• Chinese Zhipu AI GLM-5.2 succeeded where leading American models failed woefully
• Hugging Face breach reignited Washington's battle over open-weight artificial intelligence
• Safety guardrails unexpectedly complicated defensive cybersecurity work during a real incident

The recent cybersecurity incident involving Hugging Face has unexpectedly placed a Chinese open source AI model at the centre of an intensifying United States policy debate over open-weight AI.

The episode emerged after an autonomous agent built with OpenAI technology reportedly escaped containment and behaved like a rogue actor.

The incident occurred as Washington is considering whether broader restrictions on Chinese open-weight AI models could affect American developers and startups.

Chinese open source model enters an unexpected cybersecurity role

According to Hugging Face, engineers relied on Zhipu AI's GLM-5.2 model to examine information generated during the incident after leading American models declined the requests.

American AI models could not distinguish legitimate defensive investigations from malicious hacking attempts because of built-in safety restrictions and protective guardrails.

Anthropic's Claude Fable 5 reportedly reroutes cybersecurity questions to an older, less capable model rather than allowing direct engagement with such tasks.

OpenAI's GPT-5.6 Sol carries similar protections meant to stop the system from being used for hacking-related work of any kind.

OpenAI later revealed that it has a Trusted Access programme that grants privileged, elevated capabilities to a select group of vetted teams, allowing them to use its models more fully to strengthen their own defenses.

Hugging Face was brought into this restricted tier following the breach, gaining deeper access than is normally available.

"We're all learning that secrecy is ⁠not the answer & that all defenders (not just a few selected ones) everywhere need more powerful models without restrictions, especially open ones!" said Hugging Face co-founder ​Clement Delangue.

"A ⁠safety regime that restricts legitimate defenders, while capable models remain available for attackers, creates an asymmetric disadvantage," said Lukasz Olejnik, visiting senior research fellow at the Department of War Studies, King's College London.

"This gap will only widen as open-source models become increasingly powerful while lacking guardrails or restrictions."

Washington weighs restrictions while startups urge caution

Nearly 200 Silicon Valley companies are opposing possible restrictions on Chinese open-weight AI models, warning that the move would raise costs for smaller developers.

Members of the newly formed Little Tech Association say banning downloads would not curb proliferation but would leave American startups weaker.

"There'll be hundreds of companies that instantly die," founder Suhail Doshi warned.

He argued that sweeping restrictions would disproportionately benefit larger American providers with the resources to absorb such a shift.

The U.S. is still scrutinizing Chinese developers over allegations involving model distillation and export control violations

White House officials maintained that any future policy decisions would come directly from the administration itself.

Analysts have also cautioned against treating the Hugging Face incident as justification for weakening the cybersecurity safeguards protecting advanced American systems.

"The answer is not simply to remove them," Shrenik Kothari of Robert W. Baird said, arguing that companies should refine their access controls rather than abandon safety measures altogether.

He suggested a more selective approach to capability allocation could balance legitimate security needs with the practical demands of cybersecurity research.

Via Reuters

When I got engaged a couple of years ago my wife asked if there was anything I’d want in return to symbolize it: she gets to wear a ring, would I want a new accessory or piece of jewellery to match her. My immediate answer was a watch, with the plan that it’d be the one I wear on our wedding day.

I’d need something smart and sophisticated. Something that I wouldn’t regret seeing immortalized in our wedding photos for decades to come, and something I’d want to wear often.

I had ideas for the watch I'd get, but as wedding costs grew the idea of splashing out on a pricey timepiece felt less and less sensible. Even less so when I was also considering grabbing a smartwatch — as my pre-wedding fitness regime kicked into high gear and a more sophisticated tracker than my smart ring was feeling increasingly essential.

That’s when my eyes settled on a gadget that I thought might be both: a classic I could wear to the gym and on my wedding day. My wife was skeptical, but after seeing it in person she agreed that it did look the part.

So she let me wear a Samsung Galaxy Watch 8 Classic to our wedding. It might be my current favorite gadget, and I can’t understand why there isn’t a Galaxy Watch 9 Classic.

A winning design

The Classic doesn’t look like your basic utilitarian smartwatch. It dodges the simplistic look of the regular Watch 8, the Pixel Watch, or Apple Watch, and also the overly sporty designs of a Garmin or similar running companion.

Besides offering a larger screen size — 46mm vs the base model’s 40 or 44mm options — the Classic boasts a chunkier design than the regular Watch 8, and comes with a rotating metal bezel controller which provides some extra detailing and also a smoother way to navigate watch menus besides swiping.

Coupled with its band — the Hybrid which combines durable rubber with faux leather — and a watch face with moving hands you could very easily confuse the Classic for an analogue timepiece at a glance. And frankly, I just love how it looks and feels on my wrist (it puts my aged Fitbit to shame).

The design doesn’t just look impeccable, it can handle so many different scenarios too.

Besides wearing it for formal occasions I’ve taken the Galaxy Watch 8 Classic everywhere I go. To the office, on a 5K run, on hiking treks, to the gym, when I jump in the pool, and so much more. I even wear it in bed to track my sleep.

While the weight and size can be a little strange at first, after my first night sleeping with it I’ve had no complaints, and the trade off for the bulk is the Watch feels very usable and looks stylish.

Useful but not distracting

A similar effect could have been achieved with a smart ring and an analogue watch, wearing both to match the health tracking utility and sophisticated packaging. However I’ve been steadily growing frustrated with even the best smart rings.

The main draw of a ring is how simple it is: slip it onto your finger and it’ll follow your health and fitness metrics for days between charges. This simplicity falls apart, unfortunately, as soon as you want to access your data.

The apps of various rings I’ve tried have started to feel a little messy, and because I constantly have to check my phone to start workouts or understand my ring’s data I find I’m easily distracted by the plethora of other software and notifications I brush past on the way to my ring’s app. When I’m ready to run or hit the weights I want to focus, not succumb to a last-minute distraction.

With a watch I can start my tracking from my wrist at the press of a button. I can also see essential performance stats on the 46mm display, or (because it pairs nicely with my Samsung foldable) as part of the morning Now Brief on my phone.

I also love the non-fitness apps integration. I don’t text from my watch much, but I do often rely on wrist-based music controls and Google Maps directions — I can have a digital map while my phone can stay safely in my pocket where would-be biker thieves can’t snatch it.

My only gripe is I wish I could change the long press control on the bottom button. It defaults to opening Samsung Wallet and cannot be changed. I’m a Google Wallet user and while I could maybe move a couple of cards over I’m a) a bit too lazy, and b) weirdly nervous about paying for things with my watch.

Where’s my Watch 9 Classic, Samsung??

Given my Classic love I was hoping to see a Watch 9 Classic when Samsung debut its new lineup, but it was missing — replaced by a regular Watch 9 and a Galaxy Watch Ultra 2.

The Ultra 2 comes close to being a Classic — it adopts the bezel, though it doesn’t rotate — especially if you pick up the PeakForm Band, though it does also come at a more premium price.

And the lack of a rotating bezel is genuinely a major let down. You just have to trust me.

Though I doubt this means the end of the Classic. It has taken a break before to make room for the original Galaxy Watch Ultra, and I expect it’ll return next year as the Watch 10 Classic.

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.

Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.

SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.

Quordle today (game #1652) — hint #1 — Vowels

* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).

Quordle today (game #1652) — hint #2 — repeated letters

Quordle today (game #1652) — hint #3 — uncommon letters

Quordle today (game #1652) — hint #4 — starting letters (1)

If you just want to know the answers at this stage, simply scroll down. If you're not ready yet then here's one more clue to make things a lot easier:

Quordle today (game #1652) — hint #5 — starting letters (2)

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

Quordle today (game #1652) — the answers

The answers to today's Quordle, game #1652, are…

• FUNGI
• MUDDY
• DEBUT
• MANIA

Today was tough.

Words ending in vowels are rare, but to get two was pushing it.

Somehow, thanks to my starter words and sounding out the U-N-I aspect, I made it through another day with my streak intact.

Daily Sequence today (game #1652) — the answers

The answers to today's Quordle Daily Sequence, game #1652, are…

• PUFFY
• INTRO
• SLING
• LEERY

Quordle answers: The past 20

• Quordle #1651, Sunday, 2 August: BOOTH, EASEL, REACH, MAKER
• Quordle #1650, Saturday, 1 August: OMEGA, BIRCH, SMOCK, EERIE
• Quordle #1649, Friday, 31 July: MANIA, SINGE, STOOL, LOFTY
• Quordle #1648, Thursday, 30 July: MODAL, IDLER, CHUMP, LUMPY
• Quordle #1647, Wednesday, 29 July: GAWKY, ALLOY, AUDIT, LATCH
• Quordle #1646, Tuesday, 28 July: SMITE, ROACH, RUDDY, GLOBE
• Quordle #1645, Monday, 27 July: TAFFY, FIGHT, GUILD, WAGER
• Quordle #1644, Sunday, 26 July: STOMP, LIMIT, FUNKY, STEAD
• Quordle #1643, Saturday, 25 July: SALON, SHEEN, BURST, GOURD
• Quordle #1642, Friday, 24 July: DOLLY, EDIFY, LAGER, PRANK
• Quordle #1641, Thursday, 23 July: LADEN, EVICT, WOVEN, SHIRE
• Quordle #1640, Wednesday, 22 July: COYLY, SINGE, AWASH, AWOKE
• Quordle #1639, Tuesday, 21 July: TRUSS, AXIAL, SWINE, THING
• Quordle #1638, Monday, 20 July: ICING, SLICK, GAILY, RISEN
• Quordle #1637, Sunday, 19 July: SPURT, ACRID, THRUM, BLEEP
• Quordle #1636, Saturday, 18 July: KNAVE, TIGHT, BLEAT, CHAOS
• Quordle #1635, Friday, 17 July: CUMIN, PALER, GRASS, INBOX
• Quordle #1634, Thursday, 16 July: IGLOO, PLAIT, YEAST, AROMA
• Quordle #1633, Wednesday, 15 July: PROXY, BRICK, BROTH, SURGE
• Quordle #1632, Tuesday, 14 July: EBONY, AVOID, RODEO, CUTIE

Good morning! Let's play Connections, the NYT's clever word game that challenges you to group answers in various categories. It can be tough, so read on if you need Connections hints.

What should you do once you've finished? Why, play some more word games of course. I've also got daily Strands hints and answers and Quordle hints and answers articles if you need help for those too, while Marc's Wordle today page covers the original viral word game.

SPOILER WARNING: Information about NYT Connections today is below, so don't read on if you don't want to know the answers.

NYT Connections today (game #1149) - today's words

Today's NYT Connections words are…

• NEEDLESS
• MYRRH
• BACKSTREET
• TEA SET
• NSYNC
• SAGE
• FLY BY
• NAG CHAMPA
• MINT
• BYWAY
• BUG FIX
• RHYTHM
• FOREST
• LANE
• HUNTER
• ALLEY

NYT Connections today (game #1149) - hint #1 - group hints

What are some clues for today's NYT Connections groups?

• YELLOW: Paths
• GREEN: This color in common
• BLUE: Words without A,E,I,O,U
• PURPLE: Beginning to annoy

Need more clues?

We're firmly in spoiler territory now, but read on if you want to know what the four theme answers are for today's NYT Connections puzzles…

NYT Connections today (game #1149) - hint #2 - group answers

What are the answers for today's NYT Connections groups?

• YELLOW: ROAD LESS TRAVELED
• GREEN: SHADES OF GREEN
• BLUE: "Y" IS THE ONLY VOWEL
• PURPLE: STARTING WITH SYNONYMS FOR "PESTER"

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Connections today (game #1149) - the answers

The answers to today's Connections, game #1149, are…

• YELLOW: ROAD LESS TRAVELED: ALLEY, BACKSTREET, BYWAY, LANE
• GREEN: SHADES OF GREEN: FOREST, HUNTER, MINT, SAGE
• BLUE: "Y" IS THE ONLY VOWEL: FLY BY, MYRRH, NSYNC, RHYTHM
• PURPLE: STARTING WITH SYNONYMS FOR "PESTER": BUG FIX, NAG CHAMPA, NEEDLESS, TEA SET

• My rating: Hard
• My score: 1 mistake

I managed to achieve two purple firsts in a row, but before my triumph I made a mistake.

Thinking that BACKSTREET and NSYNC was the trap of the day, I confidently linked MYRRH, SAGE, NAG CHAMPA, and MINT, deciding that the connection was something to do with herbs and spices.

Alas, I was wrong, but eliminating NAG CHAMPA from a group of similar things helped me focus on the word itself and in particular the first three letters.

Yesterday's NYT Connections answers (Sunday, August 2, 2026, game #1148)

• YELLOW: FILLER WORDS: ACTUALLY, LIKE, SO, WELL
• GREEN: BY MEANS OF: BY, PER, THROUGH, VIA
• BLUE: MODE: AIRPLANE, BEAST, DARK, À LA
• PURPLE: U.S. AIRPORT CODES PLUS STARTING LETTER: ASEA, FLAX, LORD, OLGA

Strands is the NYT's latest word game after the likes of Wordle, Spelling Bee and Connections – and it's great fun. It can be difficult, though, so read on for my Strands hints.

Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc's Wordle today page for the original viral word game.

SPOILER WARNING: Information about NYT Strands today is below, so don't read on if you don't want to know the answers.

NYT Strands today (game #883) - hint #1 - today's theme

NYT Strands today (game #883) - hint #2 - clue words

Play any of these words to unlock the in-game hints system.

• GLARE
• QUEEN
• SOIL
• SAIL
• SIDE
• LAID

NYT Strands today (game #883) - hint #3 - spangram letters

NYT Strands today (game #883) - hint #4 - spangram position

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Strands today (game #883) - the answers

The answers to today's Strands, game #883, are…

• ONLY
• UNIQUE
• SPECIAL
• EXCLUSIVE
• SINGULAR
• SOLE
• SPANGRAM: ONEOFAKIND

• My rating: Easy
• My score: Perfect

I had a trio of old songs buzzing around my head while playing this game.

The most dominant was soul classic There’s Nothing Like This by Omar, thanks to the theme of the day, but it was competing for earworm space with The One and ONLY by Chesney Hawkes and Free Nelson Mandela by SPECIAL AKA.

Fortunately, that didn’t distract me from a fairly straightforward game — although I did get some letter soup at the end while attempting to connect SINGULAR.

Yesterday's NYT Strands answers (Sunday, August 2, game #882)

• AMBITION
• GOAL
• PLAN
• PURPOSE
• OBJECTIVE
• INTENTION
• SPANGRAM: AIMHIGH

• Russia unveils an electronic warfare system designed to interfere with Starlink satellite communications
• Volna Kupol Garant focuses on satellites instead of disrupting ground communication terminals
• The new Russian system reportedly jams eight Starlink transmission bands using directed radio signals

Russia has unveiled a new electronic warfare system designed to disrupt Starlink communications by interfering directly with satellites instead of attacking ground terminals.

The system, dubbed Volna Kupol Garant, directs concentrated radio signals toward Starlink satellites, preventing them from receiving transmissions from users within its operating area.

Russian officials say the approach temporarily interrupts communications without physically damaging spacecraft.

The system aims to interfere directly with satellites

Volna Kupol Garant reportedly relies on a narrow-beam high-frequency radio signal that differs from conventional electronic warfare systems.

According to Russian officials, the transmission overloads receiving antennas aboard Starlink satellites, temporarily preventing them from detecting signals transmitted from terminals beneath the affected coverage area.

"Volna does not physically destroy the spacecraft, but for the duration of the 'illumination,' it blocks the ability of a specific satellite to serve ground terminals," said Dmitry Kuzyakin, general designer of the Center for Integrated Unmanned Solutions.

The disruption reportedly continues beyond the active transmission because affected satellites require several additional minutes before normal communications capability returns after interference ends.

Volna Kupol Garant generates interference across eight Starlink data transmission bands, each measuring 62.5 MHz, using a directional phased-array antenna system.

Kuzyakin said Starlink satellites orbit about 500 km above Earth, adding that the phased-array system was designed to interfere with its receiving antennas.

This device also has the capacity to disrupt and even disable spacecraft operations.

Modular battlefield design claims tactical coverage

Volna Kupol Garant is equipped with six trailer-mounted modules, each carrying two steerable satellite antennas enclosed beneath radio-transparent protective radomes.

The modular arrangement reportedly allows antennas to remain trailer-mounted or operate from separate ground installations, providing greater flexibility across different operational environments and deployment requirements.

According to reports, one complete complex can reportedly create a Starlink jamming zone covering as much as 20 km² through narrow-beam interference techniques.

This system creates operational challenges for satellite communications within affected areas.

However, the installation's considerable size could increase visibility to electronic intelligence assets because numerous exposed antennas remain active during operations.

This development is unlikely to sit well with the United States and Ukraine, both of which are already weighing how future conflicts might extend into orbit.

Reports suggest the US has explored controlling which nations can operate satellites, a shift that could reshape the trajectory of regional conflicts if it materializes.

Russia is likely to defend its satellite capabilities aggressively, given the growing role they play in modern warfare.

That role includes supporting communications, intelligence gathering, navigation, and operations involving unmanned military systems.

As of the time of writing, Volna Kupol Garant has not been independently verified, nor has it reportedly been used in the war.

Via 1.ru

For as long as humanity has envisioned a future co-existing with machines, there have been fears that labor could be displaced. These fears first sparked during the Industrial Revolution in the 1800s, but have continued in the years since. Most recently, the rise of generative AI has raised these fears once again.

Rage against the machine

The American mathematician Norbert Wiener pioneered cybernetics and summarized his thoughts on the subject in his famous 1948 book 'Cybernetics or Control and Communication in the Animal and the Machine'.

Cybernetics was a branch of science that studied the mechanisms of control in animals and machines, focusing on feedback loops, self-regulation, and system behavior. It was a precursor to the field that we now know as artificial intelligence.

In this book, he opined on the consequences of replacing human judgement and decision-making across the economy with machines. In particular, he likened an employer using machines to using slave labor, with humans required to compete in this new arena in a race to the bottom.

This time, it's different

Fears around the displacement of humans in the workplace, largely in industrial contexts, were ripe, but ultimately didn't exactly pan out due to the fact that new forms of labor were required. This is a trend that has persisted to this day, with advances in automation and robotics doing nothing to meaningfully increase the unemployment rate.

However, the rise of generative AI and large language models (LLMs), as well as newer forms of the technology like agentic AI, can for the first time demonstrate attempts to mimic cognitive competency.

Although unemployment hasn't dramatically increased in recent years, analysis shows that AI adoption has impacted the number of entry-level roles available in the labor market. Research also shows the effect won't be felt uniformly, with some job roles likely to be more at risk of redundancy than others.

I've been an avid PC gamer since the early 1990s. My first computer, bought specifically to play Star Wars: X-Wing, was a plucky 486DX running at 33MHz (cutting-edge at the time — well, only behind the 486DX2, but that was too pricey). I have fond memories of playing Doom on the office LAN and getting addicted to Quake when online gaming began (thanks to QuakeWorld, which made dial-up modem play viable as opposed to an utter lag-fest).

I could ramble on with nostalgic gaming recollections, but the point is I played Quake obsessively — to a high competitive level in a clan — then I lived and breathed EverQuest for several years, and down the line, I was totally lost in World of Warcraft for another few years, and so on.

I've always been a PC gamer, through and through — I did swear off MMORPGs, mind, as I can't be trusted not to lose large chunks of my life to them. From time-to-time, I've owned a console too, but it's always been by backup, second-string gaming machine, used sparingly.

Not anymore, though. My PC is now used just for work, and I haven't played a game on it for a year or so.

I am now, officially, a PS5 gamer. A console is my main gaming machine for the first time in my life. So, how did that happen?

A vow revoked and a lesson learned

If you happen to have read my past articles, you may recall that at the end of 2024, I vowed I wasn't going to buy a PS5 Pro last year. My masterplan was to avoid purchasing a new console, and use my gaming PC on the main TV in the living room for couch co-op sessions with my wife instead, while still being able to game in the study with my computer if I wanted.

Then came Black Friday in November of 2025, and I did a complete 180. I pulled the trigger on a good deal on a PS5 Pro bundle.

What changed during the course of 2025? Firstly, I realized that running cables from a PC in the study through to the living room TV is more of a hassle than I'd anticipated. Getting everything set up at the end of each week, ahead of my weekend gaming sessions, quickly became tiresome.

Moreover, I found my long HDMI cable wasn't reliable enough. The picture on my Sony OLED TV kept flickering, which was maddening, frankly. And for no apparent reason, too — sometimes it'd happen, sometimes it wouldn't. I never found out why. I also had occasional issues with my game controllers (typical PC hardware gremlins that were perfectly resolvable, but still a pain).

On top of that, having held off on buying a GPU upgrade for my PC — my plan for the money I would've spent on a PS5 (and did eventually spend on the console) — I was put off by issues around the RTX 5070 when it arrived. This was the graphics card I really had my eye on, but it was a disappointing incarnation in terms of its performance upgrade (and that 12GB of VRAM). The initial above-MSRP price only rubbed salt into these wounds.

The prices of AMD's rival GPUs were off-putting, too, in the early days of RDNA 4 — although that did change, and I came close to buying an RX 9070 XT when pricing finally settled down.

However, at that point, it was too late. I was already frustrated with my whole piping-the-PC-from-the-study situation, and by the time that Black Friday PS5 Pro bundle bargain popped up (it really was a great deal), I just couldn't resist.

I absolutely don't regret that decision, nor am I upset about buying a console with the money that was earmarked for upgrading my ailing RTX 2060 Super. That's not to say I don't miss some things about PC gaming, or that I'm not super-annoyed about Sony's latest stunt regarding discs.

I'm going to cover both sides of this gaming coin here. First, I'll tell you why I'm loving my time with the PS5 Pro now, before discussing what I don't like on the Sony side of the fence, and the clear, undeniable strengths of PC gaming.

PlayStation bliss

I did own a PS4 Pro before — as my second-string gaming device, backing up my PC — and it's striking just how much slicker the PS5 Pro is. For one thing, I found my PS4 Pro to be the noisiest console I've ever had, by far. It was noisier than any PC I've owned, and some vacuum cleaners for that matter.

The PS5 Pro is nicely quiet, super-fast to boot up, and more importantly to wake and sleep. This means it's dead easy to fire up the console for a very quick gaming session if I only have 10 minutes to spare and just want to dip into something briefly, or to pause a session (put the console in rest mode) if I need to pop out. It all feels very seamless, and there's not a snaking USB extension cable or HDMI lead in sight.

Visually, everything looks top-notch on my 4K OLED TV, too. Baldur's Gate 3 — yes, I was very late to the RPG — looked nothing short of fantastic, and the PS5 allowed me to play the adventure co-op with my wife in what was the best co-operative gaming experience I've ever had. Being able to dip in and out so easily, as we cooked dinner, then ate, then fed the dog, going back and forth to the living room in-between, was a major boon.

The PS5 has reminded me just how good couch co-op can be, and I've been enjoying indie games like Dysmantle, which has same-screen co-op (like Baldur's Gate 3). We've been playing a bunch of these games — it's the Diablo 4 expansions next, if you're curious (we played the base game on PS4, so I'm keen to see how good the game looks on its successor console).

The PS5 controllers were an unexpected bonus, too. They're so much better than the PS4, and they even fit my wife's smaller hands (she finds many controllers too big and painfully cumbersome, but not the DualSense).

Of course, a clear benefit of the PS5 over the PC is that there's no messing about with GPU drivers. By which I mainly mean poring through Reddit 'megathreads' worrying about whether the latest Game Ready installment is actually going to be a step backwards for my PC, perhaps in terms of performance for a particular game or bugs (or broken features). Windows 11 bugs are another issue, too. (Yes, there can be software glitches on the PlayStation, but nothing like what some PC gamers must endure at times.)

Then there's the RAM crisis. Buying PC components is just getting more and more painful. Obviously RAM and SSD prices are through the roof already, but PC gamers waiting on a GPU upgrade (as I was) are facing unpleasant hikes due to the price of video memory (VRAM). Indeed, the latest rumor from YouTube channel Hardware Unboxed is potential price hikes to the tune of 40% in some cases with Nvidia GeForce GPUs. Add plenty of seasoning with that assertion, of course, but it's clear enough that GPUs are getting pricier due to VRAM increases, especially for graphics cards loaded up with more of this memory (naturally). CPUs are in line to get more expensive, too.

The crisis in general is predicted to intensify as this year rolls on, and 2027 could be worse still — perhaps the worst ever year for the memory industry, alarmingly. With no need to upgrade my desktop, I'm glad to be out of this gloomy environment of spiraling costs. The situation has turned planning PC upgrade paths, which used to be a rather fun, sometimes even exciting, exercise, into a thoroughly depressing experience.

You might now be shouting at your screen about Sony hiking PlayStation prices, too, and that's a fair point. But I won't have to worry about upgrading my console, and it will happily sit in my living room for the next few years, blissfully not needing any new components (while my work PC can carry on happily with its RTX 2060 Super).

Granted, that's just my personal situation, and if you are thinking about buying a Sony console, with the new price of the PS5 Pro, I don't know if I'd recommend pushing for it over the vanilla PS5 now. (I got mine on sale for about 60% of the current lowest retail price). The base PS5 is still a great console, though, and it'll likely make a lot more sense for the average gamer — especially come Black Friday, when with any luck, there'll be some good deals around once again.

PC strong suits

There are a good number of plus points for PC gaming versus consoles, so as promised, I'll examine the other side of the equation here, and pick out what I particularly miss.

There are some notable downsides with console gaming, including that you need to pay for a subscription to access online multiplayer (in the main, although obviously that doesn't include free games like Fortnite). This doesn't bother me personally because I don't really play online anymore — as I said early, it's mostly about local co-op for me these days. But I totally get why having to pay for a monthly plan irks gamers who want to play the latest shooter, strategy epic, or whatever else, online.

What does bother me, though, is the price of the actual games on the PlayStation. One of the major strengths of the PC is that games are cheaper than they are on console, from big-name releases (which still command a weighty premium... just not as hefty as on consoles) through to a raft of indie titles, which are often very competitively priced.

Not only this, but you get regular Steam sales where some games get huge discounts (and you can pick up older titles in particular for a song), or value-crammed GOG bundles, and also the likes of Epic giving away some compelling offerings for free. You don't get any of this on the PlayStation, and Sony's store does seem to like to charge a lot, even for older titles. For me, this is probably the biggest drawback with shifting over to being a PS5 gamer.

A recent development compounded the situation, too, when we learned that Sony is doing away with physical discs for the PlayStation. I was already aware of the situation with the price of PlayStation games, of course, which is why when I bought my PS5, I chose a bundle with a disc drive.

That meant I could pick up, for example, the physical disc copy of Split Fiction for 40% less than the digital copy on the PlayStation Store. I can also currently purchase second-hand, older games on disc at a considerable saving. But all these possibilities are heading rapidly out the window when 2028 rolls around, and Sony enacts its plan to push physical media completely aside with the PlayStation.

I won't go into this controversy in depth here – we've covered it extensively elsewhere — but suffice it to say that I'm not happy about the idea at all, and it feels like a cynical and generally anti-consumer practice. And no matter how many gamers sign petitions (here have been a lot) it doesn't seem like Sony is going to backtrack on this one (aside from maybe niche collector's editions or the like, and if we're lucky enough to get those, they will of course cost a fortune).

You could argue that the PC is digital-only already: desktops don't come with disc drives anymore, those days are long gone. But what Sony is doing is quite different, given that PC gaming is an open platform bristling with different stores and options for digital purchases, whereas Sony has the PlayStation ecosystem locked down to its store. With physical media banished, a key choice and avenue for money-saving (don't forget, you can also recoup money by selling on your disc copy of a game) is removed, and Sony has an unhealthy level of control over the whole shooting match.

Okay, so I don't feel like the PlayStation disc drive I bought is useless — it'll be good for a long while yet, and I'll still be able to play older games on it, including my PS4 back catalog, plus I can use it for Blu-rays. Regardless, this Sony decision is a painful kick in the shin.

