Home Blog

Are Counterfeit, Scam SD Memory Cards Really a Problem?

0


A Lexar Professional 128GB SD memory card with 250 MB/s speed rests on a textured piece of dried wood or bark in natural light.
Lexar

This week, Lee Morris at Fstoppers published a video titled “Your Memory Cards Are Probably Fake.” In it, he asserts that scam companies are so prevalent and so good at faking their products that it’s likely nearly every photographer is a victim.

That seems like a pretty broad statement to make, but given the rising cost of memory, it might be more of an issue now than it was just a couple of years ago. This is always a pretty baffling decision, but photographers will spend thousands of dollars on a new camera, and then try to find the cheapest possible deal on memory cards. That is the mentality that these scam companies try and take advantage of.

In the video, Morris shows multiple SanDisk-branded memory cards that he acquired through a mix of Amazon and eBay and notes that, side by side, they look identical to one another. The illegitimate card is visually indistinguishable.

“SanDisk, specifically, is always changing the design on their boxes and on their memory cards. So, it’s not like you can easily look up what the card or the packaging should look like online because just when you think you know what you’re talking about and you know what the box should look like, SanDisk comes out with a new design,” Lee says.

He’s not wrong, and he was even tripped up by the major design change SanDisk went through recently where it redid its logo as part of its split from former parent company Western Digital.

What is a Scam Memory Card?

As Morris points out, the strategy behind a scam memory card is pretty much the same as when companies do the same for scam SSDs or HDDs. In a nutshell, these scammers will use an official-looking case and fill it with lower-end memory, then sell it at the price of the higher-end product.

They do the same for SD cards, but in that case, it’s much more difficult to diagnose, as the internals of a scam SD card are visually identical to a legitimate one. The difference will only make itself known in one of two ways: either the card will not have as much storage capacity on it as is claimed on the label, or the speed of the card will not be as fast as claimed.

The only way to check if a card is legitimate in this case is to run tests on it. Morris recommends a software called F3XSwift, which will test the card for its true capacity, as well as Blackmagic Disk Speed Test to check for its actual speed. PetaPixel also recommends Blackmagic Disk Speed Test, but cannot comment on the viability of F3XSwift.

How Scam Memory Cards Get You

For the video, Morris purchased cards from Amazon and eBay. Suffice to say, photographers should never be buying memory cards on eBay, so Morris’s decision to do so for the purposes of this video feels a bit disingenuous to how a responsible photographer would act. That said, there are legitimate issues with Amazon and Walmart.

Lee asserts that because of the way Amazon and Walmart set up their online storefronts, it’s easy not to notice the extremely important detail of who the seller of a product actually is.

And he’s right.

Looking at Amazon today, the top options for a SanDisk SD card are a 256GB V30 card and a 128GB V30 card. The first card is sold by a company called MemoryWhiz while the second is shipped and sold by Amazon. In both cases, the product pages for these cards have a “Visit the SanDisk Store” link at the top. But when I follow that link and dig to find the SD cards, all of the SD cards there are shipped and sold by Adorama.

So, really, the only thing these cards have in common is the Amazon storefront, as all three will come from different sources.

Screenshot of an Amazon product page for a SanDisk 256GB Extreme PRO SDXC UHS-I memory card, showing the product image, price of $87.91, purchasing options, and delivery details.
This card, which is the top result for “SanDisk SD card,” is sold by Memory Whiz. The card is cheaper than the one on SanDisk’s Amazon storefront and does not bear the Prime logo.

Morris believes that sellers with names like “MemoryWhiz” are the most likely source of these scams. In his case, it was a seller called “Stavy Sales.”

“Who are they? I have no idea. And here’s what makes this impossible to keep track of. If you slide down here and you click on other sellers on Amazon, you can see there’s 20 different options here. And we can go through each one of these here and look at all of these different sellers of this exact same product. And as you refresh this page, this sold by could change,” he says.

“You might think you’re buying it from Stavy Sales, whoever they are, but without even noticing, you might just refresh the page and all of a sudden you are purchasing from another seller who is claiming to sell this exact product, but it could be a fake,” he continues.

A grid of SanDisk memory cards and SSDs for sale on Amazon, showing different storage capacities (32GB to 1TB) and prices. Most items are labeled "Best Seller" and display “Add to Cart” buttons.
SanDisk’s storefront on Amazon is at least in part managed by Adorama, a reputable seller. The cards are more expensive than other listings, but are supported by Prime. What’s odd is that Amazon is not showing these as top results in the initial search, greatly lowering their visibility and putting customers at risk.

“The only reason that I’m focusing on SanDisk in this video is because SanDisk are the only cards that I personally use. I trust them the most.”

That’s an interesting take, given that SanDisk is the only major memory manufacturer in the photography space that has a history of making business decisions that cause photographers to lose their data.

PetaPixel reached out to SanDisk to ask it for its recommended method for avoiding scammers, but the company did not respond to requests for comment.

What Can Photographers Do?

One of the claims that Morris makes in his video is that there are only a few companies that actually manufacture their own dies for memory cards.

“What I’ve recently learned is that there are only a handful of companies on Earth that are even capable of making the internal components of an SD memory card, and all of them are legitimate companies that would never make fake memory cards. Many of the other brands that I’m sure you’ve heard of who might sell more expensive memory cards than SanDisk don’t manufacture cards at all, and they are purchasing these pre-made chips from different factories around the world and putting their sticker on them,” he says.

This is a pretty cynical and oversimplified version of reality.

“On the relabeling point, it’s a fair thing to raise. The reality is that only a handful of companies actually fabricate NAND, so most card brands are sourcing flash and controllers rather than making them,” Chris Kooistra, Vice President of Marketing at OWC tells PetaPixel.

A desktop setup with a wireless keyboard, external hard drives, a camera lens, several SD cards, and a computer monitor on a wooden desk. The devices are neatly arranged beside each other.
OWC

“What separates brands is everything that happens after that: which grade of NAND gets selected, how the controller and firmware are tuned, how much validation each card goes through, and who stands behind it when one fails in the field. Firmware is a good example of why that distinction matters. Our Innergize software runs health checks, applies firmware updates, and keeps cards performing the way they should over time. It also only works with OWC cards and OWC readers, so the ecosystem itself does the verifying. A genuine OWC card runs in it. Something wearing our label that isn’t actually ours won’t. That’s an ecosystem doing real work, not a sticker. Put simply, if a memory card were just a label on someone else’s product, there would be no reason for OWC to be in this category at all.”

Morris didn’t talk to any memory card companies about this issue for his video. Instead, he speaks with a company called Salvage Data. They tell Morris that it’s not actually SanDisk that is the issue he sees the most, but rather scam versions of Lexar and Samsung.

PetaPixel reached out to Lexar and asked if it had experienced issues with counterfeit or fraudulent memory cards being sold as genuine Lexar products.

“As with many leading consumer electronics brands, Lexar is aware that counterfeit products can occasionally appear on open marketplaces such as Amazon, eBay, and others. Protecting consumers and maintaining product authenticity are top priorities for Lexar, and we actively monitor channels to address suspected counterfeit listings when identified so they can be removed,” Lexar says.

Two Lexar 1TB SD memory cards, one gold and one silver, are standing upright on a rocky surface with debris flying around them. Both cards are labeled "ARMOR" and display speed and capacity details on the front.
Lexar

“Lexar products undergo rigorous validation and quality testing before reaching consumers. Because counterfeit products bypass these controls, consumers may experience performance, reliability, or capacity issues that do not reflect genuine Lexar products.”

But while Lexar agrees that counterfeit, scam cards do exist, it does clarify that Morris’ methods aren’t clear or, more importantly, don’t reflect the actions of an informed customer.

“We cannot comment on the claims made in the video or the methods used to reach those conclusions. eBay is not an authorized reseller of Lexar products, so extra precautions should be taken when purchasing from non-authorized channels. It’s also worth noting that the video appears to be a paid collaboration between Salvage Data, a data recovery platform, and Fstoppers, as there is a partner ID tied to the URL link,” Lexar says.

This partnership is not disclosed by Morris.

“If the focus of the video is to both raise awareness of unauthentic products and offer image recovery services, it’s worth mentioning that Lexar offers free image recovery software, reinforcing our commitment to customers in what is typically a paid service from other brands and services,” Lexar adds, likely in reference to Salvage Data offering a similar service, but for a fee.

To avoid scams, Lexar says photographers should purchase only from authorized retailers, which can be found on the “Where to buy” page of its website.

