Sundar Pichai’s new pay package could be worth $692 million. Per a filing first spied by the FT, Alphabet has structured a three-year deal for its Google CEO that could make him one of the highest-paid executives on the planet — but most of it is tied to performance, including new stock incentives linked to Waymo and Wing, its drone delivery venture.

What’s striking is how little public fascination Pichai attracts compared to Google’s founders. Larry Page and Sergey Brin — the second- and fourth richest people in the world — have lately captured headlines for a different reason entirely; both have been snapping up lavish Miami properties, widely seen as a response to California’s proposed Billionaire Tax Act — a ballot initiative targeting the state’s roughly 200 billionaires with a one-time 5% levy on net worth exceeding $1 billion. Page reportedly spent over $173 million on two mansions in Coconut Grove, Florida, recently, while Brin was just linked to a $51 million megamansion 14 miles away, atop two earlier purchases totaling $92 million.

Pichai, by contrast, remains quietly rooted in Los Altos, California, as far as the public knows. He’s a billionaire, too — the nearly sevenfold growth in Google’s market cap since he took the helm in 2015 has made the stock he’s accumulated along the way hugely valuable. He and his wife currently hold shares worth nearly $500 million, with another estimated $650 million sold as of last summer, per Bloomberg’s calculations.

Pixar’s latest film, Hoppers, is now in theaters, and it’s excellent. Truly. It’s funny, surprising, heartfelt, and everything you’d want in a Pixar movie or any movie for that matter. Nowhere is that better seen than in a specific, semi-spoilery moment early in the film.

The moment was so shocking, so out of place, so ballsy, that the film’s director, Daniel Chong, was sure they’d have to cut it. This was a Disney movie, after all. In fact, he even came up with a simpler, less shocking version, fully expecting he’d have to replace it. But that didn’t happen. (Minor spoilers for Hoppers below.)

The moment in question, which is also the moment I chose to vaguely reference in my review, is when the Insect Queen, voiced by none other than Meryl Freaking Streep, is flying past our main character, Mabel. Mabel, who is a human in a beaver body, feels the queen fly by her and quickly, instinctively squishes her between her hands. It’s violent, it’s sudden, and it’s hilarious. It’s also a crucial moment in the film because the Queen’s death sets up the role of her son, Titus, voiced by Dave Franco.

Speaking to Slashfilm, Chong explained that while he loved the scene, he was ready to let the scene go. “I remember after our first screening or second screening … we were like, ‘You know what? There’s probably no chance we’re ever going to get away [with] doing this. So we should just change it now so that we don’t get this note and get screwed later,’” Chong said. “And so we reboarded it for the next screening, and we just had the Insect Queen die on her own, and then there was a scene where Mabel and [the mammal king] George looked like they killed her, but she didn’t.”

It works for the story but doesn’t quite have the same impact. Enter Pete Docter, the head of Pixar and director of Up, Inside Out, and Monsters Inc., as well as Andrew Stanton, director of Finding Nemo, Wall-E, and the upcoming Toy Story 5.

“Pete Doctor and Andrew Stanton were so mad at us,” Chong said. “They were like, ‘Why did you change that?!’ And we were like, ‘Well, we thought, because it’s Disney, we thought… ‘ And they were like, ‘No! Put that back in.’ And we were like, ‘Okay, but promise to make sure we don’t have to change this.’ And they were like, ‘We will fight for this.’ So those guys were on board. We were the ones that were second-guessing ourselves.”

Whether or not audiences will enjoy the moment as much as I, Docter, and Stanton did, we’ll know after the film’s opening weekend. But it’s awesome to see that they realized the moment is a perfect example of what makes Hoppers so special. It’s in theaters now.

Want more io9 news? Check out when to expect the latest Marvel, Star Wars, and Star Trek releases, what’s next for the DC Universe on film and TV, and everything you need to know about the future of Doctor Who.

Amazon’s website appears to be stabilizing after experiencing technical issues that kept users from logging in and prevented prices from displaying correctly. DownDetector reported a spike of outage reports around 2PM ET, but as of 5:56PM ET, user complaints have fallen significantly.

The Amazon.com homepage currently loads, and Engadgets staff have been able to load product pages and view prices without any problems. During the peak of the site’s issues, neither were loading consistently, and clicking through in some cases showed an error page with text that says “Sorry, something went wrong on our end.” Users also reported being unable to log into their accounts.

“We’re sorry that some customers may be experiencing issues while shopping,” Amazon said in a statement to Engadget. “We appreciate customers’ patience as we work to resolve the issue.” The company shared a similar sentiment with customers on X, confirming that it’s aware there’s a problem and acknowledging that its working on a fix. Amazon has yet to confirm whether the issue is fully resolved.

As a cloud provider through its Amazon Web Services (AWS) business, Amazon has experienced its fair share of outages, including one in October 2025 that took out services like Snapchat and Amazon’s own Alexa voice assistant for hours. The company’s website experiencing issues without a larger AWS outage seems a bit more unusual, and might suggest the problem lies outside of its cloud infrastructure.

Update, March 5, 5:56PM ET: Updated article to reflect improved performance on Amazon.com.

On Wednesday, the US Nuclear Regulatory Commission announced that it had issued its first construction approval in nearly a decade. The approval will allow work to begin on a site in Kemmerer, Wyoming, by a company called TerraPower. That company is most widely recognized as being financially backed by Bill Gates, but it’s attempting to build a radically new reactor, one that is sodium-cooled and incorporates energy storage as part of its design.

This doesn’t necessarily mean it will gain approval to operate the reactor, but it’s a critical step for the company.

The TerraPower design, which it calls Natrium and has been developed jointly with GE Hitachi, has several novel features. Probably the most notable of these is the use of liquid sodium for cooling and heat transfer. This allows the primary coolant to remain liquid, avoiding any of the challenges posed by the high-pressure steam used in water-cooled reactors. But it carries the risk that sodium is highly reactive when exposed to air or water. Natrium is also a fast-neutron reactor, which could allow it to consume some isotopes that would otherwise end up as radioactive waste in more traditional reactor designs.