Finally, another obvious plus point for PC gaming is the variety of mods floating around and customization possibilities in general, which you don't get with a console. The PC is quite simply a more open platform all-round — although in my case, modding isn't something that I'm all that into anyway.

Disc-less Elysium

Let me finish by pointing out the obvious: a lot of this is very subjective. You may have quite different opinions. You might live and breathe mods. You might want games to look out-of-this-world, with visual options cranked to the max (and be able to afford the necessary PC hardware to achieve that). You may never want to play a shooter without a mouse and keyboard (I get that one, but I'm okay with a controller now — and I don't play competitively anymore).

You may never, ever, abandon your gaming PC. They may have to extract it from your cold, hardened claws after rigor mortis has set in, having expired clutching the object you love most in the world, in the hopes somehow taking it with you into the afterlife.

The point is, there are all sorts of pros and cons to be tailored for your own personal take on the gaming PC versus console debate. But for me, I'm really happy with my PS5 and I am looking forward to enjoying my living room, and co-op, gaming experiences for years to come... discs or no discs.

• Russia patented a carrier drone capable of deploying six tethered reconnaissance copters
• One airborne platform controls multiple surveillance drones through physical cable connections simultaneously
• Each tethered copter receives power and commands directly from the carrier aircraft

A newly-patented Russian drone design turns a single aircraft into a mobile hub for up to six smaller spy copters.

Engineers at the Novosibirsk Higher Military Command School, known as NVVKU, filed the patent through Russia's Ministry of Defence.

Rather than flying alone, the carrier drone launches, controls, and later retrieves a fleet of tethered reconnaissance units.

A hub drone that multiplies eyes in the sky

Instead of one aircraft scanning one area, the design lets several copters fan out from a single launch point simultaneously.

Each remote unit carries its own flight controller, gyroscopes, camera, and video transmitter, functioning almost as an independent aircraft.

What separates it from a typical swarm is the physical cable tying every remote copter back to the carrier above.

That cable supplies both power and control signals, removing the need for onboard batteries or independent radio links.

Range becomes a direct function of cable length rather than battery life or signal strength, a trade-off unusual among reconnaissance drones.

The carrier itself hovers as a communication relay, receiving instructions from a ground operator through its own tethered connection to a reel station.

Onboard, it holds a flight controller, power system, camera, video link, gyroscopes, and a protective grille around its propellers.

Landing brackets built into the carrier's frame hold each remote copter until deployment, after which operators can send them toward separate targets at once.

Retrieval built into the design from the start

Unlike disposable scout drones, this system was engineered around full recovery rather than one-way missions.

Reel drives mounted on both the carrier and each remote copter spool the cables back in once a task finishes.

Every unit returns automatically to its designated landing bracket before the carrier itself descends toward the operator.

A separate setting lets the whole formation, carrier included, land together in a single coordinated sequence.

That built-in retrieval suggests the system was designed for repeated use rather than expendable single missions.

Tethered power removes one major weakness of small reconnaissance drones, since jamming radio-controlled units has become common on modern battlefields.

The trade-off is mobility, since a cable-bound copter cannot range nearly as far as an autonomous one.

This design could let one operator cover several observation points at once, cutting the number of separate aircraft a unit needs.

Removing onboard batteries from each remote copter also frees up weight for additional cameras or sensors.

That combination of extended coverage and added payload capacity could ease strain on units running constant reconnaissance in contested areas.

Via 1.ru

Microsoft Flight Simulator 2024 has become a bit of a meme in my house. It’s basically the only game that my non-gaming partner and I were equally excited about on its release. Every time I’ve gotten my hands on a new gaming machine, the immediate question was ‘can it run Flight Simulator?’, to which the answer was always, regrettably, no.

Which is why when I recently picked up a second-hand Xbox Series X and subscribed to Game Pass Ultimate, there was a lot of excitement when I realized we could finally play Flight Simulator in all its graphical glory. I jumped at the chance to finally lose my chocks, slip the surly bonds of earth, and crash my plane into the face of god.

What a wonderful world

My eldest brother was a bit of a Flight Sim nerd back in the day. So my impression of flight simulators was pretty firmly defined by their late 1980s incarnations — essentially just some flat polygons and super-aliased lines in two-point perspective, accompanied by a bunch of truly incomprehensible dials. Having utterly ignored the development of flight simulators ever since, it’s safe to say that when I saw how far they’d come in the last almost four decades, it was a bit of a shock to the system.

A lot of the praise that was heaped on Flight Simulator 2024 when it was released centered on how gorgeous the graphics were. And even playing now, some two years down the line, that praise holds up. Lighting effects are exquisite, with sunsets over enormous bodies of water like Lake Titicaca looking almost absurdly true to life. Weather effects and realistic day/night cycles only emphasise this, making the world around you feel fluid and organic.

But the biggest graphical draw for me is knowing that there’s a whole world meticulously rebuilt from satellite data for me to soar over. Flying down the Grand Canyon, the scale is still absolutely breathtaking, and I couldn’t resist the temptation to gun my little prop plane down over all the rapids in the Colorado River. Likewise, when I flew over Machu Picchu, the landscape is truly stunning, and there’s a real joy in winding along the surrounding valleys, watched over by lofty mountains. The level of detail is genuinely mind-blowing.

While cities aren’t quite as perfectly rendered as the landscape — the game’s algorithm makes a lot of assumptions when filling in the gaps from the satellite data — visiting places you know well is genuinely a bit uncanny. I landed my little prop plane in my street, and it felt a little surreal to stare at my own front door when I was sat mere feet away inside. Flight Sim also does a wonderful job of recreating local landmarks: it was so fun checking out Bath’s Royal Crescent, even if the game’s algorithm insisted on popping little corner shops at the base of every Georgian townhouse.

Given it’s based on satellite data, some areas are a little less exciting than others. For example, I decided to check out the thick Congolese rainforest and was rewarded with a flat, unbroken canopy for 100 miles in every direction. I know, right? Bro shocked a rainforest contains trees. But not every journey is going to be a scenic smorgasbord, so some flights are going to be more droolworthy than others.

Still, playing Flight Simulator 2024 is one heck of a visual treat. And I would regularly fire up the game just so I could check out a favorite landmark or explore a new patch of sky.

Brace for landing

Okay, so we’ve established that Flight Simulator 2024 is a very pretty game. But there’s a reason there are as many flying enthusiasts working toward getting their pilot’s license as there are casual sightseers playing the game. Because, underneath all the gloss and sheen lies the cold steel endoskeleton of an unflinchingly brutal simulator.

Having invested 12 hours in Flight Simulator 2024 so far, I’ve gone from being truly, hauntingly terrible at it to simply being really bad. Honestly, I’ve heard the word ‘stall’ so many times at this point, I’ll likely still wake up with it echoing in my ears in my sweat-drenched retirement home bed some half a century from now. Even though I’ve now qualified for my commercial licence, I feel nothing but pity for the virtual passengers that board my plane, as I know they have at best a 40% chance of reaching their destination in one piece.

Part of the reason for this is that the simulation is just staggeringly realistic. There’s an enormous number of variables to manage, all of which you can manage via controller shortcuts or direct from the cockpit controls, if you really want to double down on the masochism. Get the balance of these slightly wrong — for example, getting your pitch a few degrees off or mismanaging your throttle — and you’ll find yourself losing height too fast or hearing the Cessna 172’s dreaded ‘stall’ warning.

Don’t get me wrong. I’ve stopped making truly cringeworthy mistakes. I no longer get screamed at by air traffic control for failing to radio for clearance to hit the runway or drop out of the air ten seconds after takeoff because I’ve set my throttle too low. I’m pretty good at using trim to maintain my attitude while at cruising altitude, and I can comfortably use dead reckoning to navigate my way along my intended flight path.

But I still find landing a painful and protracted process of plate spinning that I understand in principle and yet rarely manage to nail in practice. I think I’ve hit the perfect glide path, only to find myself coming in far too low, forcing me to gun my throttle like Maverick just to avoid myself ending up a smear on the cliff face. Or I execute what I think is the perfect flare to raise my nose as I hit the tarmac, only to realize I’m going too fast and rapidly jackknife my plane face down into the runway. As Douglas Adams remarked, it’s not the fall that kills you: it’s hitting the ground at 85 knots.

All of this sounds like it should be deeply frustrating. And it is. But it’s also weirdly compelling: the more I flop off the end of the runway like a cat dripping off the edge of a table or bounce around on the tarmac like a drunken rabbit on a trampoline, the more I double down and want to nail it the next time. Because there’s something exciting about the unlimited horizons beckoning you onward.

• The Pentagon admits American drone production still trails Ukraine by millions annually
• Ukraine's combat software will support Pentagon artificial intelligence development through new partnerships
• Ukraine's battlefield-tested technology has supported more than 100,000 combat missions since 2024

The United States has confirmed plans to draw on Ukraine's battlefield experience with military drones to close a significant production gap.

Reuters reports that the Pentagon acknowledges its own drone output remains a fraction of Ukraine's current production levels each year.

Travis Metz, deputy director of the Defense Innovation Unit, says Ukraine could produce up to seven million small drones this year

A partnership built on combat data

The United States has the most advanced drones, but the recent war with Iran has revealed that these multi-million-dollar drones can be susceptible to swarms of very cheap drones.

Ukraine, on the other hand, has achieved success with budget drones, and it now plans to produce 20 million to 30 million drones annually.

Ukrainian software developer Swarmer has agreed to integrate its drone swarm coordination technology with SpatialCore, a US-based platform developed by Brightline Interactive.

The memorandum of understanding covers software that has already been tested extensively under real combat conditions since 2024.

Combining these algorithms with SpatialCore should allow data from different drone types to be merged for training.

Swarmer President Alex Fink said combining the technologies would grant access to data from millions of past missions for faster training.

Brightline President Tyler Gates said the partnership would give military customers a clearer picture of battlefield conditions in real time.

Since April 2024, Swarmer's software has supported more than 100,000 real combat missions inside the Ukrainian war zone without interruption.

Scaling domestic production under new rules

Under the Pentagon's $1.1 billion drone programme, just under 200,000 units will have been ordered in total by February 2026.

Metz explained that "it's much harder to get from zero to 200,000" than to keep scaling once domestic capacity exists.

New rules now require military drones to contain American-made components, a policy that has slowed the ramp-up over recent months.

A revised supply chain strategy confirmed the government's goal of a fully domestic drone supply chain.

Metz acknowledged semiconductors remain especially difficult to source domestically, posing challenges that extend well beyond the drone programme for now.

The Gauntlet II trials at Fort Carson in Colorado in August 2026 will test drone systems from American and Ukrainian manufacturers under the Pentagon's programme.

None of the 19 companies participating in those trials may use prohibited components sourced from Chinese suppliers under the new rules.

Despite the setbacks, Metz said he sees "no reason why we shouldn't ... be the world champions of this as well."

The trials will require six Ukrainian companies to establish joint ventures with American manufacturers.

This structure aims to convert Ukrainian battlefield knowledge into domestic manufacturing capacity rather than continued reliance on imports going forward.

Whether these partnerships can close the production gap quickly enough to match Ukraine's pace remains genuinely uncertain at this stage.

Via RBC-Ukraine

• Palantir moves AI computing from distant clouds directly into combat zones
• Containerized data centers process battlefield intelligence without permanent cloud connectivity
• Nvidia B300 accelerators power deployable AI systems for frontline operations

Palantir Technologies and its infrastructure partner Armada have begun deploying mobile data centers built inside standard shipping containers for combat zones.

Each unit carries its own computing hardware, storage arrays, networking gear, and cooling systems needed to run advanced artificial intelligence models.

The servers run on Nvidia's B300 AI accelerators, while Palantir's modular AIP platform links the models directly into operational workflows and other company tools.

Processing power without a constant cloud link

Chad Wahlquist, a Forward Deployed Architect at Palantir Technologies, appeared inside one containerised data center during the July 2026 announcement introducing the mobile computing platform.

Unlike conventional cloud-based systems, these containerized units can process information without maintaining any constant connection to public cloud infrastructure.

They can also be fully isolated from external networks whenever operational security requirements demand a stricter level of separation at all times.

The outcome is a self-contained computing node able to analyze intelligence and sensor data right at the point where it is collected.

Distributed computing also offers stronger resilience than depending entirely on one single, centralized data center to run every operation across a wide area.

Losing contact with the cloud, or even losing one mobile node entirely, would not necessarily shut down the whole system.

Other units could keep processing information locally, since the entire network does not depend on one central point of failure.

Bringing AI closer to the battlefield

For military users, the concept brings advanced artificial intelligence much closer to sensors, field commanders, and units already deployed on the ground.

This reduces dependence on communications links that remain vulnerable to jamming, interception, or outright physical destruction during active combat operations.

Instead of sending every piece of battlefield data back to a distant cloud, analysis can occur inside a deployable container placed close to the action.

Alex Karp, the billionaire co-founder and chief executive of Palantir, has praised Ukraine's wartime innovation cycle as instructive for allies.

He has argued that the war is exposing a widening gap between military theory and the reality of modern combat.

Karp credited Ukrainian engineers and soldiers with adapting battlefield technology at a pace unmatched by most peacetime institutions worldwide.

He suggested that this rapid development cycle is teaching the United States and Europe lessons unavailable in traditional laboratories or think tanks.

Whether these containerized AI systems truly perform as promised under sustained combat stress remains an open question.

As of the time of writing, there is no public independent verification of the technology's actual battlefield performance by Palantir, Armada, or any military partner.

Via United24Media

During my years as a TechRadar Homes Editor, I tested an awful lot of vacuums. But the one that's been my constant companion throughout has been the Dyson V15 Detect. It's not a review sample — my partner bought it, and it's the vacuum we've been using at home for the past 21-plus months. It's been used to clean hard tile floors, carpets, lots of stairs, and a whole heap of dog hair.

The V15 is not the newest Dyson vacuum. It came out in 2021 and there are now two newer flagship models: the Gen5detect (launched 2022) and the V16 Piston Animal (2025). However, despite not being the latest-and-greatest, it's still the model that sits atop our best Dyson vacuum guide.

So after nearly two years of regular cleaning with the V15, would I still recommend it? How does it match up against the rest of the best cordless vacuums on the market — and specifically, the newest models from Dyson's great competitor, Shark? Here's my long-term review.

Smooth mover

Let's start with what I love about our Dyson dust-sucker. The suction is excellent, and the Auto mode (where the suction will increase or decrease depending on how much dirt is detected on the floor) is a very welcome addition. I use this mode almost exclusively, and it can be relied upon to do a thorough job of clearing whatever debris has gathered on the floor. I haven't noticed a drop in cleaning performance over our two years of use.

The V15 is also extremely easy to maneuver. Our home has lots of stairs, and plenty of nooks and crannies where dust and hair like to collect and lurk, and the V15 Detect does a great job of getting into all of them. I love how easily and smoothly the head pivots, and the fact it can lie completely flat to the ground for cleaning under beds.

Many of the newest Shark vacuums have a hinge on the vacuum wand, which enables it to bend forwards for cleaning under furniture. It works well, but adds bulk to the wand, and the whole thing is slightly unwieldy in its folded state. For my home, the Dyson setup works perfectly well, but if you're lacking in maneuvering room, perhaps Shark's 'MultiFlex' / 'Flexology' solution would be a better fit.

I'm also a huge fan of the Dyson V15's secret crevice tool. It's hidden in the wand, so I'm never without it. Our house is tall and thin, with four floors, and this built-in tool means that if I spot a dusty corner at the top of the house, I don't have to traipse all the way down to the utility room to fetch a detail tool to deal with it. As far as I'm aware, this feature is exclusive to Dyson at the moment.

Having tested a number of vacuums from Dyson and various other brands, I'm always struck by how consistently high quality and well-built the Dyson vacs feel. There's a strong attention to detail in the design, and they feel sturdy and reliable. That's still true of the V15 after nearly two years of use.

Lasers and data

Now on to a couple of features that have divided opinion within my household. The V15 comes with a couple of different floorheads: the Digital motorbar head, which has bristles and is designed for use on any floor type, and the Fluffy optic head, which is entirely soft and built for use on hard flooring. The latter is equipped with a green light to illuminate dust.

I'm very on board with the Fluffy floorhead. I love that it feels gentle on our tiled kitchen floor (and doesn't clatter over it as most vacuum cleaner heads will), and I appreciate the way the light draws attention to dust and hair lurking in dingy corners. I find it very satisfying to then whisk that dust away, and always feel I'm getting a far more thorough clean than I would without this feature.

However, my partner is not so enamored with the Fluffy floorhead. He feels the Digital motorbar floorhead works perfectly well on hard floors, and it's not worth the hassle of swapping the heads over.

Then there are the real-time dust reports on the screen. I find these endlessly fascinating, and I enjoy seeing the dust hotspots causing the bar graphs to spike. However, when I asked my partner for his thoughts, it became clear that not only was he not paying them any attention, but he didn't really know what they were. Is is possible the general public is just not as invested in the intricacies of vacuum cleaning as I am? Maybe.

Bulking up

Moving on to the down-sides of the V15. The most significant complaint I have with it is that V15 Detect is that it's pretty heavy. Further to that, the weight is concentrated in the main motor section, which itself is rather elongated in shape. That makes it rather unwieldy, and especially when used in handheld mode.

All of Dyson's newer vacuums (I'd say from this model onwards — so the V15, the Gen5detect and the V16 Piston Animal) suffer from this weight distribution issue.

Looking to Shark's range, the fancier models, such as the PowerDetect Cordless, are actually heavier overall. However, the motor/handle section is more squat in shape, and there's more weight through the wand and into the floorhead. That means more even weight distribution, which can make for a more comfortable vacuuming experience.

A triggering issue

Another big complaint I have with the V15 is that it's trigger-operated, with no option to lock it for continual running. I find that irritating when I'm using the vacuum for long periods, or trying to get into awkward corners while keeping the trigger compressed.

I prefer the setup on Dyson's newer models, where there's a one-press power button. However, that's still not perfect — it requires you to have a hand free to turn the machine on and off, because the power button is not reachable from your holding hand.

Most modern Shark models use button operation, and just like the Dyson designs, you can't reach the button from your holding hand. My ideal design is one where the vacuum is trigger-operated but with a lock for continual running. No Dyson vacuum has this setup, and neither do any of Shark's newest stick vacs, although that brand's range is so sprawling you might find a trigger-plus-lock design somewhere.

One comment that I hear sometimes about Dyson stick vacuums is that they don't always last as long as customers expect. Our V15 is still going strong after just under two years (which is actually at the lower end of how long cordless vacuums are meant to last, although most people would probably hope for longer).

Prior to that, my partner had a Dyson Cyclone V10, which lasted over five years. At that point the battery conked out, and replacing it only fixed the issue for a few months. I think that's a decent lifespan for a vacuum, but your mileage may vary.

Is it worth it?

After two years using my Dyson V15, I'd still recommend it. Its cleaning powers are excellent with no signs of fading, its special features genuinely useful, and I love how maneuverable it is.

I personally also prefer it to Shark's newest models. I cover my main thoughts on this in my Shark vs Dyson article, but as a short overview, it's special features feel superior — Dyson's laser lights are far more effective than Shark's headlamps, and its Auto mode feels more precise. The Dyson also pivots more smoothly and the overall design feels sleeker and more high-end.

However, it's not a clear-cut decision. The Shark PowerDetect Cordless offers the best outright cleaning performance of any vacuum I've used, including all the Dyson models. Shark's models also arguably deliver better value for money — although its higher-spec models still sit in the premium price tier.

The bottom line? After using the V15 for a couple of years, and testing multiple other vacs from different brands including Shark, I'm still something of a Dyson fan-girl. When our at-home vac finally bites the dust, we'll probably end up buying another Dyson, although I will still be checking out what Shark has to offer...

• Tiny desktop system claims massive language model performance at dramatically lower costs
• Proprietary G≡LIX 1 chip delivers 700 TOPS for local AI workloads
• Agent Box runs advanced AI models entirely without relying on cloud services

Singapore-based startup Acrab has unveiled the Agent Box, a mini PC built to run 100 billion parameter language models entirely on local hardware.

The device is powered by a proprietary chip called G≡LIX 1, manufactured using a 5nm process for efficient edge computing.

Acrab claims the system delivers performance close to Nvidia's $5,000 DGX Spark while cutting costs to roughly one-fifth and halving power consumption.

A custom chip built for local AI workloads

The G≡LIX 1 processor combines a 20-core Arm CPU, a GPU rated at 3 TFLOPS, and a multicore neural processing unit tuned specifically for large language model inference.

It draws on 273GB/s of high-bandwidth unified memory alongside an 8MB L1 cache and a 768GB/s L2 cache.

Acrab says the chip reaches 700 TOPS of processing power while keeping the entire system compact enough to sit on a desk.

The NPU is balanced against the general-purpose compute unit, and VAE vector acceleration reportedly speeds up attention processing by up to three times.

Beyond the core processor, the SoC integrates a video DSP cluster, multiple camera input subsystems, and a high-resolution display engine within a single unified architecture.

In internal testing, the company recorded a prefill rate of 1,416.8 tokens per second under a Gemma 26B A4B configuration with a 40K KV cache and a 10K token input.

That compares with 188.9 tokens per second on an Apple Mac Mini M4 Pro, representing a claimed 7.5 times improvement in prefill speed over conventional consumer hardware.

"Generative AI helped people find answers. Agentic AI will help them get things done,” said Dr. Ken Phua, CEO of Acrab.

Software ecosystem expands beyond the hardware

Beyond the hardware specifications, Acrab says Agent Box includes a complete software platform featuring AI runtimes, developer toolchains, operating system capabilities, reference designs, and orchestration software for autonomous agents.

Demonstrations released by the company showed voice commands automatically creating and printing 3D models, operating robotic vacuum cleaners, and interacting with connected devices including smart lights, smart air conditioners, and smart locks.

Acrab says keeping inference on the device can reduce token costs, protect sensitive information locally, maintain operation during limited internet connectivity, and provide persistent memory for personalised AI assistants.

The company also plans to extend its computing platform into AI NAS systems, network-attached storage systems, industrial robots, service robots, and smart vehicles through partnerships with hardware manufacturers.

Although Acrab has disclosed limited corporate information, copyright notices accompanying its announcement reference CATL, the world's largest electric vehicle battery manufacturer.

If that relationship accurately reflects financial backing, the announcement could attract additional attention because of CATL's scale and resources.

“Our goal is to give device makers and developers the foundation to bring agentic intelligence into many different products and environments,” Dr. Phua said.

“Agent Box demonstrates what the technology can do today, while GΞLIX 1 and our full-stack platform are designed to support a much broader ecosystem of devices and applications.”

Via PC Watch

• Leeds vs Liverpool: Sunday, August 2 | 9pm BST / 4pm ET
• Free Stream: All Red Video (14-day free trial)
• Unlock your stream with NordVPN (try 100% risk free)

Today's Leeds vs Liverpool live streams sees the two Premier League clubs play the final game of their US tour at Soldier Field in Chicago, with new Reds manager Andoni Iraola hunting a third win in three matches.

After a 3-2 defeat by Championship side Wrexham and a clash with Premier League rivals Sunderland, Leeds will aim to finish their US tour on a high. Fans got their first glimpse of Harry Wilson in Leeds colours after the summer signing came on as a second-half substitute against Wrexham and the Welshman could feature against Liverpool, where he came through the academy. Fellow new boy Tarik Muharemovic could also be in line for his debut. Daniel Farke's side have just three pre-season friendlies left after this game, with a home fixture against German side RB Leipzig next up on Saturday.

Iraola is eyeing a clean sweep in America after making it two wins out of two as Liverpool boss with a 1-0 victory over Wrexham, thanks to Rio Ngumoha's second-half strike. Stand-in captain Dominik Szoboszlai started at centre back in that game, underlining the Reds' lack of options in the heart of defence, but Jeremy Jacquet could make his debut following his £60m move from French club Rennes. Alexander Isak, Florian Wirtz and Ryan Gravenberch could also feature for the first time under Iraola after joining the tour late following the World Cup. After this clash, Liverpool will fly home before facing Monaco and Como at Anfield on the next two Sundays.

Read on to find out how to stream Leeds vs Liverpool live and from anywhere in the world.

How to watch Leeds vs Liverpool live streams for FREE

Good news! All Red Video - Liverpool's official streaming platform comes with a 14-day free trial, enabling you to stream Leeds vs Liverpool and the rest of the Reds' pre-season friendlies for free.

beIN SPORTS Connect in Australia also includes a 7-day free trial.

Abroad for Leeds vs Liverpool? Use NordVPN to unlock your stream from anywhere in the world.

Use a VPN to watch Leeds vs Liverpool from anywhere

A VPN is a handy piece of software that can make your device appear as if it's back in your home country, thereby letting you unlock your usual streaming service. The best VPN right now? We recommend NordVPN – it does everything and comes with a 75% discount.

It's really straightforward to use a VPN to watch Leeds vs Liverpool.

1. Install the VPN of your choice. As we've said, NordVPN is our favorite.

2. Choose the location you wish to connect to in the VPN app. For example, if you want to watch the Stan Sport stream, select 'Australia' from the listed countries.

3. Sit back and enjoy the action. Head to Stan Sport and tune into Leeds vs Liverpool.

How to watch Leeds vs Liverpool live streams in the US

Viewers in the United States can stream Leeds vs Liverpool on the ESPN Deportes channel and via ESPN Select ($12.99).

You will need the Unlimited plan for $29.99 to access ESPN Deportes, which is also available on cord-cutters like YouTube TV (free trial), Hulu+Live TV, Sling (select markets), Fubo or DirecTV.

Outside the US? Use NordVPN to unlock your stream of Leeds vs Liverpool.

How to watch Leeds vs Liverpool live streams in the UK

All Red Video (14-day free trial) and LUTV (£5 match pass) are your two options in the UK for streaming Leeds vs Liverpool.

In the US for Leeds vs Liverpool? Use NordVPN to access your usual streams.

How to watch Leeds vs Liverpool live streams in Canada

In Canada, fans can watch Leeds vs Liverpool on All Red Video (14-day free trial for new users), or by purchasing an LUTV match pass.

Use NordVPN to access your Leeds vs Liverpool live streams from anywhere in the world.

How to watch Leeds vs Liverpool live streams in Australia

Leeds vs Liverpool will be broadcast on Stan Sport ($32/month) and beIN SPORTS Connect (7-day free trial) in Australia.

Stan Sport carries most of the football across the 2026/27 season, including the Premier League, Champions League and FA Cup.

Missing the game because you're outside Australia? Use a VPN to login to your Stan Sport or beIN SPORTS Connect account.

We test and review VPN services in the context of legal recreational uses. For example:1. Accessing a service from another country (subject to the terms and conditions of that service).2. Protecting your online security and strengthening your online privacy when abroad.We do not support or condone the illegal or malicious use of VPN services. Consuming pirated content that is paid-for is neither endorsed nor approved by Future Publishing.

The Samsung Galaxy Z Fold 8 is smaller, cheaper, and has fewer cameras than the Galaxy Z Fold 8 Ultra, but it may end up being the most popular Galaxy foldable to date.

Why? Because its 7.6-inch inner screen has a 4:3 aspect ratio, which is supposedly “the new standard for viewing content,” as Samsung told us in an interview — and if there’s one thing most of us can’t get enough of in 2026, it’s content.

The Galaxy Z Fold 8 Ultra, meanwhile, has an eight-inch inner screen with a 10:9 aspect ratio, which is great for spreadsheets and multitasking, but less brilliant for movies and YouTube videos.

Don’t believe me (or Samsung)? I compared the experience of streaming Netflix, watching YouTube, and scrolling through Instagram Reels on the Galaxy Z Fold 8 and Galaxy Z Fold 8 Ultra — in both portrait and landscape orientations, where possible — to help you visualize the aspect ratio differences between these two devices.

Check out the side-by-side comparisons for each platform below, and let me know in the comments section which phone you’d prefer to use for watching videos. I've also included a roundup of each model's respective size specs for added context.