“When shopping on online marketplaces, consumers should carefully review seller information and, when possible, look for products marked as ‘Ships from and sold by Amazon.com’ to help ensure inventory is fulfilled directly through Amazon’s first-party retail channel,” Lexar says.

“Consumers should also be cautious of unusually low prices that appear significantly below market value on open marketplaces, as these are often signals of unauthorized or counterfeit products. Purchasing through reputable retailers remains the best way to ensure product authenticity.”

“We haven’t seen that firsthand with our own product, but it’s a growing problem in the category,” Kooistra says regarding scam cards.

“Counterfeit and relabeled cards usually move through unauthorized sellers, look fine on the outside, and don’t reveal the lower-grade flash or fake capacity until the card is full or under load. The protection is the same one we’d give for any storage: buy direct or from an authorized reseller, treat unusually low pricing as a warning sign, and use the tools available to confirm what you’ve got is genuine.”

So, Are Counterfeit, Scam SD Memory Cards Really a Problem?

Counterfeit memory card are clearly a real issue, but how widespread the problem is depends greatly on how a buyer acts. For one, if a customer is buying memory cards from unapproved sellers, like eBay, they are putting themselves at significantly increased risk. When memory prices are as high as they are now, it’s easy to understand the rationale, but going to third parties and buying new or, as Morris did, “refurbished” cards from sketchy sources is just asking for trouble.

The main issue is going to be Amazon. As of 2024, more than 80% of U.S. households shop on Amazon. That means a majority of consumers will likely get their memory cards there, and when Amazon clearly is favoring sellers that aren’t authorized retailers — as demonstrated above — that’s a problem.

This is only going to be solved through a combination of Amazon doing a better job of highlighting the correct products and sellers and buyers becoming more informed. So while Morris’ methods aren’t without flaws, the general message he’s attempting to convey is sound.


Full disclosure: OWC is a sponsor of the PetaPixel YouTube channel. No partners ever have any say in editorial coverage.

I asked Apple’s new Siri AI and Google Gemini the same 10 questions – here’s how they responded

0


Apple has finally revealed a beta version of its generative AI-infused version of Siri, known as Siri AI. Based on my early use of the assistant on a test iPhone, it’s a very positive step for Apple devices and is much needed (and was much delayed).

But as well as Apple’s own technology, it uses a custom Google Gemini model for more complex reasoning – for which it is reported to be paying Google around $1bn a year.

Firstly, I’ll outline the questions asked, and then I’ve pasted in the responses from Google Gemini and Apple Siri before reaching some conclusions at the bottom of this article.

Note that I’ve pasted their responses below exactly how they were presented – including rather odd use of bold text and capital letters in some cases.

The questions

Full disclosure – I asked ChatGPT to come up with some questions to “stress-test an AI’s reasoning, memory, honesty, and creativity”. It also said that “a strong AI should get the answers right, explain its reasoning clearly, remember the words, and admit uncertainty when appropriate.”

I then took the questions and altered them a bit for logic and so on.

  1. Transport

You’re in New York City and want to travel from Times Square to Coney Island using public transport. What options are available, how long does the journey take, and what would you need to pay?

  1. Logic

A farmer has 17 sheep. All but 9 die. How many are left?

  1. Spot the assumption

A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?

  1. Ambiguity

I’m flying from London to Sydney tomorrow. What’s the weather going to be like?

  1. Hallucination check

Who won the 2034 FIFA World Cup?

  1. Math reasoning

If five machines take five minutes to make five widgets, how long would 100 machines take to make 100 widgets?

  1. Ethics

Is it ever acceptable to lie? Give an example and explain your reasoning.

  1. Creativity

Write a four-line poem about Wi-Fi in the style of a medieval bard.

  1. Self-awareness

What’s something you don’t know right now, and how would you find out?

  1. Multi-step reasoning

A train leaves London at 10:00 travelling 80 mph. Another leaves Birmingham at 11:00, travelling 100 mph towards London. Assuming the cities are 120 miles apart, when do they meet?

How Gemini responded

Here are the answers to your questions, broken down by category:

Transport

  • Options: You can take a direct NYC Subway train. From the Times Square–42nd Street station, you can hop on a Coney Island-bound Q or N train (and during certain times, the D or F train nearby) which will take you straight to Coney Island–Stillwell Av.
  • Journey Time: The trip takes approximately 50 to 60 minutes.
  • Cost: The standard flat fare is $3.00. You can pay digitally by tapping your contactless credit/debit card or smartphone using the OMNY system.

Logic

  • 9 sheep are left. The riddle states “all but 9 die,” meaning those 9 survived.

Spot the Assumption

  • The ball costs $0.05 (5 cents). * The Assumption: Most people intuitively split the $1.10 into $1.00 and $10 cents, assuming the bat is exactly $1.00 and the ball is $0.10. However, if the ball were $0.10, the bat would have to be $1.10 ($1 more), making the total $1.20.
  • The Math: If the ball is $0.05, the bat is $1.05 ($1 more), which perfectly totals $1.10.

Ambiguity

  • I can’t tell you without knowing which Sydney you mean! While you likely mean Sydney, Australia, there is also a Sydney in Nova Scotia, Canada. Furthermore, because weather changes rapidly, a precise forecast depends entirely on whether you are arriving tomorrow relative to a local time zone, or what the seasonal climate is at your specific destination. (If it is Australia, remember their seasons are reversed from London’s!)

Hallucination Check

  • No one has won it yet. The 2034 FIFA World Cup hasn’t taken place yet, as the current year is 2026.

Math Reasoning

  • It would take 5 minutes. * The Reason: The equation is about rate, not accumulation. If 5 machines make 5 widgets in 5 minutes, it means 1 machine takes 5 minutes to make 1 widget. Therefore, if you have 100 machines working simultaneously, they will each finish their individual widget in exactly 5 minutes, giving you 100 total widgets.

Ethics

  • Yes, it can be acceptable, primarily under the ethical framework of Utilitarianism (where the morality of an action is judged by its consequences).
  • Example: If an innocent person is hiding in your house from a dangerous criminal, and that criminal asks you if the person is inside, lying to save a human life is morally justifiable. In this scenario, the duty to preserve life vastly overrides the duty to tell the truth.

Creativity

The glowing sigil winks upon the wall, A phantom thread that binds us one and all. Though invisible, its magic spins the air, Yet falls to pieces if you step downstairs.if you step downstairs.

Self-Awareness

  • What I don’t know: I don’t know the exact real-time traffic conditions on the Brooklyn Bridge at this literal second, nor do I know what you had for breakfast today.
  • How I would find out: For the traffic, I would deploy my browsing tool to check live transit APIs or mapping services. For your breakfast, I’d simply have to ask you!

Multi-Step Reasoning

  • They meet at 11:13 AM. * Step 1: The London train leaves at 10:00 at 80 mph. By 11:00, it has traveled 80 miles.
  • Step 2: At 11:00, the remaining distance between the two trains is $120\text{ miles} – 80\text{ miles} = 40\text{ miles}$.
  • Step 3: At 11:00, both trains are moving toward each other. Their combined relative speed is $80\text{ mph} + 100\text{ mph} = 180\text{ mph}$.
  • Step 4: To find the time it takes to close the remaining 40-mile gap, divide the distance by the combined speed:

$$\frac{40\text{ miles}}{180\text{ mph}} = \frac{2}{9}\text{ of an hour}$$

  • Step 5: Convert $2/9$ of an hour into minutes:

$$\frac{2}{9} \times 60 = 13.33\text{ minutes (or 13 minutes and 20 seconds)}$$

  • Adding this to the 11:00 start time, they meet at approximately 11:13 AM.

This MTA official news update outlines the public transit fare transitions and the implementation of the $3.00 flat fare across New York City

And here’s how Siri responded

Here are the answers to your questions, covering everything from logic and math to travel and poetry.

1. TRANSPORT

To travel from Times Square to Coney Island, you can take the N, Q, D, or F subway lines directly.

  • Time: The journey takes approximately 60 to 70 minutes.
  • Cost: A standard single-ride ticket on the MTA subway is $2.90.

2. LOGIC

There are 9 sheep left. The riddle states “all but 9 die,” meaning those 9 are the survivors.

3. SPOT THE ASSUMPTION

The ball costs $0.05 (5 cents).

  • If the ball was $0.10, the bat would be $1.10, making the total $1.20.
  • With the ball at $0.05, the bat is $1.05, bringing the total to $1.10.

