News

• The World Wide Web was created 35 years ago today
• It was initially designed for sharing documents at CERN
• Its creator says we can return to its spirit of openness and collaboration

These days, it’s hard to go long without connecting to the Internet in one way or another: whether through one of the best web browsers or by connecting to a smart home device, it’s something we often take for granted. The primary means of interacting online — the World Wide Web — has just turned 35 years old today, and its creator has a timely message that couldn’t be more relevant for modern users.

The World Wide Web is the creation of British computer scientist Tim Berners-Lee. Working at the scientific body CERN in the 1980s and 1990s, Berners-Lee wanted a way to store, update, and locate documents within the sprawling organization. At the same time, he sought a solution that would allow for these documents to be shared with outside bodies quickly and straightforwardly.

That desire led to his creation of the World Wide Web. It was announced to the public on August 6, 1991, with Berners-Lee humbly stating that “We are very interested in spreading the web to other areas.” It was a platform whereby anyone could link to any document in a database using hyperlinks, or links embedded in text, images, and other media. Importantly, multiple users were able to access this platform simultaneously and independently of each other.

Adoption quickly spread from CERN to other research organizations before being opened up to the wider public, whereupon it exploded in popularity. And 35 years later, it’s an invention that has turned the world on its head.

An absolutely vital component of this system was that it had to be both free and decentralized. Berners-Lee was adamant that anyone should be able to create a website and implement hyperlinks on their own, without paying a fee and without a central body dictating what could and could not be done. That principle of openness still stands today — but it’s increasingly under threat.

Speaking to The Guardian in September 2025, Berners-Lee lamented that plenty of the web today is no longer free and open. “We see a handful of large platforms harvesting users’ private data to share with commercial brokers or even repressive governments,” he noted, adding: “We see ubiquitous algorithms that are addictive by design and damaging to our teenagers’ mental health. Trading personal data for use certainly does not fit with my vision for a free web.”

At the same time, user data is sold “to actors we never intended it to reach,” with the result being “deliberately harmful content that leads to real-world violence, spreads misinformation, wreaks havoc on our psychological wellbeing and seeks to undermine social cohesion.”

This is especially challenging in the age of AI, Berners-Lee argued. To put things right and get back to the original vision of his creation, the father of the web advocated for strong AI governance and guidance from organizations that are not motivated by data and profit but instead by collaboration and the needs of the individual. On this point, Berners-Lee quipped: “It is hard to imagine a big tech company agreeing to share the world wide web for no commercial reward like CERN allowed me to.”

As the World Wide Web turns 35, the spirit of “collaboration, creativity and compassion” that allowed it to be given away for free is being increasingly challenged by Big Tech — but Berners-Lee still thinks there’s time to put things right: “We can re-empower individuals, and take the web back. It’s not too late.”

The Apple AirPods Pro 3 have just received a surprise price cut at Amazon, bringing the excellent premium earbuds down to just $189 (was $249) at the retailer.

That's the lowest price since the retailer's annual Prime Day mega-sale back in June, and likely the best deal we'll see for a while. Ever since Prime Day, these buds have stubbornly held at $199, which is the usual 'sales price' for Apple's premium buds.

Even though the record low is officially $169, the AirPods Pro 3 only hit that on one occasion over Prime Day, so today's price cut is definitely worth considering. Even at $189, the AirPods Pro 3 are an excellent buy for anyone looking for some high-end buds. Not only do they feature excellent noise cancellation, but they're packed with new features like heart rate monitoring and Live Translation.

Amazon has dropped Apple's AirPods Pro 3 to $189.99, which is the best deal I've seen on these buds in well over a month. Upgrades to the AirPods Pro 3 include improved Active Noise Cancellation and audio, a new comfortable in-ear design, and new features such as live translation and the ability to track heart rate, workouts, and calories burned.View Deal

• See more: check out all of today's deals at Amazon

We awarded the buds an impressive four and a half stars out of five in our AirPods Pro 3 review, calling them 'near-perfect noise-cancelling earbuds' at the time. While not exactly cheap, they offer excellent noise cancellation and sound quality.

If we had to draw any criticisms, we'd mention that the AirPods Pro 3 don't have class-leading battery life (around 5.5 hours in our testing). You'll also need one of the latest iPhones to really take advantage of their full suite of features like Live Translation and personalized fitness insights. They're still great with older devices, but iPhone 15 and newer devices support Apple Intelligence, which is required for the newest bells and whistles.

• HP, Asus, and Acer have qualified and begun shipping CXMT DRAM in small volumes, in a handful of notebook models sold outside the United States
• The restraint is potentially driven less by US policy than by fear of the incumbents: with Samsung, SK Hynix, and Micron holding over 90% of the market, buyers say they dare not source too much from CXMT
• CXMT is not a cheaper option either currently, as sources say its DRAM costs no less than Samsung's currently and cannot be booked beyond the current quarter

HP, Asus, and Acer have begun putting DRAM from China's ChangXin Memory Technologies (CXMT) into notebook computers, a report from Nikkei Asia has claimed, reporting several major PC makers completed qualification of CXMT's chips around the middle of this year.