The reactor is also relatively small compared to most current nuclear plants (345 megawatts versus roughly one gigawatt), and incorporates energy storage. Rather than using the heat extracted by the sodium to boil water, the plant will put the heat into a salt-based storage material that can either be used to generate electricity or stored for later use. This will allow the plant to operate around renewable power, which would otherwise undercut it on price. The storage system will also allow it to temporarily output up to 500 MW of electricity.

Bad news: most Americans are about to lose an hour of sleep next week. Good news: if you have trouble falling (or staying asleep), Newegg is currently selling Anker’s Soundcore Sleep A20 earbuds for $113.99 ($66 off) when you use coupon code MMSF88 at checkout, which drops them to just $6 shy of their lowest price to date.

A couple of us here at The Verge are fans of Anker’s last-gen sleep buds, which do a good job of muffling disruptive noises (including snoring). They’re lightweight and comfortable enough to wear overnight, even while sleeping on your side, with multiple ear tips and wings for a personalized fit. In fact, in his review, my colleague Thomas Ricker said that they improved his average sleep time by nearly 30 minutes within a two-week period.

What’s even more convenient is that they offer a variety of sleep-focused features to help you rest better. For example, you can use them to play a range of relaxing sounds, from meditation exercises and nature clips to white noise. You can use them as a regular pair of Bluetooth earbuds, too, just in case you prefer to listen to audiobooks or your own curated sleep playlist. They even come with adjustable EQ as well, though we wouldn’t recommend using them as your primary earbuds for music, given that they can’t match the audio quality you’d get from a pair of midrange earbuds from Apple, Sony, or Bose.

In addition, the Sleep A20 offer up to 14 hours of battery life and sleep tracking, providing insights into how long and how well you’ve slept via a companion app that also details your sleep positions and movements. The newer Soundcore Sleep A30 feature active noise cancellation, which is more effective at masking sounds than the A20’s passive isolation, but Anker’s last-gen earbuds remain a decent, budget-friendly option that can help you comfortably tune out most nighttime distractions for nearly half the price.

AI Projects Are Failing at an Alarming Rate

Enterprise AI adoption is accelerating. Budgets are growing. Boards expect measurable outcomes. Yet most AI initiatives fail to deliver business value.

Multiple industry reports consistently show that 60 to 80 percent of AI projects fail to reach production or fail to generate expected ROI.

The issue is not model accuracy.
It is not lack of data science talent.
It is not tool selection.

The real problem is execution.

CIOs, CTOs, and founders are discovering a hard truth: AI success is not about building models. It is about embedding intelligence into real business workflows and driving measurable outcomes.

This is where most AI initiatives collapse.

And this is exactly where the Forward Deployed Engineer model changes the equation.

Why AI Projects Fail

1. Lack of Clear Business Objectives

Most AI projects fail because they are not tied to a specific business outcome. Organizations launch AI initiatives without defining measurable KPIs such as cost reduction, revenue growth, or productivity improvement. Without a clear success metric, projects drift into experimentation and lose executive priority. AI must be anchored to business value from day one.

2. Slow and Inefficient Deployment Models

Traditional AI rollouts rely on long proof of concept cycles before production deployment. These extended timelines delay ROI and reduce stakeholder confidence. By the time the solution is ready, priorities often shift or budgets tighten. Speed to deployment is critical to maintaining momentum and demonstrating impact.

3. Poor Integration with Existing Systems

AI models built in isolation rarely deliver value. When solutions are not embedded directly into enterprise workflows, CRM systems, SaaS platforms, or operational dashboards, adoption remains low. Integration challenges create friction that limits usability and scalability. AI must operate within real business environments, not outside them.

4. Unclear ROI and Executive Ownership

AI initiatives often lack a defined return on investment timeline and accountable leadership. Without executive sponsorship and financial clarity, projects struggle to secure continued funding. Decision makers need transparent ROI milestones and outcome accountability. Clear ownership ensures alignment, faster decision-making, and sustained commitment.

Traditional AI Rollout vs Forward Deployed Engineer Model

Traditional AI Rollout Model

Timeline

• Discovery: 2 to 3 months
• POC: 3 to 6 months
• Pilot: 3 to 6 months
• Scale: 6 to 12 months

Total time to measurable ROI: 12 to 24 months

Risks

• POC never reaches production
• Business teams disengage
• Model drift due to delayed deployment
• Budget overruns

Forward Deployed Engineer Model

A Forward Deployed Engineer, or FDE, is a senior engineer embedded directly within the enterprise team. They operate at the intersection of: Engineering, Product, Data, Business operations.

They do not just build models.
They deploy, integrate, optimize, and iterate in real time.

Timeline

• Embedded discovery: 2 to 4 weeks
• Rapid prototyping inside live workflows
• Deployment in parallel with validation
• Continuous optimization

Time to measurable ROI: 3 to 6 months

Why it works

• Immediate integration into real systems
• Faster feedback loops
• Reduced translation gaps between business and engineering
• No handoff delays

What Is a Forward Deployed Engineer or Developer

A Forward Deployed Engineer is a senior technical expert embedded directly within an enterprise team to design, deploy, and scale AI-powered solutions in real production environments. Unlike traditional consultants, they operate inside the business, not outside it.

They work at the intersection of engineering, product, data, and operations to ensure AI solutions are tightly aligned with measurable business outcomes. Their focus is not just building models, but integrating them into workflows, systems, and customer-facing applications.

By shortening feedback loops and eliminating vendor handoffs, Forward Deployed Engineers accelerate time to value and reduce AI project failure risk. They are accountable for deployment, performance, and ROI, not just technical delivery.

How Forward Deployed Engineers Accelerate AI ROI

Business-First Problem Framing

Forward Deployed Engineers begin with the end in mind by identifying the exact business metric that needs improvement. Instead of experimenting with AI use cases, they define clear targets such as reducing operational costs, increasing revenue, improving cycle time, or automating manual processes. This outcome-driven approach ensures AI initiatives are aligned with strategic priorities from the start.

Embedded Workflow Integration

Rather than building AI systems in isolation, Forward Deployed Engineers integrate solutions directly into existing enterprise platforms and AI workflows. They connect models to CRM systems, ERP platforms, SaaS products, and internal dashboards so AI becomes part of daily operations. This deep integration increases adoption, improves usability, and accelerates measurable impact.

