Feed aggregator

A few months ago, Chinese LLM DeepSeek-R1 made headlines. The world was introduced to an open-source AI model that matched OpenAI o1’s reasoning abilities at a fraction of the cost. A relatively obscure startup topped Apple’s App Store rankings and sent shockwaves through global stock markets.

DeepSeek’s breakthrough proved AI could be built without massive capital investment or Nvidia’s highest-grade chips. More importantly, it showed that an open-source strategy could challenge America’s AI hegemony. China is betting that widespread, open-source use offers more value than the West’s paywalled models, such as ChatGPT, Anthropic’s Claude, and Perplexity.

But if DeepSeek was a minnow, a Chinese whale is now having its own open-source moment: Baidu, China’s answer to Google, with 704 million monthly active app users and a market cap of 31.6 billion USD, has announced it will open-source its powerful generative AI model, Ernie. This release represents a dramatic policy reversal for Baidu, whose CEO Robin Li had previously advocated for proprietary, closed-source models as the only viable path for AI development.

If DeepSeek demonstrated that China could compete with the West, Baidu’s open-source pivot makes Chinese AI seem almost unstoppable. The commoditization of AI is accelerating, and China’s tech giants are redrawing the battle lines with the West from a performance race into a price war.

Here’s why Baidu’s open-source play matters more than DeepSeek's January breakthrough.

Deep Pockets

While DeepSeek was, at least in appearance, a proof of concept from a scrappy startup, Baidu brings the institutional weight, capital firepower, and, crucially, the distribution channels, to ensure widespread adoption.

In the age of AI, it pays to be a big company. Their sheer scale enables larger investments, improves resilience against market shocks and trade sanctions, and offers hyperscalers the advantage of applying this technology to their existing products.

Just look at Google. The American search engine giant is leveraging its existing broad and loyal customer base to attract traffic to its Gemini models, which are being integrated into its search function. At the May Google I/O conference, Google announced that its ‘AI Overviews’ (the AI summaries displayed next to search results) are used by more than 1.5 billion people each month. Baidu will doubtless do the same, leveraging its economies of scale to make Ernie a winner.

Meanwhile, DeepSeek’s momentum has sputtered. The smaller tech outfit has been forced to delay the release of its next-generation R2 model after struggling to procure enough of Nvidia’s high-end graphics processing units to complete training, because of new U.S. sanctions on chip exports to China, according to The Information.

As it stands, Baidu’s Ernie API has an 18% market share, still some way behind DeepSeek’s 34% share; however, Baidu’s size means it can rapidly make up ground.

From Performance Race to Price War

China is commoditizing AI faster than the West can monetize it. When developers can access high-performance AI without Big Tech’s pricing gates, it fundamentally rewrites who can afford to innovate.

DeepSeek-R1 debuted at $0.55 input/$2.19 output, undercutting the then SOTA model o1 by 90%+ on output token pricing. Since then, reasoning model prices have cratered, with OpenAI recently cutting its flagship model price by 80%, according to SemiAnalysis. Baidu already said in March that its Ernie X1 model delivers performance on par with DeepSeek-R1 “at only half the price.”

“If open-source AI becomes just as powerful as proprietary US models,” wrote June Yoon, Asia Lex Editor of the Financial Times in an article earlier this year, “the ability to monetize AI as an exclusive product collapses. Why pay for closed models if a free, equally capable alternative exists?”

The West may be forced to rethink its pricing strategies and business models when China can release equally good AI at virtually no cost.

Developer Ecosystem

China could never compete when the U.S. is making deals the size of Stargate, which is set to rise to $500 billion, and imposing tariffs on the tools used to build AI. That’s why China was forced to become self-sufficient.

By open-sourcing AI, they can bypass U.S. sanctions, decentralize development, and access global talent to improve models. Restrictions on Nvidia’s chips matter less when the rest of the world can refine China’s models on alternative hardware.

This approach may render restrictions moot, as much development occurs on Nvidia infrastructure outside China. By publishing the code of its flagship LLM, Baidu aims to foster broader adoption and a developer community around the technology.

CEO Li said as much to Chinese developers in April this year: “Our releases aim to empower developers to build the best applications — without having to worry about model capability, costs, or development tools.”

More developers using Ernie’s code will help it scale to new heights.

Game on

As more Chinese tech giants embrace open-source AI, they’re creating an alternative technology stack that bypasses U.S. control. Every developer who builds on Ernie instead of GPT-4 is one less customer for Silicon Valley and one more node in China’s AI ecosystem.

While DeepSeek faces delays and hardware shortages, established players like Baidu can take on the baton and sustain the open-source assault indefinitely. They have the capital to subsidize free AI and the scale to support millions of users.

For Western tech companies clinging to closed models and subscription fees, the message is clear: the game has changed. Adapt or become irrelevant.

We list the best IT Automation software.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

Artificial intelligence has firmly established itself as a cornerstone of modern business transformation. From enhancing customer service to streamlining operations and uncovering insights from data, AI’s role in the enterprise is no longer experimental - it is essential.

For many organizations, this journey has been powered by cloud-based infrastructure. The cloud has allowed businesses to quickly scale AI tools without investing in specialist hardware, accelerating adoption and innovation. However, as AI moves from tactical deployments to deeper integration across operations, cracks are beginning to show in the cloud-first model.

Latency, Cost and Compliance Risks

Real-time responsiveness is critical to the success of AI applications, especially for tasks that require instant decision making. Cloud processing, by its nature, depends on the transmission of data across networks, introducing latency that can affect performance. In sectors like finance, healthcare or manufacturing, even small delays can make a meaningful difference.

At the same time, the financial burden of running large-scale AI models in the cloud is growing. What was once a cost-efficient way to experiment with AI can become expensive at scale. As models increase in complexity and the demand for high-volume data processing grows, many businesses are finding that cloud service costs escalate quickly and sometimes unpredictably.

Then there is the issue of data governance. With regulations tightening globally, organizations are under pressure to ensure data remains within secure and compliant environments. Cloud-based AI often involves the transfer and storage of sensitive information on third-party servers, which can create challenges around data sovereignty and auditability.

Why On-Device AI Is Gaining Ground

These limitations are driving renewed interest in on-device AI as a model that processes data locally on endpoint devices such as PCs or mobile workstations. Rather than relying on remote servers, these devices are equipped with dedicated capacity to run machine learning models directly, offering a range of operational and strategic benefits.

Recent research from HP and YouGov, surveying over 1,000 UK workers and senior IT decision makers, underlines this shift. More than half of business leaders indicated they would be more likely to adopt AI if it were embedded directly into employees’ devices. And nearly a third of employees said they would be more likely to use AI if it were seamlessly integrated into the software they already rely on each day.

This reflects a broader theme: workers want AI that is immediate, intuitive and trustworthy. On-device AI addresses all three, offering speed, convenience and stronger data control.

Improving Adoption Through Experience

AI’s potential is not just about what it can do, but how it is experienced by users. When embedded within familiar tools - summarizing meetings, suggesting email phrasing, or analysing data on the fly - AI becomes a natural part of the workflow, not another task to manage. This frictionless integration drives deeper engagement and delivers sustained, tangible value.

By processing data locally, on-device AI also reduces exposure to external networks, helping organizations to meet stringent data protection requirements. In regulated industries, this ability to maintain control over sensitive data without compromising functionality is a powerful enabler of compliance.

Scaling AI Requires a Hybrid Future

Despite AI’s growing presence, widespread enterprise adoption remains limited. Only around 11 per cent of organizations have scaled AI across departments. Yet among those who have, the benefits are tangible. According to HP and YouGov’s survey, one in three report cost savings of over 11 per cent, illustrating that the challenge is not in proving AI’s value, but in deploying it effectively.

Going forward, AI strategy will increasingly demand a hybrid approach that combines the scalability of cloud with the performance, security and personalization of on-device intelligence.

Business leaders must evaluate not just where AI runs, but how it fits into the fabric of their organization. The goal should be to deliver intelligence where it creates the most value, in a way that enhances employee experience, strengthens compliance and makes operations more resilient.

Cloud-based AI helped prove the value of intelligent systems. On-device AI may well be the model that ensures that value is sustained.

We list the best business computers.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

In the rush to capitalize on Generative AI, software development and delivery has shifted into overdrive. Teams are moving faster, delivering more code, and automating everything from testing to deployment. In many ways, it’s a golden age for software productivity. But beneath the surface, a growing problem threatens to undo those gains: software quality isn’t keeping up.

Call it the ‘AI speed trap’. The more we trust AI to ship code autonomously without rigorous due diligence, the wider the quality gap becomes. And the consequences are already visible. Outages, security breaches, mounting technical debt, and, in some cases, millions in annual losses as the result of business disruption.

In fact, recent research shows that two-thirds of global organizations are significantly at risk of a software outage within the next year, and almost half believe poor software quality costs them $1m or more annually. There is an emerging tension in AI-driven software development: speed vs. stability.

Faster doesn’t always mean better

Modern DevOps and Continuous Integration/Continuous Delivery (CI/CD) pipelines were built to prioritize velocity; GenAI has turbocharged this further, creating more code than ever before.

But AI doesn’t ensure quality; it ensures output. We all know that AI can get it wrong. Without proper guardrails, AI-powered development becomes like a high-speed factory churning out code without accountability. So why are so many teams pushing code live without fully testing it? Because the pressure to deliver quickly outweighs the mandate for due diligence.

