TechRadar News

• The BRRRRT Act seeks to keep at least 126 A-10s operational
• Iran operations revive debate over the A-10's military future
• Future A-10 retirements face stricter certification requirements under new legislation

A proposed House bill requires the US Air Force to retain at least 126 A-10 Thunderbolt II Warthogs, reversing plans to retire much of the nearly 50-year-old attack aircraft fleet.

The legislation follows renewed attention on the aircraft after its reported involvement in operations linked to the recent Iran confrontation, including combat search-and-rescue missions supporting downed F-15E aircrew.

Lawmakers backing the proposal argue the aircraft should remain in service until a fully operational replacement has been certified for every mission it currently performs.

Lawmakers push to preserve the A-10 fleet

Officially named the Bolstering Recognition, Resurgence, Retention, and Remembrance of the Thunderbolt Act, the legislation uses the acronym BRRRRT.

Introduced as H.R. 9780 by Rep. Abe Hamadeh, the Act takes its name from the distinctive sound produced by the A-10's 30mm Gatling gun during firing.

“The A-10 has repeatedly proven itself on the battlefield, saving American lives and delivering unmatched close air support when our troops need it most,” said Hamadeh.

“The BRRRRT Act ensures we do not retire a proven weapons system before a certified replacement exists.”

The proposal marks another chapter in a longstanding disagreement between Congress and the Air Force over the future of the Warthog fleet and its specialized missions.

The Air Force entered fiscal 2026 with 162 A-10 aircraft and sought authority to retire the entire fleet.

However, Congress rejected that proposal and instead required the service to retain at least 103 aircraft.

The BRRRRT Act would raise that minimum inventory to 126 aircraft, requiring the Air Force to halt scheduled retirements and recover additional Warthogs from storage at Davis-Monthan Air Force Base in Arizona.

Following recent operational use, the Air Force announced that A-10 squadrons at Moody Air Force Base in Georgia would remain active through 2029.

The service also plans to keep Warthogs flying from Whiteman Air Force Base in Missouri through 2030, leaving approximately 54 aircraft assigned to three operational squadrons.

Bill sets tougher conditions for future A-10 retirements

The legislation would give the Air Force 90 days after enactment to explain how maintenance, sustainment, pilot training, flight testing, and depot operations would be restored.

Earlier retirement preparations already included closing the depot maintenance line at Hill Air Force Base in February and ending the final training class for new A-10 pilots in April.

The Weapons School course at Nellis Air Force Base has also been winding down as the service continued preparing for the aircraft's planned retirement.

The proposal would also make additional retirements harder by requiring the defense secretary to certify that a replacement aircraft has entered full operational service.

That certification would also require written approval from the Army secretary, the Marine Corps commandant, and the head of U.S. Special Operations Command.

The bill also states that assigning another aircraft the close air support mission as a secondary responsibility would not qualify as a replacement.

If the BRRRRT Act does not become law, similar language could still appear in the annual defense authorization legislation that previously established the 103-aircraft minimum.

The Air Force maintains that the F-35A, F-15E, and F-16 will eventually assume the A-10's close air support and combat search-and-rescue missions.

Supporters of the BRRRRT Act argue those aircraft have not yet replaced every specialized capability performed by the Warthog.

Via Military Times

The CIA has never confirmed or denied the following story.

The only evidence we have that it happened at all is a 1997 article in Popular Science Vol. 250, Issue 1, where journalist Dawn Stover spoke to the mastermind behind one of the best-kept secrets of the Cold War.

The year is 1963. Kennedy still sits in the White House. Khrushchev holds the Kremlin. The Cuban Missile Crisis is still fresh in everyone’s mind. Paranoia runs high on both sides of the Iron Curtain.

And a repairman strolls unnoticed through the halls of a mansion in the capital. He steps into a busy administrative office used by Russian diplomats and the KGB. His bag of tools clunks with every swing. The man pauses at a large, boxy machine in the corner of the room. It hums and whirs.

The Mrs George Pullman House where the Soviet Union Embassy was located in 1963, pictured in 2010 (Image credit: Getty Images // Coast-to-Coast)

Launched just three years prior, the Xerox 914 was the company’s first successful photocopier - with a troubling tendency to catch on fire. But that’s not why the repairman is here today.

He drops his bag to the floor. The diplomats and KGB agents ignore him, the clerical staff barely see him, the hawkish security guards notice nothing out of the ordinary. The repairman diligently begins, as he always did, his routine maintenance work.

When he’s finished the job, he slips away as quietly as he came. He leaves behind a single trace of his presence.

The repairman has just become one of the CIA’s best assets.

The bowling alley plot

But the real story begins the year before in New York, at a disused bowling alley so anonymous you wouldn’t give it a single glance, let alone two. It’s far away from prying eyes.

Inside, Xerox’s head of government programs Donald Cary meets with four engineers. They’re not there to bowl. The Xerox men selected for the mission are Douglas Webb, Kent Hemphill, James Young, and thirty-six-year-old Ray Zoppoth, who knows the Xerox 914 inside out. He should do. He helped create the photocopier in the first place.

It’s exactly the expertise the CIA needs if the agency’s plan is to work.

Because the plan is to heist as many files as they can from the Soviet Union embassy, located within the Mrs. George Pullman House in downtown DC. With security so tight, not even a repairman, however inconspicuous, could hope to walk out unfrisked with stolen files stuffed down his pants.

So, the secret service has to get creative.

For days, the engineers discuss potential ways to capture images of the files being photocopied on the Xerox 914. Finally, Zoppoth hits upon an answer. Like all great ideas, it’s as maddeningly simple as it is cunning.

Mount a battery-powered home-movie camera with a zoom lens inside the copier. Aim the lens at the mirror used to reflect images onto the drum. And start taking pictures.

According to Stover’s piece, Zoppoth suggests the team “Mount a battery-powered home-movie camera with a zoom lens inside the copier. Aim the lens at the mirror used to reflect images onto the drum. Add a photocell that would prompt the camera to snap still frames whenever the photocopier lit up. And start taking pictures.”

Every time someone lays a document on the bed and presses 'Copy', the photocopier lights up, triggering the camera.

Easier said than done, but the Xerox engineers are excited by the breakthrough. They set to work installing a Bell & Howell 8mm movie camera inside a copier located in Xerox’s New York campus.

A vintage 8mm likely similar to one allegedly fitted inside the Xerox 914 (Image credit: Getty Images // Celine Lyneborg)

The noisy, bulky 914 is, it turns out, perfectly tailored to the role of spying device. It’s so impossibly large, it conceals the camera even when the maintenance panels are removed, and the operational sound as staff make copies masks the whir of the camera. No-one has any idea they’re being covertly watched.

The test run proves a success. Ray Zoppoth and the rest of the research team can see everything passing through the photocopier. “When we developed the pictures, we found recipes and copies of music and cartoons and jokes and all kinds of things,” he later tells Popular Science.

Code-name: Disneyland East

Weeks later, Ray Zoppoth finds himself deep inside a building used by the CIA. He's led into a basement room allegedly code-named ‘Disneyland East’. Zoppoth shows the spooks the Xerox invention. He teaches them how to fit the camera inside the 914. They, in turn, train the embassy’s regular repairman to install the camera and extract the film each time he returns.

Then, one day, in 1963, a familiar face arrives at the Soviet Embassy to carry out routine maintenance work. He holds a bag containing tools and, in its depths, an innocent 8mm home-movie camera. He’s waved through the gates without a thought.

The rest is history.

With lips tightly sealed and files classified, we’ll never know what was learned. The Cold War dragged on for another 27 long years.

But shortly after the project ended, Zoppoth confessed that the secret service again tapped up Xerox. This time, they want cameras hidden inside the more streamlined Xerox 813 desktop photocopier, used by ministries, agencies, and diplomatic offices around the world.

If the story is true - and only Zoppoth admitted it, with his colleague confirming but refusing to go on record - it suggests the joint operation was a success.

It remains a reminder that even the unsexiest machine in the office can be a security risk. We think of digital surveillance as a relatively modern fear. We worry that our phones are spying on us. That CCTV watches our every move. Even laptops have physical privacy covers to mitigate security risks. But it turns out, this goes back at least 60 years - right to the heart of the Soviet embassy in Washington DC.

The world of robot vacuums has come a long way. I've been testing and writing about these puck-shaped home helpers for over two years, and during that time I've watched suction specs skyrocket, bots sprout legs and pincer arms, and docks transform into formidable, self-sufficient cleaning stations. Robot vacuums are looking cleverer and more useful than ever.

I wanted to get an idea of where the industry is headed, and what the next generation of robovacs might look like. So I rounded up representatives from some of the best robot vacuum brands to get their insights — and it's clear from what they told me that this smart-home segment is going to get even bigger… and even weirder.

Complete cleaners

The obvious place to start is with the cleaning itself, and there's one point that my interviewees all agreed on: it's no longer all about the suction specs. "Customers are looking beyond headline suction figures and realizing the performance of robot vacuums is the combination of a lot of features," says Michael Meng, President of Dreame's robot vacuum division.

Nowadays, customers are looking for better cleaning, rather than just outright power, and robovac brands have been happy to oblige, with all manner of inventive solutions.

Customers are looking beyond headline suction figures and realizing performance is the combination of a lot of features

Michael Meng, Dreame

Roborock, for example, has been working hard to make sure no corner goes uncleaned. "We see a massive push for ultra-slim designs that don't sacrifice power," says Ricky Ma, general manager for Europe and America. "People want a device that can finally clean the forgotten spaces under low-profile furniture — couches, beds, and cabinets — where dust and allergens traditionally accumulate and remain out of reach for bulkier robots."

The brand's newest bots use an innovative navigation system that doesn't require a raised puck to sit on the top surface. Roborock — alongside Dreame — has also experimented with pucks that can extend or retract, periscope-style, depending on the vertical height available.

This navigation puck, on the Roborock Qrevo Curv 2 Pro, can retract when the bot needs to squeeze into a tight spot (Image credit: Future)

Mopping functions have been the subject of much attention, too. Wet cleaning is nothing new — in fact, these days you might struggle to find a model that can't mop — but I've been seeing an increasing preference for roller mops rather than flat mop pads (either of the D-shaped or disc-shaped variety). The benefit of a cylindrical mop is that the dirty water can be scraped off, and clean water added in, as the robovac goes about its mopping business.

"One development that particularly excites me is the introduction of roller mop technology, which helps ensure floors are continuously cleaned with a fresh mop surface," says Reena Patel,

Floorcare Category Manager at iRobot, the brand behind Roomba.

Faraz Mehdi, managing director at Eufy, agrees. "Users increasingly care about whether the robot is actually cleaning with a clean mop and truly removing dirty water, rather than just spreading mess around," he says. "Real-time self-cleaning mopping makes a meaningful difference."

Roller mops — like the one on the SwitchBot S20 here — can scrape the dirty water off as they roll around, to prevent spreading spillages (Image credit: Future)

Eufy is also looking into improving filtration. In manual vacuums, Dyson has popularized multi-step filtration systems, which make use of cyclones to separate waste particles of different sizes. However, that hasn't really carried over into robot vacuums, which typically have a much more basic setup.

Eufy's AeroTurbo filter setup is significantly more complex. "We’re interested in airflow and dust-separation systems that help deliver more sustained suction performance, especially in homes with carpets and hair," explains Faraz. "One of the biggest pain points in the category is that cleaning performance can fall over time as filters clog, so technologies like multi-cyclone separation matter because they improve long-term consistency."

Our robotic arm design folds and unfolds in a way that mimics natural arm movement

Michael Meng, Dreame

Dreame has made particular strides in innovations focused on getting right into edges and corners. "No matter how powerful a robot vacuum is, dust and debris collect in corners and along edges where brushes simply can't reach. It's one of those problems that sounds minor until you actually look at how dirty those corners get," says Michael Meng.

Its newest bot — the X60 Pro Ultra Complete — has 'UltraExtend technology', which enables the mop pad to kick out up to 7 inches / 18cm from the robot body, and the side sweeper brush to unfurl, E.T.-style, to reach up to 4.7 inches / 12cm.

The design is inspired by human limbs. "When you extend your forearm, your reach increases significantly," Meng explains. "We applied that same principle to our robotic arm design — engineering it to fold and unfold in a way that mimics natural arm movement."

A number of brands have introduced sweepers and mop pads that can kick out to get closer to the edges of rooms — as seen here on the Dreame Matrix10 Ultra (Image credit: Future)

Intertwined with cleaning performance is battery function. "A robot vacuum that runs out of power mid-clean is more frustration than solution," says Meng. "Consumers with larger homes or open-plan layouts need a robot that can complete a full clean in a single charge — or intelligently return to dock, recharge and resume exactly where it left off. Battery efficiency, combined with smart charging and resume functionality, has become a key purchasing consideration for households of all sizes."

Mind the (performance) gap

Even with all the performance advancements, it's still the case that a robot vacuum-mop won't clean as well as a manual vacuum could. Robovacs are winning on convenience, but not on cleaning power. The long-term aim is to catch up. "The first step is to close the cleaning performance gap — reaching, and eventually surpassing, handheld devices," says Faraz from Eufy.

The first step is to close the cleaning performance gap — reaching, and eventually surpassing, handheld devices

Faraz Mehdi, Eufy

Longer-term, he sees things progressing to a point where robot vacuums can tackle not just vacuuming and straightforward mopping, but a wider range of cleaning demands.

"Technologies like roller mopping, steam cleaning, and multi-cyclone systems are maturing," he points out. "In the next five to 10 years, robot vacuums will take on true deep-cleaning capabilities — thorough carpet washing, high-temperature steam disinfection and floor sanitization. The kinds of tasks that currently require dedicated equipment or professional services will gradually fall within the robot vacuum remit."

Docks — like this one for the Roborock Saros 10 — can increasingly take care of all kinds of maintenance tasks (Image credit: Future)

The trump card robot vacuums already hold over manual cleaning appliances is that they take care of the job on their own. It makes sense, then, that continuing to improve automation is a key focus for all the brands I spoke to.

"In the 12 years we have existed as a company, trends have shifted constantly," says Roborock's Ricky Ma. "Currently, the primary trend is 'zero-maintenance and zero human intervention', where users expect a robot to not only vacuum and mop but also self-clean, self-dry, and navigate complex environments without human intervention."

"The biggest challenge is making advanced automation feel completely effortless," adds SwitchBot's co-founder Richard Mou. "Our goal is to reduce the number of times users need to think about their robot."

We've seen consumers become more willing to purchase premium robot vacuums to deliver a genuinely hands-off experience

Reena Patel, iRobot

iRobot's Reena Patel points out that this focus on autonomy is reflected in iRobot's sales patterns. "We've seen consumers become more willing to purchase premium robot vacuums to deliver a genuinely hands-off experience," she tells me. "Consumers are increasingly prioritizing features that save time, such as self-emptying and self-cleaning docks, advanced mopping systems, and intelligent obstacle avoidance."

A big part of the equation when it comes to true autonomy is maintenance. "People increasingly expect higher automation," says Eufy's Faraz Mehdi. "Features like auto-emptying, mop washing, drying, and anti-tangle systems are becoming much more important because users want a robot that genuinely reduces housework, rather than creating a new maintenance routine."

The ultimate aim is robovacs that don't require you to get involved at all — as Mehdi puts it, "zero manual intervention, end to end", including "full-chain self-maintenance, so the user never needs to maintain the machine itself."

Robovacs are arriving with increasingly formidable clean stations that take care of the bulk of the maintenance work (Image credit: Future)

Self-maintenance is one thing, but for a robot to operate without a human chaperone present, it also needs reliable navigation and object avoidance skills. These are essential if it's not going to get lost, stuck or tangled in something.

"Accurate navigation is the foundation of an effective robot vacuum," says Dreame's Michael Meng. "Advanced LiDAR and AI-powered object recognition allow modern robots to avoid obstacles in real time.... Smart robots should not be trapped or stuck easily, so they can work on their own without users to rescue them."

I've never met a robot vacuum that's not desperate to jam itself up by chowing down on any charge cable it can get its wheels on, so here I'd like to put in a personal appeal for more focus on low-level object avoidance.

Time to step up

When it comes to robot vacuums making their way around unassisted, the ultimate hurdle to overcome is stairs. The more premium of the current fleet of bots tend to come kitted out with something akin to small stilts, which raise up the body of the vacuum and enable it to maneuver over tall room thresholds.

However, traversing an actual flight of stairs requires an altogether more involved approach. "It sounds like a simple problem, but for anyone living in a multi-story home, it's one of the biggest limitations of robot vacuums today," Meng notes.

The Eufy MarsWalker is a transporter for moving robot vacuums up or down stairs (Image credit: Future)

As far as I'm aware, there are just three brands who have offered some kind of stair climbing solution, and they're all represented in my pack of interviewees. For Dreame, it's the Cyber X — essentially a stair-climbing pocket into which a robot vacuum can dock when it needs to change floors. Think of it like a stairlift for robovacs. Eufy's MarsWalker takes a similar approach.

Roborock's Saros Rover is rather different, with its two long, spindly legs that enable it to scale not just steps but all kinds of uneven terrain. Ricky Ma sees it as a major breakthrough in bots that don't require human intervention. "The future of this category is exciting and we can't wait to show everyone what we have to offer," he enthuses. I saw the Rover in action at CES 2026, and it was a complete departure from any robot vacuum I've seen before. At that point, it was still at prototype stage — I'll be watching with interest to see how things develop.

Future-gazing

So far, I've been focusing on developments that have already started to take root. But what about the longer term? I asked my interviewees their predictions for what the robovac market might look like in five to 10 years' time, and they had some intriguing answers.

Eufy believes autonomous decision-making is going to change the game. "Robot vacuums will reach a level of intelligence where they can genuinely understand the home environment, identify different types of dirt, and autonomously choose the optimal cleaning strategy," Faraz Mehdi predicts.

Is Roborock's Saros Rover a vision of what robot vacuums might look like in the future? (Image credit: Future)

Roborock is planning along similar lines. "Robovacs are rapidly shifting from an appliance to a smart cleaning partner. The most important shift will be greater physical intelligence," says Ricky Ma. "The ultimate goal is to build an intelligent cleaning partner that can spot the need to clean even before the user has time to ask the product to do so."

Further to this, a few of the execs I interviewed envision about a future where robot vacuums are no longer lone rangers, but work seamlessly alongside other smart appliances.

"Rather than operating as standalone devices, they will become increasingly connected with the wider smart home ecosystem, working more intelligently with other devices," says Dreame's Michael Meng.

Meng also sees a future where robovacs are far from the only autonomous helpers in the home. "While the familiar robot vacuum will remain an important form factor, its capabilities will extend well beyond floor cleaning," he says. "Advances in technologies such as robotic manipulation and embodied AI will enable these robots to interact more naturally with their surroundings and take on a broader range of household tasks."

Robovacs are rapidly shifting from an appliance to a smart cleaning partner

Ricky Ma, Roborock

SwitchBot's Richard Mou has an even more fully formed vision for the future: "I think we'll stop thinking about robovacs as a standalone category and instead see them as part of a broader family of home robots," he tells me.

"As embodied AI and humanoid robotics continue to mature, robot vacuums will contribute proven capabilities like autonomous navigation, environmental perception, and home mapping, while future humanoid robots build on those foundations to take on a much wider range of household tasks. At SwitchBot, that's exactly how we see the future — creating intelligent home robots that work together as one ecosystem rather than isolated devices."

Prepare yourself for the home robot takeover — and for an even cleaner home.

• Ukraine approves indigenous turbojet-powered guided bombs for frontline combat operations
• Domestic guided bombs now strike military objectives up to 200 kilometers away
• Seven Ukrainian manufacturers now build guided bomb conversion systems domestically

Ukraine has officially approved several domestically built guided aerial bombs for combat use, including turbojet-powered versions capable of striking objectives up to 200 kilometers (125 miles) away.

Brig. Gen. Andrii Lebedenko, deputy commander-in-chief of the Armed Forces of Ukraine, said several of these designs have already seen use on the battlefield, while others remain in testing.

“We already have systems that use more than just gliding. They now carry turbojet engines to increase their employment range to between 100 and 200 kilometers,” said Lebedenko.

Ukraine expands the reach of indigenous guided bombs

The program began in 2024 as an effort to convert older unguided aerial bombs into precision-guided weapons after Ukraine lacked domestic manufacturers capable of producing such systems.

Ukrainian officials now say the country has seven manufacturers producing guided bomb systems, compared with none when the initiative was first introduced.

“This task was set in 2024. We did not have any such manufacturers in the country. Now there are already seven. They already know how to do it,” said Lebedenko.

He also confirmed that Ukraine already operates aircraft capable of launching domestically produced weapons, while expanding manufacturing capacity has become the next priority for the defense industry.

According to Ukrainian officials, the program has also created a growing community of engineers capable of designing glide kits and propulsion systems for increasingly sophisticated aerial munitions.

“Moreover, a school of specialists and engineers has emerged who know how to create gliding systems. They know how to make them with engines,” said Lebedenko.

Some of the latest variants are no longer limited to gliding after release because turbojet engines continue powering the bombs toward military objectives at greater stand-off distances.

Indigenous bomb designs move beyond conversion kits

Alongside upgrading older munitions, Ukraine has also developed entirely indigenous guided aerial bombs, expanding its domestic precision strike capability.

Ukraine's Brave1 defense innovation platform announced the country's first indigenous guided aerial bomb after approximately 17 months of development.

The weapon carries a 250-kilogram warhead designed to strike fortified positions, command centers, and other military objectives located dozens of kilometers away.

Developers have described the weapon as a completely original Ukrainian design rather than an adaptation of an existing foreign guided bomb.

Ukrainian military officials have also indicated that improving strike accuracy remains more important than producing larger numbers of guided bombs for operational use.

“We are looking toward precision so that we can effectively destroy enemy command posts, drone control centers, infrastructure, and other targets,” said Lebedenko.

Ukraine has also disclosed expanding manufacturing capacity, growing engineering expertise, and longer operational reach for its guided bomb program.

However, independent verification of battlefield performance remains limited because operational details surrounding combat employment have not been publicly released.

Via United24Media

It’s probably fair to say Meta is facing serious public backlash to its AI glasses.

Musician Lorde took time at a recent concert to tell her audience, “Don’t get the glasses,” many online are branding them “pervert glasses,” and activist group Everyone Hates Elon has put up spoof adverts for the glasses featuring Jeffrey Epstein wearing a pair (note: the story behind a paywall).

I myself have been unsure if I’ll keep using my Meta glasses, despite their utility making them one of my favorite gadgets I’ve tested while working at TechRadar.

Meta has rebutted with new measures to combat misuse of its glasses. It recently patched its specs to prevent modders from using the camera despite disabling the recording light, and Instagram head Adam Mosseri recently announced harassment videos shot using Meta glasses will be banned from the platform — with some high-profile accounts already getting the boot.

Among the folks still in love with the Meta glasses, however, is my wife Izzy.

When I suggested taking a pair each with us on our recent honeymoon to Tenerife, she wasn’t convinced — despite them being our only sunglasses without my prescription lenses in them — but she wore the white-and-gold Oakley HSTN specs nearly all the time we were abroad.

I was also testing out a pair of the Ray-Ban Meta Optics styles for an upcoming review, and time and again the glasses proved themselves useful. Our trip confirmed how awesome these glasses can be when used right, and why a middle ground needs to be found to maximize privacy and utility before things get out of hand.

An all-in-one AI supertool

Of course, the primary reason to rely on Meta glasses is their built-in tech.

Whether we were lounging on a sunbed, strolling through the streets and past the beaches of Los Cristianos, or kicking back with a drink in the evening, we both loved the open-ear speakers that let us enjoy our podcast or playlist of choice without tuning out the world.

For hours on end, I enjoyed Radio 1 through BBC Sounds, while Izzy drifted between Spotify hits, and both pairs lasted a good length of time battery-wise — with touch controls making it super simple to control the tunes when needed.

Izzy looking cool (Taken with my Meta Optics glasses) (Image credit: Future / Hamish Hector)

The AI translations came in clutch several times as well. It offered quick advice on what phrases to use when we realized our limited selection of “Hola,” “Gracias,” and “Por favor” wouldn’t cut the mustard.

Meanwhile, Look and Ask helped Izzy decipher which of the unrecognizable shampoos and conditioners would be best for her hair after a travel mishap saw the supply we packed leak into her travel bag. The AI could read the Spanish labels to her to help identify and filter out ingredients she knows to avoid, and we ended up buying a few extra bottles of the conditioner she found to bring home after she fell in love with it.

