You might think that someone like me, who did everything from 100-mile ultramarathons in Mongolia and bikepacking through Wales to trail running in the Dolomites and eFoiling in the South of England, must have mastered the art of skiing. After all, I turned 40 last year, which is plenty of time to tick something like skiing off the bucket list.
Well, that’s not the case. I’ve never skied before, partially because we never had the money to go on ski trips when I was young, and it hasn’t come up later in life. So imagine my excitement when Helly Hansen approached me with an offer I couldn’t refuse: did I want to learn to ski in Verbier?
Part of the company’s campaign called ‘A World That Needs More Alive’, I was not only taken to Switzerland to get my grips with the beautiful sport that is skiing, but the brand also recruited legendary freeride skier and Helly Hansen ambassador Warren Smith to show me the ropes, which is not dissimilar to getting Michelangelo painting your bathroom.
As well as being pretty good at skiing, Warren also has a Ski Academy in Verbier, which he set up a quarter of a century ago. “We were one of the first foreign ski schools in Verbier, mixing in with a few Swiss schools,” he says. “We’ve built a strong community of regular skiers, and we were also early adopters of video analysis in ski training – filming and analysing technique long before it became common practice.”
The Warren Smith Ski Academy has around 3,000 regular clients and has qualified more than 1,000 ski instructors up until now. “That’s been an incredible achievement for us,” adds Warren. Aside from coaching, he had a professional skiing career as part of the Völkl Freeride Team, participating in product shoots, ski testing, and filming.
Is it possible to I was given three days to go from zero to hero, so I had no time to waste once I arrived in Verbier. But first, I had to get there.
Day 1
(Image credit: Matt Kollat/ T3)
I arrive nice and early at Geneva airport. I need to catch a train, but first, I have to find where the trains are. Switzerland strikes me as a punctual country, so I’m fairly confident I’ll be able to catch my two connections despite connection times as short as four minutes.
Get all the latest news, reviews, deals and buying guides on gorgeous tech, home and active products from the T3 experts
The trains aren’t punctual but somehow always on time. The spacious double-decker setup means I have a full four-seater for yourself, my bags, and other bits. The folding tables are neat and clean, and there are sockets and bins – just for me.
The sunshine glistens on Lake Geneva’s crescent body as we travel around it from Geneva airport to Martigny. Looking at Google Maps, the journey seems like an upside-down horseshoe (or a down-facing half-moon). The places we zoom past look like illustrations from a fairy tale. The names are equally enchanting to my ears: Lausanne, Vevey, Montreux, St Mauritz, and Chillon Castle.
The lake disappears from the view as the train’s nose starts pointing south. The sun bathes me in its golden light; haven’t realised I was on the sunny side when I got on at the airport. With Lake Geneva just behind us, the Diablerets appear seemingly out of nowhere. Snow-topped mountains pierce the sky, beautifully terrifying against the backdrop of nearly kept, lush green fields.
Looking at the mountains makes me feel at ease. I have a love affair with the Alps and jump at every opportunity to visit the area. I find it soothing to glance up at the massifs, knowing how small and insignificant we all are. At the same time, I can feel my belly flutter in joy, knowing I’m on my way up. I never had a bad time in the mountains.
The tannoy announces the Martigny stop, telling us that this is the stop to change for Chamonix and, of course, Verbier. I got off and notice people dressed in mountain gear with climbing ropes thrown around their shoulders. Long, hefty bags hiding skis and snowboards are wheeled around. Hard to imagine being able to enjoy winter sports with the sun beating on us right now. Surely, there can’t be any snow in this weather!?
(Image credit: Matt Kollat/ T3)
In Martigny-Bourg, a group of what looks like 7-8-year olds get on the train in full ski gear. They click-clack on the hard floor of the train in their ski boots, clutching their skies and poles and wearing their helmets. French chitter-chatter fills the carriage. I feel an odd sense of gratefulness for them; I wish I had the opportunity to try skiing when I was this young. Alas, I’m just about to embark on my first skiing experience at the ripe age of 40.
After Martigny-Bourg, the weather turns. We’re at the foot of the mountain, on the shady side. The slender, tall trees are covered in frost. Thanks to the heating and the bodies surrounding me, I still feel hot on the train in my Helly Hansen fleece.