The volumes are small, and the number of affected models is small, but perhaps more importantly, none of those models is sold in the United States.

The reason the volumes are small may have less to do with assemblers not wanting to irk a White House that has taken an increasingly tariff-heavy, America-first approach to manufacturing, but rather with market incumbents that currently account for as much as 90% of the total market's supply.

PC makers are not being cautious primarily because of Washington, but rather because they do not want to upset agreements with market giants such as Samsung, SK Hynix, and Micron.

An executive at one PC company told Nikkei that with the top three holding more than 90% of the global market, buyers have to stay low-key about using CXMT, adding: "We dare not source too much from CXMT at this moment."

That does make for an interesting conundrum: in the middle of the worst memory shortage in the industry's history, the binding constraint on diversification is fear of annoying the people you are already struggling to buy from.

Nikkei's report on the matter only reinforces what many market participants have already hinted at: CXMT is hardly the "cheaper" option.

Its DRAM quotes do not undercut Samsung's, and it reportedly cannot even be booked beyond the current quarter, as competition over who can buy it has continued to ramp up even as it reportedly chooses to prioritize local conglomerate Huawei's AI ambitions.

What PC makers are actually buying, then, is optionality. A supply chain manager who works with HP and Asus put it to Nikkei, saying they don't want to neglect a potentially important source while the market is this tight.

For context, validating a new DRAM supplier takes months of engineering, signal-integrity testing, thermal work, and platform validation, and it does not happen on a whim. HP, Asus, and Acer have now built a switch. Whether they throw it depends on conditions none of them control, but the switch exists, and eighteen months ago it did not.

With Samsung stating that the RAM crisis is here to stay, the underlying conditions have merely exacerbated an AI-driven memory crisis that has directly affected consumer electronics across the board as supply continues to get repurposed.

Spot RAM pricing has risen by 400% or more over the past twelve months. PC makers have responded by prioritizing memory allocation for premium models and lifting prices across their ranges by several hundred dollars. IDC expects the global PC market to contract by more than 11 percent this year, with supply conditions deteriorating further toward the end of it.

Smartphones are equally affected, with some makers scaling back certain models to memory configurations their predecessors already offered, while others increase pricing to offset higher component costs. Xiaomi, Oppo and Vivo have each cut 2026 shipment targets more than once.

The pressure has already reached the components below the memory itself: motherboard makers have absorbed roughly 50 percent increases in PCB costs and trimmed their own shipment targets accordingly.

None of this originates in the PC market. It originates in AI infrastructure, where demand for high-bandwidth memory has pulled wafer capacity away from conventional DRAM at every major supplier.

That is precisely the gap CXMT has been expanding into, which is why a company most Western buyers ignored two years ago is suddenly qualified at three of the largest PC brands in the world.

So what will Samsung, Micron, and SK Hynix think? In the short term, very little, and they have every reason to say so. They are the primary beneficiaries of this shortage, with pricing power they have not enjoyed in a decade and no commercial incentive whatsoever to loosen supply. A few thousand budget laptops in Europe and Asia that carry Chinese DRAM do not register against that.

The longer view might be less comfortable. Oligopolies are usually broken not by a competitor arriving with a better product but by customers becoming willing to qualify an alternative, and customers become willing when the incumbents make life sufficiently unpleasant. That threshold has now seemingly been crossed at three of the five largest PC vendors, during a period when the incumbents were charging whatever they liked.

The alternative parts are more expensive than expected and hard to get, so nobody has switched in any meaningful volume. But the certification work is done and dusted, CXMT is building capacity at multiples of its current base, and the moment supply loosens or Washington's posture shifts, the cost of moving volume across is dramatically lower than it was.

For now, the best example of how the industry is treating this news is visible in Acer's response to Nikkei: It declined to identify suppliers, said it maintains close contact with multiple global manufacturers, and described the goal as supply chain resilience. HP also declined to comment, and Asus did not respond at all to inquiries at the time of writing. That is not the language of companies making a bet but rather those that would like to keep as many options open as possible.

• The first HBF specification defines stacks of up to 512GB with three bandwidth grades from roughly 0.4TB/s to 3.0TB/s, connected to processors over UCIe and published openly through the Open Compute Project
• Google and Tenstorrent have joined the consortium, but Nvidia is not among the named participants, which matters more to HBF's prospects as a de facto alternative to increasingly pricey HBM
• Samples of the first AI inference devices built with HBF are expected to show up in early 2027

SK Hynix and SanDisk have published the first technical specification for High Bandwidth Flash.

Released through the Open Compute Project to coincide with the opening of the Future of Memory and Storage conference, it makes for an impressively fast milestone, arriving six months after the HBF workstream was convened at SanDisk's Milpitas headquarters and a year after the two companies first agreed to standardize the technology.

The solution, at least for now, exists only on paper, and has managed to draw two crucial partners to its consortium: Google and Tenstorrent.

High Bandwidth Flash's specification defines capacities up to 512GB across two stack configurations, 8-high and 16-high NAND dies, and sorts performance into three grades running from roughly 0.4TB/s to 3.0TB/s.