That’s not just anecdotal. The 2025 Quality Transformation Report found that nearly two-thirds of organizations admit to releasing untested code to meet deadlines. It’s a staggering statistic, and a stark warning.

The new definition of ‘quality’

Traditional metrics like test coverage, defect rates, or system stability used to define quality. Today, speed is starting to stand in for quality, but it’s a dangerous substitution. Shipping faster doesn’t mean shipping better.

If quality becomes synonymous with velocity, teams risk ignoring deeper indicators, including resilience, maintainability, and customer experience. And when those things fail, the fallout can be major: lost revenue, compliance failures, or service outages that damage trust.

Software quality must be redefined for the AI-first world. It’s not just about finding bugs, it's about ensuring long-term performance, user satisfaction, and business continuity. In this landscape, quality is less about the absence of errors and more about the presence of confidence.

When confidence is missing

Here’s the paradox: even as organizations accelerate releases, many teams hesitate internally. Over 8 in 10 (83%) of EMEA IT teams (as well 73% in the US) say they delay launches because they aren’t confident in their test coverage. The disconnect between external pressure to move fast and internal uncertainty about product stability is a symptom of broken feedback loops and incomplete visibility.

Worse still, misalignment between leadership and delivery teams creates confusion about what quality even means. While C-suite leaders push for speed and innovation, engineering teams struggle to maintain test rigor under shrinking timelines and budgets.

This breakdown isn’t just a technical issue; it’s a cultural one. To fix it, organizations need stronger alignment around goals, clearer quality metrics, and smarter automation that doesn't just accelerate work but elevates it.

AI needs to be accountable

Trust in AI is growing, and for good reason. Used well, it can offload repetitive tasks, help developers ship faster, and even make autonomous release decisions, with nine in 10 tech leaders backing its judgment. But handing over the reins to AI doesn’t mean humans should abdicate oversight.

Autonomous AI agents making release decisions may boost productivity, but without transparency, explainability, and traceability, they can also introduce risk at scale. Responsible AI use in development means embedding governance into automation. It means having a way to audit what AI did, and why.

This starts with AI-literate teams. Developers and testers need to understand the logic behind AI-generated outputs, not just blindly accept them. Ethical awareness, systems thinking, and contextual judgment must be part of every team’s toolkit if AI is going to serve as a true partner in quality.

Closing the quality gap

If software engineers want sustainable gains from AI, leaders need to clearly define what quality means for their teams, what level of risk is acceptable within their business and build that into testing strategies from day one.

The quality gap won’t close with more speed, but with smarter systems. This means investing in autonomous software testing and quality intelligence not as an afterthought, but as a strategic function.

By leveraging AI-driven insights and real-time automation, it’s possible to proactively identify risks, eliminate bottlenecks, and embed quality throughout the software development lifecycle. This enables teams to deliver at speed without compromising reliability.

It also requires a return to fundamentals: clear requirements, continuous feedback, and cross-functional accountability. These aren’t outdated concepts, they’re the foundation for any resilient development practice. In short: if AI is the engine, quality must be the brakes and the steering.

A smarter, more balanced future

AI has given us the ability to build and deploy software at unprecedented speed. But if we don’t pair that speed with intelligent quality engineering, the risks will outpace the rewards. The future belongs to organizations that move fast and stay resilient.

That means building AI-augmented testing into every stage of the software lifecycle. It means defining quality not by how fast you ship, but how confident you are that your software can perform in the wild.

It means treating AI as a tool, not a shortcut. Because in the race to deliver, the real winners won’t just be the first to cross the finish line. They’ll be the ones who don’t crash on the way.

We've listed the best Large Language Models (LLMs) for coding.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

There’s been an era-defining race underway between two types of batteries used in electric vehicles: lithium batteries that use cobalt, and ones that use iron phosphate. Cobalt, a metal with a checkered human rights record, has been in the lead. Until recently. 

Henry Sanderson’s book on the elements that build electric vehicles is Volt Rush: The Winners and Losers in the Race to Go Green.

Related episodes: 
The race to produce lithium 
How batteries are already changing the grid 
How batteries are riding the free market rodeo in Texas 
How EV batteries tore apart Michigan (Update) 
Batteries are catching fire at sea 

For sponsor-free episodes of The Indicator from Planet Money, subscribe to Planet Money+ via Apple Podcasts or at plus.npr.org. 

Fact-checking by Sierra Juarez. Music by Drop Electric. Find us: TikTok, Instagram, Facebook, Newsletter. 

• Dates: Thursday, August 21 – Sunday, August 24
• Watch for FREE on BBC iPlayer (UK)
• Use NordVPN to watch any stream

Comprising twin festivals taking place 190 miles apart, one at Richfield Avenue in Reading, the other at Bramham Park near Leeds, R&L isn't for the faint-hearted. Travis Scott, Bring Me the Horizon, Hozier and Chappell Roan are headlining both events, so read on as we explain how to watch Reading & Leeds 2025 streams online from anywhere and for FREE on BBC iPlayer.

As the unofficial annual post-exam results blowout for teenagers up and down the UK, there tends to be an end-of-days feel to Reading & Leeds, while the twin format and north-south divide are a recipe for one-upmanship.

The audience energy is just as big a part of the music, but we have seen a huge shift from the rock heavy bands of the early 2000s. That was of course since the arrival of 50 Cent as a headliner back in 2004, who among others like Panic! at the Disco, Good Charlotte, Twenty One Pilots, My Chemical Romance and Meat Loaf have been given particularly short shrift in the past.

Aside from the headliners, further highlights include Lambrini Girls, Wunderhorse, Bloc Party, Limp Bizkit and The Kooks. The Reading Friday lineup and the Leeds Saturday lineup are the same, as are the Reading Saturday lineup and the Leeds Sunday lineup, and the Reading Sunday lineup and the Leeds Friday lineup.

Read on as we explain how to watch Reading & Leeds live streams from anywhere. We've also listed the full lineups further down the page.

Can I watch Reading & Leeds for FREE?

Yes. Viewers in the UK can watch Reading & Leeds festival for free on the BBC iPlayer streaming service.

Not at home? Use a VPN to access you usual streaming services from abroad and watch Reading & Leeds for free.

How to watch Reading & Leeds 2025 from abroad

For those away from home looking to watch Reading & Leeds, you’ll be unable to tune in due to regional restrictions. Luckily, there’s an easy solution.

Downloading a VPN allows you to stream geo-blocked services online, no matter where you are. It's a simple bit of software that changes your IP address, meaning that you can access on-demand content or live TV just as if you were at home. Our favorite is NordVPN.

Use one of the best VPNs to watch Reading & Leeds 2025 from anywhere:

Editors Choice

NordVPN – get the world's best VPN
We regularly review all the biggest and best VPN providers and NordVPN is our #1 choice. It unblocked every streaming service in testing and it's very straightforward to use. Speed, security and 24/7 support available if you need – it's got it all.

The best value plan is the two-year deal which sets the price at $3.39 per month, and includes an extra 3 months absolutely FREE. There's also an all-important 30-day no-quibble refund if you decide it's not for you.

- So, try NordVPN 100% risk-free for 30 daysVIEW DEAL ON

How to watch Reading & Leeds in the UK

The best place to watch Reading & Leeds unfold is BBC iPlayer, which is hosting a dedicated Reading & Leeds popup channel. It will run from 2pm BST until late on Friday, Saturday and Sunday, showcasing performances from Reading's Main Stage and Chevron Stage.

TV coverage is limited to highlights of the Chappell Roan and Bring Me The Horizon Reading sets, which will air on BBC One at 10.40pm on Friday and 11.55pm on Saturday, respectively. An overall highlights show will air on BBC One at 11.40pm on Thursday, August 28.

Outside the UK? If you want to watch Reading & Leeds 2025 on BBC iPlayer whilst traveling abroad you'll need to download NordVPN.

Can I watch Reading & Leeds 2025 in the US, Canada and Australia?

You won't see Reading & Leeds in any international listings, but snippets of the festival will be posted on the BBC Music YouTube channel.

Brit abroad? Use a VPN to watch Reading & Leeds on BBC iPlayer from abroad.

