A lot of the best meal kits offer the satisfaction of making a home-cooked meal, the chance to learn a little culinary craft, maybe even a chance at some pride in your handiwork. Hungryroot (7/10, WIRED Review) isn’t really like that. Hungryroot instead wants to replace your trip to the grocery store with a few good recipes.

After starting as a grocery service, Hungryroot spent years piecing together proprietary AI long before AI became the byword of everything from your electric shaver to your literal kitchen sink. What Hungryroot’s AI does is geared to creating personalized food menus and shopping lists, based on your every micro-preference and whim, whether that’s a love of olives or a hatred of figs. A box then shows up at your door, with a mix of basic provisions, silly snacks, and desserts, along with recipes and usually some pre-prepped ingredients.

Cooking is often more assembly, with recipes pieced together from premium ingredients you’d expect to find at your local organic store. A fun sauce here, a pre-cooked protein there, a corn tortilla that just wants to be livened in the pan. Most meals come together in 15 or 20 minutes, in my experience—unless they need time in the oven. A bot concierge and pre-mixed sauces do cost a bit, however, and that’s where Hungryroot promo codes come in. There are Hungryroot coupons that offer 30% off or more, and Hungryroot coupon codes that offer extra free gifts for every box, for life: a protein here, a cookie there. Here’s what’s up.

Get 30% Off Plus a Free Gift With a Hungryroot First-Time Discount

The current Hungryroot discount code is good for a cool 30% off your first week’s box from Hungryroot. To qualify for the discount, you’ll need to sign up for a plan totaling $99 or more a week, including breakfast items, snacks, and produce in addition to recipes and ingredients. The Hungryroot promo offer also includes a free gift, which usually includes an option to get an extra premium protein like chicken or beef. You can see more information here.

Take Advantage of Hungryroot Coupons for Referrals

Like a lot of meal kits, there’s also a Hungryroot discount code for referring your friends to the Hungryroot service—and this usually benefits both sides of the equation.

Your friend who orders after using your referral code gets a cool $50 off their first grocery delivery of $100 or more—which adds up, potentially, to half off. The person who refers their friend also gets $50 off their next order, once their friend finalizes their first grocery delivery of $100 or more. So basically, everybody gets $50.

Follow this link for info on the Hungryroot referral program. Unlike other offers, the referral credits seem to be stackable with other offers—so you don’t miss out on any discounts by taking a referral.

Unlock of Hungryroot Coupons for New and Returning Customers

Most subscription services have hidden levels. The way to access them is to change your mind a lot. Just like some magazine subscriptions and almost all delivery meal kits and grocery subscription services, Hungryroot will chase you down with offers if you cancel a subscription or change your mind after already giving them your email.

What’s this mean? Usually, it means that in the days or weeks after changing your mind about Hungryroot, you’ll get a coupon in the mail. So if you take Hungryroot’s quiz about your dietary preferences, then don’t order? A few days later, you’ll likely get a sweet Hungryroot coupon code in your inbox. New subscription plans usually also come with free treats.

If you cancel or pause an existing account, you may also receive a Hungryroot promo code offering your $50 or even $100 off your next order, as an incentive to return.

After months of leaks and rumours, Apple has officially unveiled its new Pro iPhones for 2026, alongside the much-anticipated iPhone Duo foldable handset and a slew of other product launches.

If you’re looking to upgrade to the iPhone 18 Pro, iPhone 18 Pro Max or iPhone Duo, I’ve put together this guide on everything you need to know about pre-ordering any of the three handsets in Australia, including when preorders start, how much you’ll pay and all the best deals we’ve found.

Aside from the official Apple Australia store, as with any iPhone launch a wide range of retailers and telcos across the country are also offering pre-orders for all three phones, and we’ll update this page as soon as the offers go live.

Want to know more about the new iPhones and other new products that Apple unveiled? Scroll to the bottom of this page or check out our full coverage of Apple’s launch event.

