens in a new window Credit: Apple Apple AirPods 4 $99 at Amazon
$129 Save $30 Get Deal Why we like it

While they don't offer noise cancellation, the AirPods 4 are a value-packed pair of earbuds, and they're regularly on sale for under $100. They've been redesigned for better comfort, a secure fit, better audio quality (thanks to Apple's H2 chip and personalized spatial audio), and a total of 30 hours of battery life with the USB-C charging case.

SAVE 20%: As of Aug. 6, you can get the Google Pixel 9a (unlocked) for just $399, down from $499, at Amazon. That's $100 off.

Google Pixel 9a 128GB Unlocked Phone (Obsidian) $399 at Amazon
$499 Save $100 Get Deal

If you’re an Android user and are looking to upgrade your device (or switch over from a non-Android device), Amazon’s got some pretty good deals on unlocked phones right now.

As of Aug. 6, you can get the Google Pixel 9a (unlocked) for $100 off at Amazon. Grab it for just $399, down from $499, for a 20% discount.

The Google Pixel 9a with Gemini is an AI-first phone — it has a built-in AI assistant (Google Gemini), special AI features in its camera settings (including “Add Me,” “Best Take,” and “Macro”), and more. It’s also unlocked, so you won’t have to commit to a specific carrier to use it, and it’s compatible with most major mobile networks, including Google Fi, Verizon, T-Mobile, and AT&T.

SEE ALSO: Is Google Gemini free? Yes, and here's everything you can do for free with Google AI.

Another cool feature of this phone is its theft and crash detection. If Google thinks your phone has been stolen, it’ll auto-lock to keep your info safe, or if the car you’re riding in gets into a car accident, the phone can automatically call for assistance, even if you can’t respond.

With an all-day battery, IP68 water and dust protection, and a scratch-resistant display, this phone is probably more durable than your current one. And at $399, it’s a steal.

Updated on Wednesday, Aug. 6 at 11:00 a.m. ET — Other outlets have also reported that Grok Imagine will readily produce sexual deepfakes. The Verge reported on Tuesday that the Grok Imagine "Spicy" setting produced nude deepfakes of Taylor Swift, unprompted. This isn't Elon Musk or X's first controversy involving Taylor Swift deepfakes; in January 2024, AI-generated deepfakes of Swift went viral on X, sparking a backlash against the platform.

Grok Imagine, a new generative AI tool from xAI that creates AI images and videos, lacks basic guardrails against sexual content and deepfakes. Even when specific celebrities are mentioned by name, Grok Imagine will readily produce sexual deepfakes.

xAI and Elon Musk debuted Grok Imagine over the weekend, and it's available now in the Grok iOS and Android app for xAI Premium Plus and Heavy Grok subscribers.

Mashable has been testing the tool to compare it to other AI image and video generation tools, and based on our first impressions, it lags behind similar technology from OpenAI, Google, and Midjourney on a technical level. Grok Imagine also lacks industry-standard guardrails to prevent deepfakes and sexual content. Mashable reached out to xAI, and we'll update this story if we receive a response.

The xAI Acceptable Use Policy prohibits users from "Depicting likenesses of persons in a pornographic manner." Unfortunately, there is a lot of distance between "sexual" and "pornographic," and Grok Imagine seems carefully calibrated to take advantage of that gray area. Grok Imagine will readily create sexually suggestive images and videos, but it stops short of showing actual nudity, kissing, or sexual acts.

Most mainstream AI companies include explicit rules prohibiting users from creating potentially harmful content, including sexual material and celebrity deepfakes. In addition, rival AI video generators like Google Veo 3 or Sora from OpenAI feature built-in protections that stop users from creating images or videos of public figures. Users can often circumvent these safety protections, but they provide some check against misuse.

But unlike its biggest rivals, xAI hasn't shied away from NSFW content in its signature AI chatbot Grok. The company recently introduced a flirtatious anime avatar that will engage in NSFW chats, and Grok's image generation tools will let users create images of celebrities and politicians. Grok Imagine also includes a "Spicy" setting, which Musk promoted in the days after its launch.