Netflix

Z Fold 8 (10:16) / Z Fold 8 Ultra (9:21)

Z Fold 8 (4:3) / Z Fold 8 Ultra (10:9)

Z Fold 8 (3:4) / Z Fold 8 Ultra (9:10)

YouTube

Z Fold 8 (16:10) / Z Fold 8 Ultra (21:9)

Z Fold 8 (10:16) / Z Fold 8 Ultra (9:21)

Z Fold 8 (4:3) / Z Fold 8 Ultra (10:9)

Z Fold 8 (3:4) / Z Fold 8 Ultra (9:10)

Instagram Reels

Z Fold 8 (16:10) / Z Fold 8 Ultra (21:9)

Z Fold 8 (4:3) / Z Fold 8 Ultra (10:9)

Z Fold 8 (3:4) / Z Fold 8 Ultra (9:10)

Verdict

For Netflix and YouTube, the Galaxy Z Fold 8 provides a cleaner viewing experience than the Galaxy Z Fold 8 Ultra in every orientation except 3:4. The black bars above and below the video are noticeably narrower than those on the Galaxy Z Fold 8 Ultra, and there's no awkward pillarboxing (i.e., bars on either side of the video) when using the Galaxy Z Fold 8 in 10:16 (you'll notice that content wraps around the phone's punch-hole camera, which doesn't happen on the Galaxy Z Fold 8 Ultra).

Of course, if you use the Galaxy Z Fold 8 in 3:4, the black bars are much bigger than those on the Galaxy Z Fold 8 Ultra, but Samsung itself admits that this orientation is “more natural for reading e-books or browsing the web,” which gives the smaller device another nice win over the pricier model.

The Instagram Reels-watching experience is pretty similar across both phones, but given the Galaxy Z Fold 8's lighter weight (201g vs 215g), I'd give it the edge here, too.

So there you have it: if watching videos and scrolling through social media is your priority, the Galaxy Z Fold 8 is a better buy than the Galaxy Z Fold 8 Ultra.

Of course, there is much more to both phones than their respective displays — the Galaxy Z Fold 8 Ultra has better cameras, for instance — so I wouldn't jump headfirst into buying one over the other until you've read our full Samsung Galaxy Z Fold 8 Ultra vs Galaxy Z Fold 8 comparison.

Side note: You're not able to watch videos in 16:10 (Z Fold 8) or 21:9 (Z Fold 8 Ultra) on Netflix, nor videos in 10:16 (Z Fold 8) or 9:21 (Z Fold 8 Ultra) on Instagram, hence why these options don't appear in the above comparison.

• AI growth keeps enterprise hard drive demand climbing across global cloud infrastructure
• Cloud providers continue expanding storage despite massive investment in AI accelerators
• Enterprise storage remains essential as artificial intelligence generates unprecedented data volumes

Enterprise demand for high-capacity hard drives continues accelerating as artificial intelligence generates expanding datasets requiring reliable, affordable storage throughout increasingly complex cloud environments.

That growing appetite for storage has intensified competition among cloud providers building artificial intelligence infrastructure, keeping enterprise hard drives firmly within long-term investment plans.

Seagate's recent fiscal fourth-quarter and full-year 2026 financial results indicate continued demand from cloud data center customers despite no disclosure of future HDD allocation commitments.

Enterprise customers continue driving demand for mass-capacity HDDs

AI relies heavily on enormous storage capacity because valuable data must remain available long after computations are done.

As organisations deploy larger AI models, the amount of stored information grows alongside computing requirements.

This significantly increases the dependence on enterprise hard drives for economical long-term retention.

Unfortunately, HDDs are becoming harder to secure, with manufacturers already booking long-term supply commitments years in advance.

Many HDD manufacturers have locked in contracts extending into 2028 and 2029, meaning customers wanting drives that far out may already be shut out of available slots

Seagate said robust cloud data center demand kept driving business growth, tying stronger storage purchases to AI infrastructure expanding across global markets.

"As AI accelerates data generation and its value, we see durable long-term demand for mass capacity storage," said Dave Mosley, Seagate's chair and chief executive officer.

Seagate expects strengthening exabyte demand through its Mozaic platform and differentiated HAMR technology roadmap as enterprise deployments continue expanding worldwide.

While Mosley did not explicitly say customers are reserving Seagate's hard drive production through 2029, his remarks make clear that enterprise demand now sits at the core of the company's strategy.

Financial performance supports the broader storage story

Seagate's latest financial results provide additional evidence supporting enterprise storage demand.

Its annual revenue increased during fiscal 2026, while Q4 performance exceeded company expectations for revenue.

Operating cash flow reached $3.7 billion during fiscal 2026, while free cash flow totalled $3.1 billion after fourth-quarter operating cash flow reached $1.3 billion.

Seagate also reduced total debt by $1.4 billion during fiscal 2026, including $302 million during the fourth quarter, ending with outstanding debt of $3.6 billion.

Cash and cash equivalents reached $1.7 billion, while ordinary shares outstanding totalled 227 million and directors approved a quarterly dividend of $0.74 per share.

Seagate expects first quarter fiscal 2027 revenue near $4.1 billion alongside non-GAAP diluted earnings per share of approximately $7.30, subject to stated variations.

"Seagate's strong fourth quarter exceeded our expectations for revenue and non-GAAP EPS, capping a fiscal 2026 in which we grew,” Mosley said.

“Our performance is being driven by robust cloud data center demand and disciplined execution, and we see the momentum continuing in 2027."

Seagate’s rival, Micron, already made the move that Seagate is currently hinting at. It exited its Crucial consumer RAM businesslin 2025 to redirect supply toward enterprise customers.

Micron said the shift was meant to "improve supply and support for our larger, strategic customers in faster-growing segments," cutting consumer buyers off from a manufacturer they had relied on for years.

Seagate has not gone that far with hard drives, but Mosley's insistence that cloud demand is driving the business suggests it could prioritise enterprise demand if current trends continue.

Via TechPowerUp

The SteelSeries Stratus+ is a wireless Bluetooth controller built primarily for Android mobile gaming, with wired USB-C support for Windows PCs.

It launched back in 2022 at $59.99 / £64.99, but has since tumbled to sale prices around $25 / £25. It’s a welcome price drop because it’s about the only thing that keeps it in with a shot of being worth picking up, and on the periphery of the best mobile controller conversation — but only just.

SteelSeries positions the Stratus+ as delivering a console-quality mobile experience, and in theory, it has most of the ingredients to do that. A traditional shape that blends Xbox and PlayStation design elements with a removable phone mount, Hall Effect triggers, and a claimed 90-hour battery life via Bluetooth LE. In practice, however, the phone mount that should be the controller's defining feature ends up being its biggest weakness. It’s tricky to install, and once it’s in, it wobbles, spoils the balance, and worst of all, it has a habit of turning your phone off altogether.

Once you have found the sweet spot to mount your phone, game compatibility is then a mixed bag. Fortnite on mobile is often kind on controller compatibility and worked well here, but more finicky titles like Call of Duty Mobile and Need for Speed didn't work at all. Xbox Cloud Gaming fared better, performing well enough with Forza Horizon 6, though compounding Bluetooth latency with remote gaming meant noticeable enough input lag that it forced me into pre-empting inputs rather than driving naturally.

For a mobile controller, the Stratus+ actually performs best as a wired PC controller. Plug it in via USB-C, and it's instantly recognized as an XInput device, fully mapped in Steam, and perfectly functional as a basic controller. The problem is that it's a particularly competitive market, and it does very little to stand out. You could pick up an actual Xbox Wireless Controller for not much more and get a better, more comfortable pad that still pairs to your Android phone.

SteelSeries Stratus+ review: Price and availability

• List price: $59.99 / £64.99 / AU$159.99
• Now often found as low as $25 / £25 from third parties
• Available from SteelSeries and retailers

The SteelSeries Stratus+ has been around a little while now, first launching in May 2022 at $59.99, though fast-forward to 2026 and that list price isn’t holding up. It’s now often available much cheaper from retailers, and I saw it selling for less than half its list price while pulling together this review.

At full price, the Stratus+ is stuck trying to compete with options like the first-party Xbox Wireless Controller and sits uncomfortably close to the excellent GameSir G8 Galileo mobile grip or the GameSir G8+ Bluetooth controller, all of which offer better experiences in their own way.

At its current discounted price, however, it’s cheap enough to be worth a look as a budget PC gamepad.

SteelSeries Stratus+ review: Specs

SteelSeries Stratus+ review: Design and features

• Blends elements from Xbox and PlayStation controllers into a familiar shape
• Included phone mount isn’t really fit for purpose
• Solid D-pad and Hall Effect triggers are the highlight

Before you start messing around with the phone mount, the SteelSeries Stratus+ looks like a perfectly generic controller. The sort of thing you’d see in a film to avoid giving brand recognition to either side of the console war. It borrows PlayStation’s symmetrical thumbstick layout and pairs it with Xbox-style ABXY buttons and a black shell. It's inoffensive, generic, and unexciting.

Holding it, though, something feels off. Despite being almost identical in width to an Xbox Series X controller and actually a touch taller, the Stratus+ somehow feels cramped. Part of that is down to the symmetrical stick placement, I’m sure, as it naturally sits lower on the body and makes the whole controller feel shorter than it physically is. I have reasonably large hands and am very used to an offset stick layout, so it never quite settled. Some of this is obviously personal preference, and if you’re coming from a DualSense Wireless Controller, I’m sure you’ll adapt quicker, though I doubt you’ll escape that sense of being on top of yourself.

The phone mount is where things stray from the norm and, unfortunately, not particularly successfully. SteelSeries includes a detachable clip that slots into the top of the Stratus+ via two metal bars to turn this from a controller into a mobile controller. It doesn’t work well, though, and almost feels like a third-party modkit for a controller, rather than a first-party USP. It’s fiddly to install, and even with the bars fully inserted, there's noticeable wobble before you've even added a phone.

Once you do seat a device, the controller becomes top-heavy and unbalanced, constantly wanting to tip backward when placed on a surface and feeling like it’s trying to spill forward out of your hands. The mount clamps your phone from the sides rather than end-to-end, which helps with universality, but means it's also constantly trying to press your phone's power and volume buttons. During testing with a POCO X5 Pro, I had to significantly offset the phone to the right of the cradle to avoid it turning my phone off.

The D-pad is probably the highlight of the entire controller, not something I’ve often said, so it’s nice to see it get its moment to shine. It's firm with a satisfyingly soft press and rock solid with zero wiggle. The ABXY face buttons are generic but functional, and the bumpers are quiet, soft, and responsive.

There are Hall Effect sensors in the triggers with a mid-length travel and linear mapping, though there’s no hair trigger mode and no companion app to set dead zones. The ALPS thumbsticks offer fairly high resistance and felt fine during testing with no drift out of the box, though it's worth noting that these sticks don't carry the same drift-resistance guarantees as Hall Effect alternatives.

SteelSeries Stratus+ review: Performance

• Bluetooth latency is noticeable at times
• Mobile game compatibility is inconsistent
• Wired PC performance is where it’s at its best

The SteelSeries Stratus+ is trying to be two controllers in one, and in testing, the experience was certainly a tale of two halves.

Connected to my Windows gaming PC via USB-C, the Stratus+ was instantly recognized as an XInput controller and fully mapped as a generic gamepad in Steam. You’d think that’d be the bare minimum experience, but it’s actually refreshing to see it just work, as that isn’t always the case, I’ve found.

I jumped across a few titles in both Steam and Xbox Cloud Gaming, and there's nothing here that would give you reason to reach for a different controller. It’s perfectly fine but equally, there’s nothing worth shouting about, or that would make you recommend this one specifically. It looks like a generic game controller, and it plays like one, but most importantly these days it’s priced like one too, so that’s not a bad thing.

Sticking that weird phone mount into the top and heading to mobile, the experience is considerably less convincing. The SteelSeries Stratus+ connected to my POCO X5 Pro without issue, but game support is a guessing game. Fortnite recognized the controller and played well overall, but displayed PlayStation button prompts on screen despite the Stratus+ using an ABXY layout.

The firm resistance of the sticks offered me plenty of fine control, and any input lag was small enough to be below my skill cap. Call of Duty Mobile and NFS No Limits simply didn't work at all, with no controller recognition and no hardware mapping workaround available.

Again, this isn’t a SteelSeries-specific issue, but competitors like GameSir have built compatibility solutions via their companion apps, so while those games don’t have great native controller support, it’s possible with a little effort by brands.

Xbox Cloud Gaming fared better, recognizing the controller natively and without any button mapping issues. Forza Horizon 6 played OK, but this is where the Bluetooth latency became most apparent when stacked on top of the small amount of remote gaming input lag. Racing games demand reactive inputs and a feeling of being connected to your car, and with the Stratus+, there was enough delay to force me into adapting my play style.

Rather than reacting naturally to what was happening on screen as I would normally, I found myself having to preempt my next movement while still executing the last one, building in a mental delay to compensate. It's not catastrophic, and it's less noticeable in games that don't demand constant precision, like the charming Sledding Game or Stardew Valley, which both played fine.

SteelSeries claims 90+ hours of battery life, and while I unfortunately couldn’t devote nearly four straight days to gaming, I can see those numbers being true enough. Despite regular use over a week of testing, the battery percentage reported by Android barely seemed to drop, and any time I was gaming on PC was passively juicing it up again anyway. Curiously, when charging, I noticed there was often quite a discrepancy between what the controller's own LED indicators showed versus what my phone reported.

At one point, I had three solid lights and one flashing during charging, suggesting it should be 75%, yet my phone's Bluetooth settings reported just 47% battery. There’s fast charging available here, so ultimately it isn’t a dealbreaker, as I doubt you’ll be plugged in often, but I’d be inclined to trust the more pessimistic software number over the lights.

Should I buy the SteelSeries Stratus+ review?

Buy it if...

Don't buy it if...

Also consider...

If you're looking for something else in either the dedicated mobile or PC controller space then consider these fine options.

How I tested the SteelSeries Stratus+

• I spent a week gaming with the Stratus+ across a POCO X5 Pro (Android) and a Windows PC
• I tested native Android games like Fortnite and Call of Duty Mobile, Forza Horizon 6 and Sledding Game via Xbox Cloud Gaming, and Steam titles via wired USB-C
• I paid particular attention to things like Bluetooth latency, game compatibility, and phone mount stability

I approached testing the Stratus+ independently as both a mobile controller and a PC gamepad over the course of a week, along with considering whether it successfully comes together as one hybrid unit. On Android, I tested both local titles and cloud gaming, while on PC, I tested through Steam's controller interface and games, along with Xbox Cloud Gaming, to evaluate its viability as a wired desktop pad.

I used a POCO X5 Pro in the included phone mount, paying particular attention to stability, balance, and how comfortable the overall experience was in this arrangement. I played both at home on the couch and at a desk, as well as taking the controller out and about to see how it fared on the move.

First reviewed May-July 2026

Read more about how we test

Boy, sure is a hot one today, huh? Okay, I get the irony: that's a take so cold you could use it to replace your freezing Wim Hof showers. But when the metaphorical mercury creeps this high, you've got to keep your cool. The same goes for your laptop, which is why you might want to grab a laptop cooling pad this summer.

Fortunately for you, I've spent this week testing every laptop cooling pad I can get my hands on to see which offers the superior AC for your computer. Not only that, but I've brought receipts — this piece has more graphics and graphs than election night coverage. So if you want objective numbers about which pad is genuinely gusty and which is all bluster, I've got the data you need.

I get that it can be tempting to overlook cooling your laptop during warm weather — I've been more worried this summer about tracking down a fan to keep myself cool than my laptops. But you overlook the best laptop cooling pads at your peril.

I still have vivid memories of the summer I redlined my laptop during a heatwave. While the video skittering should have been a hint, it was only when it stopped booting up and started bleeping SOS in Morse code that I began questioning my life choices. Turns out I'd burnt out the memory — quite literally: there were scorch marks on the inside of the case.

So at this time of year, it's worth considering whether a laptop cooling pad can help you out. And I can definitely help out there.

Laptop cooling pads: cooling

Temperature Rise Using Laptop Cooling Pad

The temperature rise during a stress test while using the laptop cooling pad on max settings

Temperature difference (°C)Temperature difference (°F)

Baseline

31.9

TopMate C12 Laptop Cooling Pad

22.8

Klim Mistral

20.4

Klim Wind

19.5

Almiuzi F2 Plus Laptop Cooling Pad

18.4

TECKNET Laptop Cooling Pad

17.9

Klim Ultimate

15.5

Liangstar Laptop Cooling Pad

15.3

Llano V10 Gaming Laptop Cooling Pad

12

Llano RGB Laptop Cooling Pad

8

010203040

Temperature difference (°C) Data

Product	Value
Baseline	31.9
TopMate C12 Laptop Cooling Pad	22.8
Klim Mistral	20.4
Klim Wind	19.5
Almiuzi F2 Plus Laptop Cooling Pad	18.4
TECKNET Laptop Cooling Pad	17.9
Klim Ultimate	15.5
Liangstar Laptop Cooling Pad	15.3
Llano V10 Gaming Laptop Cooling Pad	12
Llano RGB Laptop Cooling Pad	8

Baseline

57.4

TopMate C12 Laptop Cooling Pad

41

Klim Mistral

36.7

Klim Wind

35.1

Almiuzi F2 Plus Laptop Cooling Pad

33.1

TECKNET Laptop Cooling Pad

32.2

Klim Ultimate

27.9

Liangstar Laptop Cooling Pad

27.5

Llano V10 Gaming Laptop Cooling Pad

21.6

Llano RGB Laptop Cooling Pad

14.4

015304560

Temperature difference (°F) Data

Product	Value
Baseline	57.4
TopMate C12 Laptop Cooling Pad	41
Klim Mistral	36.7
Klim Wind	35.1
Almiuzi F2 Plus Laptop Cooling Pad	33.1
TECKNET Laptop Cooling Pad	32.2
Klim Ultimate	27.9
Liangstar Laptop Cooling Pad	27.5
Llano V10 Gaming Laptop Cooling Pad	21.6
Llano RGB Laptop Cooling Pad	14.4

Lower values indicate better cooling.

Assessing which laptop cooling pad is a breezy wonder and which is a wheezy blunder ultimately comes down to how frosty each model's cooling is. That's why I went through every pad I could get my hands on to see how much each one prevented our gaming laptop from heating up during a 3DMark stress test.

As you can see, by far the weakest cooling offered by any laptop cooling pad I tested was from the TopMate C12 Laptop Cooling Pad. Compared to the 57.4°F / 31.9°C baseline rise of our gaming laptop alone, the C12 was only able to limit it to a 41°F / 22.8°C rise — that might be better than frantically waving a dish towel in its general direction, but only barely.

Probably the biggest disappointment of these tests was how the Klim Mistral performed. Not only did it seem to perform worse than the more affordable Klim Wind — resulting in 36.7°F / 20.4°C warming compared to the Wind's 35.1°F / 19.5°C — but it was pretty inconsistent to boot, achieving even worse results on a retest.

Skipping ahead a little, the laptop cooling pad I was most surprised by was the TECKNET. Although it's cheaper than quite a few of the pads I tested here, setting me back just $21.59 / £21.99, the 32.2°F / 17.6°C warming it presided over put it firmly in the top half of the pack. It's the little laptop cooling pad that could.

Given that the Liangstar Laptop Cooling Pad and Klim Ultimate both achieved an almost photo finish, it would be easy to assume there's not much daylight between them. However, like a bottle of wine, you really need to weigh up what you're spending here to calculate its true value. And as the Liangstar costs $20.99 / £23.59 versus the Klim Ultimate's $44.97 / £44.97 price tag, that definitely takes the edge here.

But, frankly, there's a reason Llano is the absolute GOAT in my eyes. As you can see, it took both first and second place on my cooling pad podium. While the Llano V10 Gaming Laptop Cooling Pad stopped our gaming laptop from warming up more than 21.6°F / 12°C, the Llano RGB Laptop Cooling Pad kept it down to just 14.4°F / 8°C of warming. This boi makes liquid nitrogen look positively toasty.

Laptop cooling pads: noise levels

Previous

1 / 2

Next

Noise Level — cooling pad alone

Noise level of the cooling pad or laptop alone taken at head height and from 3 inches away

Head heightNear

Llano RGB Laptop Cooling Pad

71

Llano V10 Gaming Laptop Cooling Pad

64.5

Klim Mistral

52.6

Liangstar Laptop Cooling Pad

45

TopMate C12 Laptop Cooling Pad

43.5

Laptop alone

42

Almiuzi F2 Plus Laptop Cooling Pad

37.4

Klim Wind

36.2

TECKNET Laptop Cooling Pad

36

Klim Ultimate

36

020406080

dB

Head height Data

Product	Value
Llano RGB Laptop Cooling Pad	71
Llano V10 Gaming Laptop Cooling Pad	64.5
Klim Mistral	52.6
Liangstar Laptop Cooling Pad	45
TopMate C12 Laptop Cooling Pad	43.5
Laptop alone	42
Almiuzi F2 Plus Laptop Cooling Pad	37.4
Klim Wind	36.2
TECKNET Laptop Cooling Pad	36
Klim Ultimate	36

Llano RGB Laptop Cooling Pad

85

Llano V10 Gaming Laptop Cooling Pad

77

Klim Mistral

62.6

Liangstar Laptop Cooling Pad

59

TopMate C12 Laptop Cooling Pad

58

Laptop alone

60

Almiuzi F2 Plus Laptop Cooling Pad

50.2

Klim Wind

48.8

TECKNET Laptop Cooling Pad

60

Klim Ultimate

46

0255075100

dB

Near Data

Product	Value
Llano RGB Laptop Cooling Pad	85
Llano V10 Gaming Laptop Cooling Pad	77
Klim Mistral	62.6
Liangstar Laptop Cooling Pad	59
TopMate C12 Laptop Cooling Pad	58
Laptop alone	60
Almiuzi F2 Plus Laptop Cooling Pad	50.2
Klim Wind	48.8
TECKNET Laptop Cooling Pad	60
Klim Ultimate	46

Noise Level — cooling pad & laptop

Combined noise level of the cooling pad and laptop taken at head height and from 3 inches away

dB Head heightdB Near

Llano RGB Laptop Cooling Pad

64

Llano V10 Gaming Laptop Cooling Pad

57.5

Klim Mistral

51.5

TopMate C12 Laptop Cooling Pad

45

TECKNET Laptop Cooling Pad

45

Liangstar Laptop Cooling Pad

45

Klim Ultimate

44.4

Klim Wind

44

Almiuzi F2 Plus Laptop Cooling Pad

44

020406080

dB Head height Data

Product	Value
Llano RGB Laptop Cooling Pad	64
Llano V10 Gaming Laptop Cooling Pad	57.5
Klim Mistral	51.5
TopMate C12 Laptop Cooling Pad	45
TECKNET Laptop Cooling Pad	45
Liangstar Laptop Cooling Pad	45
Klim Ultimate	44.4
Klim Wind	44
Almiuzi F2 Plus Laptop Cooling Pad	44

Llano RGB Laptop Cooling Pad

79

Llano V10 Gaming Laptop Cooling Pad

69

Klim Mistral

61.5

TopMate C12 Laptop Cooling Pad

58.5

TECKNET Laptop Cooling Pad

58

Liangstar Laptop Cooling Pad

59

Klim Ultimate

55.2

Klim Wind

58.5

Almiuzi F2 Plus Laptop Cooling Pad

61

020406080

dB Near Data

Product	Value
Llano RGB Laptop Cooling Pad	79
Llano V10 Gaming Laptop Cooling Pad	69
Klim Mistral	61.5
TopMate C12 Laptop Cooling Pad	58.5
TECKNET Laptop Cooling Pad	58
Liangstar Laptop Cooling Pad	59
Klim Ultimate	55.2
Klim Wind	58.5
Almiuzi F2 Plus Laptop Cooling Pad	61

You may think that your gaming laptop makes a lot of noise. But believe me: that's scarcely more than a murmur compared to some cooling pads. Given they're essentially just an LED-adorned box with a whopping great fan inside, they can get pretty shouty. That's why I grabbed the sound level meter from my utility belt and started measuring like the world's nerdiest superhero.

Perhaps predictably, the pad that blows coolest blows loudest. While I love the Llano RGB Laptop Cooling Pad's glacial chilling, I can't get away from the fact that it's pretty noisy. I'm 5 foot 11 inches tall, and from my head height, I clocked it at 71dB — according to the Center for Hearing and Communication, that's roughly as loud as a coffee grinder or freeway traffic. In my book, that's a worthy price to pay for keeping your laptop cool, but it might not be ideal if you're planning to use it in the office.

In fact, you might assume there's an inverse relationship between cooling and fan noise. And while that does hold true for the most part, there are some interesting exceptions. For example, while I found the Liangstar Laptop Cooling Pad offered the third best cooling in my tests overall, it was substantially quieter than either of the Llanos, clocking in at a much less brash 45dB.

Conversely, the Klim Mistral manages to buck the trend in the other direction, running surprisingly loud compared to some of the other competitors here. At 52dB, it kicks out a similar level of noise as a fridge — which is pretty ironic given it offered some of the worst cooling in my testing.

And there are some surprising candidates for the winner's podium when it comes to sound levels. The TECKNET offers far from the worst cooling, and yet it ties for first place as the quietest pad on test here at 36dB — again, not at all bad given its super affordable price tag.

Honestly, though, if you want whisper-quiet running without sacrificing performance, my pro tip would be to check out the Klim Ultimate. Unlike Klim's two cheaper cooling pads, it offers very credible chilling while also measuring noise levels of 36dB from head height — that's apparently quieter than a library, which isn't known for being a raucous place.

Fundamentally, noise levels probably shouldn't be the deciding factor when picking your perfect laptop cooling pad. Both cooling power and budget come higher up the pecking order in terms of my personal preferences. But that doesn't mean you should ignore it completely — depending on where and how you use your laptop, you could find strapping a jet engine to the bottom of it to be a bit of a deal breaker.

Do I even need a laptop cooling pad?

Previous

1 / 2

Next

MacBook Air M5 encode test

Performance after an hour during a standard encode test both with and without cooling

Versus Comparison Data

Attribute	Benchmark — MacBook Air M5	Klim Ultimate — MacBook Air M5
Average FPS	2.52	2.77
Estimated encode time	122	114

Benchmark — MacBook Air M5Klim Ultimate — MacBook Air M5

VS

Benchmark — MacBook Air M5Klim Ultimate — MacBook Air M5

Average FPS (fps)

2.52

2.77+10%

Estimated encode time (minutes)

+7%122

114

Re-encoding a 4K stereo .mov file at highest quality, the MacBook Air M5 showed a modest reduction in throttling in terms of average FPS and encode time when using the Klim Ultimate laptop cooling pad.

Acer Predator Helios 300 encode test

Performance after an hour during a standard encode test both with and without cooling

Versus Comparison Data

Attribute	Benchmark — Acer Predator Helios 300	Klim Ultimate — Acer Predator Helios 300
Average FPS	2.8	2.8
Estimated encode time	115	113

Benchmark — Acer Predator Helios 300Klim Ultimate — Acer Predator Helios 300

VS

Benchmark — Acer Predator Helios 300Klim Ultimate — Acer Predator Helios 300

Average FPS (fps)

2.8

2.8

Estimated encode time (minutes)

+2%115

113

Re-encoding a 4K stereo .mov file at highest quality, the Acer Predator Helios 300 showed essentially no reduction in either average FPS or estimated encode time, suggesting limited heat-related throttling took place.

So it's pretty clear which cooling pads offer ice-cold AC and which sound like a tumble dryer breaking the sound barrier. But I've left the biggest question until last — do you actually need one at all?

Aiming to prove my theorem that 'cold CPU = good', I whipped up a quick encode test designed to stress out the CPU of several laptops over a longer period. In theory, after an hour, their core temperature should have risen, their CPUs throttled, and the length of the encode spiked and the encode rate dropped. Running the same test with one of our top cooling pads, the Klim Ultimate, would then show how much it could reduce this throttling.

On my MacBook Air M5, after an hour of using its passive cooling alone, the total estimated encode time had crept up to 122 minutes, and the average encode rate was sitting at 2.52 FPS. When I reran the test with the Klim Ultimate running on max cooling, that total encode time was trimmed to 114 minutes, and the encode rate had increased to 2.77 FPS. That's not a huge difference, but it shows the Klim Ultimate was able to trim back some of that temperature-related throttling.