Reading 2025 lineup

(All times BST)

Thursday, August 21

Chevron Stage

• 9:15pm – Cam Smith
• 10:45pm – Jack Marlow
• 12:15am – Charlie Tee
• 1:45am – DJ Battle

Festival Republic Stage

• 22:30pm – Jay Knox’s Jams
• 12am – Hun FM
• 1:30am – Swiftogeddon

Friday, August 22

Main Stage

• 12pm – Red Rum Club
• 12:50pm – Alessi Rose
• 1:45pm – The Royston Club
• 2:45pm – Bloc Party
• 4:10pm – Wallows
• 5:20pm – The Kooks
• 7:10pm – Chappell Roan
• 9:40pm – Hozier

Chevron Stage

• 12:25pm – Good Health Good Wealth
• 1pm – Charlotte Plank
• 1:50pm – Badger
• 2:45pm – Nemzzz
• 3:50pm – Still Woozy
• 4:55pm – Soft Play
• 6:10pm – Rudim3ntal
• 8:40pm – AJ Tracey
• 11:30pm – C100
• 1am – Badger
• 2am – DJ Battle

Festival Republic Stage

• 1pm – Any Young Mechanic
• 1:50pm – Glixen
• 2:40pm – Jasmine.4.T
• 3:30pm – Been Stellar
• 4:20pm – Sofia Isella
• 5:10pm – Snayx
• 6pm – The Linda Lindas
• 7:40pm – Heartworms
• 8:30pm – Mannequin Pussy
• 9:35pm – High Vis
• 10:40 – The Chats
• 12am – Pop Never Dies
• 1:30am – Fuzz Club

BBC Introducing Stage

• 12pm – Girl Group
• 12:55pm – Enna
• 1:50pm – Charles
• 2:45pm – Phoebe Green
• 4:35pm – Artio
• 5:30pm – Overpass
• 6:25pm – Jack Dean
• 7:20pm – Ashley Singh
• 8:15pm – Chanel Yates

The Smirnoff Stage

• 1:15pm – Djbe
• 2:45pm – Sofa
• 4:15pm – Stresshead
• 5:45pm – Champion
• 7:15pm – Lu.Re
• 8:45pm – Annabel Stop It
• 10pm – Zero
• The Aux Stage
• 1pm – Climate Live Takeover
• 3pm – Brittany Broski in conversation with Max Balegde

Saturday, August 23

Main Stage

• 12pm – VOILÁ
• 12:50pm – Lambrini Girls
• 1:45pm – South Arcade
• 2:45pm – Good Neighbours
• 3:55pm – Royal Otis
• 5:10pm – Conan Gray
• 6:30pm – Enter Shikari
• 7:55pm – Limp Bizkit
• 9:50pm – Bring Me The Horizon

Chevron Stage

• 12pm – James And The Cold Gun
• 12:45pm – Blanco
• 1:30pm – Issey Cross
• 2:15pm – Pozer
• 3:05pm – Example
• 4:15pm – Pale Waves
• 5:30pm – Bakar
• 6:35pm – Wunderhorse
• 7:35pm – Jazzy
• 8:50pm – Becky Hill
• 1am – Jeremiah Asiamah
• 2am – DJ Battle

Festival Republic Stage

• 12:25pm – Sunday (1994)
• 1:15pm – Rifle
• 2:05pm – Mouth Culture
• 2:55pm – Origami Angel
• 3:45pm – Ecca Vandal
• 4:35pm – Vlure
• 5:25pm – Balming Tiger
• 6:15pm – House Of Protection
• 7:20pm – Bilmuri
• 9:10pm – Snow Strippers
• 12am – Uprawr

BBC Introducing Stage

• 12pm – Mudi Sama
• 12:55pm – Keo
• 1:50pm – Eville
• 2:45pm – Finn Forster
• 4:35pm – V.I.C
• 5:30pm – Lleo
• 6:25pm – The Pill
• 7:20pm – Unpeople
• 8:15pm – Nxdia

The Smirnoff Stage

• 1pm – Johnnie Hartmann
• 2:30pm – Lleahdavies
• 4pm – Driia (DJ)
• 5:30pm – Bvnquet
• 7pm – Omar+
• 8:30pm – Megra
• 10pm – In Parallel

The Aux Stage

• 1pm – Rock Revival
• 3pm – Old & Bald
• 5pm – Jaackmaate’s Happy Hour

Sunday, August 24

Main Stage

• 12:50pm – Demae
• 1:40pm – Songer
• 2:30pm – Waterparks
• 3:25pm – Sea Girls
• 4:20pm – Suki Waterhouse
• 5:30pm – Amyl & The sniffers
• 6:45pm – Trippie Redd
• 8:05pm – D-Block Europe
• 9:50pm – Travis Scott

Chevron Stage

• 12pm – Lyvia
• 1:35pm – Late Night Drive Home
• 2:45pm – Del Water Gap
• 3:20pm – Leigh-Anne
• 4:30pm – Girl’s Don’t Sync
• 5:45pm – Lancey Foux
• 7pm – DJ EZ
• 8:40pm – Sammy Virji
• 11:30pm – BL3SS
• 1am – Millie Cotton
• 2am – DJ Battle

Festival Republic Stage

• 12:10pm – Aviva
• 1pm – Bartees Strange
• 1:50pm – Balu Brigada
• 2:40pm – Nell Mescal
• 3:30pm – Skye Newman
• 4:20pm – Antony Smierzek
• 5:20pm – Luvcat
• 6:10pm – Matilda Mann
• 7pm – Good Kid
• 8:05pm – Nieve Ella
• 9:10pm – The Dare
• 12am – For The Nightcrawlers
• 1:30am – Face Down

BBC Introducing Stage

• 12pm – DJ Cliffe
• 12:55pm – Cliffords
• 1:50pm – Wench!
• 2:45pm – Indoor Foxes
• 4:35pm – Niki Kini
• 5:30pm – Mcxxne
• 6:25pm – EV
• 7:20pm – Deyyess
• 8:15pm – Amie Blue

The Smirnoff Stage

• 1pm – Aki Oke
• 2:30pm – LILI
• 4pm – Charlie Boon
• 5:30pm – Disrupta
• 7pm – Ordley
• 8:30pm – ODF
• 10pm – Riordan

The Aux Stage

• 5pm – Josh & Moyo

Leeds 2025 lineup

(All times BST)

Thursday, August 21

Chevron Stage

• 11pm – Joe Depz
• 12:30am – Lens
• 2am – DJ Battle

Festival Republic Stage

• 7pm – Ben Ellis
• 7:50pm – Jo Hill
• 8:35pm – Freak Slug
• 9:25pm – Chloe Qisha
• 10:20pm – Mazza_I20

Reload Stage

• 7pm – Lewis Taylor
• 8:15pm – Bullet Tooth

LS23

• 10pm – Dr Dubplate
• 12:15am – DJ Semtex
• 1:45am – Anaïs

Friday, August 22

Main Stage

• 12:20pm – Demae
• 1:10pm – Songer
• 2:00pm – Waterparks
• 2:55pm – Sea Girls
• 3:30pm – Suki Waterhouse
• 5:00pm – Amyl & The Sniffers
• 6:15pm – Trippie Redd
• 7:35pm – D-Block Europe
• 9:20pm – Travis Scott

Chevron Stage

• 12pm – Lyvia
• 1:30pm – Late Night Drive Home
• 2:20pm – Del Water Gap
• 3:15pm – Leigh-Anne
• 4:25pm – Girl’s Don’t Sync
• 5:35pm – Lancey Foux
• 6:45pm – DJ EZ
• 8:10pm – Sammy Virji
• 11:00pm – BL3SS
• 12:30am – Millie Cotton
• 1:45am – DJ Battle

Festival Republic Stage

• 12pm – Aviva
• 12:45pm – Bartees Strange
• 1:34pm – Balu Brigada
• 2:25pm – Nell Mescal
• 3:15pm – Skye Newman
• 4:05pm – Antony Smierzek
• 5:05pm – Luvcat
• 5:55pm – Matilda Mann
• 6:45pm – Good Kid
• 7:40pm – Nieve Ella
• 8:40pm – The Dare

BBC Introducing Stage

• 12pm – DJ Cliffe
• 12:55pm – Cliffords
• 1:50pm – Wench!
• 2:45pm – Indoor Foxes
• 4:35pm – Niki Kini
• 5:30pm – Mcxxne
• 6:25pm – EV
• 7:20pm – Deyyess
• 8:15pm – Amie Blue

Reload Stage

• 12:30pm – CamUKG
• 2pm – SHADEV
• 3:30pm – FUZION
• 5pm – ESC
• 6:30pm – Sophia Violet
• 9:30pm – Riordan

LS23

• 10pm – ESC
• 11pm – Douvelle19
• 12:15am – P-rallel
• 1:45am – Silva Bumpa

Saturday, August 23

Main Stage

• 12pm – Red Rum Club
• 12:40pm – Alessi Rose
• 1:30pm – The Royston Club
• 2:25pm – Bloc Party
• 3:50pm – Wallows
• 4:55pm – The Kooks
• 6:40pm – Chappell Roan
• 9:10pm – Hozier

Chevron Stage

• 12:00pm – Good Health Good Wealth
• 12:35pm – Charlotte Plank
• 1:25pm – Badger
• 2:15pm – Nemzzz
• 3:20pm – Still Woozy
• 4:25pm – Soft Play
• 5:40pm – Rudim3ntal
• 8:10pm – AJ Tracey
• 12:30pm – C100
• 1:45am – DJ Battle

Festival Republic Stage

• 12:35pm – Any Young Mechanic
• 1:25pm – Glixen
• 2:15pm – Jasmine.4.T
• 3:05pm – Been Stellar
• 3:55pm – Sofia Isella
• 4:45pm –Snayx
• 5:35pm – The Linda Lindas
• 7:15pm – Heartworms
• 8:05pm – Mannequin Pussy
• 9:05pm – High Vis
• 10:10 – The Chats

BBC Introducing Stage

• 12pm – Girl Group
• 12:55pm – Enna
• 1:50pm – Charles
• 2:45pm – Phoebe Green
• 4:35pm – Artio
• 5:30pm – Overpass
• 6:25pm – Jack Dean
• 7:20pm – Ashley Singh
• 8:15pm – Chanel Yates

Reload Stage

• 2pm – Scruz
• 3:30pm – Tommy Villiers
• 5pm – Daisy
• 6:30pm – Saint Ludo
• 8pm – T-Lex
• 9:30pm – Gentlemens Club

