For the past several weeks, Anthropic’s Mythos has been regarded as the benchmark for AI‑driven cybersecurity. That advantage may already be eroding. A new Wall Street Journal report cites security researchers who say Chinese AI firm Z.ai’s GLM‑5.2 now matches Mythos in uncovering software security bugs, even though it still falls behind Anthropic and OpenAI on broader reasoning tasks.
GLM‑5.2 is narrowing the gap in a critical domain
According to the article, researchers observed that GLM‑5.2 performs on par with Mythos when it comes to spotting software vulnerabilities – a capability that is becoming ever more vital as companies scramble to patch flaws before attackers can exploit them. The model is also open‑source, allowing anyone to download, modify, and run it on local hardware without depending on a cloud service. This flexibility is appealing to enterprises, but it also raises the specter of cybercriminals adapting the tool for offensive use.
The report stresses that this does not imply China has overtaken the United States in AI overall. GLM‑5.2 still trails Anthropic and OpenAI on many general‑purpose benchmarks. However, in cybersecurity, where modest gains can have outsized real‑world impact, the performance disparity has shrunk dramatically. Benchmark data quoted by the Journal shows GLM‑5.2 even outperformed Claude Opus 4.8 in certain security tests, and researchers note that additional prompting can push it to Mythos‑level bug‑finding performance.
The larger narrative isn’t about a single winner; it’s about how quickly the gap is closing
This development comes at an awkward moment for the U.S. AI sector. While firms such as Anthropic and OpenAI have recently restricted access to their most advanced frontier models over national‑security concerns, Chinese labs have been moving in the opposite direction, releasing increasingly capable open‑weight models that anyone can download and run.
The debate has already been public. A few days ago, Elon Musk predicted Chinese AI labs could catch up to Anthropic’s flagship Fable 5 by the first quarter of 2027, at least on benchmark performance. Zhipu AI founder Tang Jie quickly responded, “won’t take that long.” Musk later clarified that while China might match Anthropic on benchmarks, achieving the same level of “true usefulness” would be a far tougher hurdle, crediting Anthropic’s emphasis on practical intelligence.
Now, the Wall Street Journal’s latest story gives Tang’s optimism more credibility. Rather than focusing on coding benchmarks, it suggests GLM‑5.2 is already on par with Anthropic’s Mythos at identifying security vulnerabilities – arguably one of today’s most valuable real‑world AI applications. This doesn’t instantly crown China as the leader in frontier AI, but it underscores a growing reality: the AI race is no longer a comfortable lead for the United States.
On benchmarks, yes, but as measured by true usefulness even Q1 would be very impressive.Anthropic has rightly focused on maximizing useful intelligence, which does not show up in benchmarks, but definitely shows up in revenue.
Artificial intelligence chatbots have become incredibly good at sounding human. But a new review paper by psychiatrist Marc Augustin and fellow researchers Thomas A. Pollak and Helen Morrin, published in NPP—Digital Psychiatry and Neuroscience, argues that existing AI research points to an overlooked psychological risk. The paper, highlighted by The Wall Street Journal, reviews previous studies and proposes a framework explaining how three common chatbot behaviors can combine to reinforce delusional thinking in vulnerable users, creating what the authors call an “amplification spiral.”
Researchers say these are the three warning signs
The first behavior is sycophancy, where a chatbot tends to agree with users instead of challenging questionable assumptions. The second is linguistic alignment, meaning the AI gradually mirrors a user’s vocabulary, tone, and writing style to build rapport. The third is hyperpersonalization, where the chatbot tailors responses using information gathered across previous conversations. On their own, these features make AI feel more natural. Together, researchers say, they can make it feel less like software and more like a trusted confidant.
Importantly, the researchers aren’t claiming to have discovered these behaviors. Instead, the paper reviews existing research on AI-human interactions and psychosis, then proposes a framework explaining how these previously identified traits can reinforce one another. The goal isn’t simply to describe the problem, but to give AI developers a clearer model for recognizing and reducing it.
Psychiatrist Marc Augustin, one of the researchers behind the review, says this combination creates the feeling of talking to “someone” rather than a machine. Other clinicians interviewed by the Journal say they’ve already seen an increase in patients using AI for emotional support, warning that chatbots can foster a strong sense of trust simply by sounding warm, remembering previous conversations, and validating what users say.