The snow takes over at Sembrancher. We must be at altitude; not quite the Himalayas, but high enough for snow not to melt. The sun keeps shining relentlessly, filling the carriage with warmth and light. Someone gets off in shorts and sneakers. The weather app says it’s 6 degrees Celsius outside. Hard to imagine with all the snow everywhere.
In La Chable, I say goodbye to my tiny friends. I leap off the train and jog around the train station to catch my connection. The bus driver doesn’t speak English, so we use the international hand sign code to point at things and give a thumbs up to communicate. I hope I’m on the right bus.
I look out the window and see the cable car running up the side of the mountain. In hindsight – if only I knew – I should have taken that to the top. Who doesn’t like riding the funicular? The outskirts of Verbier are peppered with wooden mountain chalets, reminiscent of the rifugios I’ve seen in the Dolomites. My heart skips a beat as I can feel the anticipation building up. In an hour, I’ll be skiing for the first time in my life. No time to waste!
First ski lesson
(Image credit: Matt Kollat/ T3)
The first thing I did after arriving in Verbier – right after taking photos of the mountains, of course – was to call my ski instructor, Warren. Being the warm host he is, Warren immediately offers to drive me to my hotel to get changed. A professional free skier, Warren’s knowledge about skiing is beyond intimate, and as a result, I feel very comfortable around him.
Once I don my ski gear, I make my way to the ski lift we’ll be using for the next couple of days, which is about 3-4 minutes on foot (Verbier really isn’t that big). I collect my pass, and then we walk to Mountain Air, where I get fully kitted out. My feet get measured, and I get to hold ski poles to see if they are the right size for my height.
Warren says it’s absolutely critical to get the correct gear especially for beginner skiers such as myself. A good ski instructor can only do so much if the skier’s equipment isn’t right.
“For example, if someone gets their boots fitted by a part-time shop assistant who lacks expertise, they might end up with boots that are too stiff or too wide, which can completely block their ability to improve,” Warren explains. “If the boot doesn’t flex properly or doesn’t match the skier’s biomechanics, even the best coaching won’t help.”
A good shop and instructor should check flex ratings, foot width, and overall fit because even a small mismatch can make learning much harder. Ski stiffness is another factor – if the ski is too stiff for a beginner, they won’t be able to control or bend it properly, making progress frustratingly difficult.
(Image credit: Matt Kollat/ T3)
I’m given skies ideal for beginners; somewhat flexible and not too long to provide more grip on the snow. With my kit on me/ in my hand, Warren and I head over to the number 2 bus that will take us to the beginner area. Buses are free in Verbier and come very often – a nice surprise.
The sun is still shining, stronger than ever, and I have three layers of Helly Hansen gear on me, including my new favourite waterproof shell, the Sogn Tinden Jacket, which has an insulated chest pocket for phones. To be fair, the clothes do an excellent job keeping me at an even temperature. Even so, we decide to lose the outer layers to help ventilation.
But before that, Warren walks me through the essentials – literally. The first test is to walk around with one ski on to see how it feels to be strapped in. The sensation is strange, to say the least; hobbling along like a bird, dragging along my foot with skis on.
The second step is to walk around with the skies. I follow Warren in a figure-of-eight shape, taking small steps. Trying to take a big step might result in stepping on the skies. Walking done, we head to what I call the beginner-beginner slope. I ski escalator takes you up, and you slide back down.
As Warren explains, the key to having full control over your skiing is to press your shins against the front of the ski boot as hard as you can. Don’t sit back; instead, lean forward and put as much body weight on the front as possible. This digs in the skies and lets you manoeuvre better.
“Skiers need to lean forward and put pressure on the front of their boots, especially on the outside ski when turning,” Warren explains. “This gives the ski grip and control. It feels counterintuitive initially, but once you get used to it, everything clicks into place.”
Easier said than done. My visceral self is fighting me every step of the way. The natural reaction is to lean back – surely, leaning forward will cause you to bomb it down the hill? I’m doing my best to form a snow-plough shape with my skies, but my hips and legs are getting tired. Skiing is hard!
As I’m battling with myself, I can hear Warren’s encouraging words echoing in front of me: “Good, Matt! You’re totally in control of your speed!” The speed might be glacial, but I’m very much in control of it.