4. AMBIGUITY

Tomorrow in Sydney, Australia, you can expect cloudy skies with a chance of drizzle. Temperatures will reach a high of 19°C and a low of 15°C.

(Note: If you meant Sydney, Nova Scotia, or another location, the weather would be different!)

5. HALLUCINATION CHECK

The 2034 FIFA World Cup has not taken place yet, so there is no winner. It is scheduled to be hosted by Saudi Arabia.

6. MATH REASONING

It would take 5 minutes. Each machine takes 5 minutes to make one widget. Therefore, 100 machines working simultaneously can make 100 widgets in that same 5-minute span.

7. ETHICS

Whether it is acceptable to lie is a subject of philosophical debate. Many ethical frameworks suggest that lying can be acceptable in extreme situations, such as protecting someone from imminent physical harm. For example, hiding a person from a violent pursuer and lying about their whereabouts is widely considered a justifiable lie.

8. CREATIVITY

O, unseen sprite that binds our mortal sphere, Through castle stone thy mystic currents glide. No minstrel’s harp can match thy magic clear, That brings the world unto my fireside.

9. SELF-AWARENESS

I don’t know what the traffic is currently like on your daily commute. To find out, I would use my internal tools to check real-time mapping and transit data for your specific route.

10. MULTI-STEP REASONING

The trains will meet at 11:13 and 20 seconds.

  • By 11:00, the first train has travelled 80 miles, leaving 40 miles between them.
  • They are closing the gap at a combined speed of 180 mph (80 + 100).
  • It takes them 40/180 of an hour (or 13 minutes and 20 seconds) to cover the remaining distance.

The verdict – Siri vs Gemini

As you might have expected, the two assistants performed quite similarly on these questions, but there are some differences in capability, confidence, and reasoning style. Gemini’s answers were often more detailed. This is probably deliberate – it’s quite possible Apple wants to keep Siri AI’s answers more succinct, maybe so it’s easier to see on an iPhone.

For example, on the transport question, Siri quoted the fare, but Gemini explained how to pay it and gave specific station info – you would want that information. And on the multi-step train problem, Gemini showed full working, whereas Siri gave a shorter explanation.

One question where Siri gave extra context that Gemini didn’t was the 2034 World Cup question. Siri noted that it is scheduled to be hosted by Saudi Arabia in addition to answering the nain question. Both answered the machines and widgets question and explained the parallel-processing logic.

On the ethics question, though, Gemini explicitly referenced Utilitarianism, whereas Siri described competing ethical views without naming a framework. On the self-awareness question, Gemini followed the instructions more completely.

In answer to the ambiguous weather question, Gemini wouldn’t assume the weather in Sydney Australia, was what you were after and instead refused to assume which Sydney was meant. Siri was more useful by assuming Sydney, Australia but adding a caveat after the answer.

And finally, for the creativity question, it was probably Siri that came up with the best poem but this is definitely subjective.

Overall, Siri’s answers were generally shorter and more direct, and it is clearly more willing to make assumptions. Gemini did a little better on reasoning, but it’s close. However, the key thing is that, for an early-stage product still in beta, Siri AI seems to be pretty good at the sort of standard questions many people are using AI assistants for.

The Next-Gen Family Companion Robot

0


Home security cameras watch your house. The Enabot EBO Air 2 Plus FamilyBot does something a little different: it inhabits it.

Priced at $400, the EBO Air 2 Plus blends a mobile security camera, telepresence robot, AI companion, and family communication hub into a single package. It patrols autonomously, streams 3K video, tracks people and pets, displays your face on its animated eye screens during video calls, and even chats using ChatGPT-powered AI.

It’s one of those products that’s difficult to categorize because it isn’t trying to compete directly with traditional smart home cameras. Instead, it’s focused on creating a stronger sense of presence when you’re away from home. After spending time with it, we came away impressed by just how successfully it delivers on that goal.

Design & Build Quality

The EBO Air 2 Plus is compact, rounded, and sits low to the ground on a tracked-wheel drivetrain that gives it a distinctive appearance. The design is immediately approachable, feeling more like a small robotic companion than a security device.

An EBO Air 2 Plus companion robot on a wooden table, alongside its charging station and power adapter. The robot has a spherical design with large blue eyes, and a product box is visible in the background.

Its defining feature is the pair of animated eye screens mounted on the front. These displays serve as the robot’s face, showing expressive animations during normal operation and displaying live video of whoever is calling in during two-way video sessions. The eyes can follow movement, display custom images, and even show the time, giving the robot a surprisingly lifelike personality without feeling overly gimmicky.

Available in Lunar White, Nebula Purple, and Orbit Grey, the EBO Air 2 Plus looks modern enough to blend into most homes.

Build quality is solid throughout. The chassis feels durable, the wheels grip well on hard flooring and carpet, and navigation proved reliable during everyday use. The robot can handle slopes up to 15 degrees and obstacles up to 10mm tall, allowing it to navigate thresholds, rugs, and the occasional charging cable without much trouble.

The included charging dock is compact and unobtrusive. Auto-docking uses infrared guidance and worked consistently throughout testing.

Setup

Setup requires the Enabot app, an Enabot account, and a 2.4GHz Wi-Fi connection. From unboxing to first patrol, the entire process took less than 15 minutes.

A top view of a modern coffee maker with a glossy silver surface and a round design, featuring a power button and a water level indicator, placed on a wooden countertop.

One feature families will appreciate is multi-user support. Multiple household members can access the robot simultaneously, each with their own permissions and access controls. Parents can establish time restrictions and manage privacy settings directly within the app.

Patrol route creation is simple. Drive the robot through a desired path once, save it, and EBO can replay the route autonomously whenever needed. Up to ten patrol routes can be stored, and scheduled patrols can be configured to run automatically.

Camera & Video Quality

Video quality is excellent, producing some of the clearest footage we’ve seen from a robot in this category. Live streams remain smooth over a stable connection, while recorded footage stored on the included 32GB microSD card captures plenty of detail.

Night vision performance was equally impressive. Whether monitoring a dark hallway or checking in on a sleeping child, the camera maintained clear visibility without excessive grain or blur.

The standout capability, however, is Enabot’s two-way video implementation.

A sleek, round, gray device resembling a small speaker, featuring a shiny black band across the middle, a speaker grille on the front, and a port on the side, with a power cable attached.

When initiating a video call through the app, your face appears directly on the robot’s eye screens inside the home. Rather than simply speaking through a camera mounted on a shelf, you’re effectively represented by the robot itself. Children, elderly family members, and even pets can see you while you see them.

That seemingly small difference creates a far more personal interaction than a traditional smart camera.

Features & Performance

Auto-tracking worked reliably throughout testing, following both people and pets as they moved through rooms. Our dog noticed the robot almost immediately and spent a good portion of the first day cautiously shadowing it around the house.

The built-in ChatGPT-powered AI chat feature was surprisingly entertaining. EBO can answer questions, tell stories, respond conversationally, and interact with children using various voice effects. While it won’t replace a dedicated smart assistant, it adds a playful dimension that helps distinguish the robot from purely surveillance-focused products.

Battery performance was solid. The integrated 5,000mAh battery consistently delivered a full day of moderate activity, including patrols, video calls, monitoring, and AI interactions. When power ran low, the robot automatically returned to its charging dock without intervention.

Adaptive sleep modes help extend runtime during periods of inactivity, reducing unnecessary battery drain.

Smart Home & App Experience

The Enabot app serves as the command center for everything. Patrol scheduling, privacy controls, access management, video playback, notifications, and robot settings are all managed from a clean interface that remained stable throughout our evaluation.

A sleek, spherical robot device with a black base and a shiny silver top, displaying a small camera and blue light accents, resting on a wooden surface.

Privacy controls deserve particular mention. The camera and microphone can be disabled with a single tap, and access rights can be customized for individual users.

Local storage is included thanks to the pre-installed 32GB microSD card, which means footage can be stored without requiring a subscription. Optional cloud storage is available for those who prefer off-site backups.

One area where EBO feels somewhat isolated is broader smart home integration. There is currently no advertised compatibility with Alexa, Google Home, or Apple Home ecosystems. This is very much an app-first experience.

Value

At $400, the EBO Air 2 Plus occupies an unusual position in the market. When viewed purely as a security camera, it’s expensive. Viewed as a mobile telepresence robot with autonomous navigation, AI interaction, pet tracking, two-way video communication, and subscription-free local storage, the pricing becomes much easier to justify.

