A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Saturday's puzzle instead then click here: Quordle hints and answers for Saturday, July 18 (game #1636).
Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.
Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.
SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.
Quordle today (game #1637) — hint #1 — Vowels
How many different vowels are in Quordle today?
• The number of different vowels in Quordle today is 4*.
* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).
Quordle today (game #1637) — hint #2 — repeated letters
Do any of today's Quordle answers contain repeated letters?
• The number of Quordle answers containing a repeated letter today is 1.
Quordle today (game #1637) — hint #3 — uncommon letters
Do the letters Q, Z, X or J appear in Quordle today?
• No. None of Q, Z, X or J appear among today's Quordle answers.
A new NYT Strands puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Saturday's puzzle instead then click here: NYT Strands hints and answers for Saturday, July 18 (game #867).
Strands is the NYT's latest word game after the likes of Wordle, Spelling Bee and Connections – and it's great fun. It can be difficult, though, so read on for my Strands hints.
Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc's Wordle today page for the original viral word game.
SPOILER WARNING: Information about NYT Strands today is below, so don't read on if you don't want to know the answers.
NYT Strands today (game #868) - hint #1 - today's theme
What is the theme of today's NYT Strands?
• Today's NYT Strands theme is… Big talk
NYT Strands today (game #868) - hint #2 - clue words
Play any of these words to unlock the in-game hints system.
SAIL
NAIL
HAIL
THEME
MOVE
COST
NYT Strands today (game #868) - hint #3 - spangram letters
How many letters are in today's spangram?
• Spangram has 11 letters
NYT Strands today (game #868) - hint #4 - spangram position
What are two sides of the board that today's spangram touches?
• First side: left, 8th row
• Last side: right, 8th row
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.
NYT Strands today (game #868) - the answers
(Image credit: New York Times)
The answers to today's Strands, game #868, are…
GIANT
MAMMOTH
COLOSSUS
BEHEMOTH
LEVIATHAN
SPANGRAM: SUPERSIZEIT
My rating: Easy
My score: Perfect
The letter Z helped me get the spangram — eventually, as I got “size” first, then “supersize” before finally SUPERSIZEIT.
Even though I understood the theme immediately and had an idea of the words we were searching for, I really struggled to put them together .
Although rarely used, LEVIATHAN is an incredible word (it features in the Bible, describing a fire-breathing sea serpent).
I was thinking that it really should be the name of a horror or at least a wrestler when a Google search told me that both already exist; the 1989 horror flick made less cultural impact than Dave Batista, who used the nickname Leviathan before becoming The Animal. Every day’s a school day with Strands.
Yesterday's NYT Strands answers (Saturday, July 18, game #867)
SWEET
OATY
NUTTY
CRUNCHY
SALTY
WHOLESOME
SPANGRAM: BOWLOFGRANOLA
What is NYT Strands?
Strands is the NYT's not-so-new-any-more word game, following Wordle and Connections. It's now a fully fledged member of the NYT's games stable that has been running for a year and which can be played on the NYT Games site on desktop or mobile.
I've got a full guide to how to play NYT Strands, complete with tips for solving it, so check that out if you're struggling to beat it each day.
An academic at the Australian National University has accused it of a “hysterical” response to students using AI to cheat, as tertiary institutions rush to shore up the credibility of assessments.
But a colleague warned Australia is in danger of “shipping our national intellectual capability” to companies in California and China if educational rigour is not restored.
Simply put, polling rate is the frequency at which data is sent between your mouse and keyboard to your computer. For many years, the standard polling rate of most gaming mice and keyboards was 1,000Hz, and that's still a setting that's available by default.
However, with the rise of competitive shooters, particularly tactical ones such as CS2, VALORANT, and Rainbow Six Siege, the pro scene needed a shift, and we saw the advancement towards 4,000Hz and 8,000Hz for "esports tech".
But does anyone really need an 8,000Hz polling rate when gaming? Do the best FPS players in the world even take advantage of it? I'm going through everything you need to know about how 8K works, the latency you can expect, impact on battery life, and all other things you need to consider when properly getting your setup ready for success.
How does polling rate affect latency?
The Razer Viper V4 Pro is among the best gaming mice, and it supports 8K polling (Image credit: Future)
True to its namesake, an 8K polling rate will report its position to your PC a total of 8,000 times, instead of 4,000 times or a thousand times as with 4,000Hz and 1,000Hz, respectively. An 8,000Hz polling rate responds with far lower latency than either of these two previous standards; specifically, updates are sent between a mouse or a keyboard to your computer's processor every 0.125ms, eliminating a theoretical maximum of 0.875 ms input delay, meaning a smoother cursor or movement.
The difference in response time of 0.125ms (8K), 0.25ms (4K), and 1ms (1K) can be huge when it comes to polling rate, provided you have the right hardware to properly take advantage of this. That's because an 8,000Hz polling rate can be incredibly demanding on the battery life of wireless mice and keyboards, and it's one of the main considerations as to why you may want to drop the response time down in your settings.
The drawbacks of an 8,000Hz polling rate
The Corsair Sabre V2 MG, like many top-end mice, have the option of 8K polling wirelessly, even if it comes at 1K out of the box (Image credit: Future)
In brief, an 8,000Hz polling rate for your mouse and keyboard can be incredibly demanding on your computer. That includes heavy strain on even the best processors, the aforementioned heavily reduced battery life on a mouse and keyboard, and even the potential risk of micro-stuttering. As a lifelong PC gamer and a competitive shooter enthusiast who is hopelessly obsessed with VALORANT, I've seen the battery life debate happen in real-time.
I've tested and reviewed hundreds of mice and keyboards throughout my career, and one of the biggest points of contention when it comes to enabling the polling rate comes down to battery life. As an example, my favorite gaming mouse of all time is the Corsair Sabre V2 MG, an 8K-enabled gaming mouse that's built with a Magnesium Alloy frame. It weighs just 56g and features excellent skates, making it ideal for a precision heavy shooter.
As standard, via the Corsair Web Hub, this gaming mouse will run out of the box at 1,000Hz (1K), with battery life prioritized. With the pointer used as the stock, you can expect around 100 to 120 hours of play between charges. In my experience, I am charging this mouse every two weeks or so, depending on how much I'm working day-to-day, etc. However, in my many months with this mouse, I've noticed that things fall off incredibly hard with 8K enabled; my battery life tends to last only around two to three days (or one or two nights of VALORANT Competitive) before I'm reaching for the USB-C lead.
If we take 18 hours as the average battery life from the Corsair Sabre V2 MG when used in 8K mode, that's a reduction of around 85%, which is staggering. We can see a similar drop off from the most well-used new esports mice currently on the market, with the Razer Viper V4 Pro (an excellent mouse), touting up to 180 hours of playback in 1K but only 45 hours when maxed out at 8K. Similarly, while the Logitech G Pro X2 Superstrike has around a 90-hour playback at 1K, that drops down dramatically to between 20 and 35 hours at 8K. With Logitech's pointer, though, you have to factor in the strength of the haptics here, hence the rough range rather than a more exact figure.
What about the human limitations of an 8,000Hz polling rate?