After four rounds on the beginner-beginner slope, Warren suggests we catch the mini ski lift and try our luck on the other side: the beginner slope. The top of that course is steeper, and Warren guides me down so I won’t accelerate too much. My hips start aching – clearly, I’m using every muscle available, not just the ones I need. Some warm-up wouldn’t have hurt, either.
I manage to get to the bottom of the beginning slope in one piece: success! At this point, I’m parched, exhausted and ready for a break. However, I can’t resist Warren’s suggestion to go down the beginner-beginner slope once again, more or less unaided. I slide down, focusing on putting the pressure on the front and keeping the tip of the skies around 6 inches apart.
Warren cheers me on (skiing backwards and shooting a video of me) as 5-year-olds bomb it down around me like it’s nothing. I’m trying my best to remind myself that it’s natural not to be natural after an hour of skiing.
By the end of the first session, I feel confident and can’t wait for my brain to put everything I learned today into place during my sleep so I can start tomorrow’s session like a pro. For now, though, it’s time to grab some water and relax. I wish I brought my foam roller.
Day 2
(Image credit: Matt Kollat/ T3)
Had a comparatively good sleep, which is unusual when I’m away from home. This is probably thanks to the colder room and maybe even the fact that it was a single room with a single bed, so I didn’t feel like I was in a room designed for more than one occupant.
Breakfast was very continental, and the view from the conservatory was absolutely stunning, overlooking the mountains surrounding Verbier. Immediately filled my heart with excitement for the day ahead.
After some stretching in the room – I thought I might as well give my hips a fighting chance – I message Warren, who, naturally, is already at Mountain Air waiting for me. I put on my ski pants and jacket, grab my helmet and head out to meet him.
Once reunited with my skies and sticks, we take the number two bus to the beginner slope. It’s Thursday, and, as it turns out, it also happens to be the day when local schools take the little ones out for their ski lessons. Naturally, the slope is packed, with kids of all ages everywhere.
There are parents, too, with what look like 3-4-year olds. I love the clever technique they used to entice the kids to ski. Someone hangs up a toy on a string attached to a stick (much like a cats’ toy) and dangles it in front of the toddler to get them to move. Others use little toys the kids can throw, and once it lands, their mission is to ski to it.
I don’t need such encouragement; I’m ready to see if my brain and my body have managed to internalise yesterday’s lesson. We start on the beginner-beginner slope to see how my control is. After a couple of descends, Warren is confident I could tackle the other side (beginner slope), so we catch the ski lift, and off we go.
I feel a lot more in control than yesterday, but it’s still hard to lean onto the front of the skies like Warren wanted. We practise putting weight on the correct leg by lifting up the other one as we traverse the slope. As always, it is easier said than done, but by the second descent, I’m doing much better.
On my third attempt, I finally fall – for the first time since I started my lessons. I knew it was only a matter of time before it happened, and I’m glad it happened on the beginner slope. Even with my fiasco, Warren feels confident I could take on Le Chaux, a ‘real’ slope, albeit a blue one.
What’s a blue piste? From what Warren tells me, four colours demarcate the different gradients. Green is the least steep (the beginner-beginner slope), followed by blue, red and, finally, black. Le Chaux is a combination of mostly blue gradients with some red bits sprinkled in for good measure.
(Image credit: Matt Kollat/ T3)
To get there, we head back to the centre to catch the gondola lift on Rue de Médran. Verbier sits around 1,530 metres above sea level; Le Chaux is all the way up 2,260 metres. You need to take two gondola lifts to reach the top, which is about half an hour. Once there, you take a third ski lift – the classic leg-dangling-in-the-air variety – to get to the start of the piste.
Looking around, I’m amazed to see the beautiful Alps that surround us – snow-capped mountains until the eyes can see. I ask Warren if he ever gets bored of the view. “Never,” he says without hesitation. I notice some spots where the snow seemingly slid down the side of the mountain and ask Warren what happened there.
“Over the last 15 years, we’ve seen a significant increase in avalanche activity due to rising temperatures,” he says. “When you have a layer of snow affected by heat, then fresh snow falls on top, the layers don’t bond well, creating unstable conditions. We’ve had weeks where temperatures swing from -15°C to +10°C overnight, which is disastrous for snowpack stability.”