Rapid Deployment and Continuous Iteration

Traditional AI projects often delay deployment in pursuit of perfection. Forward Deployed Engineers prioritize early production releases with controlled iterations based on real-world feedback. By deploying quickly and optimizing continuously, they shorten feedback loops and ensure improvements are driven by live performance data. This significantly reduces time to ROI.

Centralized Accountability and Execution

AI initiatives frequently fail due to fragmented ownership across multiple vendors and internal teams. Forward Deployed Engineers provide unified technical and execution leadership under one accountable framework. This reduces coordination friction, speeds up decision-making, and keeps projects aligned with business outcomes, leading to faster and more predictable ROI realization.

How ISHIR Helps Enterprises Succeed with AI

ISHIR is an AI-native digital product engineering company. We do not deliver AI as a side offering. We build AI-powered systems that operate in production.

Our approach combines:

• Forward Deployed Engineers
• Global AI engineering teams
• Product-first architecture
• Enterprise-grade security
• Measurable ROI frameworks

ISHIR’s AI Execution Framework

Step 1: Business Outcome Mapping

We define:

• Target KPI
• Automation impact
• Revenue leverage
• Risk reduction

Step 2: Embedded Engineering

Our Forward Deployed Engineers integrate into your:

• Product teams
• Data pipelines
• Cloud environment
• Executive reporting structure

Step 3: Rapid AI Deployment

We:

• Prototype fast
• Deploy inside real systems
• Monitor live performance
• Optimize continuously

Step 4: Scale with Global Engineering Support

Once validated, our distributed teams accelerate expansion.

Frequently Asked Questions About AI Project Failure and Forward Deployed Engineers

Q. Why do most AI projects fail?

Most AI projects fail due to lack of business alignment, slow deployment cycles, poor workflow integration, and unclear ROI expectations rather than technical model limitations.

Q. What is a Forward Deployed Engineer?

A Forward Deployed Engineer is a senior AI and systems engineer embedded within an enterprise team to design, deploy, and optimize AI solutions directly in production environments.

Q. How do Forward Deployed Engineers reduce AI project failure?

They shorten feedback loops, integrate AI into live systems early, align execution with business KPIs, and eliminate vendor handoff friction.

Q. How fast can enterprises see ROI with the FDE model?

Most organizations begin seeing measurable impact within 3 to 6 months depending on scope and complexity.

Q. Is the FDE model suitable for enterprise SaaS companies?

Yes. It is especially effective for SaaS platforms that need to embed AI into core product features quickly and competitively.

Q. How does ISHIR structure AI engagements?

ISHIR combines embedded Forward Deployed Engineers with global AI development teams to accelerate production deployment and scale efficiently.

Q. What industries benefit most from embedded AI engineering?

Enterprise SaaS, fintech, healthcare tech, logistics, manufacturing, and B2B platforms see strong impact due to workflow-driven automation potential.

Q. What is the biggest mistake CIOs make with AI?

Treating AI as a research initiative instead of an operational transformation program tied to measurable KPIs.

Q. How does global delivery improve AI implementation?

It provides scalable engineering capacity, cost efficiency, and continuous development cycles while maintaining strategic alignment through embedded engineers.

Q. How should enterprises measure AI success?

Measure cost reduction, revenue uplift, process automation rates, time saved, and customer experience improvement.

Several years ago, Jared Cooney Horvath’s interest in teaching took a scientific turn.

He entered teaching during a period he calls “the decade of the brain” — when much of the buzz around education and learning covered new theories about brain activity and information processing. Horvath believed that if he learned more about the brain, he’d become a better teacher.

But the education ideas that captured the popular imagination in the early 2000s had to do with catering to so-called learning styles — right- versus left-brain thinkers or visual versus word learners — and notions about how to hasten cognitive development through certain outside stimuli. Remember those moms-to-be with headphones on their bellies for their babies to experience the “Mozart Effect” in utero?

The gains from these methods proved to be short-lived or difficult to measure accurately.

Yet the science of learning persists. And what Horvath — today a neuroscientist and education consultant — now knows about human cognitive development has spurred him to join a cohort of researchers who are questioning the proliferation of technology and education software in schools.

His new book “The Digital Delusion” feels like a logical progression from Jonathan Haidt’s 2024 bestseller “The Anxious Generation,” which looked at how hours spent in front of screens, especially on social media, with its rapid-fire videos and toxic commentary, has damaged children’s overall mental health and learning.

In “Digital Delusion,” Horvath outlines research showing how digital devices and screen time, at the expense of playtime, interferes with children’s cognitive development. Then he argues how the ubiquitous use in schools of laptops and edtech, at the expense of traditional skills like handwriting and note-taking, alters, for the worse, how kids learn.

Horvath’s book arrives at a pivotal moment, with digital systems facing a cultural reckoning: Social media companies defend themselves in court against accusations that their platforms harm mental health, and lawmakers propose legislation that would severely restrict screen time for kids under 13. Meanwhile, school districts across the United States impose bell-to-bell cellphone bans, and parents push to opt their children out of using digital devices for school.

Horvath takes a pragmatic approach on that score, suggesting arguments parents can use with administrators and at school board meetings. He has chapters that include examples of letters and other tools parents can customize to mobilize action at state and federal levels.

Some educators maintain that schools should emphasize responsible use of technology, including AI, to prepare students for a technology-driven workforce. Horvath isn’t convinced. First, he argues, workforce preparation should not be education’s priority, particularly in younger grades. Second, it’s inefficient: “Teach someone to use a tool and they’ll be able to use that tool,” he writes. “Teach someone how to think and they’ll be able to use any tool.”

Even so, Horvath insists he isn’t anti-tech: “This isn’t a book about resisting devices,” he writes. “It’s a book about reclaiming education as a deeply human endeavor.”

EdSurge spoke with Horvath about “The Digital Delusion” and his work with schools around the globe, including in Australia, which at the end of last year banned social media for anyone under 16.

This interview has been edited for length and clarity.

EdSurge: You make the point that whenever a new technology is introduced to a culture, early adopters are the enthusiasts. But for any given technology to have broad acceptance, it must pass muster with skeptics. Yet that didn’t really happen with digital technology in schools, did it?