(Image credit: Future)
Now that we've covered what an 8,000Hz polling rate is, how it works, how it compares against 1K and 4K, and the technical drawbacks, we now need to go over the human limitations with 8K. As previously mentioned, 1K means a 1ms response time, and that gets cut down to just 0.125ms with 8K, but what does that mean for a human being, and not a machine?
The average human response time is around 250-280ms; that's the time it takes to see something and physically react. I am not a pro player, but I do play video games at a high level, with a response time average of around 175ms, though I've been faster. The best esports players can eclipse this effortlessly: you only need to look as far as VALORANT and CS2 players, such as TenZ (widely considered to be among the best VALORANT players), with an average reaction time of 135ms. 100ms is even possible, even though the average tends to hover around the 150ms mark.
If we take 1K as a baseline for polling rate, that means there's an average of 150 reports between the mouse/keyboard and the processor, as you tap a key or click what you see. Taking 8K, then, sees 1,200 instances. With this information in mind, is there really much of an advantage here, even at this granular level, with the best players in the world and, conversely, the only people who would even notice a proper difference between 1K, 4K, and 8K?
Do pro players use 8,000Hz polling rates?
The best gamers in the world aren't interested in using an 8,000Hz polling rate for their keyboards or mice. (Image credit: Secretlab)
Considering the far lower latency, you would expect that professional esports gamers would all dial their settings to an 8,000Hz polling rate by default when playing demanding games, but that's simply not the case. If we look back at TenZ's mouse and keyboard settings when playing VALORANT, he last used the Pulsar TenZ Signature Edition mouse at 1,600 DPI, 0.24 sensitivity, and (most crucially) a 1,000Hz polling rate as standard; no 4K, and certainly no 8K enforced here.
What about a VALORANT player from the most aggressive team in the game? Then we must look at f0rsakeN from the legendary Paper Rex. He was last using the Pulsar Sustanto-X with an 800 DPI, a 0.71 sensitivity, and a 4,000Hz polling rate. Let's move away from Riot Games' shooter and instead turn our attention to Valve, then. As far as CS2 players go, few are as well-regarded in 2026 as Team Spirit's donk.
The young Russian is widely regarded as the best player in the world. Is he using an 8K polling rate for his mouse and keyboard to do this? No, while he previously used the industry standard Logitech G PRO X Superlight 2, he's now switched to a bespoke, and as of yet unreleased/prototype Zowie X Donk mouse at 800 DPI, 1.25 sensitivity, and a 1,000Hz polling rate. What about Niko? He recently had his own Razer collaboration drop, and currently uses a Razer Deathadder V4 Pro Niko Edition with an 800 DPI, 0.9 sensitivity, and a 2,000Hz polling rate.
Let's briefly look at two games for the sake of argument, and two of their respective top players. Peterbot is widely considered the best Fortnite player in the world, and he uses a Logitech G Pro X Superlight 2 at 800 DPI with a 1,000Hz polling rate. ImperialHal is among the best Apex Legends players, and the CEO (as he's known) currently uses a Finalmouse Starlight Pro with an 800 DPI, a 1.1 sensitivity, and a 1,000Hz polling rate. So, pro players simply are not using an 8,000Hz polling rate when playing, and they're simply better than the other 99.9% of players who play these games.
Is using an 8,000Hz polling rate worth it?
Based on everything I've gone through here, it becomes challenging to wholesale recommend (or actively condemn) 8K. An 8,000Hz polling rate is mechanically and technically better for how it responds between the peripheral and the machine, even if the human who presses that key or clicks the mouse ends up being the limiting factor, or bottleneck. It's the simple, undeniable truth that the saved milliseconds are (really) indistinguishable to basically everyone on the planet.
Then we get into the debate of battery life. Do you want your gaming mouse to last over 100 hours on a single charge, or just 10? That's a tough sell, considering it undercuts the primary reason to cut the cord in the first place. Then there's the massive CPU overload, game engine incompatibility (and developers and publishers are not yet gearing their software for the tech), and the variables of the hardware itself. Hardware makes the difference: TMR and Hall Effect switches and sensors will have far more of an impact than the simple reporting time between the processor and peripheral.
So why am I not fully discouraging you from using an 8,000Hz polling rate in its entirety? Because it ultimately comes down to personal preference, and to feel. If you're running one of the best gaming PCs or best gaming laptops with a high-end processor, a 4K graphics card, and a high refresh rate monitor, then, battery life aside, it doesn't really hurt to have everything cranked up and forgotten about. Are you likely to really notice it? No, but fractional optimization is still (ever so slightly) better than none in the first place.
I'm not quite cynical enough to claim that an 8,000Hz polling rate is just a marketing gimmick; however, I will say that hardware companies have gleefully exploited the ignorance of the end consumer, the simple reductionist view that a bigger number automatically means better, without looking into the nuances of the conversation. Now that's something you will be fully clued up on when considering your next keyboard or mouse.
A new NYT Strands puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Friday's puzzle instead then click here: NYT Strands hints and answers for Friday, July 17 (game #866).
Strands is the NYT's latest word game after the likes of Wordle, Spelling Bee and Connections – and it's great fun. It can be difficult, though, so read on for my Strands hints.
Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc's Wordle today page for the original viral word game.
SPOILER WARNING: Information about NYT Strands today is below, so don't read on if you don't want to know the answers.
NYT Strands today (game #867) - hint #1 - today's theme
What is the theme of today's NYT Strands?
• Today's NYT Strands theme is… A healthy breakfast
NYT Strands today (game #867) - hint #2 - clue words
Play any of these words to unlock the in-game hints system.
COUNT
STAY
WOOLY
LUNCH
LONG
LARGE
NYT Strands today (game #867) - hint #3 - spangram letters
How many letters are in today's spangram?
• Spangram has 13 letters
NYT Strands today (game #867) - hint #4 - spangram position
What are two sides of the board that today's spangram touches?
• First side: left, 4th row
• Last side: right, 4th row
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.
NYT Strands today (game #867) - the answers
(Image credit: New York Times)
The answers to today's Strands, game #867, are…
SWEET
OATY
NUTTY
CRUNCHY
SALTY
WHOLESOME
SPANGRAM: BOWLOFGRANOLA
My rating: Easy
My score: Perfect
I go through phases with breakfasts of having the same thing for ages before getting bored and switching to something new.
In recent months I’ve been through porridge, overnight oats, fruit and yoghurt, and recently a BOWLOFGRANOLA all in pursuit of “a healthy breakfast" when in truth I would love to eat bacon and eggs every day.
I digress. After finding SWEET and OATY the search became a hunt for the obvious — although I was unaware granola could also be SALTY.
Yesterday's NYT Strands answers (Friday, July 17, game #866)
ENTERTAINMENT
NATURE
SPORTS
GEOGRAPHY
SPANGRAM: TRIVIALPURSUIT
What is NYT Strands?
Strands is the NYT's not-so-new-any-more word game, following Wordle and Connections. It's now a fully fledged member of the NYT's games stable that has been running for a year and which can be played on the NYT Games site on desktop or mobile.