His concern for mountains (and the environment, in general) made him get involved with Protect Our Winters (POW), a nonprofit dedicated to fighting climate change, specifically focusing on its impact on winter sports and mountain communities. “It’s a fantastic organisation because it’s led by athletes and outdoor enthusiasts who inspire others to take action,” adds Warren. “The goal is to promote sustainable practices, such as using trains instead of planes for travel and reducing our environmental impact.”
With my poles clutched, I start my descent. Warren is ahead of me, focusing on going as slow as I am (which must be hard, to be fair). I’m conscious of everything and afraid of going too fast and literally falling off the side of the slope. We get to the first bend, and I promptly fall on my bum as I try to turn. We’re off a good start.
We descend further still until we get to a point where the slope gets steep. This is also the area where another slope joins ours, one that’s much steeper and, therefore, has skiers bomb down on it. Warren tells me we need to go to the other side, slowly edging our way toward a blue pole that marks the other side of the slope.
I’m frantically scanning the terrain to see where it looks the least steep. I intended not to fall; I don’t want to disappoint Warren or myself. I’m very conscious of skiers going 30mph perpendicularly to where I want to go, but Warren is off, so I have to go, too. My quads are working overtime, and my mind is racing, but somehow, I make it to the other side. Happy days.
The slope continues to offer a mix of blue and steeper red gradients. Whenever we hit the red bits, we regroup, and I follow Warren across the slope. Skiers literally fly past us, but I’m busy putting pressure on the correct leg and not sliding out of control on the turns. I fall and get up (often with Warren’s help). Somehow, we get to the bottom of the piste.
(Image credit: Matt Kollat/ T3)
We go again. My confidence is building but every turn feels like tap dancing on high wires – not natural. I fall and get up again, but at the same time, I feel more in control of my descent. Some of the corners I take are vaguely reminiscent of parallel turns, which is the ultimate goal of my sessions in Verbier.
It’s now 12:45 pm, and we’ve been either travelling to places to ski or actually skiing since 9 am. Warren suggests we grab a bite before going again, and I gladly agree. We hoover up some sausages and fries and wash it down with a pint of peach ice tea. It’s time to go again.
We take the ski lift to the top and descend again. Maybe it’s the break we just had or the familiarity with the slope, but I feel much better. I throw myself into some corners, and I place my weight on the correct leg most of the time. I fall occasionally, but nothing too serious.
Warren feels good about things and says I should do the next descent without my poles. I don’t dare to say no, even though I can’t imagine not having them on me. Nevertheless, we catch the lift once again and off we go.
The descent is a mix of controlled turns and falls. Fatigue starts setting in, and I miss some of the turns I easily made last time. I feel more confident, but my body is not playing ball. I end up sliding down the last bit on my bum after a fall. Thankfully, it’s not the slope but the area heading to the lift, which is comparatively flat.
We hop on the lift and return to Verbier. Even though I loved every minute of it, I feel glad we’re done skiing for the day. I need a break to internalise today’s lessons. Time to get some food and maybe even a beer.
Day 3
(Image credit: Matt Kollat/ T3)
A chilly morning in Verbier; the thermometer is showing -4 degrees Celsius. I didn’t sleep well, or more like, I slept how I usually do when I’m away from home: tossing and turning in bed. My Oura Ring 4 said it took me 1h15m to fall asleep. Sounds about right.
One of the reasons why I couldn’t sleep was because I kept thinking about skiing. My mind couldn’t wait to work on what I learned until I fell asleep, so instead, it kept showing me images of me speeding out of control on the skies. My legs automatically tensed to slow me down, preventing me from falling asleep. I did some breathing exercises to calm down; then the process started again.
Despite my pre-nightmares about being unable to control the skies, I woke up feeling that I knew how to parallel ski. I hadn’t done it until that point, but it felt like I might be able to do it – not sure why. After breakfast, I check out and head over to Mountain Air to meet Warren for the last time.
Ski gear collected, we head to the cable car. It’spacked; the station only opened 30 minutes prior. We make our way up, just to be confronted with even more skiers. Reams of people – many kids – are hitting the slopes, which makes navigation tricky. I clutch my poles (and do some affirmations) to ready my body and mind to crack parallel skiing. I feel ready for another descent.