Horvath: If I invented something, I had to convince you. This [product] will get rid of that stain on your shirt. This will keep your iceberg lettuce crisp in the fridge. If you promised something you had to live up to it, because for the few people who adopted it to begin with, if you didn’t clean their stains, they’re not coming back.

Digital technology never made a claim to anything. It just kind of appeared and people just started using it. When AI came out, the developers flat-out said, we don’t know what this does. Why don’t you guys tell us what it does? And for some reason we shoved it into schools and said, instead of me telling you what it does, why don’t I let my kids tell you what it does?

Something very weird happened where they made no claims to efficacy and then we jumped in and started using it. Our job now is to start to pull some of those weeds rather than protect before planting. And unfortunately that means there’s been a lot of victims along the way.

A lot of kids have suffered due to our rush to just put things in their hands, unfortunately.

I think we have this love affair with digital technology. I don’t know if it’s because of sci-fi or “Star Trek” or what. We intuitively think this is going to be helpful.

And now we’re just scrambling back.

You explain that children need to play for optimum cognitive development, but ordinary childhood play and behavior has been disrupted by screens. Is there evidence that if we take the technology away from children whose brains are still forming that they can bounce back?

Yes, absolutely. The good thing about human biology is it is wickedly malleable.

There’s two aspects to keep in mind. One, biology is also wickedly conservative. It changes all the time, but it never forgets anything. So if you have had a habit at one point and you drop that habit, you can move your biology a different way, but if you come back to that habit even once, your biology will have held onto that entire circuit. It’s a survival mechanism. Our genes, our brain, hold everything.

So when it comes to these tech habits, if you’ve already formed them as a kid, they will always kind of be there. If you think, I’m over this, and you pick up your phone, you will move much faster back into that habit than you did before.

The other thing to recognize here is everything we know about learning, and most of what we know about biology, basically starts after the age of 5. That’s when what we call human biological learning mechanisms really kick in.

From birth to about 5, you’re in a totally different world. The brain is basically in input mode. Gimme, gimme, gimme. And I’m going to hold onto everything. This is why if a kid grows up in a house with two languages, they will easily learn two languages because the brain just says gimme, gimme, gimme.

So that’s where I think the super danger zone comes in. If you develop habits or problems before the age of 5, when you hit 5, the brain locks itself down. You won’t be able to consciously remember what happened before the age of 5, but all of that [input] forms the foundation upon which further learning is going to occur.

My fear is if you form a habit before the age of 5 and then your brain locks down, are you now stuck in a spot where it will be very hard to get that out? If you’ve already addicted your kid before age 5, be careful. I don’t know what that’s going to mean when they get older.

There is data that says around 40 percent of 2-year-olds have tablets.

Why? My question is just why? There are a lot of states right now putting forward bills to limit screen time in primary years: K through [grade] 2, 90 minutes; [grades] 2 through 5, two hours a day. To which I always reply, why any hours?

I could easily make a case they don’t need any of this at any moment. It makes no sense for learning and development why [technology] needs to interface with anything they’re doing.

But by banning, aren’t we setting up a mystique around technology — causing a different kind of distraction around the yearning to use it?

That’s what you want. By banning and building a mystique, you give kids aspirations. I think back to my generation, when we turned 16, you couldn’t stop us from driving. Why? Because with our parents, that was the hold: you want to go to your friend’s house? You got a bike, you got feet, I’m not driving you. You want to get to school? There’s a bus, you got feet, I’m not driving you. So by the time we knew we could drive, that’s the first thing we did.

If by banning tech, that makes kids say when I’m 18, I’m using tech — then, good, that means I have 18 years to train you to be ready to use that machine.

Can schools realistically go back to paper? Textbooks, for instance, are expensive and take longer to update than websites, which are dynamic.

It’s funny, this is where you get the clash between different masters. In a good rule of thumb you can only serve one master at a time. So we’ve got issues of, I want my kids to learn, but I have monetary constraints and I have administrative bureaucracy that I’ve got to wend my way through.

When you’ve got multiple masters, eventually you’ve got to settle on one because if you try and serve many, no one’s going to be happy. And I would hope that in education we choose learning as our ultimate master. If that means, look, we have to devote more of our budget to textbooks and that means we won’t be able to do X this year, then so be it.

If that means, look, we’re going to only use the website for the last two years of history, but we’re going to have the book for the rest because it’s better for learning, then so be it.

I don’t know how much more research we need on this. People learn more from hard copy text than they do from digital text. It’s done. That battle is over. So if learning is our outcome, why not go back to what we know works best for that?

Can you explain the findings around taking notes by hand?

Most students think note-taking is something they do while they learn. So [they think] if AI does it for me — cool! But they miss the point. Note-taking is the learning, not something that’s happening in parallel to learning. That is the learning. Because that’s where you’re doing your transformation: Your teacher said it. I now have to analyze it, think about it, organize it, get it out.

That requires friction. Your brain is going much faster. So the handwriting is constraining the speed with which you can think, which in turn is forcing you to focus on ideas, which in turn is transforming those ideas as you’re going along.

That is the definition of learning.

The act of handwriting is arguably the most complex thing we do. When it comes to motor skills, there might be nothing more complex than that.

We talk about the difference between gross- and fine-motor movements. Name one skill we do that is so minutely fine as handwriting and so varied as handwriting. If you’re using a pen versus a pencil versus a crayon versus a marker, you’re doing very subtly different movements.

Those develop so much more awareness and understanding of the body in a way that then translates into other fields in ways we’ve never seen from any other skill before.

If you know how to write, you will become better at reading. If you know how to write, you will become better at recognizing faces. Why? We don’t know. But everything seems to be correlated back to that skill.

So when people debate [whether] handwriting is still worth teaching? Of course. Is cursive still worth teaching? Of course. No one’s going to use cursive as an adult. That’s not why we’re teaching it, baby. It has nothing to do with what you’re going to do as an adult. ’

You were just in Australia. What is the feedback from the social media ban?