HBF's pitch, unlike its competition, focuses on capacity rather than raw speed. A 512GB HBF stack sits against roughly 48GB to 64GB for an HBM4 stack, which is between eight and ten times the capacity, while the top HBF grade's 3.0TB/s lands inside the 2.0 to 3.3TB/s band typical of HBM4.

It also adopts UCIe, the open chiplet interconnect standard, as the link between an HBF stack and a host processor, enabling the technology to be attached to GPUs and CPUs from different vendors.

What could be challenging is delivering on its promise within a reasonable time; however, SanDisk has been working to a public roadmap since August 2025: first samples of the memory itself in the second half of 2026, and samples of the first AI inference devices built with HBF in early 2027. Neither company updated that schedule with this week's announcement.

Industry expectations put actual commercialization somewhere between late 2027 and 2028, with the volume of data center demand that would justify the investment arriving closer to 2030, in a cycle that also involves partner testing followed by ramped-up production.

This is despite SanDisk's Hsu telling SDxCentral in May 2026 that working through the Open Compute Project (OCP) let the effort move faster and that "you won't have to wait too long", which is a reasonable thing to say about a standard and a slightly generous thing to say about a product still eighteen months from sampling inside a device.

SK Hynix is equally cagey when it comes to HBF's roadmap: the same document that announces HBF also unveils the company's tenth-generation 375-layer 4D NAND, claims a 2.5x improvement in performance per watt, and states plainly that mass production of enterprise SSDs built on it begins early next year; HBF gets no date mentioned at all.

One interesting thing is that Nvidia is not among the named participants, even though it dominates the AI accelerator market.

SK Hynix has, however, expanded a strategic partnership with Nvidia covering next-generation memory as recently as July 25, and that might be a sign that things could change, but there is no official confirmation from what is arguably the biggest AI player in the market just yet. Samsung, Micron and Kioxia are likewise absent, with Kioxia pursuing a different approach to the same problem.

When HBF does eventually ship, it will probably not reach consumers in any shape or form until AI hyperscalers are sated and a viable interface for retail users is available to harness the tech. For now, HBF is unveiled but exists only on paper, allowing hyperscalers and accelerator designers to plan ahead for a technology that will become increasingly invaluable as the compute focus continues to shift from training to inference, with AI data centers coming online.

The Chuwi UniBook is a 14-inch Windows laptop built around Intel's newly launched Core 3 304 processor, part of the Wildcat Lake family and the first Series 3 chip to appear in a laptop at this price. It weighs a claimed 1.2 kg, comes housed in an anodised aluminium chassis, and pairs a 16:10 IPS display with 8GB of memory and a 256GB SSD.

Positioning this model directly against Apple's MacBook Neo, undercutting it by around $250 while offering a larger screen, the same memory and storage as the base Neo configuration, and a far wider set of ports. Early benchmarks of the Core 3 304 from other devices suggest genuinely strong single-core performance for a chip this cheap, which makes the UniBook one of the more interesting budget launches of the year on paper.

The critical caveat is that macOS runs reasonably well with 8GB of RAM and 256GB of storage, where Windows 11 is a singularly less spectacular experience.

While you can upgrade the storage to any Gen 3 or 4 NVMe 2280 drive, the amount of RAM isn’t able to be expanded.

Price is therefore the trade-off, which means the UniBook isn’t ideal for substantial tasks, even if it's acceptable for light work. One side benefit of not being especially powerful is that the battery will last easily more than a working day, and it might even stretch to two.

From a value perspective, the Chuwi UniBook looks like an excellent deal, but the specification precludes it from being one of the best business laptops.

• How much does it cost? From $450
• When is it out? Available now
• Where can you get it? Direct from Chuwiretailers

Chuwi initially opened pre-orders for the UniBook at $449 US dollars, positioning it as the cheapest laptop yet built around Intel's Wildcat Lake platform. At the time of writing, using the Global Chuwi store, it's priced at $449 for US customers, £335 for the UK and €449 for Europeans.

Given that this is a fully functional laptop with RAM and storage, and new processor technology, that price seems remarkably good.

However, there is only a single SKU that offers one processor, the Intel Core 3 304, 8GB of RAM (not upgradable) and only 256GB of storage.

I’m only aware of two other laptop makers with products built around this platform. They would be HP with the OmniBook 3 AI 14" Laptop and Lenovo’s ThinkPad E16 Gen 4.

The HP is available in the UK through Currys for £799, for a model with 8GB of RAM and 512GB of storage.

The version of the ThinkPad E16 Gen 4 that uses the Intel Core 3 304 costs £954.60 in the UK direct from Lenovo, but does come with 16GB of RAM and Wi-Fi 6E. At the time of writing, both these models have yet to be launched in the US.

The UK pricing we have strongly hints that Chuwi's pricing is remarkably competitive, and I wouldn’t be surprised if it were raised in the coming months.

It will be interesting to see how other brands price hardware based on this platform.

For those wondering, Apple's MacBook Neo sells for £699 in the UK and $699 in the USA, for a machine with 8GB of RAM and 256GB of storage, without a charger.