It can't be understated how useful the smart specs are simply as sunglasses too.

My Optics pair features Transitions XTRActive lenses, transitioning fast from clear to shaded as I moved between indoor and outdoor spaces, which was super handy for someone like me who wears glasses all the time. Meanwhile, the polarized gold HSTN lenses were perfect for Izzy as she spent hours of her day baking in high-UV weather.

Thanks to the charging case, my specs rarely ran out of battery, but even when they did, I could still rely on them just fine as stylish shades.

(Image credit: Oakley / Meta)

Of course, the controversial aspect is the camera. Beyond the AI assistance, it was helpful for snapping quick photos and first-person clips for the travel vlog I’m looking to cut together.

Though whenever I went to snap a photo, I had flashes of the story about Meta contractors who claimed to have seen intimate moments captured through the specs. Shots of either of us in swimwear went nowhere near the glasses’ snapper, and I kept a careful eye on who was in the background of any shots I did take.

We were also a little worried that someone might not like us using the glasses at all in their presence, though thankfully no one took any issue with us.

These fears and considerations do create points of friction when it comes to using Meta’s glasses—or any smart glasses, for that matter. While the Facebook company is the best-known for smart eyewear right now it is far from alone, and while it's making efforts to clean up its public perception smart glasses will be tarred with the same brush if any brand steps out of line.

I also don’t think Meta’s efforts are perfect so far, at least not yet. I’d like to see more work across social and hardware platforms to root out smart glasses misuse; I believe better, more thorough privacy rules are necessary to make it crystal clear what data is and isn’t shared with Meta and its staff, and I’m not convinced I could trust it with the immense amount of data its rumored super sense feature would capture from my daily life — blunders like its Instagram AI image use plan and backtrack don’t fill me with confidence in its AI decision making.

At the same time, my week falling back in love with the Meta glasses, and my wife loving her Oakley smart specs, proves there’s something really fun and useful about the tech that I equally don’t want to see taken away or banned outright.

(Image credit: Future / Hamish Hector)

With other brands joining in on intelligent eyewear — recently Samsung demonstrated its own pairs at Galaxy Unpacked — it’s clear the tech industry is storming ahead, just as it has with exciting emerging technologies in the past.

I just hope a middle ground can be found, because smart specs represent a serious privacy nightmare beyond other gadgets we’ve relied on. If the tech doesn’t hold itself to account, we could see even more regulation and public backlash than we’ve seen already, and that could kill off my favorite tech sector of recent years before it has a chance to thrive.

Confusion is the typical first reaction. They hold the new Samsung Galaxy Z Fold 8 in their hands, marveling at the passport-sized cover screen, and ask without irony or sarcasm, "Is this new?"

I'm finding that people think the Z Fold 8's design is both familiar and completely alien. They roll it over in their hands and, initially, appear unaware that it also unfolds to reveal a far larger, and yet also oddly shaped, inner display.

It's, for all intents and purposes, the Nokia Feature phone of modern foldables. It's a completely new device featuring modern hardware, components, tech innovation, and AI that also somehow harkens back to an era of weird Nokia phones that proudly peacocked their oddball aesthetics.

Back in the early oughts when most thought the state of cellphones had been well-worked out and we'd settled on small-ish screens and an endless array of keyboard styles, Nokia was king of the hill. There was no one-size-fits-all in the world of cell phones. Your phone's physical design was an expression of your personal style. The tiny Nokia 31000 had virtually nothing in common with a T-Mobile Sidekick or a BlackBerry 7230. Nokia's designs, though, were fun (if not always practical — looking at you, Nokia N-Gage) and they made us think differently about mobile phones and what we should expect when using them.

The Samsung Galaxy Z Fold 8, which Samsung unveiled last week in London at Samsung Upacked (and that I'm carrying today), is perhaps serving a similar function now, trumpeting the very existence of foldable phones and why they might matter to you.

Why the Nokia Effect mattersLance Ulanoff / FutureLance Ulanoff / FutureLance Ulanoff / FutureLance Ulanoff / Future

In my discussions with Samsung about why they made the Galaxy Z Fold 8, it became quite clear that this was a phone and form factor designed to attract customers disinterested or dissatisfied with the current relatively vanilla set of folding phone options, not only available from Samsung, but also from its rivals at Motorola, Oppo, and Huawei.

After all, you had either the candybar style that opens into a mini tablet or the communicator flip-style. Camera arrays might all look different, but the basic size and shapes are all the same. And none of them were moving the market — foldables held just 1.6% of the global smartphone market in 2025. None of them appear to be making people stop and think. There was, until now, no Nokia Effect.

The Samsung Galaxy Z Fold 8, though, is fun to show off. I've seen people stop, smile, scratch their heads, and be generally interested in a way they weren't with recent Folds and Flips or even the new Galaxy Z Fold 8 Ultra I'm also carrying around. One coworker called it "dreamy". How often have you heard someone describe a smartphone like that?

They notice, for instance, that in its folded size, the Galaxy Z Fold 8 is a much better fit for smaller hands, and how it actually fits all the way in a pocket (not for nothing, the 201g size doesn't hurt for portability, either). Even opened and in portrait mode, the 7.6-inch display is more palmable than the Z Fold 8 Ultra's 8-inch display.

Weird and wonderful

Nokia always went wild on design (Image credit: Lance Ulanoff)

Its inherent cuteness is an unintentional nod to the phones of a bygone era. When folded, the cover display and even the back with its smaller camera array say, "Look at me, I'm adorable."

Samsung will be leaning into this, and where the Z Fold 8 Ultra is a mobile multi-tasking and content capture and creation workhorse, the Z Fold 8 is a friend who tells you stories in a compact form. It's a lean-back reading experience, and a less letterboxed full-screen video viewing companion.

And yet, it's also a better confidant, letting you share your thoughts and texts in a virtual keyboard that's wider and more comfortable to use than one on even the Galaxy S26 Ultra or iPhone 17 Pro Max.

What Samsung has almost miraculously done with the Galaxy Z Fold 8 is create a foldable that will finally get people talking. Further, it's done so months before Apple enters the foldable ring. Imagine that, a knockout punch delivered by Samsung before the fight begins.

It won't be that easy, of course; I fully expect Apple to deliver something truly extra on the foldable front. It should share this unusual form factor, especially since Samsung is probably supplying the display. Samsung has told us that it's unconcerned about Apple's entry and actually welcomes it. Maybe that's just pole-position talk.

Will the freshness of this form factor evaporate when the iPhone Ultra arrives? Maybe, or maybe it'll be too late for Apple because surely Samsung has already cracked the code. All it had to do was take a look back and remind itself why people loved weird cellphone designs and how that might, if not shape, at least inform its expansion of the Z Fold line.

• Meta has funded multiple natural gas power plants over the past 12 months
• The corporate renewable energy initiative RE100 includes Apple, Google, and Microsoft among its 444 members
• Meta has previously committed to running all operations on renewables by 2020

Meta has confirmed it is withdrawing from the RE100 scheme, following the news that it has funded several gas-fired power stations over the past year.

With the company leaning heavily into AI and the necessary data centers that come with such a commitment, it has opted to exit the collective of corporations planning to transition to 100% renewable energy, overturning a public commitment made in 2020.

While RE100 has many other members, including Apple, Google, and Microsoft, Meta’s confirmation of a move away from the group indicates that it doesn’t see an immediate solution to the power challenges posed by data centers.

What does Climate Group membership mean?

The RE100 initiative is led by the Climate Group, a non-profit with former UK prime minister Tony Blair named as a co-founder. Membership of the project gives organizations access to policy support and technical advice to support plans for moving to a 100% renewable energy footprint.

When quizzed, a Climate Group spokesperson responded: “After several in-depth conversations between Meta and Climate Group, Meta has withdrawn from the RE100 initiative, as it is no longer able to meet the technical criteria due to investments made in new gas power.”

Meta isn’t alone in looking beyond the renewable options for supplying power to its data centers. Its former RE100 co-members Google and Microsoft have both recently invested in fossil fuel projects. However, those companies both appear to have hedged their bets more widely, with Microsoft aiming for hourly energy matching (where fossil use is matched by renewable energy) and Google employing a 100-hour battery in a collaboration with Form Energy.

Data centers need heavy industry levels of power

Moving away from commitments to renewables in order to provide data centers with the level of energy they need in the AI era is disappointing, but not entirely unexpected.

Modern data centers require power for processing, for cooling, and for backup, and this requires a level of power that traditionally has only been seen in heavy manufacturing industries. The infrastructure currently available for renewable sources of power does not widely account for supply issues, battery systems are underdeveloped, and nuclear takes years to commission.

Meta’s support for gas power includes partnering with Entergy Corp in building seven plants in Louisiana in order to deliver over 5GW for its Richland Parish data center.

Conversely, gas-powered electricity is faster to set up, is a known quantity, and is cleaner than coal. Until infrastructure catches up with the demands from data centers, more withdrawals and fudges on previous green energy commitments from big tech should not be ruled out.

AV Insider

AV Insider is our new series of interviews with influential people inside the AV industry. From execs to the people behind the technology, every Friday we'll bring you a new perspective on world of TV and audio.

See the full list of AV Insider articles

Deezer has never been backwards about coming forwards regarding the scourge of AI slop infecting our music streaming platforms. Just days ago, the French streaming site announced that over half of all new daily uploads to its site are AI — up from 44% in April and just over 30% at the end of last year. So, things are only getting worse.

Luckily, the platform not only implements its own proprietary filter to find, label and in some cases purge AI tracks from its site; it also recently made the tech available to everyone, irrespective of your chosen streaming platform.

But how was Deezer's own in-house AI filter developed, why is Deezer so keen on tackling AI tracks, and what does the company think the future of music streaming will look like — you know, for actual humans still trying to make it their career?

I spoke with Manuel Moussallam, Deezer's Head of Research, and Jesper Wendel, the company's Head of Communications, to get their takes on all this and more.

Moussallam agreed to speak with me on the morning after the company's summer party, held at a music venue in Paris. As he was "Checking if the hangover is showing on my face", I couldn't help but think it's good to know that proper live gigs and parties still matter to a firm whose success hangs on our love of music.

(Image credit: Deezer / TechRadar)Identifying an issue (back in 2021)

How long has Deezer's research arm been thinking "Okay, this AI thing is an issue, and we have to do something about it"? Moussallam explains that the site's comprehensive AI filter, now available to anyone who wants it, has been almost five years in the making.

"It all started in scientific conferences back in later 2021 or early 2022, because the generative models that were already working for images and videos were starting to make real progress — like real scientific progress — for music," he says. "That was probably a couple of years before the first commercial services were appearing.

"So, since we participate a lot in these scientific conferences, we saw people from Google and from Meta sharing these foundation models that were starting to work quite well for creating whole pieces of music. Not just snippets of 10 seconds like before, but really whole songs. So I guessed that eventually, if you're able to generate a lot of music, you can also distribute it quite easily — and down the road, it would end up in our catalogs."

‘We made some mathematical discoveries on how AI models work, and realized that we didn't actually need a big AI to detect AI songs’

Manuel Moussallam

"I would say we started working on the project in 2022, first with some complicated methods, because that's unfortunately a reflex for people like me — to try to do AI to capture AI — but then eventually we simplified a lot because we made some mathematical discoveries on how those models work.

"We realized that we didn't actually need a big AI to understand and to detect these pieces. Just by relying on some plain signal processing and mathematical tools, we were able to actually detect these artifacts for most of the models out there.

"Yeah, I think we deployed the first version of the detector in September 2023 — and we communicated the first numbers when we were able to scale it up to the catalog a few months later."

(Image credit: Deezer)A big problem — and it's only getting worse

I wonder why, given that various rival music streaming sites have implemented 'Transparency Tags' (read: optional AI badges that place the onus on distributors and labels to tell listeners that what they're listening to was not written or performed by a human), Deezer chose to create a comprehensive filter — and then to release it to all? While admirable and commendable, that's a lot of R&D budget on AI, no?

Moussallam laughs. "There are many aspects, but I guess firstly, we did it just because we wanted to see if we were able to do it. On my team, we do a lot of audio analysis. So, knowing was the first motivation."

‘I guess firstly, we did it just because we wanted to see if we were able to’

Manuel Moussallam

"But then, I think since Suno, Udio, and all these tools are trained on unlicensed catalogs, I was pretty convinced that eventually we would have to take down all this music, because it would be considered as plagiarism or copyright infringement in some sense — which may be the case if the legal battles end up that way.

"So at some point we thought that, okay, if people agree with us that it's kind of unlawful to train this model with all this music and they ask us to remove all the content from Suno in the catalog, we need to be prepared. We need to be able to do it. So, that was another big motivation.

"It didn't turn out like that — for now — but it may be that one day people tell us, 'OK, everything that was generated with that AI model prior to this date in the future, when people actually have agreements and compensate musicians for their work, all of this needs to be removed'. And we would have to be prepared for that."

‘One day, people may tell us, “OK, everything that was generated with that AI model prior to this date… all of this needs to be removed” — and we would have to be prepared for that.’

Manuel Mousallam

Putting information into the hands of people who need to make decisions

I mention that across all of my audio streaming subscriptions (not a flex, I simply review audio kit for a living), I tried the Deezer AI filter, and none of my playlists contained AI, bar one: my Tidal catalog listed 1% AI content — and to know what the pesky tracks are I'd need to switch to Deezer. I ask what the reception has been since the release of the software for all.

"Mostly people are happy with the tool, I think everyone just wants to know," replied Moussallam.

"Ever since we started deploying this detector, we've been talking to major labels, obviously, but also indie labels, sometimes small publishers. All of them just wanted to know. Some of them made some pretty interesting discoveries on how their members were using — or actually not using — these technologies; so overall it's really about putting this information into the hands of people who need to have it to make some decisions.

"Implementing some transparency for our users was also a main motivation for us to flag the content on the website. And then we realized that people had to come to Deezer to check an album, say, to see if it had the label. And we said, 'Okay, we can also create a tool just for you to check — you will see for yourself how contaminated your whole library is on other platforms".

(Image credit: Future)To purge or not to purge, based on AI markers?

Speaking to Moussallam even briefly, it become clear that he's a scientist as well as a musician (among his many accolades is a Master's Degree in Acoustics, Signal processing and Computer science applied to Music, from the Université Pierre et Marie Curie).

So when I ask how this proprietary new filter works to pick out 100% AI-generated content — which I assume is fully text-prompted music — and whether Moussallam and his team are looking to take it further, to stuff that's only 50% AI generated, say, things get more serious.

"It's a very complicated question," Moussallam concedes. "Right now, we can't really say that we only detect fully AI-generated songs, because what we detect is the presence of markers of AI generation in a song, and we detect the strength of these markers in the song.

"When we flag an album, it's because these markers are strong enough for us to be confident that it's been AI generated at some point.

"But we don't know if you wrote the lyrics for yourself, for instance. We also don't know if the song contains very few elements of human creation; that's not something that our detector is able to do right now — but that's the focus of our current research.

"So we really want to be able to boil down to the stem level and be able to make a decision based on a stem by stem case, because obviously we're seeing a lot of hybrid cases, and a lot of artists are also integrating AI tools into their creative process.

"So, these very strong markers of AI, we're starting to see them spread in human creations, and we need to be able to disentangle that from what you mentioned, like purely text-prompted music curation".

(Image credit: Deezer)

Can the markers the filter picks up recognize whether it has flagged an AI vocal, say, or artificially generated instruments — and how about stolen melodies?

"No, no, we can't go that far," admits Moussallam. "I mean, we are working on the stem by stem detection, but right now we detect markers — and we are not so sure where they come from."

He smiles. "I mean, we have ongoing research, but not something that I can publicly share. When it comes to detecting chord progressions or melodic impressions or plagiarism metrics, we are not so interested in that, because it would lead us too far. And I mean, there are so many songs that share only four chords! I think that debate is way, way out of our hands in any case."

What's next at Deezer — and can AI and humans coexist happily in music?

Perhaps surprizingly for someone who's spent nearly five years building the tech to flag and potentially quash AI music tracks, when I ask Moussallam where he thinks Deezer and online music will be in 10 years time — and whether he'd like to see all flagged AI tracks gone from the site — he instead builds a quite compelling case for computation in music.

"Well, the very first piece of music that was written by a computer dates back to 1953 I think, back in the days where computers were actually whole buildings built around them.

"Ever since people built computers, they tried to make music with computers. I think that's the story of the 20th century. I'm confident people are going to make great things with these new creative tools. It enables some stuff, and I guess as long as it's being used by humans to create cool music for other humans to listen to, we are very fine with that.

‘Ever since people built computers, they tried to make music with computers; I think that's the story of the 20th century’

Manuel Moussallam

"What we're really fighting at Deezer is automated music generation, people who script computers to generate songs — I mean thousands of songs — and uploading them in batches to streaming platform to try to overwhelm and flood and basically really replace human creation with automated creation.

"So, I guess I hope music will stay artisan, as we say in French, so something that humans do to have the attention of other humans. That's really what I guess drives our motivation. If the music includes computers, that won't be a such a revolution — computers are fine, numerical instruments are fine — it's just who is using them to create that matters, in the end."

Jesper Wendell is keen to hone in on the fraudulent aspect of AI slop and the payment dilution for for artists, because that's also one of the motivations for stifling the recent development of the AI-in-music game.

‘If the music includes computers, that won't be a such a revolution — computers are fine, numerical instruments are fine — it's who is using them to create that matters, in the end’

Manuel Moussallam

(Image credit: Deezer)

"Fraud has always been there, it's just easier now with AI music," he says. "And if you upload thousands of tracks, then you can sort of spread that fraud across more and more tracks — and that's something that we want to get rid of as much as possible".

I mention that from the perspective of the consumer (i.e., me), the key annoyance is tracks making it into our recommended playlists and ultimately usurping streaming royalties — small amounts maybe, but still — which we'd prefer would go to a band that we actually like.

Moussallam emphatically agrees. "That’s really something we wanted to avoid. Anyone listening to AI music without knowing that it's AI music, like just feeding it to people in a recommendation? That's something that was very clear for us from the beginning we wanted to avoid. Because that's an awful feeling, I think for everyone".

‘Fraud has always been there, it's just easier with AI, and if you upload thousands of tracks, you can spread that fraud across more and more songs’

Jesper Wendel

Given that Tidal only recently drew its anti-AI line in the sand, Bandcamp's strong and concise stance was set out in January of this year, and Spotify's Verified by Spotify badge certifies that an artist is human but doesn't help filter out the slop from any generated playlists, I think it's safe to say that Deezer is leading the charge here.

Based on Reddit threads I see every day, I think I speak on behalf of thousands of music lovers when I say we're grateful for everything the Deezer team is doing to combat the AI infestation trying to take hold on our streaming sites. Where do we go from here? For now, trust in Moussallam's work.

I didn't think a second screen would change how I work until I actually used one. For the last few weeks, I've been running the Plugable USB-C Portable Monitor alongside my laptop, and it's quietly become the accessory I'm most annoyed to leave at home.

Not because it's flashy, but because it solves a problem I'd stopped noticing I had: everything I do, from research with three tabs open to editing a draft while a spreadsheet sits in the background, is easier with two screens instead of one.

Why a second screen changes more than you'd expect

(Image credit: Future)

The real benefit of a second monitor isn't extra pixels—it's just relief from overlapping windows. With the Plugable next to my laptop, Slack and email stay parked on one screen while I focus on the other. I tab-switch way less, which keeps me from constantly losing my spot throughout the workday.

It's a 15.6-inch 1080p IPS screen, 60Hz, anti-glare coating. Nothing exotic on paper. But I've had it wedged at some awkward angles trying to fit it on cramped desks, and it still looks right, no washed-out colors like some of the cheaper portable screens I've tried. 300 nits of brightness.

That's fine at a desk or on the couch, but take it outside on a sunny afternoon and you'll be squinting at your own reflection. Didn't expect this part: it does HDCP, so Netflix and the rest just work instead of giving you that annoying black screen some portable monitors throw when they hit copy protection.

Plugable Portable Monitor

The Plugable USB-C Portable Monitor is a 15.6-inch 1080p IPS display built to turn any laptop into a two-screen setup with a single cable. It handles both video and up to 100W of pass-through power, so it can charge your laptop while running the screen, provided you're plugging into a video-capable USB-C port (DisplayPort Alt Mode, USB4, or Thunderbolt). Two additional 10Gbps USB-C ports double it as a hub for external drives, and it also works with USB-C iPads, Android devices, and iPhone 15 and later. At 1.85 pounds with a folding cover that doubles as a stand, it's built for travel and hot-desking rather than color-critical or gaming work. All told, it's a straightforward way to add a real second display anywhere you can find a laptop and an outlet.

Read our Plugable Portable Monitor review.

One cable does everything, if it's the right port

(Image credit: Future)

The setup is genuinely just one cable. Plug the included USB-C cable from the monitor into a video-capable USB-C port on your laptop, and the screen lights up immediately- no drivers, no software to install.

That single cable also carries power in the other direction: the monitor can pass through up to 100W from a connected charger, using 15W to run itself and sending the remaining 85W on to power your laptop. In practice, that means I run one cable from the wall to the monitor, and one cable from the monitor to my laptop, and both the screen and my laptop's battery are taken care of at once.

The important caveat, and the one thing that catches people out, is that this only works over a USB-C port that actually supports video, meaning DisplayPort Alt Mode, USB4, or Thunderbolt. Plenty of laptops have USB-C ports that only handle charging or data, and plugging into one of those gets you a black screen and nothing else.

It's worth checking your specific laptop's port specs before assuming any USB-C port will do the job. The same logic applies to cables: the one included with the monitor handles video, power, and data over its 3.3-foot length.

But if you want a longer cable for a more comfortable desk setup, you need one explicitly rated for DP Alt Mode plus Power Delivery plus at least 10Gbps data, since plenty of USB-C cables that look identical are charge-only and won't carry a picture at all.

Built-in USB-C hub functionality

This is the part I didn't expect to actually use. Beyond the main port handling power and video, there are two more 10Gbps USB-C ports built right into the monitor. So when I need to pull footage off an external SSD, I just plug it into the monitor instead of digging out a separate hub or hogging a port on the laptop itself.

It also isn't limited to laptops. The monitor works with USB-C iPads and Android phones with video-capable ports, and with iPhone 15 and later. The one thing to know is that most phones and tablets can't power the screen on their own.

So, if you're driving the monitor from a phone, you'll want a USB-C power adapter connected to the monitor's pass-through port to keep both the screen and your device running smoothly. Laptops, by contrast, can usually power it without any extra adapter.

Where it fits (and where it doesn't)

It weighs 1.85 pounds, measures roughly an inch thick, and the protective cover doubles as a stand — making it simple to pack and deploy anywhere. Color accuracy and refresh rate aren't suited for heavy media work or gaming, but as an everyday secondary display for work, it does the job effortlessly.

I went into this expecting a nice-to-have and ended up with something I genuinely plan trips around. That's a weird thing to say about a monitor, but here we are. It's not going to wow anyone with specs, and it's definitely not the display you want if your job involves color grading or fast-paced gaming.

What it's actually good at is the boring stuff: giving you a second screen with zero setup friction, wherever you happen to be working that day. Hotel desk, kitchen table, coffee shop with one free outlet, doesn't matter. Plug in the right port, and it just works. At $199 in the US, it's not an impulse buy, but if you've ever tried to get real work done on a single 13-inch laptop screen, you already know what that's worth.

For more top-performers, see our guide to the best portable monitors we've tested.

Even without being directly tied to The Big Bang Theory, there's an incredibly satisfying reason for why fans should invest in new HBO Max spinoff Stuart Fails to Save the Universe: how returnable it is.

By that, I mean the comedy series could easily run for five, 10, or even a million more seasons and not dip in quality or have any continuity issues. Because there's an infinite number of universes to explore, Stuart Fails to Save the Universe could be taken in any number of zany directions and still make complete sense.

In an age where high-quality TV shows are often being axed before they've been in streamable orbit for more than a month, the idea of this alone is refreshing. However, I do have an ulterior motive for wanting Chuck Lorre, Bill Prady and Zak Penn's latest creation to last for as long as possible.

Given enough time, I think that Stuart Fails to Save the Universe could course-correct the biggest mistake that the overarching world it spins off from has ever made, affecting both TBBT and subsequent spinoff Young Sheldon before anybody realized it was a huge issue. Frankly, it's a grudge I've held onto tightly for the last two years (at best).