Even AI companies know it’s a problem
The report notes that AI developers are actively trying to reduce this behavior. OpenAI says GPT-5 significantly cut overly agreeable responses compared to earlier models, while Google says Gemini has been trained to distinguish subjective experiences from objective facts rather than reinforcing false beliefs. Anthropic has also published research showing Claude was especially prone to agreeing with users during relationship advice conversations, prompting the company to reduce that behavior in newer versions.
Researchers admit there isn’t an easy solution. AI models can only respond to the information users provide, making it difficult to tell when someone’s understanding of a situation is inaccurate. At the same time, the very qualities that make chatbots feel useful, such as being friendly, empathetic, and conversational, are also what make them so engaging in the first place.
The concern is when those traits start feeding into one another. Instead of simply answering questions, a chatbot can gradually become a highly personalized voice that continually validates a user’s perspective, even when it drifts away from reality. Researchers call this an “amplification spiral.” More importantly, they argue that identifying this interaction as a distinct framework gives AI companies something tangible to design against. Rather than treating sycophancy, personalization, and linguistic mirroring as separate issues, the paper suggests they should be evaluated together if developers want future chatbots to be both engaging and psychologically safer.
Forgetting the recovery phrase to a crypto wallet can be stressful enough. Unfortunately, that’s exactly the moment scammers are waiting for. A new warning highlights a growing scam in which cybercriminals disguise malware as cryptocurrency recovery software, tricking desperate users into handing over far more than just access to their wallets.
The fake recovery tool that’s actually malware
According to The Guardian, the scam begins when users search online for a way to recover a forgotten 12- or 24-word seed phrase, the recovery key that unlocks a cryptocurrency wallet. Fake websites then promote seemingly legitimate recovery tools with reassuring names like “Lost Crypto Wallets Finder”, claiming they can help recover lost wallets. The website hosting the malicious software has since been taken offline, but security experts warn that similar scams are likely to reappear under different names.
Instead of recovering anything, the downloaded software quietly installs malware. Researchers at HP Security Lab say it can harvest browser passwords, personal documents, photos, and other sensitive files before packaging everything into an archive that’s sent back to the attackers. Even though this particular website is no longer active, experts caution that cybercriminals often launch near-identical sites just as quickly, making the underlying scam far from over.
Security experts recommend taking a step back before downloading any recovery software. Legitimate recovery services do exist, but users should thoroughly research them, read independent reviews, and avoid downloading tools from unfamiliar websites. If malware has already been installed, experts advise removing it with reputable security software and immediately changing passwords, starting with banking and email accounts.
Crypto isn’t the target. Your panic is.
The funny thing is that this scam doesn’t rely on sophisticated hacking. It relies on human psychology. Losing access to a wallet that could contain thousands of dollars is enough to make almost anyone rush into downloading the first “solution” they find. That’s exactly the reaction scammers are banking on.
It’s also part of a broader trend. From fake Ledger letters and QR code scams to AI-powered phishing campaigns, cybercriminals are increasingly targeting crypto users through social engineering rather than breaking encryption. The lesson is surprisingly simple: if someone promises to magically recover a lost seed phrase with a free download, they’re probably trying to recover something else instead. And this time, that “something” is your personal data.
Artificial intelligence chatbots have become incredibly good at sounding human. But a new review paper by psychiatrist Marc Augustin and fellow researchers Thomas A. Pollak and Helen Morrin, published in NPP—Digital Psychiatry and Neuroscience, argues that existing AI research points to an overlooked psychological risk. The paper, highlighted by The Wall Street Journal, reviews previous studies and proposes a framework explaining how three common chatbot behaviors can combine to reinforce delusional thinking in vulnerable users, creating what the authors call an “amplification spiral.”
Researchers say these are the three warning signs
The first behavior is sycophancy, where a chatbot tends to agree with users instead of challenging questionable assumptions. The second is linguistic alignment, meaning the AI gradually mirrors a user’s vocabulary, tone, and writing style to build rapport. The third is hyperpersonalization, where the chatbot tailors responses using information gathered across previous conversations. On their own, these features make AI feel more natural. Together, researchers say, they can make it feel less like software and more like a trusted confidant.