LS23

• 10pm – Daisy
• 11:15pm – Badger
• 12:30am – TBA
• 1:45am – Disrupta

Sunday, August 24

Main Stage

• 12pm – VOILÁ
• 12:45pm – Lambrini Girls
• 1:35pm – South Arcade
• 2:25pm – Good Neighbours
• 3:25pm – Royal Otis
• 4:40pm – Conan Gray
• 6pm – Enter Shikari
• 7:25pm – Limp Bizkit
• 9:20pm – Bring Me The Horizon

Chevron Stage

• 12pm – James And The Cold Gun
• 12:35pm – Blanco
• 1:15pm – Issey Cross
• 1:55pm – Pozer
• 2:45pm – Example
• 4:05pm – Pale Waves
• 5:10pm – Bakar
• 6:15pm – Wunderhorse
• 7:10pm – Jazzy
• 8:20pm – Becky Hill
• 12:30am – Jeremiah Asiamah
• 2am – DJ Battle

Festival Republic Stage

• 12pm – Sunday (1994)
• 12:50pm – Rifle
• 1:40pm – Mouth Culture
• 2:30pm – Origami Angel
• 3:20pm – Ecca Vandal
• 4:10pm – Vlure
• 5pm – Balming Tiger
• 5:50pm – House Of Protection
• 6:50pm – Bilmuri
• 8:40pm – Snow Strippers

BBC Introducing Stage

• 12pm – Mudi Sama
• 12:55pm – Keo
• 1:50pm – Eville
• 2:45pm – Finn Forster
• 4:35pm – V.I.C
• 5:30pm – Lleo
• 6:25pm – The Pill
• 7:20pm – Unpeople
• 8:15pm – Nxdia

Reload Stage

• 2:30pm – Megan Wroe
• 4pm – Tomike
• 5:30pm – Gee Lee
• 7pm – Omar+
• 8:30pm – Bushbaby
• 10pm – Dj Jackum

LS23

• 10pm – Auramatic
• 11:15pm – n4tee
• 12:30am – MPH
• 2am – [IVY]

What devices is BBC iPlayer available on?

You can use BBC iPlayer on all of the following devices and platforms:

Amazon Fire (Tablets, Cube, Stick, TVs)
Android TV (please note: some models aren’t supported)
Android (Mobile & Tablet) - Android 7.0 and above
Apple TV (tvOS 14 or later)
Google TV
Freely
Freesat (please note: some models aren’t supported)
Freeview Play (TVs and set-top boxes)
iOS (iPhone & iPad) - iOS 14 and above
LG Smart TVs (2016-2024)
NOW Smart Sticks and Boxes (minimum firmware v11.5.0)
PlayStation (PS4 and PS5)
Roku (Stick & Roku-OS powered TVs, minimum firmware v11.5.0)
Samsung Smart TVs (2017 and above - 2016 models only offer on demand TV)
Sky Q, Sky Glass and Sky Stream
Virgin Media (360, Stream, TiVo)
YouView (BT, Humax, Sony, TalkTalk)
Xbox (One, Series X, Series S)

We test and review VPN services in the context of legal recreational uses. For example:1. Accessing a service from another country (subject to the terms and conditions of that service).2. Protecting your online security and strengthening your online privacy when abroad.We do not support or condone the illegal or malicious use of VPN services. Consuming pirated content that is paid-for is neither endorsed nor approved by Future Publishing.

• Mullvad will remove OpenVPN support from January 15, 2026
• The move comes as the VPN provider looks to prioritize WireGuard
• Mullvad has also announced the arrival of in-app updates for Windows and Mac

Mullvad, the Sweden-based VPN provider known for its strong privacy focus and flat-rate pricing, will soon be removing OpenVPN support in favor of WireGuard.

The change will take effect on January 15, 2026, providing existing Mullvad subscribers with time to make the switch before OpenVPN servers and support are removed.

Already the default protocol for all Mullvad apps, WireGuard is employed by many of the best VPN services on the market.

Reminder about removing OpenVPN: we are removing support for OpenVPN entirely on 15th January 2026, in six months time.OpenVPN servers will be completely removed on this date, and support for it within our app will disappear.Please reconfigure your configurations to use…August 18, 2025

"We are removing support for OpenVPN entirely on 15th January 2026, in six months time. OpenVPN servers will be completely removed on this date, and support for it within our app will disappear," wrote the provider in an announcement shared on X.

With Mullvad’s Android and iOS apps already supporting WireGuard exclusively, the transition only impacts desktop users who will receive reminders within Mullvad’s Windows, Mac, and Linux apps.

Mullvad has reminded subscribers to adjust existing OpenVPN configurations to avoid potential interruptions to the service.

Although the news may come as a surprise, Mullvad co-founder Fredrik Strömberg wrote that "WireGuard is the future" as early as 2017.

Why WireGuard?

(Image credit: WireGuard)

A VPN protocol is a set of rules and algorithms that dictate how your device encrypts, tunnels, and transports internet traffic through a VPN server.

WireGuard, a newer VPN protocol first launched in 2016, has been adopted by major VPN providers including Surfshark, IPVanish, and Proton VPN. NordLynx, the proprietary protocol of NordVPN, uses a custom implementation of WireGuard.

Mullvad’s decision to drop OpenVPN marks a significant shift, as most VPN providers still offer multiple protocols for flexibility across networks and devices. As Jan Jonsson, Mullvad CEO, told TechRadar’s Chiara Castro in November 2024, less than 7% of Mullvad users are using OpenVPN.

“By moving to a single protocol, we will be able to focus our resources where they can make a difference”, said Mullvad in a November 2024 blog post.

WireGuard has just 4,000 lines of code, compared to more than 70,000 lines for OpenVPN. Mullvad has praised WireGuard’s simple design, which, aside from making for easier integration of new features, allows for faster speeds. This doesn’t come at the cost of security either, with Mullvad citing WireGuard’s “state-of-the-art cryptography.”

Such is its belief in WireGuard, Mullvad made “a sizable donation” in 2017 to support the open-source protocol’s development. To make things easier for subscribers looking to make the switch, Mullvad’s website provides guides on how to use WireGuard.

Keeping up with Mullvad’s updates is now easier than ever

Mullvad’s decision to retire OpenVPN and fully focus on WireGuard isn’t its only recent change. The VPN provider has also announced in-app updates for desktop.

Instead of having to visit Mullvad’s website to download and install the latest updates, Windows and macOS users can now view and download them directly within the app, making it quicker and easier than ever to stay up-to-date.

Mullvad states that Linux users can also benefit from automatic updates via its Linux app repositories and provides instructions to make the process more straightforward.

You might also like

• ExpressVPN adds WireGuard to support industry transition to post-quantum VPNs
• You can now verify if your Mullvad VPN app is legit
• Mullvad VPN promises to be even better at protecting you against AI surveillance

Thirty years ago, park rangers reintroduced grey wolves into Yellowstone National Park. They wanted to restore the ecosystem and get the elk population, which had decimated the plant community, in check. And it worked – or so the popular narrative suggests. But is it really so simple? Today on the show, we explore how the Yellowstone ecosystem has changed since wolves returned and whether those changes can really be pinned solely on wolves. Plus, how the narrative of the Yellowstone wolf legacy could affect wolf reintroduction elsewhere.

Curious about other science controversies? Email us at shortwave@npr.org.

Listen to every episode of Short Wave sponsor-free and support our work at NPR by signing up for Short Wave+ at plus.npr.org/shortwave.

(Image credit: mtnmichelle)

Earlier this year, the 4 Day Week Foundation announced that 200 UK companies signed up for a permanent four-day working week with employees on full pay. With three new pilot programs planned in 2025 to further explore the benefits and challenges of a reduced working week, there is growing momentum behind the concept. However, whilst adoption can be observed across various sectors and regions, large enterprises have notably remained cautious.

It’s important to consider what's holding back Britain's biggest businesses from embracing this shift to more flexible work. At the heart of their hesitation are concerns about alignment, productivity, and the fear that fewer working days could mean reduced momentum. However, businesses cannot tackle this on their own and AI powered tools can help alleviate many of these concerns and obstacles.

AI's potential to cut out mundane work

A modern approach to work doesn't necessarily mean working less, it means working smarter. A central component to this is using AI tools to streamline activities, boost productivity and take on the heavy lifting of repetitive tasks that drain time and focus. The Asana Work Innovation Lab found that 67% of UK workers use AI at least weekly for their individual work, with 37% using it daily.

Embracing flexible working arrangements doesn’t have to mean sacrificing efficiency. AI can substantially alleviate these concerns by automating repetitive, mundane tasks such as data entry, preliminary research and report generation. Instead, AI frees up employees to focus on high-value activities offsetting the perceived loss of a working day.

Moreover, AI can help ensure continuity in projects and operations by providing instant access to consolidated, real-time data and predictive analytics. With these tools, businesses can remain agile and aligned, regardless of a shortened week.

To achieve this impact, AI must be embedded where work happens, within business-critical workflows. Only then can AI function alongside humans as an autonomous team member, meaningfully reducing overwhelm and increasing productivity.

What’s holding back larger companies

Continuity becomes especially critical for large enterprises where operational consistency is non-negotiable. These organization's leaders, often grappling with teams operating from different locations, complex workflows, and deeply entrenched cultures, worry that flexible working might create more gaps than bridges.

Against this backdrop, the hesitance from leaders towards the four-day week is understandable. Leaders at large companies are tasked with maintaining continuous operations, consistency across departments, and productivity levels. Losing a traditional workday can appear disruptive, potentially causing alignment issues or slowing critical projects. This is why, rather than being a "nice to have," AI should be seen as a “must have” in alleviating these concerns.