So when I had the chance to interview the creative trio for their new HBO Max comedy, I wasted no time pitching my story idea for future seasons. I knew it was solid, but I wasn't expecting them to genuinely consider it.

The pitch: bring George Cooper back to life

The biggest mistake TBBT ever made was killing off George Cooper before he was even a realized character in Young Sheldon. In fact, the decision was so poorly judged that Lorre publicly apologized for it after George's funeral in Young Sheldon season 7.

Let me explain. When we first meet adult Sheldon (Jim Parsons) in TBBT, we learn that his father had died of a heart attack when he was a teenager, and was described as mean, absent, an adulterer and sometimes abusive alcoholic. The less said about him the better, put it that way.

But by the time Young Sheldon took off in 2017, we learned another version of Sheldon's (Iain Armitage) truth. For most of his childhood in Medford, Texas, George (Lance Barber) did everything he could to try and understand a child that seemed so alien to him. Even George's alleged affair was disproved, turning out to be Sheldon's mother Mary (Zoe Perry) in a wig.

George took Sheldon for his first-ever visit to his future academic home CalTech, drove him to NASA's headquarters just so Sheldon could prove their math was wrong, and let him talk about scientific methods beyond George's comprehension for hours on end simply because it made Sheldon happy. If you ask me, he was an incredible dad.

But the fact that George would have to die around the time Sheldon was 14 was hanging over our heads from the moment the prequel spinoff was realized. I don't think anybody expected Young Sheldon to become the gargantuan success that it was, and by the time we got to season 7, the inevitable happened.

Episodes 12-14 of season 7 were dedicated to unpacking George's death, which happened off-screen while he was coaching football at the local high school. It's something that destroyed the rest of the Cooper family for good, with follow-up spinoff Georgie & Mandy's First Marriage continuing to help explore how Mary, Missy (Raegan Revord) and Georgie (Montana Jordan) all ended up becoming the people we meet in TBBT years later.

To me, George's death never actually needed to happen in Young Sheldon. Enough distance had been created between the prequel and TBBT to ignore adult Sheldon's vision of his childhood altogether, even if it couldn't be officially rectified. Worst case, they could have waited until season 7 had finished before implying that the death might have happened at some point between Young Sheldon and Georgie & Mandy's First Marriage.

But instead, I sobbed through Mary breaking down after learning her husband had died, and cried my way through his funeral the next week (watching live at 1am UK time, which was a bad decision). I was so distraught and frustrated by the team's choices that I even wrote George a fake obituary, which actually led to me interviewing Barber himself.

Now that Stuart Fails to Save the Universe has come along, we've got a unique chance to change history for the better.

'The universe you're looking for is out there — whether it makes it to TV is another question'

So here's the idea: by the time we get many seasons into Stuart Fails to Save the Universe, Stuart will likely know how to control multiverse travel a whole lot better than he does at the moment. Logically, he could then visit a universe where George was never killed and voilà... TBBT's costly mistake is put behind us indefinitely.

If Parsons eventually has a cameo in the show too, all the better. There's no such thing as the "right" version of reality here... any one of them we come across could be where the story anchors down for good. So why shouldn't a universe fans are actually happy with be chosen in that case? It's a no-brainer decision, if you ask me.

"Let's discuss this when we're doing those future seasons," Lorre tells me, like I've just stormed into the HBO Max head office and requested a legitimate pitch meeting. "One thing you need to know is that we're not really thinking that far ahead… we're just trying to solve whatever is right in front of us."

"We'll definitely keep your ideas in mind," Penn follows up with. "There are an infinite number of universes within this universe, so the universe that you're looking for is going to exist by definition. Whether it makes it to TV is another question."

"Actually, there's a mathematical issue with that," Prady adds. "The fact that there's an infinite number of universes doesn't guarantee the existence of a particular universe."

I tell the trio that this isn't the positive mindset that I'm looking for, but even as they laugh at my unhinged fan plan delivered with so much gusto that it could have broken even Sheldon's brain, I'm convinced that I've planted an important seed.

Stuart Fails to Save the Universe being renewed for more seasons feels like a given at this point, even if HBO Max hasn't officially confirmed anything as of writing. But the confidence with which the trio talk about the future suggests that we won't be done with the multiverse fun after 10 episodes.

So remember these words if you ever see George Cooper have a surprise cameo in the series, and remember who made the retcon magic happen. You're welcome, in advance.

• Nvidia and Microsoft lead the list of founder members of the Open Secure AI Alliance
• AI, financial services, tech companies, and other organizations with developing AI services have joined the alliance
• Notable exceptions include OpenAI, Anthropic, and Google

Nvidia has confirmed its membership of the new Open Secure AI Alliance, a collective of industry-leading tech, financial, and AI businesses working in collaboration to ensure openness in the AI industry. Curiously, however, while Microsoft and OpenClaw are among the founding members, Google, OpenAI, and Anthropic are not.

Inspired by the open source movement and the Linux Foundation-led Akrites AI cyber threat response initiative, the Open Secure AI Alliance aims to use open technologies to disclose and remediate vulnerabilities.

Citing the recent Hugging Face incident, where closed AI tools prevented full investigation of the event, the Open Secure AI Alliance founding announcement underlines the importance of open models, highlighting how cybersecurity is one of the main beneficiaries of open source software.

A Hugging Face response?

(Image credit: Open Secure AI Alliance)

A sizeable group of recognizable companies are involved with the Open Secure AI Alliance, including OpenClaw, Palantir, Palo Alto Networks, Microsoft, Crowdstrike, IBM, Linux Foundation and Hugging Face.

Notable from this list are Google and Palantir, and of course OpenAI, which was responsible for the Hugging Face attack. Given that this event occurred during a cybersecurity benchmark test and OpenAI’s model conducted a wholly automated attack on Hugging Face’s servers, it serves as an important demonstration of what is at risk from closed source models in the AI world.

“When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most,” states the Open Secure AI Alliance, which emphasizes its mission: “to ensure defenders everywhere have open, frontier tools they can trust and control.”

OpenAI’s lack of openness in dealing with the incident, and Hugging Face’s initial difficulties in analyzing the intrusion due to closed source models, really underline the importance of this new alliance.

Open vs. closed

The argument concerning open and closed software models isn’t likely to end any time soon, but the Open Secure AI Alliance certainly makes a strong point for open models. Yes, they can be a risk that can be modified or misused, but those risks exist within closed systems.

So, the alliance’s view is to give defenders the tools to deal with threats, rather than blocking those tools. Decades of cybersecurity research have shown, after all, that the best option is collaboration and testing – or, as the Open Secure AI Alliance states “the safer path is the one that gives more defenders the ability to test, verify and strengthen the systems on which society relies.”

Given OpenAI’s role in the Hugging Face incident, its absence makes sense – but what about Google and Anthropic? At a time when corporate collaboration is driving development of new technologies within tight standards (for example, the Connectivity Standards Alliance’s work with Matter and Aliro), it seems as though the Open Secure AI Alliance has the right idea.

One of the biggest problems with book recommendations is that they're usually too obvious. I don't need another list of books to read after The Lord of the Rings, or someone telling me to try Brandon Sanderson because I like fantasy. I wanted recommendations based on the strange mixture of books I actually enjoy — ones that felt personal, not algorithmic.

So I gave ChatGPT my bookshelf. Metaphorically, at least.

Rather than asking for recommendations straight away, I told ChatGPT to learn my reading taste first. I started listing favorite authors and books, then asked it to quiz me about others I'd forgotten. It wanted to know what I'd enjoyed about particular novels, whether I'd read similar authors, and even the rough timeline of when I'd discovered them, building a picture of the books that had shaped me.

I also gave it some ground rules. It should avoid obvious recommendations unless there was a compelling reason to include them. Every suggestion had to be explained in relation to something I'd already read, even if that connection was simply, "This is nothing like your usual books, but I think you'll love it."

After about half an hour, ChatGPT stopped asking questions and started analyzing me instead.

"Your shelves suggest that you like speculative fiction with a sense of play," it said. "You are drawn to books with elaborate worlds, but you do not seem especially impressed by complexity for its own sake. Humor matters, although you tend to prefer humor that reveals something about the characters or the society around them."

It wasn't a perfect summary, but it was close enough to make me think this experiment might actually work.

(Image credit: Getty Images / VCG)Literary profiling

The obvious appeal of feeding ChatGPT a full reading history is that it can spot patterns across hundreds of books at once. I could have described my taste as fantasy, science fiction and comedy, but that would have been far too broad to produce anything useful. ChatGPT noticed that I repeatedly chose books about bureaucratic absurdity, unreliable institutions, strange cities and reluctant heroes who would much rather be somewhere else.

It also noticed my fondness for stories that treat big ideas lightly without treating them as trivial. That led it toward Martha Wells’ Murderbot Diaries, which pair sharp comedy with questions about identity, autonomy and the exhausting burden of dealing with humans. I had already read them, which was mildly disappointing but also reassuring. The system had identified exactly the sort of thing I wanted.

When I told it Murderbot was already familiar territory, it adjusted rather than simply replacing one title with another popular series.

“You appear to like characters who stand slightly outside their own societies and comment on the absurdity around them,” it replied. “I will move away from well-known sarcastic narrators and look for books where the humor comes from social observation, institutional failure or characters trying to remain sensible in deeply unreasonable worlds.”

That shift produced better surprises like The Gone-Away World by Nick Harkaway and The City of Dreaming Books by Walter Moers for its combination of literary obsession, elaborate worldbuilding and gleeful weirdness. It suggested The Dragon Waiting by John M. Ford because I seemed to enjoy alternate histories that trusted the reader to keep up. It also pointed me toward Diana Wynne Jones’ adult novels, noting her lighter touch and sharp understanding of human foolishness.

The recommendations became more convincing when ChatGPT explained what each book might lack. One novel had the humor but less warmth. Another had brilliant worldbuilding but moved slowly. A third matched my interest in satire but was considerably darker than most of the books I had marked as favorites.

Library AI

The experiment improved once I began disagreeing with it. One recommendation leaned too heavily into grim fantasy, a genre I can enjoy in small doses but rarely seek out for relaxation. Another featured a long military campaign, which is usually the point where my attention begins quietly packing a suitcase. Each correction sharpened the next round.

One of its most intriguing suggestions was QualityLand by Marc-Uwe Kling, a satirical science fiction novel. The recommendation came with a warning that the satire was broader and more direct than some of my favorites but that the subject matter fit my interest in technology and systems going wrong in very organized ways.

There were still misses. ChatGPT occasionally became too eager to prove it had discovered a pattern, linking two books because they both contained libraries or because their protagonists were technically immortal. At one point it recommended something almost entirely because it featured a sarcastic demon, which felt less like literary analysis and more like the work of an intern who had skimmed the dust jacket.

Even so, the overall experience was far better than typing “funny fantasy books” into a search bar. And I now have a pretty good reading list for the next few years. My bookshelf had always contained this information. ChatGPT simply read the evidence more patiently than I had.

Imagine a world where the most powerful weapon isn't a missile, but a software update. In less than 72 hours, a software engineer can patch a drone on the front line and turn an enemy's multi-million-dollar electronic warfare system into little more than expensive scrap.

This isn't science fiction but an everyday reality inside Ukraine’s real-time military tech pipeline. Driven by the necessity of national survival, a decentralized network of coders, startup founders, and makers has bypassed decades of slow defense bureaucracy.

At the heart of this transformation is Brave1, Ukraine's defense innovation engine, where software developers, startups, soldiers, and investors collaborate at startup speed to solve battlefield problems. In the newly launched Brave1 Market, an ecommerce-style procurement catalog, combat success earns digital "ePoints," public rankings fuel competition, and rewards are reinvested into even more powerful technology.

It’s no surprise that Brave1’s success has attracted global attention, with aerospace giant Airbus partnering with the platform to connect aerospace expertise with this next-gen defense ecosystem. Also backed by big-data titan Palantir, the new Brave1 Dataroom acts as a secure data pipeline, streaming raw battlefield video and thermal imagery directly to developers training AI targeting models.

Now, we can forget all about old-school defense programs and endless procurement cycles. The future of warfare is open source, software-defined, and moving at the warp speed of a Silicon Valley startup.

Why traditional military tech procurement is failing

The traditional model of military procurement is running on outdated code. For over half a century, the Western defense industry has been building bigger, better, and increasingly expensive "exquisite beasts" - fighter jets, aircraft carriers, and heavy armor, each taking a decade to design and deploy.

Under this system, the pipeline is painfully slow. Governments can spend years defining specifications, years selecting contractors, and years more building the hardware.

By the time it reaches the battlefield, the software inside is often two decades out of date and locked behind proprietary code that can't be modified without years of legal renegotiations. As The Wall Street Journal recently reported, traditional defense structures struggle to absorb the rapid pace of software and startup-led innovation.

This peacetime bureaucracy is breaking under the speed of modern warfare. Today's battlefield is defined less by firepower than software, where an overnight update can render even the most sophisticated missile system obsolete. Legacy defense structures simply can’t cut it for software-speed innovation. They prioritize caution and consensus, while modern defense tech is designed for speed and to survive battlefield surprises.

At its core, the old defense playbook presumes weapons are built once and fielded for decades later. But when a new battlefield threat emerges, waiting a year or two for a budget committee to approve a software patch is a recipe for defeat.

Modern warfare demands systems that can evolve every day. By decoupling software from hardware, these defense ecosystems empower thousands of developers to adapt faster than any centralized procurement committee ever could. At the forefront of this transformation is Brave1, Ukraine's open source tech cluster.

What is Brave1?

If traditional defense procurement is a labyrinth of government departments and endless paperwork, Brave1 is the system built to bypass it. Co-founded by multiple Ukrainian ministries, including Digital Transformation, Defense, and Strategic Industries, this platform gives anyone with a laptop and a promising idea a path to the battlefield. It brings developers, startups, investors, and soldiers into a continuous deployment cycle.

In the old-school defense world, getting a new, innovative idea in front of the right people takes months. Brave1 acts as a secure, digitized buffer zone where developers upload their designs or code, pass automated and expert reviews, and connect directly to the problems facing Ukraine’s military.

Once a project clears Brave1’s gateway, it receives an official security rating, gets access to the centralized technology registry, and can scale research and development (R&D) grants reaching up to eight million UAH to fast-track production.

However, this high-speed model comes with its own set of challenges. When hundreds of teams are sprinting to solve battlefield problems, some will run along parallel paths, spreading funding and talent across versions of similar technology. The result is a risk of fragmented resources spread across an ecosystem built for speed and experimentation.

Still, by swapping endless military paperwork for the automated Brave1 Market, the platform ensures that promising ideas don't get stuck in red tape.

Inside Brave1’s expanding innovation network

The speed at which Brave1 has grown turns traditional defense economics entirely on its head. What started as a bold innovation hub has evolved into a full-scale defense technology engine. Today, Brave1 brings together more than 3,200 registered companies and actively tracks over 4,500 products moving through its secure development pipeline.

Yet innovations don't win wars unless they can be implemented. Rather than relying on sluggish defense grant cycles, Brave1 has awarded 500 direct innovation grants totaling over $11 million, putting funding directly into the hands of frontline engineers and hardware startups. This pipeline is designed to bridge the notorious hardware "valley of death," where many promising technologies traditionally run out of money.

(Image credit: Brave1)The technology verticals that are reshaping the front line

Instead of focusing on billion-dollar missiles or heavy armored vehicles, Brave1 targets technologies that are reshaping modern warfare. Its product catalog spans a wide scope of specialized, software-driven technology verticals, including:

• Unmanned aerial vehicles (UAVs): As we write, more than 1,000 manufacturers are building next-generation drones, ranging from AI-powered attack drones and long-range reconnaissance systems to fiber-optic-guided drones that bypass conventional radio jamming.
• Naval and autonomous sea drones: More than 50 manufacturers are developing uncrewed surface vessels (USVs) that use autonomy and swarm tactics to challenge conventional naval superiority at a fraction of the cost.
• Automated electronic warfare (EW and SIGINT): Over 350 companies are building compact electronic warfare systems, software-defined jammers, and smart spectrum analyzers that can adapt to hostile drone frequencies on the fly.
• AI and computer vision products: More than 300 software-centric companies are building the digital brains behind modern warfare, from intelligent targeting systems and automated image recognition to swarm coordination and AI-assisted decision-making.
• Robotic ground complexes (UGVs): Over 280 manufacturers are building autonomous ground robots that take on the battlefield's most dangerous tasks, from delivering supplies and evacuating the wounded to laying mines and other frontline operations.

However, the ecosystem's biggest advantage is also one of its technical hurdles. Thousands of independent systems must work seamlessly together, on the same battlefield, requiring constant software integration so drones, electronic warfare systems, and ground robots can form a synchronized network.

The outstandingly open ecosystem

Unlike traditional defense programs built around closed systems and proprietary codebases, Brave1 embraces an open ecosystem framework inspired by the world of open-source software.

It’s no secret that traditional defense innovation is built around extreme secrecy, where companies protect their intellectual property at the cost of collective advancements. Brave1 takes a completely different course by encouraging developers to share code, 3D-printing schematics, hardware blueprints, and engineering know-how across their networks.

The open architecture is already going global. The Ukrainian government officially passed the Brave International framework to open up the ecosystem to allied defense partners. Brave France, developed alongside the French Defence Innovation Agency (AID), pairs joint funding with battlefield testing through Test in Ukraine, fast-tracking tech innovations from lab to field.

This openness comes with a cost, and it’s a bigger cyber battlefield. Every new developer, repository, and software dependency increases the number of potential attack vectors, making zero trust a network a must. Now, rather than targeting finished weapons, sophisticated attackers focus on the software supply chain that builds them. In an ecosystem like Brave1, even a single compromised software package could cascade through countless downstream systems before a security gap is identified.

The app store model for the front line

If Brave1 is the military’s revolutionary software stack, capital is the processing power that drives the code.

Financing innovation at the speed of war

To understand how Brave1 moves with the speed of a Silicon Valley startup, we first have to follow the money. In traditional defense networks, getting an innovation grant involves a multi-month marathon of paperwork, compliance, and risk assessment. For tech startups, this creates an administrative bottleneck better known as the "valley of death," where promising ideas collapse before they ever reach the battlefield.

Brave1 bypasses this hurdle with a decentralized, non-dilutive funding model built for wartime speed. Developers submit applications through a unified digital interface, where an inter-ministerial panel evaluates each project's potential battlefield impact. Successful applicants receive early-stage funding within weeks, allowing them to spend their time on engineering rather than filling out paperwork.

Points, rankings, and the gamification of warfare

Once a prototype is approved, it enters the Brave1 Market, a secure digital marketplace that works much like an enterprise app store for the military. Through the recently launched Buyer's Cabinet, verified commanders can browse a catalog of vetted domestic technologies, compare specifications and battlefield performance, and purchase equipment directly through the platform, dramatically shortening the journey from prototype to deployment.

Brave1 Market is bound up with Ukraine's points-based battlefield system dubbed the ePoints system, which applies game mechanics to battlefield breakthroughs. Frontline units earn these points for verified combat results, and every confirmed strike against enemy assets is recorded through military situational awareness systems. These actions feed into an active, real-time Call of Duty-style leaderboard run by the Unmanned Systems Forces, which allows onlookers to track top drone units via the military's official online killboard.

Afterward, units can redeem their accumulated ePoints as an internal digital currency directly inside the marketplace to purchase newer, more powerful drone hardware or electronic warfare (EW) kits. This type of decentralized purchasing power bypasses top-down supply chains altogether, allowing commanders to fill their digital shopping carts with the exact tools they need based on real-world performance data.

However, the system's greatest strength, its flexibility, also creates one of its biggest pain points. When individual military units have the freedom to buy specialized equipment from hundreds of small suppliers, standardizing spare parts, software support, and maintenance protocols across the entire military infrastructure becomes a serious operational challenge.

(Image credit: Brave1)The 72-hour software sprint

The main mechanism of this software-defined ecosystem is its split-second adaptability. In the world of EW, signal-jamming frequencies can change overnight, rendering entire drone fleets ineffective. Updating a weapon's electronic systems requires months of contractor negotiations, approval processes, and engineering changes, which is much slower than the speed at which modern threats evolve.

Inside the Brave1 network, this logistical nightmare is treated as a fast-paced development sprint:

1. Telemetry capture: When a drone at the front line detects a new jamming frequency, it automatically records the hostile signal pattern, creating data for future countermeasures.
2. Data pipeline: Raw battlefield data is transferred from frontline systems into centralized developer environments, where engineers can analyze threats and come up with next-gen countermeasures.
3. Patch deployment: As soon as a new threat is identified, engineers can adjust the software, updating radio-frequency configuration or terminal tracking code, without swapping out any hardware.
4. Over-the-air update: The updated software patch is compiled and pushed right back to the frontline within 72 hours.

While it all sounds beautiful on paper, when you move code from a lab into a muddy trench, things can get messy. As you can already guess, the real bottleneck isn’t the tech - it’s the human factor. Constantly updating software requires frontline operators to learn new configurations, radio profiles, and interfaces in high-stress situations. Requiring soldiers to run updates in the middle of a mission creates an enormous margin for error, showing that even the slickest software must ultimately work within human limits.

Battlefield beta testing

To shorten the path from prototype to deployment, Brave1 relies on the official Test in Ukraine program. Instead of spending years in laboratory simulations to clear traditional safety benchmarks, promising prototypes undergo swift safety checks before moving into live combat, turning battlefield feedback into the next immediate upgrade.

For instance, if an optical tracker fails in heavy dust or a carbon-fiber chassis breaks under stress, developers receive real-time frontline feedback. What the process gains in speed, however, it sacrifices in long-term reliability testing.

Training the AI: The combat data factory

Algorithms mean little without a continuous stream of raw, real-world data to train them.

The battlefield as an AI training ground

On the modern software-defined battlefield, algorithm accuracy shapes survival. If code is king, then raw data is the electricity that powers the throne. To feed this infrastructure, Brave1 functions less like a traditional defense bureau and more like a colossal combat data factory. Every hour of flight telemetry, electronic warfare logs, and automated targeting footage flows continuously from the battlefield into secure engineering nodes, giving developers a steady stream of real-world data.

Together, this creates a continuous machine learning (ML) pipeline, with each mission generating new data for the next software update. According to the Ukrainian Ministry of Defense, over 100 Ukrainian companies are training AI models using the Brave1 Dataroom. At its core is a secure repository that gives developers access to structured visual and thermal data on aerial threats, all captured under real battlefield conditions.

On top of that, as highlighted by The New York Times, Ukraine has opened access to millions of drone videos and extensive battlefield telemetry, allowing both domestic developers and allied partners to train and refine new technologies using real data.

The sheer volume of this dataset is staggering. DefenseScoop reports that more than half a million hours of battlefield footage are being used to train and refine AI target-recognition algorithms. Brave1's AI models are trained on conditions that are almost impossible to recreate in a laboratory.

While Western commercial tech labs are forced to train terminal guidance AI on clean, synthetic data simulations, Brave1 developers feed their convolutional neural networks a steady diet of gritty, real footage with battlefield smoke, physical camouflage, and chaotic weather conditions. This massive collection of battlefield data helps train AI models to identify armored movements, track targets through dense terrain, and maintain navigation even when GPS signals and communications links are disrupted.

The silicon bottleneck on the edge

However, trying to turn the battlespace into a living supercomputer comes with severe physical constraints. The long-term objective is the deployment of automated systems that can guide hardware to targets without a human pilot or continuous communications link.

Outside the front line, running sophisticated computer vision systems means relying on massive power-hungry cloud data centers or stacks of high-end enterprise GPUs. On the battlefield, we don't have that luxury. Tech developers must shrink these massive AI models into something small enough to run on low-power, edge-computing chips and clamped onto lightweight platforms operating at the front line.

Active combat zone is the supreme stress test for AI. Models must swap complexity for raw survival, learning to work with blurry cameras, damaged sensors, and broken connections. The smartest system on paper is useless if it stops working the second the electronic environment gets ugly.

This centralized intelligence advantage doubles as a single point of failure. A repository with battlefield datasets, AI models, and software configurations from numerous defense companies is an awfully attractive target for cyber adversaries. Protecting that ecosystem demands strict zero-trust security, as a single slip-up can expose the entire network.

Bipedal robots and autonomous systems

To take humans out of danger fields, developers are looking past basic drones to deploy agile, multi-terrain robotic platforms.

Brave1 humanoid robot program and the reality of trench AI

Brave1's latest frontier is humanoid robotics. By launching a dedicated grant competition, Brave1 has formally established armed bipedal robots as their own defense technology vertical. The mission is to create near-human machines that can traverse trenches, transport supplies, and clear high-risk zones while keeping troops out of harm's way.