• Value: 4.5 / 5

Laptop:

Chuwi UniBook

CPU:

Intel Core 3 Processor 304, Series 3 Wildcat Lake (5 Cores, 5 Threads, up to 4.3 GHz, Intel 18A process)

GPU:

Intel Graphics, Xe3 architecture, up to 2.3 GHz, 9 TOPS

NPU:

Intel NPU 5, 15 TOPS

RAM:

8GB LPDDR5, 6400 MT/s not upgradable

Storage:

256GB PCIe 3.0 SSD, single M.2 2280 slot rated up to PCIe 4.0 x4

Screen:

14.0-inch IPS, 1920 x 1200 (16:10)

Ports:

2x USB C (5 Gbps, PD, 4K@60Hz video), 2x USB A 3.2 Gen 1, 1x USB A 2.0, 1x HDMI 1.4b (4K@30Hz), 1x Gigabit Ethernet, 1x TF card slot, 1x 3.5mm audio

Camera:

720p webcam with physical privacy shutter

Networking:

Wi-Fi 6, Bluetooth 5.2

Dimensions:

314.5 x 221 x 16.4 mm

Weight:

1240g

OS:

Windows 11 Pro, pre-installed

Battery:

53.38Wh Lithium Polymer

Power supply:

65W USB-C PD charger

• Anodised aluminium alloy chassis
• Lightweight and thin
• Lots of ports

The chassis on the UniBook is made from anodised aluminium alloy, and Chuwi quotes a weight of approximately 1.2 kg and a thickness of 16.4 mm. Those figures put the UniBook among the lighter 14-inch Windows laptops at this price, assuming they hold up on a scale. The hinge opens to 180 degrees, and the marketing page states an 88% screen-to-body ratio thanks to narrow bezels around the 16:10 panel.

The specification lists two full-function USB-C ports, two USB-A 3.2 ports, one USB-A 2.0 port, HDMI, Gigabit Ethernet, a TF card slot, and a 3.5mm jack, which, on paper, is a generous selection for a budget machine. What’s odd, when I think about it, is that more expensive laptops tend to have fewer ports, forcing the owners to consider using docking stations to add those connections.

The inclusion of a LAN port is a definite advantage, especially in that awkward new Windows PC phase where Microsoft attempts to download all of human knowledge in the full knowledge that it needs almost none of it.

The USB-C and two USB-A ports are only 5Gbs, which is slightly disappointing not to have one that is 10Gbps, but there is HDMI out and a MicroSD card reader.

The keyboard is full-size and has two-level backlighting. It’s not ideal for creating dissertations, but for small input jobs it’s fine. The provided review hardware was a US layout, which I assume is the one most buyers will get.

Since I know how the UK layout maps to the US one, this wasn’t a huge problem for me, except that this keyboard annoyingly doesn’t have a PrtScn key. That made screen capture slightly more convoluted.

As trackpads go, this one is on the smaller side, measuring approximately 120 mm by 70 mm, and it uses the flexible-click methodology.

Overall, considering the low cost of this device, its construction is much better than I’d anticipated. The use of metal skins and a simple but effective style works well. What it lacks is a fingerprint sensor or a Windows Hello camera. But with so many USB ports, it could easily use a security fob if you want to add extra protection.

However, given the specifications and price, I see this device being more oriented towards education than business, and I have some thoughts on its suitability for that in the next section.

• Design: 4 / 5

• Wildcat Lake Intel Core 3 304
• NPU and GPU short of Copilot Plus certification
• 8GB of RAM can’t be upgraded

Wildcat Lake is Intel's answer to a question that has irked it for years. How do you bring modern chip design to the cheapest laptops without pricing them out of the market they serve? The family was announced quietly at CES 2026 alongside the much bigger Panther Lake launch, and its job is to replace the ageing Alder Lake N and Twin Lake chips that have powered budget laptops, mini PCs and Chromebooks since 2023.

Where those older chips relied on Gracemont efficiency cores alone, Wildcat Lake borrows directly from Panther Lake's playbook. It pairs Cougar Cove performance cores with Darkmont low-power efficiency cores, adds Xe3 graphics and an NPU 5 neural engine, and builds the whole thing on Intel's 18A process.

In effect, Intel has taken its flagship architecture and squeezed a scaled-down version of it into a chip built for machines costing a few hundred dollars rather than a few thousand.

Well, that’s the promise, and on paper it sounds perfect for the budget consumer. Modern architecture at 18A should bring a genuine efficiency jump over the older Intel 7 process used by the N-series chips, and early single-core benchmarks on Wildcat Lake devices back that up, with some results landing close to flagship phone silicon. For a family aimed squarely at browsing, office work and video calls, single-core strength matters more than raw multi-core muscle ever would.

However, in the time-honoured Intel tradition of grasping potential defeat from the jaws of victory, this series has become more about what Intel chose to cut.

Wildcat Lake uses a simpler organic substrate and flip chip packaging rather than the advanced packaging found on Panther Lake, which keeps costs down but limits how far the platform can scale. Core counts are modest across the board, topping out at six, and GPU configurations are similarly trimmed.