In large companies, senior leaders often understand the strategic value of AI, but may underestimate the practical challenges employees face in using new technologies. Without adequate training, support, and a clearly articulated vision of how AI can improve their work experience, employees struggle to integrate these tools effectively.

In contrast, when employees can clearly see the benefits AI brings to their daily tasks, adoption rates increase, leading to wider organizational efficiency and smoother workflows. This “showing, not telling” approach is a key way for senior leaders to enable their teams to move forward.

Scaling AI use across teams

Once AI is effectively deployed alongside comprehensive, accessible training programs, the next challenge is scaling this effectively across the entire organization to truly unlock flexibility without compromising productivity.

This is important because companies that have broadly integrated AI across their operations are seeing much greater returns on investment compared to those still experimenting with it in isolated teams or limited use cases. Our research found that AI ‘Scalers’ are 43% more likely to report revenue gains and 40% more likely to report boosted productivity.

Despite these evident advantages, there remains a striking lack of AI use beyond simple individual experimentation. Research shows nearly half of AI workflows are built for individual use, driving only 6% of AI adoption by colleagues and peers. Furthermore, data suggests that AI adoption is trapped in a ‘leadership bubble’, with senior leaders being 66% more likely to be early AI adopters than their employees. The net result is as many as 67% of companies fail to scale AI tools effectively across their organizations.

Before companies can truly scale AI within their organization, they must first examine how teamwork happens. If teams are operating in silos, workers are more likely to continue using AI for solo use rather than unlock AI use within teams – and crucially, across different team functions, where we are seeing the strongest impact.

The majority of AI workflows being designed for individual use needs to change if AI is to achieve widespread adoption and add tangible value. Companies that achieve this transition will be best positioned to offer initiatives like the four-day week without sacrificing alignment across teams.

The road to enhanced work is built on smarter tools

The desire for flexible working is clear, with our research showing that currently only 16% of British workers stick to the traditional 9-5 schedule. The next stage is overseeing how a four-day week can be unlocked - and the answer may just lie in AI tools.

Companies that successfully integrate AI and automation into their operations won't just sustain productivity levels, they'll boost them. Employees, freed from the burden of routine tasks, can contribute more meaningfully to their organizations.

The path toward adopting the four-day week isn't simply about cutting hours, it's about enhancing efficiency through smarter work practices enabled by AI.

We've listed the best employee management software.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

The rise of sovereign clouds has become inevitable as regulatory demands, and geopolitical pressures push enterprises to rethink where their data resides. Localized cloud environments are increasingly becoming essential, allowing organizations to keep their data within specific jurisdictions to meet compliance requirements, and provide risk mitigation.

But sovereign clouds can’t succeed without data portability, which is the ability to move data seamlessly between systems and locations. Today organizations shouldn’t wait to be pushed by regulations, they need to be ahead of the game.

Enterprises need to address the reality that data migration across hybrid environments is far from straightforward. It’s not just about relocating primary data, you also must keep it protected while considering associated datasets like backups and the information used in AI applications.

While some may need to address the protection of Large Language Model (LLM) training data, many organizations are instead turning to Retrieval-Augmented Generation (RAG) or AI agents to bring intelligence to their proprietary data without building models from scratch.

Either way, data sovereignty is a valid approach to the pressures facing organizations today, but the focus should always be on data resilience first, no matter where it's stored.

It’s a familiar tale - the cloud will give businesses more options and flexibility, but to take advantage of these properly, they’ll need some joined up thinking.

Today’s forecast

Regulators around the globe are driving organizations to look at their data differently, appearing at pace in response to increasing data globalization as countries try to get a better grasp on their data. The European Union (EU) has been particularly stringent, introducing the comprehensive General Data Protection Regulation (GDPR) that stipulates data sovereignty.

Under it, the laws of the country where data is stored or processed are now applicable to the data, regardless of where the data was originally collected. Special attention is also being paid to the chain of custody of data in the EU with both the NIS2 and DORA regulations demanding robust risk management for data, especially when held or handled by third parties.

As data, including highly sensitive and classified data, is being increasingly handled by these third parties, namely cloud providers, keeping it bound under privacy laws has become a priority for both organizations and governments as their data moves across borders.

With this increased movement of data between countries, or even continents, global instability concerns have become unavoidable, especially for governments. Some have already adopted sovereign clouds to protect their most sensitive data from potential malevolent access. And some have taken it one step further.

With cloud services completely reliant on data center infrastructure, some governments have started to divest their interest in foreign cloud and IT infrastructure, reinvesting instead in their own. This way, they can avoid storing their most sensitive data with foreign providers.

But cloud sovereignty is not a silver bullet. For those utilizing multinational cloud providers, there might be the option to stipulate where your data is ultimately stored and what countries’ laws it will be held under, but there is no guarantee that it will not change. The issue isn’t just solved by relocating the primary data.

Sure, it needs to be protected, but what about all the related data? Cloud backups and Large Language Model training data sets for example, all need to be carefully considered to meet data sovereignty – or alternatively, organizations can utilize RAG or AI agents to level up their data without having to deal with reams of AI training datasets in the first place.

Freedom of Movement

But to do all of this, organizations need to ensure that data portability is enshrined in their data resilience planning. After all, there’s a fine line between protecting your data and inadvertently restricting it beyond the point of use. If organizations are unable to ensure data portability, then moving to a hybrid cloud environment to take advantage of both sovereign clouds and localization of data storage is a non-starter.

There are SaaS (Software-as-a-Service) and DRaaS (Disaster Recovery-as-a-Service) providers that can simplify the process, but it’s not a task that can be completely offloaded to a third party. Organizations still need to take a hands-on approach, planning and managing it thoroughly to ensure data remains secure throughout the process. Otherwise, organizations will be unable to utilize sovereign clouds to adhere to the myriad data residency and sovereignty regulations.

It’s not without caveats though. For those larger, multinational organizations operating across countries and continents, multiple cloud environments will be required to house multiple sovereign clouds. But this also brings increased complexity for both the monitoring and management of data across jurisdictions.

Not only will multiple cloud environments need to be considered, but also multiple sets of data regulations across countries. And, for those organizations that do get it right, the benefit of enhanced data resilience comes with the added risk of data fragmentation.

There’s no easy win, but what’s certain is that data portability will be an essential part of any solution that organizations settle on. Whichever approach is taken, being able to move data seamlessly across platforms and clouds will be a necessity as organizations wrestle with data sovereignty. And, as regulations continue coming down the line, data portability will give organizations a head start on future compliance, allowing them to flex more easily to meet regulations, where less portable counterparts will struggle.

Tying down the cloud

Data globalization is showing no signs of slowing, with information flows now their own form of trade, even generating their own economic value. And with global instability as an ever-present factor, data sovereignty will only move higher up the priority list. But it’s not a task that can just be passed onto a third-party provider.

Although organizations may not fully realize it yet, data sovereignty and operational clarity are closely linked. To start securing your data, you need to know exactly where it’s stored and how. Then, with this comprehensive understanding of your data landscape, you can pinpoint those operations or processes where data resilience might be lacking and tackle your data portability.

By reworking data resilience from the ground-up, organizations can cement security, compliance, and sovereignty into their operations, and actively manage them through risk assessments, compliance audits, and strategies that take into account multiple suppliers.

With this in place, organizations can start leveraging hybrid cloud environments effectively, perhaps storing the most sensitive data on-premises under precise data sovereignty regulations, while offloading less critical data to the cloud.

But this can only be utilized by organizations that have prioritized data portability. Rather than waiting for regulations to enforce it, organizations need to be proactive to take advantage of the flexibility, longevity, and most importantly, security of the cloud.

We've listed the best IT management tool.

This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro

The risk from the recalled shrimp is "quite low," said Donald Schaffner, a food safety expert at Rutgers University. Cesium-137 is a byproduct of nuclear reactions.

(Image credit: Charles Krupa)

If you frequently call or visit countries outside the US, these prepaid phone plan improvements could help your travel budget.

• OpenAI is retiring ChatGPT's Standard Voice Mode in September
• Only the faster, more expressive Advanced Voice Mode will be available
• Many users are upset about the change, preferring the sound and approach of the voice OpenAI is removing

The voice people have come to associate with ChatGPT is retiring come September 9, and not everyone is happy about it. ChatGPT’s “Standard" voice is going away in favor of the “Advanced” voice option first released to a limited selection of ChatGPT users last year. Rebranded simply as “ChatGPT Voice,” it will be the only choice going forward.

The original “Standard” voice mode debuted in 2023, built on a simple pipeline: you’d speak, OpenAI’s servers would transcribe your input, generate a response using the GPT model, then read it back using a relatively neutral synthetic voice.

ChatGPT's Advanced Voice Mode is designed to respond more quickly, to be more human in its tone and manner of speech, and to generally perform at a higher level than its predecessor. Nonetheless, plenty of people think it's a mistake.

"The standard voice offers a warmth, depth, and natural connection that the advanced voice simply doesn’t match," one user wrote in a post on OpenAI's forum. The advanced voice comes across as robotic and detached, lacking the soulful and understanding tone I value."

More than one person described the new voice as less engaging to speak with. There were also complaints that the new model speaks too quickly, as if it were trying to get the interaction over with.