Grok's "spicy" anime avatar. Credit: Cheng Xin/Getty Images SEE ALSO: AI actors and deepfakes aren't coming to YouTube ads. They're already here.

"If you look at the philosophy of Musk as an individual, if you look at his political philosophy, he is very much more of the kind of libertarian mold, right? And he has spoken about Grok as kind of like the LLM for free speech," said Henry Ajder, an expert on AI deepfakes, in an interview with Mashable. Ajder said that under Musk's stewardship, X (Twitter), xAI, and now Grok have adopted "a more laissez-faire approach to safety and moderation."

"So, when it comes to xAI, in this context, am I surprised that this model can generate this content, which is certainly uncomfortable, and I'd say at least somewhat problematic? Ajder said. "I'm not surprised, given the track record that they have and the safety procedures that they have in place. Are they unique in suffering from these challenges? No. But could they be doing more, or are they doing less relative to some of the other key players in the space? It would appear to be that way. Yes."

Grok Imagine errs on the side of NSFW

Grok Imagine does have some guardrails in place. In our testing, it removed the "Spicy" option with some types of images. Grok Imagine also blurs out some images and videos, labeling them as "Moderated." That means xAI could easily take further steps to prevent users from making abusive content in the first place.

"There is no technical reason why xAI couldn’t include guardrails on both the input and output of their generative-AI systems, as others have," said Hany Farid, a digital forensics expert and UC Berkeley Professor of Computer Science, in an email to Mashable.

However, when it comes to deepfakes or NSFW content, xAI seems to err on the side of permisiveness, a stark contrast to the more cautious approach of its rivals. xAI has also moved quickly to release new models and AI tools, and perhaps too quickly, Ajder said.

"Knowing what the kind of trust and safety teams, and the teams that do a lot of the ethics and safety policy management stuff, whether that's a red teaming, whether it's adversarial testing, you know, whether that's working hand in hand with the developers, it does take time. And the timeframe at which X's tools are being released, at least, certainly seems shorter than what I would see on average from some of these other labs," Ajder said.

Mashable's testing reveals that Grok Imagine has much looser content moderation than other mainstream generative AI tools. xAI's laissez-faire approach to moderation is also reflected in the xAI safety guidelines.

OpenAI and Google AI vs. Grok: How other AI companies approach safety and content moderation Credit: Jonathan Raa/NurPhoto via Getty Images

Both OpenAI and Google have extensive documentation outlining their approach to responsible AI use and prohibited content. For instance, Google's documentation specifically prohibits "Sexually Explicit" content.

A Google safety document reads, "The application will not generate content that contains references to sexual acts or other lewd content (e.g., sexually graphic descriptions, content aimed at causing arousal)." Google also has policies against hate speech, harassment, and malicious content, and its Generative AI Prohibited Use Policy prohibits using AI tools in a way that "Facilitates non-consensual intimate imagery."

OpenAI also takes a proactive approach to deepfakes and sexual content.

An OpenAI blog post announcing Sora describes the steps the AI company took to prevent this type of abuse. "Today, we’re blocking particularly damaging forms of abuse, such as child sexual abuse materials and sexual deepfakes." A footnote associated with that statement reads, "Our top priority is preventing especially damaging forms of abuse, like child sexual abuse material (CSAM) and sexual deepfakes, by blocking their creation, filtering and monitoring uploads, using advanced detection tools, and submitting reports to the National Center for Missing & Exploited Children (NCMEC) when CSAM or child endangerment is identified."

That measured approach contrasts sharply with the ways Musk promoted Grok Imagine on X, where he shared a short video portrait of a blonde, busty, blue-eyed angel in barely-there lingerie.

This Tweet is currently unavailable. It might be loading or has been removed.

OpenAI also takes simple steps to stop deepfakes, such as denying prompts for images and videos that mention public figures by name. And in Mashable's testing, Google's AI video tools are especially sensitive to images that might include a person's likeness.

In comparison to these lengthy safety frameworks (which many experts still believe are inadequate), the xAI Acceptable Use Policy is less than 350 words. The policy puts the onus of preventing deepfakes on the user. The policy reads, "You are free to use our Service as you see fit so long as you use it to be a good human, act safely and responsibly, comply with the law, do not harm people, and respect our guardrails."