I've got a full guide to how to play NYT Strands, complete with tips for solving it, so check that out if you're struggling to beat it each day.
A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Friday's puzzle instead then click here: Quordle hints and answers for Friday, July 17 (game #1635).
Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.
Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.
SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.
Quordle today (game #1636) — hint #1 — Vowels
How many different vowels are in Quordle today?
• The number of different vowels in Quordle today is 4*.
* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).
Quordle today (game #1636) — hint #2 — repeated letters
Do any of today's Quordle answers contain repeated letters?
• The number of Quordle answers containing a repeated letter today is 1.
Quordle today (game #1636) — hint #3 — uncommon letters
Do the letters Q, Z, X or J appear in Quordle today?
• No. None of Q, Z, X or J appear among today's Quordle answers.
What letters do today's Quordle answers start with?
• K
• T
• B
• C
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.
Quordle today (game #1636) — the answers
(Image credit: Merriam-Webster)
The answers to today's Quordle, game #1636, are…
KNAVE
TIGHT
BLEAT
CHAOS
This was already a tricky game, with two words featuring silent letters. I navigated those okay, then crashed out — ending my win streak on the final word.
I had four possibilities for words ending I-G-H-T and I picked three wrong ones before I ran out of guesses.
Why I went with the obscure “wight” rather than TIGHT I can’t explain.
Daily Sequence today (game #1636) — the answers
(Image credit: Merriam-Webster)
The answers to today's Quordle Daily Sequence, game #1636, are…
Amazon's AWS cloud services billing system went wild on Friday
Customers saw bills in the trillions
AWS apologized and is working on a fix
AWS, the cloud platform that serves millions who use Amazon's servers to run websites and businesses you use every day, just freaked out many of its customers with usage bills that literally went into the trillions.
"I just saw $1.5 trillion on my AWS bill and my soul left my body," wrote Bharath_uwu on X (formerly Twitter) on Friday, and he wasn't alone.
The Guardian originally spotted the Tweet and multiple social posts relating to the seemingly widespread glitch.
AWS doesn't mail out invoices. Instead, customers logging into the Billing and Cost Management console spotted the billion- and trillion-dollar bills and expressed emotions ranging from concern to panic.
I just saw $1.5 trillion on my AWS bill and my soul left my body https://t.co/EgfQKJTHVl pic.twitter.com/L0gXYbDio7July 17, 2026
As Dan Harvey, who heads marketing at Learning Through Landscapes, told the Guardian, "I almost had a heart attack when I received an email alert from Amazon Web Services with the billing for our charity’s school grounds audit app."
Some tried to take the billing error in stride. Chinmay on X encouraged others to post their highest AWS bills. As of 10 hours ago, he was at $333B.
Others called out AWS for the error with something less than good humor. "I bet someone must have had a real heart attack. This shouldn't be legal...," wrote Mr Doob on X.
Amazon, to its credit, has been delivering multiple updates starting at around 1:30AM PDT on Friday, July 17, when they first spotted the error: "We are investigating issues with Cost Explorer reflecting inaccurate estimated billing data."
A few hours later, AWS reported it had found the culprit: "We have identified the root cause as an issue with unit pricing within the estimated billing computation subsystem, and we are working on a mitigation," they wrote on the AWS Status page.
However, hours later, AWS admitted, "Our efforts to backfill corrected estimated cost and usage data are still underway. We are progressing slower than anticipated."
The good news is that AWS is not asking anyone to pay these incorrect fees and that, while this has been a big frustration for services that run their websites and businesses on AWS, the cloud system has been working flawlessly, and there have been no reports of outages or performance issues on Friday (if you don't count the years shaved off freaked-out customers' lives).
"Typo alert: Some customers saw quadrillion-dollar AWS billing estimates today. Slight miscalculation on our end (very slight ). We're fixing it now. No action needed on your end. Sorry for the confusion. Real question: what will you do with those trillions instead?"
Yes, everyone can all laugh now, though I'm not certain customers like Harvey, Bharath, and others are chuckling along with them.
Was your business affected? How big was your AWS Billing mistake? Let us know the "damage" in the comments below.
Nvidia's RTX 5000 Super models are being readied, and board makers have the chips, a new rumor reckons
However, these graphics cards are 'on hold' due to the high price of VRAM
The Super refreshes pack a lot of VRAM, and so that cost could have an unfortunate impact on MSRPs – and none of this sounds particularly comforting for would-be buyers
Nvidia's RTX 5000 Super refreshes are apparently being readied, but they're on hold for now – and the reason why makes me even more nervous about the pricing of these GPUs when (well, if) they finally emerge.
Of course, these RTX Super models are still rumors – albeit so persistent at this point, there must be at least some validity to them – and the latest assertion comes from VideoCardz.
The tech site has apparently spoken to a source at one of Nvidia's board-making partners who claims that their company has already received 'graphics cards', which presumably refers to GPU chips (which is what Nvidia supplies, along with VRAM, to partners that make the boards).
At any rate, while this board maker supposedly has at least some of the necessary chips, it has been told by Nvidia that the RTX 5000 Super refreshes are now 'on hold'.
Why? Apparently, the delay is down to the price of the video RAM, specifically the new 3GB GDDR7 modules that are required (as per previous rumors) for the refreshed graphics cards.
Compared to the 2GB GDDR7 modules currently in use with Nvidia RTX 5000 models, the 3GB memory chips are around three times more expensive right now. Ouch.
Analysis: Super stumble?
(Image credit: Future)
This presents an obvious stumbling block because, due to these Super variants being packed with VRAM, the total cost of materials gets expensive quite quickly.
Remember that the grapevine believes that the RTX 5080 Super and RTX 5070 Ti Super will bristle with 24GB of VRAM, and even the RTX 5070 Super will have a weighty 18GB on board. Gamers are going to love the latter as an upgrade to the lean 12GB of the vanilla version – but not if it pushes the price of the GPU through the roof.
Going by what VideoCardz tells us about pricing in US dollars as the market currently stands, in the case of the RTX 5070 Super, the greater quantity of RAM, and more expensive modules, would mean 18GB costs in the order of $360 (maybe a bit more), versus $120 for the current RTX 5070's 12GB loadout, whacking an extra $240 on the price (or the regional equivalent elsewhere).
These graphics cards were already expected to be pricey, but that kind of additional premium lumped on top, this could make them real wallet-worriers.
The good news for those hoping for new GPUs from Nvidia is that at least these RTX 5000 Super models do seem to be happening. It's just that we may have to wait some time for them yet – although another recent rumor did suggest they were due this year. The latter speculation also theorized that an RTX 5060 Super 12GB model was incoming, but that didn't make a lot of sense, and I note that it isn't mentioned in this latest rumor. As ever, don't go lightly with the seasoning on all this.
Users says they've observed LG gaming monitors automatically installing unwanted software on PCs
LG's terms of service warn that conversations may be "captured and processed" on TV with the latest version of webOS
The terms say you must now warn guests they may be recorded "in compliance with applicable wiretapping… laws"
How smart should a smart TV be? According to LG's latest TV terms and conditions, the answer is "not quite smart enough to comply with wiretapping laws", because that's now your responsibility if LG captures the voice of a guest in your house through its AI voice services. Though the situation with LG monitors appears to be even more dramatic.