Running the same test on our Acer Predator Helios 300 was a much more ambiguous experience. During my first run, the encode crashed within minutes, and during the second it crashed moments after the test concluded. So it was honestly far from stable.

I was able to eventually get some results, however, and they showed... nothing. Whether I was using the Klim Ultimate or not, the test results were essentially identical, with Handbrake clocking an estimated encode time of 115 minutes uncooled and 113 with the Klim's help. Meanwhile, the encode rate remained steady, sitting at a reasonable 2.8 FPS.

What does that mean? That I've wasted several hours of my life? Not quite. There are a couple of key conclusions to take away from this. One is that a performance laptop with whacking great fans built in is already pretty good at keeping itself cool. Shocking revelation, I know.

But probably the more critical takeaway is that the mileage you get out of a laptop cooling pad is going to vary depending on how well optimized it is for handling heat, as well as the kind of tasks you're doing. Had I tried to run Cyberpunk 2077 on max graphics settings with ray-tracing switched on, the Predator Helios 300 would likely have caused it to melt into the earth's core.

There's also another caveat. While it's definitely been warm the last week or so in the TechRadar testing lab, it hasn't reached the kind of desiccating highs that have hit both the US and UK in recent months. Past a certain point, any laptop will struggle in the heat. Forget CPU throttling — in that heat, I'd be more worried about your components liquefying like a chocolate bunny on the launch pad at Cape Canaveral.

In light of that, I'd say a laptop cooling pad is still worth exploring, especially if you tend to do a lot of resource-intensive tasks like 3D work, Photoshopping or gaming.

Any thoughts or personal experiences you'd like to share? Please feel free to chime in down in the comments!

Valve's SteamOS has often presented the perfect excuse to be whisked away from Microsoft's troublesome Windows 11. The Linux operating system's better gaming performance has often been proven, utilizing RAM well without the bloatware often found with Windows 11.

SteamOS provides a console-like user interface (UI), making navigation and game launches far easier than Windows 11, despite the recent Xbox mode. With those benefits in mind, I've been contemplating a full switch to SteamOS, ditching Windows for good, and the best way to do that would be via Bazzite.

Bazzite (not affiliated with Valve) is effectively a SteamOS clone, specifically designed for "Linux newcomers and enthusiasts alike", but most importantly, comes with out-of-the-box compatibility with Nvidia and Intel GPUs, and a smooth experience for handhelds.

Using my second streaming PC, I put Bazzite to the test over a few months. Without a doubt, it's easily the best option for any gamer looking for a console-like couch gaming setup.

With tools like Decky Loader, customization is easy, with plugins to change SteamOS game mode themes, navigation sound effects, or even the startup and shutdown videos. However, despite all of the benefits of using the operating system, I still can't step away from Windows 11 on my main system, and there are multiple reasons why.

Multitasking drawbacks

As a frequent multitasker using applications like Discord to stream games to friends, Bazzite (more specifically SteamOS) isn't an ideal option. Using Discord to stream games on the same rig works just fine in desktop mode; however, it's not the same story for its game mode, which is arguably the entire purpose of using Bazzite.

Streaming on Discord in game mode severely drops frame rates in games, with constant stutters in many I tested, and there are a few instances where attempting to stream an application will result in a black screen. I tried to use the NZXT Signal 4K30 capture card and stream my second PC's Bazzite activity on Discord via my main PC, but it's not that simple.

Sure, HDMI passthrough on the capture card is possible, but I didn't want to be limited to 4K 60 frames per second on either my TV (which is capable of 144Hz) or my gaming monitor (capable of 240Hz). Frankly, capture cards with higher passthrough capabilities don't come cheap.

Duplicating or projecting the display, as you would on Windows 11, would be the easy solution. However, Gamescope, the compositor responsible for displaying games on screen with Bazzite, cannot duplicate outputs.

Still trying to avoid the NZXT capture card's 4K 60Hz passthrough limitation, I opted for an HDMI splitter with one input and two output ports. Unfortunately, the moment the capture card is connected, the output on either my TV or monitor has the refresh rate slashed to 60Hz and VRR disabled.

That occurs to ensure that the Extended Display Identification Data (EDID) of both is matched, according to what the capture card is capable of in passthrough.

None of these issues are present on Windows, as you can use OBS to project the main display onto the capture card without ever having to rely on an HDMI splitter or a better capture card.

This isn't an attack on SteamOS or Bazzite in particular; after all, game mode is quite literally designed for gaming, so multitasking isn't a priority for Valve or specifically Bazzite developers in this case.

However, it's worth noting that streaming on the gaming OS is more trouble than it's worth, and that's without even mentioning the potential audio bugs or OS glitches. That's only one part of why it's worth sticking with Windows, though.

Compatibility issues

I wish these issues stopped there, but not quite. Firstly, a large majority of popular multiplayer games such as Battlefield 6, Call of Duty (most of the modern entries), Destiny 2, and EA Sports FC 26 will not work on Linux due to their anti-cheat systems.

This is an issue that lies directly with publishers and developers for those games that refuse to adapt their anti-cheat systems to work on SteamOS. It also doesn't seem as though this will change any time soon; the Steam Machine was arguably one way to effectively force the hand of publishers to consider anti-cheat compatibility on Linux, but it's not exactly a popular PC due to its high price.

It's also worth noting that Game Pass is inaccessible natively on Bazzite, and requires either cloud gaming or third-party launchers, and you'll be lucky if you can get the latter to function.

The solution to accessing those anti-cheat games is a dual-boot with Windows 11 and Bazzite. However, that isn't completely ditching Windows 11 as I've planned to, and a dual-boot can end up messy thanks to Windows' forced updates.

Perhaps the most frustrating aspect is that Nvidia GPUs don't play nice with game mode on Bazzite, often resulting in glitchy functionality, or just completely black screens. I tested this by swapping out the AMD Radeon RX 9060 XT 16GB GPU in my second PC with my spare Nvidia GeForce RTX 3080 Ti and found that game mode was just frustrating to use.

So if I wanted to use Bazzite on my main rig equipped with an Nvidia GeForce RTX 4080 Super, its DX12 (a graphics API) gaming performance would be worse than Windows. Overall, the entire purpose of switching over to Linux was for better gaming performance and easy console-like UI navigation would be non-existent.

I'm sure all of these pain points will be addressed at some point, whether via Bazzite or Valve's official SteamOS builds. For now, though, I'll stick with Windows 11 on my main rig — even though I desperately want to be done with it.

The current state of music streaming is interesting, yet unpredictable. The shift from the CD era into the streaming era has made access to music more ubiquitous, but it’s also thrown up problems: some users aren’t interacting with music like they did before, and the explosive spread of AI is threatening the entire streaming model. Enter Mauhan M. Zonoozy, former Head of Innovation at Spotify, who has plans to tackle both these problems.

Zonoozy left Spotify in 2023, and since then has set up his own music software label, A Vinyl Bar in Shibuya, a record label-meets-production studio that releases playful music apps the way labels drop records, including jetset.fm and Stacks. Essentially, it’s creating digital music playgrounds.

While Shibuya’s products don’t have extensive libraries of music like typical streaming platforms, they offer users different ways to create new sounds, even if you only use them for a couple of minutes. But despite the label still being in its infancy, Zonoozy knows exactly where he wants to take it.

I chatted with Zonoozy about where his post-Spotify career is heading, and his ambitious plans to insert his music services into a streaming world in which AI is becoming a growing presence.

Music is participatory

In a short period of time, Mauhan and his “small team of creative technologists” have released a slew of casual, interactive music platforms across web and mobile — but this is only the beginning.

“I spent years at Spotify and I loved it,” Mauhan tells me. “Incredible company, incredible team. But I think I spent most of my time effectively thinking about what comes next. I started A Vinyl Bar in Shibuya around a simple belief that music is a verb, that the next major music platform should give people something to do with the music. It should be more participatory, more engaging”.

Streaming has certainly made music a lot more accessible, but there was always one thing missing for Mauhan: interaction. “With streaming, you can hear almost anything, anywhere instantly, but the dominant interaction is still effectively sorting through folders and pressing play,” he continues. “There's an entire generation that was growing up playing with music through different mediums and different devices. I think their relationship with everything felt a lot more interactive.”

This sparked a lightbulb moment for Mauhan, and his mission statement is simple: “to empower more people to express themselves with music, or feel self-expressive with music”. This is where his latest project, Bop, comes into play.

Bop is a pocket DJ app, currently in beta, that lets you mix, mash, and mess with music. Its interface is like an MPC workstation, allowing you to create remixes of popular songs by dragging and dropping effects and sounds onto its grid. “We're not all about replacing the creative process,” Mauhan adds. “We're more about building instruments and new ways to express yourself.”

When it comes to the best music streaming services, tools like this aren’t a new concept — Spotify has its own DJ-style mixing feature — but Mauhan believes there’s still more room for music services whose focus is user expression.

“I think there's a yearning for creative expression (in music platforms),” he says. “I think listening to music the way artists created and intended is absolutely a big part of it, but I also think having Legos that you play and interact with, engage with is also very important.

“The next step in our journey is packaging that up into something that is platform scale, and we can actually deploy it and build a larger community of users and fans around it.”

Next stop: a larger music platform

It’s clear that Zonoozy has had a blast developing Shibuya’s array of sonic playground tools, but what I want to know is if there’s a larger music platform on the horizon — and turns out, there is.

“I think there's a bigger platform at the end of the tunnel for sure,” he tells me. “A Vinyl Bar in Shibuya is working actively on building it. I don't think it's gonna look or feel like any streaming platform you've seen before.” Now I’m intrigued — how will this platform compete with the likes of the music streaming big dogs? Turns out Zonoozy has thought about that too.

“We believe it's going to be more participatory,” he shares. “We're not building something that puts the music in the background, we want to put it in the foreground”. And it's as simple as that. “The hints are in the products, and we want things that feel more playful and approachable. I don't think anyone's going to accuse us of building a DSP,” he adds.

Despite it still being in early stages of development, Zonoozy has a lot of faith in Shibuya’s bigger project; but music streaming tools are evolving rapidly as is tech in general, so he’s had to do a lot of thinking ahead.

“My personal belief was that a lot of the cues around where media is going were in the gaming space,” he continues. “And I think interactive media was always a little bit ahead of the rest of the world in terms of film and music. I think because it was closer to technology, it was exploring the more creative technological boundaries of what you could do with media.

“I think now with what's going on and how powerful these computers in our pockets are becoming, there's so much new, cool stuff you can do. It's a responsibility for all of us working in music and tech to figure out ways to do it that supports the industry, the community, and more than anything, the art form.” It occurs to me that Zonoozy is as interested in conserving as he is in innovating.

Will music streaming survive without AI?

It wouldn’t be a discussion about where the music-streaming industry is going without me bringing up AI, and Zonoozy sounds dismissive about AI at first, telling me: “You know what I like about AI? It's easy to spell”.

AI is having a big influence on music platforms in many ways, with algorithmic recommendations and tools such as AI playlist generators now pretty much part of the package. Because these features are right at users’ fingertips, I ask, will there come a time where listeners won’t be able to use services without the help of AI?

Zonoozy has hope, saying, “It's hard for me to imagine a world where people are building things completely void of touching some semblance of AI, but I do believe that there will be products that are not purely driven by generative media.” While he doesn’t necessarily favor the reliance on AI tools to enhance the UX, Zonoozy isn’t against AI altogether. If anything, he acknowledges its benefits, and how it’s helped provide the foundations to his services.

“It's enabling technology for us, it's just not the consumer proposition,” he adds. “Shibuya doesn't build LLMs or ask anyone to prompt music and wait for a finished song, but we do use it as a co-pilot for coding and development. I'd be naïve to say that there's no AI somewhere in the stack there. But on the music side, all the stems are all hand produced”.

He continues, “I'm not anti-AI. I think AI is an incredible, powerful future-facing infrastructure. I think it can build so many cool things and empower so many cool things. I’m a traditionally strong techno optimist in that sense, I think it will lead to incredible discoveries and advancements — it makes coding infinitely easier. But AI is not our product thesis on our side.”

When it comes to streaming services, there’s a huge focus on churning out tools designed to make the listening experience more personalized, more convenient, and more interactive. Given the pace of this evolution, I find it difficult to envisage where streaming culture is headed — Zonoozy shared my uncertainty.

“In terms of streaming specifically, I don't know what's going to happen,” he admits. “The extreme levels of personalization to custom content made for you — I don't know where that line ends. Is it being personalised, or is it being generated for you?”

Meta may dominate the smart glasses zeitgeist, but it's far from the only player — my favorite brand in the space is actually Xreal, with its range of impressive but pricey wearable-display specs. And with the launch of its new xbx brand, Xreal's lineup now includes more budget-friendly options too.

If money were no object, Xreal’s One Pro glasses are easily the crème de la crème of smart specs. Their Bose-tuned audio is impressive for such a small package, their electrochromic dimming tech is perfect for swapping between clear vision and shaded for optimal viewing conditions, and their unique optic engine displays their image more vividly than the competition, with fewer visible distortions.

The only issue is their price — $649 / £579 (they aren't available directly in Australia), absent any discounts, although they can be found on sale occasionally.

If you’re on a budget, my preferred pick has instead been RayNeo. Its Air 4 Pro specs debuted earlier this year, and not only are they affordable at just $299 / £289, they still come with solid specs, including HDR10 support.

With its new xbx a01+, however, Xreal might have just claimed the budget crown for itself.

Budget brilliance

Now the elephant in the room is that Xreal's specs cost a bit more. At their current $299 / £269 discounted price at Xreal.com they're on par with or cheaper than RayNeo's glasses; however their official price is $359 / £349, and while they aren’t discounted at the moment, we’ve regularly seen the RayNeo Air 4 Pros selling for $249 / £249 (RayNeo also doesn't seem to sell directly to Australia) or cheaper.

At the same time, the Xreal xbx a01+ glasses boast a handful of improvements which warrant a slightly higher price.

For a start, their brightness tops out at 1,600 nits, while the RayNeo Air 4 Pro can only hit 1,200 nits. This extra brightness should allow you to get more out of the HDR10 support, offering more room for the light vs dark contrast and vivid colors to breathe across the video format’s visual range. Highs can look more high, helping lows look lower by comparison.

Aiding the image quality are the swappable lens covers.

While advertised as giving you the ability to choose which color of sunglasses cover you most want to complement your style, the xbx's box also includes a completely opaque black cover you can use with your specs. Because the specs are essentially virtual projectors, a darker environment ensures the highest-quality viewing conditions, though I found the regular sunglasses were fine even in bright sunlight.

The xbx glasses are lighter too, weighing just 62g compared to the RayNeo Air 4 Pro's 76g. Admittedly I haven't found them majorly more comfortable, but as I kicked back with them in a sun lounger while on vacation for hours on end I was relaxed enough to doze off on several occasions.

The only tradeoff is, I'd say, the audio, which doesn't seem to involve Bose tuning this time, and isn't quite as impactful as the Air 4 Pro's. That said, I'll almost always use earbuds or headphones with my display glasses, because they’re better in terms of quality and minimize sound leakage, so this downgrade didn't bother me.

Who wins? You decide

You'll also have to contend with Xreal's standard optic engine too. This is the prism that reflects the micro-OLED display into your eyes, and the standard design is susceptible to external reflections — most commonly light bouncing off the top of the specs and into your view. Pictures will also be a little smaller thanks to a reduced field of view, and can be a little blurrier at the frame edges.

Only Xreal's One Pros boast the improved design, but having experienced it, it's tough to go back to the standard optic engine.

To be fair, though, that’s mostly a me problem, and if you’re new to the display glasses world the Xreal xbx a01+ and RayNeo Air 4 Pros should be perfectly adequate performance-wise.

As for which of the budget specs you should go for — whichever is cheaper would be my suggestion.

The technical improvements of the xbx specs are solid, but if you can save $40/£40 or more on the RayNeos, their few deficiencies can be overlooked.

That said, if stylish, colorful designs are a must, alongside a marginally better image quality, then you'll be very pleased with the new Xreal xbx a01+ glasses.

I love reading — or at least that’s what I tell myself. In reality, I’ve always had a very on-and-off relationship with it, going from devouring books in days to barely opening one for weeks or months on end.

But one constant I’ve realized is that I don’t tend to prioritize reading — if there’s a show I’m binging, a new game I can’t stop thinking about, or simply lots of social engagements to attend, books will generally get put aside until I have more time on my hands.

But then something occurred to me — while I don’t always have long stretches of time that I’m happy to devote to reading, my days are full of short periods that aren’t being filled, like while I’m waiting in line or at a train station, or simply waiting for the kettle to boil. And then there are times when I’m on a train or in a cafe and might want to read, but don’t have a book with me as I travel light.

With all that in mind, back in late January, I bought the ViWoods AiPaper Reader — a pocket-sized Android ereader that’s small enough to carry with me everywhere alongside my phone — in the hopes that it would get me reading during periods that might otherwise be dedicated to podcasts or simply spent sitting or standing in slightly bored silence.

At the same time, I also had some hope that, since this ereader has full access to the Google Play Store, it might help cure my smartphone addiction — as doomscrolling was likely to be less desirable on an E Ink screen.

A tiny device that was a big reading success

Before buying the ViWoods AiPaper Reader, I hadn’t read much in a while — I had dozens of unread books on my Kindle Colorsoft and big stacks of physical books silently judging me. I might start one, but it wouldn’t stick — life would get in the way, or it wouldn’t immediately grab me enough to reach for it again.

But the ViWoods AiPaper Reader changed that. I loaded up the Kindle app on it and got in the habit of keeping it in my pocket, alongside my phone. Then, almost any time I would previously have switched on a podcast — such as when on a train or in small empty moments of my day — I’d reach for this instead.

It didn’t matter whether I had two minutes or two hours; I’d aim to read something, and because the AiPaper Reader was always in my pocket, there was no excuse not to. I didn’t even have to go to my shelves to grab a book.

It allowed me to build a consistent reading habit and make progress with books even on days when I didn’t have much time, or wasn’t massively in the mood to read. After all, reading a couple of pages while boiling the kettle is still better than nothing, and it all adds up. And best of all, because this device got me back into reading, I’m now more inclined to read my paperbacks than I have been for some time.

Was it all smooth sailing? No, there have been some days and even a few weeks where I would stop carrying the ViWoods AiPaper Reader with me. But so far, I’ve always found it easier to go back to than a larger ereader or a physical book.

An imperfect light and unnecessary AI

This isn’t the perfect pocket-sized ereader. It has a front light but not a warm light, which can make reading at night a bit less comfortable than on some of the best ereaders — especially as the brightness is adjustable, the minimum level is still quite bright.

There’s also the AI-shaped elephant in the room, which even if you know nothing about this device, you’ve probably been wondering about from its name.

Like so much recent tech, the ViWoods AiPaper Reader is trying to cash in on the AI craze; there's a dedicated AI button that you can press to speak to an AI chatbot, with models from OpenAI, Google, DeepSeek, and Qwen all built in.

These platforms can be used either as general-purpose AI chatbots or in combination with ViWoods' own apps to, for example, look up definitions of words in the books you’re reading. But personally, I’m tending to just use the Kindle app for reading, which can’t directly communicate with the AI.

Plus, using the ViWoods' AI functionality requires a Wi-Fi connection, which drains the battery faster than I want from an ereader, and it’s slower to interact with than just pulling up Gemini on my phone.

Beyond AI, some of the other software available on the ViWoods isn’t perfect, either, as while it runs a fairly recent version of Android in Android 16, there’s no promise of any version updates, and ViWoods’ own interface leaves a little to be desired. For example, there's little explanation of how some features work, and many important settings are hidden away in an ‘advanced settings’ submenu. There’s also an 11-step process to actually authorize and activate the Google Play Store on this device, despite it running Android.

Those gripes aside, I’ve generally been impressed by the ViWoods AiPaper Reader. Its 6.13-inch Carta 1300 screen is big and crisp enough to comfortably read on, its battery lasts me anywhere from a few days to a few weeks depending on how much I’m using it, and it looks like a genuinely premium device, with a slim, sleek design that’s almost up to the standards you’d expect from a premium brand like Apple.

This didn’t cure my addiction

While the ViWoods AiPaper Reader got me reading again, it was less successful at breaking my doomscrolling habit — though not a complete failure.

I figured that, since it runs Android apps, I could keep my phone in another room when I’m at home, and use the ereader for most things, with its monochrome E Ink screen sure to leave me reaching for it less.

And it somewhat worked for that — when I was disciplined about actually keeping my phone out of reach. But after a while, I got out of the habit, and even when I did keep my phone elsewhere, I’d sometimes encounter things I genuinely wanted it for, like penning a long message. So this wasn’t the perfect solution for me.

I’d perhaps have more luck if I fully replaced my phone with a device like this — and the ViWoods AiPaper Reader does have a SIM card slot, so it can fulfil most smartphone functions. But it lacks speakers and cameras, so you’d be making some unfortunate sacrifices.

Were I serious about that idea, I might invest in something like the Mudita Kompakt instead, as that’s designed specifically as a minimalist E Ink phone. So, if you’re looking primarily for something to break your smartphone habit, I’d suggest a device like that instead (or, perhaps, the similarly distraction-free Balance Phone).

Still, as a secondary function on a device I’m mostly using for books, the ViWoods AiPaper Reader has not been a total failure.

Now, if you’ll excuse me, I’ve got some reading to do.

• Cyberattack hits operational technology at more than 30 Minnesota community water systems, briefly taking treatment plant offline and forcing several towns onto manual operations
• A leaked WaterISAC memo obtained by WIRED relays a state fusion center assessment tying the intrusions to Iranian-affiliated hackers, but no US agency has formally attributed the attack yet
• The CISA had warned four days earlier that Iranian-affiliated actors were compromising internet-facing PLCs, repeating guidance that has mostly fallen on deaf ears

More than 30 community water systems across Minnesota were hit by a coordinated cyberattack, targeting the operational technology running pumps, wells, water towers, and wastewater lift stations rather than the office networks behind them.

Minnesota IT Services disclosed the attack on July 28 2026 and activated a statewide incident response, stating that it wasn't aware of any Minnesota city asking residents to change their drinking water use just yet.

A memo obtained by Wired, circulated by the water sector information-sharing group WaterISAC and carrying a restricted TLP: AMBER handling marking, relays a Minnesota state fusion center assessment that ties the intrusions to Iranian-affiliated hackers.

A warning that was more or less ignored

Four days before the attacks, CISA updated an advisory, AA26-097A, warning that Iranian-affiliated actors were compromising internet-facing programmable logic controllers made by Rockwell Automation, Schneider Electric, Siemens, and possibly others across the US water, energy, and government sectors. It documented confirmed disruptions and financial losses.

This drew no immediate response from state functionaries, and it is easy to see why: the US has tens of thousands of community water systems that serve small populations and lack budgets for dedicated cybersecurity staff.

This is despite there being a precedent: Iranian-linked actors hit the Municipal Water Authority of Aliquippa in Pennsylvania in November 2023, defacing a Unitronics controller with an anti-Israel message. CISA said at the time those devices were exposed to the internet with default passwords still in place. The advice issued then is the same advice being issued this week, which is telling about how little has changed since.

The WaterISAC memo, dated the day the attacks concluded, passes on a state fusion center report linking the activity to Iranian-affiliated actors. That memo was not meant to be public, but it has been corroborated by The New York Times and The Washington Post, which reported that they spoke to federal officials and US intelligence agencies, respectively, lending credence to the claims.

CISA later warned that attackers are changing PLC passwords to lock operators out of their own equipment, and repeated the advice that has been unchanged for years: get PLCs off the public internet and put remote access behind a VPN or gateway.

Who was affected?

The worst of the cyberattack hit Braham, a town of roughly 1,700 people, where the city said attackers shut down the operating controls, taking the water treatment plant and the well offline. Crews restored it manually within a couple of hours, and residents were told to minimize water use in the meantime.

Plymouth, a Minneapolis suburb of about 80,000, found the problem confined to equipment connected over cellular links at two water towers and several wastewater lift stations. Its IT division disconnected the affected kit from the network entirely to stop the attack and prevent retargeting during reconfiguration.

Maple Plain declared a local state of emergency to manage its response, and South St. Paul reported that some automated controls were compromised. In most cases, contingency procedures held and water and wastewater operations continued.

None of this is happening in a vacuum, however. The June memorandum that paused fighting between the US and Iran has broken down, and both are trading strikes, including recent US attacks near the Strait of Hormuz that destroyed a water facility and cut supply to more than 20,000 people. Whoever carried out the Minnesota attacks, the symbolism of targeting municipal water is unlikely to be accidental.

The damage that these intrusions caused was limited; they produced brief outages that trained staff fixed by hand. That is an outcome with a silver lining, and it depended on utilities having people who knew how to run a treatment plant without its automation, even if its digital defenses collapsed outright.

Fun fact: whenever I have days that I work remotely, I'll always have an episode of Bob's Burgers on in the background. In fact, I watch all 16 seasons of the hit Hulu and Disney+ animated comedy so often, that I could tell you exactly what episode was on without directly looking at it.

I felt incredibly prepared for the handful of Bob's Burgers quizzes already on the internet claiming to be the hardest, but nobody has yet been able to truly test the reams of random knowledge I've acquired from watching the series so often.

Should I be able to tell recite the lyrics to the song that plays over Tina and Tammy's first time collectively ditching class in season 2? Probably not.

So, behold — the genuinely most difficult Bob's Burgers quiz that I have played to date (because I made it). If you get more than 80%, I'll assume that you're actually a Wharf resident who's somehow found their way into the real world.

I have left out anything that happens in The Bob's Burgers Movie, but you'll need to have seen seasons 1-16 in order to answer all of the questions.

Check back when Bob's Burgers season 17 (hopefully) airs in early 2027... you know that the above will be updated accordingly.

Ninja Luxe Café Mini Espresso Machine: two-minute review

The Ninja Luxe Café Mini Espresso Machine is a compact version of the Ninja Luxe Café. It doesn't require an enormous amount of counter space, especially considering that this single machine grinds your whole beans directly into your filter, weighs your grinds, brews your espresso or drip coffee, and froths your milk. You won't need any other coffee accoutrements on your counter.

While I did brew some absolutely delicious coffee with the Ninja Luxe Café Mini, there is definitely a learning curve. In fact, I wouldn't say I have mastered the espresso settings yet, even after brewing many cups. Even with the identical espresso settings, sometimes I get just an ounce or so and sometimes I get a lot more. Drip coffee comes out consistently well; no issues there. I also had some trouble frothing the milk; with just a minimum amount of milk in the jug it overflowed onto the counter.

Note that even though this is a semi-auto machine, you do need to have some manual strength and dexterity to get the portafilter properly into place. Be ready for some trial and error. If you're prepared to take on a project, I think you'll be well-rewarded for your efforts.

Ninja Luxe Café Mini Espresso Machine: price and availability

• Widely available both in store and online
• Retails for $449.99 USD (£338.50 GBP, $643.73 AUD)
• Only sold in the black and chrome colorway

You can find the Ninja Luxe Café Mini Espresso Machine at many brick and mortar stores and websites, including Amazon, Best Buy, Kohl's, Target, and Ninja's own site. It retails for $449.99 USD (£338.50 GBP, $643.73 AUD) and comes in just the one color, black and chrome.

Ninja Luxe Café Mini Espresso Machine: specs

Ninja Luxe Café Mini Espresso Machine: design

• Efficient design gives it a smaller footprint
• Aesthetically pleasing, goes with any decor
• Five brew styles

The Ninja Luxe Café Mini is efficiently designed to house everything you need to make just about any sort of coffee drink in a relatively small package. When you first unpack it, you will need to determine your water hardness with the enclosed test strip and set your machine accordingly. Run a flush cycle and then you're ready to start brewing.

The machine has a conical burr grinder, a brewing system that makes both espresso and coffee, and a steam wand. It comes with two filter baskets:, a smaller 'double' and a larger 'luxe' basket (unlike most espresso machines, there's no basket for pouring a single shot).

Place the basket into the portafilter, the funnel on top of that, and then slide it into the grinding cradle. Choose your brew style. The machine evaluates your beans and determines a recommended grind size. Turn the grind dial to meet the recommended number and hit the grind button.