Apple iPhone 18 Pro and Pro Max pre-orders in Australia: What you need to know

The iPhone 18 Pro starts at AU$2,099 for the base 256GB model, and the iPhone 18 Pro Max starts at AU$2,299. Retailers JB Hi-Fi and The Good Guys, as well as telcos Vodafone, Optus and Telstra, have confirmed they will all carry the new handsets.

Aside from Apple’s official Australian store, JB Hi-Fi, Vodafone and Optus will also accept trading in your old phone for a credit towards the iPhone 18 Pro or Pro Max. Apple is offering up to AU$1,595 in trade-in credit – that’s based on an iPhone 17 Pro Max, while a comparatively ancient iPhone X will net you a AU$70 credit.

If you’re going the trade in route, you can calculate your old phone’s credit value at Optus, Vodafone or JB Hi-Fi to get an idea before pre-orders go live.

Apple is also offering a new financing program in Australia via Afterpay, allowing monthly payments over 6-, 12- or 24-month terms instead of the service’s usual four payments spread out fortnightly. If you’re using Afterpay, the iPhone 18 Pro starts at AU$87.46 per month over 24 months, while the 18 Pro Max starts at AU$95.79 per month. This monthly payment option is only available through the Apple Australia website and stores — you won’t find it at third-party stores.

Apple iPhone Duo pre-orders in Australia: What you need to know

Prices begin at a slightly eye-watering AU$3,499 for the base 256GB model (for reference, Samsung’s equivalent Galaxy Z Fold 8 starts at AU$2,699), and go as high as AU$5,899 for the 2TB configuration.

Like the 18 Pro and Pro Max above, you can also trade in your old phone to receive up to AU$1,595 off the price. You can get an estimate on your trade-in credit at Optus, Vodafone or JB Hi-Fi to get an idea before pre-orders go live.

Financing via Afterpay is also available for the iPhone Duo, starting at AU$145.79 per month for 24 months.

Headlining Apple’s new handset range for 2026 is the iPhone Duo, the company’s first attempt at a foldable and its answer to the Samsung Galaxy Z Fold 8. Featuring a 5.4-inch outer display with a 7.6-inch Super Retina XDR inner display, the device looks somewhat like a squat iPhone mini when folded, and an iPad mini when opened.

It features the new A20 Pro chipset and a titanium frame, which Apple says is reinforced by ceramic fibres. The Duo, however, also has a few core iPhone features missing, like Face ID, a telephoto camera and the action button, but is still undeniably the star of the show. There are also some exclusive software features that look to take advantage of the larger folding display, like some new Liquid Glass animations and multitasking features. Moreover, the Duo will also get Apple Pencil support, which is a first for any iPhone.

As for the iPhone 18 Pro and Pro Max, the biggest upgrades are the main cameras, with each getting a new variable aperture to let users adjust the light depending on what they need to photograph, for example.

They also get the new A20 Pro chip, which Apple claims will have a 20% faster CPU and 40% faster GPU than the A19 Pro found in the iPhone 17 Pro, Pro Max and iPhone Air. Battery life is also said to be improved, with Apple claiming the iPhone 18 Pro will have 36 hours of video playback — matching the 17 Pro Max — and the 18 Pro Max now rated for 45 hours.

We’ll be posting our reviews for all three devices over the next few weeks, so stay tuned for our thoughts and hands-on experience.

Sony has just made a decision that feels straight out of a “how to lose fans” playbook. In a statement posted on X, PlayStation Studios confirmed it is stepping away from Hideo Kojima’s PHYSINT after what it called three decades of “industry-leading, genre-defining” collaboration. Barely minutes later, Xbox confirmed it would publish the game instead, expanding its existing partnership with Kojima Productions.

What exactly happened between Sony and Kojima?

Kojima cleared up the timeline himself. He revealed that PlayStation Studios quietly canceled PHYSINT back in mid-June, catching his studio off guard. His team then spent the next three months searching, negotiating, and asking around before landing on a partner who shared their vision for the future. 