The response is overwhelmingly positive. Basically every school I worked at, the kids are fine with it. Teachers are fine with it. All of a sudden, behaviors are getting so much better in school. They said the biggest problem is with parents, oddly enough, who basically have to hang out with their kids and they don’t know what to do. If that’s our biggest problem, we’ll solve that. Hang out with your kid.

Any time you remove something from your kid’s heart, you’re going to have to fill it with something else. You’re going to have to fill it with yourself, which means you’re going to have to take some of your own tech out of your own life to devote more of your time to your kid.

“In a normal Layer-2 switch, the switch learns the MAC of the client by seeing it respond with its source address,” Moore explained. “This attack confuses the AP into thinking that the client reconnected elsewhere, allowing an attacker to redirect Layer-2 traffic. Unlike Ethernet switches, wireless APs can’t tie a physical port on the device to a single client; clients are mobile by design.”

The back-and-forth flipping of the MAC from the attacker to the target, and vice versa, can continue for as long as the attacker wants. With that, the bidirectional MitM has been achieved. Attackers can then perform a host of other attacks, both related to AirSnitch or ones such as the cache poisoning discussed earlier. Depending on the router the target is using, the attack can be performed even when the attacker and target are connected to separate SSIDs connected by the same AP. In some cases, Zhou said, the attacker can even be connected from the Internet.

“Even when the guest SSID has a different name and password, it may still share parts of the same internal network infrastructure as your main Wi-Fi,” the researcher explained. “In some setups, that shared infrastructure can allow unexpected connectivity between guest devices and trusted devices.”

No, enterprise defenses won’t protect you

Variations of the attack defeat the client isolation promised by makers of enterprise routers, which typically use credentials and a master encryption key that are unique to each client. One such attack works across multiple APs when they share a wired distribution system, as is common in enterprise and campus networks.

In their paper, AirSnitch: Demystifying and Breaking Client Isolation in Wi-Fi Networks, the researchers wrote:

Although port stealing was originally devised for hosts on the same switch, we show that attackers can hijack MAC-to-port mappings at a higher layer, i.e., at the level of the distribution switch—to intercept traffic to victims associated with different APs. This escalates the attack beyond its traditional limits, breaking the assumption that separate APs provide effective isolation.

This discovery exposes a blind spot in client isolation: even physically separated APs, broadcasting different SSIDs, offer ineffective isolation if connected to a common distribution system. By redirecting traffic at the distribution switch, attackers can intercept and manipulate victim traffic across AP boundaries, expanding the threat model for modern Wi-Fi networks.

The researchers demonstrated that their attacks can enable the breakage of RADIUS, a centralized authentication protocol for enhanced security in enterprise networks. “By spoofing a gateway MAC and connecting to an AP,” the researchers wrote, “an attacker can steal uplink RADIUS packets.” The attacker can go on to crack a message authenticator that’s used for integrity protection and, from there, learn a shared passphrase. “This allows the attacker to set up a rogue RADIUS server and associated rogue WPA2/3 access point, which allows any legitimate client to connect, thereby intercepting their traffic and credentials.”

A recent Omegle interaction between American streamer IShowSpeed and a Mongolian woman has gone viral.

As IShowSpeed matched with the soundly sleeping woman on the video chat platform, he immediately started shouting, “Hey, wake the fuck up, wake up, wake up, yeah, wake up, it’s that time.” The startled woman woke up, initially appearing confused by the abrupt awakening.

During their exchange, IShowSpeed told her it was “four o’clock in the evening, I mean afternoon,” to which she remarked that it was actually five in the morning in Mongolia.

Upon learning her location, the streamer revealed he had previously visited the country. The snoozing woman quipped, “Oh, I know.“

IShowSpeed swiftly apologized, saying,

“Oh, I’m sorry I interrupted your sleep. I hope you have a sweet dream and dream of sweet things.”

The woman politely thanked him and said goodnight before ending the call.

https://x.com/IShowSpeedHQ/status/191058787763484682

Meanwhile the internet erupted with memes

pic.twitter.com/MCFlAhSjIE
— Lutaa (@LutochirT) February 26, 2026

pic.twitter.com/HdX7H86t8m
— Aman Singh (@Amanravimaurya) February 26, 2026

The girl rn pic.twitter.com/YEOuHzXjae
— Mobor James (@Emuobor_) February 25, 2026

pic.twitter.com/BqSrIP82Rz
— FinegirlDami (@_FinegirlDami) February 26, 2026

Please don’t go around waking up the Mongolians, the last thing we need right now is one of them motherfuckers deciding it’s time to rerun the glory days.
— enterfan (@JustWantedPax) February 26, 2026

Girl: what do you want pic.twitter.com/Q79UWwgrG1
— PRAKASH (@prakash_21oct) February 26, 2026

In 2025, IShowSpeed visited Mongolia as part of his international travel content. The 20-year-old streamer, whose real name is Darren Jason Watkins Jr., was greeted by a swarming crowd of fans in Ulaanbaatar. During his trip, he explored Mongolian culture, trying traditional clothing and food, visiting the Chinggis Khan Museum, wrestling with local fighters, and experiencing horse riding. He called Mongolia “underrated.” 100%

| WATCH: Speed matched with a girl on Omegle who was asleep and woke her up by screaming pic.twitter.com/xPeU8lbGin
— Speedy HQ (@IShowSpeedHQ) February 25, 2026

See Also: IShowSpeed’s Reunion With 2021 ‘Kinda Homeless’ Teens Leaves The Internet In Stitches

See Also: ‘I need a therapist’ IShowSpeed Opens About Mental Health Amid USA Tour; Says ‘I Feel Alone’

Cover: X/Twitter

My first post made the case for what a semantic layer can bring to the modern enterprise: a single source of truth accessible to everyone who needs it—BI teams in Tableau and Power BI, Excel-loving analysts, application integrations via API, and the AI agents now proliferating across organizations—all pulling from the same governed, performant metric layer. The promise is compelling. But what happens when organizations actually build and deploy one? To find out, I interviewed several early adopters who’ve moved semantic layers from concept to production. Four themes emerged from those conversations: some surprising, some predictable, and a few that will sound familiar to anyone who’s ever shipped data infrastructure.