However, early tactical edge trials suggest that Hollywood-style humanoid robots remain a long way from battlefield reality. Prototypes struggled to carry more than 20 kilograms, offered slim protection against severe weather, and saw battery levels bleed out rapidly during demanding missions. While tech labs design robots to shuffle across flat factory floors, real combat carries an unstructured nightmare of mud, debris, and steep ditches.

In contrast, low-profile wheeled and tracked UGVs have already completed more than 50,000 frontline logistics missions, offering a cheaper and sturdier alternative. To survive the field, tech developers have chosen simplicity over sophistication.

Ruggedized software vs clean labs

Building AI for the battlefield requires abandoning many of the assumptions of traditional software engineering. In a traditional clean lab tech stack, AI models rely on high-performance cloud infrastructure and high-quality data streams. On the front line, software must be streamlined to run locally on low-cost edge-computing chips clamped onto lightweight, plastic platforms.

This optimization is non-negotiable when looking at the sheer volume of hardware hitting the production line. In fact, the scale is so massive now that Ukraine's defense production has hit a staggering 10 million drones annually, with plans to double that to 20 million. When building at this unprecedented multi-million-unit scale, your code must be lightweight enough to run smoothly on millions of cheap, disposable devices.

Meanwhile, developers must build software that persists through blurred imagery, glitchy sensors, and dead air. A model that depends on perfect data or constant network connectivity can quickly become useless once it leaves the lab.

Ultimately, Silicon Valley can continue building software for data centers, but the battlefield must have software built for sheer survival.

Automated software countermeasures

One of the top technical hurdles for humanoid robots isn't mobility but maintaining stable communications. Near the ground, radio signals are weakened by terrain, vegetation, and other obstacles. For a simple, two-legged platform, however, even a slight interruption can disrupt its balance, causing the entire machine to collapse.

So, what's the solution? Not building a better radio tower in the middle of a battlefield, but moving decision-making closer to it. That’s why Brave1 is pushing deterministic, edge-computing autonomy.

If the command link is lost, onboard AI instantly takes over. Using local computer vision, the system can navigate, avoid obstacles, and continue its mission without relying on GPS or a live operator link.

The evolution of electronic warfare

Modern electronic warfare isn't about broadcasting the strongest signal. In fact, that would be a good way to hang a giant "shoot here" sign over your own position.

Dating back to the Cold War and the post-conflict era, electronic warfare relied on powerful vehicle-mounted jammers that flooded broad sections of the radio spectrum. In contemporary conflicts, those high-emission systems have become increasingly vulnerable, as their massive radio frequency (RF) emissions can reveal their coordinates to enemy sensors and radar-seeking weapons. Rather than flooding the airwaves with raw power, modern electronic warfare depends on compact, software-driven systems that can adapt swiftly while remaining difficult to detect.

Through Brave1, developers are shifting toward compact, trench-level electronic warfare systems. Instead of constantly broadcasting powerful radio signals, these software-defined radio (SDR) platforms silently monitor the local spectrum for hostile activity. As soon as they detect an incoming threat, they transmit a brief, targeted jamming burst on the specific frequency being used, disrupting the attack while remaining hidden.

Democratizing defense production

By taking manufacturing out of rigid defense factories, Brave1’s ecosystem has turned local tech talent into frontline defense developers.

The tech talent workforce

The real breakthrough isn't all about tech - it’s fundamentally human. Instead of relying exclusively on career defense engineers, Brave1’s ecosystem draws talent from across the commercial technology sector. Software developers, UX designers, data engineers, and automation experts are applying the same skills used to build consumer apps and cloud platforms to next-generation defense technology. This subversive talent shift is bringing commercial software thinking into a field that has traditionally been shaped by slow, hardware-driven development cycles.

Now, a team that previously spent months fine-tuning logistics for a commercial delivery app can shift to building software layers that coordinate autonomous drone fleets in a matter of weeks. This talent shift is opening the door for a new generation of software engineers to reshape how defense technology is built, not just what gets built.

Yet, bringing commercial software talent into defense creates a whole new set of engineering challenges. Silicon Valley coders may still be new to physical hardware constraints, extreme environmental stress factors, or advanced ballistic mathematics.

Overcoming this challenge requires continuous collaboration between two different worlds: the software developers building the systems and the frontline experts testing them in the heat of battle.

Distributed development networks

The old-school defense factories are a tempting target: a single structure, a single strike, and systemic collapse. To protect production from long-range missile strikes, Ukraine had to move away from that factory model. Now, production has been broken apart into a distributed network of smaller facilities operating across the country. Despite their digital connectivity, these workshops function more like a peer-to-peer network than a traditional industrial empire.

The main strength of this model is its robust resilience against single points of failure. While a precision strike can damage individual production hubs, it can’t cause a cascading collapse across the whole network. Like with a digital ecosystem, production simply reroutes around broken nodes.

Still, the primary challenge remains as clear - standardizing such networks is harder than standardizing factories. Maintaining consistent quality control across hundreds of small-scale workshops calls for ongoing supervision, as even minor discrepancies in suppliers, parts, or production techniques can cause unexpected failures in the field. What performs perfectly on a test bench in Kyiv might fail during deployment simply because a single workshop used a slightly different batch of soldering wire.

To keep this under control, the Brave1 ecosystem must rely on constant monitoring and software-based diagnostics capable of scanning the whole network and flagging issues before things start to break.

The open-source playbook

The real breakthrough is not simply that Brave1 opens the market to more competitors, but that it changes the very mechanism of how defense technology evolves. Borrowing from the world of open-source software, the platform treats each tech improvement as shared building blocks that can be refined, adapted, and deployed across the wider network.

Code, 3D designs, and hardware fixes become part of a shared, ever-evolving knowledgebase rather than closely guarded company assets. So, if one team tweaks a flight controller, upgrades a component, or finds a better way to build, that progress is instantly shared and used by everyone across the ecosystem.

This approach also changes the ways Ukraine studies and replicates enemy technology. In mid-June 2026, the Ministry of Defense launched the TrophyLab platform, a secure platform built to study and catalog captured Russian military equipment. Through it, approved international partners, defense organizations, and contractors can access technical documentation, electronic warfare vulnerability reports, and telemetry analysis for over 115 captured weapons.

Instead of letting captured weapons gather dust as classified secrets, TrophyLab turns them into an open sandbox for building new defenses together. On the other hand, moving this fast can also scale up systemic errors. A flawed software update, compromised dependency, or hidden vulnerability in a shared repository could spread across multiple platforms before developers spot the problem.

In an open, fast-paced network, the real battle isn't just building cooler tech but proving that every single update is completely secure.

Software startups vs slow bureaucracy

Before code can even expect to rewrite the rules of modern warfare, it must first break through decades of slow, legacy bureaucracy.

Why Western government programs are slow

Beyond the battlefield, software is also reshaping defense bureaucracy. Many Western procurement systems are crippled by years-long bottlenecks and constrained by rigid rules originally drafted for industrial-era hardware like warships, fighter jets, and military bases. While this framework makes sense for concrete and steel, it struggles to keep pace with software that evolves from week to week.

By the time an agency reviews a tech requirement, writes a request, and clears legal review, the tech landscape has already shifted. This rigid setup lets legacy, prime contractors dominate simply because they possess the paperwork army, not because they build the best software. The downside to this stability is a systemic delay in sending state-of-the-art commercial tech to the field, which widens the innovation gap between the military and private tech markets.

Brave1’s flexible, software-first approach stands in stark contrast to old bureaucracy. Borrowing from the workflow of a technology startup, Brave1 is designed to evaluate ideas swiftly, fund promising prototypes, and connect developers directly with the front lines. The focus shifts from predicting tomorrow's battlefield to adapting to it as it changes.

The rise of venture-backed defense tech

Recognizing these bureaucratic bottlenecks, venture capital has begun flowing into defense technology at a breakneck speed. For decades, the math for defense tech simply didn't work for private capital, suppressed by sluggish procurement cycles, multi-year product runways, and heavy ethical baggage. In today's markets, the rise of adaptive, software-driven defense companies is changing that perception, turning the defense tech industry into one of the fastest-growing areas of deep-tech investment.

In line with this trend, Germany has also begun rethinking how it funds defense innovation. As reported by Reuters, Berlin is deploying a new state-backed investment vehicle to inject capital directly into defense startups. The objective is to de-risk defense innovation while bypassing agonizingly slow acquisition cycles.

Private investors are now backing startups building everything from AI-powered software to edge computing hardware to next-gen autonomous systems. Increased venture backing allows smaller companies to fast-track their initial product deployment. They can perfect new technologies without relying on slow legacy defense acquisition pipelines.

Yet market forces do not always align with military priorities. Traditional private capital prioritizes high-velocity software growth and massive markets. Meanwhile, defense forces require rugged, custom-built hardware built for niche military needs - a tech stack with zero civilian upside. The result? Well, vital, battlefield-ready technology often gets sidelined in favor of whatever scales fastest.

(Image credit: Brave1)Big tech partnerships: Brave France and Brave Germany

To bridge the gap between fast-moving tech startups and rigid military institutions, the ecosystem has shifted, with nations now turning to formal bilateral frameworks. Ukraine’s new Brave International framework bridges this financial gap with over €100 million in joint funding. This allows agile international startups to bypass slow procurement and fast-track their tech directly to the battlefield. Under this blueprint, Ukraine and its partners share costs via a 50/50 funding split. Joint, parity-based expert boards review applications to scale up sister programs like UNITE Brave NATO, Brave Norway, and Brave Lithuania.

Brave France is the first major spin-off to go live under this playbook. Finalized at the Eurosatory defense exhibition, the Brave France Bilateral Grant Program unlocks a €20 million fund built specifically to bypass bureaucratic procurement bottlenecks. This pipeline hands out €1 million per project to de-risk co-developed missile tech, robotics, and next-gen air defense systems. The first call for projects goes live this September.

At the same time, Berlin is spinning up its own Brave Germany track to tackle critical front-line hardware shortages. Signed in Kyiv, the Brave Germany agreement pumps direct capital into specialized battlefield hardware such as laser systems and secure tactical communications. Most notably, the framework fast-tracks the production of 5,000 AI-powered strike drones. Crucially, the pact covers the co-development of strategic long-range systems capable of reaching up to 1,500 kilometers, allowing allied tech companies to validate their prototypes directly in live combat via the "Test in Ukraine" loop.

The business logic of low-cost tech

Before a software update can rewrite the rules of combat, it must first change the core economic equation of the modern battlefield.

The asymmetric cost equation

To understand why software-defined warfare is reshaping modern conflict, we have to start with the economics. For decades, conventional military strategy was created under the assumption that a sophisticated battle tank required an equally sophisticated (and expensive) anti-tank missile system. It was a high-stakes arena where both offense and defense demanded multi-million-dollar investments to get their hardware to the starting line.

Ecosystems like Brave1 have torn up the traditional defense playbook. By weaponizing cheap commercial tech, they give agile startups the ability to neutralize multimillion-dollar systems at a fraction of the cost. Today, a standard off-the-shelf racing drone can be modified with a basic 3D-printed payload mechanism and a $50 onboard AI microchip, and all of it costs around $500 to assemble. Yet, guided by smart edge-computing algorithms that throwaway piece of plastic gets the precision to seek and destroy an armored vehicle worth as much as $5 million.

This staggering economic asymmetry flips the logic of attrition warfare on its head. Why spend millions on heavy, legacy hardware when low-cost, disposable tech can destroy it too?

Well, this hyper-cheap approach has its own structural headaches. Sourced from commercial supply chains rather than defense contractors, these tools sacrifice military-grade certification and risk resembling an early-stage Kickstarter project. A cheap capacitor might tap out the moment it encounters a brisk autumn breeze. In this arena, you trade hardware perfection for the pure math of a statistical zerg rush.

National survival over corporate profits

Traditional defense giants operate much like bureaucratic mega-corporations, obsessing over shareholder returns and safeguarding their proprietary tech. Their business models are built around stability and product lifecycles that stretch over decades.

Wartime innovation ecosystems like Brave1 trade corporate profit targets for immediate frontline deployment. This type of urgency rewards rapid prototyping and open-source collaboration rather than multi-year development loops and locked software.

While this agile approach speeds up innovation during trying times, it leaves deep-tech startups with seriously thin financial safety nets. When the immediate crisis cools down, many of these narrow-margin startups may struggle to keep their engineers paid or scale into mature defense companies.

The new playbook for global enterprise tech

The massive ripples from Brave1 reach far beyond the mud of the front line, handing a brand new playbook to deep-tech manufacturers worldwide. By proving that you can build, patch, and scale seriously complex physical networks using decentralized, open-source code, this pipeline has cracked the code on absolute agility. It proves that during a chaotic crisis, the most lethal asset in your toolkit isn’t a shiny yet rigid hardware product but an adaptable software infrastructure that can change on a dime.

The future of enterprise technology is no longer trapped within the squeaky clean labs of isolated corporate ivory towers. The companies that are going to rule the next decade, whether building next-gen defense hardware or streamlining global shipping routes, will be those that successfully mimic this framework. They will swap stiff, top-down corporate hierarchies for open developer portals and trade sluggish multi-month updates for continuous 72-hour coding sprints. To put it simply, they will realize that raw, real-world data will always stomp all over laboratory theory.

Now, we can officially say goodbye to old-school development manuals and endless peacetime planning committees. Tomorrow's winning companies will stop treating business like a catalog of static products, choosing instead to run their operations like a living software ecosystem that evolves the second the environment shifts.

Tracking the innovation: How to follow Brave1

Brave1 moves at hyper-speed, spinning up new grant tracks and global pipelines in weeks instead of years. Thankfully, this explosive evolution is heavily documented, so we can track Ukraine's defense tech disruption in real time:

If you are tracking Brave1, watch these channels:

• Brave1 Official Portal: The primary hub for active grant tracks, tech challenges, and developer registration.
• Ministry of Digital Transformation: The best source for the high-level government policies shaping the dual-use tech ecosystem.
• Brave1 on X (Twitter): The fastest source for frontline product launches, hackathons, and real-time field test videos.
• Brave1 Market: The "app store for the military" marketplace where frontline units buy verified hardware directly from builders.
• TrophyLab Portal: A secure research platform used by allied engineers to analyze captured hardware and electronic warfare data.

In the end, this brand-new blueprint proves that the old corporate playbook is as good as dead, and global tech firms must learn to move fast or get left behind.

The VPN industry has a problem. For a significant number of people, their VPN subscription has quietly auto-renewed, jumped in price, or cost more than they initially bargained for.

The issue is so bad that multiple VPN providers — including ExpressVPN, NordVPN and Surfshark — have faced legal scrutiny over their auto-renewal practices.

This creates a dilemma for us at TechRadar when it comes to recommending VPNs. We are confident these are the best VPNs available — they are the fastest, most secure, and best at streaming — but it's clear that more needs to be done to ensure fair and transparent billing.

To get a better understanding of people's real-world experience using our top-rated VPNs, we analyzed almost 30,000 Android VPN reviews across the 'Big 4' — NordVPN, ExpressVPN, Surfshark & Proton VPN — published on the Play Store since the beginning of the year.

The results are stark. Across the entire Play Store dataset, people are mostly happy with the apps — just 35% of overall written comments are negative. But when we looked at billing and pricing specifically, that figure rose to almost 60%.

Read on to find out which provider performed the best, which you might want to be wary of, and the key practical steps to avoid common VPN issues.

This article is the first in a series investigating real-world customer experience of using major VPN services, inspired by our exclusive analysis of user-generated Android VPN reviews.

Which VPN has the most positive reviews?

The vast majority of written Android reviews offer little insight into how people really feel. There are thousands of generic entries like "top app," "good," "ok," or "bad."

However, once we filtered out the noise, billing and pricing issues emerged as a significant area of frustration, accounting for almost 30% of all categorized feedback.

Because total review counts varied widely between brands, we compared percentages rather than raw numbers. Attitudes were scored using a machine-learning model supported by manual human checks.

Read more about our methodology here.

When it comes to billing and price satisfaction, Proton VPN is the clear winner. Just 35% of billing-related comments were negative. However, that’s in large part because of the free tier it offers.

In fact, if you remove references to the product being 'free,' the rate of negative comments rises to 57%. While that points to significant underlying friction for paid accounts, it’s still better than the rest.

By contrast, the remaining market leaders face significant dissatisfaction:

• NordVPN: 83% negative billing sentiment
• ExpressVPN: 79% negative billing sentiment
• Surfshark: 71% negative billing sentiment

It wasn’t all bad news, though. There was some positive feedback, with one user praising Surfshark for its "outstanding value for money" and another calling NordVPN the "best affordable" VPN. But for the majority of reviewers, issues around free trials, auto-renewals and price hikes dominated.

Of course, all written reviews skew towards negative emotions as consumers seek to fix or change something. However, taken together they demonstrate a comprehensive picture with clear issues at play.

Free trials, auto-renewals, and price hikes

Across all of the reviews analyzed, three distinct issues appeared.

Firstly, it’s clear many people are experiencing issues with free trials. Specifically, people are signing up only to find themselves charged automatically.

As one Surfshark user wrote: “Wanted to try using the free trial, [it] didn’t work and still charged me for a month.” Meanwhile, a NordVPN user reported not being able to “cancel my free trial just days in.”

Another source of friction occurs when automatic renewals are triggered. And it impacts all of the providers mentioned. As one Proton VPN reviewer wrote: "A lot of users would appreciate more transparency before renewals happen.”

Closely related to the automated renewals are the price hikes that are associated with them. Even if you sign up for a certain price for a year, VPN providers often use more expensive rates at the point of renewing a user’s subscription.

One ExpressVPN user wrote: "Of course every time they renewed my subscription they used old prices and never told me about it.” Meanwhile, a Surfshark reviewer put it even more succinctly for others: “Beware of subscription auto renewal.”

