When you’re driving your car, you’re probably regularly looking at the speedometer to make sure you comply with the local speed limits. The method by which it works is simple enough: the rotation of the wheels is sent mechanically via a cable to a dial on the dash, or an electronic sensor counts the rotations of the drivetrain and an electronically-controlled needle or display shows the speed.

But what about if you were in an aircraft, and the wheels had nothing to do with how fast you were going? How would you even begin to measure speed? There are two ways: there’s a convenient solution to this problem rooted in simple fluid mechanics, and a far-more-complex modern solution. Today, we’ll explore how planes and helicopters are able to figure out how fast they’re going, by the old ways and the new.

Classical Methods

A key thing most aviators want to know is how fast their aircraft is going. Specifically, it’s nice to know how fast it’s moving relative to the airstream around it, which is referred to as airspeed. This is important, because it’s the aircraft’s velocity relative to the flow, such as wind, that determines the performance of the airfoils, how much lift is generated, and whether or not the aircraft is approaching a stall condition where it might fall out of the sky.

Measuring airspeed is most commonly achieved with the use of a device called a Pitot tube. The pitot tube is a tube with a hole in one end that points directly into the airflow in the direction of travel of the aircraft.

As air flows in, it reaches a dead end and the flow slows to a stop, or stagnates, since it has nowhere to go. This allows a pressure sensor or a manometer or other device to measure the stagnation pressure at this point. The stagnation pressure measurement is related to the flowspeed of the incoming air since the kinetic energy of the flow is converted to pressure as the flow comes to a halt.

A secondary tube, pointing perpendicular to the airflow, is then used to measure the static pressure of the surrounding air, without the ram effect of the air being forced in by the aircraft’s forward motion. Then, it’s possible to calculate the velocity of the aircraft relative to the airstream by plugging the stagnation pressure and static pressure into a rearranged Bernoulli’s equation.  If the pitot tube and static tube are hooked up to electronic sensors, the airspeed can be calculated electronically, and fed to a display or digital gauge.

Alternatively, it’s possible to effectively do this “calculation” mechanically. In earlier days, static and stagnation pressure captured by each tube would be fed to a gauge. Inside, the stagnation pressure would be fed to a diaphragm which moved due to the difference relative to the static pressure which is fed into the gauge body, and the movement of the diaphragm would, via a simple mechanism, shift the needle on the gauge.

A small General Aviation aircraft might mount a single pitot tube on the aircraft, feeding the air speed instrument in the cockpit. Commercial aircraft might mount two or more for safety’s sake, in case one becomes inoperable, while large airliners may have four or even more to provide a high level of redundancy and error checking. Heaters are commonly included on pitot tubes to ensure they can be kept free of ice, which can otherwise completely block a tube and make it impossible to obtain an airspeed reading.

For pilots, not knowing how fast (or slow) the aircraft is going can be highly dangerous, as it can lead to entering unstable flight regimes such as stall. Thus, it’s imperative that the pitot tubes remain unobstructed and functional for safe flight. Many aircraft accidents have occurred because of blocked or malfunctioning pitot tubes or airspeed instruments.

The New Way

Of course, you could fuss about with pitot tubes and pressure sensors and deicing measures, but that’s all very fiddly and old hat. There is an entirely different way to figure out a plane’s speed, though it’s only been available for the last few decades. It’s as simple as throwing a GNSS receiver on the aircraft.

Yes, whether your particular poison is GPS, Baidou, GLONASS, or Galileo, any major satellite navigation system will be able to tell you the speed of your receiver. Simply measuring the change in the receiver’s position over time is enough to calculate out the speed, and any off-the-shelf receiver will present this information as standard. It’s generally not used as a primary indicator in aircraft, because it reports ground speed, not airspeed, the latter being more relevant for aviation purposes. Still, it can prove to be a useful sense check when traditional airspeed indicators are non-operative or reporting confusing data, and GNSS devices are widely used on many aircraft today.