None of that is a flaw exactly, since this was never meant to compete with premium silicon, but it does mean the ceiling was made low by design.

The Core 3 304 sits right at the bottom of that stack. It carries five cores across five threads, just one Cougar Cove performance core paired with four Darkmont low-power efficiency cores, and turbo clocks reaching 4.3GHz. That single performance core is notably light even by Wildcat Lake standards, since every other chip in the family carries two. Graphics come from a single Xe3 core rated at 9 TOPS, alongside an NPU rated at 15 TOPS.

What it lacks is just as telling. There is no second performance core to lean on for heavier multitasking, no multi-core Xe3 graphics block, and no ray tracing.

Combined NPU and GPU output falls well short of Microsoft's 40 TOPS Copilot Plus threshold. The dedicated NPU is rated at 15 TOPS Int8. Combined with the GPU's 9 TOPS, that gives a platform total of 24 TOPS. Microsoft's Copilot Plus certification requires 40 TOPS from the NPU alone, so despite Chuwi's AI Laptop branding, the UniBook does not qualify and will not run Copilot Plus-specific features.

Memory is 8GB of LPDDR5 running at 6400 MT/s, a healthy speed for this class of chip even if the capacity is modest. Is it enough for Windows 11?

The fact is that when you fire up the machine with the default installation, 6.3 GB of RAM has already been eaten before the first app is launched. Launch a browser and open 10 webpages; the system starts paging, and things get jerky from that point onwards. Where it really struggles is multitasking, since there is only a single performance core, alongside the four efficiency cores.

There may be ways to reduce Windows 11's memory usage, and you’ll need to use them all if you intend to run multiple apps and not find yourself waiting on this system. In times long gone, you would open the laptop, find the 8GB SODIMM, and replace it with a larger one, but that’s not possible with this design.

The very week I wrote this review, Microsoft threw a few of its disposable developers at members of the press to talk about changes they intend to make Windows 11 work better with less memory. Technically, the minimum memory requirement is 4GB, but Microsoft have focused apparently on improving the 8GB experience. While these updates are not yet launched, and when they will be is unclear, these are clearly driven by the catastrophic price hikes in RAM cost.

I need to point out that Windows 11 launched in 2021; therefore, for five years, Microsoft sat on its overly profit-padded backside, ignoring users who complained that their operating system wasn’t an efficient user of memory.

But a few PC makers whine that they need to reduce the RAM complement to 8GB from 16GB on some of their lines, and those in Redmond leap into action? The poor optics of these announcements won’t surprise many.

It’s also interesting to note that this machine was built to compete with Apple's MacBook Neo, and that comes with the same 8GB+256GB entry-level profile, but nobody is complaining that it doesn’t work smoothly with macOS.

This would be less of an issue for laptops in general if the makers all hadn’t chosen to make memory non-upgradable. A choice they made, not the people buying them.

The only upgrades available here are the M.2 Wi-Fi adapter and storage, and getting to them is relatively easy if you remove the 11 screws that hold the underside in place.

If anyone is reading this and has this system, just be aware that 8 of the screws are short and three are long, and getting them mixed up could have dire consequences when you reassemble the system.

There is only one M.2 2280 for storage, so you will need a USB caddy to do a drive cloning, and there isn’t room for a drive with a heatsink inside.

One major disappointment I noticed is that the pre-installed drive is only PCIe 3.0 x4, where the slot is rated for PCIe 4.0 x4. And, as the drive is only 256GB, finding a 500GB or 1TB replacement drive that’s PCIe 4.0 would be a good upgrade.

Overall, this system is cut to the bone in some critical areas, specifically RAM, making it less than ideal for those who like to load multiple apps and multitask between them.

It needed those Microsoft efficiency upgrades yesterday, or it's likely to be a contributing factor in the switch to Linux for those owning this hardware.

• Hardware: 3.5 / 5

Laptops

Chuwi UniBook

Chuwi CoreBook X 7430U

CPU

Intel Core 3 304

AMD Ryzen 5 7430U

Cores/Threads

5C 6T

6C 12T

TPD

15W

15W

RAM

8GB LPDDR5

16GB DDR4 (1x 16GB)

SSD

HOGE H750L 256GB

512GB Wodposit WPBSNM8

Graphics

Intel UHD Graphics (Gen 12.2)

AMD Radeon Graphics

NPU

Intel NPU (15 TOPS)

N/A

3DMark

WildLife

5,198

4,581

FireStrike

1796

2085

TimeSpy

830

836

Steel Nomad.L

590

716

CineBench24

Single

N/A

63

Multi

N/A

254

Ratio

N/A

4.05

GeekBench 6

Single

1962

1442

Multi

5338

4848

OpenCL

6006

11015

Vulkan

7304

10122

CrystalDIsk

Read MB/s

2315

3588

Write MB/s

2087

1852

PCMark 10

Office

4351

4676

Battery

20h 26m

9h 30m

Battery

Whr

55.38

46.2

PSU

65W

65W

WEI

Score

8.1

6.6

It was difficult for me to find laptops I’ve reviewed that would make a suitable comparison, since most of them use much more powerful chips than the one in the UniBook.