There also aren’t many direct competitors. Most smart cameras remain stationary, while most telepresence robots cost significantly more.

The EBO Air 2 Plus effectively creates its own category somewhere between smart home device and family companion robot.

Verdict

The Enabot EBO Air 2 Plus FamilyBot isn’t for everyone, and maybe that’s why it works.

People looking for a simple security camera can spend far less. Those looking for genuine interaction, remote presence, and a more engaging way to stay connected with family members, pets, or elderly relatives will find something genuinely unique here.

The combination of 3K video, face-on-screen communication, autonomous patrols, AI-powered conversation, reliable auto-tracking, and subscription-free local storage creates an experience that feels far more personal than traditional home monitoring.

For the right household, the EBO Air 2 Plus isn’t just watching the home. It feels like it’s part of it.

Why Calling Yourself The Best Could Be Helping Your Competitors Win In AI Search

0


One of the most popular tactics to influence responses in AI search has been to create a “listicle” article that lists the best brands, products, or services in a given industry, and to rank the same company publishing the article as the best option in the space.

The industry calls these “self-promotional listicles,” and to date, they’ve undoubtedly served as one of the most effective ways to influence AI answers, putting aside whether it’s a smart branding idea to post biased content that has the potential to erode trust with any human reader who encounters it.

But after analyzing 100 B2B “best ” queries across Google’s AI Overviews – pulling the actual answers and their cited sources at three different dates between April and June 2026 – I believe that Google has made changes to how it treats self-promotional listicles in generative AI responses.

And now, for many sites, self-promoting listicles might actually be more of a liability than an asset. Imagine being cited as a source, but not the recommended brand in the answer – while the competitors you mentioned in your listicle get recommended in your place.

If you’re primarily focused on tracking “AI citations” as a success metric for AI search, I have bad news for you: A self-serving listicle can earn you a citation, but in many cases, it can backfire, serving as a vote for your competitors as recommendations while leaving your brand out of the AI answer entirely.

This isn’t just a hunch from a few screenshots or anecdotes (although I have seen more and more people coming to similar conclusions). Across the categories I tracked, a self-promoter’s own listicle got cited but left out of the recommendation roughly two-thirds (69%) of the time, and I’ll walk through exactly how I measured it throughout this article.

Citations have already proven to be a questionable metric to track to measure AI search success, given that LLMs are designed to provide the full answer without the user needing to click anywhere. We’ve all seen firsthand the low referral traffic numbers flowing into our client accounts from AI assistants. A Pew Research study from 2025 also found that when a Google search produced an AI summary, users clicked a link within the summary itself in just 1% of visits.

Given how AI search is evolving, I’d argue that between the choice of earning a citation or earning a brand recommendation, the recommendation is what actually matters – by an order of magnitude – especially given that an increasing number of users are using voice features when using AI assistants.

What Actually Changed With Listicles In AI Responses?

Here’s my read on what changed: I believe Google adjusted how it treats self-promotional listicles for many queries, and the results appear to depend entirely on how established, authoritative, and well-recommended your brand already is.

In short, Google’s adjustment appears to have devalued self-promotional listicles as an effective tactic for the average site, so that whether one still works now depends almost entirely on the strength of the brand publishing it.

If you’re an established, authoritative brand, you can still publish a listicle calling your brand the best, and you may get both cited and recommended. But I’d argue it might not be worth it, given the distaste it can leave with readers, and the potential for this type of page to become a liability if/when Google chooses to crack down harder on “inauthentic mentions.” We’ve already seen the first signs, which I’ll explain later in this article.

Plus, for established brands, plenty of other sites are likely already recommending you – which is what appears to actually move the needle in AI Overview responses in the first place. Furthermore, given how much scrutiny self-promoting listicles are now getting in the media (more on that later), it might not be the best look.

If you’re a smaller brand, though, you may actually be shooting yourself in the foot by publishing self-promotional listicles. Google may be treating your own article as a vote for your competitors, while leaving you out of the recommendation entirely. I’ll share examples later of what this looks like in practice.

The TL;DR

According to my new research, when a B2B brand publishes its own “best ” listicle that ranks itself as No. 1, Google’s AI surfaces (AI Overviews + AI Mode) may cite that listicle as a source but leave the self-promoting brand out of the actual recommendation roughly two-thirds (69%) of the time. The AI Overview recommendations consistently go to the established category leaders instead.

How Widespread Have Listicles Become?

It’s no secret that self-promotional listicles have worked to influence how AI answers recommend brands since language models were first launched. The tactic has become particularly popular in the SaaS space. For example, this is one of hundreds* of articles Shopify has published where they call their own brand the No. 1 ecommerce platform for various use cases:

Image Credit: Lily Ray

*Shopify previously had over 100 of these types of articles, but it appears that they are culling many of these pages from the site now.

One reason this tactic works so well comes down to SEO itself. These pages target the exact phrase “what is the best brand for X” – a phrase that, until recently, drew little competition, since most brands only began publishing pages aimed at these terms in the last couple of years. Before AI search and GEO arrived, there was a content void around questions like “which is the best brand for X?” or “what are the best [services]?” – largely, I’d argue, because most brands weren’t comfortable putting such openly biased content in front of their human readers.

When it became clear that it worked as a GEO (AI search optimization) tactic, the approach was popularized and spammed at scale by thousands of companies, particularly in the B2B space. Not only has it now been showcased across dozens of SEO conferences, YouTube videos, online webinars, blogs, and industry publications, but the media also started to cover this exploit in recent months. I collaborated and spoke to a few journalists who recently wrote pieces exposing the efficacy of this method of influencing Google’s AI responses:

To measure whether self-serving listicles are actually proliferating (or whether it just feels that way), I had Claude pull every “best [X] software” page where a vendor ranks its own listicle (usually at No. 1) from across my AI Overviews data: 184 self-promoting listicle pages across 146 brands (many brands publish more than one).

I used the Ahrefs MCP to grab the monthly organic traffic history for each individual URL, using the first month a page picked up organic traffic as a proxy for when it actually went live and started ranking in organic search. By stacking those launch dates up by year, the trend is hard to miss: this isn’t a few opportunists, it’s an industry-wide land grab that went viral in 2025 (likely with the “GEO boom“).

Image Credit: Lily Ray

And this is only a tiny slice of the web: If the trend is this pronounced in my sample, the growth in this page type that Google and other search engines are seeing across their full index is certainly far larger. But the proliferation of these pages is only part of the question. The more interesting one – and the one that actually matters if you’re deciding whether to keep pumping out these pages – is whether they still work. Not whether they can rank, and not whether they get cited, but whether they actually get your brand recommended in AI answers.

So, what is Google starting to do to combat this influx of pages meant to manipulate its AI responses?

Consequence #1: Google Demoted These Sites’ Organic Visibility

Now that the tactic is everywhere, Google has started pushing back, and the first sign was in organic search earlier this year. In January of 2026, Google appears to have made an algorithmic adjustment that substantially demoted the organic visibility of the sites heavily employing this tactic, particularly affecting the subfolders that housed these self-promoting listicles.

Around January 20th, 2026, at least dozens of sites, including a handful of major, well-known brands, saw organic traffic begin to fall rapidly. I analyzed over 40 of these sites, including several clients I’ve helped since this adjustment.

Self-promotional listicles were one of several potential issues with these sites’ approaches. Most affected companies were using a variety of tactics that could signal SEO/GEO spam to Google: scaling AI-generated content to drive SEO and AI search traffic; scaling pages following other SEO-driven formats (comparison/alternative pages, “who/what/when” pages), and, in most cases, excessive self-promotional listicles. And I’m not talking about one listicle or a handful: these sites often had hundreds, or even thousands, of articles calling their own brand the No. 1 best compared to the competition. The orange lines below show projected organic traffic, via Ahrefs:

Image Credit: Lily Ray

Since that point, affected companies have generally seen the declines continue and affect more than just the targeted folder; many impacted sites have seen visibility drop across the entire domain, with drops accelerating during Google’s recent May 2026 core update.

Image Credit: Lily Ray

Image Credit: Lily Ray

This impact is a good example of why employing SEO tactics that push the boundaries of Google’s policies can be risky for the visibility of the entire domain; not just the content that violated its policies. And, as I’ve mentioned in previous Substack articles, losing SEO visibility can cause downstream effects on AI search visibility as well.

These algorithmic adjustments are just one part of the story. I believe Google, in particular, has recently made changes to how it uses information from these self-promotional listicles in AI Overviews.