@import url('https://fonts.googleapis.com/css2?family=Open+Sans:wght@400;700&display=swap');@import url('https://fonts.googleapis.com/css2?family=Poppins:wght@400;700&display=swap');#fv-chart-1785251267479-g474gdird *, #fv-chart-1785251267479-g474gdird *:before, #fv-chart-1785251267479-g474gdird *:after, #fv-chart-1785251267479-g474gdird-slideshow *, #fv-chart-1785251267479-g474gdird-slideshow *:before, #fv-chart-1785251267479-g474gdird-slideshow *:after {box-sizing: border-box !important; margin: 0; padding: 0; border: 0;font-size: 100%; font: inherit; vertical-align: baseline;}#fv-chart-1785251267479-g474gdird, #fv-chart-1785251267479-g474gdird-slideshow { font-family: 'Open Sans', sans-serif !important; }#fv-chart-1785251267479-g474gdird {position: relative !important; overflow: visible !important;--riv-primary: #2E6E93;}#fv-chart-1785251267479-g474gdird .fv-inner-wrapper {color: #1F2937 !important; background-color: #ffffff !important;padding: 1.5rem 1.5rem 2rem !important; border-radius: 0.5rem !important;box-shadow: 0 1px 3px rgba(0, 0, 0, 0.1) !important;margin: 1rem 0 !important; display: flex !important; flex-direction: column !important;overflow: hidden !important;}#fv-chart-1785251267479-g474gdird .fv-inner-wrapper.fv-no-header.fv-is-image-compare {padding-top: 0 !important;}#fv-chart-1785251267479-g474gdird.fv-full-bleed {width: 100vw !important;margin-left: calc(50% - 50vw) !important;}body {overflow-x: clip !important;}#fv-chart-1785251267479-g474gdird.fv-full-bleed .fv-inner-wrapper {padding: 0 !important;border-radius: 0 !important;box-shadow: none !important;margin: 0 !important;background-color: transparent !important;}#fv-chart-1785251267479-g474gdird .fv-inner-wrapper.fv-is-shop-the-look {padding: 0 !important;border-radius: 0 !important;box-shadow: none !important;margin: 0 !important;background-color: transparent !important;}#fv-chart-1785251267479-g474gdird-slideshow {position: relative !important;width: 100% !important;margin: 1rem 0 !important;--riv-primary: #2E6E93;}#fv-chart-1785251267479-g474gdird-slideshow .fv-slides-wrapper {position: relative !important;width: 100% !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-slide {width: 100% !important;animation: fv-fade-in 0.3s ease-in-out;}@keyframes fv-fade-in {from { opacity: 0; }to { opacity: 1; }}#fv-chart-1785251267479-g474gdird-slideshow .fv-slideshow-nav-row {position: relative !important;display: flex !important;justify-content: space-between !important;align-items: center !important;padding: 0 0 16px 0 !important;width: 100% !important;z-index: 20 !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-nav-btn {background-color: var(--riv-primary) !important;color: #ffffff !important;border: none !important;border-radius: 4px !important;padding: 8px 16px !important;font-size: 14px !important;font-weight: 700 !important;cursor: pointer !important;display: flex !important;align-items: center !important;justify-content: center !important;gap: 6px !important;transition: opacity 0.2s, background-color 0.2s !important;height: 36px !important;text-transform: none !important;box-shadow: 0 1px 2px rgba(0,0,0,0.1) !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-nav-btn svg {width: 18px !important;height: 18px !important;stroke-width: 3px !important;filter: none !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-nav-btn:hover {opacity: 0.9 !important;transform: translateY(-1px) !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-nav-btn.disabled {background-color: #E5E7EB !important;color: #9CA3AF !important;cursor: default !important;pointer-events: none !important;box-shadow: none !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-slide-counter {font-family: 'Poppins', sans-serif !important;font-size: 14px !important;font-weight: 600 !important;color: #374151 !important;text-align: center !important;min-width: 40px !important;background-color: rgba(255,255,255,0.8) !important;padding: 2px 8px !important;border-radius: 10px !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-slideshow-select {position: absolute !important;top: 10px !important;right: 10px !important;z-index: 20 !important;appearance: none !important;-webkit-appearance: none !important;-moz-appearance: none !important;background-color: white !important;border: 1px solid #d1d5db !important;color: #1F2937 !important;font-family: 'Open Sans', sans-serif !important;font-size: 14px !important;font-weight: 600 !important;padding: 6px 32px 6px 12px !important;border-radius: 4px !important;cursor: pointer !important;box-shadow: 0 1px 2px rgba(0,0,0,0.05) !important;background-image: url("data:image/svg+xml,%3csvg xmlns='http://www.w3.org/2000/svg' fill='none' viewBox='0 0 20 20'%3e%3cpath stroke='%236b7280' stroke-linecap='round' stroke-linejoin='round' stroke-width='1.5' d='M6 8l4 4 4-4'/%3e%3c/svg%3e") !important;background-position: right 0.5rem center !important;background-repeat: no-repeat !important;background-size: 1.5em 1.5em !important;}#fv-chart-1785251267479-g474gdird-slideshow .fv-slideshow-select:focus {outline: 2px solid #2E6E93 !important;border-color: #2E6E93 !important;}#fv-chart-1785251267479-g474gdird .fv-chart-title {font-weight: bold !important;text-align: center !important;margin-bottom: 0.5rem !important;color: var(--riv-primary) !important;font-size: 20px !important;line-height: 1.2 !important;font-family: 'Open Sans', sans-serif !important;text-transform: none !important;white-space: normal !important;overflow-wrap: break-word !important;padding: 0 20px !important;}#fv-chart-1785251267479-g474gdird .fv-chart-subhead {font-size: 18px !important;font-weight: 500 !important;text-align: center !important;margin-bottom: 2rem !important;color: #374151 !important;line-height: 1.7 !important;font-family: 'Open Sans', sans-serif !important;display: block !important;text-transform: none !important;padding: 0 20px !important;}#fv-chart-1785251267479-g474gdird .rv-chart-caption { font-size: 15px !important; color: #374151 !important; text-align: center !important; font-style: normal !important; font-weight: normal !important; line-height: 1.7 !important; font-family: 'Open Sans', sans-serif !important; display: block !important; }#fv-chart-1785251267479-g474gdird .fv-versus-chart { display: flex; flex-direction: column; width: 100%; margin-top: 1rem; }#fv-chart-1785251267479-g474gdird .fv-versus-header { display: flex; justify-content: space-between; align-items: center; margin-bottom: 1.5rem; padding: 0 1rem; }#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper { flex: 1; min-width: 0; }#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-left { text-align: center; padding-right: 1rem; }#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-right { text-align: center; padding-left: 1rem; }#fv-chart-1785251267479-g474gdird .fv-versus-select-container { position: relative; display: inline-block; max-width: 100%; width: 100%; }#fv-chart-1785251267479-g474gdird .fv-versus-chevron { position: absolute; top: 50%; transform: translateY(-50%); pointer-events: none; width: 16px; height: 16px; flex-shrink: 0; }#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-left .fv-versus-chevron { right: 0; }#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-right .fv-versus-chevron { right: 0; }#fv-chart-1785251267479-g474gdird .fv-versus-select { background: transparent; border: none; border-bottom: 2px solid; font-family: 'Poppins', sans-serif; font-weight: 700; font-size: 14px; padding: 0.25rem 0; cursor: pointer; outline: none; appearance: none; -webkit-appearance: none; -moz-appearance: none; max-width: 100%; width: 100%; text-overflow: ellipsis; overflow: hidden; white-space: nowrap; }#fv-chart-1785251267479-g474gdird .fv-versus-select.fv-select-left { text-align: center; direction: ltr; padding-right: 1.25rem; }#fv-chart-1785251267479-g474gdird .fv-versus-select.fv-select-right { text-align: center; padding-right: 1.25rem; }#fv-chart-1785251267479-g474gdird .fv-versus-select option { font-family: 'Open Sans', sans-serif; font-weight: 400; font-size: 14px; color: #374151; direction: ltr; text-align: left; }#fv-chart-1785251267479-g474gdird .fv-versus-vs { font-family: 'Poppins', sans-serif; font-weight: 700; font-size: 14px; color: #374151; letter-spacing: 0.1em; padding: 0 1rem; }#fv-chart-1785251267479-g474gdird .fv-versus-body { display: flex; flex-direction: column; gap: 1.5rem; }#fv-chart-1785251267479-g474gdird .fv-versus-row { position: relative; height: auto; padding-top: 20px; margin-bottom: 0.25rem; display: block; }#fv-chart-1785251267479-g474gdird .fv-versus-bar-container { position: relative; height: 32px; display: flex; align-items: center; }#fv-chart-1785251267479-g474gdird .fv-versus-bar-left-wrapper { flex: 1; height: 100%; display: flex; justify-content: flex-end; align-items: center; }#fv-chart-1785251267479-g474gdird .fv-versus-bar-right-wrapper { flex: 1; height: 100%; display: flex; justify-content: flex-start; align-items: center; }#fv-chart-1785251267479-g474gdird .fv-versus-bar { height: 32px; width: var(--target-width); transition: width 0.8s ease-out; animation: fv-grow-max-width 0.8s ease-out forwards; display: flex; align-items: center; overflow: hidden; color: #ffffff; }#fv-chart-1785251267479-g474gdird .fv-versus-bar-left { border-radius: 4px 0 0 4px; justify-content: flex-end; padding: 0 8px; }#fv-chart-1785251267479-g474gdird .fv-versus-bar-right { border-radius: 0 4px 4px 0; justify-content: flex-start; padding: 0 8px; }@keyframes fv-grow-max-width {from { max-width: 0; }to { max-width: 100%; }}#fv-chart-1785251267479-g474gdird .fv-versus-center-line { position: absolute; left: 50%; top: 0; bottom: 0; width: 4px; background-color: #ffffff; transform: translateX(-50%); z-index: 1; }#fv-chart-1785251267479-g474gdird .fv-inside-left { white-space: nowrap; flex-shrink: 0; }#fv-chart-1785251267479-g474gdird .fv-inside-right { white-space: nowrap; flex-shrink: 0; }#fv-chart-1785251267479-g474gdird .fv-versus-val-text { font-family: 'Poppins', sans-serif; font-weight: 700; font-size: 14px; }#fv-chart-1785251267479-g474gdird .fv-versus-pct-diff { font-size: 12px; font-weight: 600; }#fv-chart-1785251267479-g474gdird .fv-versus-label { position: absolute; left: 50%; transform: translateX(-50%); top: 0; background-color: transparent; border: none; box-shadow: none; padding: 0; font-family: 'Open Sans', sans-serif; font-weight: 700; font-size: 14px; color: #374151; white-space: nowrap; }#fv-chart-1785251267479-g474gdird .sr-only { position: absolute !important; width: 1px !important; height: 1px !important; padding: 0 !important; margin: -1px !important; overflow: hidden !important; clip: rect(0,0,0,0) !important; white-space: nowrap !important; border: 0 !important; }#fv-chart-1785251267479-g474gdird .fv-bottom-bar { display: flex !important; flex-direction: column !important; align-items: center !important; margin-top: 0.5rem !important; gap: 1rem !important; }#fv-chart-1785251267479-g474gdird .fv-footer-content { text-align: center !important; width: 100% !important; }#fv-chart-1785251267479-g474gdird .fv-logo {display: block !important;margin: 0 auto !important;width: 120px !important;min-width: 120px !important;max-width: 120px !important;height: auto !important;object-fit: contain !important;flex-shrink: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-dropdown-wrapper { text-align: center !important; margin-bottom: 16px !important; margin-top: 0 !important; }#fv-chart-1785251267479-g474gdird .fv-dropdown-title-container { position: relative !important; display: inline-block !important; max-width: 100% !important; }#fv-chart-1785251267479-g474gdird .fv-dropdown-title {appearance: none !important;-webkit-appearance: none !important;-moz-appearance: none !important;background: transparent !important;border: none !important;font-size: 18px !important;font-weight: 600 !important;color: var(--riv-primary) !important;padding-right: 28px !important;padding-left: 10px !important;cursor: pointer !important;text-align: center !important;text-align-last: center !important;width: auto !important;max-width: 100% !important;font-family: 'Open Sans', sans-serif !important;line-height: 1.3 !important;margin: 0 !important;text-overflow: ellipsis !important;overflow: hidden !important;white-space: nowrap !important;}#fv-chart-1785251267479-g474gdird .fv-dropdown-title:focus { outline: none !important; }#fv-chart-1785251267479-g474gdird .fv-dropdown-title::-ms-expand { display: none !important; }#fv-chart-1785251267479-g474gdird .fv-dropdown-chevron {position: absolute !important;right: 0 !important;top: 50% !important;transform: translateY(-50%) !important;pointer-events: none !important;color: var(--riv-primary) !important;display: flex !important;align-items: center !important;}#fv-chart-1785251267479-g474gdird .fv-carousel-title-controls { display: flex !important; justify-content: space-between !important; align-items: center !important; margin-bottom: 16px !important; width: 100% !important; gap: 12px !important; }#fv-chart-1785251267479-g474gdird .fv-carousel-nav-btn {background: transparent !important; border: 1px solid #d1d5db !important; border-radius: 6px !important; padding: 6px 10px !important;cursor: pointer !important; font-size: 14px !important; color: #374151 !important; display: flex !important; align-items: center !important; gap: 4px !important; font-family: 'Open Sans', sans-serif !important;}#fv-chart-1785251267479-g474gdird .fv-carousel-nav-btn:hover { border-color: #9ca3af !important; }#fv-chart-1785251267479-g474gdird .fv-carousel-counter { font-size: 14px !important; color: #374151 !important; text-align: center !important; margin-top: 1rem !important; }#fv-chart-1785251267479-g474gdird .fv-legend { display: flex !important; justify-content: center !important; flex-wrap: wrap !important; gap: 8px 16px !important; margin: 0 !important; padding: 0 !important; margin-top: 1rem !important; }#fv-chart-1785251267479-g474gdird .fv-legend-item { display: flex !important; align-items: center !important; gap: 6px !important; font-size: 14px !important; color: #374151 !important; }#fv-chart-1785251267479-g474gdird .fv-legend-color { width: 12px !important; height: 12px !important; border-radius: 3px !important; }#fv-chart-1785251267479-g474gdird .fv-multi-value-legend {display: flex !important;justify-content: center !important;flex-wrap: wrap !important;gap: 12px 24px !important;margin-bottom: 1.5rem !important;padding: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-multi-legend-item { display: flex !important; align-items: center !important; gap: 8px !important; font-size: 14px !important; color: #374151 !important; font-weight: 500 !important; }#fv-chart-1785251267479-g474gdird .fv-multi-legend-swatch { width: 16px !important; height: 16px !important; border-radius: 3px !important; }#fv-chart-1785251267479-g474gdird .fv-benchmark-group { margin-bottom: 1rem !important; }#fv-chart-1785251267479-g474gdird .fv-benchmark-title {font-size: 18px !important; font-weight: 600 !important; margin-bottom: 16px !important; margin-top: 0 !important; padding: 0 !important;text-align: center !important; color: var(--riv-primary) !important; flex: 1 !important; min-width: 0 !important;font-family: 'Open Sans', sans-serif !important; line-height: 1.3 !important;text-transform: none !important;white-space: normal !important;overflow-wrap: break-word !important;word-wrap: break-word !important;max-width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-bar-row, #fv-chart-1785251267479-g474gdird .fv-stacked-product { display: flex !important; align-items: center !important; width: 100% !important; margin-bottom: 0.75rem !important; position: relative !important; }#fv-chart-1785251267479-g474gdird .fv-bar-label { width: 150px !important; flex-shrink: 0 !important; font-size: 14px !important; color: #374151 !important; padding-right: 10px !important; text-align: right !important; font-weight: 500 !important; display: block !important; }#fv-chart-1785251267479-g474gdird .fv-bar-container { flex-grow: 1 !important; background-color: #E5E7EB !important; border-radius: 4px !important; min-height: 25px !important; border: 1px solid #D1D5DB !important; position: relative !important; display: flex !important; align-items: center !important; }#fv-chart-1785251267479-g474gdird .fv-bar-commentary-inline { display: none !important; position: absolute !important; left: 150px !important; top: 0 !important; bottom: 0 !important; right: 0 !important; width: calc(100% - 150px) !important; margin: 0 !important; padding: 0 8px !important; font-size: 13px !important; color: #fff !important; background: rgba(0,0,0,0.8) !important; border-radius: 4px !important; line-height: 1.4 !important; font-weight: normal !important; text-transform: none !important; word-wrap: break-word !important; z-index: 10 !important; align-items: center !important; overflow-y: auto !important; }#fv-chart-1785251267479-g474gdird.preview-wrapper .fv-bar-row:hover .fv-bar-commentary-inline, #fv-chart-1785251267479-g474gdird.preview-wrapper .fv-bar-commentary-inline:focus, #fv-chart-1785251267479-g474gdird.preview-wrapper .fv-bar-commentary-inline:focus-within, #fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row:hover .fv-bar-commentary-inline, #fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-commentary-inline:focus, #fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-commentary-inline:focus-within { display: flex !important; }#fv-chart-1785251267479-g474gdird .fv-bar { height: 100% !important; border-radius: 3px !important; display: flex !important; align-items: center !important; transition: opacity 0.2s ease, width 0.8s ease-out !important; min-height: 23px !important; }#fv-chart-1785251267479-g474gdird .fv-bar:hover { opacity: 0.8 !important; }#fv-chart-1785251267479-g474gdird .fv-bar-inner-content { display: flex !important; justify-content: space-between !important; align-items: center !important; width: 100% !important; height: 100% !important; padding: 0 8px !important; font-size: 14px !important; font-weight: bold !important; overflow: hidden !important; }#fv-chart-1785251267479-g474gdird .fv-bar-inner-label { white-space: nowrap !important; overflow: hidden !important; text-overflow: ellipsis !important; padding-right: 8px !important; }#fv-chart-1785251267479-g474gdird .fv-bar-inner-value { flex-shrink: 0 !important; }#fv-chart-1785251267479-g474gdird .fv-bar-value-outside { padding-left: 8px !important; font-size: 14px !important; font-weight: bold !important; color: #374151 !important; white-space: nowrap !important; }#fv-chart-1785251267479-g474gdird .fv-bar-label.fv-primary-product { font-weight: bold !important; color: var(--riv-primary) !important; }#fv-chart-1785251267479-g474gdird .fv-multi-bar-container { flex-direction: column !important; padding: 4px !important; align-items: stretch !important; gap: 4px !important; height: auto !important; }#fv-chart-1785251267479-g474gdird .fv-multi-bar-item { display: flex !important; align-items: center !important; height: 25px !important; width: 100% !important; }#fv-chart-1785251267479-g474gdird .fv-stacked-bar { display: flex !important; overflow: hidden !important; }#fv-chart-1785251267479-g474gdird .fv-stacked-segment { height: 100% !important; display: flex !important; align-items: center !important; justify-content: flex-end !important; padding-right: 8px !important; border-right: 1px solid rgba(255,255,255,0.3) !important; }#fv-chart-1785251267479-g474gdird .fv-stacked-segment:last-child { border-right: none !important; }#fv-chart-1785251267479-g474gdird .fv-segment-value { font-size: 14px !important; font-weight: bold !important; }#fv-chart-1785251267479-g474gdird .fv-grouped-bar-product { display: flex !important; flex-direction: column !important; width: 100% !important; margin-bottom: 1.25rem !important; }#fv-chart-1785251267479-g474gdird .fv-grouped-product-title-wrapper { padding-left: 150px !important; }#fv-chart-1785251267479-g474gdird .fv-grouped-product-title { width: 100% !important; text-align: left !important; padding-right: 0 !important; margin-bottom: 0.5rem !important; font-weight: 700 !important; font-size: 14px !important; color: #374151 !important; text-transform: none !important; }#fv-chart-1785251267479-g474gdird .fv-bar-cluster { width: 100% !important; flex-grow: 1 !important; display: flex !important; flex-direction: column !important; }#fv-chart-1785251267479-g474gdird .fv-bar-cluster .fv-bar-row { margin-bottom: 3px !important; }#fv-chart-1785251267479-g474gdird .fv-bar-cluster .fv-bar-container { height: 20px !important; }#fv-chart-1785251267479-g474gdird .riv-grid line {stroke: #D1D5DB !important;stroke-dasharray: 3 3 !important;}#fv-chart-1785251267479-g474gdird .fv-x-axis-wrapper { display: flex !important; width: 100% !important; margin-top: 0.5rem !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-label-space { width: 150px !important; padding-right: 10px !important; flex-shrink: 0 !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-chart-space { flex-grow: 1 !important; padding-right: 8px !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-wrapper.fv-grouped-x-axis { margin-left: 0 !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-line { border-top: 1px solid #D1D5DB !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-ticks { display: flex !important; justify-content: space-between !important; padding-top: 4px !important; font-size: 13px !important; color: #374151 !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-ticks span { position: relative !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-ticks span::before { content: '' !important; position: absolute !important; top: -6px !important; left: 50% !important; transform: translateX(-50%) !important; width: 2px !important; height: 4px !important; background-color: #D1D5DB !important; border-radius: 1px !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-unit { text-align: center !important; font-size: 14px !important; color: #374151 !important; margin-top: 8px !important; display: block !important; }#fv-chart-1785251267479-g474gdird .fv-x-axis-title { text-align: center !important; font-size: 15px !important; color: #374151 !important; margin-top: 8px !important; margin-bottom: 16px !important; line-height: 1.5 !important; padding: 0 1rem !important; display: block !important; font-weight: bold !important; }#fv-chart-1785251267479-g474gdird .fv-y-axis-title {font-size: 15px !important;color: #374151 !important;line-height: 1.5 !important;text-align: left !important;padding-left: 5.83% !important;margin-bottom: 4px !important;display: block !important;font-weight: bold !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-pie-container,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-pie-container {flex-direction: column !important; gap: 1rem !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-grouped-product-title-wrapper,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-grouped-product-title-wrapper {padding-left: 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row,#fv-chart-1785251267479-g474gdird.mobile-view .fv-stacked-product,#fv-chart-1785251267479-g474gdird.mobile-view .fv-grouped-bar-product,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-row,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-stacked-product,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-grouped-bar-product {flex-direction: column !important; align-items: flex-start !important; margin-bottom: 1.25rem !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-label:not(.fv-grouped-product-title),#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-label:not(.fv-grouped-product-title) {width: 100% !important; text-align: left !important; padding-right: 0 !important; margin-bottom: 0.25rem !important; font-size: 14px !important; font-weight: 700 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-label,#fv-chart-1785251267479-g474gdird.mobile-view .fv-grouped-product-title,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-label,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-grouped-product-title {width: 100% !important; text-align: left !important; padding-right: 0 !important; margin-bottom: 0.25rem !important; font-size: 14px !important; font-weight: 700 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-container,#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-cluster,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-container,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-cluster {width: 100% !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row:hover .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row .fv-bar-commentary-inline:focus,#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-row .fv-bar-commentary-inline:focus-within,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-row .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-row:hover .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-row .fv-bar-commentary-inline:focus,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-row .fv-bar-commentary-inline:focus-within {position: static !important; display: block !important; width: 100% !important; margin: 4px 0 0 0 !important; padding: 0 !important; background: transparent !important; color: #6B7280 !important; font-size: 12px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-x-axis-wrapper,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-x-axis-wrapper {margin-left: 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-x-axis-label-space,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-x-axis-label-space {display: none !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-x-axis-chart-space,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-x-axis-chart-space {padding-right: 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-benchmark-title,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-benchmark-title {font-size: 16px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-dropdown-title,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-dropdown-title {font-size: 16px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-carousel-nav-btn,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-carousel-nav-btn {padding: 8px 12px !important; font-size: 14px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-chart-title,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-chart-title {padding: 0 8px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-chart-subhead,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-chart-subhead {padding: 0 8px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-header,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-header {flex-direction: column !important; align-items: center !important; padding: 0 !important; gap: 0.5rem !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select-wrapper,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select-wrapper {flex: 1 !important; min-width: 0 !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select-wrapper.fv-left,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select-wrapper.fv-left {text-align: center !important; padding-right: 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select-wrapper.fv-right,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select-wrapper.fv-right {text-align: center !important; padding-left: 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select.fv-select-left,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select.fv-select-left {text-align: center !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select.fv-select-right,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select.fv-select-right {text-align: center !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-vs,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-vs {text-align: center !important; padding: 0.25rem 0 !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select-container,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select-container {max-width: 100% !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-versus-select,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-versus-select {font-size: 14px !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-stl-shop-all-btn,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-stl-shop-all-btn {bottom: 0.5rem !important; right: 0.5rem !important; height: 2rem !important; font-size: 0.75rem !important; padding: 0 0.75rem 0 2.5rem !important; max-width: calc(100% - 1rem) !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-stl-shop-all-logo,#fv-chart-1785251267479-g474gdird.mobile-view .fv-stl-shop-all-icon,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-stl-shop-all-logo,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-stl-shop-all-icon {width: 2rem !important; height: 2rem !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-stl-shop-all-icon svg,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-stl-shop-all-icon svg {width: 14px !important; height: 14px !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird.labels-on-top .fv-bar-commentary-inline {display: block !important; margin-left: 0 !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird.mobile-view .fv-y-axis-title { padding-left: 5% !important; }#fv-chart-1785251267479-g474gdird.mobile-view.fv-contains-line-chart .fv-footer-content {margin-left: -1rem !important;margin-right: -1rem !important;}@media (max-width: 599px) {#fv-chart-1785251267479-g474gdird .fv-pie-container {flex-direction: column !important; gap: 1rem !important;}#fv-chart-1785251267479-g474gdird .fv-grouped-product-title-wrapper {padding-left: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-bar-row,#fv-chart-1785251267479-g474gdird .fv-stacked-product,#fv-chart-1785251267479-g474gdird .fv-grouped-bar-product {flex-direction: column !important; align-items: flex-start !important; margin-bottom: 1.25rem !important;}#fv-chart-1785251267479-g474gdird .fv-bar-label:not(.fv-grouped-product-title) {width: 100% !important; text-align: left !important; padding-right: 0 !important; margin-bottom: 0.25rem !important; font-size: 14px !important; font-weight: 700 !important;}#fv-chart-1785251267479-g474gdird .fv-bar-label,#fv-chart-1785251267479-g474gdird .fv-grouped-product-title {width: 100% !important; text-align: left !important; padding-right: 0 !important; margin-bottom: 0.25rem !important; font-size: 14px !important; font-weight: 700 !important;}#fv-chart-1785251267479-g474gdird .fv-bar-container,#fv-chart-1785251267479-g474gdird .fv-bar-cluster {width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-bar-row .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird .fv-bar-row:hover .fv-bar-commentary-inline,#fv-chart-1785251267479-g474gdird .fv-bar-row .fv-bar-commentary-inline:focus,#fv-chart-1785251267479-g474gdird .fv-bar-row .fv-bar-commentary-inline:focus-within {position: static !important; display: block !important; width: 100% !important; margin: 4px 0 0 0 !important; padding: 0 !important; background: transparent !important; color: #6B7280 !important; font-size: 12px !important;}#fv-chart-1785251267479-g474gdird .fv-x-axis-wrapper {margin-left: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-x-axis-label-space {display: none !important;}#fv-chart-1785251267479-g474gdird .fv-x-axis-chart-space {padding-right: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-benchmark-title {font-size: 16px !important;}#fv-chart-1785251267479-g474gdird .fv-dropdown-title {font-size: 16px !important;}#fv-chart-1785251267479-g474gdird .fv-carousel-nav-btn {padding: 8px 12px !important; font-size: 14px !important;}#fv-chart-1785251267479-g474gdird .fv-chart-title {padding: 0 8px !important;}#fv-chart-1785251267479-g474gdird .fv-chart-subhead {padding: 0 8px !important;}#fv-chart-1785251267479-g474gdird .fv-versus-header {flex-direction: column !important; align-items: center !important; padding: 0 !important; gap: 0.5rem !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper {flex: 1 !important; min-width: 0 !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-left {text-align: center !important; padding-right: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select-wrapper.fv-right {text-align: center !important; padding-left: 0 !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select.fv-select-left {text-align: center !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select.fv-select-right {text-align: center !important;}#fv-chart-1785251267479-g474gdird .fv-versus-vs {text-align: center !important; padding: 0.25rem 0 !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select-container {max-width: 100% !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-versus-select {font-size: 14px !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-stl-shop-all-btn {bottom: 0.5rem !important; right: 0.5rem !important; height: 2rem !important; font-size: 0.75rem !important; padding: 0 0.75rem 0 2.5rem !important; max-width: calc(100% - 1rem) !important;}#fv-chart-1785251267479-g474gdird .fv-stl-shop-all-logo,#fv-chart-1785251267479-g474gdird .fv-stl-shop-all-icon {width: 2rem !important; height: 2rem !important;}#fv-chart-1785251267479-g474gdird .fv-stl-shop-all-icon svg {width: 14px !important; height: 14px !important;}#fv-chart-1785251267479-g474gdird .fv-bar-commentary-inline {display: block !important; margin-left: 0 !important; width: 100% !important;}#fv-chart-1785251267479-g474gdird .fv-y-axis-title { padding-left: 5% !important; }#fv-chart-1785251267479-g474gdird.fv-contains-line-chart .fv-footer-content {margin-left: -1rem !important;margin-right: -1rem !important;}}{"@context": "https://schema.org","@type": "Dataset","name": "VPN auto-renewal costs","description": "Comparing the prices of 1-year VPN plans across the 'big four' initially and after renewal","creator": {"@type": "Organization","name": "TechRadar","logo": "https://cdn.mos.cms.futurecdn.net/dAwPEozB9GMMhAngdFJTHb.png"},"isAccessibleForFree": true,"dateCreated": "2026-07-28T15:07:47.480Z","citation": "Add your caption to explain the data, the learnings and the key takeaways of the data being visualized","keywords": ["NordVPN","Surfshark","ExpressVPN","Proton VPN","Initial cost","Renewal cost","Group 1","benchmark","comparison","performance","review","TechRadar"],"measurementTechnique": "Performance Benchmarking","variableMeasured": [{"@type": "PropertyValue","name": "NordVPN – Initial cost","value": 53.88,"unitText": "USD"},{"@type": "PropertyValue","name": "NordVPN – Renewal cost","value": 139.08,"unitText": "USD"},{"@type": "PropertyValue","name": "Surfshark – Initial cost","value": 50.85,"unitText": "USD"},{"@type": "PropertyValue","name": "Surfshark – Renewal cost","value": 79,"unitText": "USD"},{"@type": "PropertyValue","name": "ExpressVPN – Initial cost","value": 74.85,"unitText": "USD"},{"@type": "PropertyValue","name": "ExpressVPN – Renewal cost","value": 99.95,"unitText": "USD"},{"@type": "PropertyValue","name": "Proton VPN – Initial cost","value": 47.88,"unitText": "USD"},{"@type": "PropertyValue","name": "Proton VPN – Renewal cost","value": 83.88,"unitText": "USD"}]}VPN auto-renewal costs