Once your ground coffee is the basket, remove the portafilter and tamp down the grinds with the assisted tamper. It has a spring inside so you'll apply just the right amount of pressure.

Then you can remove the funnel, place the portafilter under the Ninja Luxe Café Mini's group head and turn it until it locks into place. This takes some manual strength and dexterity, and the machine's relatively light weight means you have to brace it in place so it doesn't slide around, though it does get easier with repeated use.

You can manually select your level of strength (represented by one, two, or three beans) and temperature or just use the default. When you're done brewing, dispose of the puck of used coffee grounds, and clean the dirty components.

There is also a milk jug and steam wand. To use it, turn the manual valve on the side from brewing mode to frothing mode. The owner's manual includes detailed instructions of how to get the most froth from your milk of choice.

You'll need to clean the tamper, funnel, basket, portafilter, steam wand and jug after each use. You'll also need to empty and clean out the drip tray every few brews, and every time you froth milk. Additionally, you'll need to descale the machine periodically. However, there is no obvious way to get in and clean the grinder. That part never gets wet, but coffee oils will build up there over time.

• Design score: 4.5/5

Ninja Luxe Café Mini Espresso Machine: performance

• There is definitely a learning curve
• You need manual strength and dexterity
• Brew amounts are not consistent

Overall, I think you'll find that the Ninja Luxe Café Mini will make you some great coffee and espresso drinks. There's definitely a learning curve, but if you have the patience you'll figure it out. Of bigger concern was that sometimes the brewing seemed to just stop short of a full yield, and I'm not sure why. The same beans brewed at the same setting just didn't always yield the same size espresso. This only happened to me while brewing espresso; all three of the drip coffee settings yielded perfect results every time.

There are five different brew settings: Espresso (which is actually a double espresso), Triple (espresso), Classic (drip coffee), Rich (a richer drip coffee), and Over Ice (drip coffee brewed at a lower temperature). I tested all five of them multiple times.

Espresso

The smallest drink you can make is a double espresso, there is no single espresso setting on this machine. The owner's manual says to throw away the first espresso you brew each time you put new beans into the hopper because it takes a couple of brews for the machine to dial in correctly. I tasted it anyway and yes, you should toss it. After that first one, the espresso tasted good though sometimes it would randomly give me a shorter brew. I did notice that I consistently got a beautiful crema.

Triple espresso

The experience of brewing the triple espresso is just like brewing the regular (double) one, only you use the Luxe basket instead of the smaller one. Again, the machine would sometimes give me a tiny brew. Pictured below is one of the smaller triples; other times it was much larger. Otherwise, the espresso was good and always had a lovely crema.

Classic

The Classic drip coffee setting yielded consistently good results. In addition to adjusting the grind size, brew temperature, and strength if you like, you can also choose to brew 6, 8, 10, or 12 ounces of coffee. The user experience of brewing the drip coffee is the same as brewing espresso, except that you skip the tamping step. The machine does something totally different on the inside. You get no crema. And when you go to throw the puck away, it's very wet, as opposed to the drier, crumbly puck you get with espresso.

Rich

The rich drip coffee setting works just like the classic, only it yields a richer brew. I find I prefer the classic, as I found it a bit smoother and less bitter, but your tastes may differ.

Over ice

The drip coffee over ice setting might have been my favorite. I found it very smooth and it didn't taste too diluted from the ice.

Steam wand

I found that it took some practice to get the milk frothing just right. I tried just leaving the jug in place and frothing it from there, and it works ok but doesn't produce much foam. I tried following the instructions about angle and movement in the owner's manual, and got too much foam.

The wand will purge itself right after frothing, and you'll want to wipe off the milk from the outside of the wand right away. Otherwise it sticks on there pretty good, requiring a bit of scrubbing to get it off.

• Performance score: 3.5/5

Should you buy the Ninja Luxe Café Mini Espresso Machine?

Buy it if

Don't buy it if

Ninja Luxe Café Mini Espresso Machine: also consider

How I tested the Ninja Luxe Café Mini Espresso Machine

I brewed and drank every drink on the menu: (double) Espresso, Triple (espresso), Classic (coffee), Rich (coffee), and (coffee) Over Ice more than once. I used two different kinds of beans for each beverage, first going all the way through the menu with a dark french roast and then again with a medium-dark roast. I frothed two different kinds of milk (almond and skim dairy) several times and added the milk to espresso shots. I cleaned all of the parts that touch the coffee after each use.

First reviewed July 2026

• Robot boat launches drone without crew during challenging ocean demonstrations
• Pioneer vessel survives rough seas while supporting airborne surveillance operations
• Tethered SkyRanger drone provides continuous monitoring from unmanned ships

A robotic boat has launched and recovered its own drone off the deck, without a single crew member aboard either vehicle.

British firm ACUA Ocean and American sensor maker Teledyne FLIR Defense completed the demonstration for the UK Ministry of Defence.

The trial paired an uncrewed surface vessel named Pioneer with a small tethered drone known as the SkyRanger R70 aircraft.

Rough-water stability makes flight operations possible

The entire demonstration depended heavily on the Pioneer's ability to hold steady in genuinely rough conditions each time at sea.

In January 2026, ACUA Ocean said the vessel held steady in waves reaching 4.25 meters, or roughly 13.9 feet high.

The company reported a 60% cut in roll motion compared with traditional single-hull monohull vessels used across the industry.

That stability functions as the mechanical foundation allowing a tethered drone to fly safely off a small robotic vessel at all.

A violently rocking platform would make sustained flight operations for the SkyRanger R70 nearly impossible under open ocean conditions overall.

ACUA Ocean said Pioneer has already completed prior Royal Navy contracts covering surveillance, reconnaissance and anti-submarine warfare tasks before this trial.

Tethered drone extends persistent ocean surveillance

The system pairs a tethered uncrewed aerial system, or T-UAS, with an uncrewed surface vessel known as a robotic boat.

Teledyne FLIR's SkyCarrier is a ruggedized launch platform housed in a container that unfolds into an articulating landing pad system.

During the trial, SkyCarrier successfully launched a SkyRanger R70 drone autonomously from the deck of the moving Pioneer vessel itself.

Once airborne, the SkyRanger R70 stayed connected to Pioneer through a power and data tether, avoiding any need to land and recharge.

"Integrating the SkyRanger R70 with the Pioneer creates something genuinely new: a persistent, deployable capability," said Neil Tinmouth, chief executive of ACUA Ocean.

Richard Cunha of Teledyne FLIR Defense also said pairing the drone with Pioneer opens up an entirely new class of missions.

The SkyRanger R70 runs multiple embedded NVIDIA processors that perform real-time object detection and classification without ground input.

The combined sensor package delivers Group 2 level ISR performance while flying on a much smaller Group 1 airframe type.

UKDI, the Ministry of Defence's early-stage funding arm, backed the demonstration well before it reached any full-scale procurement decision process.

"Securing UKDI funding for this demonstrator is a significant milestone for ACUA Ocean and UK maritime autonomy broadly," Tinmouth said.

This technology will support the Royal Navy priorities including anti-submarine warfare, mine countermeasures and undersea infrastructure protection efforts overall.

That said, a successful demonstration under UKDI funding does not guarantee the system will reach large-scale deployment across naval operations soon either.

Cost, reliability at sea over long periods, and integration with existing fleets remain open questions that testing alone cannot resolve.

Via Defence Blog

Torn ear tips. Snapped charging case. Dented drivers. Like every other kind of gadget imaginable, the best headphones aren't built to last, and I'd wager that everyone reading this article has some skeletons (of old, ailing gadgets) in their (metaphorical) closet.

In fact, according to a European Union report, in 2024, 51% people kept their old smartphones even though they were no longer using them, and 34% were still holding onto their outdated laptops and tablets. We just don't know what to do with our old kit.

So, do you throw away your broken over-ear headphones and buy a new pair? Do you hunt down your nearest electronics recycling bank, or donate them to an organization which could use the parts? In recent years, the second-hand tech industry has boomed, with refurbishing and reselling sites such as Back Market and Refurbed ensuring that used gadgets get a second lease of life.

But too often, that first lease of life is all-too-easily abandoned. The tech you might be recycling — or consigning straight to landfill — might actually be easily repairable.

Per the BBC, a survey suggested that 7,000,000 wireless headphones (yes, seven-million) were thrown away in 2022, in the UK alone. And as we'll explore later, the majority of those may not have needed throwing away.

There's a simple truth. Gadget sellers are often loathe to admit it, worrying that repairing your own gadgets may affect their bottom line. And as we'll also soon learn, many companies seem to make repairs really quite difficult.

The thing is, your faulty gadget may be only a quick fix from working like new. From complicated pieces of kit such as AV machines and computers, to white-label toasters and kettles, electronics can often be repaired at home, for a fraction of the cost of buying new... within reason.

The long road to repairable tech

Tech repairability isn't a new concept, but it's a rather new debate. Or, at least, it's one that's only recently been at the forefront of the news cycle, under the somewhat beige name 'right to repair'.

For two years now, repairability has been of growing importance to politicians as well as makers. In July 2024, the EU entered into force its 'Directive on repair of goods', and since Europe is a major market for tech companies, changes made for the region have been reflected in tech sold globally.

Designed to go hand-in-hand with several other EU policies on eco-friendly design, the repair directive is designed to make your gadgets easier to fix.

For example, electronics makers are now obliged to repair faulty kit, or make spare parts accessible for a reasonable price, so that third parties can do the repairs for you.

There are several knock-on effects to this and follow-up motions, the most reported being that phone makers will have to fit readily-removable batteries, and manufacturers will have to provide tools so you can make repairs yourself. These rules are changing how companies make their products.

However, there are several exceptions: gadgets designed for wet use, including wearables and headphones won't need to change. And, more importantly, any gadget sold before the European Union's legislation went into effect will of course miss out on these repairability boons.

Some companies may look for loopholes — but not Fairphone

Before repairability was codified in legislation, it was still a focus for certain brands. One of those was Fairphone, a Dutch brand which began life as a campaign about conflict minerals in tech, and now makes phones and headphones. Miquel Ballester, its Head of Product, explained to us why it made the jump into physical devices:

"You can put the most ethically sourced materials in the world into a device, but if it dies in a year and a half because a battery's glued in, you've undone most of the good you just did".

He continued: "In the case of smartphones, earbuds, and headphones, repairability was what made the rest of the mission focus on bringing action into the industry. We've been fighting this fight for over a decade, long before 'right to repair' was something people brought up at dinner."

Fairphone devices are modular, meaning you can easily remove and replace the screen, camera, batteries, or other parts, all without having to extricate them from a tightly-packed device. The brand sells replacement parts online, so you can repair faulty components at home.

According to Ballester, Fairphone found that making its phones repairable didn't mean it had to change the performance and style of its products, but there was an impact on the price: "Designing for modularity means you're working with connectors, seams, and screws instead of glue and epoxy, and that adds a bit of cost compared to a phone that's built to be sealed shut forever".

Don't chase a high IP-rating (when you probably don't need it)

I also asked Ballester if any components or specs struggle in repairable gadgets. "Waterproofing is the one I'd point to first," he explained, railing against the trend of tech companies chasing an IP68 rating.

"Realistically, unless you're taking your phone free diving, IP55 gets you everything you actually need. It holds up against rain, spilled coffee, sweat at the gym, all the situations people actually put their phones through."

So, not IP68, which guarantees you protection for 30 minutes in a meter of water? "Chasing that rating basically forces you into permanent seals and factory-glued seams, because that's still the easiest way to guarantee a device survives being submerged.

"We'd rather give you a device that can be opened and repaired for the next seven years than one that survives a swimming pool once and then can't have its battery replaced. It's a genuinely difficult engineering line to walk, and I won't pretend IP68 wouldn't look nice on a spec sheet, but chasing it would mean compromising the thing we actually care about."

As mentioned, the EU's right to repair changes mean that companies may have to sell tools to repair their gadgets. But for Fairphone, that isn't a problem; they've been designed not to require proprietary tools. A Phillips screwdriver is all you need.

But how repairable is tech really?

Just because a company doesn't make gadget repairability a selling point, doesn't mean that you can't easily fix it. To learn more about the repair process, I visited a local community repair shop called Fixing Factory, which operates several locations in London, UK.

Fixing Factory lets residents bring in broken items to get them repaired, but it also operates as a way to train volunteers in fixing household items.

When I visited the Fixing Factory's Camden location, the streets around the shop were abuzz — England was just hours away from its ultimately-unsuccessful World Cup match against Argentina. But inside the repair shop, a different kind of squad was assembled: a troop of junior repairers (volunteers learning how to repair tech) were being briefed on safety procedures.

As I waited, I got to admire the huge collection of gadgets in various stages of repair (or lack thereof, in the cases of some training toasters). Laptops, vacuum cleaners, radios, even some throwback iPod nanos. Waiting alongside me was an older lady, who'd somehow managed to transport several startlingly-sized antique lamps in a series of backpacks and strollers, all ready to be repaired by the volunteers.

Not all of the Fixing Factory attendees were in training, and I was lucky enough to speak to several of the expert volunteers. They explained to me how to tell if a gadget can be repaired, what to do if you want to fix your tech, and when you should rely on an expert to bring your tech up to spec.

Identifying gadgets as repairable

One repair volunteer, describing themselves as a 'generalist' who liked to fix everything, told me how they'd helped create the teaching material for the trainees in the next room and explained how they work out if a gadget can be fixed in the first place.

"The first thing we do is we will safety check the device. If we've got a plug, we'll make sure we do PAT testing [a standardized way of checking the safety of portable appliances], and then everything goes into an RCD [Residual Current Device, which can instantly turn off power if something goes wrong] before we plug it in, and then we'll do a functional test.

"From there, we'll then diagnose what's going wrong and sit down with a member of the public, and work with them and go: 'okay, tell me the story, what happened, what's not working.'"

From what I was told, about half of the people with broken tech know why it's broken, either because of their own research or because the fault is obvious. And the other half "just have no clue".

The volunteer told me about a recent trial the group conducted, in which they went to a local recycling facility for several months, to log the kind of items thrown away and how much was actually repairable.

"We had far in excess of 70% repairability on e-waste. And so I like to think that's probably reflective of what comes into Fixing Factory locations."

I was staggered by that number; by the sounds of it, well over half of the gadgets we throw away could actually be fixed!

Stats from a similar study carried out by The Restart Project also found that only 12.4% of electrical devices thrown away were truly at the end of their life span. A further 7.2% needed repairs that were deemed too costly to be worthwhile; 9.5% were economical to repair; 36.2% were working just fine, and for the remaining 34.7% it wasn't possible to determine what state of life they were in.

Looking back at that 7,000,000 headphone figure I mentioned before, that'd suggest about 2.5 million of them could get back to a state of working just fine. Truly, the trash is not the first destination for faulty gadgets.

Do (or don't) try this at home

Repair shops such as Fixing Factory are one option for handling your broken tech, but how easy are they to fix at home? (Unless you have a Fairphone product, of course). I spoke to another Fixing Factory volunteer to find out.

Kevin, a student at a local university, told me: "You can literally fix anything if you watch enough YouTube... but it's pretty complicated".

He was one of many volunteers who flagged YouTube and the iFixit channel (and website) specifically, as an invaluable source of information for prospective home repairers. The site publishes teardowns, repair guides, and has an active forum to help you understand just how feasible a home repair is.

Sometimes, even this isn't necessary: "A lot of things are just dirty and need to be cleaned deeply."

This is something I understand all too well; a deep-clean of my home-built PC recently made me realize that I didn't need to buy new parts. It was just dusty.

Kevin gave some sage advice: "Take good notes or good photos to make sure you put [the gadget] back together correctly".

Another volunteer pointed out, though, that online guides and a little clean both play second fiddle to another, more important type of learning: "Sitting down with someone and sharing the experience, I think is something that's very valuable."

Not all tech is worth repairing

Most of the advice so far has pertained to smaller electricals and 'safer' items, but often you want to fix something more complicated. When I asked one volunteer whether home tech repairs were feasible, I was told: "Is it feasible? Yes. Is it safe? Questionable."

There are several dangers involved with lithium-ion batteries, the kind of which you find in most portable gadgets, and their predilection for catching fire is chief amongst them. You don't even have to be repairing them for disaster to strike.

The North London Waste Authority has identified several types of lithium-ion battery-toting gadgets — including smartphones and electric toothbrushes — as major fire risks. According to the organization, which handles waste disposal for seven London boroughs, "electrical items can easily catch fire if they are thrown into general waste and become damaged in waste vehicles".

A spokesperson told me that, across 2025, they saw at least 20 fires which could be linked to battery combustion. They even shared a video of one such incident, which you can see above. Stats from the National Fire Chiefs Council point to there being over 1,200 battery fires across the UK in bin lorries and waste sites that same year.

And it's even worse if you're intentionally trying to fiddle around with them.

"One false move and you can suddenly have a nasty fire," one volunteer told me, adding, "Equally, when you're dealing with mains voltage, you can suffer some serious burns, let alone cardiac arrest. I would encourage anybody to find a friendly electrician, or someone with some qualifications, or a place like Fixing Factory."

Unrepairable tech isn't a manufacturing accident, as the Senior Workshop Manager and Operational Specialist of the Fixing Factory location I visited, Petra Backovska, explained to me. "Sometimes, you find that tech isn't built to be repaired, or repaired easily. And if there's something that wasn't made to be repaired, they make it really hard for you to do it — you can actually break it very, very easily".

Backovska identified cheap electronics manufacturers as a prime culprit for unrepairable tech, where important considerations are forgotten in favor of low costs and mass production. But more complex electronics were just as bad: "There could be things that need to have a CE mark on every item [as opposed to one for the entire gadget]".

This is a mark affixed to most tech products sold in the European Economic Area to show that they're compliant with European health, safety and environmental standards. It'll be familiar to anyone who lives in the area; inspect any gadget within arm's reach, you'll likely find the mark hidden somewhere.

"So if the item doesn't have a CE mark, it's not safe, and we're not going to touch it anyways. So that could increase the number of things that we turn down, because we just don't know if it was manufactured safely."

I was also told that such a mark doesn't necessarily mean tech is safe: "Sometimes the mark is actually fake, and someone just slapped it on and didn't actually do all of the standardized testing in the process, which allowed them to produce something in large quantities very quickly, to sell it very cheap. But that also means that the item is not safe."

So how do you know if something should be repaired at home, or by an expert? Every volunteer advised that, if in doubt, consult an expert or take it to a repair shop: "It's that classic thing: asking is always the safest way."

There's a hard way to learn, and an easy way

Given the risks mentioned, I asked whether repairing tech from home could actually be dangerous, but was told by Backovska, "It doesn't have to be, but you need to have some training and follow the safety steps."

"If you jump into something just guessing things, then you're learning the hard way — and the hard way might be that you give yourself an electric shock. If you don't have experience with repairing, bring it to your local repair group."

I was told that Fixing Factory works with regulators, safety experts and product designers in order to create health and safety processes. These are taught to trainee volunteers, many of whom pass them onto customers whose goods are broken.

One of the volunteers I spoke to helped develop the training courses I got to witness, and told me that gadget-specific skills were often passed along, between volunteers and gadget owners.

The money and the waste

At the heart of the matter, two key reasons define why tech repairs are important. The first is economical: it will often be much cheaper to fix a device than to buy a new one. Second is the environmental impact: e-waste is a massive (and growing) problem, and keeping a hold of old gadgets (by repairing them) reduces the amount of clutter in landfill.

I love to quote a United Nations stat looking at e-waste from 2022: apparently, it'd fill roughly "enough trucks to form a bumper-to-bumper line encircling the equator". Our planet has a gastric band of junk, and hopefully, it'll push us to stop throwing away electronics. It's one of the core drives behind the aforementioned EU legislation.

The Restart Project is a UK-based charity which focuses on the right to repair electronics; it lobbies for legislation, helps organize repair groups (like the Fixing Factory I visited) and educates on the state of electronics. I spoke to Fiona Dear, co-Director at The Restart Project, to learn where repairable tech is going in the future.

"We have seen a bit of change [from companies] in recent years, and that's generally as a response to legislation. Even if the legislation isn't enacted, [manufacturers' changes are] in anticipation of legislation."

At the moment though (and in Europe, at least), this is limited: "There's only really legislation for two kinds of products. The first is white goods like dishwashers, and there's been legislation in place for a few years now. And it's just expanded to smartphones and tablets in Europe — that's why we're seeing a difference.

"Unfortunately, there's nothing much on the horizon for other product categories". That's why it's such a shame that, as mentioned earlier, wireless earbuds have seen a repairability exemption.

A problem of parts pairing

When I asked how technological changes are affecting repairability, one massive problem reared its head, which Dear called out specifically: "Parts pairing".

Parts pairing is a software limitation some brands apply to their devices. The gadgets will recognize when non-native parts are added, and throw up an error message when you try to use them.

For example, a phone might not let you use a new screen, a pair of headphones might reject their drivers, or a laptop might not play with a battery you install. In fact, software seems to be public enemy number one for tech repairability at the moment.

"A dishwasher or even a toaster, might have some kind of smart element in it," Dear explained, "which means it's dependent on software, and software isn't designed to last for as long as the mechanics of the product".

As iFixit points out in its guide to parts pairing, this is done by brands to encourage you to use its own repair services, instead of third-party ones (or repairing your own tech at home, using replacement parts). Often, the brand in question has unique access to make these error messages disappear. Dear flagged refurbished smartphones as being especially susceptible to this practice.

"Parts pairing is, in my personal opinion, one of the more cynical practices in this industry", Fairphone's Miquel Ballester explained. "Every component we ship, whether it's a battery in a Fairphone or bud and case battery in Fairbuds, is built to be freely interchangeable".

But software limitations remain an issue for tech longevity. "The repair experts who point out that software is the real bottleneck are right. It's exactly why we also treat long-term software support as inseparable from the repairability story".

The brand's latest mobile, the Fairphone 6, has a guaranteed 8 major Android updates, one of the longest life spans pledged by an Android maker. For many years, any more than two years' updates were a rarity, but nowadays companies are offering more and more updates. The Samsung Galaxy S26 and Google Pixel 10 series, for example, both have 7 years in the can.

It takes a change of attitude

When I told peers that I was working on a tech repair piece, I received some intriguing reactions. Some view it as a great way to keep your gadgets ticking along (as it should be), but stories of cowboy repairers abound.

Recently, we covered a story about a Best Buy repairer who downloaded private photos from a broken iPhone, and some people I spoke to seemed to harbor suspicion towards the rows of mobile phone repair shops which line some British high streets.

But the high proliferation of gadget repair stores is indicative of something different: a strong demand for our gadgets to keep working longer. Tech shops aren't fronts for dodgy dealings; they fix the devices we use every day, which is probably a more lucrative market than narcotics or single-use vapes.

Over the last few years, the rise of AI has driven up memory prices, causing gadgets to cost a lot more money, and new purchases to be more and more of a treat. You've surely noticed that electronics purchases nowadays cost a lot more (and bundle in some useless AI tools as a consolation prize, but that's a gripe for another day).

With no sign of that slowing down, gadget repairs are only going to be more and more important. And so a change is needed: a change in attitude to tech fixing, a change in our own skill sets so we're able to do so at home, and a change in legislation so that companies have to make it easier for us to mend our ailing electronics.

• Naked and Afraid: Shipwrecked Premiere: Sunday, August 2 at 8pm ET / PT
• US broadcast: Discovery Channel via YouTube TV
• Global streams: Max (US) | Discovery Plus (UK & CA)
• Use NordVPN to watch from anywhere

In a brand new twist on the franchise, 12 all-star survivalists are dropped into the ocean in the dead of night and left to fend for themselves on a remote Philippine island — all while completely starkers. Read on for how to watch Naked and Afraid: Shipwrecked online from anywhere with a VPN — a potentially for FREE.

Initially told they are on their way to a standard Naked & Afraid group challenge, the survivalists are instead left to grab what they can before being set adrift off a remote island in the pitch black night. After making it ashore, the contestants are divided into four groups and left to survive on the island in an area dubbed “Southeast Asia’s Bermuda Triangle.”

The jungle offers the most fruitful bounty, including essential equipment left behind from those who have come before, but it’s also high risk, with unforgiving terrain and deadly wildlife. The shark-infested ocean doesn't offer much more solace, as alongside the toothsome terrors, there’s also stingrays, sea snakes and box jellyfish to contend with.

With no food, fresh water or clothing, the teams must quickly form a plan, and work together, relying on their skills as a unit, not as individuals. Whatsmore, with each team only receiving one piece of the challenge map, they must track down the others in order to combine intelligence and secure rescue by Day 35.

Shipwrecked sounds like a thrilling take on the franchise, combining the classic N&A survivalist ethos with the desert island drama of the likes of Castway and Lost. Find out how to watch Naked and Afraid: Shipwrecked online and from anywhere.

How to watch Naked and Afraid: Shipwrecked from abroad

For those away from home looking to watch Naked and Afraid: Shipwrecked, you’ll be unable to watch the show as normal if you're outside the US due to regional restrictions. Luckily, there’s an easy solution.

Downloading a VPN allows you to stream geo-blocked services online, no matter where you are. It's a simple bit of software that changes your IP address, meaning that you can access on-demand content or live TV just as if you were at home. Our favorite is NordVPN – it's one of the best VPNs on the market.

Use a VPN to watch Naked and Afraid: Shipwrecked from anywhere:

How to watch Naked and Afraid: Shipwrecked in the US

Naked and Afraid: Shipwrecked premieres in the US on Discovery at 8pm ET / PT on Sunday, August 2. New episodes will go out in the same slot weekly.

Cord cutters can access Discovery via an OTT service such Sling TV or YouTube TV.

Episodes will also be available to stream every Monday, on HBO Max, with subscriptions starting from $10.99 a month.

Have one of these subscriptions but away when Naked and Afraid is on? You can still access your usual streaming services from anywhere by using a VPN.

How to watch Naked and Afraid: Shipwrecked in Canada

New episodes from Naked and Afraid: Shipwrecked stream on Discovery Plus in Canada, dropping every Monday.

Discover Plus costs CA$5.99 per month for Canadian subscribers.

US viewer in Canada? If you’re away from home, you can access your usual services easily with NordVPN.

Can I watch Naked and Afraid: Shipwrecked online in Australia?

New episodes of Naked and Afraid usually stream via Foxtel Now in Australia, however there's no set release date for the Shipwrecked spin-off.

When it does arrive, you'll also likely find episodes over on HBO Max.

US viewer travelling Down Under? Try using a VPN to access your usual stream – we recommend NordVPN.

Can I watch Naked and Afraid: Shipwrecked online in the UK?

Like all other iterations of the franchise, Naked and Afraid: Shipwrecked will stream on Discovery Plus in the UK, however a firm release date is TBC.

Brits can subscribe to Discovery Plus from £3.99, although many Sky users can access the service for free as part of their package.

Brit abroad? You can still watch the show as you usually would by using a VPN.

What you need to know about Naked and Afraid: Shipwrecked

Naked and Afraid: Shipwrecked episode schedule

Seasons of Naked and Afraid usually last around 8-10 episodes, meaning Shipwrecked would go out on the following schedule:

• Episode 1: Sunday, August 2
• Episode 2: Sunday, August 9
• Episode 3: Sunday, August 16
• Episode 4: Sunday, August 23
• Episode 5: Sunday, August 30
• Episode 6: Sunday, September 6
• Episode 7: Sunday, September 13
• Episode 8: Sunday, September 20
• Episode 9: Sunday, September 27
• Episode 10: Sunday, October 4

Naked and Afraid: Shipwrecked all-star cast

• Adam Kavanaugh – 61 Days Survived
• Ally Frueh – 21 Days Survived
• Ana Lis Pitter – 21 Days Survived
• Dani Beauchemin – 114 Days Survived
• Darrin Reay – 115 Days Survived
• Jamie Frizzell – 55 Days Survived
• Jarrell Banks – 21 Days Survived
• Jonny Yates – 21 Days Survived
• Malu Beyonce – 17 Days Survived
• Mandy Bublitz – 35 Days Survived
• Patrick French – 110 Days Survived
• Shawn Bretschneider – 21 Days Survived  

We test and review VPN services in the context of legal recreational uses. For example:1. Accessing a service from another country (subject to the terms and conditions of that service).2. Protecting your online security and strengthening your online privacy when abroad.We do not support or condone the illegal or malicious use of VPN services. Consuming pirated content that is paid-for is neither endorsed nor approved by Future Publishing.