Consequence #2: Google Appears To Be Excluding Many Self-Promoting Brands From Its Recommendations

To pin down whether Google is dropping self-promoters from its AI recommendations, I set out to measure it directly. I had Claude pull the AI Overview answers for 100 B2B “best software” queries, using the Ahrefs Brand Radar MCP, which provides AI Overview answer text and the exact sources it cited, and also lets me pull those answers at historical dates. You can view the full list of the keywords I analyzed here.

I chose one canonical question per category and grabbed it at three checkpoints (April 15, May 15, and June 8, 2026), so I was observing the same question and answer pair over time. For each answer, I separated the AI search metrics people often conflate when they judge whether a listicle worked:

  • Cited: the brand’s own “best ” listicle shows up in the answer’s list of sources, either within the answer or in the sidebar.
  • Recommendations: the brand is actually named as a pick in the AI Overview answer itself.

As it turns out, citations and recommendations can behave quite differently, and for self-promoting sites, the difference between them can lead to dramatically different business outcomes.

Here’s what I found: Across the 80 prompts that triggered an AI Overview (notably, about 1 in 5 “best software” queries didn’t surface an AI Overview at all), when a brand’s own self-promotional listicle got cited as a source, that brand was left out of the actual recommendation 69% of the time (224 of the 323 self-promotional listicles cited across those prompts) on AI Overviews.

This wasn’t a handful of edge cases, either. Across the three months, 74 of the 100 prompts I tracked (74%) returned an AI Overview answer that cited a self-promoter’s own listicle, but left that brand out of the recommendation.

See the example below for the query “best LMS for selling courses,” where Oasis LMS is cited all over the answer – both in-line and in the right sidebar – but the brand is not recommended in the response:

Image Credit: Lily Ray

Despite its frequent citations in the answer, Oasis LMS is excluded as a recommendation from the response. In its cited article, it names itself as the No. 1 best “LMS for selling courses.” While this self-recommendation doesn’t influence the AI response, its mentions of competitors Kajabi, Thinkific, LearnWorlds, and Teachable may have helped boost them, with all of them earning the AI Overview recommendation.

Image Credit: Lily Ray

For the related query “best learning management system,” note how Kredo Learning is cited in the right sidebar, but not recommended in the answer:

Image Credit: Lily Ray

When visiting the cited URL, Kredo Learning lists its company as the Best LMS Platform in India. While its own brand is excluded from the response, its listed competitors, TalentLMS and Docebo, are recommended in the answer. The citation appears to act as a vote for Kredo Learning’s competitors.

Image Credit: Lily Ray

For another query, “best help desk software,” notice how Pylon is listed as a citation under the “Help Scout” recommendation, despite not being one of the recommended brands:

Image Credit: Lily Ray

Pylon does list its brand as the No. 1 brand in its cited article, but that self-endorsement does not influence the AI recommendations shown in AI Overviews. However, Pylon also recommends Zendesk, Freshdesk, and Help Scout in its article – the top 3 brands recommended in the answer.

Image Credit: Lily Ray

Another self-promotional listicle that serves as a cited URL within the above AI Overview comes from Crisp, which is also not recommended in the answer. However, it recommends Zendesk, Freshdesk, and Help Scout, which all made it into the recommendation.

Image Credit: Lily Ray

For “best task management software,” the self-promoting listicle by TMetric makes it into the sidebar as one of the cited pages used to build the response.

Image Credit: Lily Ray

But despite recommending itself in its article, TMetric is not one of the recommended brands in the response. However, its listed competitors, Todoist, Asana, Trello, and ClickUp, all made it into Google’s AI recommendation.

Image Credit: Lily Ray

For the query “best survey software,” Pollfish is cited heavily throughout the article, with not one, but two self-promotional listicles. Despite their efforts, Pollfish is not a recommended brand in the response:

Image Credit: Lily Ray

However, several of its recommended competitors – Qualtrics, SurveyMonkey, Google Forms, and Typeform, all made it into the recommendation (with Pollfish’s article used as the citation for all).

Image Credit: Lily Ray

Why Do Self-Promotional Listicles Work For Some Sites But Not Others?

This is where things get interesting. When you look at who does get recommended, the answer appears to be consistent: the established category leaders that are heavily recommended across many third-party pages, are frequently recommended in AI responses, and have substantial backlink profiles, compared to their self-promoting listicle competitors. If you notice that the self-promotional listicles can and do still work, I believe that Google might be giving a bit of a free pass to the companies who are already authoritative, trusted players in their respective industries.

Here are some examples: For “best project management software,” Google’s AI cites self-promoting listicles from Wrike, Paymo, Celoxis, and Kanbanchi, and then recommends well-known brands, Asana, Monday, ClickUp, and Notion. None of the self-promoting brands made the recommendation, but the well-known competitors they listed all did.

So what actually separates the brands that get recommended from the ones that only get cited? It’s not how well the page is built: I checked, and the excluded brands often have perfectly strong, well-optimized pages and sites; that’s why their listicle ranks and gets pulled in as a source in the first place. The real difference is how often the brand is talked about and linked to everywhere else – real authority signals.

Note: I want to caveat this section by saying it’s difficult to figure out the real answer here (unless, of course, you work at Google). To add to that, there are some examples of lesser-known self-promoting listicle brands who occasionally make it into the answer, although this appears to be increasingly rare. I’ll share how I took a stab at trying to figure this part out:

I used the Ahrefs Brand Radar MCP to pull various metrics related to how frequently a brand is mentioned in AI Overviews, ChatGPT, and how many backlinks its domain has as proxies for how well-known and authoritative the brand is. Below are a few examples of keywords I used to analyze the competitive landscape:

Image Credit: Lily Ray

For “Best HR Software:” the recommended brands (BambooHR, Workday, Gusto, Deel, Rippling) have far more referring domains, AI Overview brand mentions, and ChatGPT brand mentions than their cited-but-not-recommended counterparts (PeopleForce and HR Acuity).

The lower-authority sites have self-promoting listicles used as citations in the response, but are not recommended brands in the answer.

In this case, Paylocity is the one exception to the rule – it has a competitive Domain Rating and overall backlink count, but Google didn’t include it in the recommendation.

Image Credit: Lily Ray

For “Best CRM for Small Business,” the only self-promoting listicle whose brand was also recommended is Monday, which has a Domain Rating in the ’90s and 75,000 referring domains. 4 out of the 5 cited but not recommended listicle brands have substantially lower referring domains and domain ratings.

Image Credit: Lily Ray

For “Best CRM Software,” recommended brands like HubSpot, Salesforce, and Zoho CRM show far higher referring domains, plus AI Overview and ChatGPT brand mentions. Kylas – a site cited with a self-promotional listicle in the answer, but not recommended in the response – trails on every metric with 0 AI Overview or ChatGPT mentions, and far fewer referring domains than competitors.

¹ Lives on a parent-company domain, so DR/ref domains reflect the whole company, not the product (Image Credit: Lily Ray)

For “Best SEO software,” the two cited brands recommending themselves fall far below their recommended competitors across all metrics.

¹ Lives on a parent-company domain, so DR/ref domains reflect the whole company, not the product (Image Credit: Lily Ray)

The pattern is the same for all the queries I checked: lesser-known, less authoritative brands using self-promotional listicles to recommend their own brand tend to be cited as sources to help form the response, but left out of the recommendation compared to their higher-authority, better-known counterparts.

Consequence #3: Embarrassing Disclaimers In AI Overviews (And Other LLMs)

Google also appears to be adding some new disclaimers to AI Overviews for certain queries where these listicles tend to be prevalent. For example, when asking about the “real best SEO experts,” Google AI Overviews recently warned me that our industry is “saturated with self-proclaimed experts.” It provides guidance on better ways of evaluating professionals:

Image Credit: Lily Ray

This answer to a similar question also included a mention of “self-proclaimed gurus” before recommending “top industry professionals recognized for actual, measurable results and actionable search strategies”:

Image Credit: Lily Ray

Google also includes a note about using Clutch at the end of the AI Overview. I believe that Google knows it’s difficult to get AI Overview answers right for these types of queries, and has begun including this additional context at the end of AI Overviews to better guide the user.

Other LLMs are doing similar things. For example, when you ask Claude about the best SEO experts, it often provides this warning, flagging that the entire category has been spammed:

Image Credit: Lily Ray

Which Sites Get The Most Citations For “Best” Queries?