That partner turned out to be Xbox. Xbox CEO Asha Sharma called it an honor that Kojima “chose Xbox,” while Kojima confirmed the collaboration now extends beyond games into film and television, and covers both OD and PHYSINT, the ambitious action-espionage title he has described as genre-defining. Sony, for its part, insists its bond with Kojima “remains strong,” even as it walks away from his newest project.

from the inevitable dept

I want to start this whole thing off with a disclaimer. Nothing in this post should be misconstrued as gloating, as being pleased, or as having a connotation that is anything beyond sadness and the need for change. Nobody is happy to have the obvious confirmed when it comes to a child’s death. There is still a very real family in very real pain as a result of a life that ended far, far too short. All that matters here is ensuring that pain isn’t replicated elsewhere as a result of the many mistakes that led to this death.

That being said, it is with the soberest thoughts that we learned that one of the two recent deaths of children in Pennsylvania was in fact due to measles, as confirmed by the local coroner there. When these deaths were announced by state government officials a few weeks ago, the response from HHS Secretary RFK Jr. was, predictably, to dive into conspiracy theories in stead of confronting the reality of what his decades of anti-vaxxer nonsense has produced. He claimed that there had been no confirmation that the deaths were caused by measles. He pointed out that one of the children has a cause of death listed from a ruptured spleen, not bothering to acknowledge that is a potential effect of having measles at birth. He suggested obliquely that some deaths may have simply been made up by state government officials.

All he had to do was wait and he wouldn’t have made a fool of himself.

Stephen Diamantoni, a Republican elected coroner in 2007, told Lancaster Online that the baby died at home. “I believe it was the 18th of August,” he said.

He also noted that the baby had a genetic condition called Amish lethal microcephaly, in which babies are born with unusually small heads and underdeveloped brains. The condition is caused by a mutation in the SLC25A19 gene, which codes for a protein involved with energy-producing enzymes in the mitochondria and is thought to be important for brain development. About 1 in 500 babies in the Old Order Amish population of Pennsylvania is born with the condition. Infants with the disorder only survive for about six months.

Advertisement

Diamontoni confirmed to Lancaster Online that while microcephaly was present, measles was the cause of death.

And here is where you’re going to hear the anti-vaxxers start their spin game. Some may claim that the coroner is wrong and the presence of microcephaly means that’s what killed the child. Some may dismiss the death entirely due to the likely early death this child would have suffered as a result of the same, which is an absolutely evil thing to suggest.

And I have no doubt that others will point out that none of this is RFK Jr.’s fault, because both children were members of the Amish community and they don’t let their children get vaccinated due to their religion. And, just so we’re all clear about this, that isn’t true. There is no religious prohibition in the Amish community against being vaccinated for measles. The Amish are wildly under vaccinated, to be certain. But Amish settlements typically have something like a 10%-30% vaccination rate for measles.

This information all came out last week. Amazingly, the CDC’s measles tracking site still has an asterisk next to the number of reported deaths from measles in 2026 at the time of this writing. The most recent reporting is that those deaths are being obfuscated at Kennedy’s direct request, wrapped in a bullshit excuse that local coroners haven’t confirmed the cause of death to be measles yet. Well, that excuse is gone, yet the website hasn’t been updated.

But what really needs to be gone is the scourge that is Kennedy’s tenure at HHS. For that to happen, Congress must act. Sadly, not enough of them seem to want to.

Filed Under: cdc, hhs, measles, pennsylvania, rfk jr., stephen dimantoni

Norton Antivirus is the only anti-virus software I’ve ever used, in part because many years ago I switched to Linux, which eliminates the need for antivirus software. If you’re still on macOS or Windows, though, Norton Antivirus is still a solid choice. None of these services is something anyone is excited to run, but they can save you bacon if you make a bad decision downloading questionable files or clicking links you shouldn’t have. The best way to stay protected without spending a fortune is to use a Norton coupon code to get yourself a deal.