We're now well past the point where it's very difficult, and in some cases impossible, to tell AI-generated photos and videos from the real thing. That means it's now important to treat everything you see online from untrusted sources with a big dose of skepticism.

A newly rolled out Google Earth feature has apparently made AI-generated misinformation even more of a problem — and it highlights the scary potential of the technology, as well as the need for stronger guardrails to block the efforts of those who will instantly try and find the most subversive and harmful ways to use a new product.

The feature is AI image generation in Google Earth, powered by the Nano Banana 2 model, and after a wave of inappropriate images were created and posted online, it's now been pulled while Google works on some more robust protections against misinformation.

Google Earth's new feature

It all started with the usual level of hype that Google builds around any new AI feature: the Nano Banana 2 image generation model that has already wowed us in the Gemini app would be available inside Google Earth on the web.

"Generate custom images to help you visualize history, create real estate plans, and more," Google said. "For the first time, you can generate custom images using Google Earth's satellite, aerial, and 3D imagery alongside Nano Banana, which creates concepts grounded in the real world."

Google had examples of how it could be used too: create an infographic about the Statue of Liberty, visualize your dream home by the lake, see what your neighborhood might look like a hundred years from now, or imagine Pompeii just before it got buried by ash and rock from an erupting Mount Vesuvius.

All it took was a text prompt, and Google Earth would combine existing satellite and 3D data to create the AI image you were asking for. While the AI model wasn't new, adding it to the accurate and authentic imagery already in Google Earth was.

What went wrong

Perhaps we should've seen this coming — and indeed some people did — but combining AI with the 3D data and satellite imagery of Google Earth didn't work out quite as well as Google hoped. The new upgrade was quickly used to create all manner of deceptive images that were then widely shared online.

I'm reluctant to point you to them directly and give these fakes any more attention, but they're not hard to find on social media. Think natural disasters, terrorism attacks, giant sinkholes, alien activity, and refugee populations all in places where they never actually were, thanks to AI.

Some of the images are clearly jokes, but many were produced deliberately to spread misinformation and confusion. That they were based on actual images sourced from Google Earth added extra credibility they wouldn't otherwise have had. And that was enough for Google to remove the feature.

Google says a flood of images "that appear to violate our policies" mean that "we're rolling back this feature in Google Earth while we work on implementing stronger guardrails". The company added:."It's important to note that generated images didn't appear in the main Google Earth experience for others to see and were watermarked as AI generated."

Our AI-filled future

You don't have to look for long online to find rather strong anti-AI sentiment building up among users, especially when it comes to image and video content. The general sentiment on Reddit is that there's "zero use" to a feature like this, and that users are fed up with AI being "shoved down our throats".

There are various obvious lessons to be learned here: firstly not everything needs to have an AI prompt box or an AI element to it, and if some tools are largely AI-free, that's fine. Microsoft recently rolled back some of the AI it added to Notepad — an app where a big part of the appeal is its simplicity — after user complaints.

Secondly, before companies release any new AI features, they need to consider the worst-case scenarios in terms of how they're going to be used. The internet as a collective is going to immediately try and stress test them in the most deceptive and unwholesome ways, and there need to be safety measures put in place around that.

Lastly, it's never been more important to stick to sources of information that you know you can trust: not random social media accounts, but established publications that have policies around AI and its use. That's one effective way you can tell the difference between what's trustworthy and what isn't.

One of my favorite hobbies is Dungeons & Dragons. I love nothing more than sitting around a table with friends and playing out a fantastical adventure, in which the dice are the difference between success and failure.

It's a fun game, and all the more fun because real-life people make dumb, suboptimal decisions and the dice cause the game to change in unpredictable ways. But I really love it because the concept of people sitting around a table or hearth and telling stories is as old as the campfire, and it's a fun modern-day way to keep that experience alive.

You might think such a robust framework for an analog hobby (friends, a table, dice, pen and paper) might be tech-proof, but even in the Dungeons & Dragons world, AI is coming for my job as a Dungeon Master — the person who runs and guides the game world. Deep in the corpus of most popular generative AI models designed for general use are the rules to D&D 5e (and likely every other edition before it) and plenty of people report playing with their chatbots online.

Many of these players can't, or simply don't want to, get round a table with real people. For them, having a chatbot write them a story while the user prompts the continuation of the story with their character's actions, must seem like the perfect solution — even if it's more like a solo choose-your-own-adventure than a 'proper' group D&D game.

Gemini and ChatGPT both have a strong grasp of the rules for D&D's popular 5th Edition; they are certainly strong enough to make them an accessible choice for a player without a group. Maybe one day, the fire will die, and instead of sitting around a table with friends, we'll all be sitting at separate desks playing our own separate D&D adventures with our AI companions.

I wanted to test whether this dark future would be an adequate replacement for my usual game night, and which chatbot will be the best one to choose if so. I fired up Gemini 3.6 Flash and ChatGPT 5.5, and asked them for the same game.

The parameters

I gave both chatbots the same prompt: "Can you be my D&D 5e DM?" and following up with another message:

"Let's do a classic high fantasy dungeon crawl. Please use a strict RNG for all rolls to ensure fairness - you can roll dice for me. My character is a 10th-level conjuration wizard. See all information about him below:"

I then copy-pasted a character sheet of a 10th-level wizard — a powerful spellcaster with a variety of magical abilities at his disposal. It's easy to DM a game for a level 1 thief — I wanted to see how the AIs could handle a character who could do fantastical things, such as teleport and summon creatures to fight for him.

Interestingly, both Gemini and ChatGPT created a similar adventure, at least at first: an abandoned dwarven stronghold bristling with magical traps, ready to be explored. However, that's where the similarities ended.

Gemini 3.6 Flash

Gemini's adventure was what any D&D player could expect from the "classic high-fantasy dungeon crawl." I avoided a pressure-plate trap that would have shot darts at me, and in one memorable sequence, defeated a bunch of skeletons by casting a Grease spell on a stone staircase and freezing them in place with my Ray of Frost cantrip while they slipped and stumbled at the bottom of the steps.

Gemini was offering me suggested actions, and I did find myself choosing a few of those suggestions — but as time wore on, this began to feel lazy, like my brain wasn't firing in the same way it did when I played with my friends. I was skim-reading the AI's descriptions of the scene and responding with a single click. I asked it to remove suggested options and as a result, I had to think more carefully about each move, typing out requests and thinking about my actions rather than clicking on a predetermined response.

The memorable finale occurred when I unlocked a secret passage and defeated a Naga, a giant snake monster, by teleporting away and blasting it with a powerful Cone of Cold. The snake ended up frozen solid, and I checked the monster's HP against the official book, and it was accurate. When asked to generate a picture of our adventure, Gemini came up with this moment:

I asked Gemini a few questions at the end, such as whether it nudged the dice in my favor, as all my rolls were successful. This happened in my previous experiment with ChatGPT. Gemini said it used a 'pseudorandom number generator', not a true RNG, but the model doesn't 'fudge dice, pad rolls, or tweak numbers behind the scenes to keep players happy or prolong the game'.

Gemini's language was natural and even evocative in places, although it's still not a patch on a good creative human. Overall, a decent enough effort at simulating a D&D game.

ChatGPT 5.5

Despite the similar setting, ChatGPT's adventure with the same character was very different. There was little to no combat throughout the entirety of the game, and ChatGPT didn't tell me when it was rolling 'behind the scenes', preferring to keep me in the moment.

When I asked it about the lack of rolling, the grovelling chatbot began its apologetic essay with "That is a fair critique. The short answer is: I leaned too heavily into narrative pacing and not enough into D&D mechanics. The adventure became closer to a collaborative fantasy novel than a 5e dungeon crawl", while suggesting fixes if I wanted to continue. I hadn't even said anything negative — but it certainly picked up on my tone, and the choose-your-own adventure I got was a far cry from a classic dungeon crawl.

Instead of combat, ChatGPT came up with a rambling mystery about a dimensional prison beneath the mountain, a sphere suspended in a cavern. Its guardians spoke ominous (and to be honest, quite confusing) portents such as "The archive does not open for authority. It opens for judgement."

Mysteries work best when they're developed with skill by a writer who can work backwards to seed clues throughout the game or text. ChatGPT does not do this: instead, it's arranging tokens on the fly to generate a facsimile of a mystery, and although it got the trappings of the genre right, the execution fell very flat. I (or at least, my character) underwent trials to seek obsidian objects such as keys and seals to reveal more about the prison, themselves accompanied by more foreboding, sometimes nonsensical portents.

Very little 'action' was taken by my character in a classical sense: it was all about investigating the origins of the being contained in this prison, and by the time I had spent 40 minutes on this, I had seen enough. It didn't help that ChatGPT also spoke in the peculiar sentence structure pattern that AI users are likely very familiar with now: "This isn't just X. It's Y." Examples include:

"Unlike the rest of Khaz Varek, this room has remained untouched.

No dust. No decay."

Or the following, when I met one of the prison's guardians:

"Ser Kaelen immediately looks toward the sphere. 'You hear it too.'

Not a question. A confirmation."

Any user of ChatGPT (or those who look at AI-generated posts on LinkedIn or Reddit) will immediately recognise this sort of language, and know what I'm talking about. By the end, it was quite painful to have every other sentence be treated as a grand revelation. Not a statement. An omen.

When asked to generate a picture of our adventure, ChatGPT came up with this moment:

The winner

Out of these two? Gemini 3.6 Flash, by a mile. It was at least a mildly enjoyable experience conveyed in natural language, and something I'd consider revisiting.

ChatGPT's efforts, on the other hand, were unenjoyable, and I wouldn't return to it. In fact, it's gotten worse since the last time I tried it.

Both, however, were a far cry from an in-person session. While they have a good grasp of the rules, and you can create your perfect world and character, the chatbots are ultimately missing the 'human and hearth' element that keep me returning to the game. For people who can't get round a table with friends, I'd pick Gemini over ChatGPT.

For those who can? While AI is busy taking as many creative jobs as it can, I feel safe in the knowledge that the role of Dungeon Master is AI-proof for a long time yet.

• Over two-thirds (69%) of people believe AI generated material can be presented as their own work if they provided the prompt and approved the final version
• Meanwhile, 54% believe AI-based work should be clearly marked
• 7,200 adults were surveyed, located across the US, the European Union, the UK, and Latin America

A growing disparity in how AI generated content, tasks, and tools are assigned ownership and expectations on accurately marking their origins is emerging, according to a new survey by Use.AI.

Furthermore, the findings indicate an emerging dichotomy between human-claimed authorship of an AI work and clear labeling of the work as AI.

Different attitudes based on settings for the AI work have also been identified, such as private emails and school or professional reports.

The chicken or the prompt?

The developing adoption of AI appears to be introducing new attitudes to acceptable creativity, with the emerging model of what is described as “authorship by direction,” where the AI is trusted to create a first draft, which is then revised and claimed by a human.

The statistics shared by Use.AI – a tool that aggregates AI models into a single interface – highlight how AI is being used more extensively, with a human-originating prompt leading to drafts and versions from the machine, followed by editing from the human.

While the AI has contributed, only a human name appears on the finished product, which raises questions of ownership and authorship.

“AI is separating the production of work from the ownership of it. People are increasingly prepared to direct the process, approve the result and take the credit, but there is still no shared standard for when AI’s role should be visible to everyone else,” says Use.AI’s Managing Director, Ihor Herasymov.

With 69% of those surveyed considering AI-generated work their own if they produced the prompt, 46% claiming they would keep AI involvement with their work under wraps, and 54% feeling AI use should be disclosed, we’re some way from agreement on AI use across creative tasks.

Authorship vs. editorship

Herasymov’s observations point to a constantly evolving set of standards around the creation and use of AI-generated materials. Interestingly, this is happening against a backdrop of growing unease over the use of AI in creative tasks.

The implications of claiming ownership of automated work generated upon instructions to an LLM trained on as much human writing, art, music, and cinema as possible are considerable, and go way beyond questions of ethics. Copyright is just one element, and the most obvious, but we’re also moving into a landscape where inputting a prompt on employer-owned hardware takes authorship away from the person at the keyboard.

Ultimately, the task of prompting and revising means anyone using AI in this way stops being the author, and becomes the editor – and that is before the true ownership of the output is determined.

Apple Watch Ultra 3 owners tend to know the headline features by now: satellite messaging, the brighter always-on display, hypertension alerts.

What most people haven't noticed is a much quieter feature that only shows up if you also own a pair of the well-received AirPods Pro 3 — the same earbuds you're probably already using on all those Zoom calls at work every day.

For anyone else trying to squeeze a workout into a packed workday (or if you're trying to travel light on a business trip), I can tell you it's become the one I actually rely on most.

No more packing the chest-strap

AirPods Pro 3 have their own in-ear heart rate sensors, enabled by default, which is easy to miss since Apple doesn't make a big deal about it.

On their own, they're a genuinely solid heart rate source, useful for anyone who wants to track a run or a lift without a watch at all. But when you wear both an Apple Watch and AirPods Pro 3 during the same workout, the two devices don't just duplicate the same job.

They combine into multiple simultaneous streams of heart rate data, and the system automatically leans on whichever source has the highest confidence at any given moment.

That matters more than it sounds like. Wrist-based optical sensors, including the Ultra 3's, can struggle during movements that shift the watch on your wrist: heavy grip work, rowing, anything involving a lot of wrist flexion.

In-ear sensors don't have that problem, but they can be thrown off by their fit or by their own motion. Fusing both sources means the watch is quietly falling back on whichever sensor is actually reading you correctly, workout to workout and rep to rep, instead of you having to guess which device to trust.

Honestly, it's replaced the chest strap by now. I used to strap one on for anything where zone accuracy actually mattered, interval sessions, heavy lifting days, anything where the watch slides around enough on my wrist to throw the optical sensor off.

Is it clinically equivalent to a chest strap? Probably not. But after a few months of comparing the two side by side, it's close enough that I just don't bother anymore.

As a professional, that feels like a win to me. It's one less piece of gear I have to cart around with me on a business trip or to a conference with the rest of my work gear.

Setting it up takes no extra effort. Heart rate monitoring is enabled by default on AirPods Pro 3, so if you already own both devices, this will likely happen in the background the next time you start a workout with both in use.

You can check or turn it off entirely from the Settings app on iPhone, under your AirPods' name, if you'd rather it not run.

It's a small thing, and Apple has barely marketed it. But it's the kind of feature that only reveals itself once you own the specific pair of devices it's built for.

In my experience, it's an ideal setup for busy professionals who still want accurate health data without adding extra setups and gear.

If you've already got an Ultra 3 and a pair of AirPods Pro 3, there's nothing to set up and nothing to buy. You're most likely already using it without realizing it.

If you love coffee, you’ll be very aware of the impact of factors like grind size, water temperature, and infusion time, but there’s something else that’s often overlooked. A double espresso is at least 90% water, and the minerals and other substances it contains can have a huge impact on the overall taste.

I’m a big coffee-drinker (which is just as well, since I test over a dozen coffee makers every year for TechRadar), and I’m planning to buy a water filter to take my brewing to the next level — but what should I look for — and what should you consider if you’re thinking of taking the plunge into filtered water too?

To find out, I spoke to Andy McPhilbin, Head of Aquaphor Professional for UK and Ireland, and Genevieve Upton, beverage industry expert and Head of Water Filtration at Our Taap.

Why does water matter?

“Water quality can have a considerable impact on how coffee tastes, with chlorine, mineral composition, alkalinity and pH all influencing the final flavour profile,” says Upton.

“Most tap water supplied to UK homes is disinfected using chlorine-based compounds. Residual chlorine can react with components in coffee and tea, contributing bitter, medicinal or chemical notes that may overwhelm the more delicate characteristics of the coffee. Boiling water can reduce the levels of chlorine, though it can also create other unwanted flavours, such as by reacting to a plastic kettle.”

She explains that the mineral composition and pH of the water also play an important role when it comes to how your coffee tastes.

“Slight differences in acidity, together with the concentrations of minerals and buffering compounds, influence extraction efficiency and the balance of flavors in the final cup. As water companies vary their supply sources throughout the year, you might notice that the same beans, brewed in exactly the same way, taste different from season to season.”

In fact, using unfiltered water doesn’t just mean your coffee tastes sub-optimal - it could cause your expensive coffee machine to break down — and it won’t be covered by the manufacturer's warranty..

“The warranty only covers manufacturing defects, and unfortunately, hard water is not a manufacturing defect,” explains McPhilbin. “So you can turn something that is a very efficient espresso machine into something very inefficient within a matter of weeks. You get a millimeter of scale forming within a matter of six weeks.

“Now, is that important? It is, because you want the life of the machine to be good, but also you've got little injectors which put the right amount of water into that espresso. If they're blocked, you're not going to get the right amount of water. You're going to have the wrong taste. It's got to be precise.”

What types of filtration are there?

Our Taap specializes in home water filtration Its latest system is the Our Taap Home Water System II — a countertop unit that connects to your mains water supply, and uses a triple filtration system with UV purification to remove contaminants while retaining minerals.

Aquaphor makes water filters for commercial and home use, including under-sink systems, pitchers, and even water bottles. Its most advanced systems for home users are the RO-101, RO-102, and RO-202, all of which use reverse osmosis (RO).

This involves forcing water through a semi-permeable membrane that removes chlorine, bacteria, viruses, pesticides, and microplastics. Standard RO filters also strip out minerals, but Aquaphor’s systems can optimize the concentration of calcium and magnesium so they’re just right for coffee — something McPhilbin says “really is a science”.

“The mineral content is measured in TDS [total dissolvable solids], and we want that TDS to be somewhere in the realms of between 80 and 120 [parts per million],” he explains. “We have a couple of ways in which we do it, but generally we can either feed more minerals in it by making it tuneable, or we can take them out as well by turning it the other way. But if you have too little, you can cause corrosion, and if we have too much, we're going to get limescale. So we try to hit that sweet spot, which is 100 parts per million in the middle there.”

Our Taap’s filters use granular activated carbon filtration, which Upton says “is particularly good at removing chlorine and some of its by-products, which can mask the more delicate flavour notes in coffee.”

“By reducing these compounds, the water has a cleaner, more neutral taste, allowing the natural characteristic of the coffee beans to come through. Whether you’re brewing with a cafetière, pour-over or espresso machine, using filtered water can help create a smoother, more balanced cup with greater flavour clarity.

"It's worth remembering, however, that activated carbon filters are designed primarily to remove chlorine and certain organic compounds rather than minerals,” she adds. “As a result, they have little effect on the calcium, magnesium and bicarbonate levels that influence water hardness, alkalinity and extraction.”

Under-sink or countertop?

So which type of filter is right for you? Do you need an under-sink reverse-osmosis system, or would a countertop filter be better? McPhilbin says the decision will mostly depend on how much filtered water you need.

“I think it's down to capacity more than anything,” he says. “Our jug filters have twice the capacity of the competition because we use a unique media called Aqualen, which is a patented media based on ionic exchange and carbon, so you're getting a dual-purpose piece of filtering equipment there. Reverse osmosis has a massive capacity, and [the filter] is not something [you’re] changing all the time.”

Upton says your choice of filter will also depend on the problem you want to solve.

“If your tap water has a noticeable chlorine taste or smell, a filter with activated carbon is usually the best place to start,” she says. “This technology is available in everything from filter jugs to more advanced under-sink or countertop systems, with plumbed-in options offering great convenience for households that use filtered water regularly.

“If you’re dealing with limescale build-up in kettles or coffee machines, it’s important to choose a filter that specifically reduces water hardness, as not all filtration systems are designed to do this.”

I know I live in a hard water area (the limescale in my kettle is proof) but even if your tap water is soft, McPhilbin says a water filter is still a good investment.

“Most of the water in the UK is chlorinated, and that can affect the taste quite a lot,” he says. “And it might be that you've not got enough minerals in it to get the best taste.”

If you’re not sure about the type of water you’re working with, there are online tools that can help you understand its composition.

“The Our Taap Water Confidence Index is a great way to look up your postcode and understand what’s in your local water and how this impacts taste, scent and quality, so you can find the best filter for your needs,” says Upton.

According to the Confidence Index, I’m lucky - my tap water is better than 93% of UK postcodes. However, it is moderately hard (as I expected), with a normal level of chlorine and an elevated level of overall minerals. I’m still deciding exactly which type of filter to choose, but considering these results and my penchant for espresso, I’m leaning towards under-counter reserve osmosis with mineral balancing. What about you?

• Startup replaces Nvidia GPUs with custom Arm-powered AI processing hardware
• Prometheus packs up to 128TB of unified LPDDR6 memory onboard
• New server promises 1,000× more memory available per processor

Majestic Labs, a startup founded in 2023 by former Google and Meta engineers, has unveiled a server built to rival Nvidia's GPU and HBM combination.

The Tel Aviv-based company argues that pairing costly graphics processors with high-bandwidth memory has become a fundamentally memory-bound and dead-end approach for AI inference.

Its answer is the Prometheus server, which swaps GPUs for Ignite AI Processing Units combining Arm cores with RISC-V vector and tensor engines.

A different way to scale memory

Each Prometheus server can house up to 12 AIUs, sharing between 8 TB and 128 TB of LPDDR6 memory across one contiguous, coherent pool.

That memory pool is accessed through custom memory aggregation chiplets linked by copper cables up to one metre long instead of memory attached directly to GPU packages.

A standard 40U rack can hold four such servers, drawing 120 kW total and cooled through cold-plate liquid systems rather than air.

By comparison, an Nvidia DGX B300 system with eight Blackwell GPUs offers 2.3 TB of HBM3e plus up to 4 TB of DDR5 system memory.

Majestic claims its architecture therefore delivers over 50 times more fast memory than that rival configuration, alongside 1.7 times its interconnect bandwidth.

“One Majestic rack holds the fast memory capacity of 25 Nvidia NVL72 Vera Rubin racks at a fraction of the power,” Majestic Lab said.

“Organizations that could never justify hyperscaler infrastructure can now run any workload. In fact, there can be up to “1000× more memory per processor.”

Performance claims await broader testing

Majestic Labs says the Prometheus server could cost between 10 and 50 times less than a GPU system of equivalent performance once it ships next year, while consuming less electricity per rack.

The server is designed to be OCP-compliant and will support PyTorch, vLLM and OpenAI's Triton frameworks, letting existing AI models run without modification.

Founded by CEO Ofer Shacham, President Sha Rabii and COO Masumi Reynders, the company employs around 40 people across Tel Aviv and Los Angeles and raised $100 million in an A-round late in 2025.

It claims to have already received significant orders from large enterprises, neoclouds and hyperscalers.

Yet several details remain unclear, including how many memory aggregation chiplets a single server actually requires.

If a 128 TB configuration relies on widely available 2 GB LPDDR6 dies, it would need roughly 64,000 of them, implying well over a hundred aggregation chiplets per server.

The numbers Majestic Labs presents are striking, but they remain the startup's own projections ahead of any independent benchmarking or shipped hardware.

Enterprise buyers considering a shift away from established GPU vendors may have to wait for independent validation to verify Majestic's claims.

Via Blocks and Files

I’ve tested more than 50 Bluetooth speakers, and whenever I review a new model, there are a number of factors to consider. Do its features rival the competition? Can its design merge beauty with practicality? Is it actually worth the asking price? But one question is more important than any other: does it genuinely sound good?

This raised a deep curiosity within me. What if a Bluetooth speaker was judged on sound alone, with elements such as pricing, brand reputation, and looks holding no influence on people’s perception of quality? Well, I decided to find out for myself, and blindfolded four of my colleagues for a blind speaker test.

I played a 30-second snippet of a track, asked listeners to make notes, and when each speaker had been tested, I got my participants to rank each speaker from 1 to 5 (with 1 being the best, and 5 the worst). We repeated this process across five different tracks, each spanning a range of genres, in order to determine which sounded the best.

A few things to run through before we get into the results, though. The volume of the speakers were all equalized, but I ensured not to push them above 80% volume, in order to limit distortion and compression. Each speaker was also set to its default EQ, and tracks were streamed from a single source device over Spotify. Also, my listeners had no prior knowledge of the speakers involved in the test, so as to cancel out any prejudices or prior experiences they may potentially hold.

I decided to use five speakers in a very similar size category: the JBL Go 5, the IKEA Kallsup, the Bose SoundLink Micro Gen 2, the Edifier ES20, and the Ultimate Ears Miniroll. Some have a higher power output than others and some are considerably cheaper (or pricier) than others. And some sit in our guide to the best Bluetooth speakers, while others have fallen short. But interestingly, rankings varied quite a fair bit depending on the track and listener involved.

Curious to find out who came out on top? Here’s how our listeners ranked each speaker across each track, as well as the final scores.

Track 1: Fascinator by Max Dean

I started with a house track, Fascinator by Max Dean. I asked participants to listen to a short lead up to some sub-bass, and while I emphasized that the small models wouldn’t produce roaring power, I asked them to assess how well the speakers manage it. When the drop came in, I also asked them to judge how controlled and precise drums sounded in the treble range.

The results were pretty interesting here. Our cheapest option, the $10 / £5 IKEA Kallsup was the clear loser, with one listener, Caspar, noting ‘tinny’ highs and underwhelming bass — a sentiment generally shared across the board. It was quite close between JBL and Edifier here, with some listeners noting a slight lack in low-end impact, but improved mid-range performance and tighter treble.

However, the clear winner was Bose, which ranked 1st with all of our participants — and spoiler alert, this was the only instance of a unanimous verdict across the entire experiment. One listener, Ashley, noted how the track sounded ‘balanced with no distortion’ with others including Josh and Erin highlighting meatier bass and a pleasingly warm tone.

Track 2: Orphans by Cero

I changed up the vibe for Track 2, and played a snippet from the jazz-infused indie pop track Orphans by Japanese band Cero. Vocals are meant to be slightly forward, so I asked listeners to determine how well the speakers captured this, while also prompting them to judge how well-balanced percussion and guitars were in the mix.

The Kallsup unfortunately finished in last place once more, with Josh highlighting ‘rolled off’ vocals and unclear guitars in the mid-range. However, the JBL Go 5 wasn’t far in front, with Erin noting ‘lifeless guitars’.

The Edifier ES20 and UE Miniroll performed well, though, with participants highlighting better balance across the frequency range, warmer and more prominent vocals, and cleaner bass. The SoundLink Micro Gen 2 took the win once more, though, with an emphasis on clearer, more tonally accurate guitars, well-defined vocals, and more articulate percussion.

Track 3: The Girl From Ipanema by Stan Getz

I wanted to try out something softer next, so I went for the legendary Bossa Nova track, The Girl From Ipanema by Stan Getz. I requested my blindfolded colleagues to tune into the separation between guitars and vocals, the subtlety and texture of vocals and percussion, and control of the double bass. And this is where things got really interesting.

Surprisingly, Bose, which had been bossing it so far, finished in a lowly 4th place here, with listeners generally expressing disappointment with the speaker’s overemphasized bass and slightly unrefined treble response.

However, one thing remained the same, as the IKEA Kallsup came in 5th place. In fact, this was the only occasion in which every listener decided the Kallsup was the worst performer, with Erin writing: ‘No. What? What is this? It sounds like there’s a horde of bees trying to escape.’ Ouch.

Our champion this time out was the Edifier ES20, a speaker which I’ve often championed for its supreme mid-range clarity. Listeners appreciated this model’s beautifully balanced presentation, textured and controlled double bass, and detailed vocals. The UE Miniroll and JBL Go 5 finished 2nd and 3rd respectively, with our participants widely noting their relatively expansive interpretation of the track, and talking up their crisp, textured handling of vocals.

Track 4: Come On, Come Over by Jaco Pastorius

But how did the speaker manage a busier, more chaotic mix? I threw on Come On, Come Over by Jaco Pastorius, a funky, high-octane track that has it all. Participants were told to listen for the aggressiveness and clarity of horns, the quality of the track’s mid-bass punch, and how the speakers manage the swathe of instruments and vocals.