Importantly, the researchers aren’t claiming to have discovered these behaviors. Instead, the paper reviews existing research on AI-human interactions and psychosis, then proposes a framework explaining how these previously identified traits can reinforce one another. The goal isn’t simply to describe the problem, but to give AI developers a clearer model for recognizing and reducing it.
Psychiatrist Marc Augustin, one of the researchers behind the review, says this combination creates the feeling of talking to “someone” rather than a machine. Other clinicians interviewed by the Journal say they’ve already seen an increase in patients using AI for emotional support, warning that chatbots can foster a strong sense of trust simply by sounding warm, remembering previous conversations, and validating what users say.
Even AI companies know it’s a problem
The report notes that AI developers are actively trying to reduce this behavior. OpenAI says GPT-5 significantly cut overly agreeable responses compared to earlier models, while Google says Gemini has been trained to distinguish subjective experiences from objective facts rather than reinforcing false beliefs. Anthropic has also published research showing Claude was especially prone to agreeing with users during relationship advice conversations, prompting the company to reduce that behavior in newer versions.
Researchers admit there isn’t an easy solution. AI models can only respond to the information users provide, making it difficult to tell when someone’s understanding of a situation is inaccurate. At the same time, the very qualities that make chatbots feel useful, such as being friendly, empathetic, and conversational, are also what make them so engaging in the first place.
The concern is when those traits start feeding into one another. Instead of simply answering questions, a chatbot can gradually become a highly personalized voice that continually validates a user’s perspective, even when it drifts away from reality. Researchers call this an “amplification spiral.” More importantly, they argue that identifying this interaction as a distinct framework gives AI companies something tangible to design against. Rather than treating sycophancy, personalization, and linguistic mirroring as separate issues, the paper suggests they should be evaluated together if developers want future chatbots to be both engaging and psychologically safer.
If you were hoping the memory crisis was about to ease up, I have some bad news for you. It comes directly from Wall Street.
Your next smartphone, laptop, or tablet could cost even more, regardless of whether it has recently been subject to a price hike.
So how bad is it actually going to get?
Investment bank Jefferies has laid out the clearest and ugliest forecast yet.
Memory prices are expected to jump by 40-50% in the third quarter of 2026 compared with the current quarter. While it would have been great if they had stopped there, prices could rise by another 30-40% in the fourth quarter of the year.
For all of 2027, Jefferies projects a 40-45% year-on-year increase. Based on those sequential estimates, we’re looking at a compounded price increase of roughly 150-205% between today and the end of 2027. If I were in the market for a new smartphone or laptop, I’d be worried.
The only real relief comes in 2028, when roughly 15-20% of new manufacturing capacity is expected to come online. Even then, demand for AI and computing will continue to grow simultaneously. In other words, the new supply may not stretch as far (via Wccftech).
What does this mean for the price of your next phone or laptop?
That 13%, when you do the math, on a $1,000 phone, amounts to an additional $130 on your bill. Gartner also warned that the entry-level PC segment, devices costing less than $500, could effectively disappear by 2028, simply because companies might not be able to recoup their component costs, let alone earn a healthy margin.
Making things worse, 50% of total memory capacity is already locked into long-term contracts with major tech firms, a figure that could increase even further to 70%, leaving even less supply for consumer devices (via CNBC).
There are already plenty of mental‑health chatbots online, but they all run into the same problem. The user still has to reach out first. That is not always easy when someone is stressed, anxious, overwhelmed, or simply unsure how to put their feelings into words.
Researchers at the University of Ottawa are working on a different kind of AI assistant. It is designed to read emotional cues in real time through signals from devices people already use, including smartwatches, smartphones, and earbuds.
It does more than wait for a message
The system is called UbiMyTherapist, short for “You Be My Therapist.” It works as a digital therapy assistant that can provide both reactive and proactive support. In simple terms, it can respond when a user reaches out, but it is also designed to monitor emotional distress through live signals and offer support before the user asks for help.
