Tech
Two pizzas and a prototype: How agentic AI is rewiring Amazon’s teams and upending its traditions

[Editor’s Note: Agents of Transformation is an independent GeekWire series, underwritten by Accenture, exploring the adoption and impact of AI and agents. See coverage of our related event.]
Amazon is legendary for its process of “working backwards.” Start with a customer problem, imagine a future in which it’s solved, draft a press release and FAQs as if it had already happened, obsess over the document until it’s just right, and then go make it a reality.
But sometime last year, it dawned on Swami Sivasubramanian, Amazon Web Services VP of agentic AI, that new coding tools had suddenly made it easier for his teams to develop a demo — actual working software — than to write the classic six-page Amazon “PRFAQ.”
So they began starting with the prototype instead.
If something is “a low-risk bet where we just want to prove our intuition, then I actually say, let’s first go build the demo, and then iterate,” Sivasubramanian said in an interview last week, in advance of his keynote address Wednesday at the AWS New York Summit.
It’s an illustration of how agentic tools are reshaping even the most entrenched workplace practices and traditions. But it’s just one of the ways that the AWS agentic AI team is departing from the company’s established norms, and in some ways returning to its roots.
Inside Amazon, CEO Andy Jassy says he wants the company to run like the world’s largest startup. Sivasubramanian’s division may be the closest thing to what that looks like in practice.
Back to two pizzas
The AWS agentic AI division is organized into dozens of small teams, many of them just large enough to feed with two pizzas. That was the organizing principle that Amazon pioneered in its early days and that much of the company outgrew as it scaled to 1.5 million employees.
When Matt Garman, the CEO of AWS, carved out agentic AI as its own division last year, Sivasubramanian went with small teams on purpose. It matches the new reality of the AI era: projects that once required 30 to 40 people, he said, can now be done by teams of six to eight.
Case in point: the Amazon Quick desktop app, which connects to a user’s email, calendar, Slack, documents, and other apps in a single workspace, and uses AI to search across them, answer questions, and perform tasks. It’s Amazon’s entry in a market where Anthropic, Microsoft, Google, and OpenAI have captured much of the attention.
It traces its roots to late January of this year, when Sivasubramanian said it became clear to him and others on the team that the underlying models had gotten good enough that the main missing ingredient was connecting them to the systems where people actually work.
He pulled together a team of about six engineers to build it. Six weeks later, 200 people inside Amazon were using it. Ten weeks in, it was up to 10,000 internally. The team circled back to write the PRFAQ after the product was already in beta, to help refine their approach to the external launch. They shipped on April 28, three months after they got started.
Under the old system — writing the PRFAQ, routing it through layers of review — the paperwork alone could have taken as long as building and shipping the actual product.
Similar stories are playing out across the division.
- One team open-sourced Strands, an AWS software development kit for building AI agents, after a member of Sivasubramanian’s team messaged him at 7 a.m. with the idea. After a quick call with Garman, they decided to go ahead. Within days, it was done.
- Kiro, the AI coding tool, was built by a deliberately small team, using Kiro itself to build it. One engineer prototyped a complex cross-platform notification feature for Kiro that had been estimated at four weeks of work, and shipped it in a day and a half.
- The internal Amazon team that rebuilt the inference engine for the company’s Bedrock platform for AI models did it with six engineers in 76 days, a project originally expected to take 30 developers 12 to 18 months.
Smaller teams everywhere
What’s happening inside Amazon’s agentic AI division is part of a trend across the tech industry toward smaller teams and flatter organizations, driven by AI and agents.
Microsoft’s 2026 Work Trend Index, a survey of 20,000 workers in 10 countries, found that the biggest factor behind AI’s real impact in the workplace isn’t individual skill but whether the organization has restructured around the new technologies.
Vijaye Raji, OpenAI’s CTO of applications, said during a recent Technology Alliance event that the company’s “ambitions are growing faster than we can hire people” — but the profile of who gets hired is changing. OpenAI increasingly looks for engineers who work with AI tools natively, and the gap between those who do and those who don’t is stark: the top engineers at OpenAI use roughly 100 times more AI tokens than the median.
All of this leads to a natural question: what does this mean for jobs? Amazon has cut roughly 30,000 corporate jobs since late 2025 as part of what Jassy has described as an effort to reduce bureaucracy. He has said he expects AI to shrink the corporate workforce over time.
Similar cuts are playing out across the industry, from Meta to Block to LinkedIn, as companies rethink not only the roles they need to fill but also how many people they need overall.
Bigger goals, same team
Sivasubramanian describes the shift differently: In his division, the same number of people are now pursuing a bigger charter. With the new structure, they’re able to take on more projects, and faster, accomplishing things in weeks that would have taken much longer in the past.
The nature of the roles inside those teams is changing, too. Increasingly, product managers write code, and engineers make product decisions. On the Kiro team, for example, a product manager built the first version of a cost analysis dashboard using Kiro itself.
This also requires leaders to operate differently. For example, Sivasubramanian said he is careful to monitor which decisions need his approval, even when traveling. At the current pace, even four or five days of delay can add as much as 10% to a team’s shipping timeline.
Managing these teams also raises new questions. Sivasubramanian said his division has started tracking how much it spends on AI tokens — the basic unit of interaction with an AI model — the way it would track any other operating cost.
So far, the numbers have been manageable: tools like Kiro invest upfront in defining specs and pulling in the right context before generating code, which makes them more efficient with tokens rather than burning through them in aimless back-and-forth.
Even the heaviest users consume only a few thousand dollars a month, he said. But he expects that over time, companies will need a full picture of their operating expenses that includes not just headcount but the cost of the AI agents working alongside them.