The first theme: Semantic layers are showing up in unexpected places. Most discussion positions them as enterprise-level infrastructure—a single location capturing all company metrics for centralized access and governance. That’s still the primary use case. But practitioners are also deploying semantic layers for narrower purposes. One organization, for example, built their semantic layer specifically to power a targeted chatbot application—letting users query data conversationally without any traditional BI tools in the mix. No Power BI, no Excel, just an AI interface pulling from governed metrics. The rationale for these smaller deployments is straightforward: Semantic layers deliver high accuracy on structured data, even with lightweight models. The core value drivers remain speed, accuracy, and access—but organizations are finding more ways to extract that value than the enterprise-wide vision suggests.

The second theme: AI is the reason organizations are moving now. The other benefits still matter—single source of truth, multitool compatibility, true self-serve access, cost reduction in cloud environments—but when I asked practitioners why they prioritized a semantic layer today rather than two years ago, the answer was consistent: AI. Whether it was a specific chatbot project or enabling AI-driven analytics at scale, AI requirements were the catalyst. This tracks with what I discussed in my first post: Structured data alone isn’t enough for reliable AI analytics. Adding semantic context—field descriptions, model definitions, object relationships—dramatically improves accuracy. The data industry has noticed. Semantic layers have moved from niche infrastructure to strategic priority: Snowflake, Databricks, dbt Labs, and Microsoft have all made significant investments in the past year.

The third theme: Semantic layers reduce work for developers while making trusted data easier to access. Multiple practitioners cited the value of maintaining metrics and business logic in a single location. Any analyst knows the pain of metric sprawl—leadership requests a change to a core KPI, and you discover it’s been defined a dozen different ways across databases, BI tools, and spreadsheets scattered through the organization. The semantic layer eliminates the chase. One engineering lead described a financial metric that had accumulated over 60 versions across the company. After deploying the semantic layer, there was one.

Want Radar delivered straight to your inbox? Join us on Substack. Sign up here.

Access simplifies too. Instead of provisioning controls across warehouses, BI workspaces, individual dashboards, and cloud storage locations, users connect directly to the semantic layer and pull data into the tool of their choice. One organization was surprised to find that after deployment, the most common access point was Excel. But with the semantic layer, that wasn’t a problem: The data served in Excel was identical to what powered their AI tools, Power BI dashboards, and application integrations via API.

The fourth theme will sound familiar to anyone who’s shipped data infrastructure: The biggest challenge isn’t the technology—it’s the data itself. Every practitioner I spoke with identified the same bottleneck: consistency, availability, and accuracy of the underlying data. Engineers and analysts can build the semantic layer, but they can’t will clean data into existence. Success requires close collaboration with business stakeholders, clear ownership of metrics, and leadership alignment to prioritize the work. None of that is new. But despite these challenges, everyone I interviewed reached the same conclusion: The semantic layer is worth the effort.

Semantic layer technology is still early. The tools, vendors, and best practices are evolving fast—what works today may look different in a year. But these conversations revealed a clear signal beneath the noise: semantic layers are becoming critical AI infrastructure. The practitioners I spoke with aren’t experimenting anymore. They’re operationalizing. And despite the expected challenges around data quality and organizational alignment, they’re seeing real returns: fewer metric versions to maintain, simpler access controls, and AI tools that actually produce trusted answers.

My first article made the case for what a semantic layer could be. This one asked what happens when organizations actually build them. The answer: It’s hard, it’s worth it, and for companies serious about AI-driven analytics, the semantic layer is no longer a nice-to-have. It’s the foundation.

The decision was made following a review of “country-specific” conditions

Food delivery service Deliveroo will cease operations in Singapore after Mar 4.

In a statement on its website today (Feb 25), the platform said that it was exiting the market and would begin an orderly wind-down process.

“This is a difficult decision and follows a review of country-specific conditions, and our focus on investing where we see the clearest path to sustainable scale and long-term leadership,” said the statement.

The company added that it would “work closely” with local teams to support customers, partners and riders through the transition.

It will also exit three other markets

Deliveroo first launched in Singapore back in Nov 2015.

According to the company, the exit was part of a broader review of the company’s international portfolio.

Apart from Singapore, its parent company, DoorDash, said in a separate press release that it will cease operations in three other markets: Qatar, Japan, and Uzbekistan.

It is also implementing limited operational changes in select locations, including investing in certain engineering roles in the United Kingdom.

DoorDash acquired Deliveroo in May 2025, in a deal valued at US$3.85 billion. The move was aimed at helping DoorDash grow its market share in Europe, competing against Just Eat and Uber Eats. Britain, Ireland, France and Italy are among Deliveroo’s major markets.

Vulcan Post has reached out to both Deliveroo and DoorDash for comments.

Featured Image Credit: Victor Velter/ Shutterstock.com

Technology shapes every aspect of our daily lives, yet the stories behind the code and the circuits often remain untold. While the tech industry has historically been male-dominated, women have played a crucial role in its development since the very beginning. From writing the first computer program to leading some of the world’s most influential tech giants, women have consistently pushed the boundaries of what is possible.

We believe it is essential to recognize these achievements. Celebrating the women who have made substantial contributions to the technology industry isn’t just about looking back; it’s about understanding the foundation upon which our modern digital world is built. This post highlights the remarkable achievements of key figures, the challenges they overcame, and the lasting impact they have had on the tech world.

The Early Visionaries

Before modern computers even existed, women were already theorizing about their potential. Their work laid the groundwork for everything we use today, from smartphones to space exploration software.

Ada Lovelace: The First Programmer

Often cited as the first computer programmer, Ada Lovelace saw potential in Charles Babbage’s Analytical Engine that went far beyond simple calculation. In the mid-1800s, she wrote an algorithm intended to be processed by a machine, effectively creating the first computer program.

Her contribution was visionary. She understood that computers could do more than just crunch numbers; they could manipulate any symbol, including music and art. We have seen time and again how foundational her insights were. Every time we write a line of code or deploy a new software solution, we are building on the conceptual framework she established nearly two centuries ago.

Grace Hopper: The Queen of Code

Rear Admiral Grace Hopper was a pioneer who fundamentally changed how we communicate with computers. She was one of the first programmers of the Harvard Mark I computer and developed the first compiler for a computer programming language.