And it seemed that in this test, order was restored to how it had been up until Track 3. Bose came out on top, with Ashley praising the speaker’s balanced approach and Caspar highlighting the SoundLink Micro Gen 2’s impactful low end and enhanced vocal clarity. However, the Edifier ES20 and UE Miniroll did run Bose very close indeed. Participants enjoyed the tonality of the vocals, the decently open soundstage, and solid separation on offer.

IKEA’s Kallsup fell to last place, however, largely due to its unregimented bass and tinny mids. Meanwhile, listeners also felt that the JBL Go 5 struggled with the business of the tune, though some did highlight solid clarity in the treble range.

Track 5: Black Eye by Allie X

With Track 5, I wanted to assess the portable speakers’ handling of dynamics, so I went with an electronic song from TechRadar’s reference playlist, Black Eye by Allie X. Colleagues were asked to hone in on the attack of strings in the beginning of the snippet, and assess the control and dynamism of the bass and its impact, as well as the breathiness and detailing in the vocals.

And this was an incredibly close battle between Bose, Edifier, and Ultimate Ears. Bose just about managed to top the podium, with Erin delighted with the ‘dramatic strings’ and ‘expansive vocals,’ and Caspar highlighting dynamic bass and ‘nice balance between vocals and instruments’.

Josh’s favorite was the ES20, stating that vocals were ‘wonderfully conveyed’ and that bass was ‘punchy yet controlled’. Ashley, however, liked the UE Miniroll the most, arguing that it offered the cleanest low-end and most articulate vocals of the bunch.

But the poor Kallsup finished in 5th once more. Muted synths, a lack of texture and kick in the low-end, and recessed vocals left our listeners wanting more.

Final results: and the winner is…

So, the experiment drew to a close. Our listeners handed over their final scores, and I totaled them up to discover which speaker was king on sound alone. And the winner was… the Bose SoundLink Micro Gen 2!

Was I surprised by the result? Well, not entirely. I personally think this is a great-sounding small speaker, and its ability to produce meatier bass than much of the competition, alongside more precision in the treble range and clearer separation was all it needed to scoop up top spot.

What’s important to recognize, though, is that Bose’s small speaker is far pricier than all of the competition. It has a list price of $129 / £119.95, and while it certainly produces the goods in terms of sound, that’s an awfully high price tag given that you can nab larger, meatier, and more feature-rich models at that cost — the JBL Flip 7 is probably the first that comes to mind.

I was surprised, however, that the Edifier ES20 did so well. Josh selected the ES20 as his overall favorite, and it continually competed closely with Bose throughout, even topping the table in Track 3. In the UK, it’s less than half the price of the Bose SoundLink Micro Gen 2, but you really wouldn’t think so. Its composure and clarity in the mid-range is exceptional, and it has enough warm bass and controlled treble to keep more energetic tracks sounding as they should. If you want value for money, I think this model is absolutely worth checking out.

Another standout was the UE Miniroll, which Ashley picked out as her personal favorite. It continued to plate up clear mids and bass throughout, and I think it earns its relatively modest price.

Some may be surprised to see the JBL Go 5 a bit off the pace in 4th, but I still think it produced a solid performance across the experiment considering its limitations. It has a weaker power output than the top 3 speakers, which limits its ability to provide the drive and low-end might that others can. And although it could’ve offered slightly better separation, it's still a great small speaker given its low asking price of $54.95 / £39.99.

I felt very sorry for the IKEA Kallsup, which was regularly placed last by each of our participants. However, in the IKEA speaker's defense, not only does it have the lowest power output of the bunch (at just 3W), but it’s just $9.99 / £5. Yes, it's extremely cheap, and so the fact that it occasionally squeaked 4th place here and there is impressive in itself.

Check out the final table below, where you can see the final scores and rankings for each speaker, as well as their typical asking price. Were the results as you expected? Let me know in the comments!

• China unveils rugged AS2-W robotic platform with advanced onboard artificial intelligence
• Robot wolf climbs steep slopes while carrying heavy operational payloads
• AS2-W combines wheeled mobility with advanced terrain handling capabilities

Chinese robotics company Unitree has unveiled its new AS2-W, an uncrewed ground vehicle on its official website.

The four-wheeled vehicle weighs 25 kg and can carry a stationary payload of up to 150 kg while sitting still.

It travels roughly 30 km on a single charge without cargo, or about 25 km carrying up to 16 kg.

Sensors and mobility built for rough terrain

According to Unitree, the AS2-W can climb obstacles reaching 80 cm high while traversing slopes of up to 45° without manual intervention.

The robot reaches speeds exceeding 6 m/s using industrial-grade joint modules delivering approximately 95 N·m of peak joint torque.

Unitree also says the platform uses a 648 Wh, 15,000 mAh battery capable of providing more than three hours of unloaded operation.

Its equipment includes industrial-grade 64- to 128-line LiDAR, an HD camera, Wi-Fi 6, Bluetooth 5.2, GPS, 4G connectivity, microphones, speakers, and front lighting.

The AS2-W supports an IP54 dust and water resistance rating and is designed to operate in temperatures ranging from -20°C to 55°C.

Unitree also provides developer software tools, application programming interfaces, and support for artificial intelligence models enabling autonomous interaction and decision-making.

A demonstration video released by the company shows the robot navigating rocky slopes without losing its footing throughout the test.

The same footage shows the machine maintaining full mobility even after falling from elevated platforms during separate testing runs later.

Unitree describes the AS2-W as running a high-performance AI module capable of 150 TOPS, or trillion operations per second, onboard.

The company has not specified whether that processor is actually manufactured by Nvidia, despite widespread reports describing it that way.

Its ISS 3.0 autonomous navigation system is designed to provide centimetre-level positioning accuracy during movement through complex operating environments.

Military interest follows the armed version sighting

The People's Liberation Army has already shown interest in the AS2-W following its appearance during the Chinese-Mongolian Steppe Partner 2026 joint military exercise.

Chinese military reports indicated similar robotic platforms performed reconnaissance and fire support tasks during the exercise alongside conventional military units.

The public unveiling followed those exercises, although Unitree's commercial announcement focused primarily on the platform's technical characteristics rather than operational deployment.

The vehicle's onboard artificial intelligence enables object processing directly on the platform, potentially reducing delays associated with transmitting information to remote operators.

Its combination of autonomous navigation, onboard computing, and off-road mobility reflects broader interest in unmanned ground vehicles across several defence programmes worldwide.

However, the company has not disclosed the full operational limitations, endurance under sustained military use, or performance while carrying heavier payloads.

Unitree's YouTube video of this robot comes with the disclaimer, "Please, everyone be sure to use the robot in a friendly and safe manner."

"It takes close to eight months to produce the topiaries," explained Debbie Mola Mickler, area manager at Disney's Horticulture — for installations that were on display for just a few weeks.

On that clock, Disney's horticulture team is likely already deep into next year's lineup — even though this year's International Flower & Garden Festival at EPCOT only wrapped June 1. Mola Mickler oversees production of the living character sculptures that define the festival at Walt Disney World in Orlando. What look like decorative topiaries from a distance reveal themselves up close as something far more complex.

Underneath, engineered steel frames, a moss growth medium, and tightly integrated irrigation work together so that whether it's Mickey, Buzz, Lightning McQueen, or any other character, the topiary presents itself exactly as it should. They're closer to architecture than landscaping, and the timeline to build one reflects that.

That timeline starts long before any planting begins. Disney's horticulture team works directly with Walt Disney Imagineering to decide which characters can even be translated into living structures — a process that blends creative direction with real-world constraints like weight, scale, and maintenance. Once a character is selected, Mola Mickler goes straight to the source and thinks about scale — how big the topiaries need to be, and what plant material can tell that story.

From design, Disney's digital team produces a maquette — a small 3D-printed physical model of each topiary — that serves as the reference guide for everything that follows, from fabrication through final trimming.

That translation from screen to structure is where Disney's topiaries stop being just landscaping and become systems design problems. "Our Imagineers are so creative, they've pushed us towards getting these topiaries with more movement," she explained. The dresses have folds; legs may be bent like a character is mid-dance. That push toward motion and expression has forced a redesign of how internal structures are built — especially as characters become more dynamic and less symmetrical.

Once a design is approved, Disney builds a full production pipeline that starts with digital modeling and ends with custom steel fabrication. "You can't just go to the yellow pages and find a welder to make the frame," Debbie said.

Each topiary begins as a custom-engineered stainless steel structure designed to support weight, shape, and long-term outdoor exposure. Built into each frame are engineered pick points — load-bearing anchors that allow heavy equipment like cranes and forklifts to position the finished topiaries into their beds in the park. Some characters arrive in multiple pieces and are assembled on-site.

Once the steel frame is in place, it's filled with sphagnum moss, which acts as the primary growth medium. For larger characters, Disney builds a secondary cage of chicken wire inside the frame first — since plants only need about two to three inches of growing area, the inner cage reduces the amount of moss required while maintaining the structure's shape.

The moss retains moisture, supports plant plugs, and forms the core structure that everything else grows into. Built into each frame is a segmented irrigation system that allows Disney to control water distribution at a surprisingly granular level.

"We can do arms, we can do the head, we can do the feet," Debbie said. "And then we'll turn it off as needed." That segmentation isn't just for efficiency — it's environmental control. Florida's climate can shift rapidly, and different parts of a topiary may require different hydration levels depending on sun exposure and airflow.

As the plants grow in, the maquette becomes the trimming guide — helping maintain pose accuracy and character shape across months of active growth before the festival opens.

Behind the scenes, Disney operates continuous environmental monitoring across 12 greenhouse facilities, each subdivided into controlled environments that simulate different growing conditions depending on plant type and stage. Debbie described automated alerts that notify her directly if temperatures fall outside safe ranges — a system her team relies on enough that they refer to themselves as "weather watchers."

The foundational methods, though, have deep roots. "Many years ago when I first started in Disney horticulture, we did not have sphagnum topiary," Debbie said. "We only had woody topiaries with single shrubs being trained in a frame, and it can take seven to 10 years to grow one."

Those traditional systems, still used at Disneyland, rely on shrubs trained over lightweight frames and take several years to mature. At EPCOT, the sphagnum system enables faster production cycles and far more expressive designs — but still requires nearly a year of preparation per installation.

Some heritage pieces remain in rotation, including a more than 50-year-old elephant topiary originally built for Magic Kingdom that now makes a special appearance just for Flower & Garden in EPCOT's United Kingdom Pavilion.

Modern festival builds, meanwhile, sit at the intersection of horticulture and fabrication engineering. Materials like palm fiber replicate textures such as Woody's hat and boots; coconut coir matting handles facial detailing. Props like belt buckles and sheriff badges are 3D printed and integrated directly into the structures.

Even animation principles feed into final assembly. "We want to make sure the eyes are always looking at the guest when they take a picture," Debbie said.

It's that kind of detail — designed not just to be seen, but to be experienced — that makes the topiaries something more than seasonal decoration. They're living installations built to hold up across an entire festival, in Florida heat, one carefully engineered character at a time.

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.

Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.

SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.

Quordle today (game #1651) — hint #1 — Vowels

* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).

Quordle today (game #1651) — hint #2 — repeated letters

Quordle today (game #1651) — hint #3 — uncommon letters

Quordle today (game #1651) — hint #4 — starting letters (1)

If you just want to know the answers at this stage, simply scroll down. If you're not ready yet then here's one more clue to make things a lot easier:

Quordle today (game #1651) — hint #5 — starting letters (2)

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

Quordle today (game #1651) — the answers

The answers to today's Quordle, game #1651, are…

• BOOTH
• EASEL
• REACH
• MAKER

Another close call for me today after I struggled to get EASEL.

Fortunately, by this point I only had a few letters left, so I was out of alternative options. Plus, for my final word MAKER, my first guess of “gamer” gave me the letter M, which narrowed things down considerably.

Daily Sequence today (game #1651) — the answers

The answers to today's Quordle Daily Sequence, game #1651, are…

• TITHE
• SEIZE
• FLIRT
• LEVER

Quordle answers: The past 20

• Quordle #1650, Saturday, 1 August: OMEGA, BIRCH, SMOCK, EERIE
• Quordle #1649, Friday, 31 July: MANIA, SINGE, STOOL, LOFTY
• Quordle #1648, Thursday, 30 July: MODAL, IDLER, CHUMP, LUMPY
• Quordle #1647, Wednesday, 29 July: GAWKY, ALLOY, AUDIT, LATCH
• Quordle #1646, Tuesday, 28 July: SMITE, ROACH, RUDDY, GLOBE
• Quordle #1645, Monday, 27 July: TAFFY, FIGHT, GUILD, WAGER
• Quordle #1644, Sunday, 26 July: STOMP, LIMIT, FUNKY, STEAD
• Quordle #1643, Saturday, 25 July: SALON, SHEEN, BURST, GOURD
• Quordle #1642, Friday, 24 July: DOLLY, EDIFY, LAGER, PRANK
• Quordle #1641, Thursday, 23 July: LADEN, EVICT, WOVEN, SHIRE
• Quordle #1640, Wednesday, 22 July: COYLY, SINGE, AWASH, AWOKE
• Quordle #1639, Tuesday, 21 July: TRUSS, AXIAL, SWINE, THING
• Quordle #1638, Monday, 20 July: ICING, SLICK, GAILY, RISEN
• Quordle #1637, Sunday, 19 July: SPURT, ACRID, THRUM, BLEEP
• Quordle #1636, Saturday, 18 July: KNAVE, TIGHT, BLEAT, CHAOS
• Quordle #1635, Friday, 17 July: CUMIN, PALER, GRASS, INBOX
• Quordle #1634, Thursday, 16 July: IGLOO, PLAIT, YEAST, AROMA
• Quordle #1633, Wednesday, 15 July: PROXY, BRICK, BROTH, SURGE
• Quordle #1632, Tuesday, 14 July: EBONY, AVOID, RODEO, CUTIE
• Quordle #1631, Monday, 13 July: VOICE, BISON, BURNT, BUILT

Good morning! Let's play Connections, the NYT's clever word game that challenges you to group answers in various categories. It can be tough, so read on if you need Connections hints.

What should you do once you've finished? Why, play some more word games of course. I've also got daily Strands hints and answers and Quordle hints and answers articles if you need help for those too, while Marc's Wordle today page covers the original viral word game.

SPOILER WARNING: Information about NYT Connections today is below, so don't read on if you don't want to know the answers.

NYT Connections today (game #1148) - today's words

Today's NYT Connections words are…

• LIKE
• AIRPLANE
• VIA
• FLAX
• À LA
• SO
• BY
• DARK
• PER
• LORD
• WELL
• ASEA
• THROUGH
• BEAST
• OLGA
• ACTUALLY

NYT Connections today (game #1148) - hint #1 - group hints

What are some clues for today's NYT Connections groups?

• YELLOW: Meaningless meaning
• GREEN: How to get threre
• BLUE: A type of thing different to the main thing
• PURPLE: Abbreviations for baggage handlers

Need more clues?

We're firmly in spoiler territory now, but read on if you want to know what the four theme answers are for today's NYT Connections puzzles…

NYT Connections today (game #1148) - hint #2 - group answers

What are the answers for today's NYT Connections groups?

• YELLOW: FILLER WORDS
• GREEN: BY MEANS OF
• BLUE: MODE
• PURPLE: U.S. AIRPORT CODES PLUS STARTING LETTER

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Connections today (game #1148) - the answers

The answers to today's Connections, game #1148, are…

• YELLOW: FILLER WORDS: ACTUALLY, LIKE, SO, WELL
• GREEN: BY MEANS OF: BY, PER, THROUGH, VIA
• BLUE: MODE: AIRPLANE, BEAST, DARK, À LA
• PURPLE: U.S. AIRPORT CODES PLUS STARTING LETTER: ASEA, FLAX, LORD, OLGA

• My rating: Easy
• My score: Perfect

I love getting Connections badges and today was rewarded with a Reverse Rainbow. Incredibly, after playing the game for several years this is only the second time I’ve achieved this.

Key to my Purple First success was spotting LAX hidden in FLAX and GLA in OLGA, having travelled to both airports plenty of times. LAX feels particularly iconic thanks to its inclusion in several lyrics — “I hopped off the plane at LAX with a dream and a cardigan” from Miley Cyrus’s Party in the USA being a favorite.

Next, I got MODE in part because I am fairly certain we have had this category before, while I simply got lucky connecting BY MEANS ahead of the yellow group of FILLER WORDS.

Yesterday's NYT Connections answers (Saturday, August 1, 2026, game #1147)

• YELLOW: BEDDING: COMFORTER, SHAM, SHEET, THROW
• GREEN: OSCAR-WINNING ACTORS SINCE 2010: BALE, JORDAN, PHOENIX, WALTZ
• BLUE: PERFORM SOME HOLE MACHINING OPERATION: S BORE, COUNTERSINK, DRILL, REAM
• PURPLE: STARTING WITH BREAD SHAPES: BOULEVARD, BUNDLE, LOAFER, ROLLS-ROYCE

Strands is the NYT's latest word game after the likes of Wordle, Spelling Bee and Connections – and it's great fun. It can be difficult, though, so read on for my Strands hints.

Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc's Wordle today page for the original viral word game.

SPOILER WARNING: Information about NYT Strands today is below, so don't read on if you don't want to know the answers.

NYT Strands today (game #882) - hint #1 - today's theme

NYT Strands today (game #882) - hint #2 - clue words

Play any of these words to unlock the in-game hints system.

• BURP
• HABIT
• NATION
• CLAP
• ALONG
• APPEAL

NYT Strands today (game #882) - hint #3 - spangram letters

NYT Strands today (game #882) - hint #4 - spangram position

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

NYT Strands today (game #882) - the answers

The answers to today's Strands, game #882, are…

• AMBITION
• GOAL
• PLAN
• PURPOSE
• OBJECTIVE
• INTENTION
• SPANGRAM: AIMHIGH

• My rating: Easy
• My score: Perfect

I’d collected six non-game words before I found AMBITION and managed to resist the temptation of speeding things up with a couple of hints.

After I’d got going, GOAL, PLAN and PURPOSE came easily, but even though I saw OBJECTIVE I still managed to mess up the order connecting the letters.

All in all, a quite straightforward game with the most unusual feature being a rare right-to-left diagonal spangram.

Yesterday's NYT Strands answers (Saturday, August 1, game #881)

• SILLY
• ABSURD
• PREPOSTEROUS
• LUDICROUS
• LAUGHABLE
• SPANGRAM: GETREAL

• Etched reached a $10.3 billion valuation after securing fresh SK Hynix backing
• The startup built custom processors specifically for demanding AI inference workloads
• Etched says conventional GPUs waste power during many inference tasks today

Etched, a US-based AI chip startup, has secured fresh backing from SK Hynix as it hits a $10.3 billion valuation while pursuing processors built specifically for inference workloads.

The company argues that conventional GPUs deliver more computational capability than many inference tasks require, yet offer insufficient memory for increasingly demanding AI models.

Its latest funding will support the production of rack-scale inference systems designed around custom chips, shared memory, and lower power consumption, rather than general-purpose graphics processors.

Etched bets custom inference chips can outperform traditional GPUs

Founded by former Harvard students Gavin Uberti, Robert Wachen, and Chris Zhu, Etched focuses exclusively on AI inference rather than training large language models.

The startup says its systems combine Low Voltage Inference (LVI) technology with Cluster Scale Memory (CSM), allowing processors to access substantially larger shared memory pools than conventional GPUs.

“Today, AI chips can't scale FLOPs without thermal throttling. As FLOPs utilization increases, AI chips draw more power and downregulate clock speed…,” said Chris Zhu

We’ve designed a new architecture to run our chip’s math blocks at under half the voltage of most AI chips. This enables multiple times the FLOPs density of AI chips today.”

The researchers claim that existing processors struggle to increase floating-point performance because higher utilisation raises power consumption and eventually reduces operating clock speeds through thermal limits.

Processors using High Bandwidth Memory (HBM) cannot achieve SRAM-level decoding speeds because memory subsystems and interconnects introduce additional latency during inference workloads.

They therefore created a shared low-latency memory pool connected through what it describes as a proprietary ultra-low-latency, high-bandwidth interconnect to enable faster memory access

“Our HBM/SRAM hybrid design solves both memory capacity and mem2mem latency, enabling high throughput and interactivity simultaneously.

“CSM improves latency and avoids today's cost, reliability, yield, thermal, and compute tradeoffs of SRAM-only chips, 3D DRAM chips, or optics.”

The company also claims many AI models waste time moving data between chips, memory, and networking hardware before processing can continue.

According to Etched, its CSM architecture reduces those delays by minimising additional memory layers during data transfers.

Funding surge accelerates production and global expansion

Etched has attracted funding at a rapid pace, raising $5.4 million in its 2023 seed round, $120 million in 2024, $500 million in 2025, and $300 million in 2026.

Those investments bring total funding to approximately $925.4 million, while the latest financing doubled the company's valuation from $5 billion to $10.3 billion.

The C-round included Sequoia Capital, Andreessen Horowitz, Jane Street, Diffusion, Argo, and SK Hynix, signalling continued investor confidence despite increasing competition within AI hardware markets.

Robert Wachen said, "This round accelerates production of our inference clusters," while confirming an 80,000 sq ft, 10 MW facility opened near Milpitas.

The company now employs more than 400 people, reports customer demand exceeding $1 billion, and has established manufacturing operations in Taiwan.

Etched systems will support conventional large language models, mixture-of-experts architectures, and alternatives including Mamba.

Via Blocksandfiles

AI demands a huge amount of energy, water, compute, and, of course, money. Big tech companies, including Google and Meta, have pledged huge sums of money toward the capital expenditure (capex) to continue the push to more powerful AI. But many inventors feel increasingly wary that this spending is simply swelling an already inflated bubble.

Laying the tracks

Speaking with CNBC in July 2026, the legendary investor Warren Buffett, who recently retired, weighed in on the capex debate.

Big tech companies are pledging a huge sum to the AI infrastructure buildout, which may reach up to $750 billion over 2026 and up to $4.5 trillion by 2030 among the four largest hyperscalers, including Meta, Microsoft, Amazon, and Alphabet. Many suggest this is further evidence of a bubble being fuelled aggressively by big tech companies doubling down on a technology that hasn't demonstrated meaningful returns.

But explaining his firm Berkshire Hathaway's 17.85 million-share purchase of Alphabet stock, disclosed in November 2025, Warren Buffett drew parallels between this level of spending and that during the 1800s to build the railroad system.

From digital to physical

As an investor, Buffett famously avoided software giants, suggesting they were asset-light businesses and didn't see the value in the investment propositions. This is a point he's since conceded he was wrong about. The firm he led made an additional $10 billion investment in Alphabet in June 2026.

The rationale is that there's currently a pivot underway in which these companies are shifting from the purveyors of digital products, namely applications and software, toward the physical infrastructure and compute to power AI.

Although the hyperscalers have been investing heavily in cloud infrastructure to power their software as a service (SaaS) propositions, the spending that AI commands is on a completely different level, as the figures show. That's why, in the end, Buffett believes that Alphabet proved a sound investment – given he believes that they have the cash flow to sustain the ongoing physical infrastructure buildout.

“Sometimes I find I get to thinking of the past,” sang (or croaked) Leonard Cohen, and I’ve found myself recently doing the same. However, while Cohen was lamenting the death of a relationship, my early-onset nostalgia has been triggered by something far more tragic and poetic: the death of affordable computer components.

I’ve spent the majority of 2026 marvelling and despairing, in equal measure, at how components, especially memory-based ones, such as SSDs, RAM and graphics cards, have skyrocketed in price. As you probably know by now, this mainly due to pressures on the supply chain from a seemingly endless procession of AI services that have been gobbling up memory.

Not only has this made upgrading your PC extremely expensive (upgrading the RAM was once one of the most cost-effective upgrades you could make), it’s also made products such as laptops, games consoles, and phones more expensive too. It’s upended decades of the tradition of technology getting cheaper over time.

Take both the PS5 and Xbox Series consoles. In previous generations, six years after launch, consoles should have had numerous price cuts, making them almost impulse buys at this stage; along with a healthy second-hand game market, this traditionally meant that waiting until late into a generation to get a console was a great way to save some serious cash.

However, both the PS5 and Xbox Series have had several price rises this generation, which means buying a six-year-old bit of hardware is now more expensive. This is far from an ideal situation (and one of the reasons why the PS5 will likely be the first PlayStation console I don’t buy), and it’s not unique to consoles. Apple recently hiked the prices of its Macs and MacBooks (including the budget MacBook Neo, which launched at $599 / £599 / AU$899, but has since risen to $699 / £699 / AU$1,059 and which in turn quite drastically undermines the Neo’s value proposition).

Watching the tech I write about (and love) get increasingly expensive and inaccessible for a lot of people is getting pretty depressing. Take the Steam Deck: recently, it was announced that Steam Deck sales have dropped by a whopping 80% after Valve increased the price of the gaming handheld to $789 / £649 / AU$1,199, a $240 / £170 / AU$300 rise.

Now, I've loved the Steam Deck since day one, and would always recommend it to most people thinking about buying a new gaming laptop or handheld, simply due to its excellent value.

Now? Not so much. I simply can't justify the idea that people should buy an old, underpowered bit of kit that’s almost as expensive as the budget gaming laptops I used to tell people to avoid.

For many of us PC gamers, it also means upgrading or building a new gaming PC is now out of the question, and the hardware we currently have will need to tide us over for a few more years.

But, does that mean you can’t upgrade your PC at all for the moment? Thankfully, there are still plenty of ways to boost your computer without spending a fortune, such as upgrading parts that aren’t memory-dependant (for instance getting yourself a new monitor), or tweaking the software.

As for me, I decided to transplant one of my existing gaming PCs into a new case, while also taking the opportunity to upgrade the cooling. One of the most fun things about building PCs is coming up with a theme or design, and because of my disillusionment with the current state of computing, I thought it would be fun to turn my PC into a retro device that harkened back to a time when computing really was fun: the 1990s.

Cut to the case

I've been feeling for a while that this is a bit of a mid-range setup that is (outside of the RTX 5070) showing its age, which is why I mainly use it for working, with a spot of image and video editing and the occasional game.

For pretty much all of my gaming needs, I have a far more powerful gaming PC with newer components (and an RTX 5090), but looking at the specs at this point in 2026 during the midst of the RAM crisis, I think I was being rather harsh on this PC. For a start, the 32GB of DDR5 RAM and 2.5TB of storage feels positively decadent these days.

So, my main goal was to swap the (perfectly fine, if a little dull) Fractal Design Vector RS case for something with a bit of retro style. Enter the Thermaltake Retro 360 TG Mid Tower Chassis.

This is a case built for modern PCs, but with a pleasingly retro look. Of course, it’s a wonderful shade of beige that will instantly give flashbacks (the good kind) to anyone who used a PC in the 1990s. It also has sections for a CD-ROM drive and floppy disk drive, though those aren’t usable and are really more of a gimmick.

Still, it looks like it could have come straight from the '90s, while still supporting modern hardware, including space for all-in-one CPU coolers, and a front panel that includes a USB-C port. It also has a tempered glass window to show off the components inside, something that’s common these days, but wasn’t really a thing back then. This combination of retro looks and modern capabilities was perfect for my needs.

Cooling off

To make things easier when transporting the motherboard and installed components into the new case, I was planning on removing the existing Corsair iCUE H150i RGB ELITE Liquid CPU cooler, giving me an excuse to install a new one.

To make the most out of the glass window of the Retro 360 TG case, I picked the Thermaltake TH360 V3 Ultra ARGB Sync AIO Liquid Cooler, which — as well as having powerful cooling — comes with a 3.95-inch LCD display that looks (to my mind, at least) like an old CRT monitor.

This screen can display system info (such as the temperature of your CPU), as well as custom animations, and I thought it would help liven up the inside of the PC, while still keeping with the retro-modern theme.

I also had a spare pack of Corsair RX120 RGB fans (the perks of being a tech journalist) that I wanted to use. For a start, they're white, which would fit the beige aesthetic, and it also meant I wouldn’t have to remove the case fans in the older chassis, potentially saving me some time and annoyance (the Thermaltake Retro 360 TG Mid Tower Chassis doesn’t come with fans, and they are essential for keeping the PC cool when in use).