As Gamers Nexus reports, some LG monitors appear to be installing adware on Windows PCs without asking for permission: in addition to the LG Monitor App Installer, they also install McAfee Scam Detector.
LG's own app requires full access to all system resources, which potentially includes all your online activity, logins, hardware, location and more — while McAfee has a long history of being installed on devices as 'bloatware', and people are not reacting positively to suddenly finding it on their PC.
There may be a perfectly innocent explanation for all of this, but when big tech firms keep getting caught doing bad things because they thought they could get away with it, it's no wonder people are assuming the worst.
The bit that's causing consternation regarding smart TVs is part 6(d) of the new LG Electronics terms of service, headed Voice Recognition and Privacy Compliance.
As Notebookcheck's Hannes Brecher notes, the section states that it's your responsibility "to obtain all necessary consents from any third parties whose voices may be captured by the Product and to notify household members and guests that their voices may be captured and processed, in compliance with applicable wiretapping, eavesdropping, and privacy laws."
There are three ways around that. One, you can turn off all microphone-based features. Some people won't mind that, but they can be useful — especially asking it for settings you don't know how to find.
Two, you can avoid installing the latest software — but that means you won't get any security updates, which are important (to protect your privacy, ironically, among other things).
Or you can disable your TV's connection to the internet so it can't send information back, but that obviously makes it less useful, and will also disable the voice controls anyway.
I think the terms and conditions are an attempt at corporate ass-covering rather than something sinister: the preceding paragraph talks specifically about when "a product with voice recognition functionality is used" and it's possible that "family members, guests, children, and bystanders" might be overheard; if you're choosing to activate AI-based voice features then of course voices are going to be captured and processed in order for those features to work.
At the same time, it does seem broad enough to let LG use people's voices as AI inputs, rather than just accidental capture. And in the context of what LG's doing with PC monitors (not to mention AI training's very, ahem, casual relationship with consent and copyright), that's a worry — and people online increasingly have no tolerance for anything like this, as the responses on Reddit and Gamers Nexus' original YouTube video have shown.
We've approached LG for comment on these claims, and will update the story when we hear back.
Designers say that as well as offering a degree of protection from surveillance, their clothes make a powerful fashion statement about the importance of privacy
As facial recognition technology is rolled out across Britain’s public spaces, a new generation of designers say privacy could be the next big fashion trend.
Companies have started incorporating “adversarial patterns” in their garments – carefully designed arrangements of shapes, colours and repeated motifs said to exploit weaknesses in some computer vision systems.
Meta is reportedly planning a cloud business that would sell access to AI computing power and models, extending its internal infrastructure into a commercial service for outside developers and enterprises. Reuters, citing Bloomberg’s reporting, noted that the planned offering would allow customers to access AI models hosted on Meta’s infrastructure and pay based on usage, effectively positioning the company in the infrastructure-as-a-service and AI platform markets. On the surface, this seems like a logical next step. If you are already spending enormous amounts of money to build AI infrastructure, there is a natural temptation to ask whether some of that investment can be monetized beyond your own internal use.
I have seen this pattern before. A company builds sophisticated internal systems, recognizes their value, and then begins to imagine that becoming a cloud provider is simply a matter of exposing those capabilities to external customers. It sounds straightforward, especially given the excitement around AI and the demand for high-performance infrastructure. But cloud computing is not just another distribution model. It is not simply a matter of offering on-demand multitenant services and charging a fee. It is a deeply operational, trust-based business in a market that punishes companies that do not fully understand what enterprise customers require.
A crowded neocloud market
The first problem Meta faces is that this is not an open opportunity. The neocloud space, meaning purpose-built AI infrastructure delivered as a service, is already crowded and increasingly difficult to enter. Amazon, Microsoft, and Google dominate the conversation for obvious reasons. They have years of cloud operating experience, broad service portfolios, global reach, mature ecosystems, and deeply established enterprise relationships. Oracle remains a serious player as well, especially in enterprise applications, data platforms, and performance-sensitive workloads. IBM still matters in hybrid cloud, operations, and industries where governance and regulatory rigor remain central.
That list alone should give Meta pause. These companies are not just infrastructure vendors. They are experienced cloud operators. They have spent years building not only the underlying platforms, but also the native capabilities enterprises now expect by default. Those capabilities include security, governance, identity management, observability, support, compliance, billing controls, resilience planning, and integration with the broader enterprise technology estate. These are not secondary features. They are part of the core value proposition.
This is why late entry into the cloud market is so hard. A new provider is not just competing on price or capacity. It is competing against accumulated trust. Enterprises are not casual buyers. They are selecting long-term operating environments for applications, data, AI models, and business-critical processes. They want confidence that the provider understands how these services will be consumed, governed, and supported over time. Meta is entering a market where the incumbents already have a major head start on all of those fronts.
Harder than it looks
Over the years, I have had many technology companies come to me and say they wanted to reposition their technology in the cloud space, either as software as a service or infrastructure as a service. In the beginning, enthusiasm is always high. The technology is impressive. The market size looks attractive. The revenue models appear compelling. Investors love the story. Then we begin to walk through what it really means to operate as a cloud provider, and the optimism usually fades fast.
The questions become very practical and very uncomfortable. How will tenants be isolated? How will identity and access controls work across different kinds of customers? What governance models will be built in natively? How will workloads be monitored, optimized, and secured? What does support look like 24 hours a day, across regions, across industries, across compliance boundaries? How will outages be handled, communicated, and remediated? How will the platform integrate with existing customer tools for operations, policy management, and security response? How much investment will it take just to become credible before you even begin to differentiate?
Once companies fully understand the complexities, market dynamics, and the capital and execution required to compete even with secondary players, many of them back off. They realize that cloud technology is not a packaging exercise. It is a transformation in how a company designs, operates, supports, sells, and evolves technology. That is why I remain skeptical when any company assumes it can translate internal infrastructure excellence into external cloud success without a very long, disciplined commitment.
Meta’s market readiness
Of course, Meta is not lacking in financial resources. If any company can afford to spend aggressively in this space, it is Meta. The company has the capital to build infrastructure, absorb losses, hire experienced talent, and stay in the market long enough to make a serious attempt. I would never argue that Meta is too small or too poor to try. Quite the opposite. If there is any non-traditional entrant with the financial scale to force itself into the conversation, Meta would be high on the list.
But money does not erase complexity. It only gives you the chance to confront it. The real question is not whether Meta can afford to become a cloud provider. The question is whether Meta has what it takes to become an excellent cloud provider. Those are two very different things. Enterprises are not going to move meaningful workloads to a new platform simply because the company behind it is wealthy or technically famous. They are going to ask whether the provider understands enterprise consumption patterns, enterprise risk, enterprise governance, and enterprise operations.