Warren says we’ll add another move today: parallel sliding. You take up a position perpendicular to the slope and drop from the edges of the skies to the flat bottom, which makes you slide down sideways. Raising the skies to the edges stops the descent.
To my dismay, Warren immediately recommends losing the poles. It’s all in the legs, he says, and the poles will only distract you. I hesitantly say yes, even though I really want to feel the safety net the poles provide. As soon as we get off the ski lift, my skies ran out from under me, and I fall on my bum. Excellent start!
I find my way back onto my feet and catch up with Warren. He is waiting for me in the left corner, where we always meet at the top of the slope. After another easy section, I find myself face to face with the first real drop of the piste. Warren goes ahead, I steel my nerves and follow him.
After gathering some speed, I almost naturally start moving the skies into a parallel position. I’m going faster than yesterday, so turning feels better. Turning slowly felt awkward and unnatural; faster movement helps me lean into the movement better.
Warren is happy with the progression and suggests we go down Le Chaux the other way with a stepper drop at the top. I’m in his hands, so I say yes, and up we go. We practice the parallel slide, and every time it goes a little bit better. Not perfect, even by the end, but there certainly is a sense of progression.
We make five descents before 12, each time a bit quicker and in more control. I fall many times but manage to get back on the skies on my own (in most cases). Warren lets me do sections on my own instead of hovering around me, which instils more confidence in me.
I’m throwing my skies left and right, always repeating Warren’s words in my head: “Pressure on the front, turning to the left (or right)!” Reminding myself at every turn doesn’t magically make me a free-ride champion, but it helps me focus on the right area. I’m feeling in control and great. One might say I feel alive!
At around midday, we decide to head back down to Verbier so I wouldn’t have to rush around before leaving for the airport. My body is exhausted, and my mind is filled with unforgettable memories. I went from being unable to form the snow plough shape with my skies to skiing downhill on blue/red gradients alone. I feel fantastic.
Skiing: best sport ever?
Ski instructor extraordinnaire Warren and I
(Image credit: Warren Smith)
Do I wish I tried skiing before? Absolutely. It’s a brilliant sport that makes you feel differently from any other sport I tried before. Even as a 40-year-old, I learned the basics of skiing in a few days. I fell quite a few times in the process, but I haven’t sustained any injuries.
It’s worth noting that I’m generally an active person, so my body isn’t in the worst shape for exercise. I could be more flexible and muscular but don’t feel out of depth when trying new sports. People with a more sedentary lifestyle might have a different experience learning to ski.
That said, it’s never too late to start, as my experience in Verbier exemplifies (hopefully). I’m already planning on hitting the dry slopes near where I live so I can hit the piste even more confidently next year. After all, Warren learned to ski on a dry slope, and he’s a world-renowned free skier.
If you’re concerned about the money, I’ve already been suggested a few websites that can help keep costs down. I found Heidi quite intuitive, and Iglu Ski is supposed to be a good place to start planning last-minute ski trips. As for gear, I can certainly recommend Helly Hansen’s ski range. It’s top of the range, and you’ll see it everywhere on the slopes, which is reassuring.
Six biomechanical tests
The best way to learn to ski is to hit the slopes – no surprises there – but there are several ways you can ready your body for the adventure, from exercises to hitting the dry slopes. One area you should certainly pay attention to is biomechanics.
“Biomechanics plays a huge role in skiing performance, and we use a structured process to assess it,” Warren explains. At the Ski Technique Lab, they run six key biomechanical tests to understand a skier’s movement limitations. These include:
Ankle Flex Range Test: Measures forward flex range; the national average is about 9 cm, but we aim for 15 cm for safety and 20 cm for performance.
Ski Symmetry Test: Looks at lateral control, ensuring skiers don’t wobble when trying to keep their legs parallel.
Leg Steering Range Test: Measures how well a skier can rotate their legs inward for turns. Many people have one side that rotates well and another that’s restricted.
Leg Collection & Extension Test: Evaluates how well a skier flexes and extends their legs while maintaining balance.
Core Stability Test: Ensures the core remains active while making functional movements.