"Standard Voice is thoughtful and has a voice and cadence that is natural and comforting. Poignant," A Reddit user posted. "Advanced Voice doesn’t have the same characteristics, doesn’t give thoughtful answers, has restrictive content limits, and always sounds like they’re trying to rush through a mediocre response. "

Advanced voices

Even if you don't mind how the new voice sounds, some ChatGPT users are annoyed because they've found it won't even perform the same as the earlier voice.

Advanced Voice Mode integrates your voice, the AI’s responses, and its vocal expression in one real-time process. The integrated process means the AI doesn't quote the written response verbatim. Instead, it expresses ideas more conversationally, sometimes skipping phrases, condensing clauses, or adjusting tone based on context. Technically impressive, but not what some ChatGPT users want.

"The Standard voice would literally read out the exact response that ChatGPT would normally give you. It was a direct line, you know?" read one example post on Reddit. "But this new one? It sounds like it's paraphrasing or summerizing [sic] it instead. It skips over the little details and makes the whole conversation feel way more disconnected."

That might sound minor in the grand scheme of AI progress, but it echoes a broader trend in tech where people are upset when there's a big change, even if it's ostensibly an upgrade.

Not everyone dislikes the new voice option, of course. Some like its realism and speed, and how it makes for a more fluid conversation. OpenAI has promised more improvements to come, as well. But, given that complaints about GPT-4o's removal when GPT-5 debuted led to the older model's return, I wouldn't be too surprised to see the Standard Voice Mode stage a comeback too.

You might also like

• I compared ChatGPT 5’s three model options, and the results explain why people miss GPT‑4o
• ChatGPT users are still fuming about GPT-5’s downgrades – here are the 4 biggest complaints
• 'We hear you' – OpenAI’s Sam Altman responds to user concerns at a GPT-5 AMA

• Many AI models aren't as effective as they're marketed to be, report claims
• 95% of surveyed companies have seen very little impact from their LLMS
• Specialisation is the key to successful AI adoption

New research by MIT’s NANDA initiative has claimed the vast majority of GenAI initiatives attempting to drive rapid revenue growth are ‘falling flat’.

Of those sampled, 95% of companies deploying Generative AI are stalling, “delivering little to no measurable impact” on profit and loss.

It seems to be an all-or-nothing game, as the 5% of companies who are benefiting from generative AI are excelling - these are primarily, the lead author says, startups led by 19 or 20 year olds, who have seen revenues ‘jump from zero to $20 million in a year’.

GenAI tools on the rise

It seems the key to success with AI models is specialisation. Successful deployment is about picking ‘one pain point’ and executing this well, and carefully partnering with companies using tools.

Specialised vendors have success around 67% of the time, but internally built models succeed only around a third as often. Highly regulated sectors like the financial industry see many organizations build their own AI systems, but the research suggests the companies are much more prone to failure when they do so.

When line managers are empowered to drive the adoption, they see success because they are able to choose tools that can adapt over time.

Allocation is important too, as most GenAI budgets are dedicated to sales and marketing - but the biggest ROI was seen in back-office automation.

This isn’t the first time that research has suggested that AI models aren’t working as they should. A significant number of companies have introduced layoffs of lower level workers and brought in AI systems - but over half of UK businesses who replaced workers with AI regret their decision.

Tangible benefits from these models are increasingly difficult to find, and security risks linked with the models are concerning organisations - as well as AI models making ESG goals much more difficult to reach.

You might also like

• Bad news - most devs believe AI is going to kill 'most or all' of their company's marketing department
• Check out our choice for best endpoint protection software to keep you safe
• 70% of people are sick of talking to AI – where did all the humans go?

Here are the answers for The New York Times Mini Crossword for Aug. 20.

• Grammarly has debuted a collection of specialized AI tools for aiding students
• The AI agents can simulate grading, find citations, and predict reader reactions
• The tools are part of the new Grammarly Docs platform

Grammarly is looking to take a more active role in helping students write with a new set of AI agents that go well beyond simply ensuring you use semicolons correctly. The company has released a set of eight AI agents built for specific writing support, embedding them directly into a new writing platform called Docs (not the Google kind).

The new features combine AI abilities with quieter digital assistance. You don't need to write a prompt asking for specific help, just tap on the right tool, and Grammarly's AI will help find sources, predict how a professor will react to your wordplay, and make sure you don't sound like an AI yourself (RIP em-dash).

Grammarly has been augmenting its services with AI for a while, including rewriting tools and an AI chatbot last year. The new agents go beyond that reactive approach by using the context of your writing and the reason you're writing it to offer advice without you having to explain it explicitly.

Grammarly is pitching the new tools to both students and professionals, but the academic demand seems particularly dire. Students are currently torn between doing whatever it takes to succeed academically and compromising that success with unethical AI use that bypasses the actual learning.

Grammarly's bet is that students who don't want to cheat themselves with AI can use these tools to help them learn, not do it for them. They could use the AI Grader for feedback that mimics a real instructor's assessment using course-specific materials and details of what the teacher is looking for. The Citation Finder agent can check your sources and help you find better ones, formatting them properly, too. Plus, the Expert Review agent can offer domain-specific feedback on writing in fields like law and medicine, measuring arguments against professional standards.

And if you perhaps inadvertently hew too closely to a source, the Plagiarism Checker will help flag unintentional copying done when you're up late. And the AI Detector checks to make sure your fatigue hasn't made your writing seem machine-generated.

Learning for a world of AI

The tools can help students succeed in the long term, according to Grammarly, by teaching them how to research and write well without compromising ethics, even though AI tools and shortcuts are everywhere.

According to the company’s internal research, only 18% of college students feel “very prepared” to use AI professionally after graduation, despite the demand for AI literacy among employers. Grammarly wants to become the training ground for those skills without undermining academic integrity along the way. You can access all these tools in Grammarly’s new docs platform, both as a free and paid subscriber.

Of course, Grammarly’s not the only player chasing this idea. Microsoft Copilot in Word offers some similar features to Grammarly, as does Google's Gemini AI in Google Docs. But Grammarly’s approach is both more comprehensive and streamlined because of its focus on avoiding having an AI write everything for the user. That’s what might make this update stick.

Because while Grammarly could have just been another AI writing tool, it made its AI support take a step back. As imperfect as all AI tools are, at least this approach tries to address the very real crisis of people having no idea how to use AI writing aides ethically, if they even want to.

You might also like

• ChatGPT is the homework helper for more than a quarter of teens – and the trend is accelerating
• Can Grammarly's AI bloodhound sniff out text written by ChatGPT?
• 3 ways to use the new ChatGPT Study mode that aren’t for homework

Norton has long been a household name for antivirus software, yet despite launching its first VPN, Norton WiFi Privacy, in 2017, it’s never achieved the same pedigree status in the space. Its original iteration, Norton Secure VPN, offered little to shout about, though a recent revamp seems to have set the service on a much more promising path.

Although not yet rivaling the likes of NordVPN, Norton VPN offers surprisingly quick speeds, a super-friendly interface, and a great array of features you’d expect from a top VPN service. While some gaps in the service remain, it’s clear the Norton VPN team is striving to bring the same security pedigree seen on its antivirus software to its VPN, and pretty quickly, too.

We’ve put this new and improved service to the test. Both in day-to-day scenarios and more rigorous, proprietary tests, we’ve uncovered where the service has made great strides, and where it should look next to ensure it reaches the heights its aiming for, all to help you decide whether Norton VPN is the choice for you.

Features

Norton VPN’s recent revamp has seen a host of expectable but mightily important features arrive across several platforms. Plus, it continues to improve its fundamentals – even if some areas still have some work to do.

As far as VPN protocols go, Norton offers OpenVPN, WireGuard, IPSec, and Mimic, its proprietary protocol designed for obfuscation. iOS and Mac users are limited to only IPSec and Mimic, which is disappointing given OpenVPN and WireGuard’s improved capabilities – although IPSec can be fast, WireGuard offers a faster, less device-intensive experience. Luckily, wider support for these protocols is said to be in the works, though an exact release date is unknown.

Users on iOS and macOS have a limited protocol choice, with only IPSec and Mimic available (Image credit: Future)

As for extra features built into Norton VPN, there aren’t any particularly unique choices, though each is beneficial and often seen among top VPNs. Wi-Fi detection allows you to auto-connect to the VPN when connecting to public Wi-Fi. On macOS, it also auto-connects on “compromised networks”. It’s unclear how Norton decides if a network is compromised, but all wireless networks are considered public by default.

Norton also offers ad and tracker blocking. Tracking is blocked at a DNS level to reduce the risk of your data being used for targeted ads while you browse online, and can be turned on by simply toggling the option. Although the tracker appeared to work, its ‘Trackers Blocked’ counter seems to run on a delay, rather than in real time, so while we know 50 trackers were blocked across a period of our testing time, we don’t have any indication of when each was blocked.

Ad blocking is undertaken by a browser extension. Although this means another download and sign-in process, we did find the ad blocker to be effective – even if not quite as effective as dedicated ad-blocker services. Scoring 77% in our tests, it is definitely worth turning on should you use Norton VPN, though don’t expect a flawless display. NordVPN and Surfshark only achieved 84% in our latest tests, and ExpressVPN only achieved 90%, so there isn’t far for Norton VPN to go before leading VPNs in this field.