That is where the challenge becomes much more serious. Meta has extensive experience running infrastructure for itself. That is valuable, but internal operating excellence is not the same thing as external service maturity. Running systems for your own workloads allows a high degree of control over architecture, standards, priorities, and operating assumptions. Running systems for paying customers requires flexibility, consistency, transparency, and support across a wide range of use cases that you do not control. Those are very different disciplines, and companies often underestimate the gap between them.
What exactly is Meta?
Another concern here is strategic clarity. Meta already has a complicated market identity. It is a social media company, an advertising platform company, a hardware company, an AI company, and still, in the minds of many, the company that spent billions pursuing the metaverse. If it now wants to be viewed as a serious cloud infrastructure provider, it will need to explain not only what it is offering, but why customers should believe this is a durable long-term commitment and not just another adjacent experiment.
That uncertainty can be damaging. Customers want stable providers with clear strategic intent. They do not want to architect important systems around a platform if they suspect the provider may lose interest, shift direction, or reframe the business after a few years of uneven results. Cloud computing requires patience, consistency, and deep customer orientation. It is not a market where strategic ambiguity helps.
This could become confusing for Meta internally as well. Building a true cloud business demands focus. It demands years of investment in areas that may not be glamorous but are absolutely necessary, such as governance, operations, controls, support frameworks, partner programs, and enterprise sales alignment. If the company is not willing to make those sacrifices fully and for the long term, the initiative will struggle.
Since 2017, Iason Gabriel has worked at the tech giant, trying to anticipate – and think through – the impact of AI. But as commercial and geopolitical pressures escalate, can ethicists make any difference?
ARK augmented reality describes a form of spatial interaction in which artificial intelligence does more than place digital objects in the physical world. It uses knowledge from foundation models to understand unfamiliar scenes, generate useful information and adapt its response to the person, task and environment.
The concept comes from a 2023 research paper titled “ArK: Augmented Reality with Knowledge Interactive Emergent Ability.” Its authors proposed a system that transfers knowledge from large foundation models into new physical and virtual environments. Instead of training an AI agent separately for every room, object or activity, ArK uses existing knowledge memory to interpret what it sees and generate appropriate 2D or 3D content.
Traditional augmented reality places information over reality. ARK augmented reality tries to understand reality before deciding what should appear.
That distinction creates a compelling direction for Apple. Vision Pro already blends digital content with physical space. Apple Intelligence adds language, image understanding and personal context. Visual Intelligence lets users look at objects, text and scenes before asking for information. Siri AI adds a conversational layer. ARKit, RealityKit and visionOS provide the spatial foundation.
Apple now has many of the components required to build a knowledge layer over the physical world.
ARK Augmented Reality Goes Beyond Digital Overlays
Most familiar augmented reality experiences are relatively fixed. A furniture app places a digital sofa in a room. A navigation app displays arrows over a street. A museum app identifies an artwork. A repair app shows where a component belongs.
These experiences can be useful, but they usually depend on predefined content and rules. The system recognizes a known object or location, then presents prepared information.
ARK augmented reality proposes something more adaptive. A knowledge-enabled system could enter an unfamiliar environment, interpret what is present and generate assistance based on a broader understanding of objects, relationships and possible actions.
A user looking at a complicated electrical panel would not only see labels. The system could identify the panel type, explain the purpose of each section, compare the arrangement with standard configurations and warn the user not to touch dangerous components.
A student examining a machine could ask how it works and receive a spatial explanation attached to its real parts. A surgeon could review anatomical information positioned around a medical model. A technician could receive repair guidance even when the exact equipment had not been included in a previous training dataset.
The display becomes less like a floating screen and more like an informed interpretation of the environment.
Image Credit: Apple Inc.
Why Apple Is Positioned to Build It
Apple’s advantage would not come from inventing every part of ARK augmented reality. It would come from combining hardware, software, artificial intelligence and personal context into one controlled ecosystem.
Vision Pro already understands rooms, surfaces, hands, eyes and objects well enough to place digital content convincingly in physical space. Its cameras and sensors continuously map the environment while visionOS manages windows, volumes and immersive experiences.
Apple Intelligence provides another layer. Foundation models can interpret language, summarize information, generate content and connect requests with app actions. Visual Intelligence can identify text, images and real-world objects. Siri AI can act as the conversational interface.
The Apple ecosystem also provides personal context that a generic AR platform may not have. Calendar knows the user’s schedule. Maps understands location. Photos contains visual history. Notes and Mail hold project information. Health contains personal trends. Home understands rooms, accessories and automations.
An Apple-built knowledge layer could combine the environment in front of the user with the information already available across trusted devices and apps.
A person preparing for a meeting could look at a conference room and see the agenda, participant names and relevant documents arranged around the table. A traveler could look across a train station and receive directions connected to the correct platform, departure time and digital ticket. A homeowner could examine a room and see live information from temperature, air-quality and energy sensors.
The physical environment becomes the interface.
Vision Pro Is the Laboratory
Vision Pro is the obvious development platform for this idea because it already offers the cameras, displays, sensors and processing required for advanced spatial understanding. Its size and price limit broad adoption, but those same characteristics allow Apple and developers to test experiences that would be harder to build first on lightweight glasses.
ARK augmented reality on Vision Pro could become especially useful in enterprise environments. Apple already promotes the headset for employee training, product design, field service, collaboration and specialized visualization.
A worker could enter an unfamiliar facility and receive contextual instructions attached to real equipment. An engineer could inspect a physical prototype while design changes, performance data and maintenance history appear around it. A new employee could learn a workflow through spatial guidance generated from company manuals and operational knowledge.
The system would not need to display every available fact. It would need to understand which information fits the current task.
That is where AI knowledge becomes more valuable than a traditional AR overlay. Too much information can be as disruptive as too little. A strong system should filter, prioritize and explain rather than cover the room with floating labels.
The iPhone Could Bring ARK to More People
Vision Pro may be the development environment, but iPhone could provide the scale. The iPhone already combines cameras, LiDAR on Pro models, location, motion sensing, Visual Intelligence and Apple Intelligence.
An iPhone-based ARK experience would be less continuous because users would need to hold up the device. It could still deliver useful knowledge interaction without requiring a headset.
A user could point the camera at a plant and receive care guidance based on its species, visible condition, local weather and placement in the room. Someone assembling furniture could receive generated guidance positioned over the actual parts. A shopper could compare products by looking at them rather than manually searching for each model.
The difference from ordinary visual search would be interaction. The user could ask follow-up questions, change the goal and let the system update the visual guidance.
Future glasses could eventually make the experience persistent. The iPhone would help Apple develop the intelligence, privacy controls and app ecosystem before lightweight eyewear becomes ready.
A World That Explains Itself
The most ambitious version of ARK augmented reality is a world in which information is attached to objects, places and situations only when requested or needed.
A city could become understandable without searching through maps, review sites and separate transit apps. Buildings could reveal history. Streets could explain temporary closures. Restaurants could surface reservations, dietary options and accessibility details. Public transport could show the correct entrance and platform.
Inside a home, Apple’s knowledge layer could explain energy use, identify maintenance problems and guide repairs. In a classroom, physical models could become interactive lessons. In healthcare, patients could receive visual instructions for approved home procedures or medication devices.
The experience would depend on context rather than menus. The user looks, asks and receives information positioned where it belongs.