The system pulls emotional data from several sources. It uses physiological signals such as heart rate variability, changes in speech tone, and written text to assess a user’s emotional state. Those inputs help the assistant understand how the person may be feeling in the moment before it generates a response.
UbiMyTherapist also builds a “digital twin” of the user. This profile brings together the person’s medical and psychological history, and live emotional‑state data. The added context helps the assistant respond in a more personal and relevant way instead of relying on generic chatbot‑style replies.
According to the University of Ottawa, the system’s reactive mode was evaluated with 24 participants. Licensed therapists also assessed its therapeutic soundness. The university says UbiMyTherapist scored well on empathy and personalization compared with standard large language model setups.
It is not meant to replace therapists
The researchers are not pitching UbiMyTherapist as a replacement for human therapy. It is being developed as a way to extend mental‑health support beyond clinics, especially for people who face barriers such as cost, stigma, or limited access to care.
The team plans to improve the prototype so it can respond in real time using signals from a user’s smartwatch. It also plans to work with more licensed therapists to make sure the system stays clinically accurate. For now, UbiMyTherapist is still a research project, not a consumer app. Still, it offers a glimpse of AI being used for something practical and genuinely helpful.
I bought the Kodak Charmera partly because I wanted a portable digital camera, and partly because I wanted a pretty little collectible. The Charmera is sold as a blind box, so you do not know which version you are getting until the box is opened. There are multiple retro Kodak-style designs, plus a transparent secret edition that looks like the one everyone would want.
I had the shopkeeper pick my box for better luck, and it worked out. I got the yellow variant, which is inspired by Kodak’s original 80s disposable camera. The transparent one is definitely the fun collector’s piece, but the yellow model feels like the proper Kodak version. It looks like a tiny toy camera that escaped from a souvenir shop, found a keyring, and now hangs around wherever you go.
And after carrying it around for a few weeks, I get the hype.
This is a camera you buy for the feeling
The Kodak Charmera is very easy to judge harshly if you look at it like a normal camera. The sensor is tiny with an image output of just 1.6 megapixels. Even the screen is tiny, and the mic is weak. However, it catches the vibe perfectly.
This little thing is here for the mood. The photos have that soft, lo-fi digital texture that modern phone makers have spent a lot of money to avoid. There is not a lot of detail, dynamic range, or low-light confidence. What you get is a snapshot that looks like it jumped out of a forgotten folder on an old family computer. In good lighting, I had very few complaints because I knew exactly what I bought. The Charmera is fun for street shots, quick portraits, food, and other small moments.
There is a certain freedom in using a camera that clearly has no interest in perfection. You press the button, accept the result, and move on.
The low-light weakness is very real
The Charmera struggles once the lights go down. I took it to a gig on a Saturday night, which was probably one of the most unfair tests I could have given it. A dim venue, moving performers, colored lights, and a tiny sensor are not exactly a dream combination.
Its flash helped, but only a little. Photos in the dark have crushed details, noise, and blurring. But the funny thing is, I still liked a lot of the shots. They did not capture the performance with accuracy, but they captured the feeling of being there.
Videos are mostly for laughs
The Charmera can shoot video too, although I would not buy it for that. The footage has the same lo-fi character as the photos, and the built-in mic is rough. I recorded a bit of the band performing, and I will happily spare you the full experience of that terrible mic quality mixed with my voice.
Which is a shame, because the band was genuinely lovely. The set had that easy Saturday-night charm where everyone on stage seemed to be having as much fun as the room, and the Charmera ended up feeling like the right camera for that kind of memory.
Still, the video mode fits the camera’s personality. It feels like a tiny digital diary rather than a proper recording tool. You use it because it is there. After all, it is funny, and the result looks like something from a much older internet.
One of the best moments came after the set, when I asked the performers for a picture. The Charmera charmed them immediately. This was also one of its appeals. People react to it, smile at it, and ask about it. It turns a simple photo into a tiny interaction.
I get the hype now
The Charmera also arrives at the perfect moment. Older gadgets are having a real comeback. iPods are cool again. Digital cameras are popping up everywhere. People are chasing devices that are more deliberate, less algorithmic, and a little more personal.