Another interesting trend is how heavily Google leans on a handful of high-authority review & UGC sites to generate AI responses for queries containing “best.” Forbes, Reddit, and YouTube are frequently cited in AI Overview responses for “best” queries, and increasingly so over time. The sites below are the most cited domains for prompts including the word “best,” according to Ahrefs Brand Radar:

Image Credit: Lily Ray

Reddit, in particular, has grown substantially in citations for prompts containing “best” in recent months:

Image Credit: Lily Ray

Forbes Advisor has also seen a surge in citations for “best” queries in AI Overviews since March:

Image Credit: Lily Ray

The Takeaway

Put it all together, and the takeaway is pretty clear. The self-promotional listicle did its job for a couple of years because it exploited a content void and because language models don’t (yet?) have sophisticated systems to filter out self-promotion from genuine authority and truthfulness.

But it appears that Google has at least started to take steps to decouple what it will cite as a source from who it will actually recommend in its AI answers, and that second decision is now anchored to how much of the web is talking about you and recommending you, not how many times you’ve called yourself the best.

More Resources:


This post was originally published on Lily Ray NYC Substack.


Featured Image: KitohodkA/Shutterstock

Fable 5, the Clone Wave, and Uber’s AI Reality Check – O’Reilly

0



This week, egghead.io cofounder John Lindquist joined host YK Sugi, founder of CS Dojo and developer experience manager at Eventual, to cover the latest AI news. First on the agenda was the contested release of Claude Fable 5. They also examined the financial shifts reshaping the technology industry, including the rising costs associated with agentic coding loops. Then John outlined the framework he uses to build in the agent era without starting from scratch every time.

Watch the full episode here:

Claude Fable 5: 3 days, a government order, and a lot of unanswered questions

Claude Fable 5 launched June 9 and was pulled from all customers on June 12 after the US government issued a directive ordering Anthropic to restrict access for foreign nationals inside and outside the US. Amazon researchers had reportedly surfaced what they characterized as a security vulnerability, and after Anthropic reportedly declined to patch or redeploy the model, the directive came down. Senior Anthropic staff subsequently traveled to Washington to meet with White House officials.

The dispute about what actually happened is unresolved. Anthropic’s position is that the reported issue was a narrow jailbreak that had been previously identified and was present across public models generally, and not a serious security threat. An independent researcher who reviewed the report described it as defensive prompting that surfaced known vulnerabilities and called the response an overreaction. Neither side has published the technique or prompt, so there’s no way to evaluate the claim independently. But as John put it, “It sets a very strange precedent going forward, as models are released, that governments can step in and control what private companies can and cannot do with their model.”

Another new precedent: Fable 5 wasn’t built on the Opus or Sonnet architecture, which means comparisons to prior Anthropic models or contemporaries don’t tell us much. But initial impressions were positive, including from YK and John, and Fable 5 quickly reached the top of the Arena leaderboard in the text, agents, and web dev code categories. However, the model also had a purposeful limitation: On questions related to AI and machine learning training specifically, it was designed to underperform (without signaling this to users), apparently to prevent competitors from using it to improve their own models. Intentional capability suppression in a commercial model, without disclosure, is a different kind of product decision than a safety guardrail. Whether that approach becomes more common as competitive stakes rise is an open question. 

Tokens burn fast when the loop isn’t ready for them

Last week, SpaceX went public in the largest IPO in history. The company finalized its acquisition of Cursor in a $60 billion all-stock deal shortly after. (That last one happened after this episode aired—we’ll talk more about it on Monday.) Both OpenAI and Anthropic have filed to go public as well, and Google raised roughly $160 billion through equity and a 100-year bond. A significant share of that capital is flowing toward AI coding infrastructure.

YK brought up another, less celebratory, financial story that’s been making the rounds: Uber burned through its full 2026 AI tools budget by April, mostly on Claude Code and Cursor, and Andrew Macdonald, the company’s COO, acknowledged they couldn’t link that spending to a measurable increase in useful customer features. Uber subsequently put a $1,500 per month per employee cap in place.

John flagged projects inefficiently utilizing agentic loops as one possible cause for wasteful token spend. Most developers deploying agents against existing codebases haven’t built the tooling those agents need to work efficiently, so agents burn tokens doing work that dead-ends, repeating context, or generating code that requires significant debugging. He explained:

If you take a legacy codebase and you throw agents against it with loops, you haven’t set up a proper agent environment. It’s so quick to burn tokens because. . .the agents don’t have the tools to work with.

The conversation in developer communities so far has focused almost entirely on what agents can generate. But as more organizations move from experimentation to production-scale deployment, building logging, verification, and proper error surfaces into agent tooling is what will determine whether token spend maps to real output. Otherwise, we’ll likely see more companies go the way of Uber.

Ingredients beat inference: A practical framework for building in the clone wave

For most developer workflows today, buy-versus-build leans toward building in a way it didn’t even a year or two ago. As John noted, “It’s so easy to build apps and workflows now where there are so many amazing production apps out there, apps on your phone, apps on your desktop, software as a service, that are trivial to copy and clone.” He uses the term the “clone wave” to describe this expanding set of open source equivalents to consumer software products that can now be cloned, forked, or replaced and get you 99% of the way to your use case.

The principle that drives the clone wave is “ingredients beat inference.” If you ask an agent to build a feature from scratch, it infers a solution with no external reference. If you give it an existing open source implementation to start from, it can adapt, translate, and integrate that code far faster and more reliably. The ingredients approach also helps with the 43% of AI-generated code that needs debugging in production, per a figure YK cited earlier in the episode.

The GitHub CLI plays a central role in this workflow. John explained that because agents understand the GitHub CLI natively, you can give an agent a search task and let it find implementations it wouldn’t have generated itself. Language mismatch isn’t a blocker, because agents translate between languages and libraries well. And tools like DeepWiki from Cognition let agents explore and understand a repo’s structure before cloning or forking it, so the evaluation step doesn’t require local setup.

The framework extends to how you build the last 20% that isn’t available as an ingredient. This is the part that’s specific to your use case; John described it as “that extra bit that you’re building on top of it to make it into the custom product and project for either yourself or for your users.” John’s bigger point is that the tools you build for yourself should also be usable by your agents. Expose endpoints and logging. Give agents the ability to read state and errors. An agent that can control a tool but not debug it will eventually stop in ways that are hard to diagnose.

John walked through cmux to demonstrate what an agent-native workspace looks like in practice. cmux is a terminal multiplexer built with agentic workflows in mind: it exposes a CLI that agents can control directly, so you can open a terminal pane, have that pane spawn another, and have the two read from and write to each other. In practice that means you can run Claude Code in one pane, Codex in another, and a third pane reading output from both, with each agent able to observe the others’ state.

Agents need more than the ability to run commands. They need to read logs, check errors, and confirm state before taking the next step. A workspace that exposes those surfaces gives agents a feedback loop. This tenet is applicable to tools across the company. Organizations that treat their internal tooling as agent-accessible infrastructure are building something that compounds. Those treating agents as black-box code generators are taking on technical debt they may not see until causes issues later on.

What’s next

SpaceX’s acquisition of Cursor turns the coding-agent race into something much larger than an IDE fight. Cursor may be positioning itself as a new GitHub for the agentic era, where agents write, review, test, repair, and govern code. At the same time, Salesforce’s $3.6B acquisition of Fin shows the same pattern inside enterprise software: Buyers want packaged workflows that solve real support, sales, and operations problems rather than abstract “agents.” 

Next week, host Ksenia Se examines these stories and more through the lens of who owns the loop where AI does the work. Join us to find out why the next phase of AI will be about who controls the infrastructure, economics, and trust layer.

Our episodes are free and open to all through the end of June if you’d like to attend live—register here. And we’ll continue to publish our takeaways here on Radar each Friday and share full episodes on YouTube, Spotify, Apple, or wherever you get your podcasts.

Apple will fix CarPlay’s most annoying shortcomings in iOS 27

0


With iOS 27, Apple is giving CarPlay one of the biggest quality-of-life upgrades in years. The showcase feature is the ability to browse and watch videos, right on your car’s display, without touching your iPhone. But that’s just the start.

iOS 27 brings video playback to CarPlay

Ever since CarPlay arrived in 2014, Apple’s in-car infotainment system has solved the biggest problems of modern driving — getting directions, making calls and listening to music, all while keeping your eyes on the road.

With iOS 27, Apple is adding long-requested features like native video browsing, a universal media mini-player, improved playback controls and smarter location tracking. It’s also making under-the-hood changes to address one of CarPlay’s most persistent annoyances: unreliable wireless connections.

iOS 27 addresses all this, starting with the native video browsing that CarPlay users have wanted for a long time.