This is the transition from augmented reality as decoration to augmented reality as understanding.
Privacy Will Decide Whether It Works
A knowledge layer capable of understanding the physical world also creates serious privacy questions. The system may see homes, workplaces, documents, faces, screens and private conversations. The more context it receives, the more useful it becomes. The same access can also make it invasive.
Apple’s privacy architecture could become one of its strongest advantages. On-device processing would allow some scene understanding without sending continuous camera feeds to external servers. Private Cloud Compute could handle more complex requests while preserving stricter controls than an ordinary cloud AI service.
Apple would also need visible controls over what is analyzed, remembered and shared. A user should know when the system is interpreting the environment and whether information is temporary or retained. Workplaces and public spaces would need rules for recording, identification and sensitive data.
ARK augmented reality cannot become socially acceptable if people feel that nearby devices are constantly cataloging them.
The successful version would behave less like surveillance and more like intentional assistance. The user asks for understanding. The system processes only what is needed. The knowledge disappears when the task ends unless the user chooses to save it.
Image Credit: Apple Inc.
Developers Could Build Knowledge Worlds
Apple would not need to create every knowledge experience itself. visionOS, ARKit, RealityKit and App Intents could let developers connect specialized knowledge with physical environments.
A medical company could build guidance for approved equipment. A manufacturer could create spatial maintenance systems. A university could attach interactive explanations to laboratories and museums. A retailer could create product comparisons based on inventory and customer preferences.
Apple could provide the shared intelligence layer: scene understanding, object recognition, language interaction, privacy permissions and spatial presentation. Developers would provide the trusted domain knowledge.
This division would be essential. General AI can explain many things, but professional environments require verified information. A generated answer about a historical building has different consequences from an answer about industrial machinery or medical treatment.
Apple’s platform could distinguish between general knowledge and certified domain sources, allowing users to see where information comes from before acting.
From Spatial Computing to Contextual Computing
Apple describes Vision Pro as a spatial computer because digital content can occupy physical space. ARK augmented reality suggests the next progression: contextual computing.
A spatial computer knows where digital content is placed. A contextual computer understands why the content is useful in that location, what the user is trying to accomplish and which knowledge should appear.
That would connect several Apple strategies that currently look separate. Vision Pro supplies spatial awareness. Apple Intelligence supplies reasoning and generation. Siri AI supplies conversation. Visual Intelligence supplies real-world understanding. The ecosystem supplies personal context.
Together, they could create a system that does not wait for users to open the correct app and search through the correct menu. The environment itself becomes the starting point.
Apple’s opportunity is to make that interaction restrained rather than overwhelming. The company does not need to cover reality with digital noise. It can use design, privacy and contextual intelligence to make information appear only when it improves the moment.
ARK augmented reality offers a useful description of that future. Reality remains visible. Knowledge becomes interactive. The computer begins to understand the world in front of the user instead of asking the user to translate everything into a search box.
If you're starting to plan for the upcoming school year and are finding your current laptop a little lacking, the good news is retailers have already launched some solid deals across a vast range of devices from the likes of Dell, Apple, Lenovo, HP, and more. So, I've taken a look and picked out what I think are eight of the best laptop deals if you're shopping for a new device for the year ahead.
I've taken a variety of budgets and needs into consideration: whether you want a cheap Chromebook for basic schoolwork and light use, or a powerful machine for more demanding needs.
One of the standouts has to be this Dell XPS 13 that drops to $599.99 with a student discount. That's a fantastic price for a brand-new version of one of our favorite laptops here at TechRadar, which boasts excellent all-around performance and a stylish design.
Apple fans will also want to check out the manufacturer's Education Store, which has just launched its college offer for the year. As well as special pricing on several devices, including the much-loved and impressive value MacBook Neo, you can also get gift cards of up to $150 on select Macs and iPads.
As someone who finds the best laptop deals regularly for TechRadar, you can browse all of my top picks below, and stick with TechRadar over the next few weeks for even more of our back-to-school coverage, featuring top tech recommendations and the best offers on must-have products.
If you want a cheap Windows laptop, then I think this Lenovo IdeaPad Slim 3i is a good value-for-money option for basic needs. I like it as a simple and effective device for school or college, with enough power for browsing, word processing, and other general admin tasks.View Deal
If storage is important to you, then I think this HP OmniBook 3 could be worth the price, as it features a 512GB SSD — something I don't regularly see at this price. Of course, that means some compromises have been made elsewhere, with a slightly weaker Ryzen 3 CPU and only 8GB of RAM. Still, that's perfectly fine for light use. Just be aware that the larger 17-inch form factor makes it a less portable option, too.View Deal
If you want a flexible device that's just as easy to use for work and admin jobs as well as casual web browsing and streaming media, then this Acer Chromebook Plus Spin 14 is an excellent buy. It mainly features entry-to-mid-level components, but there's enough power to keep up with light daily use. Plus, you get the added benefit of a versatile 14-inch touchscreen, effectively giving you a laptop and tablet in one.View Deal
Display: 13 inches Processor: A18 Pro RAM: 8GB Storage: 256GB
The MacBook Neo just got a price bump, but students can get it cheaper again thanks to Apple's education discount. The latest MacBook packs a 13-inch Liquid Retina display, an A18 Pro chip, 8GB of RAM, 256GB of storage, and up to 16 hours of battery life, all for under $600, which is fantastic value. Discounts have been better in the past, but I think these will be rarer to see over the coming months and this price likely won't be beaten over Black Friday.View Deal
This is a solid discount on a modern laptop powered by one of the efficient Snapdragon X CPUs. As well as excellent levels of all-around performance, this device boasts a huge all-day battery life, so you won't need to worry about being attached to a charger at all times. There's even 512GB of SSD storage, which you don't often see in combination with those other strong performance components at this price point. This is an all-around excellent laptop for everyday, college and business use.View Deal
The Dell XPS 13 has regularly been one of our favorite laptops here at TechRadar, and this brand-new version continues that trend. With an Intel Core 5 processor, 8GB of RAM and a 512GB SSD, it has everything you need for solid all-around performance at work or college. Students can also save an extra $100, bringing the price down to a ridiculously low $599.99.View Deal
Here's one of the best deals on a modern, all-purpose, and well-rated laptop in Dell's latest sale. You can now get $550 off this machine, which features a powerful Intel Core Ultra 5 processor, a healthy 16GB of RAM, and a decent 512GB SSD. With all of those components, you'll have no issues with all of your everyday computing and work needs. And at 14 inches, it's an easily portable device, too. Altogether, it's an excellent value all-around option and a laptop we scored 4.5/5 in our review.View Deal
One of Samsung's latest premium laptops just got its biggest discount yet at Best Buy. While similarly powered devices can be picked up for less, you do get a gorgeous 16-inch LED screen with smooth and responsive 120Hz support with this Galaxy Book6. It's powerful elsewhere, too, with an excellent battery life of up to 24 hours, making it a great pick for college, work, and productivity needs.View Deal
Palit has launched a new RTX 3060 model with 12GB of VRAM
This continues a recent theme of old GPUs being resurrected
These GPUs may pack 12GB of video RAM at a relatively affordable price, but they don't make sense compared to current-gen pricing
We might not have seen any new graphics cards so far from Nvidia this year — and no RTX 5000 Super refreshes are in sight (although there are rumors again) — but we're continuing to witness the resurrection of some old GeForce GPUs, with Palit officially revealing a new RTX 3060.