This gets to a bigger point: “The bottleneck is not about the time it takes to build something,” Sivasubramanian said. “The bottleneck is about crafting the right specification and the tests and the right product and customer experience.”
In a blog post published last week, Sivasubramanian wrote that teams across the company that restructured their workflows around AI saw a median 4.5x productivity gain, with some exceeding 10x gains. The teams that simply added AI tools to their existing way of working didn’t see the same results.
Coding and testing
That shift has created its own challenges. Teams can generate code faster than ever, but if they don’t define what success looks like up front — the specs, the tests, the edge cases — the agents don’t have as much chance of success.
Amazon is now pushing testing to the moment of coding rather than handling it in stages, so agents can check their own work before anything reaches production.
Sivasubramanian learned this first-hand, the hard way. Earlier this year, jet-lagged and unable to sleep in his hotel room on a trip to India, he decided to try a fun project: He used Kiro to rebuild a piece of AWS infrastructure he’d originally developed by hand nearly 20 years ago — a replication engine that still underpins core services like S3 and DynamoDB.
He and one of Amazon’s earliest distinguished engineers, Allan Vermeulen, had spent four months on the original. Sivasubramanian figured the agent would make quick work of it. Instead, he spent four nights going back and forth, babysitting each step.
On the fifth night, he realized the problem: he hadn’t given the agent the tools to test its own output. Once he wrote the right spec and set up the testing environment, it was done in about two hours. Asked what he did with his rebuilt version of the engine, Sivasubramanian laughed. He never shipped it. “Maybe I should have,” he said.
With the right team and a couple of pizzas, maybe he still can.
Tech
Court Partially Reverses Trump Attacks On Law That Tried To Make Sure Broadband Deployment Isn’t Racist
from the affordable-fiber-optics-is-woke dept
Earlier this year the Trump administration decided to illegally dismantle the 2021 Digital Equity Act, which was intended to help push internet access into long-neglected parts of the U.S. The Act took very vague aim at digital redlining, or the longstanding practice by telecom giants of refusing to upgrade (or at times even timely repair) broadband service in minority and low-income neighborhoods.
Big ISPs like AT&T have long been caught not only refusing to upgrade or repair broadband access in minority areas of cities like Detroit and Cleveland, but charging minority neighborhoods more money for slower service than their less diverse, more affluent counterparts.
Here’s the thing: the Digital Equity Act barely mentions race; it simply included some vague language stating that deployments and broadband grants must be even and non-discriminatory. The law identified minority status as one of eight nonexclusive indicators of barriers to digital access, while separately prohibiting discrimination in programs receiving funds.
As it has done with numerous other programs of this kind aimed at lowering broadband bills, the Trump administration clumsily — and quite illegally — tried to dismantle the whole law last year, insisting it was somehow racist against white people.
Last week, the DC District Court issued a ruling that allowed the Act to survive, but stripped out the already modest race-based components of the law, declaring them unconstitutional.
Groups like the National Digital Inclusion Alliance, which had done a lot of good studies on broadband redlining, celebrated the decidedly mixed bag:
“We are proud to have pushed to keep the Digital Equity Competitive Grant Program alive. This crucial program provides communities across the country not just with access or technology, but the skills, confidence, and pathways necessary to fully participate and thrive in our digital age. We fundamentally object to the government’s position that empowering Black and Brown communities is unconstitutional.“
So the competitive grant program at the heart of the law will continue, but there’s no real consensus on what that will look like or how helpful it will be under a federal government too racist and corrupt to function in the public interest. And there’s not much left to address the very real issue of broadband digital discrimination, which runs parallel with racial discrimination in other U.S. infrastructure sectors like energy.
The Infrastructure Act not only featured $42.5 billion to expand broadband access, it featured a lot of included (and adjacent) legislation intending to address racism in broadband and broadband affordability more generally. Most of that’s been brutally stripped away by the Trump administration, which is instead funneling billions of dollars to Elon Musk for costly Starlink service, then declaring the problem solved.
It’s a lovely bundle of corruption, racism, and regulatory/court capture all thrown into a stew by a bunch of zealots keen to pretend they’re engaging in policy reform and serious legal analysis.
Filed Under: broadband, digital discrimination, digital equite act, fiber, illegal, racism, redlining, ruling, telecom
Tech
Tech sector pours $1T into AI and sends customers the bill
AI AND ML
History’s biggest infrastructure build-out is pushing up hardware and software prices, analyst says
Spending on AI infrastructure is pushing tech sector expenditure to historic levels, and enterprise customers are already footing the bill through higher software and hardware prices.
John-David Lovelock, Distinguished VP Analyst at Gartner, told The Register that tech companies’ own technology spending already amounted to around $1 trillion and was set to grow by 34.7 percent in 2026.
The colossal splurge is driving global sales, leading Gartner to raise its 2026 estimates to $6.37 trillion, a surge of 14.2 percent year-on-year. That’s up from April‘s forecast of $6.31 trillion and February’s $6.15 trillion.
Lovelock said overall growth was accelerating, although tech spending was moving at three different speeds. Devices, which include consumer purchases as well as business laptops, are set to grow by 9.8 percent. However, a significant chunk of that increase comes from higher prices as memory and chips become more expensive. Services and telecoms had lower growth, at 5.3 percent and 4.4 percent respectively.
Infrastructure as a service – one segment of cloud computing – is on pace to grow by 29.3 percent this year to reach $287 billion. In 2025, the market grew by 25.3 percent, Gartner said.
Much of the acceleration is being driven by technology companies equipping datacenters to provide capacity for the expected AI boom. Gartner’s spending figures exclude the buildings themselves and their cooling systems.