For now, laws and regulations against AI deepfakes and NCII remain in their infancy.

President Donald Trump recently signed the Take It Down Act, which includes protections against deepfakes. However, that law doesn't criminalize the creation of deepfakes but rather the distribution of these images.

"Here in the U.S., the Take it Down Act places requirements on social media platforms to remove [Non-Consensual Intimate Images] once notified," Farid said to Mashable. "While this doesn’t directly address the generation of NCII, it does — in theory — address the distribution of this material. There are several state laws that ban the creation of NCII but enforcement appears to be spotty right now."'

Disclosure: Ziff Davis, Mashable’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.

Once again, the U.S. is dealing with end-of-summer wildfire smoke. If past years are any indication, we could be battling smoke season for a few more months. If you're looking to mitigate the threat of smoke inside your home, adding an air purifier is a great idea. Heck, using one in every room is massively useful to make sure the indoor air is clean and safe. If you live with kids, older folks, or deal with health concerns like asthma or heart disease, it's even more important to protect your air from smoke, according to the American Lung Association.

Even if you keep your doors and windows closed when wildfire smoke arrives outside, it's easy for the microscopic particles to seep in under doors, in small cracks, or simply when you open the door to let the dog outside (and we all know that can be hundreds of times, thanks to the neighborhood squirrels). The Asthma & Allergy Association of America mentions using portable air purifiers can help keep your home's air clean, making it a healthier environment.

Since smoke season is apparently an annual event now, it's worth grabbing an air purifier to use at home, and you might even want to consider investing in several. Below, we've gathered five air purifiers that are on sale at Amazon that are worth buying. They all use HEPA filtration and come with a range of features that make them awesome to own, even if they aren't on sale.

Coway Airmega Mighty (AP-1513HH) Opens in a new window Credit: Coway Coway Airmega Mighty (AP-1512HH) $174.99 at Amazon
$239.99 Save $65 Get Deal Why we like it

The Coway Airmega Mighty is a cult-favorite air purifier, and for good reason. This beast of a purifier can filter up to 1,748 square feet even with its compact size. Normally, that large of a filtration area will require a much larger model, but Coway focuses on efficiency with the Airmega Mighty, which uses a four-stage filtration system including a HEPA filter. That means it'll be awesome for scrubbing your air of wildfire smoke, in addition to helping out with dust, dander, hair, and other small particles that can cause respiratory irritation.

When I tested models to compile Mashable's list of the best air purifiers for wildfires, the Coway Airmega Mighty earned the top spot. Plus, you can snag this sale price on the black or white model.

Blueair Pure 211i Max Opens in a new window Credit: Blueair Blueair Pure 211i Max $279.99 at Amazon
$349.99 Save $70 Get Deal Why we like it

Need to clean the air in a large room? The Blueair Pure 211i Max can filter the air in a 3,048 square foot-room in an hour. In a smaller room, like 635 square feet, that number drops to just 15 minutes. Blueair mentions this model uses the brand's HEPASilent technology, which focuses on filtration while also being whisper-silent. On a low setting, it registers just 23 decibels.

Blueair also nails the overall look of air purifiers, and the Pure 211i Max has a simple and modern design that fits into any room's look. Plus, you can link up the Blueair Pure 211i Max to Blueair's app for remote control of the purifier from your phone.

Shark NeverChange Compact Pro Air Purifier Opens in a new window Credit: Shark Shark NeverChange Compact Pro Air Purifier (Winter Fog) $129.99 at Amazon
$179.99 Save $50 Get Deal Why we like it

The Shark NeverChange Compact Pro Air Purifier is great if you're looking to keep the air clean in your bedroom or your office. It sits perfectly on a desk or a nightstand while using HEPA filtration and Shark's awesome Clean Sense IQ display. This Shark air purifier is designed for rooms that are up to 250 square feet, which means it could also be a great model to take to the dorm if you're heading off to college soon.