Flying High

If you’ve ever wondered how an aircraft measures its speed as it floats through the amorphous gas cloud we call an atmosphere, now you know. Even to this day, where electronics and computer wizardry control our fanciest aircraft, airspeed measurements are still done with the same simple physics, just with some fancier sensors for help. The fundamentals haven’t changed at all. Now you know, you can always dig deeper into the many other rich applications of Bernoulli’s equation and fluid mechanics in general. Happy learning.

Meta wants its own AI coding tools. To get there, it is telling its engineers to be careful with the rival tools they lean on today.

Meta has placed strict limits on how engineers in its applied AI division use Anthropic’s Claude Code and OpenAI’s Codex, The Information reported. The worry is inadvertent distillation. One internal memo even told some teams to pause tasks that used the outside tools. It warned that the rivals’ output could seep into Meta’s training data and trigger “serious escalations with partner companies”.

What distillation means here

Distillation is when one model learns from another model’s outputs. A company feeds a strong model’s answers into its own system, and the smaller model picks up the bigger one’s skills. The method is cheap, fast, and legally fraught.

That is the heart of Meta’s problem. The company is building its own coding tool, called MetaCode, to replace Claude Code and Codex. If its engineers rely on those rival tools while shaping the replacement, Meta could end up training on a competitor’s model by accident. That could breach the rivals’ terms of service and hand them a lawsuit.

The bind Meta is in

The situation is awkward. Meta still needs the best coding tools to move fast. For now, the best ones belong to Anthropic and OpenAI. So Meta is asking staff to keep using the very products it wants to leave behind, only with more caution. The rules sit inside its new applied AI engineering division, the unit Meta built to catch up in the model race.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

Cost is the other half of the story. Meta is trying to wean itself off expensive outside coding tools. It is not alone. Amazon is weighing cheaper alternatives after Anthropic raised its prices. The pressure to cut the AI bill is everywhere.

Anthropic keeps gaining leverage

This is the latest sign of Anthropic’s growing clout. Its Claude models have become a default for coders, which gives the company room to push. It recently struck a half-price deal to put Claude across California’s state agencies. It is also winning paying customers at pace.

The flip side is friction with the very firms that depend on it. Anthropic has already accused Alibaba of distilling Claude into a rival model. Meta clearly does not want to be next in line.

Squeezed on every side

Meta’s pinch is not only about Anthropic and OpenAI. Google has capped how much Meta can use its Gemini models for coding and chatbots, Engadget reported, citing a lack of capacity. So Meta faces limits from three rivals at once. It must build its own tools, and fast.

That is a strange place for a company of Meta’s size. It spends billions on AI talent and chips. Yet on coding tools, it still depends on the labs it is racing against. The new rules try to close that gap without tripping a legal wire.

Why it matters

The episode shows how the AI business is maturing. The model makers are no longer just selling access. They are guarding their outputs as prized training data, and they are watching who learns from them.

For Meta, the lesson is sharp. Owning the frontier means more than raw compute and big hires. It means controlling the tools your own engineers use every day. Until Meta’s in-house coding system is ready, it has to borrow from rivals while trying not to copy them. That is a tightrope, and the memos show Meta knows it.

Don’t get hit with the Apple tax.

Apple’s App Store is convenient for managing subscriptions, but it’s not worth paying extra for YouTube Premium

Reyanaska/Shutterstock

I love cordless vacuums, and I don’t just say that because I’m one of CNET’s primary testers of the category. I say it because switching from corded vacuums to cordless vacuums was a big quality-of-life upgrade, thanks to the ease of use and maneuverability around my apartment. The downside is that cordless vacuums don’t usually match the suction power and cleaning capabilities of upright or canister models. 

Shark’s PowerDetect Transformer 3-in-1 is the company’s attempt to address this trade-off without forcing you to buy multiple vacuums.