That popular workhorse of a gaming GPU has already made something of a comeback, returning to the shelves of retailers like Newegg as an alternative budget graphics card in a RAM-starved climate where prices of many of these current-gen boards are being hiked up (alongside other PC components, of course).
Although to be fair, even today, the RTX 3060 remains a very popular GPU, ranking second on the Steam hardware survey for June 2026 (behind only the RTX 4060 laptop GPU in a rather bizarre twist I won't get into here).
Palit observes: "Combining a massive 12GB GDDR6 memory capacity with a clean, all-black dual-fan shroud, the Infinity 2 delivers honest, dependable performance without the premium price tag."
Analysis: let's hope for a Palit-able price tag, but so far, RTX 3060 pricing has been out of whack
(Image credit: Palit)
The key selling point for these RTX 3060 models is that they're loaded with 12GB of video RAM (VRAM), as Palit emphasizes with this new Infinity 2 OC model. Where this is side-stepping the RAM crisis (to some extent) is that these past-gen Nvidia cards use GDDR6 VRAM, a different memory supply to tap than the GDDR7 used in RTX 5000 graphics cards (except the RTX 5050, which goes with GDDR6, but that's the outlier here).
This means Nvidia can get its partners to produce graphics cards that have 12GB of GDDR6 VRAM (a loadout a good few gamers put a priority on securing), but sit at the cheaper end of the market. Due to cost issues, these days it's a lot less tenable to sell lower-end GPUs which stack up the GDDR7 VRAM (like the RTX 5060 Ti with 16GB).
It's sad that the RAM crisis has forced us to this point, but that's the reality of the GPU world at the moment. What's also problematic for me here is that the RTX 3060 might have 12GB of VRAM at a relatively affordable price point, but looking at current pricing in the US as an example, it's not meaningfully cheaper than the RTX 5060 with 8GB.
Yes, you're getting a lean RAM configuration with the latter current-gen graphics card, but it's still a much faster GPU — one that benefits from DLSS 4, too — and only just over 5% more expensive (at the time of writing). At this kind of price differential, the RTX 3060 12GB doesn't make a lot of sense.
If these RTX 3060 models were a lot cheaper, then there might be more of an argument. To be fair to Palit, we don't have a price tag on this new offering yet, but pricing needs to be a lot more compelling than what I'm seeing right now (especially in the UK, where the RTX 5060 8GB is actually cheaper than the 3060 at some retailers, although there isn't much stock of the latter).
Nichirei Co confirms cyberattack disrupting refrigerated warehouse and frozen food logistics, impacting customers including KFC Japan
KFC warned of delivery delays leading to possible menu restrictions, shortened hours, or temporary suspension of online orders
No data leaks or attacker claims have surfaced; disruptions suggest ransomware, but both companies expect operations to normalize by week’s end
Nichirei Co, a large Japanese company which produces frozen and processed foods, and operates cold-chain logistics to distribute them across the country, has confirmed it suffered a cyberattack which affected some of its customers, including KFC.
The company confirmed the news in a notice stating, “Today, Nichirei Co experienced a system failure caused by unauthorized access."
Nichirei said that it is currently investigating the incident, but that there is no evidence of personal or customer data leaking outside the company. However, due to the nature of the incident (which hasn’t been explained in detail), multiple operations have been affected, including refrigerated warehouse inbound/outbound operations, and frozen food shipping services.
No claims yet
These disruptions affected, among others, KFC - one of the largest fast food chains in the world. In a separate announcement, KFC said that a system failure at Nichirei Logistics Group, which happened on July 13, affected the company’s logistics and delivery sites. “From Tuesday, July 14, 2026, food delivery to KFC stores is expected to be affected,” it said.
“As a result, each store may suspend operations depending on some products going out of stock, menu restrictions, shortened business hours, or ingredient stock availability. Additionally, online orders from the official app and website have also been temporarily suspended.”
The nature of the attack, or the identity of the attackers, were not disclosed. Since Nichirei had to shut down some of its services, it could mean this was a ransomware attack and that some company data was stolen after all. The Register reports that KFC Japan hasn’t posted information about store closures and continues to promote summer menu items.
At press time, no threat actors claimed responsibility for the attacks, and no Nichirei/KFC data surfaced on the dark web. The companies expect to resume normal operations by the end of the week.
For a decade, Kubernetes was the right answer. It organized containers, scaled services horizontally and gave platform teams a shared vocabulary for running software in production. It abstracted away enough of the underlying complexity that engineers could stop thinking about servers and start thinking about services. Most cloud-native infrastructure today is built on top of it, directly or in spirit, and EKS made that model the default for the majority of enterprise teams running workloads on AWS.
The workload that defined that era was the stateless HTTP request, fast in, fast out, disposable. A user action triggers a request, the request hits a service, the service returns a response and the container is done. Kubernetes was optimized for that pattern down to the scheduler internals: Bin-pack containers onto nodes, autoscale on CPU and memory, evict and reschedule when something goes wrong. The whole system is tuned around the assumption that individual units of work are short, stateless and interchangeable.
That assumption no longer holds for the workloads that matter most right now.
The agent workload is structurally different
Agents are long-running, stateful processes. They reason across time, call external tools, spawn subprocesses, write and execute code, and make decisions that depend on what happened five steps earlier in the same task. A single-agent workflow might run for minutes or hours, touching a dozen external systems and generating intermediate outputs that subsequent steps depend on. The compute layer for that kind of work needs to do things the old model was never asked to do. That is the new pattern: Execution infrastructure designed around agent semantics rather than request semantics.
The Kubernetes community itself has acknowledged this mismatch. In March 2026, Kubernetes SIG Apps published an introduction to Agent Sandbox, a new CRD-based abstraction designed specifically for singleton, stateful agent workloads. The framing is direct: The ecosystem is moving from short-lived, isolated tasks to deploying multiple, coordinated AI agents that run continuously, and mapping those workloads to traditional Kubernetes primitives requires an entirely new abstraction. The fact that the Kubernetes maintainers built a dedicated primitive for this, rather than recommending teams compose one from existing resources, is itself the clearest signal that agent execution does not fit the old model.
What agent execution actually requires
Concretely, it requires four things. First, isolated execution environments that provision in milliseconds, not minutes, so each agent task gets its own sandbox for code execution and tool calls without blocking the reasoning loop. The difference between a two-second environment and a two-minute environment is not a performance optimization; it determines whether the architecture is viable at all. Second, durable state management across the full task lifecycle, so an agent can pause, hand off or resume without re-initializing from scratch and burning tokens to reconstruct context it already built. Third, coordination primitives for multi-agent work: The ability to spawn subagents, pass structured outputs between them and track task dependencies across a graph of concurrent processes. Production agent systems are rarely single agents; they are pipelines of specialized agents with handoffs that need to be reliable and inspectable. Fourth, credentials and secrets management that travel with the execution context, so agents can authenticate to external services securely without exposing credentials in the task definition, logs or the environment variables of a shared container.