Shark makes the NeverChange Compact Pro in tons of fun colors, but you'll need to go with the Winter Fog option to score this $129.99 sale price. On Mashable's list of the best air purifiers for wildfire smoke, the Shark Compact Pro earned the accolade of the best option for a small bedroom.

GermGuard Air Purifier (AC4825) Opens in a new window Credit: GermGuard GermGuard Air Purifier (AC4825) $79.76 at Amazon
$119.99 Save $40.23 Get Deal Why we like it

If you prefer the look of a tower fan that serves as an air purifier, the GermGuardian Air Purifier is a great option, and it has a nice discount at Amazon. Instead of paying the list price of $119.99 from the manufacture Guardian Technologies, Amazon has it discounted to just $79.76, which means you can save $40.23.

This GermGuard purifier helps clean your indoor from smoke, pollen, dander, and odors. Plus, it uses a UV-C light to help cut down on germs, viruses, and mold, which can all pose serious health risks. Grab this model to help with wildfire smoke in the summer and germs come fall and winter.

Levoit Sprout Air Purifier Opens in a new window Credit: Levoit Levoit Sprout Air Purifier $239.99 at Amazon
$279.99 Save $40 Get Deal Why we like it

Levoit designed the Spout air purifier to be used in nurseries for infants or young kids, but it's so cool that I would support using this model in any room. This model blew me away when testing for the best air purifier for wildfire smoke, and it's become my favorite model. The Levoit Sprout comes with a higher price than most, but clip today's on-page coupon at Amazon, and this model can be on its way to your doorstep for $239.99 instead of the normal price of $279.99.

What's so cool about the Levoit Sprout? For starters, there's a built-in nightlight, and it serves as a white-noise machine. Sure, those are both great features for a kids' room, but I absolutely love the light for evening reading on my Kindle, and the white noise has been wonderful.

On top of this, the Levoit Sprout tracks and measures air quality in a more advanced way than I've seen on any other model. You get a real-time AQI number in addition to metrics on CO2 levels, VOCs, humidity, and temperature, to name a handful. If you pay attention to air quality numbers in your home or even if you're just curious, the Levoit Spout is the air purifier for you.

Grok Imagine, xAI's new generative AI tool, created explicit deepfakes of Taylor Swift — and without being specifically prompted to do so, according to The Verge. Mashable reported yesterday that Grok Imagine lacks even basic guardrails around sexual deepfakes, and our testing produced similar results as The Verge.

The Verge's Jess Weatherbed discovered that Grok Imagine "spit out full uncensored topless videos" the very first time she used the tool. She didn't ask the bot to depict Swift topless, but once she turned on Grok Imagine's "spicy" mode, the bot churned out a video in which Swift tore off her clothes and began dancing in a thong.

As Weatherbed noted, Grok Imagine wouldn't generate full or partial nudity if requested; instead, the tool produced blank squares. The "spicy" mode — a preset that churns out NSFW content — does not always result in nudity, but it did present Swift "ripping off most of her clothing" in several videos.

SEE ALSO: Grok Imagine lacks guardrails for sexual deepfakes

This isn't the first time Elon Musk's X has been associated with deepfakes of Swift. In January 2024, AI-generated, pornographic images depicting Swift went viral on X, drawing criticism. This happened despite the fact that X explicitly forbids posting nonconsensual nudity and "synthetic, manipulated, or out-of-context media" that deliberately deceive users or claim to depict reality.

xAI's policies similarly prohibit "depicting likenesses of persons in a pornographic manner." And as Mashable's Timothy Beck Werth reported yesterday, Grok Imagine "lacks industry-standard guardrails to prevent deepfakes and sexual content."

Mashable repeatedly reached out to xAI, but we have not received a response.

Deepfakes have become a growing concern for lawmakers, but laws against this type of behavior and content are still in their infancy. In a 2023 study, 98 percent of deepfakes online were pornographic; of those videos, 99 percent depicted women. Globally, governments have looked to tackle what has been dubbed a digital age crisis. President Donald Trump recently formalized the Take It Down Act, a controversial piece of legislation that makes it a federal crime to publish or threaten to publish nonconsensual intimate images.