Lovelock told us: “The AI infrastructure build-out is the largest infrastructure project humanity has ever undertaken. Bigger than the US highways, bigger than European rail, bigger than the Great Wall of China, and the International Space Station combined.
“That’s how big this sucker is. It is transformational in that sense. We are shifting from a world where we spend on information technology to a world where we’re going to spend on intelligence technology. And right now, you can have your head in the sand and try and avoid that reality, but it’s coming.”
As enterprise software companies embed AI into their products and partner with foundation model builders such as OpenAI and Anthropic, organizations buying IT are concerned about price increases.
“CIOs are extremely concerned about price increases coming at them from all of their vendors, and they are pushing back hard in every area where they can. But the only place that they’re being successful is in the IT services area, where when a service provider adds AI to their product offering, the service provider is rewarded with a lower price point from their customers,” Lovelock said.
There were also unanswered questions about whether the market can sustain the increases, or whether higher prices are a defensive move by vendors trying to protect their market share. For example, by adding AI model Gemini to a search engine, it could be argued Google is defending its dominant position in the market from the threat of AI, as opposed to gaining new revenue.
There is also the problem of users trying to manage AI costs in response to price increases from model builders, several of whom have switched from capped subscription to usage-based billing.
Lower-cost models are coming onto the market from China, while developers are looking to use open source models where appropriate to curb their use of proprietary foundation models.
Whether the price crunch will leave the tech industry able to continue paying for its AI infrastructure building program is “the big open question,” Lovelock said. “But it’s not being investigated well or answered incredibly well.” ®
Tech
Microsoft Defender for Endpoint leaves some Linux boxes defenseless after update
PATCHES
One bug disabled the security service on restart, another blocked installation on hardened RHEL systems
Not content with broken Windows updates, Microsoft has disclosed two problems with Defender for Endpoint on Linux – one that could disable the security service after a reboot, and another that prevents updates on FIPS-enabled Red Hat Enterprise Linux 8 and 9.
The more serious problem affected versions 101.26042.0000 through 101.26042.0009 across all supported Linux operating systems. After an upgrade or reinstall followed by a reboot, “the Defender service might be disabled on some devices,” according to Microsoft.
“If you use Defender for Servers (Plan 1 or 2) with Defender for Cloud and have the MDE [Microsoft Defender Endpoint] integration enabled, automatic updates for the MDE.Linux extension are enabled by default, which means your machines could have received an affected version automatically,” it explained.
“If an affected version was installed, the issue might impact active protection on rebooted devices until remediation steps are taken.”
Microsoft did not specify what caused Defender to become disabled, but anything that could knock out endpoint protection will give administrators sweaty palms.
A separate problem affected RHEL 8 and 9 systems running in FIPS mode: the 101.26042.x update could fail to install, leaving devices on their previous version. FIPS refers to US Federal Information Processing Standards, which in this context impose requirements on the cryptography used by government and other regulated systems.
Although Microsoft’s alert did not mention an available update, its release notes direct users affected by the disabled-service bug to build 101.26042.0011. The separate FIPS installation problem is fixed in version 101.26052.0011 and later.
Microsoft Defender for Endpoint on Linux protects server workloads on-premises and in the cloud. According to Microsoft, “it helps you prevent, detect, investigate, and respond to advanced threats with unified visibility through the Microsoft Defender portal.”
Other endpoint security platforms are available, but where an organization has gone all-in with Microsoft, the unified management offered by Defender for Endpoint on Linux can be difficult to resist.
Microsoft has an unfortunate habit of shipping broken updates for its flagship operating system, Windows. An update that breaks software specifically designed to protect a device takes things to another level, particularly given the relentless rise in attacks and the need to both fend them off and monitor activity. Hence the appeal of unified visibility through the Microsoft Defender portal.
However, an update that could leave Defender disabled after a reboot – while also refusing to install on some security-hardened systems – is less than ideal. ®
Tech
How Much Does An Alternator Replacement Cost? Here’s What To Know
An alternator is an important part that enables you to power your car’s electrical system and maintains the batteries charge. While the battery stores enough of a charge to get the vehicle’s engine running, it’s not doing much once the vehicle is in motion.
The alternator, which gets its power from the engine, runs systems like your dashboard lights, power windows, and even electric power steering. So, you can’t operate a car without a healthy alternator (at least not for long). There are ways to tell if your car’s alternator is bad, especially if your electronic components start flickering or dimming, and this isn’t something to ignore.
An alternator typically lasts up to a decade, though there are ways to extend their lifespan, so, older vehicles will eventually need а new alternator. The cost of a new alternator depends on several factors, such as your vehicle model and your mechanic’s labor rate. According to J.D. Power, typical prices fall between $100 and $350, not including installation.
If you include the cost of the part along with installation, you’re looking at $230 on the low end and north of a $1,000 on the upper end. In addition, since the alternator is belt driven, you may need to invest in a new belt, pulley, and tensioner, which can run anywhere from around $30 up to $300. Of course, you can always save money by making the repair yourself.
Is replacing an alternator a challenging do-it-yourself project?
If you have some familiarity around vehicle engines, replacing an alternator is a fairly straightforward job. You’ll need some basic tools like a socket set in order to remove the bolts that hold the part in place. However, depending on the engine configuration, the ease of accessing it can vary. Before getting anywhere near the alternator, you must first disconnect your car’s battery, which helps protect the electrical system and prevents you from getting shocked.
Next, take a picture of the belt going around the alternator so you’ll have an idea of how it wraps around the various pulleys. Remove the belt going around the alternator by loosening the tensioner, which holds it taut. You can find various instructional clips showing this process, such as O’Reilly Auto Parts’ YouTube video, “How To: Replace an Alternator.”