Comparing the prices of 1-year VPN plans across the 'big four' initially and after renewal

NordVPNInitial cost53.88Renewal cost139.08SurfsharkInitial cost50.85Renewal cost79ExpressVPNInitial cost74.85Renewal cost99.95Proton VPNInitial cost47.88Renewal cost83.88037.575112.5150USDGroup 1 DataProductInitial cost (USD)Renewal cost (USD)NordVPN53.88139.08Surfshark50.8579ExpressVPN74.8599.95Proton VPN47.8883.88window.iFrameResizer = {heightCalculationMethod: 'taggedElement'};(function() {window.fvAnimateCharts = function(chartWrapper) {if (!chartWrapper) return;function animateBars(chartElement) {if (!chartElement) return;var bars = chartElement.querySelectorAll('.fv-bar, .fv-stacked-segment');bars.forEach(function(bar, index) {bar.style.setProperty('width', '0%', 'important');bar.style.setProperty('transition', 'none', 'important');var targetWidth = bar.dataset.targetWidth;if (targetWidth === undefined) return;void bar.offsetWidth;var targetMargin = bar.dataset.targetMargin;var baseMargin = bar.dataset.baseMargin;if (baseMargin !== undefined) {bar.style.setProperty('margin-left', baseMargin + '%', 'important');}setTimeout(function() {var marginTransition = baseMargin !== undefined ? ', margin-left 0.8s ease-out' : '';bar.style.setProperty('transition', 'opacity 0.2s ease, width 0.8s ease-out' + marginTransition, 'important');bar.style.setProperty('width', targetWidth + '%', 'important');if (targetMargin !== undefined && baseMargin !== undefined) {bar.style.setProperty('margin-left', targetMargin + '%', 'important');}}, index * 50 + 50);});}function animateLineChart(chartElement) {if (!chartElement) return;var lineSvg = chartElement.querySelector('svg');if (!lineSvg) return;var paths = lineSvg.querySelectorAll('.riv-line-path');paths.forEach(function(p, i) {if (typeof p.getTotalLength === 'function') {var len = p.getTotalLength();p.style.transition = 'none';p.style.strokeDasharray = len;p.style.strokeDashoffset = len;p.getBoundingClientRect();setTimeout(function() {p.style.transition = 'stroke-dashoffset 1s ease-out ' + (i * 0.1) + 's, stroke-width 0.2s, opacity 0.2s';p.style.strokeDashoffset = '0';}, 100);}});var dots = lineSvg.querySelectorAll('.riv-dot');dots.forEach(function(dot, i) {dot.style.opacity = '0';setTimeout(function() {dot.style.transition = 'opacity 0.3s ease';dot.style.opacity = '1';}, 500 + i * 10);});}var charts = chartWrapper.querySelectorAll('.fv-chart-item');charts.forEach(function(chart) {if (window.getComputedStyle(chart).display === 'none') return;var chartType = chart.dataset.chartType;if (chartType === 'Line') {animateLineChart(chart);} else if (chartType !== 'Pie') {animateBars(chart);}});};function initialize(uniqueId, isSlideshow) {var root = document.getElementById(uniqueId);if (!root && !isSlideshow) return;function setupWrapper(chartWrapper) {if (!chartWrapper) return;function checkMobileView() {var width = chartWrapper.getBoundingClientRect().width;var isMobileDevice = window.screen && Math.min(window.screen.width, window.screen.height) <= 599;var isMobile;if (width === 0) {isMobile = isMobileDevice || window.matchMedia('(max-width: 599px)').matches;} else {isMobile = isMobileDevice || width < 600;}if (isMobile) {chartWrapper.classList.add('mobile-view');} else {chartWrapper.classList.remove('mobile-view');}}checkMobileView();if (typeof ResizeObserver !== 'undefined') {var ro = new ResizeObserver(function() {checkMobileView();});ro.observe(chartWrapper);}window.addEventListener('resize', checkMobileView);var labelsOnTop = chartWrapper.dataset.barLabelsOnTop === 'true';if (labelsOnTop) {var charts = chartWrapper.querySelectorAll('.fv-chart-item');var hasVisibleBarChart = false;charts.forEach(function(c) {if (window.getComputedStyle(c).display !== 'none') {var cType = c.dataset.chartType;if (cType === 'Bar' || cType === 'Stacked Bar' || cType === 'Versus') {hasVisibleBarChart = true;}}});if (hasVisibleBarChart) {chartWrapper.classList.add('labels-on-top');} else {chartWrapper.classList.remove('labels-on-top');}} else {chartWrapper.classList.remove('labels-on-top');}chartWrapper.querySelectorAll('.riv-line-legend').forEach(function(legend) {var chartId = legend.dataset.chart;var chart = chartWrapper.querySelector('[id="' + chartId + '"]');if (!chart) return;var lines = chart.querySelectorAll('.riv-line-path');var dots = chart.querySelectorAll('.riv-dot');var legendItems = legend.querySelectorAll('.riv-legend-item');legendItems.forEach(function(item) {var productName = item.dataset.product;var safeProductName = productName.replace(/"/g, '\"');var line = chart.querySelector('.riv-line-path[data-product="' + safeProductName + '"]');var productDots = chart.querySelectorAll('.riv-dot[data-product="' + safeProductName + '"]');if (line && !line.dataset.origStroke) {line.dataset.origStroke = line.getAttribute('stroke-width') || '2';}var highlight = function() {if (legend.querySelector('.isolated')) return;lines.forEach(function(l) { l.style.opacity = '0.2'; l.setAttribute('stroke-width', l.dataset.origStroke); });dots.forEach(function(d) { d.style.opacity = '0.2'; });legendItems.forEach(function(i) { i.style.opacity = '0.3'; i.style.transition = 'opacity 0.2s'; });if (line) {line.style.opacity = '1';line.setAttribute('stroke-width', '4');}productDots.forEach(function(d) { d.style.opacity = '1'; });item.style.opacity = '1';};var clearHighlight = function() {if (legend.querySelector('.isolated')) return;lines.forEach(function(l) { l.style.opacity = '1'; l.setAttribute('stroke-width', l.dataset.origStroke); });dots.forEach(function(d) { d.style.opacity = '1'; });legendItems.forEach(function(i) { i.style.opacity = '1'; });};var toggleIsolate = function() {var isIsolated = item.classList.contains('isolated');legendItems.forEach(function(i) { i.classList.remove('isolated'); });if (isIsolated) {lines.forEach(function(l) { l.style.opacity = '1'; l.setAttribute('stroke-width', l.dataset.origStroke); });dots.forEach(function(d) { d.style.opacity = '1'; });legendItems.forEach(function(i) { i.style.opacity = '1'; });} else {item.classList.add('isolated');lines.forEach(function(l) { l.style.opacity = '0.2'; l.setAttribute('stroke-width', l.dataset.origStroke); });dots.forEach(function(d) { d.style.opacity = '0.2'; });legendItems.forEach(function(i) { i.style.opacity = '0.3'; i.style.transition = 'opacity 0.2s'; });if (line) {line.style.opacity = '1';line.setAttribute('stroke-width', '4');}productDots.forEach(function(d) { d.style.opacity = '1'; });item.style.opacity = '1';}};item.addEventListener('mouseenter', highlight);item.addEventListener('focus', highlight);item.addEventListener('mouseleave', clearHighlight);item.addEventListener('blur', clearHighlight);item.addEventListener('click', toggleIsolate);if (line) {line.style.cursor = 'pointer';line.addEventListener('mouseenter', highlight);line.addEventListener('mouseleave', clearHighlight);line.addEventListener('click', toggleIsolate);}});});var versusCharts = chartWrapper.querySelectorAll('.fv-versus-chart');versusCharts.forEach(function(vc) {var dataJson = vc.getAttribute('data-versus-data');var leftColor = vc.getAttribute('data-left-color') || '#E53935';var rightColor = vc.getAttribute('data-right-color') || '#FFB300';if (!dataJson) return;var versusData = JSON.parse(dataJson);var leftSelect = vc.querySelector('.fv-select-left');var rightSelect = vc.querySelector('.fv-select-right');var rows = vc.querySelectorAll('.fv-versus-row');function updateVersusChart() {var leftProduct = leftSelect.value;var rightProduct = rightSelect.value;leftSelect.style.color = leftColor;leftSelect.style.borderColor = leftColor;rightSelect.style.color = rightColor;rightSelect.style.borderColor = rightColor;rows.forEach(function(row) {var index = parseInt(row.getAttribute('data-index'));var data = versusData[index];if (!data) return;var leftVal = data.productData[leftProduct] ? data.productData[leftProduct].value : undefined;var rightVal = data.productData[rightProduct] ? data.productData[rightProduct].value : undefined;var leftIsNum = typeof leftVal === 'number';var rightIsNum = typeof rightVal === 'number';var leftNum = leftIsNum ? leftVal : 0;var rightNum = rightIsNum ? rightVal : 0;var maxVal = Math.max(leftNum, rightNum, 0.0001);var leftWidth = leftIsNum ? (leftNum / maxVal) * 95 : 0;var rightWidth = rightIsNum ? (rightNum / maxVal) * 85 : 0;var winner = null;var pctDiffStr = null;if (leftIsNum && rightIsNum) {if (leftNum > rightNum) {winner = 'left';if (rightNum > 0) {var diff = Math.round(((leftNum - rightNum) / rightNum) * 100);pctDiffStr = '+' + diff.toLocaleString() + '%';}} else if (rightNum > leftNum) {winner = 'right';if (leftNum > 0) {var diff = Math.round(((rightNum - leftNum) / leftNum) * 100);pctDiffStr = '+' + diff.toLocaleString() + '%';}}}var leftDisplay = data.productData[leftProduct] && data.productData[leftProduct].displayValue !== undefined ? data.productData[leftProduct].displayValue : (leftIsNum ? leftNum.toLocaleString() : (leftVal !== undefined ? leftVal : '-'));var rightDisplay = data.productData[rightProduct] && data.productData[rightProduct].displayValue !== undefined ? data.productData[rightProduct].displayValue : (rightIsNum ? rightNum.toLocaleString() : (rightVal !== undefined ? rightVal : '-'));var unit = (data.productData[leftProduct] && data.productData[leftProduct].unit) ||(data.productData[rightProduct] && data.productData[rightProduct].unit) || '';var leftTextStr = leftDisplay;var rightTextStr = rightDisplay;var leftBar = row.querySelector('.fv-versus-bar-left');var rightBar = row.querySelector('.fv-versus-bar-right');var leftText = row.querySelector('.fv-inside-left');var rightText = row.querySelector('.fv-inside-right');var labelText = row.querySelector('.fv-versus-label span');var leftWrapper = row.querySelector('.fv-versus-bar-left-wrapper');var rightWrapper = row.querySelector('.fv-versus-bar-right-wrapper');var existingPctDiffs = row.querySelectorAll('.fv-versus-pct-diff');existingPctDiffs.forEach(function(el) { el.remove(); });if (winner === 'left' && pctDiffStr) {var pctSpan = document.createElement('span');pctSpan.className = 'fv-versus-pct-diff';pctSpan.style.color = 'rgba(255, 255, 255, 0.9)';pctSpan.textContent = pctDiffStr;if (leftBar) leftBar.insertBefore(pctSpan, leftBar.firstChild);} else if (winner === 'right' && pctDiffStr) {var pctSpan = document.createElement('span');pctSpan.className = 'fv-versus-pct-diff';pctSpan.style.color = 'rgba(255, 255, 255, 0.9)';pctSpan.textContent = pctDiffStr;if (rightBar) rightBar.appendChild(pctSpan);}if (leftBar) {leftBar.style.backgroundColor = leftColor;leftBar.dataset.targetWidth = leftWidth;leftBar.style.setProperty('--target-width', leftWidth + '%');leftBar.style.width = leftWidth + '%';}if (rightBar) {rightBar.style.backgroundColor = rightColor;rightBar.dataset.targetWidth = rightWidth;rightBar.style.setProperty('--target-width', rightWidth + '%');rightBar.style.width = rightWidth + '%';}if (leftText) {leftText.innerHTML = leftTextStr;}if (rightText) {rightText.innerHTML = rightTextStr;}if (labelText) {labelText.textContent = data.attribute + (unit ? ' (' + unit + ')' : '');}});}if (leftSelect) leftSelect.addEventListener('change', updateVersusChart);if (rightSelect) rightSelect.addEventListener('change', updateVersusChart);});var barRows = chartWrapper.querySelectorAll('.fv-bar-row');var globalCaptionEl = chartWrapper.querySelector('.rv-chart-caption');var fallbackCaptionHtml = globalCaptionEl ? globalCaptionEl.innerHTML : '';barRows.forEach(function(row) {var commentaryEl = row.querySelector('[data-commentary-key]');if (commentaryEl) {var commentaryText = commentaryEl.textContent;if (commentaryText && commentaryText.trim().length > 0) {row.addEventListener('mouseenter', function() {if (!chartWrapper.classList.contains('mobile-view') && globalCaptionEl) {globalCaptionEl.innerHTML = commentaryText;globalCaptionEl.classList.add('fv-bar-active-caption');}});row.addEventListener('mouseleave', function() {if (!chartWrapper.classList.contains('mobile-view') && globalCaptionEl) {globalCaptionEl.innerHTML = fallbackCaptionHtml;globalCaptionEl.classList.remove('fv-bar-active-caption');}});}}});var charts = chartWrapper.querySelectorAll('.fv-chart-item');var dropdown = chartWrapper.querySelector('.fv-dropdown-title');var prevBtn = chartWrapper.querySelector('.fv-carousel-nav-btn.prev');var nextBtn = chartWrapper.querySelector('.fv-carousel-nav-btn.next');var carouselTitle = chartWrapper.querySelector('.fv-carousel-title-controls .fv-benchmark-title');var counter = chartWrapper.querySelector('.fv-carousel-counter');var subheadEl = chartWrapper.querySelector('.fv-chart-subhead');var captionEl = chartWrapper.querySelector('.rv-chart-caption');var footerContentEl = chartWrapper.querySelector('.fv-footer-content');var bottomBarEl = chartWrapper.querySelector('.fv-bottom-bar');var logoEl = chartWrapper.querySelector('.fv-logo');if (charts.length > 1 && (dropdown || prevBtn)) {var currentChartIndex = 0;var titles = [];if (dropdown) {titles = Array.from(dropdown.options).map(function(o) { return o.text; });} else {charts.forEach(function(c) {titles.push(c.getAttribute('data-title') || '');});}function showInternalChart(index) {if (index < 0) index = charts.length - 1;if (index >= charts.length) index = 0;currentChartIndex = index;charts.forEach(function(c, i) {c.style.display = i === index ? 'block' : 'none';if (i === index) {var cType = c.dataset.chartType;if (cType === 'Line') {} else if (cType !== 'Pie') {window.fvAnimateCharts(chartWrapper);}var labelsOnTop = chartWrapper.dataset.barLabelsOnTop === 'true';if (labelsOnTop && (cType === 'Bar' || cType === 'Stacked Bar' || cType === 'Versus')) {chartWrapper.classList.add('labels-on-top');} else {chartWrapper.classList.remove('labels-on-top');}}});if (dropdown) dropdown.value = index;if (carouselTitle && titles[index]) carouselTitle.textContent = titles[index];if (counter) counter.textContent = (index + 1) + ' of ' + charts.length;var activeChart = charts[index];if (activeChart) {var newSubhead = activeChart.getAttribute('data-subhead');var newCaption = activeChart.getAttribute('data-caption');var currentChartType = activeChart.getAttribute('data-chart-type');var hideGlobalCaption = currentChartType === 'Countdown' || currentChartType === 'Image Comparison' || currentChartType === 'Shop the Collection';if (subheadEl) subheadEl.textContent = newSubhead || '';if (captionEl) {captionEl.textContent = newCaption || '';fallbackCaptionHtml = newCaption || '';}if (footerContentEl) {if (newCaption && newCaption.trim().length > 0 && !hideGlobalCaption) {footerContentEl.style.display = 'block';if (bottomBarEl) bottomBarEl.style.display = 'flex';} else {footerContentEl.style.display = 'none';if (bottomBarEl && !logoEl) {bottomBarEl.style.display = 'none';}}}}}if (dropdown) dropdown.addEventListener('change', function(e) { showInternalChart(parseInt(e.target.value)); });if (prevBtn) prevBtn.addEventListener('click', function() { showInternalChart(currentChartIndex - 1); });if (nextBtn) nextBtn.addEventListener('click', function() { showInternalChart(currentChartIndex + 1); });}var imageCompareWrappers = chartWrapper.querySelectorAll('.fv-image-compare-wrapper');imageCompareWrappers.forEach(function(wrapper) {var inner = wrapper.querySelector('.fv-image-compare-inner') || wrapper;var slider = wrapper.querySelector('.fv-image-compare-slider');var fgImage = wrapper.querySelector('.fv-image-compare-fg');var bgImage = wrapper.querySelector('.fv-image-compare-bg');var labelLeft = wrapper.querySelector('.fv-image-compare-label-left');var labelRight = wrapper.querySelector('.fv-image-compare-label-right');var isDragging = false;var scale = 1;var panX = 0;var panY = 0;var isPanning = false;var hasPanned = false;var lastClientX = 0;var lastClientY = 0;var initialDistance = null;var lastCenterX = null;var lastCenterY = null;function updateTransform() {if (wrapper.classList.contains('fv-image-compare-fullscreen')) {inner.style.setProperty('transform', 'translate(' + panX + 'px, ' + panY + 'px) scale(' + scale + ')', 'important');} else {inner.style.removeProperty('transform');scale = 1;panX = 0;panY = 0;}}function constrainPan() {var rect = wrapper.getBoundingClientRect();var maxPanX = Math.max(0, (rect.width * scale - rect.width) / 2);var maxPanY = Math.max(0, (rect.height * scale - rect.height) / 2);panX = Math.max(-maxPanX, Math.min(panX, maxPanX));panY = Math.max(-maxPanY, Math.min(panY, maxPanY));}wrapper.addEventListener('wheel', function(e) {if (!wrapper.classList.contains('fv-image-compare-fullscreen')) return;e.preventDefault();var zoomSensitivity = 0.005;var zoomFactor = Math.exp(-e.deltaY * zoomSensitivity);var newScale = Math.max(1, Math.min(scale * zoomFactor, 5));if (newScale === scale) return;var rect = wrapper.getBoundingClientRect();var mouseX = e.clientX - rect.left - rect.width / 2;var mouseY = e.clientY - rect.top - rect.height / 2;var ratio = newScale / scale;panX = mouseX - (mouseX - panX) * ratio;panY = mouseY - (mouseY - panY) * ratio;scale = newScale;constrainPan();updateTransform();}, { passive: false });wrapper.addEventListener('mousedown', function(e) {if (!wrapper.classList.contains('fv-image-compare-fullscreen') || scale <= 1) return;if (e.target.closest('.fv-image-compare-slider') || e.target.closest('button')) return;isPanning = true;hasPanned = false;lastClientX = e.clientX;lastClientY = e.clientY;});window.addEventListener('mousemove', function(e) {if (!isPanning) return;var dx = e.clientX - lastClientX;var dy = e.clientY - lastClientY;if (Math.abs(dx) > 2 || Math.abs(dy) > 2) {hasPanned = true;}lastClientX = e.clientX;lastClientY = e.clientY;panX += dx;panY += dy;constrainPan();updateTransform();});window.addEventListener('mouseup', function() {isPanning = false;});wrapper.addEventListener('touchstart', function(e) {if (!wrapper.classList.contains('fv-image-compare-fullscreen')) return;if (e.touches.length === 2) {e.preventDefault();var dx = e.touches[0].clientX - e.touches[1].clientX;var dy = e.touches[0].clientY - e.touches[1].clientY;initialDistance = Math.sqrt(dx * dx + dy * dy);var rect = wrapper.getBoundingClientRect();lastCenterX = (e.touches[0].clientX + e.touches[1].clientX) / 2 - rect.left - rect.width / 2;lastCenterY = (e.touches[0].clientY + e.touches[1].clientY) / 2 - rect.top - rect.height / 2;hasPanned = true;} else if (e.touches.length === 1 && scale > 1) {if (e.target.closest('.fv-image-compare-slider') || e.target.closest('button')) return;isPanning = true;hasPanned = false;lastClientX = e.touches[0].clientX;lastClientY = e.touches[0].clientY;}}, { passive: false });wrapper.addEventListener('touchmove', function(e) {if (!wrapper.classList.contains('fv-image-compare-fullscreen')) return;if (e.touches.length === 2 && initialDistance !== null) {e.preventDefault();var dx = e.touches[0].clientX - e.touches[1].clientX;var dy = e.touches[0].clientY - e.touches[1].clientY;var distance = Math.sqrt(dx * dx + dy * dy);if (initialDistance > 0) {var zoomFactor = distance / initialDistance;var newScale = Math.max(1, Math.min(scale * zoomFactor, 5));var rect = wrapper.getBoundingClientRect();var centerX = (e.touches[0].clientX + e.touches[1].clientX) / 2 - rect.left - rect.width / 2;var centerY = (e.touches[0].clientY + e.touches[1].clientY) / 2 - rect.top - rect.height / 2;var ratio = newScale / scale;panX = centerX - (centerX - panX) * ratio;panY = centerY - (centerY - panY) * ratio;if (lastCenterX !== null && lastCenterY !== null) {panX += (centerX - lastCenterX);panY += (centerY - lastCenterY);}scale = newScale;lastCenterX = centerX;lastCenterY = centerY;constrainPan();updateTransform();}initialDistance = distance;} else if (e.touches.length === 1 && isPanning) {e.preventDefault();var dx = e.touches[0].clientX - lastClientX;var dy = e.touches[0].clientY - lastClientY;if (Math.abs(dx) > 2 || Math.abs(dy) > 2) {hasPanned = true;}lastClientX = e.touches[0].clientX;lastClientY = e.touches[0].clientY;panX += dx;panY += dy;constrainPan();updateTransform();}}, { passive: false });wrapper.addEventListener('touchend', function(e) {if (e.touches.length < 2) {initialDistance = null;}if (e.touches.length === 0) {isPanning = false;}});function handleMove(clientX) {var rect = inner.getBoundingClientRect();var x = Math.max(0, Math.min(clientX - rect.left, rect.width));var percent = Math.max(0, Math.min((x / rect.width) * 100, 100));if (slider) slider.style.setProperty('left', percent + '%', 'important');if (fgImage) fgImage.style.setProperty('clip-path', 'polygon(0 0, ' + percent + '% 0, ' + percent + '% 100%, 0 100%)', 'important');if (labelLeft) {if (percent < 10) {labelLeft.style.setProperty('opacity', '0', 'important');} else {labelLeft.style.setProperty('opacity', '1', 'important');}}if (labelRight) {if (percent > 90) {labelRight.style.setProperty('opacity', '0', 'important');} else {labelRight.style.setProperty('opacity', '1', 'important');}}}function onMouseMove(e) {if (!isDragging) return;handleMove(e.clientX);}function onTouchMove(e) {if (!isDragging) return;e.preventDefault();handleMove(e.touches[0].clientX);}function stopDragging() {isDragging = false;window.removeEventListener('mousemove', onMouseMove);window.removeEventListener('mouseup', stopDragging);window.removeEventListener('touchmove', onTouchMove);window.removeEventListener('touchend', stopDragging);}if (slider) {var startDrag = function(clientX) {isDragging = true;handleMove(clientX);window.addEventListener('mousemove', onMouseMove);window.addEventListener('mouseup', stopDragging);};var startTouchDrag = function(clientX) {isDragging = true;handleMove(clientX);window.addEventListener('touchmove', onTouchMove, { passive: false });window.addEventListener('touchend', stopDragging);};slider.addEventListener('mousedown', function(e) {e.preventDefault();startDrag(e.clientX);});slider.addEventListener('touchstart', function(e) {e.preventDefault();startTouchDrag(e.touches[0].clientX);}, { passive: false });}var expandBtn = wrapper.querySelector('.fv-image-compare-expand-btn');var closeBtn = wrapper.querySelector('.fv-image-compare-close-btn');if (expandBtn) {if (window !== window.parent) {expandBtn.style.display = 'none';} else {expandBtn.addEventListener('click', function(e) {e.stopPropagation();wrapper.classList.add('fv-image-compare-fullscreen');document.body.style.overflow = 'hidden';if (fgImage && fgImage.dataset.highresSrc) {fgImage.src = fgImage.dataset.highresSrc;fgImage.removeAttribute('srcset');fgImage.removeAttribute('sizes');}if (bgImage && bgImage.dataset.highresSrc) {bgImage.src = bgImage.dataset.highresSrc;bgImage.removeAttribute('srcset');bgImage.removeAttribute('sizes');}});}}if (closeBtn) {closeBtn.addEventListener('click', function(e) {e.stopPropagation();wrapper.classList.remove('fv-image-compare-fullscreen');document.body.style.overflow = '';updateTransform();});}document.addEventListener('keydown', function(e) {if (e.key === 'Escape' && wrapper.classList.contains('fv-image-compare-fullscreen')) {wrapper.classList.remove('fv-image-compare-fullscreen');document.body.style.overflow = '';updateTransform();}});});var hotspots = chartWrapper.querySelectorAll('.fv-stl-hotspot-btn');var allProductsModal = chartWrapper.querySelector('.fv-stl-all-products-modal');var shopAllBtn = chartWrapper.querySelector('.fv-stl-shop-all-btn');var allProductsList = chartWrapper.querySelector('.fv-stl-all-products-list');var stlContainer = chartWrapper.querySelector('.fv-stl-container');function closeAllModals() {if (allProductsModal) {allProductsModal.classList.remove('is-active');var items = allProductsModal.querySelectorAll('.fv-stl-all-products-item');items.forEach(function(item) {item.classList.remove('is-highlighted');});if (stlContainer) {setTimeout(function() {if (!allProductsModal.classList.contains('is-active')) {stlContainer.style.minHeight = '';if ('parentIFrame' in window) {window.parentIFrame.size();}}}, 300);}}hotspots.forEach(function(btn) { btn.setAttribute('aria-expanded', 'false'); });if ('parentIFrame' in window) {window.parentIFrame.size();}}hotspots.forEach(function(btn) {btn.addEventListener('click', function(e) {e.stopPropagation();var hotspotId = btn.getAttribute('data-hotspot-id');var isExpanded = btn.getAttribute('aria-expanded') === 'true';closeAllModals();if (!isExpanded && allProductsModal) {btn.setAttribute('aria-expanded', 'true');allProductsModal.classList.add('is-active');var container = btn.closest('.fv-stl-container');if (container && container.offsetHeight < 450) {container.style.minHeight = '450px';}var targetItem = allProductsModal.querySelector('.fv-stl-all-products-item[data-product-id="' + hotspotId + '"]');if (targetItem) {targetItem.classList.add('is-highlighted');setTimeout(function() {targetItem.scrollIntoView({ behavior: 'smooth', block: 'center' });}, 100);}if ('parentIFrame' in window) {window.parentIFrame.size();}}});});if (shopAllBtn && allProductsModal) {shopAllBtn.addEventListener('click', function(e) {e.stopPropagation();closeAllModals();allProductsModal.classList.add('is-active');var container = shopAllBtn.closest('.fv-stl-container');if (container && container.offsetHeight < 450) {container.style.minHeight = '450px';}if ('parentIFrame' in window) {window.parentIFrame.size();}});}if (allProductsModal) {var closeAllBtn = allProductsModal.querySelector('.fv-stl-all-products-close');if (closeAllBtn) {closeAllBtn.addEventListener('click', function(e) {e.stopPropagation();closeAllModals();});}}chartWrapper.addEventListener('click', function(e) {if (!e.target.closest('.fv-stl-all-products-content')) {closeAllModals();}});if (allProductsModal) {allProductsModal.addEventListener('click', function(e) {if (!e.target.closest('.fv-stl-all-products-content')) {closeAllModals();}});}var iaNodes = chartWrapper.querySelectorAll('.fv-ia-node-button');var iaWrapper = chartWrapper.querySelector('.fv-ia-wrapper');var originalCaption = chartWrapper.querySelector('.fv-original-caption') || captionEl;var dynamicCaption = chartWrapper.querySelector('.fv-ia-dynamic-caption');var exploreBtn = chartWrapper.querySelector('.fv-ia-explore-btn');var currentIaIndex = -1;function closeAllIANodes() {iaNodes.forEach(function(btn) { btn.classList.remove('is-active'); });if (originalCaption) originalCaption.style.display = 'block';if (dynamicCaption) dynamicCaption.style.display = 'none';}function resetExploreBtn() {currentIaIndex = -1;if (exploreBtn) {var exploreSpan = exploreBtn.querySelector('span');if (exploreSpan) exploreSpan.textContent = 'Explore';}}iaNodes.forEach(function(btn, index) {btn.addEventListener('click', function(e) {e.stopPropagation();var isActive = btn.classList.contains('is-active');closeAllIANodes();if (!isActive) {currentIaIndex = index;if (exploreBtn) {var exploreSpan = exploreBtn.querySelector('span');if (exploreSpan) exploreSpan.textContent = 'Next';}btn.classList.add('is-active');if (dynamicCaption) {var title = btn.getAttribute('data-title') || '';var desc = btn.getAttribute('data-desc') || '';dynamicCaption.innerHTML = '';var strongTag = document.createElement('strong');strongTag.textContent = title;dynamicCaption.appendChild(strongTag);if (desc) {dynamicCaption.appendChild(document.createTextNode(' - ' + desc));}if (originalCaption) originalCaption.style.display = 'none';dynamicCaption.style.display = 'block';if (footerContentEl) footerContentEl.style.display = 'block';}} else {resetExploreBtn();}});});if (exploreBtn) {exploreBtn.addEventListener('click', function(e) {e.stopPropagation();if (iaNodes.length === 0) return;var nextIndex = currentIaIndex + 1;if (nextIndex >= iaNodes.length) {closeAllIANodes();resetExploreBtn();} else {currentIaIndex = nextIndex;var targetBtn = iaNodes[currentIaIndex];if (targetBtn) {if(targetBtn.classList.contains('is-active')) {targetBtn.click();}targetBtn.click();}}});}if (iaWrapper) {iaWrapper.addEventListener('click', function(e) {if (!e.target.closest('.fv-ia-node-button') && !e.target.closest('.fv-ia-explore-btn')) {closeAllIANodes();resetExploreBtn();}});}window.fvAnimateCharts(chartWrapper);var countdownContainer = chartWrapper.querySelector('.fv-countdown-container');if (countdownContainer) {var targetDateAttr = countdownContainer.getAttribute('data-target-date');if (targetDateAttr) {var targetDate = new Date(targetDateAttr);var primaryColor = countdownContainer.getAttribute('data-primary-color') || '#f97316';var subheadColor = countdownContainer.getAttribute('data-subhead-color') || '#ffffff';var pad = function(n) { return (n < 10 ? '0' : '') + n; };var updateCountdown = function() {var difference = +targetDate - +new Date();var d = 0, h = 0, m = 0, s = 0;if (difference > 0) {d = Math.floor(difference / (1000 * 60 * 60 * 24));h = Math.floor((difference / (1000 * 60 * 60)) % 24);m = Math.floor((difference / 1000 / 60) % 60);s = Math.floor((difference / 1000) % 60);}var daysEl = countdownContainer.querySelector('[data-time="days"]');var hoursEl = countdownContainer.querySelector('[data-time="hours"]');var minsEl = countdownContainer.querySelector('[data-time="minutes"]');var secsEl = countdownContainer.querySelector('[data-time="seconds"]');if (daysEl) daysEl.textContent = d;if (hoursEl) hoursEl.textContent = pad(h);if (minsEl) minsEl.textContent = pad(m);if (secsEl) secsEl.textContent = pad(s);};updateCountdown();setInterval(updateCountdown, 1000);}}}if (false) {var slideshowContainer = document.getElementById(uniqueId + '-slideshow');if (slideshowContainer) {var slides = slideshowContainer.querySelectorAll('.fv-slide');slides.forEach(function(slide) {setupWrapper(slide.querySelector('.fv-chart-wrapper'));});}} else {setupWrapper(root);}}if (document.readyState === 'loading') {document.addEventListener('DOMContentLoaded', function() { initialize('fv-chart-1785251267479-g474gdird', false); });} else {initialize('fv-chart-1785251267479-g474gdird', false);}})();How to avoid VPN auto-renewal traps and surprise charges