Norton Coupons: Up to 58% off 360 and Antivirus Plans

While I’ve only ever used the Antivirus offering, the real way to save is to bundle several services and use a coupon code for the lot. With Norton that means Norton 360, which gets you the Antivirus, malware, ransomware, and hacking protection, along with a VPN and dark web monitoring, for up to three devices. You also get a password manager and 2GB of online storage. With this Norton coupon code, you get all that $40 for the first year, which is about the same as paying for a VPN alone. Norton 360 Deluxe gets you everything that comes with the regular version, but it works with five devices and gives you 50GB of online storage.

Get Unbeatable Norton With LifeLock Protection From $8.33 per Month

LifeLock is Norton’s usual online protection suite, but adds up to $1 million in potential coverage if you experience financial loss due to identity theft. With this Norton coupon code, Norton 360 with LifeLock is just $8.33 per month.

Norton Anti Virus Promo Codes: 50% Off 360 Antivirus Plus

If you want to get (and stay) serious about your privacy and protection, and the overall health and longevity of your computer, I’d suggest investing in an anti-virus protection plan for your whole setup. With Norton 360 Antivirus Plus, you’ll get powerful antivirus protection for your device, including AI-powered Scam Protection, Password Manager, and 2 GB Cloud Backup. And you can get even more peace of mind with a 60-day money back guarantee. Best of all, Norton 360 Antivirus Plus is 50% off right now, taking it from $60 per year to only $30—which works out to being just $2.50 per month—a small investment worth so much more in protection and peace of mind.

​​Save $100 on Norton Cybersecurity for Small Businesses

If you want to protect more than a few devices, the small business plans are the way to go. These coupons will get you $100 off the Small Business Premium plan, and $80 off Small Business. These plans cover two devices per user and up to five users.

Test Out Norton 360 for Yourself With a Free Trial

Not sure if Norton 360 is worth the money? You can try it out for seven days, though set yourself a reminder because after the initial 7-day free trial, your paid subscription will begin unless you cancel before.

SECURITY

Toothless security

PWNED Welcome back to PWNED, the weekly column where we highlight examples of how not to handle your security. This week’s tale of woe comes from a very unhealthy part of the healthcare sector.

Have a story about someone leaving a gaping hole in their network? Share it with us at pwned@sitpub.com. Anonymity is available upon request.

Our story comes courtesy of Chris Kirksey, founder and CEO of Direction, a digital marketing and SEO company that works in the healthcare industry. He also does security audits of his clients’ systems.

Last year, Kirksey was checking out a dental practice’s systems and noticed something strange. There were three accounts that had admin access to the patient database, including one that belonged to a scheduling company the dentists had stopped using all the way back in 2021. 

The account had been active for at least three years and could access 4,000 patient records. Leaving an unnecessary account with access to protected health information created a potential HIPAA compliance risk, particularly if someone no longer authorized to view the data could still get to it.

The office manager responsible for using the system didn’t even know that this dangerous login existed. Apparently, a contractor who set up the account never told anybody, then left the company. Because no one knew that the account existed, no one knew to kill it.

Kirksey immediately set about getting rid of all three admin accounts he found on the dental practice’s system. He then set up new policies for his client.

“I built a permanent rule after that,” he said. “Every vendor relationship that ends now triggers an automatic access shutdown and the full list gets reviewed twice a year no matter what.”

Since the incident, Kirksey has found similar security holes at six other healthcare practices he has worked with. Yikes!

“Everyone worries about the sticky note with a password on it or the file just called passwords.xls, because those get caught fast and make a good story,” he told us. “Nobody worries about the login they forgot even exists, and that is usually the one still wide open years later, causing real, unseen damage.”

The lesson here is pretty straightforward. You need to see all of the accounts that have access to your data and make sure that they all have a reason to exist. Conduct regular audits, even if nothing seems wrong.

And, as we’ve seen before, zombie accounts can kill. When an employee or contractor leaves, check not only which accounts they used, but also which accounts they created while doing the job. ®

Apple could be about to give the iPhone 18 Pro a much more capable camera system, with several features discovered in iOS 27 pointing towards a proper pro-style shooting mode.

MacRumors forum member pdfu found references to five camera upgrades in the iOS 27 code, including manual focus, focus peaking, waveforms, histograms and an overexposure warning. These tools would give photographers considerably more control than the standard iPhone camera app currently offers.