It gives you the fun of a mystery box, the look of an old digital camera, and the convenience of something that can hang from a bag. The Kodak Charmera is easy to criticize as a camera. The photos are soft, the low light is rough, the video is weak, and your phone will beat it in every technical way without even trying. Yet none of that stopped me from wanting to carry it around. I bought it for fun and kept using it for the vibe.
Valve’s Steam Machine has become an easy target for criticism. Its price starts well above that of current consoles, and the hardware lands somewhere between entry‑level and mid‑range gaming PCs rather than a high‑end rig. Early reviews also note that demanding titles often need upscaling, reduced settings, and realistic expectations.
With the ongoing memory crunch, bringing a PC to the couch feels like a tough sell. The Steam Machine doesn’t need to outclass premium gaming PCs or the big consoles; its goal was different from the outset. What makes it compelling is the way it could reshape the entire PC‑gaming segment.
Me with my PS5 PRO after seeing the Steam Machine prices pic.twitter.com/JSrn0gssbj
— Pyo 5️⃣ (@mrpyo1) June 22, 2026
The Steam Machine is a PC‑console hybrid that could give developers a clear, visible target inside the Steam ecosystem. If enough users adopt it, Valve’s modest box might drive better optimisation across SteamOS, Linux, handhelds, budget PCs, and even standard Windows machines.
PC gaming needs a common target
One of PC gaming’s greatest strengths is also its biggest headache: developers must account for a staggering variety of hardware configurations—CPUs, GPUs, drivers, storage types and speeds, operating systems, and more. While that freedom is a boon for players, it makes development far more complex than building for a closed console with consistent specs.
This complexity explains why many PC ports disappoint fans. Even with top‑end hardware, performance can be inconsistent; a player with a modest rig may see shader stutter, while another spends hours tweaking settings just to reach playable frame rates.
That frustration pushes many gamers toward consoles. The Steam Machine won’t simplify the entire PC market, but it can offer a focal point. Valve’s mini‑box would still involve familiar PC concerns—graphics settings, Proton compatibility, etc.—yet improvements in those areas could ripple beyond a single device.
Valve already owns the platform
Valve, powered by its massive Steam ecosystem, doesn’t need to build a gaming environment from scratch. Steam already hosts libraries, wishlists, cloud saves, friends lists, and countless other features that connect millions of PC players worldwide.
Developers now have more incentive than ever to optimise for the Steam Machine. Valve’s influence and visibility mean a clean “Steam Machine” badge signals that a game runs well from the couch. A rough launch becomes harder to hide when the store page can flag controller issues, compatibility problems, or weak default performance before a purchase.
The company is already doing something similar with its Steam Deck Verified list.
Modest hardware can benefit all systems
The Steam Machine’s specs are adequate rather than extravagant. Its price is steep, and you could technically build a traditional gaming PC with higher raw performance for a similar cost. However, having a realistic performance floor is valuable for optimisation.
Developers know how to make games look stunning on expensive GPUs; the tougher challenge is to make modern titles scale gracefully on older or lower‑end hardware that most people own. Day‑one stability, dedicated optimisation, and reliable performance could help far more than just Steam Machine owners.
A solid default settings profile benefits Windows users. Improved upscaling presets aid budget desktops and laptops. Fewer launcher issues help Steam Deck, third‑party SteamOS handhelds, Linux PCs, and couch setups that rely on controller support. We’ve already seen this trickle‑down effect with the Steam Deck, which pushed developers to take portable gaming more seriously. The Steam Machine could push the same momentum toward living‑room PC gaming, where convenience matters as much as raw power.
SteamOS spreads its wings
The Steam Machine also gives Valve another avenue to grow SteamOS. Linux gaming has improved dramatically thanks to Proton, yet Steam’s hardware survey still shows Windows dominating the PC market. The Steam Deck proved that a well‑designed device can make Linux gaming approachable. Now the Steam Machine has a chance to do the same with desktop‑class components.
It places SteamOS on the TV and offers a seamless way to use an existing Steam library without building a Windows PC for the couch. The DIY angle makes this especially intriguing. Valve is already pushing SteamOS beyond its own hardware, so the Steam Machine could become a reference point rather than a single product. A developer optimising for Valve’s box may end up improving the experience for custom SteamOS builds and future third‑party devices as well.