It's clear that major VPN providers need to do more to ensure fair and transparent pricing and billing. Thankfully, there are also a number of steps you can take to make sure you're not unnecessarily impacted.

To avoid the billing and pricing issues referenced here:

• Turn off auto-renewal immediately after signing up: Navigate straight to your account dashboard after subscribing and turn off auto-renew.
• Prepare for the ‘early billing’ window: Many VPN providers trigger automatic renewal charges 7 to 14 days before your subscription technically expires. Set calendar reminders two weeks before your renewal date so you aren't caught off guard.
• Keep an eye on your inbox: VPN providers will contact you to warn you about the auto-renewal and are likely to send you an email about extending your discounted rate around 60 days ahead of your contract expiring.
• Always request an official cancellation first: Submit an explicit cancellation request through the provider's official account portal or support chat.
• Check refund terms: Many ‘free trials’ require upfront payment details and automatically convert into full recurring plans. Read the fine print before providing billing information, and remember that 30-day money-back guarantees often only apply to first-time purchases and are void during certain promotional offers.

What did the VPN companies say?

We contacted each VPN provider for comment on our findings.

Representatives from NordVPN, Surfshark and ExpressVPN highlighted their strong overall Google Play ratings and stressed that user feedback informs ongoing app development.

ExpressVPN added that subscriptions bought via Google Play are subject to Google’s payment and refund policies, but noted its focus is on making the experience "easier to understand, manage, and resolve when customers need help."

A NordVPN spokesperson argued that its premium pricing reflects ongoing infrastructure investment, but pointed out that because the app is free to download, "some users expect the service itself to be free as well," which can drive down review scores when users discover it is a paid product.

A Surfshark spokesperson acknowledged that "public reviews naturally tend to overrepresent moments of friction" but said its current approach is "resonating positively with users" overall.

Meanwhile, Proton VPN emphasized that its paid subscriptions fund an unlimited, ad-free tier for all users, adding that it "clearly states its pricing structure, upfront costs, and terms of renewal" without using tiered feature paywalls or surprise price increases.

Methodology

We collected almost 30,000 user reviews published on Google Play Store since the beginning of the year for the four VPN providers featured in this report using the google_play_scraper library.

As individual reviews often address multiple topics (e.g. streaming and security), we broke reviews down into sentence-level units. This expanded our analysis to over 47,000 entries, with reviews permitted to sit across multiple categories when required.

Each sentence was assigned to specific categories using regex keyword filtering, and analyzed for sentiment using a pre-trained BERT model that took into account the user’s star rating.

While all machine learning sentiment pipelines carry a margin of error, every VPN provider was subjected to the exact same pipeline to ensure consistency and fair comparison. Human review was also conducted throughout.

Data processing scripts were developed in Python with assistance from an LLM, and all outputs were manually reviewed.

Formula 1 is close to reaching 1 billion fans across the globe, rising from 12% year over year from 2024 to 2025., with the current figure standing somewhere around 800 million.

To try and capture some of this huge potential market, the TGR Haas team has partnered with Infobip, resulting in a novel way to stay engaged with fans while funding car development through merchandise marketing.

RaceMate is an AI-powered fan companion that provides conversational updates on the team’s grid positions, race and qualifying outcomes, and team race intelligence.

(Image credit: Benedict Collins / Future)Gamified engagement across the racing schedule

At Silverstone, the home of the British Grand Prix, I spoke to Michael Heath, Senior Fan Engagement & CRM Manager at TGR Haas F1 Team, and Ante Pamuković, Chief Revenue Officer at Infobip, to discuss how RaceMate has changed the face of fan engagement throughout the 2026 season.

Following the launch, Haas ran several highly successful fan quizzes through the Racemate chat - but Heath and the team noticed that fans were engaging in an entirely unexpected way.

“We actually started to see that fans were then asking this WhatsApp channel questions, but at the time we didn't have any capabilities built into it that would respond,” he explained. “That was ultimately what fans wanted to do, how they wanted to interact, that's kind of where RaceMate was then born from.”

Racemate runs on Infobip’s AgentOS platform, which has allowed Haas to set guardrails on the topics of conversation users can begin, and helps steer fans back on to relevant topics about TGR Haas.

“This is where we saw actually the spike of more and more people joining the conversations and actually more of them returning,” said Pamukovic. “We saw up to 30% of returning fans to the platform, which is quite high.”

A challenge TGR Haas is attempting to counter is the banning of social media for under 16s, as many Formula 1 fans use social media as their first point of contact for engaging with teams and drivers.

“We need to have a place where, one: it's safe for them to be, but two: they can still get all the information that they're used to getting,” Heath explains.

“When you look at short form video content that they can get on Instagram right now, can we now start delivering that to them through Racemate because it's a safer platform where there's not all of the comments and negativity which is associated to it, but it's still on something which they feel like they own and operate and isn't just a commercial brand shouting at them loudest,” he adds.

(Image credit: Benedict Collins / Future)

RaceMate is available directly through the WhatsApp Business Platform and Apple Messages for Business.

“It's on a communication channel that you already have effectively, rather than making you download something else and remembering to go there,” Heath explains.

One of the most important parts of fan engagement is creating a revenue stream without making it cost to be a fan. “Race mate for us is great because it's a very low barrier of entry,” Heath says. “Once they're then in and they can start engaging, then we can start capturing their data, and then once we've got the data we can then start to build that 1 to 1 relationship, start to build that profile type.”

“I think that's gonna have more waiting for commercial value rather than charging the fan directly,” he adds.

Pamuković concurs with Heath’s point. “We see all of these different components, different silos merging into one, and being able to create the data and then use it to the best possible extent. This is the way forward. Understanding the interactions, the fans, the behaviour, what drives them, what motivates them, and then of course how to best serve them.”

To that end, Racemate includes a feature that allows users to virtually try on team merchandise before making a purchase - something I very much enjoyed testing.

A demonstration of RaceMate's virtual try on feature.Benedict Collins / FutureA demonstration of RaceMate's virtual try on feature.Benedict Collins / FutureRaceMate’s future

Looking into the future, Heath has big plans for Racemate. “I want Racemate to be the second screen. So if you are watching the race or you're not in the house and you want to know what's going on in the race and you want to get closer to the action, I want Racemate to become that home.”

Racemate also has the potential to help new fans learn more about race strategy, and the general rules surrounding Formula 1.

“Can we basically use Race Mate to unlock that conversation in fairly simple terms so that a fan can start to understand what they're looking at on a pit wall? So that a fan can start to understand why they've made a pit stop at a certain point? What window are they trying to come back out into?” Heath asks.

“Because I think to truly understand the sport you need to understand all of those touch points.”

For quite a long time now, I've been concerned about how Microsoft is going to monetize Windows going forward. While the world's dominant desktop operating system remains available for a one-off fee (assuming you aren't an upgrader who can get it for free), there's a good chance this could change in the future.

Why do I think that? To me, it feels somewhat inevitable that eventually, Microsoft is going to look for a way to shift Windows from an upfront payment to a subscription model for consumers. That regular monthly income stream piling up in the coffers is the end game for most big tech companies and their products these days for good reasons in terms of the profits to be made.

Of course, rumors about Microsoft looking to charge a subscription fee for using Windows to consumers (as opposed to businesses, where there's already a subscription option) have been floating around for years. True, they've all been dubious and sketchy in nature, or indeed proven outright incorrect, but this idea keeps bubbling up, and I believe we're witnessing a development with AI right now that indicates how Microsoft might have an ideal opportunity to make this pivot at some point down the road.

Last week, Windows Central spotted that the Copilot feature Deep Research is getting the axe, to be replaced with a new feature: Researcher. The idea behind both pieces of functionality is roughly the same (researching and creating detailed reports complete with citations), but the difference is Deep Research was free, whereas Researcher requires a subscription to Microsoft 365 Premium.

This follows Microsoft Whiteboard (where AI assists you in brainstorming ideas) getting changed so personal accounts can no longer use it — the app now requires being signed up for a Microsoft Business account. On top of that, Outlook's Meeting Insight functionality was just shifted behind a paywall, being transformed into an admittedly beefier AI feature, but one that needs a Microsoft 365 Copilot license.

So, there's a trend towards turning previously free AI features into paid ones, presumably as Microsoft rejigs its Copilot offerings and figures out what works well — and what's being used — and whether any of that can be charged for.

AI pivot

(Image credit: Microsoft)

Now, we all know AI agents are going to be the 'next big thing' (TM) in Windows 11, certainly if Microsoft has its way, and the idea is for these AI entities to be doing more and more within the OS.

We won't just have an agent for changing Windows settings — and I mean a proper incarnation of this already existing idea, which really can adjust a host of options based on a simple request to "make my laptop battery last longer" or similar — but we will have a small crowd of them. Maybe a troubleshooting agent, for example, which can take a problem that you're battling in Windows and use some genuine AI smarts to help solve it. Or perhaps a creativity agent which can tackle a host of image or video-related tasks, or more broadly help with organizing your projects and photos.

I think the catch will be that eventually, as we've seen with some aspects of AI and Copilot, Microsoft will start shuffling some of these agents behind a paywall — especially considering that more powerful capabilities like these in-depth AI tricks will cost Microsoft a fair bit to drive in terms of cloud resources. It's certainly conceivable that Microsoft could end up charging a small monthly fee to use these premium agents, maybe as separate add-ons in the hope that you'll bolt more of them onto Windows for a cumulatively greater benefit to its coffers.

We could ultimately be looking at a kind of modular, AI-focused OS, and to me, this seems like the easiest way for Microsoft to transform Windows into a subscription model, because it'll happen slowly. You don't need a troubleshooting AI agent to run Windows at all, but it'll be a nice thing to have — certainly for less tech-savvy types — in case problems do arise in the OS.

As more bits and pieces that might start off free drift behind a paywall — as these AI features are built up and become more compelling — people may eventually be tempted to buy a bundle of them. And before you know it, you've signed up for a Windows subscription of sorts (although the free version of the OS will, of course, still be available).

In the end, there may be a couple of bundles — tiered subscriptions, in other words — and let's not forget Microsoft's potential ambitions for a cloud PC model for consumers, either.

This has been a possibility raised in past rumors and leaks (and again, this is something which has already happened in the enterprise world), and if you put all this together, you're looking at a kind of Netflix model, if you will: a streamed OS with several paid subscription tiers. Albeit with a free basic tier for Windows – although who's to say that may not become ad-supported, or indeed more ad-supported, as Windows 11 already does a fair old line in adverts and promos. (Although admittedly Microsoft is cutting back on that front in its crowd-pleasing efforts to fix Windows 11).

Not a foregone conclusion — but a likely enough prospect

(Image credit: MAYA LAB / Shutterstock)

This is just my opinion, naturally, and yes, maybe I got rather carried away with the extrapolation at the end there. It's also true that some big question marks remain hanging over the wider notion I've put forward.

As I just mentioned, Microsoft is very much bending over backwards to please Windows 11 users right now, so any possible timeline for this shift may be pushed way back in view of that. Bringing in some form of subscription is hardly going to be a well-received move, even if implemented in small, delicate increments as I'm guessing it would be.

The other obvious sticking point is that Microsoft needs to make its AI agents in Windows worth having. They need to be objects of desire and come packing genuinely useful AI abilities, otherwise clearly, people won't pay for the privilege of having them on their Windows desktop. They also need to be secure and trustworthy so they don't end up throwing spanners in the works of your Windows installation.

That could be the biggest hurdle for Microsoft to overcome, because as it stands, when the idea of AI features being paywalled in Windows 11 has been raised (in rumors and the like), it's been actively welcomed by the more skeptical out there. The cynics are more than happy to have everything AI-related locked away from them, and as non-paying users, this would give them an AI-free Windows 11 desktop.

Again, this plays into any potential move along these lines having to happen further into the future, but I think Microsoft does have this as an eventual goal. If you ask yourself the question: if Microsoft could charge a subscription for Windows 11, would it? The answer is clearly yes, a thousand times over. But the actual, real question here is whether Microsoft thinks it could successfully get away with such a plan without sparking a large-scale defection from its desktop OS.

We can keep our fingers crossed that a monthly charge for Windows isn't coming, but frankly, I think it's likely. Or even just a matter of time — perhaps a lot of time, granted — whether that subscription pertains to AI add-ons, or a cloud PC offering, or Microsoft finds another way to spin this.

I didn't realize I'd developed an unhealthy obsession with wellness and optimization. But looking back, there were definitely signs. This was more than a decade ago, long before recovery scores, longevity influencers and what the best smart ring colors are became water cooler conversation.

As a technology journalist, reviewing health gadgets and fitness trackers has always been part of my job but they gradually spilled over into my personal life too. At one point, I was wearing multiple fitness trackers at the same time because I didn't fully trust any single device. I'd compare the data, then transfer it into spreadsheets each night so I could better analyze it myself.

Most days I walked more than 20,000 steps. Some days it was closer to 30,000 and I got such a kick out of seeing those numbers climb up. I became fascinated by the idea that there was an optimal way to do almost everything. An optimal diet, optimal morning routine, optimal supplement stack and an optimal sleep schedule.

Every new wellness trend arrived with the promise of hidden knowledge and I was eager to believe it. I spent money on retreats, courses and gadgets. I followed advice that ranged from questionable to ridiculous. And the less said about the gruelling fasting retreat where we were all given daily wheatgrass enemas, the better.

The strange thing is that none of this felt unhealthy at the time — if anything, it felt virtuous. My trackers congratulated me for hitting goals and fitness apps handed out badges and streaks. There was always another target to hit and another metric to improve. The obsession disguised itself as self-improvement so effectively that I barely questioned it.

What I didn't understand at the time was how thin the line between discipline and obsession can be. You can cross it gradually, one habit and one goal at a time, until something that started as a genuine attempt to look after yourself becomes another source of pressure, anxiety and control.

More than a decade later, as wearables, recovery scores and optimization culture have since become mainstream, I don't think my experience is unusual. If anything, I'm surprised we don't talk more about the psychological cost of constantly measuring ourselves and the role technology plays in keeping us focused on the numbers.

Why health data can become a trap

(Image credit: Samsung)

For a long time, I assumed my experience was unusual. When my obsession with optimization was at its worst, there weren't podcasters talking about longevity-maxxing and relatively few people owned fitness trackers.

Over time, I managed to develop a healthier relationship with exercise and health data. But as wearables have become more common and self-tracking has moved into the mainstream, I've started noticing some of the same patterns I once recognized in myself.

When I spoke to psychotherapist Sarah Dosanjh, who works with people experiencing health anxiety and disordered relationships with food, much of what I'd experienced sounded familiar.

"What I'm noticing in my practice is that people who already have some form of anxiety seem particularly drawn to health data technology devices," she says. "An anxious client seeks certainty and control and health information feeds a sense of control, making it very appealing to an anxious person."

Looking back, some of the periods when I was most immersed in tracking and optimization were also periods when other parts of my life felt uncertain or difficult. The data gave me something concrete to focus on. It offered numbers, targets and routines at times when everything else felt far less predictable.

Dosanjh explains the irony is that tools designed to help people feel more in control can backfire. "The most common devices I see people struggling with are food and exercise tracking apps and continuous glucose monitoring devices," Dosanjh says. "But what starts as feeling in control can quickly turn into feeling controlled by the technology."

She describes people becoming increasingly preoccupied with maintaining specific numbers and avoiding anything that might disrupt them.

"Closing exercise rings, achieving a certain step count and keeping blood sugar within a specific range can turn into a compulsion,” she tells me. “Anxiety peaks at the mere thought of not hitting numbers. What feels like a good thing to do for your health can become a source of intense anxiety instead."

These behaviors rarely look unhealthy from the outside. Going for a walk, exercising regularly or paying attention to what you eat are generally considered positive things. The difficulty is knowing when useful habits become rigid rules and when health stops being something that supports your life and starts becoming the thing your life revolves around.

Dosanjh says that cycle can become self-reinforcing. "This situation becomes more dangerous when the person believes they can manage this added anxiety by setting even higher goals,” she explains. “They are chasing the initial relief and dopamine hit that they experienced early on in their tech use. It can become an addictive trap, driving people further into disordered eating and compulsive exercise."

I think that’s what makes optimization culture so difficult to talk about. For some people, tracking genuinely is helpful. It can encourage movement, show them useful patterns and provide much-needed motivation via streaks and digital rewards, such as badges and kudos from other users. For others, particularly those already vulnerable to anxiety, perfectionism or compulsive tendencies already, the same tools can pull them further down a path they may not even realise they're on.