Once the belt is lifted off the pulley on the alternator, you have two different connections to remove; a plug which regulates voltage, and a positive cable, which is held in place by a nut. Finally, there are a few bolts holding the alternator in place that will need to be removed. Once you extract the old part, simply install the new alternator and follow these steps in reverse.
Can you repair an alternator instead of replace it?
You can choose to have an alternator fixed rather than replace it with a new unit, but it only makes sense in certain situations. The age of the part, complexity of the repair and your skills are the determining price factors when opting whether to fix or replace it. If the alternator is only a few years old for instance, it might be worth fixing yourself, especially if you want to save some money.
Current must flow only one way in an alternator, and this is accomplished with diodes, which can sometimes develop faults. This is something you can check using a multimeter as demonstrated by the maddoxmechanic YouTube short called, “Quick Check: How to Test Alternator Diodes with a Multimeter.” The diodes themselves typically run around just under $50, but they do require you to take apart the alternator, which demands more skill. Optionally, you can also find alternator rebuild kits, which can be between $28 and $60.
However, if you’re not comfortable with more in-depth automotive tasks or are short on time, having a mechanic do the repair work might not save much money at all. This is especially true if the part is old and worn, which would be better to replace rather than salvage. Labor costs for replacing the diodes could run between $150 and $400, per Partcatalog.com.
Tech
How Many Lumens Do You Need For An Outdoor Projector?
A backyard movie night can be a great way to get family and friends together over the weekend. However, the experience can also be an exercise in frustration if the picture is difficult to make out due to ambient light pollution. When picking out the best possible portable projector for your needs, factors like resolution, smart features, and audio quality may top your list. But you shouldn’t neglect brightness, which is measured in lumens.
In terms of projectors specifically, ANSI lumens are used, as they provide a more precise evaluation of brightness. One ANSI lumen is around the strength of 2.4 lumens and measures brightness from nine separate points in order to arrive at an overall figure. Conversely, lumens are measured from a single point, the source of the light.
So, how bright does your projector need to be for a well-defined picture that isn’t washed out? Typically, a unit rated above 2,000 ANSI lumens should be sufficient in all but the most challenging of scenarios. The brighter the projector, the better the experience can be across a wide range of conditions. But unfortunately, more ANSI lumens can mean higher costs as well, so you’ll need to consider the best options within your budget.
Can outdoor projectors work in daylight?
Unfortunately, forcing a projector to compete with the Sun isn’t a recipe for success. According to LBC, the Sun puts out an estimated 6,840,000,000,000,000,000,000,000,000 or 6.84 octillion lumens. For context, the number of lumens you should look for in a camping flashlight is between 300 and 1,000. So, as you may imagine, there isn’t a projector on Earth that can overpower the brightness of the Sun, meaning that watching movies outdoors during the day will be difficult or even unfeasible in some situations.
However, if you are determined to set up a projector during the day, there are ways to help improve your chances of achieving a watchable picture. Of course, a projector with a rating north of 3,500 ANSI lumens can help, but contrast is also a vital component. How intensely a projector can produce bright white colors versus dark blacks is measured using contrast ratio. The higher the number on the left, the better, especially in daylight, making a projector with something above a 6,000:1 contrast ratio desirable.
While the projector itself is a factor in terms of improving daylight viewing, it can’t do all the work. Investing in a quality reflective screen can also help keep your projector’s image looking bright and sharp. Other considerations include things like the time of day, as viewing experiences are often better around sunset. Strategically using your yard’s available shade can help, too. Weather, is another variable, with a cloudy day providing a clearer viewing experience than a sunny one.
Tech
Dopl raises $6.3M to bring remote robotic ultrasounds to rural patients as it pursues FDA clearance

Dopl Technologies announced $6.3 million in new funding to support the development and commercialization of a remotely operated robotic ultrasound system. The Bothell, Wash.-based startup has now raised more than $8 million.
Dopl is pursuing FDA clearance for its platform, which pairs off-the-shelf ultrasound probes with commercially available robots. The system facilitates communication between a remotely based ultrasound technician (or sonographer) and the robot, allowing the sonographer to conduct the exam with haptic feedback.
The platform also includes a video connection to allow conversations between the sonographer, patient and onsite health personnel.
The goal, said CEO Ryan James, is to target underserved and typically rural communities “to improve access to care and the time it takes for patients to get that care.”

While Dopl could eventually create remote interfaces for a variety of health services, ultrasound has widespread applications.
“People say that ultrasound is the new stethoscope,” James said. While perhaps best known for its use in obstetric exams during pregnancy, the diagnostic tool is also employed for cardiac and vascular exams and can be used to detect stroke, cancer, heart disease and other conditions.
The startup was co-founded by James, Chief Operating Officer Steve Seslar, and Chief Medical Officer Wayne Monsky, who began researching novel care delivery methods together at the University of Washington in 2017.
The system pulls together a suite of technologies including fine-motor robotic control, haptics, computer vision, AI, and advanced networking.
The new investment allows the company to hire staff to help it navigate the FDA approval process. Dopl is conducting clinical evaluations of its platform and aims to submit an application for FDA clearance next year.
Dopl currently provides in-person ultrasounds using its technology through traveling sonographer services. FDA clearance would allow it to perform remote robotic exams and scale its platform.
The startup has a letter of intent with a rural health care collaborative that includes 31 critical access hospitals in Washington and has partnerships across the nation, including on the East Coast.
The company also expects to take advantage of opportunities through the Rural Health Transformation Program, a $50 billion federal effort that will distribute funds in every state from fiscal years 2026 to 2030. The program is targeting innovations that expand access to care.