Waveforms and histograms would provide a more detailed look at an image’s exposure, making it easier to spot areas that are too dark or too bright before taking a photo. Focus peaking could also highlight the edges of subjects that are in sharp focus, which should be particularly useful when manually controlling focus.

Through the looking glass: A new camera with the first-ever aperture controls on an iPhone is the main star of this year’s iPhone presentation. With the iPhone 18 Pro, Apple is introducing a more versatile camera, more powerful photo-customization software, and yet another tech giant’s attempt to distinguish authentic images from AI-generated creations.

Pre-orders for the iPhone 18 Pro and Pro Max open on September 12, with availability beginning September 18 starting at $1,199. As previously rumored, Apple seems to have delayed the standard iPhone 18 until next year.

The iPhone 18 Pro is the first iPhone to support aperture control. With six blades and a new sensor, users can smoothly switch between ƒ/1.48, ƒ/1.8, ƒ/2.8, and ƒ/4.0 apertures to control lighting and depth of field, as well as enhance low-light photography.

Users can also apply new texture styles and cinematic effects to videos recorded at up to 60fps. Additionally, time-lapse mode now supports 4K and Dolby Vision HDR recording. Meanwhile, the new iPhone adjusts the Dynamic Island to show up to three Live Activities simultaneously while still enabling Face ID.

Like Google and other developers, Apple also aims to provide a tool for watermarking AI-edited photos and verifying unaltered images. Apple Reference Image creates a duplicate of each new photo with cloud-processed, signed sensor data to prevent editing. Users can compare it with the copy they normally use to check for changes. A new API also makes reference images available to third-party apps, but the feature’s resilience against ever-evolving AI tools in the wild remains untested.

The processor in this year’s iPhone also receives another die shrink. The A20 Pro features TSMC’s new 2-nanometer process, increasing memory bandwidth by 50% and GPU performance by 40% over the iPhone 17 Pro’s A19 Pro. A new dual 16-core Neural Engine also enhances photography and AI performance, powering the iPhone 18’s new photo features and iOS 27’s AI-powered Siri.

A new vapor chamber improves thermal performance by 40%, while a new battery design extends the iPhone 18 Pro’s maximum battery life to 36 hours and the iPhone 18 Pro Max’s to 45 hours. The standard model can charge to 50% in approximately 15 minutes via a wired connection or 30 minutes wirelessly.

Storage options include 256GB, 512GB, 1TB, and a new 2TB tier. Available colors include burgundy, glacier, silver, and black.

LG on Wednesday denied that its TVs eavesdrop on customers after a gaming site appeared to show that the TVs can be used to listen to conversations in real time.

Also read: You’ve Heard About Smart Home Hacking: Here’s How It Works and How Likely It Is

“The claims made in the recently published video are not true,” an LG representative said in a statement to CNET. The company said LG TVs process voice data only when the voice button on the remote control is pressed and held or the TV’s “wake word” is uttered by the customer.

“Other than these instances, the TVs do not collect or record ambient conversations,” the statement said.

The company said that if the device doesn’t recognize the wake word, no voice data is transmitted to the server, and that the wake word is detected locally and immediately deleted.

Gaming site GamersNexus reported earlier this week that LG TVs can be hacked by a third party to become listening devices, even if a user disables the onboard microphone.

GamersNexus says it used a man-in-the-middle attack on an LG G5 TV to enable it to listen to audio from a remote computer, and says the TV was able to record audio without the TV being connected to the internet. In a man-in-the-middle attack, the hacker intercepts sensitive data by putting itself in between two other entities, such as a user and a web application.

The 2-hour video, compiled with the help of the Level1Techs site, says that hackers could listen to a USB webcam or an HDMI input even when the physical switch located next to the mic was switched off. It said this alleged vulnerability was possible due to the TV’s Automatic Content Recognition feature, which transmits viewing habits and voice requests back to LG.