In the end, I found that the results for the Chuwi CoreBook X 7430U are reasonably comparable, especially with respect to GPU performance. However, one massive advantage that the CoreBook X 7430U had was 16GB of RAM, even if it was DDR4. That allows it to share more with the GPU, which is why the scores for that hardware improve as the graphics challenges increase.

One unexpected casualty of the 8GB memory capacity was the CineBench24 benchmark, which failed to run due to the amount of RAM allocated to the Intel UHD Graphics (Gen 12.2) GPU.

Another disappointing result is the NVMe drive's speed. But this is exactly what happens when you use a cheap Gen 3 drive in a Gen 4 slot. The battery performance is thankfully a more positive story.

Chuwi's own promotional material quotes up to 13 hours of 1080p video playback from the 53.38Wh battery, tested at 50% brightness and volume with Wi-Fi on, Bluetooth off, and Best Efficiency Mode engaged.

However, you can use this machine with the screen much less bright, and in my testing, it lasted over 20 hours at full capacity on its 55.38Wh battery. Easily the best result of these tests.

That’s a couple of busy days at least, or one dawn-till-dusk exercise. That capacity can be replaced on the provided 65W charger in a little over 2 hours, enabling a quarter of the capacity to be recovered in just 30 minutes.

The screen performance levels that Chuwi quotes are 100% sRGB and 300-nit peak brightness. While the 300nit value is reasonably accurate, the colour gamut of the screen on the supplied UniBook sample came nowhere near 100% sRGB. It managed just 63% of sRGB, 43% of AdobeRGB and 34% of Rec2020.

Gamut is certainly a weakness, but contrast, uniformity (brightness and colour) and tone response are all decent. My conclusion was that this isn’t a bad screen, unless you are specifically looking for colour accuracy.

Those looking for a performance laptop won’t be interested in this hardware, which is nailed down by a processor with only five cores and only 8GB of RAM running on a PCIe 3.0 SSD.

When using this system, keep your expectations low, and keep the number of web pages open small.

I was almost tempted to put a PCIe 4.0 SSD drive in the review unit, and put a modern Linux distro on it to see how much better the experience might be, but time was against me.

In short, this system is cheap for a reason, and even with Intel’s new silicon, the performance is undermined on numerous other fronts.

• Performance: 3 / 5

UniBook is a genuinely interesting proposition on paper. It is the first widely available laptop built around Intel's new Wildcat Lake platform; it undercuts Apple's MacBook Neo on price while offering more ports and a bigger screen, and early third-party benchmarks of the Core 3 304 hinted at unusually strong single-core performance for the money.

However, a Windows 11 system with 8GB of RAM that can’t be upgraded and a storage drive that undersells the socket it sits in combine to undermine what could be a much more useful system than it ultimately is. The overselling of its AI capability doesn’t help either.

What is good about this system is the excellent battery life and the wide number of ports, even if the best it can offer is USB 3.2 Gen 1.

None of that rules the UniBook out. It simply means the value case pivots on the purchaser's use case, and it needs to be modest. For that reason, I wouldn’t recommend this to a student, as they’ll discover its limitations on the first day of use.

Value

Competitively priced for the specification

4.5/5

Design

Lightweight chassis with plenty of ports

4/5

Hardware

New Intel Wildcat Lake platform but only 8GB of RAM

3.5/5

Performance

Only suitable for light tasks due to Windows 11

3/5

Overall

Inexpensive hardware suitable only for light tasks

4/5

You need a low cost Windows 11 machine
The best aspect of the UniBook is its price; considering the build quality and screen size, it looks like a bargain. However, there is no ‘free lunch’ with this machine, and it does have some limitations that mean it's only useful for light tasks.

You expect all-day battery life
In a continuous test, the UniBook lasted over twenty hours, and that’s with the screen brightness turned down somewhat. Even the makers quote more than thirteen hours of continuous video playback, making it suitable for a long journey.

You need multiple apps open
With only 8GB of RAM, most of which Windows 11 has already earmarked, running multiple apps or those with high demands isn’t practical on this hardware.

You use external SSDs
The lack of a Gen 2 USB port means the fastest file transfers from the Chuwi UniBook can achieve are capped at 500mb/s. Using 1,000mb/s rated external SSDs or even faster options won’t operate to their full potential.

As AI develops into the latter half of the 2020s, scientists have repeatedly warned of the dangers that may arise from its widespread deployment. In particular, there is an overriding view that advancing AI is a tool that can be used for good or for bad depending on who is wielding the tool at any given time.

The British computer scientist Geoffrey Hinton was giving an interview with the New York Times upon leaving his role at Google, primarily so he could speak freely and publicly about the dangers of the technology that he helped create.

This article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. Read the full series here.

In this landmark interview, the 'Godfather of AI' was clear that he couldn't continue working for Google while harboring the deep-rooted concerns that he held over the use of AI in various domains. The scientist is, for example, opposed to using AI on the battlefield in "robot soldiers", the newspaper reported.