The seed round was led by SpringTide Ventures, which was joined by WRF Capital, Tacoma Venture Fund (pre-seed round lead), HeartX, Transform Health Ventures, Precursor Ventures, and additional early-stage institutional investors.
Dopl was a finalist for startup of the year at the 2026 GeekWire Awards.
There is competition in this market, including traditional ultrasound companies and startups. A notable rival is AdEchoTech, which launched in France in 2008 and has since received FDA clearance to sell in the U.S.
James said Dopl’s haptic technology sets it apart, allowing remote technicians”to actually feel the patient as they’re scanning.”
The remote robotic care delivery industry is rapidly growing, James said, and “represents the next wave of care delivery and shift in mindset in terms of how clinicians operate.”
Tech
What Was The Top Speed Of The Original 1965 Ford Mustang?
Answering the question of the top speed of the original Ford Mustang that was made in 1965 is tricky — it depends. For our purposes here, we will use the highest possible top speed that was produced by the highest-performing Ford-produced Mustang of this particular year. We are not including Shelby Mustangs, which were initially made for racing and were not production-line Mustangs. Instead, we will focus on the famous K-code Mustang, about which we have covered how to identify one and what one’s worth today.
The K-code Mustang was available between 1964 1/2 and 1967. The K-code engine was a high-performance 289 cubic-inch V8 that produced 271 horsepower, thanks to upgrades to its four-barrel carburetor, cylinder heads, pistons, connecting rods, and lifters. A handling package and dual red-stripe tires were also included, with a fender badge proclaiming “High Performance 289.” A four-speed manual transmission was required. In addition, you couldn’t get air conditioning or power steering on a K-code, while the car’s warranty was shrunken to three months or 4,000 miles. Ford seemed to know exactly who the buyers of these “hi-po” Mustangs would be, which had a lot to do with what made these early Ford Mustangs so iconic.
The performance of the Mustang with the K-code engine was revealed in a road test contained in a 1965 special Mustang edition of Car Life magazine. It did 0-60 mph in 8.3 seconds, made it through the quarter-mile in 15.9 seconds at 85 mph, and hit a top speed of 120 mph at 6500 rpm in top gear.
What else should you know about the 1965 Mustang?
The 1965 Ford Mustang was a runaway sales success, with a total of 559,451 sold that year alone, or 680,989 if you include the 1964 1/2 models that are considered to be 1965s. The Mustang was initially introduced at the 1964 New York City World’s Fair, while Ford also ran Mustang commercials on all three TV networks, reaching 29 million viewers with the Mustang message. Over four million people visited Ford showrooms during the first weekend, while 22,000 placed orders for a Mustang. Four months later, sales surpassed 100,000. It was March 2, 1966, when one million Mustangs had been produced, with a retail value of $2.8 billion. Ford had a winner on its hands.
The Ford Mustang was an iconic vehicle for the Baby Boomers, who were just reaching college age. This was also a time when many families were moving to suburban areas, where they found a need for an extra car. This “youth market” had a set of needs and desires that were unique, yet they did not have much cash, which drove Ford to keep the Mustang’s price low. The Mustang prototype that Ford developed to gauge public reaction was intended to be, “…demure enough for church-going, racy enough for the dragstrip, modish enough for the country club.” While the 1965 Mustang represents the car’s first iteration, we have done a deep dive into every generation of the iconic pony car. The Ford Mustang made history back in 1965, and continues today as the only “pony car” that has been in continuous production since it debuted.
Tech
The 10 Best WIRED-Tested Handheld Vacuums of 2026
Compare Our Picks
Others We Tested
Photograph: Molly Higgins
Riccar Gem Handheld Vacuum With Tools for $70: This model is very similar to the Simplicity corded model above, and although it’s portable and handheld, this big boy needs to be plugged in for power. I like how basic it is: It just has a top switch for on/off, and a button on top to release the dust canister, which is at the front of this model rather than behind. Once disengaged, you just need to pull up the semi-spherical fabric mesh filter and dump out debris. Unfortunately, lots of fur and dust get stuck to the filter, so you’ll need to make sure you wipe it with a semi-damp cloth to clean. It comes with several attachments, including a bendy hose, long crevice nozzle with detachable brush, upholstery brush, and dusting brush. Instead of twisting or snapping, each attachment fits snugly into the opening, making it easier for those with fine-motor issues. I liked that the long, flexible hose reached easily into tight spaces like between my counter and fridge and behind my automatic litter box. The corded machine is loud and gets hot quickly due to its sheer power. While I don’t think this is the best handheld vac for everyone, it’s a solid option with lots of power.
Photograph: Molly Higgins
Tineco Go Mini Cordless Hand Vacuum for $200: A lightweight vacuum that’s fairly straightforward and charges through the end of the handle, the Tineco Go must charge for five to six hours before use, which gets you 10 to 20 minutes of use time. The main things I wanted to focus on cleaning in my everyday life were cat litter and pet hair. This vacuum spit out roughly 80 percent of the litter I tried sucking up, and the pet hair just sort of rolled around on the carpet rather than being sucked back into the vac. (I’m not sure if it was the angle or shape of the mouth that aided in both of those problems.) The suction wasn’t as powerful as others tested, and it tended to work best on hard surfaces. This vac wasn’t bad, per se, but others on this list at similar price points worked a lot more effectively on a lot more various surfaces (with multiple head attachments) than this was seemingly capable of.