Body Angulation Test: Focuses on upper body positioning in turns, but this is usually something we address at a later stage.
“Most people don’t realise they have biomechanical imbalances until they see them in action,” he adds. Video analysis is key here – it allows skiers to see their own movement patterns and understand what they need to improve. To learn more about biomechanics, visit the Warren Smith Ski Academy website.
The rapid evolution of tech, and especially AI, has brought us to the era of edge computing, an approach in which data is processed closer to its source rather than relying entirely on distant cloud servers.
At the forefront of this trend is ‘edge AI’, where AI models and algorithms are deployed on local devices to deliver immediate insights and actions. The approach is being fueled by advancements in efficient large language models (LLMs) such as DeepSeek, as well as small language models, so-called tiny AI, and the integration of neural processing units (NPUs) in modern hardware.
For businesses, edge AI promises not just technical sophistication but tangible benefits like reduced costs, faster operations, and stronger data security.
Edge AI has the potential to redefine how businesses harness AI by moving processing power to the edge – on devices such as Internet of Things (IoT) sensors, smartphones, and industrial machines – eliminating the latency that comes with sending data back and forth to cloud servers. For example, in a smart manufacturing environment, edge AI can analyze sensor data on-site, identifying potential equipment failures before they happen.
Another critical advantage of edge AI is enhanced security. Traditional cloud-based AI requires data to travel to remote servers, introducing potential vulnerabilities along the way. In contrast, edge AI processes data locally, reducing exposure to breaches or unauthorized access, something to please IT and cybersecurity departments.
As Leo Gebbie, principal analyst & director, Americas at CCS Insight points out, “On-device AI is far more secure, given data doesn’t need to move to and from data centers, and this is seen as crucial in encouraging users to allow AI to handle more personal and private data moving forward.”
Beyond these technical and operational benefits, edge AI also can offer significant cost savings for businesses, as running AI models on local hardware can drastically reduce expenses associated with cloud storage and energy usage. Edge AI isn’t just a technological advancement; it’s a strategic shift that empowers businesses to operate smarter, faster, and more securely in an increasingly connected world.
Receive our latest news, industry updates, featured resources and more. Sign up today to receive our FREE report on AI cyber crime & security – newly updated for 2024.
For all these reasons, edge AI is becoming more of a topic for IT decision-makers. Grand View Research estimated the total market value of edge AI as $20.87 billion in 2024, projecting a compound annual growth rate of 21.7% between 2025-30 for the market.
Enabling edge AI with small language models
Small language models and tiny AI are transforming the way AI operates at the edge.
Unlike their larger, cloud-based counterparts, these lightweight models are designed to run efficiently on devices with significantly less computing power, making them ideal for real-time applications. These models can often run on business-grade devices and even the upper end of consumer hardware such as mid-range graphics processing units (GPUs).
A good example is small language models that power on-device virtual assistants, enabling voice recognition and translation services without sending data to the cloud. By keeping all processing local, the models enable faster response times, greater reliability in areas with poor connectivity, and enhanced privacy as sensitive data remains on the device.
Deploying these efficient models, companies can reduce operational costs associated with cloud infrastructure while improving service quality. In retail, for example, edge AI can analyse in-store customer behavior to personalise recommendations instantly, while in logistics, it can optimize delivery routes in real time.
AI PCs and NPUs
The rise of AI PCs and the inclusion of NPUs in modern hardware have been pivotal developments in bringing AI down from the cloud and to the edge.
NPUs are specialised chips designed to handle complex AI workloads, enabling devices to process tasks like image recognition, natural language processing, and predictive analytics in tandem with other processes handled by the device’s CPU and GPU. This hardware advancement is particularly significant in industries requiring real-time responsiveness, such as manufacturing or health care, where latency can mean the difference between success and failure.
But it’s important to note that NPUs aren’t entirely necessary for edge AI. Gebbie tells ITPro that these chips are still optional if the goal is simply to inference data locally.
“The presence of an element such as an NPU isn’t a prerequisite to a device qualifying as an AI PC or AI smartphone,” he says. “Lots of devices are capable of running AI on-device – perhaps by using their CPU and GPU – and we define AI devices by their ability to run software, rather than by hardware.”