Her work led to the development of COBOL, an early high-level programming language that is still in use today. Hopper famously said, “The most dangerous phrase in the language is, ‘We’ve always done it this way.’” This spirit of innovation drives the industry forward. We have witnessed firsthand how her legacy of user-friendly programming languages allows businesses to adapt and grow with ever-demanding technology changes.

Contemporary Leaders

The legacy of Lovelace and Hopper continues with modern leaders who are not only advancing technology but also reshaping the culture of the industry.

Sheryl Sandberg: redefining Leadership

As the former COO of Meta (formerly Facebook), Sheryl Sandberg has been a transformative figure in the tech world. Her leadership helped scale one of the most influential social media platforms in history. Beyond her executive role, her book Lean In sparked a global conversation about women in the workplace.

Sandberg’s impact extends beyond corporate strategy. She has provided a roadmap for women aspiring to leadership roles in male-dominated fields. Her work emphasizes that diversity in leadership leads to better decision-making and more innovative solutions. We have observed that companies embracing these leadership principles are often better positioned to handle complex challenges.

Reshma Saujani: Closing the Gender Gap

Reshma Saujani founded Girls Who Code, a nonprofit organization dedicated to closing the gender gap in technology. Recognizing that the number of women in computer science was declining, she took action to change the trajectory for future generations.

Since its inception, Girls Who Code has reached hundreds of thousands of girls, equipping them with the computing skills needed to pursue 21st-century opportunities. Saujani’s work ensures that the future of tech is inclusive. We know from experience that when diverse voices are included in the development process, the resulting technology serves a broader audience more effectively.

Overcoming Challenges

The road has not been easy for these pioneers. Women in tech have historically faced significant barriers, from lack of access to education to workplace discrimination.

Systemic Barriers

Early pioneers often had to fight just to be in the room. Grace Hopper was initially rejected by the Navy due to her age and size. Ada Lovelace’s work was often overshadowed by her male contemporaries. Even today, women in tech report facing bias in hiring, promotion, and venture capital funding.

Despite these hurdles, their resilience is a testament to their dedication. They have proven that talent and innovation are not defined by gender. We have worked closely with many organizations that have actively dismantled these barriers, and the result is invariably a more dynamic and successful team.

The Importance of Diversity

Diversity drives innovation. When teams are composed of individuals with different backgrounds and perspectives, they are more likely to identify unique solutions to complex problems. A diverse tech workforce is essential for creating products that cater to a global user base.

We have seen that diverse teams are more creative and effective. By fostering an inclusive environment, the tech industry can unlock untapped potential and drive progress at an accelerated rate.

Paving the Way for the Future

The contributions of women in tech have opened doors for the next generation. Today, more women are entering STEM fields than ever before, inspired by the role models who came before them.

Mentorship and Support

Programs like Girls Who Code and various STEM initiatives are critical. They provide the mentorship and support networks that were often missing for earlier generations. We have had the pleasure of seeing young women thrive when given the right resources and encouragement.

A Call to Action

The journey is far from over. To ensure a vibrant and innovative future for the technology industry, we must continue to support and promote women in STEM.

Here is how you can help:

• Mentor: Offer your time and expertise to young women interested in tech.
• Advocate: Push for inclusive hiring practices within your organization.
• Educate: Encourage schools to provide equal access to computer science education.
• Celebrate: Continue to highlight the achievements of women in technology.

Since the early days of computing, women have been integral to technological progress. From Lovelace’s algorithms to Saujani’s advocacy, their impact is undeniable. We are confident that by supporting diversity and inclusion, the tech industry will continue to reach new heights, solving the problems of tomorrow with the brilliance of today’s innovators. Let’s commit to ensuring that the future of tech belongs to everyone!

The now-viral X post from Meta AI security researcher Summer Yue reads, at first, like satire. She told her OpenClaw AI agent to check her overstuffed email inbox and suggest what to delete or archive.  

The agent proceeded to run amok. It started deleting all her email in a “speed run” while ignoring her commands from her phone telling it to stop. 

“I had to RUN to my Mac mini like I was defusing a bomb,” she wrote, posting images of the ignored stop prompts as receipts.  

The Mac Mini, an affordable Apple computer that sits flat on a desk and fits in the palm of your hand, has become the favored device these days for running OpenClaw. (The Mini is selling “like hotcakes,” one “confused” Apple employee apparently told famed AI researcher Andrej Karpathy when he bought one to run an OpenClaw alternative called NanoClaw.) 

OpenClaw is, of course, the open source AI agent that achieved fame through Moltbook, an AI-only social network. OpenClaw agents were at the center of that now largely debunked episode on Moltbook in which it looked like the AIs were plotting against humans.  

But OpenClaw’s mission, according to its GitHub page, is not focused on social networks. It aims to be a personal AI assistant that runs on your own devices.  

The Silicon Valley in-crowd has fallen so in love with OpenClaw that “claw” and “claws” have become the buzzwords of choice for agents that run on personal hardware. Other such agents include ZeroClaw, IronClaw, and PicoClaw. Y Combinator’s podcast team even appeared on their most recent episode dressed in lobster costumes. 

But Yue’s post serves as a warning. As others on X noted, if an AI security researcher could run into this problem, what hope do mere mortals have? 

“Were you intentionally testing its guardrails or did you make a rookie mistake?” a software developer asked her on X.  

“Rookie mistake tbh,” she replied. She had been testing her agent with a smaller “toy” inbox, as she called it, and it had been running well on less important email. It had earned her trust, so she thought she’d let it loose on the real thing. 

Yue believes that the large amount of data in her real inbox “triggered compaction,” she wrote. Compaction happens when the context window — the running record of everything the AI has been told and has done in a session — grows too large, causing the agent to begin summarizing, compressing, and managing the conversation.  

At that point, the AI may skip over instructions that the human considers quite important.  

In this case, it may have skipped her last prompt — where she told it not to act — and reverted back to its instructions from the “toy” inbox. 

As several others on X pointed out, prompts can’t be trusted to act as security guardrails. Models may misconstrue or ignore them. 

Various people offered suggestions that ranged from the exact syntax Yue should have used to stop the agent, to various methods to ensure better adherence to guardrails, like writing instructions to dedicated files or using other open source tools. 