Photograph: Molly Higgins
Bissell AeroSlim Handheld Vacuum for $45: Not the most powerful I’ve tested, but it’s lightweight and super easy to maneuver. It comes with a multi-accessory tool that combines a long, slim crevice tool with a moveable brush (the brush moves up and down on the tool depending on need), but you can also just use the vac without the slim attachment for more general surface cleaning. The tiny 0.1-liter front of the vac twists off to empty, and a small dual filter can also be removed for cleaning/replacement. When testing, I wasn’t able to successfully pick up bigger items like wood shards or litter, and only picked up a fraction of pet hair as some of the more powerful models with a roll brush, but I wasn’t expecting it to—that’s not this cutie’s job. It’s light, simple, and easy to use—making it a great lightweight portable option (just don’t expect it to handle bigger messes).
Photograph: Molly Higgins
Hoto AutoCare Air Duster & Vacuum for $60: Looking for a lightweight (0.78 pounds) vacuum that can also blow debris/dust, inflate and deflate, and vacuum-seal? The Hoto may be a fit. It has tons of attachments: a pointed crevice nozzle, dust brush, dust cup nozzle, air nozzle, and swim ring nozzle. It can hold a little over 3 ounces, has 20,000 Pa suction, and each attachment easily twists off and on. I found that the grooming brush head tended to push around fur/hair rather than sucking it up. Like many gadgets we test here at WIRED, this one fell into the pitfall of trying to do a lot of things, rather than doing any one thing really well. This is a great deal for a device with so many capabilities (if that’s what you need), but if you’re just looking for a really great handheld vacuum, there are others that are more effective.
Photograph: Nena Farrell
Black and Decker Dustbuster Flex for $74: WIRED reviewer Nena Farrell thought this cordless handheld vacuum was great for cars and hard-to-reach spots because of its 4-foot hose. It has a crevice tool and pet hair brush attachments for a variety of messes, and runs on a 20-volt lithium-ion battery that delivers high suction power. Plus, it has a handy charging mount that the accessories clip into. However, we aren’t fans of the short 15-minute battery life.
FAQs
I factored in charge and actual power time in overall value and ranking, considering some charge for four hours and last less than half an hour of run time—that just won’t be practical for certain people. Since most of these come with attachments for specific needs, I made sure to test each on a myriad of surfaces, like rugs, carpet, hardwood, stairs, counters, crevices, inside my car, on furniture, and even in between vents and blinds. I also tested on various types of messes (no liquids though), especially focusing on my personal major mess gripes—cat litter and pet hair.
Definitely pay attention to specs like charge and usage time, as well as power (like voltage) and attachments provided. If you’re working construction and need a handheld vac for debris, I’d recommend something like a Ryobi, while if you’re looking for a portable lightweight travel vacuum, the AeroSlim might be better suited for you. It’s also important to factor in charging versus usage time. Some of these last less than 15 minutes and have very small waste capacities, so if you have a bigger project, you’ll need to take that into consideration.
Power up with unlimited access to WIRED. Get best-in-class reporting and exclusive subscriber content that’s too important to ignore. Subscribe Today.
Tech
This dentist built a S$500M dental empire from one clinic. Now, Q&M is betting S$146M on Asia.
Q&M Dental Group is eyeing 300 clinics with a S$146M expansion plan
If you’ve walked around Singapore long enough, you’ve probably seen one: a neighbourhood dental clinic with bright green lettering and a giant tooth logo.
Chances are, you never thought much about it. But behind those familiar storefronts sits a SGX-listed company worth over S$500 million, operating 110 clinics across Singapore and now pursuing acquisitions in Australia, Thailand, China, and even of another local rival.
That company is Q&M Dental Group—and after nearly three decades of quietly dominating Singapore’s heartlands, it’s now betting big on becoming a regional dental giant.
Built for the heartlands


Q&M was established at Bukit Batok in Nov 1996 by Dr Ng Chin Siau, an NUS dental surgery graduate who had worked as a Dental Officer with the Ministry of Health before joining a private clinic at Bukit Batok—and eventually buying it over after selling his three-room Clementi flat for S$180,000.
From the outset, Dr Ng had little interest in building a premium dental brand. Instead, he focused on something far more scalable: providing affordable dental care to everyday Singaporeans.
That philosophy was even reflected in the company’s original Chinese name, “全民” (Quan Min), meaning “for all the people.”
Demand grew quickly and outpaced what one dentist could handle.
By 1998, fellow dentist Dr Ng Jet Wei had joined the practice. A year later, together with Dr Chong Kai Guan, the trio had opened another four clinics. By 2000, Q&M was already operating 10 clinics with 20 dentists.
Rather than chasing prime shopping malls or affluent districts, Q&M planted its clinics where Singaporeans actually lived—in HDB estates, neighbourhood centres, and suburban malls.
The strategy was simple: keep prices accessible, expand steadily, and acquire smaller dental practices whose dentists and patient bases came as part of the deal.
Along the way, Q&M also built supporting businesses, including its own dental laboratory to produce crowns, bridges, and dentures in-house, giving it greater control over costs and operations.
By the time the company listed on the SGX Mainboard in 2009, it already had a proven expansion playbook. The IPO simply gave it the capital to execute it at a much larger scale.
Building the Q&M empire


In the years that followed the IPO, Q&M expanded aggressively through acquisitions, and revenue climbed from roughly S$60 million to S$155 million by 2016.
By then, its outlet count had grown to 76 clinics across Asia—a number that would nearly double over the following decade.
This brought Q&M Dental Group’s market capitalisation to S$557.6 million that year.
Q&M’s growth has not been limited to opening more outlets. Alongside the clinic network, the group has built a set of adjacent capabilities that reinforce the core business.
In 2019, it launched the Q&M College of Dentistry at City Square Mall—Singapore’s first private postgraduate diploma programme in clinical dentistry, designed to train general dentists with an expanded range of advanced skills. The college earned a full four-year EduTrust certification from Feb 2026, allowing it to enrol international students.