“LG TVs are able to be utilized as eavesdropping listening devices, in part because of the advertising functionality that’s built into them. We were able to retrieve plain text transcripts of a voice conversation and inputs from the TV,” the video says.

The video also reports that the researchers were able to see a list of the IP addresses of all the other devices on the network, even if those devices weren’t connected to the TV.

“LG TVs feature the ability to scan for and connect to nearby devices on the same network. This is a standard function commonly provided by smart TVs and smart home devices,” an LG statement said.

The LG TV microphone, like many smart devices, is designed to listen for a wake word or phrase such as “Hi TV” and then perform the requested task. 

Meanwhile, Apple says its new “always listening” Audio Intelligence feature doesn’t let the company access users’ raw audio. 

Almost all TVs and streaming devices use automatic content recognition to target ads to customers based on viewing habits, but users can usually disable this feature in the device’s settings.

It’s worth noting that any smart speaker will record audio for a limited time after hearing its wake word. If you somehow disclose your Social Security number during that time, for instance, it will be recorded and transmitted.

Smart-home attacks are rare, but the researchers recommend removing LG TVs from your network to prevent unauthorized data collection and privacy breaches. Unprotected computers and routers, not televisions, are the most likely targets for hackers.

Senior editor Ty Pendlebury has been reviewing TVs and home entertainment for CNET since 2006. He has a cat named after one of the best TVs ever made.

See full bio

Ask your LLM AI about the project you’ve spent months building, and it may remember your name, your writing style, and the budget you mentioned last week. For a moment, that feels like being remembered. Then open its memory settings and look at what’s actually there. You won’t find any evidence that the underlying model learned how to work with you. What you’ll find instead are fragments assembled from your chats, files, and connected apps. Puzzle pieces that can resemble a complete picture from a distance, even though the model never built the full picture.

The leading LLM assistants now describe this in their own support material. One calls its current memory a continually updated synthesis and says its earlier version worked like a notepad stored separately from chat history. Another searches old conversations and builds a memory you can view, edit, disable, export, or replace.

Those features are useful, and they can spare you from having to repeat your preferred writing style, the name of a project, or the fact that you hate morning meetings, but they also show the difference between retrieving details about you and understanding how those details fit together.

While you hope that the screen you’re looking at fresh today may remember, the fact is that the model at the center of the experience remains the same pretrained model that everyone else gets to experience in much the same way and with the same limitations.

Saved context is not learned experience

So what’s all that saved memory actually doing, if it isn’t teaching the model anything? Less than it feels like. Because while it seems to reach back into your earlier chats, pick out whatever looks relevant, and lay some of it down beside your next question, and the answer that comes back does feel more personal, it’s easy to mistake that as a sign that the thing is finally getting to know you. It isn’t. It just has better notes in front of it than it did a few minutes ago, and because people want it to understand them, they may start actually believing that it’s getting to know them better. 

And that difference stays out of sight as long as everything’s going smoothly and the assistant remembers your project by name, it brings back the preference you mentioned yesterday, and it all feels like continuity. And that’s the trap, because you only catch the illusion when the work turns complicated.

Say it remembers that Julie approved a budget, but it never caught that Julie moved to a different role three weeks back. So it hands you the decisions she made, and yet it loses the position and facts that forced a newer budget to be created in the first place. Or it gives you an instruction that a newer iteration of the developing budget already cancelled. And while each of those memories is perfectly true on its own and for when they occurred, the story they add up to and present to you as today’s fact is still wrong.

And that’s the thing about storage. A filing cabinet remembers too. It keeps whatever you put in it, and it gives a folder back when you know what to search for. But nobody’s ever confused the cabinet with the human co-worker standing next to it, the one who knows why that folder mattered, what changed after it was filed, and which part of it you should still be listening to today.

Long conversations expose the limit

Now see what happens when you ask that filing cabinet to follow a single thread stretched thin across hundreds of folders with dozens of turns and hundreds of separate and sometimes subtle sessions. Julie approves the budget, and then she changes roles, and then she pushes back on something you decided later, and then she’s gone for weeks on vacation before she turns up again. So just pulling her name out of the pile you’re working on isn’t going to work. The real work is knowing which facts about her work still hold, what came before what, and why the whole relationship bent the way it did along the way.