The mismatch shows up fast in production
Kubernetes and EKS expose the mismatch quickly in practice. Pod eviction terminates an agent mid-task with no clean recovery path. Autoscaling reads CPU utilization as the load signal, but an agent holding a long inference connection looks idle to the scheduler even when it is doing the most consequential work in the pipeline. Provisioning a new environment takes 45 seconds to two minutes on a well-tuned cluster; agent workloads need that in under two seconds or the reasoning loop stalls and the user experience degrades visibly. These are not edge cases or misconfigurations. They are the normal operating conditions for production agent workloads running on infrastructure that was not designed for them.
The utilization data makes the broader cost picture even starker. The 2026 State of Kubernetes Optimization Report from CAST AI, drawn from analysis of over 23,000 production clusters across AWS, Azure and GCP, found average CPU utilization at 8 percent, down from 10 percent the year prior. Memory utilization fell from 23 to 20 percent. CPU overprovisioning jumped from 40 to 69 percent year over year. These numbers reflect clusters running traditional workloads, and the pattern is worsening, not improving, as environments scale. Agent workloads compound this problem further. An agent holding an open inference connection or waiting on a tool call registers as idle to a scheduler that reads CPU and memory as the only meaningful load signals. The infrastructure responds to the wrong metric, overprovisioning capacity for demand it cannot measure, while the actual bottleneck, environment provisioning latency and state continuity, goes unaddressed.
Security is not the same problem it was before
Agent workloads change the threat model at the infrastructure level. A compromised stateless service exposes a narrow surface defined by its API contracts. A compromised agent exposes every system it can reach, every credential it holds and every action it is authorized to take on behalf of the user. Agents generate and execute their own code, make non-deterministic tool-call decisions and accumulate context across long-running sessions. Standard container namespacing does not contain that kind of risk. Kernel-level isolation, default-deny network egress, scoped credentials per session and agent-aware observability are not optional hardening steps. They are baseline requirements for running agents in production.
What teams that ship agents have already figured out
Some of the clearest evidence for this shift comes not from infrastructure vendors but from product engineering teams running agents at scale on their own code. In late 2025, Ramp’s engineering team published a detailed account of building Inspect, their internal background coding agent. Each Inspect session runs in a sandboxed VM with a full-stack development environment and deep integrations across their observability, CI, and deployment tooling. The architecture requirements map almost exactly to the four primitives above. Filesystem snapshots keep sessions starting in seconds rather than minutes. Sessions are isolated and stateful. The agent can run tests, review telemetry, query feature flags and visually verify frontend changes in a real browser. And the whole system supports unlimited concurrency, so engineers can spin up ten parallel sessions exploring different approaches to the same problem without contention.
The results speak for themselves. Within months of launch, roughly 30 percent of all pull requests merged to Ramp’s frontend and backend repositories were written by Inspect. That level of adoption was not mandated. It happened because the execution environment was fast enough, capable enough and well-integrated enough that the agent was strictly better than a local workflow for a meaningful share of tasks. The key insight from the Ramp case is not about the model. It is about the execution layer. As their team put it, session speed should only be limited by model-provider time-to-first-token; everything else, like cloning and installing, needs to be done before the session starts. That is a statement about infrastructure, not intelligence.
The ecosystem is catching up, but defaults are sticky
None of that is a criticism of the tools. Kubernetes solved exactly the problem it was designed for, and it solved it well. The issue is that infrastructure defaults are sticky. Teams inherit them, build on top of them and optimize within their constraints long after the underlying workload has changed. The Kubernetes community’s own response, the Agent Sandbox project under SIG Apps, validates the thesis that a new abstraction is necessary. The new primitives the community is building include warm pools for near-zero cold starts, lifecycle management for suspending and resuming idle agents without losing state, and pluggable kernel isolation for secure execution of untrusted code. These are not incremental improvements to existing resources. They are net-new abstractions that acknowledge the old model does not stretch to fit.
But adoption of purpose-built agent infrastructure remains early. Enterprises building agent pipelines today are largely running a request-oriented orchestration model against an execution-oriented workload, and the mismatch shows up in task failure rates, runaway costs and debugging cycles that have no good tooling because the observability layer was also designed for stateless services.
The structural advantage is available now
The infrastructure to close that gap exists now. The prerequisite is recognizing that agent execution is a first-class compute pattern with its own primitives and its own requirements, not a variant of the stateless service model that defined the last decade. Teams that make that shift early will have a meaningful structural advantage. The ones that do not will spend the next two years wondering why their agent systems are unreliable at a scale that should be tractable.
This article is published as part of the Foundry Expert Contributor Network. Want to join?
Psychologist Candice Odgers has studied adolescent mental health for 25 years. She fears the current debate around smartphones obscures some of the biggest issues facing teenagers – from the impact of Covid to the health of their adult caregivers
The quickest way to make being online safer for children and teens would be to kick all adult men off the internet, the Canadian psychologist Candice Odgers believes. Men are the biggest perpetrators of sextortion and most likely to spread misinformation, she says.
Odgers is not recommending this as a policy for governments to adopt: “That would be crazy, right? It would be unfair.” But she is on a drive to puncture the prevailing narrative that the best way to address online harms is a social media ban for teenagers.
Whistleblower suggests internal security services deployed spyware from 2017 against key domestic and foreign targets
A former member of Morocco’s domestic intelligence service has helped to provide an unprecedented insight into how the north African state used hacking software – including Pegasus spyware – to target journalists, human rights defenders, French politicians and Spanish cabinet ministers and police officers.
Pegasus, which is manufactured by the Israel-based NSO Group, allows its operator to access everything on a target’s mobile phone, including emails, text messages and photographs. It can also activate the phone’s recorder and camera, turning it into a listening device.
A new Quordle puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Wednesday's puzzle instead then click here: Quordle hints and answers for Wednesday, July 15 (game #1633).
Quordle was one of the original Wordle alternatives and is still going strong now more than 1,500 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today — or scroll down further for the answers.
Enjoy playing word games? You can also check out my NYT Connections today and NYT Strands today pages for hints and answers for those puzzles, while Marc's Wordle today column covers the original viral word game.
SPOILER WARNING: Information about Quordle today is below, so don't read on if you don't want to know the answers.
Quordle today (game #1634) — hint #1 — Vowels
How many different vowels are in Quordle today?
• The number of different vowels in Quordle today is 4*.
* Note that by vowel we mean the five standard vowels (A, E, I, O, U), not Y (which is sometimes counted as a vowel too).
Quordle today (game #1634) — hint #2 — repeated letters
Do any of today's Quordle answers contain repeated letters?
• The number of Quordle answers containing a repeated letter today is 2.
Quordle today (game #1634) — hint #3 — uncommon letters
Do the letters Q, Z, X or J appear in Quordle today?
• No. None of Q, Z, X or J appear among today's Quordle answers.
Login