This is a crucial move with Singapore’s dentist supply constraint. Singapore’s only domestic dental school graduates around 50 to 60 dentists per year, and foreign-trained dentists must pass a rigorous licensing exam before practising. Any competitor trying to build a rival chain faces the same thin talent pool.
Q&M’s training college gives it a head start in sourcing and retaining the practitioners it needs to keep opening clinics, a structural advantage that competitors cannot easily replicate.


When COVID-19 hit, Q&M pivoted part of its operations, acquiring Acumen Diagnostics to distribute test kits and run laboratory PCR testing in 2021—a business that briefly became a significant revenue contributor before demand evaporated as the pandemic receded.
As pandemic revenues evaporated, results retreated sharply in 2022 and 2023. But the core dental business proved its resilience: core healthcare EBITDA held steady at around S$37 to S$40 million throughout, even as the group-level numbers were distorted. By FY2025, the group had returned to profit, recording a16% growth.
In 2023, it opened a Free Dental Clinic at Chai Chee Road, offering essential dental treatment at no cost to underprivileged patients. While charitable in nature, the initiative also reinforces Q&M’s standing in the communities where its business is most deeply rooted.
More recently, Q&M has turned its attention to technology.
In 2024, the group invested in EM2AI, a dental technology firm developing AI-powered diagnostic and treatment-planning tools, including a cloud-based practice management system called EM2Clinic.
The tools are designed to reduce the administrative burden on dentists and standardise clinical workflows, which is useful as Q&M grows toward a network where consistency across hundreds of clinics matters more than before.
Why scale matters


With 110 clinics, Q&M operates roughly 8% of Singapore’s nearly 1,300 licensed dental clinics. According to the company, it also serves around one-third of all patients who visit private dental groups.
That scale creates advantages that go beyond simply having more outlets.
As Singapore’s largest private dental chain, Q&M has greater bargaining power with landlords and suppliers, while its size also allows it to spread fixed costs across a much larger network.
Government healthcare policies could further strengthen that position.
The enhanced Community Health Assist Scheme (CHAS) subsidies introduced in Oct 2025 expanded coverage for restorative dental procedures and extended eligibility to 1.7 million cardholders. As the country’s largest CHAS-accredited private dental chain, Q&M stands to benefit more than smaller competitors. Analysts estimated the changes lifted the group’s revenue by around 3% in the second half of FY2025.
More support is also on the way. From mid-2026, seniors will be able to use up to S$400 a year from Flexi-MediSave for dental treatment at CHAS clinics. The policy aligns neatly with Singapore’s ageing population, with one in four residents expected to be aged 65 or older by 2030. As dental needs typically increase with age—and many seniors live in the heartland estates where Q&M has built its network—the demographic trend could provide another long-term tailwind for the group.
Meanwhile, Q&M continues to consolidate its position. In Mar, the group announced plans to fully acquire an unnamed Singapore dental chain, backed by a profit guarantee of up to S$34 million over five years. The deal would further strengthen its presence in its home market even as it looks overseas for growth.
Now betting big on the region


For most of its history, Q&M’s overseas ambitions effectively stopped at the Causeway.
Over the years, the group steadily expanded its presence in Malaysia, where it now operates 38 dental clinics alongside a dental supplies and equipment distribution business. Beyond that, however, its growth remained largely concentrated in Singapore.
That is now changing.
In recent months, Q&M has unveiled plans to enter three markets almost simultaneously, marking the group’s most ambitious expansion programme to date
The biggest move is Australia. In Jul 2026, Q&M signed binding agreements to acquire Experteeth Group for A$119.64 million (S$107.83 million). The deal would add 40 clinics and around 120 dentists across New South Wales, Victoria, Queensland, Tasmania, and the Australian Capital Territory, making it Q&M’s largest acquisition ever and its first entry into a market outside Asia.
In Thailand, it signed a deal to acquire a group of 30-plus clinics focused on cosmetic and aesthetic dentistry, a sector closely tied to the country’s medical tourism industry.
In total, the acquisitions in both countries will amount to a combined US$113.2 million (S$146.26 million) investment to build a pan-Asian dental company.
China forms the third pillar of Q&M’s expansion strategy. The group also owns Chinese dental operator Aoxin Q&M, which it has now fully consolidated as a subsidiary. It plans to use the business as a platform to acquire dental chains in southern China, expanding beyond Aoxin’s traditional base in the country’s northeast.
If all three acquisitions close, Q&M’s footprint could grow from over 160 dental outlets today to more than 300 within two years.
Not every growth initiative, however, has gone according to plan.
In Apr 2025, the group also proposed a secondary listing on Bursa Malaysia, which would have given Malaysian investors direct access to the stock and strengthened Q&M’s capital markets presence in its second-largest operating market. The plan was later shelved, with the company citing prevailing market conditions.
What could go wrong?


Three concurrent acquisitions in markets Q&M has limited or no experience operating in are an ambitious programme for a management team whose track record has been built almost entirely in Singapore and Malaysia.
China is the cautionary tale here. Q&M first entered China back in 2013, and later spun that business off as a separately listed company, Aoxin Q&M, in 2017. Aoxin broke even for its first couple of years, then racked up losses for several years running—around RMB142 million (S$27.09 million) in total losses from 2020 to 2024—as it battled COVID-19 shutdowns and China’s strict lockdown policies.
It only turned a profit again in 2025, roughly six years after the losses began.
Now Q&M has folded Aoxin fully into its own accounts, and in doing so has added S$77.0 million of goodwill to its balance sheet—essentially the premium it’s paying on the bet that these businesses will earn enough in future to justify the price.
If they don’t, that goodwill may eventually need to be written down, hitting profits the same way it did for Aoxin’s own past investments. Australia and Thailand are brand-new markets for Q&M with no comparable track record to lean on, so if that expansion underperforms, the money put into it may not pay off.