A key area of improvement has been in Norton VPN’s security-focused features. Firstly, Norton VPN offers a simple-to-use kill switch, ensuring your internet traffic stops immediately when you lose connection to the VPN, preventing you from broadcasting unencrypted traffic over public Wi-Fi networks. If you’d prefer, you can choose whether you’d like to keep your access to local devices over a LAN even when the kill switch is active. The kill switch isn’t enabled by default and requires you to dig through a few menus to set it up, but if you’re on macOS, Norton gives you an easy set of guidelines to follow so you can quickly set it up.

The new Norton VPN app houses all of its improved features in an easy-to-use design resemblant of other Norton products (Image credit: Future)

Outside the kill switch, Norton VPN has added an array of features. In April 2025 alone, Norton VPN gained Double VPN, IP Rotation, Pause VPN capabilities, and the ad-blocking browser extension mentioned above.

Double VPN is available across 8 servers. While your connection options are fixed, for example, USA via Canada, they are bi-directional, and give you access to Australia, Canada, Germany, Japan, New Zealand, Taiwan, the United Kingdom, and the United States. While it’s not completely customizable, like Surfshark’s Nexus technology allows, its 8 locations put it only two behind NordVPN, which offers 10.

IP rotation also features, something not even NordVPN can say. Only Surfshark also offers IP rotation among the best VPNs. While Norton’s offering isn’t quite as expansive, it does offer servers in the US, Japan, Canada, Australia, the United Kingdom, and Germany to use. What’s more, Norton’s offering is much simpler to access, and connects in almost no time at all, no matter the server you pick. In our testing, our IP seemed to change every few seconds, meaning it should be good enough to give you a new IP for every website you visit.

Pause VPN is the final feature worth mentioning. Overlooked by the likes of PIA and ExpressVPN, pause VPN gives you the choice to temporarily end your VPN connection for a set time, with the VPN reactivating once the time ends. This is a great tool should you briefly wish to search for something or use an app outside of the connection without risking forgetting to turn your VPN back on. Norton VPN’s implementation is extremely simple to use, and is easily accessed, though it’d be great to see a minimum pause time shorter than 15 minutes as, frankly, most times when we’ve needed it we’ve been finished in five minutes rather comfortably.

Server Network

Norton VPN’s server network isn’t up there with the very best VPNs quite yet, but it’s seen massive expansion in recent months. Since our last review, Norton VPN has expanded from servers in 29 countries to servers in 65, with 104 total locations. For comparison, Norton VPN now has more locations than Mullvad, which offers 89, and only a few less than Windscribe, which offers 112.

What’s good is the spread Norton VPN offers. 25 locations in the US are great for streaming enthusiasts, plus 6 countries in South America and 5 countries in Africa is superb given the lack of coverage these areas traditionally recieve. Asian coverage is weaker, with only 12 countries, though the most popular locations are included among these.

Given Norton VPN’s rate of increase, it wouldn’t surprise me if its country spread increased further over the coming months. If so, it’d be good to see the service fill the various gaps left in Asia and, since it already has a strong presence in South America and Africa, bulk out its presence in Europe and the US to account for key sites and services located in currently omitted regions.

Norton VPN's global coverage is especially impressive in South America and Africa, though its Asian coverage could be better (Image credit: Future)Apps

Norton’s VPN app is available for Windows, macOS, Android, and iOS. It’s also available on Apple TV and Android-based smart TVs. For Android TVs, you need version 10 or later, whereas Apple tvOS requires version 17 or later.

Notably, there’s no Fire TV or Linux support, so if you need to cover your Ubuntu or Linux Mint install, Norton won’t cut it quite yet - there’s not even support for a command-line VPN. Norton does not offer configuration files for OpenVPN or WireGuard either, so there’s no way to use Norton with an unsupported device.

It’s worth mentioning that Norton VPN’s apps have several instances of feature disparity. As highlighted already, iOS and macOS users are already limited in protocol choice, but these platforms also lack split tunneling and some auto-connect functionality. The lack of protocol choice is especially disappointing in this instance, given that other providers have offered OpenVPN and WireGuard on these platforms for a while now.

Otherwise, there’s nothing really to dislike about Norton VPN’s apps on any platform. They’re simple to look at, easy to navigate, and look so unapologetically Norton-like that anyone new to VPNs who’s familiar with its antivirus tools will quickly gain the sense of security often felt when using other Norton tools.

Ease of use

Norton VPN is a breeze to install, likely thanks to the years of experience Norton has with its other products in making complex systems simple to introduce.

Once you’re in, the experience is impressively simple, though at the expense of a couple of handy extra tools. The menus aren’t overcomplicated with features, settings, and data – though the option to set favorites or see the best servers at any time would be nice – the settings are all explained in simple enough terms for beginners, and there’s easy access to any extra tools you might have in your plan, or guides you may need to help set up your VPN connection how you’d like it.

In true Norton fashion, the experience you have is as close to identical as possible across any device you might have, too. This makes Norton VPN a superb choice should you be new to VPNs and looking for easy, quick access to the settings you need, without worrying that your usual server, connection type, or setting may be hiding somewhere new.

Even Norton VPN’s Advanced Servers, meaning its P2P-optimized, double VPN, and IP rotation optimized servers, are easily accessed. With dropdown menus giving you the information you need to understand where you’re connecting to, and any additional routing your connection might take.

Speed and performance

Norton’s speeds are its biggest area of improvement since our last round of testing. Starting with the headlines, we recorded an average speed of 909Mbps download using WireGuard from our testing server in the UK, rivalling the likes of NordVPN, ExpressVPN and Proton VPN as a result.

Our connection to the US wasn’t quite as impressive, as we only recorded speeds of 463Mbps, which is around the middle of the pack. It’s still more than enough to watch multiple 4K streams simultaneously, but it’s a pretty big dropoff compared to our initial UK tests.

How we perform speed tests

(Image credit: Shutterstock)

We test from two different virtual PCs, one in the UK and one in the US. We’re expecting big things out of NordVPN, as both of these servers have a 1 Gbs connection. You can find out more in our VPN testing methodology.

As for OpenVPN performance, Norton didn’t do as well. Our average speeds were around 260-275Mbps, whether connecting to the UK or US, which is also more than enough for most internet activities, but pales in comparison to the speeds we’ve seen elsewhere.

Our latency recordings were about what we’d expect for a top-tier VPN connecting to UK servers. In Norton’s case, 3.5ms. Most of the providers we’ve tested clock in around the 2-5ms mark, which is barely perceivable. One or two milliseconds' difference won’t make a difference to most internet apps or online gaming sessions.

As for the US connection, Norton tops our list in terms of low latency at 66.6ms, just barely beating out NordVPN to take the top spot. If you’re gaming or running a video call and want your connection to be as responsive as possible while connected to the US, Norton is the ideal solution.

Unblocking

We’ve tested Norton VPN works with them with loads of your favorite streaming services and it worked with them all! Netflix is cracking down on VPNs, but we were able to watch Netflix US, UK, Australia, Canada, and Japan all from our local testing centers. The same goes for Amazon Prime and Disney Plus, however, we weren’t able to access US Youtube.

As for regional providers, it’s a mixed bag but mostly positive results. We were able to access BBC iPlayer, ITV, Channel 4, 7Plus, and 9Now, but TVNZ+ intermittently failed, and we couldn’t access 10Play at all.

Norton VPN does have P2P capabilities, and our testing showed it works reasonably well. It’s worth mentioning that Norton doesn’t support port forwarding, which means that you won’t get incoming requests for sharing when torrenting, limiting your connectivity.

Norton only has two P2P servers, one in the Netherlands and one in Dallas. It’s a far cry away from the full P2P connectivity of a provider like PIA, but even Avast’s competing SecureLine VPN offers eight P2P servers.

On the plus side, Norton does allow you to connect automatically when you boot up a supported P2P app, but you’ll have to enable this option from the settings menu first.

Privacy and security

Norton is clearly taking the necessary steps to ensuring its VPN is secure and private. Its no-logs policy is extensive yet clear, outlining the data collected by the VPN app at any instance, your browsing data is never stored on their servers, including DNS requests, which are instead served by a private DNS server run by Norton, preventing ISP spying, and it’s had its no-logs policy audited to ensure trust.

In August 2024, VerSprite audited Norton’s policy, noting two issues that could result in sensitive user information being disclosed. Norton took the necessary steps to address these issues, and once remedied, VerSprite agreed the no logs policy was both accurate and implemented correctly.

However, Norton does collect some anonymized information from the VPN client. This includes connection timestamps, platform details such as OS and timezone, and crash logs. They also aggregate overall data transmission for network planning. So, if you’re extremely concerned about your privacy, Norton might collect slightly too much information for comfort. As it stands, Norton’s acceptable for day to day browsing, but you might consider Proton VPN instead if you need rock-solid privacy guarantees.

Norton is clearly taking the necessary steps to ensuring its VPN is secure and private."

Rob Dunne - VPN Editor, TechRadar

Something that may ease some privacy concerns would be if Norton VPN implemented RAM-only servers. These servers wipe when rebooted, meaning you cannot store any data on them, thus eradicating the risk of any user data being available should a data request be submitted by authorities. Not having RAM-only servers isn’t an issue as such, though it adds a layer of trust for users, and is becoming increasingly common among top VPNs.

Given its privacy focus, we wanted to know more about Norton VPN’s proprietary Mimic protocol. In addition to offering obfuscation, Mimic is powered by TLS 1.3 ciphers (AES-256 and ChaCha20), as well as CRYSTAL-Kyber-512 for post-quantum cryptography. It’s good to hear that Norton is already thinking ahead when it comes to quantum security, which puts them significantly in front of most of the VPN industry.