“The upright vacuum has looked and worked the same way for decades,” said Petra Oman, vice president of marketing at SharkNinja, in a press statement. “We saw an opportunity to rethink the category by eliminating the bulky hose and creating a system that adapts to the way people actually clean. Transformer delivers the deep-cleaning performance consumers expect from an upright, with the flexibility and reach needed to clean everything from floors and carpets to stairs, furniture, ceilings and the car.”

As the name suggests, the PowerDetect Transformer is three vacuums in one. It’s a full-size upright vacuum that’s intended for deep cleaning carpets and hardwood floors with the strongest suction. Shark says you can remove the main canister with one click, and it’ll turn into a slim stick vacuum for regular, lightweight cleaning, getting under furniture and into other tight spots. One more click turns it into a handheld vacuum, making it easier to clean stairs, upholstery, corners and cars. 

In terms of specs, the Transformer will have key features from Shark’s most popular models, including LED lighting to help you find debris and automatic detection of dirt levels, flooring types, edges and movement to automatically adjust suction and cleaning performance. It features anti-tangle brushrolls, odor-neutralizing tech like the Shark Stratos and HEPA filtration. It’ll also come with an auto-emptying system that empties the debris into the main dustbin when the handheld clicks back into place. 

I haven’t had a chance to go hands-on with this vacuum yet, and I’m not entirely sure how the system breaks down. I’ll be testing it both at home and at CNET’s Louisville lab. The most interesting question will be whether the PowerDetect Transformer can truly deliver the cleaning performance of an upright vacuum without compromising elsewhere. 

Price and availability 

The Shark PowerDetect Transformer 3-in-1 will be available on SharkNinja and TikTok Shop for $529. It’ll also come to Amazon, Walmart, Best Buy, Target, Costco and Sam’s Club. 

The Japanese group plans to mass-produce data-centre battery cells in Kansas by fiscal 2028, redirecting a large slice of its AI infrastructure investment toward storage.

The companies that built batteries for electric cars are discovering a new and hungrier customer: the data centre.

Panasonic plans to localise production of data-centre battery cells in the United States, its energy unit’s chief executive has said, building the cells at a plant in Kansas rather than shipping them in, as the Japanese group chases a market that barely existed a few years ago.

Mass production at the Kansas site is scheduled for the financial year ending March 2029, which Panasonic counts as fiscal 2028.

The plant gives the company a domestic base to supply American data-centre operators directly, a meaningful advantage at a moment when tariffs, supply-chain anxiety, and the sheer speed of AI build-out have made onshore manufacturing a competitive asset rather than a cost to be minimised.

The money behind the move is substantial. Panasonic is directing about 350 billion yen, roughly $2.18 billion, of a previously announced 500 billion yen AI infrastructure investment over fiscal 2026 to 2028 to its Energy unit, the division that also supplies Tesla, with the remaining 150 billion yen going to its Industry segment.

The split tells you where the company thinks the growth is: the battery business that grew up around electric vehicles is being retooled to feed the server hall.

The ambition is sized accordingly. Panasonic Energy chief executive Kazuo Tadanobu described the unit’s 950 billion yen sales target for data-centre-related energy storage in fiscal 2028 as a “minimum commitment,” with the business aiming to push sales past 1 trillion yen.

For a target to be framed as a floor rather than a goal is a sign of how quickly the company expects demand to climb.

The logic is grounded in how modern data centres actually run. The facilities training and serving AI models draw enormous, spiky loads, and they cannot tolerate even a flicker of interruption, which makes large-scale battery storage essential for smoothing supply, bridging outages, and managing the gap between what the grid can deliver and what the racks demand at any given instant. As AI compute scales, the storage attached to it scales with it.

The cells these facilities need are also a different specification from the ones that go into cars, tuned for grid-style duty cycles rather than the range and weight constraints of a vehicle, which is part of why an established battery maker still has to build dedicated capacity rather than simply repurpose its existing lines.