That’s exactly what a group of researchers set out to measure with a benchmark called LoCoMo, built around conversations that run an average of 300 turns across as many as 35 sessions. They gave the LLM models every advantage, longer context windows and better retrieval, and the models improved. But people still ran circles around them. The systems trailed human performance by 56 percent overall, and that shortfall stretched to 73 percent once the questions leaned in more heavily on timing and sequence. Which is to say, they did fine when remembering meant finding a sentence, but then fell apart when remembering meant actually understanding how one thing changed everything that came after it.

So adding on more and more storage was never going to finish the job, because holding onto every sentence was never the hard part. The hard part is judgment. It’s knowing which details belong together, and which one rewrites the meaning of another, and which source has earned some weight while another hasn’t. It’s knowing what’s still true this morning, and what should be allowed to go soft and fade away because it was only ever the noise of a passing exchange.

A useful memory should separate what matters from what doesn’t. It should keep the key decision that turned a whole project around and let the small talk dissolve. And when two facts arrive and contradict each other, it should hold onto the one that explains the change instead of letting the newest sentence walk in and take over the direction.

Memory has to participate in thought

This is where it stops being a storage problem and starts being an architecture one. 

For most LLM AI, memory is a drawer off to the side that the model opens when it needs something. But Vertus takes a different architectural approach. It isn’t an LLM with a bigger scrapbook taped to the side. Vertus calls its architecture a Cognitive Reasoning Superintelligence, and in that design memory isn’t a drawer at all. It’s the bedrock the whole system is built on and reasons from.

Vertus calls this Cognitive Memory, with people, projects, decisions, evidence, and contradictions held as connected knowledge inside a persistent structure called the Cognitive Knowledge Lattice, rather than scattered as loose scraps waiting to be fetched. 

So when Julie changes positions, her old role isn’t wiped and the new one isn’t dropped in beside it like a stranger. The link between them carries the span of time it held, as well as the evidence that backed it, and the moments it changed. When an instruction gets replaced, the system keeps the change and the reason for it. And when the same claim shows up a thousand times over with nothing new behind it, all that repeating still doesn’t get to pass itself off as somehow elevated importance.

It also can tell the difference between the fresh and the old in the same way that you do. Whatever you’re working on right now stays close, while the routine back-and-forth loosens its grip over time, and the moments that genuinely mattered keep shaping what comes next. Nobody’s trying to remember everything forever, because remembering everything just builds an attic nobody can walk through. The point is to keep the shape of what’s been learned, tightly enough that memory actually changes how the next problem looks when it arrives, instead of sprinkling a familiar name on top of an old answer that was already written.

And that’s the real turn. When Vertus builds a fresh structure to think a question through, Cognitive Memory shapes that structure right from the first move. It isn’t fetched later and then somehow laid beside the answer after the important thinking’s already underway. It helps decide what the problem even is, and whose history matters to it, and which way through the evidence makes sense right now. The memory is doing the thinking, not watching it.

Tomorrow reveals what the memory is worth

Now, several of the big LLM Cartel assistants can remember more than they could a year ago. But that only makes the next question even more important. What kind of memory sits underneath the improvement?

Is it a tidy summary the system keeps about you, a transcript it searches, a few details it sets down next to your next request? Or is it experience hardening into structure, and changing how the request is understood before an answer even begins to form? Both can feel like continuity. Both can feel like being known. But one merely retrieves what you said while the other lets what it learned change what comes next.

And you feel the difference when the tab closes, and the week rolls over, and the project moves on, and yesterday’s simple little fact turns into today’s complicated quandary. 

Any notepad can tell an LLM what you said. A living memory helps it understand what those words actually mean and what they’ve come to become since.

A better test isn’t whether an AI can read yesterday back to you word for word. But rather whether what happened yesterday changes how it approaches today.

Digital Trends partners with external contributors. All contributor content is reviewed by the Digital Trends editorial staff.