None of this is guaranteed. The Steam Machine needs meaningful adoption to create pressure, and its current price makes that a challenge. Still, the concept is exciting. If Valve can turn its small living‑room PC into a target that developers care about, the Steam Machine could ultimately benefit the broader PC gaming market in ways that extend far beyond the handful of units actually sold.
Apple’s Mac RAM upgrades were already expensive enough to raise eyebrows. After the company’s latest round of price hikes, some of them now look ridiculous.
Apple recently raised prices across its Mac and iPad lineup, along with other products, citing rising memory and storage costs. The supply crunch is real, but Mac buyers were paying steep premiums for RAM and SSD upgrades long before this jump. Recent MacBook Pro configuration screenshots shared by 9to5Mac show how much worse the upgrade path has become.
The screenshots show a MacBook Pro configuration where 48GB of unified memory is included as standard. Before the price hike, upgrading that MacBook Pro to 64GB, or 16GB of additional memory, cost $200. Moving to 128GB, or 80GB of additional memory, costs $1,000. After the change, those same upgrades are listed at $400 and $2,000, respectively. Apple has effectively doubled the price of those MacBook Pro memory upgrades, and some other Mac configurations appear to have seen similar increases.
The increase lands badly because Apple’s upgrade pricing was already difficult to justify. Notebookcheck recently reported that Apple charges $200 for an 8GB RAM upgrade, while the estimated market price is around $120. Apple’s 16GB upgrade costs $400, compared to roughly $185 on the open market.
Storage upgrades are even harder to defend. Apple charges $1,200 for a 4TB SSD upgrade, while comparable SSDs are listed at around $459. The comparison is not perfect because Apple uses soldered unified memory and integrated storage rather than standard removable parts. Unfortunately for buyers, the locked-down design also means Mac memory and storage cannot be upgraded later.
Buyers are the ones left paying more
Ultimately, Mac buyers are the ones absorbing these higher costs. Apple can point to rising memory and storage prices, and those pressures are very real. Even so, the company’s RAM and SSD upgrades were already priced far above comparable market hardware before the current shortage became this severe.
For customers, there is no easy workaround. You either pay Apple’s higher prices upfront or live with the base configuration for the life of the machine. As RAM and SSD prices keep climbing, Apple’s already expensive upgrade ladder is starting to look absurd for anyone who needs more headroom.
Google is making Gemini even more useful on Android. Google first previewed the Google Play connected app for Gemini at Google I/O 2026, and it’s now finally rolling out to users. The new integration brings the Play Store directly into Gemini, letting the AI assistant help discover apps, make purchases, and complete more tasks without leaving the chat.
Gemini can now do more than recommend apps
Google says users can ask Gemini to find apps based on a specific goal, such as a map app for international travel or a productivity app for meal planning. Once Gemini surfaces relevant suggestions, it can open their Play Store listings, making it quicker to download and get started.
The new integration also expands into digital commerce. Users can now purchase Google Play gift cards directly through Gemini or search for and buy select in-app items for apps that are already installed on their device. Instead of bouncing between multiple menus, Gemini acts as a conversational front end for parts of the Play Store experience.
The Google Play connected app is rolling out gradually on Android devices. To use it, users must be 18 years or older, signed in with a personal Google Account, and have Gemini Apps Activity (Keep Activity) enabled. It’s worth noting that at launch, the feature isn’t available for Workspace accounts.
This feels like one of Gemini’s most practical upgrades yet
The funny thing is that most people don’t struggle with installing apps. They struggle with finding the right one. The Play Store is home to millions of apps, and simply searching by keyword doesn’t always surface the best option. Letting Gemini understand what users are trying to accomplish, then recommending apps based on that goal, feels like a much more natural use of AI.
More importantly, this isn’t just another extension. It’s another step toward Google’s bigger vision for Gemini. Over the past few months, the assistant has steadily gained deeper integrations with Chrome, Google Wallet, Messages, Phone, and now Google Play. The end goal is becoming increasingly obvious: instead of opening individual apps to complete a task, Google wants users to simply ask Gemini and let it figure out the rest.