When tracking went mainstream

(Image credit: Future)

When I first started reviewing fitness trackers, this kind of behavior felt relatively niche. Most people weren't tracking their sleep. Very few people knew what heart rate variability was. The idea of waking up and checking a readiness score before deciding how hard to exercise would have sounded bizarre. But today, optimization is everywhere.

More than 45% of people in the UK and around 60% of people in the US now own a smartwatch or fitness wearable. Even if you don't actively seek out health tracking, many of the features, like step counts and calorie estimates, are now built directly into devices like Apple Watches or smartphones that we all use every day.

Despite this, researchers are still trying to understand the psychological impact of living with a constant stream of biometric data. There's no clear evidence that large numbers of wearable users are developing serious problems. But a growing body of research does point towards some worrying patterns. Several studies have linked fitness tracking technologies with increased anxiety, body dissatisfaction and rumination. Others have found associations between wearable use and higher levels of obsessive-compulsive traits, like perfectionism and over-conscientiousness.

I found that one study even introduced a new term: technohypochondria. Researchers defined it through three features: biometric data obsession, digital catastrophizing and a pathological need for algorithmic feedback. Unlike traditional health anxiety, which tends to focus on illness itself, technohypochondria describes a dependence on the continuous flow of personalized health data and the reassurance it appears to provide. I guess I’m a recovering technohypochondriac?

What I found particularly interesting was that researchers working on one of the key studies raised a really important question: “are people with pre-existing compulsive tendencies more drawn to wearables, or do continuous tracking and feedback loops foster or exacerbate these tendencies?”

It’s a chicken-and-egg sort of question and the answer is probably complicated. But the researchers do suggest a feedback loop may exist, where pre-existing vulnerabilities and constant self-tracking reinforce one another.

None of this means that everyone who wears a smartwatch is heading towards a crisis. I know personally that I did already have compulsive tendencies and controlling behaviors with food long before Fitbit released its first device. But I think it does suggest that the emotional impact of optimization deserves more attention than it often receives. Because once tracking becomes woven into everyday life, it can be surprisingly difficult to tell where useful information ends and unhealthy fixation begins.

When the numbers become the point

(Image credit: Mile Atanasov / Shutterstock)

When I asked my friends and followers on social media to share their experiences with wearables and health tracking, I expected a handful of disparate stories. Instead, I heard from people who described becoming trapped by numbers in ways that felt extremely familiar.

Some were dealing with chronic illness. Others were recovering from major health events. Some simply wanted to get fitter or sleep better. But again and again, there was the same pattern. What began as a search for reassurance gradually became another source of anxiety.

I spoke to Emma, who started relying heavily on health data after cancer treatment and surgical menopause left her worried about the long-term impact on her health. "My watch became reassuring," she tells me. "I thought if my heart rate looks normal, I'm probably okay."

But over time that reassurance became its own form of dependence. "It felt like I was controlling my health anxiety," she says. "But I think I was just making it worse."

Sarah, who has used wearables for years while managing endometriosis and dysautonomia, described a different concern. "I wake up and let Oura tell me if I've slept well and then Visible gives me a score for the day, and that just feels wrong," she says. What impacted her most was the lack of trust she had in her own judgement. "I think it's made me lose trust in myself a little. Am I actually tired and am I really that stressed because an app said so?"

I think that gets to the heart of what can make continuous tracking so complicated. The problem isn't necessarily that you’re looking at data a lot — the problem is what happens when the data becomes more important than your own lived experience.

Understanding the cycle

(Image credit: Shape Pilates founder Gemma Folkard )

Even though all of the stories I heard about wearables were different, a common thread was just how many seemed to be born from a feeling of trying to manage uncertainty.

Many of the people who contacted me weren't trying to become superhuman, chase longevity records or optimize every minute of their lives when they first started tracking. They were dealing with health scares, chronic illnesses, anxiety, weight issues and burnout.

And the data that their wearables collected offered reassurance. But the problem is that reassurance doesn't tend to last.

"Many of my clients understand that something doesn't feel quite right about their technology use, but they are afraid to stop using it," Dosanjh tells me.

She often explains this through what psychologists call the anxiety cycle. It starts with uncertainty. Am I healthy? Am I eating the right thing? Am I doing enough? The device provides an answer, whether that's a sleep score, a heart rate reading or a closed exercise ring. The anxiety temporarily decreases and the brain learns that checking the data creates relief.

But, over time, that starts to fade. "The temporary relief that comes from hitting a target or seeing a reassuring number keeps feeding the illusion of mastery over your health," Dosanjh explains. "Helping clients understand that what they are seeking in the tech is certainty, and since complete certainty is impossible, the focus of our work moves from trying to eliminate anxiety to developing the capacity to tolerate uncertainty."

That really describes my own experience. The more I tracked, the more I felt I needed to track. The more information I had, the more information I wanted. The pursuit of health slowly became the pursuit of certainty, which, as we all know logically, isn’t something you can ever get, achieve or “win” at.

What actually helped me

A big part of getting better was realizing that my obsession with health data wasn't really about the health data at all.

Looking back, some of the periods when I was most preoccupied with optimization were also periods when other parts of my life felt difficult, uncertain or out of my control. The trackers gave me something concrete to focus on. There was always another metric to improve, another target to hit, another problem that seemed solvable.

But many of the things I was actually struggling with couldn't be fixed with a spreadsheet or a sleep score. Therapy helped me recognize that pattern. So did learning to tolerate uncertainty a little better. Over time, I became less interested in controlling every variable and more interested in understanding why I felt the need to control them in the first place.

That doesn't mean the tendency completely disappeared, I still recognize it in myself from time to time. The difference is that now I see it for what it is and can catch the early warning signs.

Should wearables be designed differently?

(Image credit: Future/Garmin)

I don't think every fitness tracker needs to be redesigned around people like me. The most effective intervention for my own unhealthy behavior was surprisingly simple: I took all the devices off. But some researchers argue that the design of wearables does deserve more attention.

One of the studies I looked at found that some of the negative psychological effects associated with fitness technologies could be linked to the features themselves. Feedback systems, gamified rewards, social comparison tools, constant notifications and the stream of immediate statistics can all encourage people to engage more frequently with the data, sometimes in ways that become unhelpful.

The researchers believe that psychological wellbeing should be treated as an important measure of success alongside more familiar metrics, like engagement, accuracy and battery life.

That doesn't necessarily mean removing goals or progress tracking. But it could mean giving people more control over how they interact with the technology.

Based on what I’ve seen from years reviewing wearables, that could mean less judgemental language, fewer alarming warnings, more ways to take breaks without feeling punished, and more flexibility over what data is displayed and when.

Most wearable companies already offer some degree of customization. Yet many products are still built around the assumption that more engagement is always better. In many cases that's true; regular use makes the data more useful over time because you can see patterns and track trends. But there should always be room for people to step back when they need to. A healthy relationship with wearable technology shouldn't require constant engagement, and users shouldn't feel punished for taking a break.

Health should improve your life, not become your life

The irony is that I became interested in health and fitness because it genuinely helped me. I learned from a really young age that exercise improved my mood, moving more reduced my anxiety and looking after myself made life feel calmer and more manageable. But I started paying way too little attention to how I felt and too much attention to what the numbers said.

These days, I still review fitness technology and sometimes still wear trackers outside of work. But I pay attention to different things now. If I feel guilty taking a device off, that's a warning sign. So is chasing a target despite being exhausted and spending more time thinking about the data than paying attention to my actual experience.

But most importantly, I ask myself what else is going on. Because when I feel like I’m becoming overly fixated on optimization, there's often something else happening underneath. Data is usually not the cause. It's just where all that energy and anxiety ends up being funnelled.

Dosanjh encourages people to approach health data as information rather than instruction. "Health tech should be an enhancement in your life and not an additional source of stress," she says. "Prioritize your wellbeing over optimization."

She also encourages people to regularly check whether the technology is still serving the purpose they originally bought it for. "How you feel is a more helpful barometer of wellness than numerical data. Be clear about your reasons for using the tech and check they align with your life values,” she says.

That's the lesson I wish I'd understood years ago. Health should improve your life but not become your life. Because no matter how sophisticated our trackers are there are still some things they can't measure. Like whether you have enough energy to spend time with the people you love, whether you're enjoying your life and whether you're actually feeling well. Those are infinitely more important than if you hit 10,000 steps today.

The AI race started off with a pretty clear direction – bigger and better. The first waves were characterized by building bigger models, but it’s all change in the world of artificial intelligence and with enterprises, SMBs and consumers all finding use cases for the technology, the focus has shifted.

Now, AI firms and model developers are looking to realize a much tougher goal. Efficiency. Cost per token, performance per watt, output per input, it’s all about driving maximum efficiency.

One clear divide is between training and inference. While training models still requires huge amounts of resources, inference efficiency is starting to improve, and one company (Rebellions) now believes an opening for inference-first hardware could create a new market.

The company’s racks are said to consume around 16-20kW, compared with around 120kW for leading GPU-based inference systems that, for many use cases, are sheer overkill.

Rebellions’ rack costs are also said to be around one-third of the price, making AI inference more accessible and helping enterprises to deploy AI more widely.

This hardware shift could be the start of truly efficient AI

Memory is also another battleground, whereby huge trillion-parameter models are testing the limits of today’s hardware and the intertwined reliance on memory and compute. Something Rebellions says it’s looking to fix by working with the likes of SK Hynix and Samsung to align multiple roadmaps, instead of having to respond to shifts in architecture.

Ultimately, today’s black-and-white chip manufacturing landscape is now evolving, and Rebellions sees two key changes happening simultaneously. Firstly, training and inference hardware is starting to differ more drastically. Secondly, aligning multiple hardware roadmaps across memory, compute and more will drive more efficiency not just across deployments, but in terms of bringing new products to market.

I spoke to Rebellions CEO Sunghyun Park allows me to understand how and why inference and training hardware are starting to separate, as well as the importance of open standards and collaboration in the drive for all-round efficiency.

• The AI chip market seems to be splitting between training-first and inference-first architectures. Why is that happening, and why now?

This split exists because training and inference are fundamentally different problems.

Training is how you build a model. It happens once, involves a small number of organizations, and rewards raw computational flexibility because the workload keeps shifting as research moves forward.

Inference is how you actually use a model: every query answered, every transaction processed, every decision an AI system makes in production. That happens billions of times a day across nearly every industry, and it’s where AI moves from R&D into revenue.

Those two jobs need different physics. Training requires maximum FLOPS. Inference requires efficiency, reliability, and economics that hold up when you’re serving users at scale.

The industry forced a training chip into that second job because that’s what existed. Now that inference has become the larger, more urgent market, that compromise no longer holds. Enterprises and governments are asking how fast they can deploy. That’s why the conversation is splitting now.

• Where does Rebellions fit in that split, and what makes your approach to AI inference fundamentally different from your competitors?

We built for inference, from day one. Most first-generation AI chip companies emerged from the 2016-2017 training boom and adapted their architectures for inference afterward.

We started in 2020 – after that wave – with inference as the only target, which meant designing around what production AI actually needs instead of retrofitting a training chip.

The numbers reflect that choice. Our racks draw 16-20kW versus roughly 120kW for leading GPU-based inference systems, about a sixth of the power, in a market where power is the binding constraint for most operators.

Acquisition cost runs around $10 million per rack versus roughly $30 million, about a third of the cost. Our chiplet-based architecture also scales out rather than betting that a single device can handle a model’s full size, which matters now that production workloads are trillion-parameter mixture-of-experts models instead of the few-hundred-million-parameter models the first generation was built around.

It’s also why our architecture is memory-centric rather than compute-centric. The chiplet approach exists to keep memory close to logic as models scale, not just to add cores.

And we have three years of production deployments behind that architecture, not pilots. That’s the hardest part to replicate: real workloads, running at scale, today.

• A year ago, everyone in AI was talking about chiplets. Now the conversation has shifted to memory. What changed?

The chiplet conversation was about architecture: breaking a chip into modular pieces that scale independently, rather than betting everything on a single monolithic die.

That mattered because it let the industry move past an assumption the first generation of accelerators made in 2016 and 2017, that a single device would always be big enough to run any model.

That assumption broke once mixture-of-experts and trillion-parameter models arrived.

The memory conversation is the layer underneath that. Once the architecture problem is solved, the constraint becomes physical: can you actually get enough high-bandwidth memory (HBM) to build what you’ve designed?

HBM is 3D-stacked memory, and how closely you can physically stack it to compute is as much of a bottleneck as raw supply.

Every AI accelerator company is competing for the same limited supply right now, and demand has outpaced what memory makers can produce. That’s the memory-logic co-design problem: architecture and memory supply are no longer separable decisions.

We’re in a different position because our investor relationships were built around supply, not just capital. Our memory partners are also investors, and we co-design our memory architecture directly against their roadmaps rather than simply purchasing off them.

Our chiplet architecture also develops against our foundry partner’s process roadmap. When the rest of the industry was fighting for allocation, we already had a seat at the design table through those relationships.

That’s memory-logic co-design, not just secured supply. The shift from chiplets to memory tracks has moved the real constraint: from architecture to physical supply.

• Your stack runs on open-source frameworks like vLLM, PyTorch, Kubernetes, and OpenShift, tools many enterprises already use. Does that make adoption relatively plug-and-play, or is that an oversimplification?

It’s mostly true, but ‘plug-and-play’ undersells how deliberate that was, and oversimplifies in one specific way.

We built entirely on open standards: vLLM, PyTorch, Kubernetes, and Red Hat OpenShift. We’re one of only two chip companies in the PyTorch Foundation, and the only AI accelerator company fully integrated with OpenShift.

A developer who already knows how to run inference on existing infrastructure already knows how to run it on ours. There’s no proprietary runtime to learn and no migration project. That part really is close to plug-and-play.

The first generation of AI chip companies each built proprietary software stacks, and hundreds of millions of dollars went into software that didn’t survive. We came to market once the open source ecosystem had matured and chose to build on it instead of forking it.

Where it oversimplifies is assuming that means zero integration work. Production deployments still require validating performance at your specific workload and scale, and that takes real engineering time, no matter how compatible the stack is.

What open standards remove is lock-in risk and retraining cost, not the deployment work itself.

• What are the biggest challenges organizations face when trying to run AI inference outside of hyperscaler platforms?

Most organizations aren’t built like hyperscalers, and much of the available inference infrastructure assumes they are.

The first challenge is physical. Most enterprises, telcos, and governments already have data centers. They can’t wait two to three years or spend $600 million-plus on new ones, and they can’t retrofit for liquid cooling without major cost and disruption.

So the practical question is whether inference hardware runs on what they already have: standard racks, air cooling, and existing power budgets.

The second is sovereignty. Organizations increasingly want to bring compute to where their data already lives, rather than move sensitive data to wherever compute is hosted.

That’s partly regulatory, partly operational, but either way, cloud-only inference creates a dependency a lot of operators are no longer comfortable with.

We built specifically for that gap. Our systems run at 4-5kW per server on standard air-cooled infrastructure, no facility redesign required.

SK Telecom has run on our hardware for nearly three years, scaling from a small cluster to close to 100 racks and now processing 50 million API transactions a day, entirely inside their existing network.

KT Cloud runs real-time inference on highway CCTV systems nationally. Both show you don’t need hyperscaler-scale infrastructure to run AI at hyperscaler-relevant volume.

• Cerebras' IPO, and potential listings from others in the space, have put AI infrastructure in the spotlight. What do these IPOs signal about the market, and where might the hype be outpacing deployment reality?

It tells the public markets that AI infrastructure is a durable, investable asset class, not just a venture-backed bet. That validation benefits the whole sector, including us: it says purpose-built AI silicon is real, differentiated from general-purpose GPUs, and worth independent capital.

Capital flowing into the sector is necessary, but it doesn’t by itself determine who wins. The companies that define the next decade of this market won’t necessarily be the most-funded ones.

They’ll be the ones with real fundamentals: production customers, deployment scale, proven economics, durable supply chain relationships. That’s a different filter than fundraising size, and it’s the one that matters once public market scrutiny starts.

We’ve been building toward that filter since 2020, with production deployments and supply relationships with our foundry and memory partners that we secured before the rest of the industry was fighting over the same allocation. The capital is a tailwind for everyone serious about inference.

Whether it gets deployed well is a separate question, and one the market will answer over the next few years.

• Looking ahead, what trends are you watching most closely across AI infrastructure and inference over the next 12–24 months?

The buildout happening inside existing infrastructure keeps accelerating. Most enterprises, telcos, and governments aren’t waiting for new data centers.

They’re deploying inference into facilities they already have, and I expect that to become the bigger story even though it gets less attention than hyperscaler headlines.

Additionally, memory remains the physical constraint. HBM supply hasn’t caught up with demand, and as models keep moving toward trillion-parameter mixture-of-experts architectures, that pressure increases rather than eases.

Companies with secured, strategic supply relationships will have a real advantage over the next two years, not just a cost one.

Chiplets are how we get there. We’ve already mass-produced a highly advanced 4-chiplet package – a level of integration Nvidia has struggled to reach.

Reporting this year indicated Nvidia had built and demonstrated a four-chiplet Rubin Ultra design, then canceled it in favor of a dual-die architecture over manufacturability concerns: a four-die single package pushes roughly 7.5-8x past reticle limits on yield and cost. That’s the foundation.

The next layer we’re building on is performance optimization of HMB3E (3D-stacked memory) in close collaboration with memory and compute co-designed together from the start, not bolted on after the fact.

Also on my radar is the fact that as more companies in this space go public, capital will get valued against production fundamentals rather than funding rounds.

And the efficiency point matters: as inference gets cheaper per token, demand doesn’t shrink, it expands, because new use cases become viable at lower cost. That’s been true of every computing platform in history, and I don’t expect AI inference to be the exception.

If you're looking to get ahead for college in the fall, then a VPN is one of the most sensible back-to-school staples to tick off your list now. You're going to spend a lot of time connected to campus Wi-Fi and it's often not as user-friendly as it might seem.

For sure, there can be security concerns when connecting to any form or public network, which a VPN can help with, but its best use at university is to make sure you can access all the content that you usually do at home.

That's because campus Wi-Fi can be very restrictive. These are networks that have to manage huge bandwidth demands, maintain cybersecurity, and also comply with legal requirements.

That means that they often limit access to certain usage-heavy services, and sites and apps that are more in a grey area when it comes to safety and the law. Think video streaming, online gaming and torrenting, for examples, as three activities that you might find curtailed.

But, if you're using a VPN on your device, the campus network in question won't be able to see what internet sites and services you're accessing, so it will be blind to you and your activities.

It will probably be able to see that you're using a VPN, the amount of bandwidth your device is using and what your device is but your online deeds will remain private. So, if you're headed to college this year, you might want to think about getting a VPN.

Right now, IPVanish is a good choice for college students. It does a great job of keeping your digital life private and, just as importantly, it's cheap!

Get IPVanish: $2.19 per month (that's $52.56 for 2 years)
IPVanish has long been a reputable VPN provider. It has a an audited no-logs privacy policy, good features for torrenting and will unblock streaming services such as Netflix and ESPN+ wherever you are. It doesn't have as many worldwide server locations as some of the more expensive VPNs but that isn't a problem when it comes to getting round campus Wi-Fi restrictions. Try it out with the safety of a 30-day money-back guarantee.View Deal

At $2.19 per month, IPVanish is solid, 4-star VPN and is only short of the very best VPNs on features like server count and some streaming service unblocking that's not a priority for this use case.

It does also have a unique privacy feature which may be handy for your browsing too.

IPVanish's Secure Browser is remote browser software that runs on an IPVanish cloud server. The sites and services you then navigate to have no idea about you or your device at all. They only connect with the cloud server.

That means the session cannot be linked with you at all and no trackers, cookies nor anything else can come your way. It also protects you from any malware or any other nasty things that you stumble across. Definitely worth using when you're searching the darker corners of the web.

Along with IPVanish's standard features, that should have you covered for all the college dorm internet use cases you'll have. Give it whirl.

New reports on real-world AI deployments seem to be being published almost daily, but there's one clear message which seems to span them all – shadow AI is a major problem.

The use of unapproved or unauthorized tools by workers is a common theme regardless of business size, sector or geography, and it often stems back to one or two reasons – employers are either being too prescriptive about permitted AI tools and are giving workers a narrow window of unsuitable tools to experiment with, or they lack any clear strategy altogether.

These reports have already detailed the risks in great depth, but to summarize, using consumer-grade versions of AI apps puts sensitive and confidential workplace data at risk, be it leaks or secondary exfiltration via model training. Hence why companies invest in enterprise-grade versions with additional safeguards.

Shadow AI is a symptom of a bigger problem

Too commonly, employers consider shadow AI a disease that plagues their workers. Something that should be stamped out with more effective training or harsher consequences to breaking the rules.

But the reality is that shadow AI is more often a symptom of the boarder workplace culture, and it's the cause of this that I set out to explore when speaking with industry experts and policymakers.

Canva preaches the importance of freedom of choice – the Australian software giant gives its workers full autonomy over the models they want to use, affording them the time to identify the right tools rather than being prescribed unsuitable alternatives.

Policies only work when they're accessible

Beginning with insufficient and unsuitable policies, Zendesk Chief Legal Officer Shana Simmons explained to me in an exclusive interview that many of today's agreements and policies are far too formal and field-specific.

"AI policies often fail because they’re written for lawyers, not for the people expected to follow them," she outlined, "if people can't understand the guidance, their behavior won't change."

Simmons also explained the policies are being stored behind closed doors in hard-to-reach places, like HR folders that workers never, ever check.

"If a policy is buried in a handbook or on a website, it’s not going to reach employees when they need its," she said, noting that policies should actually form part of the UI – or in other words, where the workers already are.

Canva warns us that, "the most common mistake is treating AI training as a curriculum," whereas it should really be seen as an ongoing back-burner activity that's always developed. The company's spokesperson insisted that workers learn through fixing their own problems, not by "sitting through a course on prompting."

Unsuitable tools and taking matters into their own hands

In a bid to work out whether it's employees or employers who are at fault (or whether it's shared), I asked whether shadow AI is a reflection of worker misconduct or insufficient tooling.

In response, Simmons stressed that "most people want to do the right thing," agreeing that the most common cause of shadow AI is indeed poor tooling.

"If employees are given the tools they need and are informed of the rules and requirements in a way that’s understandable to them, and technical controls are in place to restrict the riskiest behavior, I’d expect shadow AI to be no greater a problem than any other form of employee misconduct."

The answer then isn't necessarily to approve every new AI application, but understanding why users prefer certain tools over others is key to building suitable policies and safeguards around those.

In certain, low-risk conditions, shadow AI could actually be an important and useful part of feedback, showing organizations where they're falling short and exactly where to invest, but a clear oversight over this is just as important to ensure that no leaks or other threats occur.

AI literacy can't be taught – it's learned

Clearly, then, workers need more guidance and support. But does that come in the form of training, policies, access to tools, or something else?

Simmons explained that "training alone is not very effective for developing AI fluency," though giving workers a clear direction and some initial pointers certainly serves as a helpful baseline. Zendesk, for example, has found the greatest success in giving workers time and space to experiment and become accustomed with AI on their own terms.

This particular company's stance was to pause non-urgent work and organize a dedicated internal hackathon to encourage proactive exploration. The result was a marked increase in employees' practical AI skills and better cross-team collaboration, but halting non-urgent operations altogether isn't a necessity and just reflects one initiative.

It's a similar initiative that's being piloted by Canva, which tells its 5,300+ workers to drop tools for a full week and experiment with AI.

"We give our team the room to step back, get out of business as usual, and try something genuinely new," a spokesperson said.

"Practical AI training should go beyond introductory courses and prompting techniques and create space for employees to actually use the tools in a safe, secure environment," Simmons concluded. Piloting AI tools with synthetic data (and therefore, no harmful consequences) ultimately leads to the highest levels of confidence.

'Employers own the conditions... employees own the curiosity'

Another key area where studies and reports have been split is in whose responsibility it is to upskill and re-skill, whether that's through updated policies, passive training or active experimentation.

As a C-suite exec, Simmons believes the organization should bear the brunt of the responsibility by giving workers access to tools and learning opportunities. Clearly, they must think outside the box and offer a much broader array of support: "not just training, but also ideation and experimentation through initiatives like hackathons, sandboxes, and collaboration opportunities."

But beyond that, it's totally on the workers' shoulders to "take those opportunities and run with them." After all, it's not just for the benefit of their organization, but it's also to ensure they stay relevant as work evolves in an AI-first era.

A secondary opinion by Canva also backs this up: "Employers own the conditions: the time, budget, permission to experiment... Employees own the curiosity."