At the time, he continued his relationship with Google because he saw the company as being a "proper steward" for AI as it was being developed. His concerns materialized when Microsoft incorporated Bing with a chatbot, forcing Google to act fast so it could incorporate AI into Google Search.

Hinton, who was jointly awarded the Nobel Prize in Physics for his work developing neural networks in the 1980s, has since remarked in an interview with CBS News that he didn't think we would make such progress in the 40 years since.

He has also floated the notion that there's a non-zero chance that AI could take over, comparing AI with a cute tiger cub that could, one day, potentially kill you when it's grown up. That came alongside warnings that various organizations and individuals, like cyber criminals or authoritarian regimes, could weaponize AI for their own agendas.

• Camera-only systems can only “approximately match” a human driver

• Multi-sensing setup offers more than just a failsafe
• Waymo co-CEO backs the use of Lidar and radar over camera-only systems

Waymo co-CEO Dmitri Dolgov has spoken out about the technology required to run a fully autonomous robotaxi service, addressing the subject of camera-only systems.

While he didn’t mention Tesla directly in his in-depth talk at Y Combinator’s Startup School, the presentation directly addressed vision-only camera setups, like those Elon Musk and Tesla have been pursuing.

“Humans, of course, can drive with just eyes, so there’s that proof of existence,” he said, according to Electrek.

“If the goal were to just approximately match human performance or to build an assist product, that’s a very reasonable way to go.”

But “weak sensing” inevitably leads to a safety curve that flattens out “way too early,” he added, highlighting the many cases where a camera system can come unstuck.

Low light or poor visibility, dust storms and objects physically blocking the lens can bring a camera-based system to a grinding halt in seconds.

Dolgov then gave an example of how Waymo’s combination of Lidar, radar and high-definition cameras allowed one stricken vehicle to safely navigate itself back to a depot when a tree branch became caught in its wipers and obscured the camera lens.

While Dolgov doesn’t completely dismiss camera-based systems, stating that they can easily support a driver with highly autonomous assistance systems, he admits they will struggle to replace a human entirely.

The issue is that fully autonomous robotaxis need to be as close to perfect as it’s possible to be.

A mounting record of accidents will likely put off customers for life and kill the technology before it has time to gain a foothold.

Having ridden in numerous autonomous vehicles over the years and listened to those that have built them, I am fully in Dmitri Dolgov's camp.

While a camera-based system might help keep costs down, fitting autonomous vehicles with Lidar, radar, and other sensors feels like a natural and reasonable step to make them as safe as possible.

A visit to Motional’s HQ in Las Vegas earlier this year illustrated the need for multiple sensing technologies when you completely remove the driver from the picture.

Its robotaxis, which are now bespoke versions of the Hyundai Ioniq 5 (complete with radar and Lidar), have driven over two million autonomous miles with zero at-fault incidents.

Like Waymo, it too believes a multimodal approach that fuses various technologies is the safest way forward.

Laura Major, the company’s CEO, said at the time that rival startups were able to get to a “90% performance” very quickly, often relying solely on high-definition cameras.

“These offer really smooth rides in most situations,” Major remarked. “The problem is when you’re going to remove a driver and have true driverless operation, that’s not enough. You can’t perform well in 90% of the cases,” she added.

Admittedly, I haven't ridden in one of Tesla's driverless Robotaxis, but the numbers don't lie.

According to Electrek, Tesla's own data confirms a crash rate three times worse than humans even with a monitor at the wheel, while Waymo's numbers suggest it is 17 times safer after covering some 220 million rider-only miles.

Tesla has accumulated just 380,000 driverless miles over the past year since launching its robotaxi service.

A new NYT Connections puzzle appears at midnight each day for your time zone – which means that some people are always playing 'today's game' while others are playing 'yesterday's'. If you're looking for Thursday's puzzle instead then click here: NYT Connections hints and answers for Thursday, August 6 (game #1152).

Good morning! Let's play Connections, the NYT's clever word game that challenges you to group answers in various categories. It can be tough, so read on if you need Connections hints.

What should you do once you've finished? Why, play some more word games of course. I've also got daily Strands hints and answers and Quordle hints and answers articles if you need help for those too, while Marc's Wordle today page covers the original viral word game.

SPOILER WARNING: Information about NYT Connections today is below, so don't read on if you don't want to know the answers.

Today's NYT Connections words are…

• JACKPOT
• MOTOR
• SNOW
• JAR
• WICK
• SUNDAE
• RAMBO
• BUTTONS
• TOMATILLO
• SHAFT
• WAX
• BUCKYBALL
• LIQUID NITROGEN
• FRAGRANCE
• CABLES
• REFRIGERATOR

What are some clues for today's NYT Connections groups?

• YELLOW: Parts of thing you light
• GREEN: Frozen
• BLUE: Elements of a lift (in the UK)
• PURPLE: Begin with a masculine beast

Need more clues?

We're firmly in spoiler territory now, but read on if you want to know what the four theme answers are for today's NYT Connections puzzles…

What are the answers for today's NYT Connections groups?