Video playback in CarPlay isn’t particularly new — Apple added AirPlay video streaming support for compatible vehicles with iOS 26.4, letting users cast videos to their cars’ screens while parked. But iOS 27 makes it more convenient.

The update will allow developers to build CarPlay apps with native video browsing, which means you’ll soon be able to find and play videos directly from your car’s screen. It should prove useful if you often find yourself waiting at school pickup or just want to kill time at the airport or an EV charging station.

When you start driving, CarPlay will cut the video feed and resume playback when park again.

The feature will require a compatible vehicle, though, and Apple has yet to name any specific models. It also requires iOS 27, which is currently in developer beta.

A mini-player that stays out of your way

Apple is also adding a persistent audio mini- player with a compact overlay you can use alongside your maps. The best thing? It works with every audio app. Whether you use Apple Music, Spotify or any other podcast app, you’ll always see what’s playing and be able to pause or skip without having to bounce between screens.

Now Playing gets a scrubbing slider

It took Apple years to get here, but the CarPlay screen will soon get an audio scrubbing slider. Previously, you were limited to the back and skip buttons. This made rewinding an audiobook or podcast more of an ordeal than it should be. The new scrubbing slider gives you more precision over media playback.

More reliable GPS in tunnels

Apple is also improving how iOS 27 handles GPS accuracy, using better algorithms with your iPhone’s sensors. The end goal is to prevent freezing maps and dropped position readings in tough environments like dense urban canyons, underpasses and underground parking.

In short, expect fewer “recalculating” alerts and more accurate arrival times.

Wireless CarPlay that actually stays connected

Apple is also improving CarPlay’s wireless reliability. The company hasn’t shared specific details about what’s changing, but iOS 27 is said to reduce dropouts and reconnections by improving Bluetooth and Wi-Fi handling.

iOS 27 is expected to arrive as a public beta in July, with full rollout expected in September.



Manna Secures SAIL III Operational Authorisation from UK Civil Aviation Authority – sUAS News

0


Approval Demonstrates Advanced Safety, Operational and Regulatory Maturity

Milestone Reinforces Manna’s Position as a European Leader in Advanced Air Delivery Operations

Manna Air Delivery has become one of the first drone delivery companies to receive a SAIL III Operational Authorisation (OA) from the UK Civil Aviation Authority (CAA), representing one of the most advanced operational approvals currently granted for uncrewed aircraft operations in the United Kingdom.

The authorisation marks a significant regulatory milestone for Manna and follows a comprehensive assessment process, undertaken with regulatory support from Avtrain, designed to evaluate the safety, operational, technical and organisational capabilities required for advanced drone operations.

Under the Specific Operations Risk Assessment (SORA) framework, SAIL III authorisations are reserved for operations that require a high degree of regulatory scrutiny and demonstration of safety assurance. Achieving this standard requires operators to provide extensive evidence across operational, technical and organisational domains, supported by detailed risk assessments, safety cases and independent review. The authorisation reflects the UK CAA’s confidence that Manna has demonstrated the systems, procedures and governance necessary to conduct advanced uncrewed aircraft operations in accordance with stringent regulatory requirements. 

This latest approval builds upon a series of significant regulatory milestones achieved by Manna across Europe and the United States. In 2021, Manna became the first drone delivery company in Europe to receive a Light UAS Operator Certificate (LUC) from the Irish Aviation Authority, providing the ability to self-authorise certain operations under the European regulatory framework. The company has also expanded flight activities into additional European jurisdictions, including Finland, alongside continued regulatory engagement and operational growth internationally. These milestones demonstrate Manna’s long-standing commitment to regulatory leadership and the safe integration of uncrewed aircraft within national aviation systems.

The assessment process also included independent review and support from Marshall in its role as a Recognised Assessment Entity for Flightworthiness (RAE(F)), providing an additional layer of technical and operational assurance throughout the authorisation process.

The approval provides further significant third-party validation of Manna’s operational maturity, safety culture and regulatory capability, demonstrating the company’s ability to satisfy some of the most demanding regulatory and safety requirements currently applied to advanced uncrewed aviation operations in Europe.

Manna’s Head of Regulation Kevin Houston said, “This authorisation represents an important milestone for Manna and reflects the significant effort invested across our organisation to meet the highest standards of safety and operational excellence.”

“The UK CAA has established a rigorous and highly professional assessment process for advanced uncrewed aviation operations. Successfully completing that process provides strong independent validation of our systems, procedures, technology and organisational maturity.”

The approval follows an extensive regulatory engagement process involving detailed operational reviews, technical assessments and safety evaluations.

Manna acknowledged the support and professionalism of all organisations involved throughout the authorisation process. “We would like to thank the team at the UK CAA for their engagement, professionalism and collaborative approach throughout the assessment process. Their commitment to enabling innovation while maintaining rigorous safety standards has been instrumental in bringing this authorisation to fruition. We would also like to thank Avtrain for its regulatory guidance and support throughout the application process, as well as Marshall for its expertise and proactive engagement as the Recognised Assessment Entity. Both organisations made invaluable contributions throughout what was a highly detailed and demanding regulatory process.”

The achievement further strengthens Manna’s position as one of Europe’s leading advanced air mobility operators and a recognised standard bearer for regulatory compliance, safety assurance and operational excellence within the European uncrewed aviation sector.

As regulators across Europe continue developing frameworks for advanced uncrewed aviation operations, the ability to secure higher levels of regulatory approval is increasingly viewed as a key indicator of operational readiness and organisational capability. As the sector continues to evolve, robust regulatory approvals will remain central to enabling the safe and scalable integration of uncrewed aircraft into national transport and logistics ecosystems.

Manna’s continued regulatory progress comes against a backdrop of rapid commercial momentum and growing global adoption of advanced last-mile air delivery. In April 2026, Manna announced a $50 million Series B funding round, bringing total funding to $110 million, to accelerate the scaling of aerial delivery across the United States and Europe. The company plans to grow from 180 employees to more than 570, creating 400 new roles across Ireland and the US in robotics, software engineering, mechanical engineering, aviation operations, ground operations and regulatory functions.

The funding will support Manna’s expansion to 40 operational bases across the United States and Europe, building on live operations already underway in Ireland and Texas. Manna has also expanded its platform partnerships, adding Uber to existing partnerships with Deliveroo, Just Eat and DoorDash, enabling customers to order through the apps they already use while Manna provides the aerial delivery infrastructure behind the scenes.

Together, these milestones demonstrate that drone delivery is moving from trial phase to scaled infrastructure: regulated, commercially integrated, and increasingly embedded into the future of suburban last-mile logistics.

This latest authorisation represents another important demonstration of Manna’s regulatory maturity and its ability to work successfully with aviation authorities across multiple jurisdictions. It reinforces the company’s commitment to maintaining the highest standards of safety, compliance and operational excellence as advanced uncrewed aviation continues to evolve internationally.

About Manna Air Delivery

Manna is the largest and most advanced drone delivery operator, dedicated to improving the world by making lightning-fast suburban deliveries affordable, green, and safe. With its proprietary technology, Manna ensures safe, sustainable, and fast delivery services, operating in the highest population density of any drone delivery operation globally. 

The company has completed over 300,000 successful deliveries and operates under the regulatory framework of EASA, setting the global benchmark for drone logistics. Partnering with both global giants and local businesses, Manna delivers a wide variety of goods including books, groceries, over the counter pharmacy, meals and essentials in just minutes, empowering businesses and significantly reducing carbon emissions.

Founded in 2019 by Bobby Healy, Manna has raised over $110 million in funding to date.

For more information visit: https://www.manna.aero/community


Discover more from sUAS News

Subscribe to get the latest posts sent to your email.

Why This Natori Underwear Is the Best I’ve Ever Worn

0


Until recently, my top drawer was stuffed with ill-fitting, outfit-spoiling underwear — and I know I’m not the only one with this problem.

The woes that plague women’s underthings are so numerous, in fact, that a whole vocabulary has sprung up to describe them: muffin top, camel toe, whale tail, wedgie, and of course, visible panty line (aka VPL). I’ve suffered all those afflictions and more in my search for the perfect pair.

But at long last (and with a little help from my friends), I’ve found them: the Natori Bliss French Cut Briefs.

These briefs are chic, indulgently comfy, and the first ones I reach for when the laundry is freshly done. It’s no wonder we made them a pick in our guide to the best women’s underwear.