Meanwhile, Norton's standard VPN protocols, OpenVPN and WireGuard, use AES-256-GCM and ChaCha20-Poly1305 encryption, respectively. These are considered the top encryption methods used by all of the best VPNs, a strong demonstration, therefore, of Norton’s intent to found its service on strong privacy staples.

Customer Support

Norton’s customer support staff are very helpful. There’s a community forum where you can post issues you’re having with Norton VPN, where other members and support staff can pitch in with their own advice. From the interactions we’ve had, they’re very knowledgeable and quite prompt in returning with information.

However, the support materials on the site are a different matter. When you search for help topics on Norton’s website, you’re immediately given an AI prompt for your search, which looks like it’s powered by Gemini. It’s not totally useful when you’re trying to find specific help on a topic, and you have to scroll past it to get to the actual results.

As for the knowledge base, it’s not particularly in-depth - most of the articles consist of bullet point lists and some are thinly-disguised marketing material. If you want help with Norton, we’d stick to contacting their customer support directly through the forum or over the phone. There’s also a 24/7 helpdesk upgrade if you need around-the-clock customer service.

Pricing and plans

Norton offers three tiers of pricing. The standard VPN package starts at $39.99 for the first year, which works out at roughly $3.33 per month. After the introductory offer is over, it renews at $79.99 per year. That makes Norton one of the cheapest VPNs we’ve see.

However, there’s one major limitation: you’re only allowed five device connections (or worse, only one if you’re in some regions including the UK), significantly under the 8-10 you’ll find from most other providers. Surfshark and PIA both go even further, offering unlimited device connections on their cheapest subscription plans. It should also be clarified that Norton’s device limit isn’t a traditional simultaneous connections limit. Instead, Norton’s five device limit refers to the number of devices with the VPN installed. Should you wish to install on a sixth device, you’d instead be asked to remove one of the previous five devices, rather than merely disconnecting from the VPN.

Upgrading your subscription costs an extra $10 per year, making your subscription cost $4.17 per month initially and $109.99 every year after that. There’s no meaningful change to the VPN, but you get extra features from Norton’s security suite, including virus protection, password management, dark-web scanning, 10GB encrypted cloud storage, and AI-powered scam detection.

The Ultimate subscription package will set you back an extra $20 per year, making it $5 per month initially and $129.99 per year after the first. In addition to extra family-safety features for monitoring your child’s devices, the device count is bumped up to 10, so you can take full advantage of Norton’s security features on most of your household’s devices, and you get 50GB of secure storage (or 150GB should you activate auto-renewal).

To its credit, Norton offers an above-average 60-day money-back guarantee and a true 7-day free trial when you sign up. We love 7-day free trials as an entry point into VPNs as they offer a ‘try before you buy’ solution, without some of the hassles caused by a 30-day money-back guarantee. Seeing Norton VPN be one of the first to offer this, then, puts it in good stead to be among the best VPNs for beginners in the future.

Switching from its current install-based device limit also seems like an easy win Norton VPN could take advantage of. While increasing the device limit with more premium plans does help this slightly, removing the confusing of registering and removing devices is an easy way to make the service more accessible to newer users and takes away the sting of the small device limit on the standard plan.

Should you use Norton VPN?

Norton VPN is a rapidly improving VPN. In the space of a few months, it’s brought in a spread of features you’d expect to see in a top VPN, alongside some features some top VPNs don’t think to include. Its performance has risen to a point where it can rival top VPNs like Surfshark and NordVPN, it has apps simple enough for anyone to use, and it brings Norton-pedigree security to make anyone trust its privacy guarantees.

That said, there are several areas it still needs to address. Primarily, sorting the device install limit will add to its already high-value package by reducing connection roadblocks for users. Outside of this, expanding the feature pool, adding Linux and Fire TV support, and eradicating the feature disparities for macOS and iOS users will quickly put Norton VPN among the best value VPNs available if done right.

For many, now might not be the right time to pick up Norton VPN due to any one of the limitations mentioned. That said, it’d be wise to keep an eye on Norton VPN over the coming year or so, as, from what we’ve seen already in 2025, it looks as though the provider could quickly become a high-value, high-security VPN from a name renowned for its security capabilities.

Norton VPN alternatives

1. NordVPN – The best VPN overall
NordVPN combines security, speed, and usability in a single VPN package that can’t be beat on value. In addition to lightning-fast servers, NordVPN offers some unique features like Threat Protection Pro, which integrates anti-virus, anti-phishing, and ad-blocking capabilities into a single product. Try it today with a 30-day money-back guarantee.View Deal

2. Surfshark – get the 7-day free trial
When you buy a Surfshark subscription, you get unlimited simultaneous device connections, 100+ countries to connect to, and advanced IP rotation features powered by their own Nexus software-defined network. It's perfect if you need to cover a household full of devices. Try Surfshark risk free with a 30-day money-back guarantee or a 7-day free trial.View Deal

3. ExpressVPN – The best secure VPN
ExpressVPN consistently ranks as our most secure VPN thanks to its forward-thinking security features, like quantum-resistant encryption. It's all thanks to their proprietary Lightway VPN protocol, designed from the ground up for mobile-first support. The ExpressVPN app is perfect for beginners with a simple interface across all platforms, plus 24/7 support on hand to walk you through setting it up. Express also comes with a 30-day money-back guarantee.View Deal

Here are hints and answers for the NYT Strands puzzle for Aug. 20, No. 535.

Here are hints and the answers for the NYT Connections: Sports Edition puzzle for Aug. 20, No. 331

My M1 Mac mini has been one of the best purchases I’ve made in recent years. It’s performed incredibly well since I bought it and has served me well over the years despite having nothing more assuming than an entry-level chip. Throw in the fact that it only cost me $699, and it’s left me one happy bunny.

Now, AppleInsider says it’s been told by sources inside Apple that the company is testing an M5 Mac mini. This follows up on the outlet’s report earlier this year that an M5 Pro Mac mini was in the works and suggests we could be getting close to seeing the next generation of Apple’s tiny desktop computer.

As well as my M1 Mac mini has performed, I’m sorely tempted to upgrade to an M5 model. By all indications, it’ll come at the right time and with the right features to make the outlay worth it.

An upgrade that’s built to last

(Image credit: Apple)

The Mac mini is already the best mini PC you can buy, and it’s not even close. Its powerful, efficient chips deliver significantly more performance than you would expect from a computer this small. And it really is small, leaving a tiny footprint on your desk, even by the standards of these small-scale devices. It’s made from high-quality materials with a solid aluminum chassis, yet its $599 price tag - $100 less than I paid for the M1 model – means it offers tremendous value.

That’s all with the M4 chip, but with an M5 on the way, we can expect things to get even better. The M5 chip is rumored to bring a performance increase of 15-25% over the M4, as well as Thunderbolt 5 in the entry-level models and Wi-Fi 7 connectivity. It’ll be a modest but meaningful upgrade.

There won’t be a new design for the Mac mini, but that really doesn’t bother me. Considering the current look is less than a year old – and still looks fantastic – there’s no need to change it. If it ain’t broke, don’t fix it.

All this talk of a new Mac mini comes at the ideal time for me to upgrade. My M1 Mac mini is five years old and can’t always keep up with more heavy-duty tasks. My MacBook Pro with M1 Pro chip, meanwhile, is doing OK, but I’m increasingly having to turn down the settings in more demanding games like Cyberpunk 2077. In both instances, my Macs are starting to show their age.

A Mac mini with an M5 or M5 Pro chip would be perfect for my more performance-intensive work (such as reviewing high-end games), allowing my MacBook Pro to focus on less-demanding work on the go. Given how much I’ve managed to eke out of my M1 Mac mini, an upgraded model could last me well into the future.

Pro possibilities

(Image credit: Apple)

I recently took the macOS version of Cyberpunk 2077 for a spin on two different Macs: my MacBook Pro with M1 Pro chip, and an iMac with M4 chip. And thanks to its extra graphics cores, the M1 Pro actually outperformed the M4, despite being several generations out of date.

That makes me excited to think about what might be possible with an M5 Pro Mac mini and the extra graphical oomph granted by its increased core count. Throw in frame generation – Apple has its own MetalFX implementation of this in the works – and the next Mac mini could achieve frame rates I never thought possible from a Mac. And that’s all for $599, which makes the Mac mini a pretty spectacular bargain.

For now, we don’t yet know when the M5 Mac mini is going to launch. AppleInsider previously suggested that it would arrive in 2025, but early 2026 is also a possibility.

Either way, it’s going to come at a potentially expensive time for me. I’m eyeing up the MacBook Pro with M6 Pro chip, as its expected combination of high-grade performance and OLED display could make it a gaming dream machine. At the same time, I also want to upgrade my desktop PC from 2021, which isn’t quite able to hit the levels I want with its RTX 3070 graphics card.

But given what could be coming to the Mac mini, it’s going to be hard to resist prioritizing Apple’s dinky desktop Mac. All in all, it feels like a great time to be an unashamed computer nerd.

You might also like

• Apple Mac mini (M4, 2024) review: smaller, faster, better
• These could be the next 17 Mac laptops and desktops that Apple launches over the next year, according to a new leak
• I love my M1 MacBook Pro and have never considered upgrading it, but these new M6 rumors might just change my mind

A Reuters/Ipsos poll finds almost half of Americans think AI is bad news for humanity.