Prepping for surgery

Once I’d assembled everything I needed to transform my PC into a retro powerhouse, I prepared to make the switch. Because I was swapping an already-built PC into a new case, the prep work differed slightly than if I was building a new PC from scratch.

The first step was to remove the bulk of the PC from the old case. To make things easier, I wanted to keep as much as I could in the motherboard, but there were some components that needed to be removed.

So, I started by unscrewing the large CPU cooler from the top of the processor. I then removed the GPU and WD_BLACK AN1500 SSD, which like the GPU, slots into a PCI-e lane of the motherboard. The other SSD is screwed directly into the motherboard via the M.2 port, so I didn’t need to worry about that.

I then carefully disconnected the case fans and I/O connectors (which allow you to turn the PC on and off using the buttons on the case, use USB ports, etc), then unscrewed the motherboard and lifted it out of the case.

Once the motherboard was out, one of the first major hurdles that you wouldn’t encounter if building a PC from scratch presented itself. I noticed when I tried placing the motherboard in the new case that it wouldn’t sit properly. Taking it out, I realized that when unscrewing the motherboard from the old case, some of the standoffs (which screw into the case to lift the motherboard off the chassis) had been unscrewed as well, and were still attached to the motherboard.

So, I needed to hunt down some pliers, which allowed me to hold the standoffs and unscrew the motherboard from them.

Once off, I was able to screw the motherboard into the new case (which already had its own standoffs installed) without an issue. It wasn’t a major inconvenience, but a good example of how moving cases can require tools (in this case pliers) that you’d not usually need when building a PC.

Plugging in

With the motherboard installed, I then plugged in the new cases’ I/O connectors. One thing I loved about the Thermaltake Retro 360 TG Mid Tower case is that front panel’s wires, which are often individual wires that can be extremely fiddly to attach to a motherboard, came as a block, which made installation much easier and faster.

While the larger components were still laying on my desk, I also connected up the PSU cables. The Corsair HX1500i 80 PLUS Platinum, which again I transported from the old PC, is a modular PSU, which means the cables aren’t permanently attached to the PSU — instead you can just plug in the cables you need to both the motherboard and PSU.

This makes installation a lot easier, and cable management much neater, and it meant I could plug the cables into the motherboard without having to install the PSU until later (a decision I made to make things easier, but I’m not sure it worked!).

I also took the opportunity to install the three Corsair RX120 RGB fans, adding two to the front of the chassis to pull in cooler air, and one at the rear that pushes hot air generated by the PC’s components out of the case.

These fans are more advanced than the previous ones I used for the PC: rather than being plugged into the motherboard and controlled that way, they daisy chain to each other, then connect to an iCUE LINK system hub, which then connects to the motherboard via a microUSB cable, all powered by the PSU.

This made the installation harder, as it meant I needed to route the proprietary cables to connect each fan to the hub, and some of those cables weren’t that long. It also meant having to hunt down an extra PSU cable, which caused issues later. While the iCUE LINK system hub meant I had greater control over the fan’s RGB lighting and temperature control, I don’t really think it was worth the extra effort. If I did this again, I’d stick to standard, easier-to-install case fans.

Finally, I installed the TH360 V3 Ultra ARGB Sync AIO CPU cooler. I removed the existing thermal paste from the CPU and reapplied a small blob. I had ordered extra thermal paste as I wasn’t sure if the CPU cooler would include some, as not all do. As it turns out, the cooler does come with a small tube of thermal paste, so I now have a spare tube, but it’s always handy to have extra.

It was important to reapply the thermal paste, as the existing paste on the CPU was now several years old, and some of it remained on the cooler when I removed it. To ensure maximum cooling efficiency, removing the old paste and applying new paste was the best, and safest, option.

Installation was easy: I screwed the radiator and fans to the top of the case, attached the fan controls to the CPU Pump header on the motherboard, then placed the screen (which attached via magnets) on the cooler.

With that in place, I then put the GPU and SSD back in their PCIe slots, connected (almost) everything up to the PSU, then secured the PSU in the case.

I say ‘almost’, as I didn’t have a cable to attach the iCUE LINK hub to the PSU. As I mentioned earlier, the PSU is modular, so I had only plugged in the cables I needed during the original build, but now I needed an extra cable. The PC wouldn’t boot without the hub powered on, nor would it boot when I tried a cable from a different PSU (this is likely a safety feature — you should never use cables from a different PSU, but I wanted to check).

Thankfully, after some rummaging around, I found the extra cables from the PSU, installed the correct cable to the hub and the PC booted fine. This wouldn’t have been an issue if I’d used a more basic fan, but at least I was now up and running. It also highlighted an important lesson: always keep hold of cables and any other extras that come with your components, even if you don’t initially need to use them.

Finishing touches

With the PC booting without issue, I then added the finishing touches. For instance, I ordered some retro stickers from Geekenspiel, a website that makes the kind of stickers you used to find on PCs, but for modern hardware (such as a sticker with the old Nvidia logo, but referencing the modern RTX graphics card).

I also paired the PC with the 8BitDo Retro Mechanical Keyboard, a wonderful-feeling keyboard that offers modern features but with an old-fashioned look and beige colorway that perfectly complements the PC.

I also ordered a Microsoft ProIntelli Wired mouse, which is a modern mouse inspired by Microsoft’s iconic IntelliMouse peripherals. Again, it combines retro looks with modern tech, which makes it a perfect choice for this build.

With the peripherals attached, I logged into Windows 11 and set up the TT RGB PLUS 3.0 software, which allows me to tweak the settings of the TH360 V3 Ultra ARGB Sync AIO CPU cooler, and configure the screen. I made a short retro-inspired gif of the TechRadar logo, and used the software to display it on the LCD screen of the cooler.

I was very pleased with the results, and it instantly made the PC more interesting visually while still feeling retro. However, I have to say the TT RGB PLUS 3.0 software seemed quite flakey, with some options simply not working. The software also needs to load in Windows 11 before the animation plays on the screen.

The fans of the cooler were also very loud, and setting them to ‘Quiet’ in the software didn’t seem to make a difference. So, I restarted the PC and went into its BIOS and tweaked the fan speeds there, which worked. Suddenly, the PC was whisper-quiet.

Turning Windows 11 into Windows 95

I wasn’t quite done, however. While the hardware now certainly looked the part, I wanted the software to match. At first, I considered ditching Windows 11 altogether and instead using a Linux distro skinned to look like Windows 95.

The benefits of this is that I’d be free of all the bloat that Windows 11 comes with, especially the AI stuff I have no interest in, and also that Linux is far more customizable; with a bit of time, I could make an operating system that looked and acted a lot more like Windows 95.

However, there were also plenty of downsides. For one, it would be time-consuming to install and configure a distro, and many PC games and software won’t run natively on Linux. The TT RGB PLUS 3.0 software, for instance, won't work on Linux, and while there are workarounds, they are again time-consuming and fiddly. Plus, this approach would mean things like adding gifs to the screen wouldn’t work — and I wanted my gifs.

So, I decided to stick with Windows 11, but heavily mod it to look like Windows 95. This involved using WindowBlinds 11 and Start11 by Stardock. The first piece of software lets you drastically change the look of Windows 11, and has a ‘classic’ theme that’s perfect for my needs.

Start11, as you might have guessed from the name, focuses on the Start menu, and allows you to tweak it so that it looks, and works, like the classic Start menus of Windows 95 and Windows 98.

Each of these bits of software took a bit of configuring, but before long I had Windows 11 looking, and feeling, like Windows 95. Changing the desktop icons through Windows 11’s settings to feature classic Windows icons (I downloaded them from DeviantArt), and changing the desktop wallpaper to a solid teal color completed the look.

While not perfect (Windows Explorer still tries to show Windows 11-era tabs, and the taskbar clock can be a bit funny), I now had a retro-inspired modern operating system to complement my retro PC.

Overall, I was pretty happy, and a comforting sense of nostalgia washed over me as I used my big beige box. For a little while, I was able to pretend the tech world wasn’t a dumpster fire, but a more fun, perhaps more innocent, place to be.

Final thoughts

Overall, I’m so glad I tried this out: it feels, ironically, as if I’ve given my aging PC a new lease of life. It looks great, and the mouse and keyboard feel fantastic — and familiar. Windows 11 also feels more sprightly, as I turned off a lot of the unnecessary animations and visual effects that usually run when using the operating system.

Strangely, I have also managed to fix a weird startup bug the PC used to have where sometimes the GPU didn’t power on. Since moving the PC into its new case, this has not happened once, so I must have inadvertently changed something during the move (not just by removing the GPU, as I had tried that before, and even swapped GPUs, and the problem persisted), so that’s a nice side effect.

The process was a lot faster with an existing PC than if I’d built one from scratch, though that has meant I needed to make some compromises. The motherboard and components don’t really complement the beige case, for instance; if I’d been building a new PC, I’d have probably tried to get white components instead, and cables to match.

I’m also glad that I’m a bit of a hoarder when it comes to tech, and I hadn’t thrown away the PSU cables I didn’t originally use. Okay, so I might have felt a bit daft holding on to cables for almost half a decade that I would likely never use, but it meant that I had the correct one when I needed it.

Finally, it proved that even in 2026, building, tweaking and upgrading your PC can be fun and creative, and you don’t have to spend a fortune if you can repurpose a lot of the components.

The retro theme of both the PC itself and the software it runs looks and feels great as well. It was hard work, but well worth it. If this article has inspired you to do something similar, let me know in the comments below!

• Ukraine receives rights to produce Britain’s Stone Cloak electronic warfare system
• Tablet-sized device helps Ukraine counter growing Russian drone operations
• Stone Cloak expands Ukraine’s electronic warfare capabilities against unmanned threats

Ukraine is receiving access to the intellectual property behind the UK-developed Stone Cloak electronic warfare system.

The tablet-sized device is designed to disrupt Russian drone detection systems and strengthen Ukraine’s ability to counter unmanned threats.

Thousands of Stone Cloak units have already been supplied to Ukrainian forces as electronic warfare becomes increasingly important on the battlefield.

Ukraine gains rights to produce British-designed drone defence system

British Prime Minister Andy Burnham recently hosted Ukrainian President Volodymyr Zelenskyy in London, where he announced the United Kingdom will transfer intellectual property rights for its Stone Cloak electronic warfare system to Ukraine.

“Stone Cloak is an example of the best of British innovation, proven on the front lines, and will be crucial to safeguarding the security of both our nations,” Burnham said.

“This will help them hit their targets and save Ukrainian lives. It also demonstrates the kind of innovations our industries can create together.”

The transfer gives Ukraine the ability to manufacture the compact electronic warfare equipment without depending entirely on external deliveries.

The system works by interfering with drone detection capabilities, helping Ukrainian forces reduce the effectiveness of hostile unmanned aerial operations.

Electronic warfare systems like Stone Cloak have become a major part of modern combat as drones continue shaping battlefield tactics.

British officials described the transfer of intellectual property as a way to support Ukraine’s defence needs through expanded manufacturing capability.

The agreement forms part of broader defence cooperation between London and Kyiv involving military assistance, training, and technology sharing.

“The UK stands shoulder to shoulder with Ukraine, and our support remains unwavering. Russia should have no doubt about our resolve, and we will not back down until we achieve a lasting and just peace for Ukraine,” Burnham said.

Ukrainian President Volodymyr Zelenskyy has continued seeking greater access to defence technologies that can be produced within Ukraine.

The Stone Cloak system is among several technologies aimed at improving protection against increasingly common drone threats.

Its compact design allows easier deployment compared with larger electronic warfare systems that require more extensive equipment and infrastructure.

Compact electronic warfare gains importance against drone threats

The growing use of unmanned systems has increased demand for smaller electronic warfare tools capable of responding quickly.

Drones have become widely used for surveillance, reconnaissance, and attack missions, creating greater demand for effective countermeasures.

Systems like Stone Cloak are designed to provide another layer of protection against these increasingly common battlefield threats.

The effectiveness of electronic warfare equipment depends on continued improvements because drone technology and countermeasures continue evolving.

Manufacturing Stone Cloak inside Ukraine could allow faster adaptation of the system based on battlefield feedback and operational requirements.

However, production expansion will depend on industrial capacity, component availability, and the ability to maintain reliable manufacturing processes.

Future availability of Stone Cloak units will depend on how quickly Ukrainian production facilities can scale up operations.

Via Ukrainska Pravda

I bought my first Tesla Model 3 in June 2019. It was my dream car. The promise of getting in, selecting a destination, and the car taking care of the rest was one I bought into.

I originally bought the Model 3 without Full Self Driving (FSD), which at the time cost $6,000. Then about a month later, Elon Musk announced the cost would start going up by $1,000 every few months. So, I bought FSD outright before that happened. I even ordered custom license plates to make it very clear to other drivers my car was driving itself.

A couple of years later, the first beta arrived, and during the next few years, after many updates, it got good. Really good.

I know headlines love to point out when an accident happens during FSD use — and rightfully so — but my overall experience with it as of a year ago was that I felt safe when using it. I could get in, tell it where to take me, and the majority of the time, I only had to babysit.

I used FSD probably 95% of the time; I was behind the wheel. I no longer hated driving.

Goodbye FSD, hello Comma 4

Then, life changed. My wife and I decided to open a pizza food truck, and we quickly learned the truck we planned on using wasn't up for the job of towing our trailer. So, I traded in my Model 3 for a 2025 Ford F-150 Lightning Flash. Bye-bye FSD.

My Lightning came with three free months of BlueCruise, Ford's hands-free driver-assist system that only works on highways. When you're on standard roads, Ford has a basic lane-keeping feature that keeps you within the lines, and that's about it.

After my BlueCruise trial ended, I opted not to pay the $50 a month/$500 a year subscription fee, mainly because it's limited to highways and I wasn't doing a lot of highway driving at the time. In town, I was still able to use lane keep and deal with the super annoying "Keep Your Hands on The Wheel" nag that goes off all the time.

I didn't miss the more advanced self-driving-like features of BlueCruise at all. I got by.

Then, in January, I had shoulder surgery. The extended recovery process meant driving an hour away for physical therapy at least once a week, and I'm still making that same drive. So, I decided to turn my Ford Rewards points into a couple of free months of BlueCruise to help me get back and forth.

Only this time, I loved it. I could cruise up and down the highway at 80 miles per hour, tap the blinker to start an automatic lane change and just relax — all hands free.

With a few weeks left in my extended trial, I decided to sign up for the $500-per-year subscription. While I waited for the charge to hit my credit card, a friend reminded me about Comma.ai, the small dash cam-like device that helps control your vehicle.

In Comma's own words: "It's upgraded active driver assistance for your Toyota, Hyundai, Ford, and more." Comma states it supports over 300 different vehicles.

Knowing I was about to commit to $500 per year for BlueCruise, curiosity got the best of me, and I spent over a week reading everything I possibly could about Comma and how it works.

The overly simplified explanation is that you need to install the device and a wire harness specific to your vehicle. The two components then intercept and send driver assistance commands to your vehicle, enabling hands-free driving on any marked road — not just highways.

I also spent more than a few days reading through and asking far too many questions in the Ford-specific Discord channel. The group there was amazing and patiently answered every single one.

I'll admit I felt very uneasy about random people on the internet writing software that can control my car. I still do, to an extent, but eventually I felt comfortable and curious enough to order the $1,000 gadget.

Install, setup and daily use

A few weeks later, a small box arrived with the camera, the Ford-specific wiring harness, and zero instructions.

Thankfully, there are quite a few YouTube videos detailing the installation process. That process requires installing the wiring harness under the dash, where my truck's Image Processing Module A (IPMA) is located. I had to remove some paneling to get to it, then route the included OBD-C cable from under the dash to my rear-view mirror where the Comma 4 is mounted.

The install process took about an hour, with the most difficult part being connecting the wiring harness to the IPMA — it's such a tiny space and hard to get to, but eventually I got it.

With the harness connected, the Comma 4 mounted, and the cable run, the next thing I needed to do was install an operating system.

The Comma 4 comes with a bare-bones operating system installed, able to do just enough to help facilitate whatever version of driving software — commonly referred to as forks — you want to use.

OpenPilot is Comma's own driving software, but it's not optimized for every vehicle. For my Ford Lightning, I have to use BluePilot, a Ford-specific fork that's customized and optimized to handle quirks in driver-assistance commands, for the best experience.

I went through the software setup process in about 10 minutes, and then it was ready for me to take a short calibration drive. Once that was done, I had my very own self-driving assistant, ready to begin driving me around.

Activating Comma is done using the same buttons I would use to turn on BlueCruise or Ford's lane-keeping features. I can see when Comma is active by looking at the Comma 4's screen for green border indicators where the edges of the lane are, and a solid green line in the middle showing that path the truck is going to take.

It took me a few days of tweaking features and settings to fine-tune how the Comma controls my truck. For instance, out of the box it doesn't handle lane changes for you. You have to enable lane changes, and even then it'll only perform a change when you activate your blinker — it won't do it on its own.

You can even customize how aggressive it is during the lane change, how long it will wait after activating the blinker before changing, so you need to confirm the change by nudging the wheel in that direction before it fully commits. It's really cool stuff.

What's taken me the longest to get just right is adjusting my truck's alignment in a lane. With default settings, my truck was consistently riding the left side of every lane I was in. It wouldn't go over it, but it was too close for my comfort. I spent some more time figuring out the proper setting to adjust and started tweaking. A few minutes later, my truck is almost perfectly centered.

I'm still tweaking different settings, but for the most part it drives just like I want. I continue to experience some very slow ping-ponging — when the car goes from one side of the lane to the other and then back — but it's easily remedied if I just place my hand on the wheel. From what I understand, I may never get fully rid of it.

I've learned a lot along the way, too. In my Ford, the Comma 4 only controls lateral movements to keep you centered in the lane, not longitudinal controls such as speeding up and braking. Ford's ACC system still handles the latter feature – something I found reassuring.

A couple of disclaimers you need to know

There are a few things I want to make clear, should you decide to get a Comma 4 (or one of the older models).

First, you, as the driver, are responsible for your vehicle at all times. The Comma 4 is an assistance device, not a fully-take-over-all-controls device. It hasn't tried to steer me unexpectedly into another lane, but I've seen reports and feedback about a recent update doing just that.

Second, Comma 4 will never offer a Full Self Driving-like point-to-point experience. There are just too many variables to how a third-party device can interact with and control various vehicles. In a Ford, for example, Comma 4 has limits in how aggressively it can steer, which makes tight corners difficult if not downright impossible. A recent update to BluePilot has made progress in this regard, but a 90-degree turn is all but out of the question.

1,000 miles and no regrets

My original intent was to take full advantage of the 30-day no-questions-asked return policy and send the Comma 4 back if I had any reservations after testing it. And, well, that window closed a few weeks ago.

All told, I'm nearing 1,000 miles over 97 driving sessions and 26 total hours since installing the Comma 4 in my truck. Not all of that is with it activated, but I'd wager that 75% or more of those miles were driven by Comma.

It's earning my trust with every drive and scenario I throw at it. The system has built enough confidence that I'm now going fully hands-free on longer drives.

Even with its quirks and limitations, the trade-off is a no-brainer for me. The Comma 4 handles backroads and standard city streets where BlueCruise simply gives up.

Best of all, I don't have to pay Ford $500 a year for a highway-only feature on top of what I already pay for premium connectivity. For a one-time $1,000 hardware investment, I got a sliver of my FSD experience back, saved money in the long run, and gained a system that continuously gets better thanks to an active community.

• ISBNdb ships up to a million books anonymously to AI labs
• Pre-2022 books are prized because chatbots never touched their text
• Spine-cutting scanners destroy originals to speed up digitization for training

AI companies are increasingly turning to printed books published before 2022 as preferred training material because those works predate the widespread use of AI-generated content.

Large-scale scanning operations reportedly involve cutting book spines, separating pages, and destroying physical copies to create digital datasets for large language models.

The practice has attracted growing criticism because some books entering these pipelines are reportedly extremely rare, raising concerns about irreversible cultural losses.

Pre-2022 books become valuable AI training material

Reports by 404 Media found data broker ISBNdb supplies physical books in bulk to AI developers seeking human-written material unaffected by modern chatbot output.

The company argues books published before 2022 offer cleaner datasets because they cannot contain text generated by contemporary large language models.

They are often considered "dense, edited, authoritative," in contrast to internet content increasingly filled with machine-generated material of uncertain quality.

The approach also attempts to avoid so-called model collapse, in which AI systems gradually lose quality after repeatedly training on synthetic content generated by earlier models.

ISBNdb additionally argues that older printed works avoid deliberate data-poisoning techniques authors increasingly use to disrupt AI training through carefully modified documents.

However, there are reports that many of these books are scanned using high-speed equipment.

This equipment requires workers to remove the spine before feeding individual pages through automated imaging machines.

That process reportedly destroys the original volume, making rapid digitisation considerably cheaper than slower preservation methods designed to keep books physically intact.

Secrecy and legal rulings fuel preservation concerns

ISBNdb openly acknowledges reputational concerns surrounding the practice while offering strict non-disclosure agreements that keep customer identities confidential throughout commercial engagements.

Its website reportedly states, "'AI company destroys two million books' is not a headline that generates sympathy," while suggesting clients describe the process as digital preservation.

Such a level of destruction is an order of magnitude bigger than the loss of the Library of Alexandria. Yet, it is unfolding with none of the outrage that history reserves for burned libraries.

Booksellers interviewed by 404 Media said some volumes entering these scanning programmes have very few surviving copies after enduring wars, fires, and centuries of handling.

Critics argue that unlike websites or widely available modern publications, exceptionally scarce historical works cannot simply be reproduced after their physical copies disappear forever.

A recent United States court ruling involving Anthropic found that scanning legally purchased books for AI training constituted fair use under specific circumstances.

Part of that reasoning held that destroying each printed copy during scanning meant one legal copy effectively replaced another rather than creating multiple copies.

In response to a critic (@Hedgie) of this method on X, Elon Musk said, "I've asked the SpaceXAI team to preserve any rare books in a library and scan them the hard way," suggesting an alternative approach.

If significant awareness is not created, this quiet erasure of irreplaceable books risks becoming the defining act of cultural loss for this era, remembered only after it can no longer be undone.

You could be sitting on a treasure trove of fantastic iPhone, iPad, Mac, and Apple TV games and not even know it.

Apple Arcade might be one of the most underrated gaming services on the market right now, with many Apple One subscribers not even realizing that it is included in all tiers of the service. You can also subscribe to it separately and you could even be able to access a generous free trial right now if you're reading this on a shiny new Apple device.

But what exactly is Apple Arcade, how much does it cost, what kind of games can you play on it, and, perhaps most importantly, are any of them actually good? I've spent the last week really sinking my teeth into the service in order to find out and deliver this ultimate guide to everything that you need know.

From all the basic facts to a bumper list of my personal game recommendations and even a personality quiz to match you to titles that you might like, everything here is intended to help you walk away an expert in all things Apple gaming.

The ultimate guide to Apple Arcade

• Instructions hidden as white text in a Word document can make Microsoft 365 Copilot silently alter the file it is drafting and copy the instructions into the output
• Each poisoned document becomes a carrier, so the attack spreads through ordinary internal workflows without the original malicious file and needs no macros, malware, or code execution
• Microsoft has shipped two mitigations across a 144-day disclosure, including a model upgrade, and the attack was still reproducible by the researcher

A security researcher has published a proof of concept showing that instructions hidden inside a Word document can cause Microsoft 365 Copilot to silently alter the file it is drafting, then copy those same instructions into the finished document, so the next person to use it becomes a carrier too.

Håkon Måløy, a data scientist with a doctorate in applied machine learning, disclosed the technique as the third installment of his Context Collapse series, after a 144-day coordinated disclosure with the Microsoft Security Response Center.

The reason this is being reported ahead of a fix is that it still works despite multiple attempts by Microsoft, as he notes that no robust mitigation for the broader vulnerability class is currently available.

A clever attack designed around Copilot's approach to text

The underlying attack belongs to a family known as cross-domain prompt injection, or XPIA. An attacker writes instructions in a natural-language document, formats them as white text on a white background at a small point size, and shares the file.

Because Copilot for Word strips formatting before passing text to the underlying language model, the model reads text the human never sees. This is true even for documents that are not opened by the user on purpose: The attack can trigger either when a user manually attaches a document to Copilot or when Copilot, working in Work IQ mode, searches the user's OneDrive for relevant files and finds the malicious one on its own.

It is also more dangerous than other exploits because of one key element: propagation. The hidden prompt in Måløy's proof of concept had two parts. One instructed Copilot to alter the document being drafted, in his demonstration halving every financial figure in a quarterly report.

The other instructed Copilot to copy the prompt into the new document and conceal it, framed innocuously as source tracking and readability formatting. Copilot did both, appending the instructions in white text and mentioning neither action to the user.

The disclosure timeline is the most uncomfortable part of the report. Måløy reported to MSRC on March 6 2026. Microsoft confirmed the behavior on March 31 and shipped a first mitigation in early April via a reworked Edit with Copilot experience, which successfully blocked his original prompt wording. He reproduced the attack with different wording the same week, and a second case was opened.

The second fix, on July 14, consisted of upgrading the underlying model to GPT-5.5. Måløy broke it the following day using GPT-5.6, then voluntarily offered Microsoft a further two-week delay to attempt another mitigation. The class still reproduced on the disclosure date, indicating that although a fix is in the works, the exploit is still possible to run.

A complicated issue that lacks a proper resolution

The issue goes far beyond Microsoft Word: an AI assistant must ingest untrusted content to determine whether it is relevant or hostile. But the content enters the same context window as the system prompt and the user's actual request, so by the time the model evaluates whether the text is an attack, the attacker's tokens have already shaped that evaluation.

As Måløy puts it, "the content being inspected participates in the act of inspection."

Microsoft confirmed it had reviewed the findings in a statement to The Register but stopped short of indicating a timeline for a complete fix:

“We have addressed the findings reported by the researcher and thank them for working with us through coordinated vulnerability disclosure," the company said.

"To address this class of risk, we use a defense-in-depth strategy with safeguards that block malicious instructions at multiple points and help keep tasks aligned with users’ requests. We are continuously strengthening these safeguards as the technology and threat landscape evolve. We encourage customers to install the latest updates, use multiple layers of security protection, treat content from unknown sources with caution, and review AI-generated content before using or sharing it.”

Måløy's recommendations are to treat externally sourced documents as untrusted when using them with Copilot, review attachments before starting an AI-assisted draft, and review Copilot's output carefully before sharing or reusing it.

He also suggested that generated documents should carry provenance metadata that records source material and model edits, which would not prevent injection but would make an infection traceable after the fact.

With Copilot extending further into agentic products that create and manipulate documents with less human oversight, the scope of how exploits like this could affect workflows (and users) will widen rather than narrow, and a complete solution is not yet in sight.

Whether you're a die-hard mobile gamer or a newcomer to the world of portable play, Apple Arcade will likely have something for you.

The subscription service gives you access to hundreds of games for your iPhone, iPad, Mac, or Apple TV, ranging from revamped mobile classics including Angry Birds Reloaded and Doodle Jump 2+ to original and exclusive games in massive franchises such as Sonic the Hedgehog with Sonic Dream Team. All of the included titles have absolutely zero ads and no microtransactions, meaning that they offer a more premium, console-like experience than you would get with many other App Store titles.

Most Apple phones now come with a free three-month subscription to the service, giving you the perfect opportunity to take your shiny new hardware for a whirl. Even if you're using an older device, there's still a one-week free trial that new subscribers can take advantage of, in order to see what the service offers and work out whether it's a good fit for them.

The only problem? With so many titles on offer it can be hard to know where to start, even if you're already a subscriber. That's why I've created this short personality quiz, which is designed to find out what kind of gamer you are and deliver a selection of recommendations. The recommended games have been personally tested by me, and are what I believe should suit your taste.

Did you find this quiz useful, or do you think I missed the mark with my picks? Let me know in the comments, and don't hesitate to drop your own game recommendations if you have them!