That demand is already straining the systems around it. The build-out has pushed electricity grids to their limits, with operators from Denmark pausing new connections to China wrestling with how to match clean power to data-centre load, a backdrop that makes on-site storage less of a luxury than a requirement.

Batteries are becoming part of the basic plumbing of AI, not an optional extra bolted on at the end.

Panasonic is not moving into an empty field. Chinese battery giants including CATL are racing into the same data-centre storage market, and the competition runs alongside the broader contest over the silicon inside those facilities, where Chinese firms are pushing domestic alternatives to Nvidia at speed.

The energy layer of the AI stack is becoming as contested as the compute layer.

The US plant is one node in a wider network. Panasonic Energy also plans a third plant in Mexico, with mass production likewise targeted for fiscal 2028, giving it North American capacity on both sides of the border.

The company has not detailed the Kansas site’s output volumes or named the data-centre customers it expects to supply, leaving the commercial specifics to emerge as production approaches.

What is clear is the direction: a battery maker that bet its future on cars is now placing a second bet, on the machines learning to think.

Austin streets now host something that looks ordinary at first glance but represents a sharp break from everything that came before. Production Cybercab units have started engineering tests on public roads, and these vehicles carry no steering wheel and no pedals. Tesla just posted video of the tests on June 30. The footage and supporting details show the first examples built for actual use rather than pure development. Earlier cars sometimes carried temporary controls. These do not.

The Texas Department of Transportation confirmed that the production design has no driver-operated controls of any type. Inside one test model, cabin footage shows the safety monitor seated in the front position, leaving an empty space where a steering wheel and pedals would normally be. The monitor’s hands rest against their legs. No controls are within reach. The huge central screen displays the current Tesla navigation interface, which includes the route, speed, and autonomous status in a familiar simple arrangement. The automobile navigates through typical traffic, curves, and downtown streets without the driver’s input.

TZYFOQN Metal Model Cars 1 32 for Teslas Cybercab Alloy Car Model Boy Birthday Vehicle Gift…

• Miniature models are not only vehicle replicas, but also spread the car model story and history and culture. Such as: track racing cars, sports cars…
• The zinc alloy construction provides the body with precise contours and substantial weight; Plastic components offer flexibility and resilience…
• The appearance of the classic car is highly restored, with special tread tires with strong grip, and the combination of retro and modern styling…

The cabin layout focuses on passengers, with two forward-facing seats in an open area created by the removal of any driving hardware. Large glass portions and a clean headliner contribute to a bright, airy atmosphere. Everything revolves around the ride rather than the act of driving. A single visible display provides both occupants with trip information and vehicle status.

Production Cybercabs are around 4.2 meters long and 1.8 meters wide, although they have usable inside space because designers were not required to package steering columns, pedal boxes, or instrument clusters. According to latest EPA data, the car has a battery capacity of roughly 48 kilowatt hours, a single front-mounted motor rated at around 219 horsepower, and a curb weight of approximately 3,113 pounds. Efficiency appears to be high, with estimates indicating 290 miles or more of real-world range in typical conditions.

Some models have a glossy metallic gold finish, which highlights the sleek body lines and futuristic lighting. The two-door form has distinct proportions that appear purposeful rather than dazzling. Doors are designed for quick access in a vehicle of this size. Tesla began producing these vehicles at Gigafactory Texas earlier this year. In February, the first production unit left the line. Volume manufacturing targets were set in April. The latest tests are the next step toward establishing that the entire hardware and software combination works on real roads with normal traffic.

There are presently 34 vehicles participating in the downtown Austin runs. During the validation process, everyone carries a safety monitor as usual procedure. The monitors observe and prepare for rare events that may necessitate human intervention, but they do not steer, brake, or accelerate whatsoever. What happens next depends on how these vehicles perform in the coming weeks and months, as well as the regulatory steps required for widespread unsupervised use. For the time being, seeing these control-free vehicles cruising through Austin traffic provides the clearest picture yet of what Tesla has built specifically for a driverless future.