In the interest of full transparency, TechCrunch could not independently verify what happened to Yue’s inbox. (She didn’t respond to our request for comment, though she did respond to many questions and comments sent her way on X.) 

But it doesn’t really matter. 

The point of the tale is that agents aimed at knowledge workers, at their current stage of development, are risky. People who say they are using them successfully are cobbling together methods to protect themselves.

One day, perhaps soon (by 2027? 2028?), they may be ready for widespread use. Goodness knows many of us would love help with email, grocery orders, and scheduling dentist appointments. But that day has not yet come. 

Back when the movie industry had fewer legitimate existential threats, studios and theaters were utterly apoplectic about the scourge of film pirates supposedly robbing them blind. At the peak of this panic, the Motion Picture Association of America (MPAA) released a controversial study claiming cam-wielding criminals had cost the industry a whopping $18.2 billion in lost revenue for 2005 alone.

The year before, industry-wide anxiety also compelled Warner Bros. and the MPAA to produce one of the most memorable PSAs of the new millennium. Titled “You Wouldn’t Steal a Car,”—and recently revealed to possibly have been made with pirated fonts—the commercials made the case that procuring a burnt DVD or torrent file of a grainy Meet the Fockers was, in fact, tantamount to a B&E robbery or grand theft auto. The two 45-second morality plays contained interstitial title cards insinuating that the presumably honest and law-abiding individual watching would never commit such real-world transgressions. So how come they’re OK with anonymously making or watching digital copies of media from the comfort of their own home?

Immediately, the public reacted to the ads with ridicule, not just at their pearl-clutching corniness, but the central premise itself. As many would point out in the years to come—sure, they’d pass on hotwiring a car in their neighbor’s driveway, but they’d be perfectly fine with downloading one. Once merely a meme, this fantasy is now one huge step closer to reality thanks to a recently reported breakthrough in the field of 3D printing.

A team at the Massachusetts Institute of Technology (MIT) has recently developed a printer with four different extruders that outputs five different materials to produce a fully functioning linear motor in about three hours. Publishing their work in the industry journal Virtual and Physical Prototyping, the team explained how by retrofitting a printer with enough extruders to handle the various materials needed to make a working motor, they decimated the usual production time for such a device and brought the material costs down to around $0.50.

In a world where the usual cost of building such a prototype would be comparatively huge and turnaround time could take anything from weeks to months, the team’s proof-of-concept has the potential to change the manufacturing world. The linear motor they fabricated, which operates in a straight line rather than spinning like a car engine’s, is primarily used in automation and manufacturing. It’s hard to fathom the scope of impact if factories floors were someday able to print out complex replacement parts rather than be at the whim of a sluggish global supply chain.

While the linear motor is a far cry from the complexity of a V12 engine, the MIT team’s development is unquestionably a significant baby step in that direction. We’re clearly still a ways off from being able to download and slice 2026RangeRover.stl files, but you might be surprised to learn that there are already hobbyists successfully 3D-printing parts and even entire cars, one piece at a time.

Big Auto may want to get a head start on brainstorming their own PSAs while there’s still time.

The Stop Killing Games campaign is evolving into more than just a movement. In a YouTube video, the campaign’s creator, Ross Scott, explained that organizers are planning to establish two non-governmental organizations, one for the European Union and another for the US. According to Scott, these NGOs would allow for “long-term counter lobbying” when publishers end support for certain video games.

“Let me start off by saying I think we’re going to win this, namely the problem of publishers destroying video games that you’ve already paid for,” Scott said in the video. According to Scott, the NGOs will work on getting the original Stop Killing Games petition codified into EU law, while also pursuing more watchdog actions, like setting up a system to report publishers for revoking access to purchased video games.

The Stop Killing Games campaign started as a reaction to Ubisoft’s delisting of The Crew from players’ libraries. The controversial decision stirred up concerns about how publishers have the ultimate say on delisting video games. After crossing a million signatures last year, the movement’s leadership has been busy exploring the next steps.

According to Scott, the campaign leadership will meet with the European Commission soon, but is also working on a 500-page legal paper that reveals some of the industry’s current controversial practices. In the meantime, the ongoing efforts have led to a change of heart from Ubisoft since the publisher updated The Crew 2 with an offline mode.

A second fueling test on NASA’s Space Launch System rocket ended Thursday night, giving senior managers enough confidence to move forward with plans to launch four astronauts around the Moon as soon as March 6.

Unlike the first attempt to load propellants into the SLS rocket on February 2, there were no major leaks during Thursday’s practice countdown at Kennedy Space Center in Florida. Technicians swapped seals at the launch pad after hydrogen gas leaked from the rocket’s main fueling line earlier this month. This time, the seals held.

“For the most part, those fixes all performed pretty well yesterday,” said Lori Glaze, acting associate administrator for NASA’s exploration programs. “We were able to fully fuel the SLS rocket within the planned timeline.”

The results keep the Artemis II mission on track for liftoff as soon as next month. NASA gave up on a series of February launch opportunities after encountering a persistent hydrogen leak during the first Wet Dress Rehearsal (WDR).

“We’re now targeting March 6 as our earliest launch attempt,” Glaze said. “I am going to caveat that. I want to be open, transparent with all of you that there is still pending work. There’s work, a lot of forward work, that remains.”

If teams complete all of that work, liftoff of the Artemis II mission could occur within a two-hour window opening at 8:29 pm EST on March 6 (01:29 UTC on March 7). NASA has other launch dates available on March 7, 8, 9, and 11, but the mission may have to wait until April. There are approximately five days per month that the mission can depart the Earth after accounting for the position of the Moon in its orbit, the flight’s trajectory, and thermal and lighting constraints.

The Artemis II mission will last between nine and 10 days, taking NASA’s Orion spacecraft with commander Reid Wiseman, pilot Victor Glover, and mission specialists Christina Koch and Jeremy Hansen around the far side of the Moon before returning to Earth for splashdown in the Pacific Ocean. Wiseman’s crew will set the record for the farthest humans have ever traveled from Earth, and will become the first people to fly to the vicinity of the Moon since 1972.