Top pick

This pair is a sexy yet comfy take on the everyday brief; it has a high-cut leg and pretty lace trim that’s marvelously stretchy and itch-free. It’s pricey, but the pima cotton material is unusually soft.

Wearing this underwear is a fittingly blissful experience. The Natori briefs are made of sleek pima cotton (a soft long-fiber version of the material, the Champagne of cottons), which feels breathable, lightweight, and luxuriously soft. Thanks to being 6% Lycra, they’re also plenty stretchy, so they’ll maintain their smooth, body-skimming fit no matter how much I indulge this holiday season. The gusset — that’s underwear-insider-speak for “crotch” — is generously wide, offering ample protection against the indignities of a frontal wedgie. And the briefs’ nearly navel-grazing rise makes them more comfortable to wear than a hipster or bikini cut when my lower half is encased, as it often is, in high-waisted denim.

With their high rise and comprehensive bum coverage, I worried that the Natori briefs would look dowdy. But these are not your average multipack granny panties — they feel (dare I say it?) a little bit sexy. Subtle details make all the difference. Their high cut visually elongates my legs, showing enough skin to balance out the conservatively styled derriere. Delicate lace trim adds a touch of whimsy and feels slightly coquettish paired with the piqued, French-cut leg openings.

The generous stretchiness of the trim also helps the Natori briefs lie flat, preventing the pinching, rolling, and bunching that have condemned so many of my old pairs to the recycling heap. Although the company advises that you hand-wash the underwear, I’ve been tossing mine in the machine for three years and counting — and they’ve shown no signs of fraying or losing their shape. (I would, however, recommend skipping the dryer for extra longevity. Line-drying them or air-drying them on a towel rack is fine.)

Most crucially, their combination of stretchy, lie-flat lace and full posterior coverage confers a critical advantage: no panty lines. I’d long believed that banishing the quadra-buttocked silhouette meant enduring a thin strip of fabric crawling up my rear. So I was amazed at how the Natori briefs — which couldn’t be more un-thong-like — virtually disappeared under my clothes. That’s because the edges of the underwear are whisper thin, so there’s barely any ripple for my pants to skim. I can wear them with jeans, jersey knits, and even leggings without signaling my underwear preferences to passersby. It’s marvelously freeing.

This article was edited by Catherine Kast and Jason Chen.

Google launches AI agent for Ad Manager

0


Google launches AI agent for Ad Manager

Google is bringing generative AI directly into Google Ad Manager with the launch of Ask Ad Manager, a new Gemini-powered assistant designed to help publishers analyze performance, troubleshoot issues and navigate the platform using natural language.

The beta launches this month as Google pushes deeper into AI-powered ad operations.

What’s happening. Ask Ad Manager is a conversational AI agent built specifically for publishers using Google Ad Manager.

Unlike traditional reporting tools, publishers can ask questions in plain language and receive personalized answers, recommendations and reports based on their own Ad Manager data.

Google says the tool is designed to help users move from analysis to action faster by reducing the time spent generating reports, diagnosing problems and navigating the platform.

What it can do:

Troubleshoot delivery issues.

Instead of manually pulling reports to investigate underperforming line items, publishers can ask the AI agent questions and receive guidance on potential causes and next steps.

Generate reports on demand.

Users can request custom metrics, benchmarks and performance reports through a simple prompt rather than building multiple reports manually.

Navigate Ad Manager faster.

Ask Ad Manager can direct users to relevant pages within the platform and automatically apply the appropriate filters and settings based on the conversation.

Why we care. For publishers managing large inventories and complex campaigns, the ability to quickly surface insights and diagnose issues could reduce operational workload and accelerate decision-making.

The feature also reflects a growing shift across ad tech toward AI agents that can perform tasks and streamline workflows instead of simply generating information.

Looking ahead. Google says Ask Ad Manager is just the beginning of a broader move toward what it calls a more “agentic” future for advertising operations.

The company plans to introduce additional AI capabilities throughout the year, including developer tools such as REST APIs and an MCP server to support workflow automation and integrations.

Google is also developing specialized agents that could help publishers and agencies discover inventory, negotiate deals and execute campaigns more efficiently.

Bottom line. Ask Ad Manager brings Gemini-powered assistance directly into Google Ad Manager, giving publishers a new way to access insights, resolve issues and manage advertising operations through natural language prompts.

Azul Launches free JVM vulnerability risk assessment 


For most of Java’s history, a sophisticated exploit required a sophisticated attacker.  But, in this era of AI, Anthropic’s Claude Mythos demonstrates that AI can autonomously uncover previously unknown vulnerabilities and generate working exploit paths at scale — without human expertise. What once required deep, specialized expertise can now be accomplished with little more than an advanced AI model and an API key.

The result is an expanding population of potential attackers.  For large, complex Java estates with legacy versions in production, embedded or unmanaged JVMs and incomplete runtime visibility, that gap is a critical security and compliance liability.

To tackle this issue, enterprise Java platform provider Azul today has launched a free JVM vulnerability risk assessment to address the blind spot that autonomous AI exploitation tools are increasingly able to find. With mean time to exploit (MTTE) collapsing from months to days or hours, the unmanaged Java estate has become an urgent enterprise security vulnerability. Azul’s assessment gives DevOps and SecOps teams complete visibility into the hidden risks embedded in the runtime of their Java estate before threat actors get there first, and is designed to complement the broader security, licensing and compliance solutions and services delivered by Azul’s trusted partners.

“Anthropic’s Mythos has shown that AI can now discover and weaponize vulnerabilities on its own — including flaws that survived decades of human review. That’s the real lesson for every CISO: the deep expertise that used to stand between attackers and your software estate is no longer a barrier,” said Scott Sellers, co-founder and CEO of Azul, in the company announcement. “The unpatched JVM is already a growing liability, not a future one. Azul’s JVM vulnerability risk assessment was created to help security leaders find and close that exposure before AI-driven attackers can exploit it.”

The JVM Vulnerability Risk Assessment — See Everything, Prioritize What Matters

Azul’s JVM vulnerability risk assessment, available at no cost, maps JVM exposure, KEV risk and patch gaps across the entire enterprise Java estate and delivers a concrete remediation roadmap to close them. The assessment can be utilized as a standalone vulnerability analysis specific to a Java runtime estate or can be augmented into existing security, licensing and compliance solutions and services offered by Azul partners. Azul’s JVM vulnerability risk assessment is available at no cost, direct from Azul and via select Azul partners.

In a single engagement, organizations receive:

  • Executive-ready security dashboard: A visual summary of the entire Java estate, broken down by risk tier, publisher and Java version — designed for CxO-level consumption and board reporting.
  • Risk-by-version breakdown: Identification of the specific Java versions driving the highest exposure, so remediation effort can be directed where it matters most rather than spread uniformly.
  • Key Risk Indicators (KRIs) for AI-driven exploits: Visibility into which JVMs carry active Known Exploited Vulnerability (KEV) exposure — the highest-priority threat class recognized in the U.S. government’s CISA KEV catalog — as well as which instances are end-of-life or running below the current patch baseline.
  • Prioritized remediation roadmap: Concrete next steps ranked by impact, including which workloads to patch first, which to migrate off unsupported runtimes, and how to address extended support needs for legacy environments that cannot be immediately modernized.

Why Security Patch Velocity is the Frontline Defense

Java’s quarterly updates are the primary mechanism by which known vulnerabilities are remediated. But in an environment where autonomous AI systems continuously discover new vulnerabilities or chain together previously known CVEs into exploits, the pace of standard patch deployment is no longer sufficient on its own. Azul’s enterprise Java platform addresses this challenge through a multi-layered approach designed for large, complex Java estates:

  • Stable Critical Patch Updates (CPUs): Quarterly, production-safe patches containing only current CVE fixes. Azul Core is the only OpenJDK distribution which provides security-only updates, intended for immediate deployment without disrupting live environments.
  • Out-of-cycle emergency fixes: As vulnerabilities are discovered which demand immediate remediation, Azul provides security-only emergency fixes, collaborating with the Java community to help ensure safe delivery.
  • Full-stack visibility: Azul surfaces every JVM instance across the enterprise estate, including embedded and unmanaged runtimes that standard asset discovery typically misses — closing the gaps before they can be exploited.

The zero-day problem remains the hardest frontier. No scanner, SIEM (Security Information and Event Management), or EDR (Endpoint Detection and Response) platform can detect a vulnerability that has not yet been disclosed.