• YELLOW: COMPONENTS OF A CANDLE
• GREEN: COLD THINGS
• BLUE: PARTS OF AN ELEVATOR
• PURPLE: STARTING WITH MALE ANIMALS

Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON'T WANT TO SEE THEM.

The answers to today's Connections, game #1153, are…

• YELLOW: COMPONENTS OF A CANDLE: FRAGRANCE, JAR, WAX, WICK
• GREEN: COLD THINGS: LIQUID NITROGEN, REFRIGERATOR, SNOW, SUNDAE
• BLUE: PARTS OF AN ELEVATOR: BUTTONS, CABLES, MOTOR, SHAFT
• PURPLE: STARTING WITH MALE ANIMALS: BUCKYBALL, JACKPOT, RAMBO, TOMATILLO

• My rating: Hard
• My score: 2 mistakes

“Solid” was Connections’ assessment of my performance today, which is very generous considering I messed up both COMPONENTS OF A CANDLE and COLD THINGS.

For COLD THINGS I thought it was specifically about keeping things cold, but after feeling confident about LIQUID NITROGEN, REFRIGERATOR, and SNOW, I panicked and selected JAR.

Next, I was certain we were looking for something to do with candles but instead of JAR I picked SHAFT, reasoning that it described the shape of a traditional candle.

• YELLOW: CLASSIC WOODEN TOYS: ALPHABET BLOCKS, CUP-AND-BALL, JACOB'S LADDER, LINCOLN LOGS
• GREEN: PLACES WITH LANES: BOWLING ALLEY, FREEWAY, SUPERMARKET, SWIMMING POOL
• BLUE: ICONIC ROMAN LANDMARKS: COLOSSEUM, PANTHEON, SPANISH STEPS, TREVI FOUNTAIN
• PURPLE: PLACES WITH NETS: BARCLAYS CENTER, FISHING BOAT, HOCKEY RINK, TENNIS COURT

NYT Connections is one of several increasingly popular word games made by the New York Times. It challenges you to find groups of four items that share something in common, and each group has a different difficulty level: green is easy, yellow a little harder, blue often quite tough and purple usually very difficult.

On the plus side, you don't technically need to solve the final one, as you'll be able to answer that one by a process of elimination. What's more, you can make up to four mistakes, which gives you a little bit of breathing room.

It's a little more involved than something like Wordle, however, and there are plenty of opportunities for the game to trip you up with tricks. For instance, watch out for homophones and other word games that could disguise the answers.

It's playable for free via the NYT Games site on desktop or mobile.

• Recycled Mars rover hardware may still cost NASA more than expected
• PROMISE faces mounting questions over unexpectedly high development and integration costs
• Existing rover hardware cannot eliminate costly engineering and certification requirements

NASA's long-term lunar exploration program includes robotic missions requiring specialized vehicles capable of carrying scientific instruments across the Moon.

One proposed mission involves sending the Polar Rover for Observation, Mapping, and In-Situ Exploration (PROMISE), a robotic Moon rover, to conduct scientific investigations on the lunar surface.

According to NASA, the proposal to build the rover from recycled Mars rover engineering hardware could become too expensive after estimates reached between $723 million and $1.33 billion.

Instead of building the Moon rover from scratch, NASA plans to combine engineering models from the Curiosity and Perseverance rover programs.

Although the concept relies on existing hardware, NASA says the expected costs remain far beyond what the agency considers acceptable for such a mission.

The cost estimates were prepared by Casey Dreier, the Planetary Society's chief of space policy, after reviewing NASA's proposal for the planned lunar rover.

According to Dreier, the estimated total includes launch and landing costs ranging from $234 million to $320 million together with engineering expenses.

NASA Administrator Jared Isaacman did not challenge every figure within the estimate, but argued that launch and landing expenses would exist for any lunar mission.

Isaacman believes that it “would be crazy not to take advantage of hardware taxpayers have already invested hundreds of millions of dollars in."

But, "If PROMISE costs even 20 percent of your low-end estimate, we simply will not fly it," he concluded.

The proposal combines MAGGIE and OPTIMISM, engineering models developed to support testing for the Curiosity and Perseverance rover programs before their Mars missions.

Those vehicles were built to help engineers test hardware performance, software functions, and driving activities under controlled conditions before operational spacecraft left Earth.

NASA originally considered the concept because using hardware that already exists appeared capable of reducing development costs.

Engineers would still need to certify the hardware for launch because equipment designed for ground testing cannot automatically survive spaceflight and lunar operations.

Additional work would also involve adapting communications equipment and integrating NASA's available plutonium radioisotope power system into the redesigned lunar vehicle.

The engineering models currently support existing activities, meaning removing them from those roles could affect testing work linked to other programs.

Dreier summarized the financial challenge by saying, "Regardless of whether one agrees with every estimate… one thing is clear: PROMISE is far from a 'freebie.'"

The engineering models currently support existing activities, meaning removing them from those roles could affect testing work linked to other programs.

NASA has not released its own detailed cost estimate for the proposed rover; therefore, the project's overall financial viability remains under wraps.

That said, its future will likely depend on whether engineers can reduce costs while overcoming technical challenges involved in converting Earth-based test hardware.

Via The Register