Dental practices are also relationship-driven businesses, where patients follow their dentist, not the brand. When a chain acquires a clinic, the real asset is the practitioners inside it.
If key dentists leave post-acquisition and take their patient books with them, the acquired revenue can evaporate quickly.
Q&M has tried to address this through 15-year service agreements and long lock-up equity arrangements with acquired dentists, but such structures have never been tested at the scale Q&M is now attempting across four markets simultaneously.
Then there is the Johor-Singapore dynamic complicating profits further. The Rapid Transit System Link scheduled to open in Dec 2026, will cut the Woodlands North to Bukit Chagar crossing to around five minutes—at a fare of roughly S$5 to S$7.
Singaporeans already save 50 to 65% on identical dental procedures by crossing the Causeway, with more than 100,000 estimated to make the trip annually. A single dental implant costs S$4,500 to S$6,500 in Singapore versus S$1,200 to S$2,000 in Johor Bahru.
When the RTS removes the main friction, being the border queue, from that equation, the maths for price-sensitive patients shifts meaningfully.
Q&M’s 38 Malaysian clinics, including in Johor, might mean some of that outflow stays within the group. But how much of Q&M’s dental revenue holds up once the crossing becomes as easy as taking the MRT is a question nobody can fully answer yet.
From a single clinic in Bukit Batok to a listed group eyeing four countries simultaneously, Q&M’s story is, in many ways, a study in patience. Three decades of unglamorous, heartland dentistry have built toward a moment where the company is finally ready to bet big on becoming Singapore’s dominant private dental chain in the region. Whether the bet pays off will be the next chapter.
- Learn more about Q&M Dental Group here.
- Read other articles we’ve written on Singaporean businesses here.
Also Read: The price of going regional: Raffles Medical’s S$600M bet is still struggling to pay off
Featured Image Credit: National University of Singapore, Velocity @ Novena Square
Tech
Google Chrome could soon let you translate pages right inside Reading Mode
Chrome‘s Reading Mode just got a new button tucked inside its settings menu. Google is quietly testing a translate option for the feature on Chrome Canary, the browser’s experimental testing channel where new ideas show up long before they’re ready for everyday use.
What does this new translate button actually do?
The option was first spotted by @Leopeva64, who shared a video demo on X highlighting the new button tucked inside Reading Mode’s dropdown menu. It exists alongside the controls for theme, font, images, and links, though it doesn’t actually do anything yet. Once this feature is live, clicking it would translate whatever page or PDF you’re reading into your preferred language, without leaving Reading Mode.
For now, clicking the button does absolutely nothing. Google appears to have added the interface piece first, well before building out the actual translation function behind it. If you need to translate a page today, the existing Google Translate browser extension already covers the same job just fine.
Is it rolling out to everyone on Canary?
Not quite yet, it seems. PiunikaWeb tested this on the latest Chrome Canary build, but the translate button was nowhere to be found. That mismatch suggests Google might be testing this in limited batches rather than pushing it to every Canary user simultaneously, or it could be a platform-specific rollout.
In the last few months, we have seen a bunch of other Chrome upgrades. Google overhauled Reading Mode into a fullscreen, distraction-free view earlier this year, and rolled out vertical tabs alongside it. The company also added split view, PDF annotations, and direct Google Drive saving. Separately, Chrome is working on native lazy loading support for video and audio, a change that could speed up page loads across the board once it ships.
-
Fashion3 days agoWeekend Open Thread: Brooks Brothers
-
Crypto World6 days agoGrayscale Files For Worldcoin ETF, WLD Registers Sharp Rise
-
NewsBeat6 days agoHow a former Blue Peter presenter stunned America’s Got Talent judges
-
Tech15 hours agoIntel is reversing course and bringing hyper-threading back to its server chips
-
Tech7 days agoSail Virtually Aboard The “Itanic” With IA-64 Emulator
-
Tech6 days ago
Turtle Beach Command Series KB7 review: a nifty screen-equipped gaming keyboard
-
Business5 days agoNew Jersey voter registration controversy explained: How 6,600 noncitizens got on the rolls, and what happens next
-
Entertainment6 days agoJohnny Depp’s R-Rated Gothic Cult Classic Gets New Release Ahead of Sydney Sweeney Remake
-
Crypto World5 days agoEthics, other provisions in crypto Clarity Act to be further discussed
-
Politics4 hours agoLuke Littler dismantles Gerwyn Price to retain title in Blackpool
-
NewsBeat7 days agoShanghai science forum photos show China’s AI and robotics advances in rivalry with US
-
Sports18 hours agoCommonwealth Games boxing: Jadumani Singh seals dominant 5-0 win over Pakistan’s Sumama Rehman to enter quarter-finals | Commonwealth Games News
-
Sports4 days ago2026 3M Open leaderboard: Scottie Scheffler finds putter in Round 1, sits three back
-
News Videos4 days agoThe Peugeot Family: How 200 Years of an “Old Money” Dynasty Died in A Boardroom
-
Politics1 day agoSpain sweeps the board at 2026 World Cup with individual awards
-
Fashion3 days ago16 Dresses for the High Summer Event
-
Crypto World7 days agoAndrew Cuomo joins OKX board as crypto exchange expands in U.S.
-
Entertainment3 days agoA New Post-Apocalyptic Gundam Anime Series Blasts Into SDCC
-
News Videos18 hours agoBITCOIN JUST ENTERED THIS CRITICAL ZONE…
-
Crypto World1 day agoXRP Ledger adds $2.6B as RWA inflows rank second






You must be logged in to post a comment Login