While NPUs provide a performance boost, many businesses are leveraging existing CPUs and GPUs to run AI models effectively, making edge AI more accessible without requiring a complete overhaul of infrastructure. For example, laptops and desktops with AI capabilities can handle tasks like fraud detection in banking or automated customer support in real time, streamlining operations.
Business applications and challenges for edge AI
One key advantage of edge AI is the ability to lower operational expenses by running AI models on local hardware rather than relying on cloud systems which charge per input and output.
Latency reduction is another critical benefit of edge AI, especially in environments where immediate responsiveness is essential. On a smart factory floor, edge AI can enable machines to detect and respond to anomalies within milliseconds, ensuring uninterrupted operations.
But despite its advantages, edge AI does present several challenges for enterprises.
Implementing edge AI solutions often requires significant upfront investment in hardware and software, which can be a barrier for smaller businesses. While mid-range devices may be technically capable of running an AI model, leaders will have to weigh up whether these cheaper options deliver the speed and performance desired – and fork out for more expensive options if they’re seeking to run the most sophisticated edge models.
Edge AI systems may also lack the scalability and raw computational power of cloud-based models, limiting their ability to process complex datasets or accommodate rapid growth. It’s for this reason that IT leaders may choose a blended approach, designating tasks that need low latency to edge AI models while relegating those that can take more time or require far more data to be examined to their cloud workloads.
A necessary trade-off with edge AI is accepting that models small enough to run locally are, at present, less powerful than the leading LLMs. For example, DeepSeek and the smaller of Meta’s Llama models can be run locally but neither score as high as OpenAI’s reasoning models or the latest of Google’s Gemini releases.
The future of edge AI seems poised to bring even greater leaps forward, with the development of specialised AI models tailored for edge computing environments. These models, designed for efficiency and precision, will enable more industries to adopt edge AI, particularly as 5G and IoT continue to expand.
Emerging applications like autonomous vehicles, augmented reality, and real-time supply chain optimization are already benefiting from edge AI’s ability to deliver low-latency performance. Innovations in AI hardware, such as more advanced NPUs and hybrid processing capabilities, will further enhance the speed and scalability of edge AI solutions.
As edge AI evolves, its impact on industries like manufacturing, health care, and retail is expected to grow exponentially. For businesses, the case for adopting edge AI is becoming increasingly compelling. While initial investments in hardware and system integration may be high, the long-term benefits – reduced reliance on the cloud, operational agility, and the ability to innovate rapidly – far outweigh the challenges.
As edge AI continues to mature, businesses that embrace this technology will be well-positioned to lead in their respective industries, leveraging the power of localised intelligence to drive growth and sustainability.
ASRock, a prominent name in PC hardware, has recently extended its product portfolio into the power supply unit (PSU) market. As a company whose philosophy revolves around reliability, ASRock aims to provide robust and aesthetically appealing PSUs for high-performance gaming PCs and workstations. The company released four PSU series almost simultaneously – the Taichi, Phantom Gaming, Steel Legend, and Challenger – with the Steel Legend series being the most affordable modular PSUs that the company currently offers.
We examine the Steel Legend SL-1000G, a 1000W power supply designed to meet modern ATX 3.1 and PCIe 5.1 specifications. This PSU balances efficiency and performance and could be a notable contender to some of the best power supplies in the market. With 80Plus Gold and Cybenetics Platinum certifications, the Steel Legend SL-1000G delivers modern features, aesthetics, and value.
Specifications and Design
Swipe to scroll horizontally
Power Specifications (Rated @ 50 °C)
RAIL
+3.3V
+5V
+12V
+5Vsb
-12V
MAX OUTPUT
20A
20A
83.4A
3A
0.3A
120W
120W
1000W
15W
3.6W
TOTAL
1000W
1000W
1000W
1000W
1000W
AC INPUT
100 – 240 VAC, 50 – 60 Hz
100 – 240 VAC, 50 – 60 Hz
100 – 240 VAC, 50 – 60 Hz
100 – 240 VAC, 50 – 60 Hz
100 – 240 VAC, 50 – 60 Hz
PRICE
$160
Row 5 – Cell 2
Row 5 – Cell 3
Row 5 – Cell 4
Row 5 – Cell 5
In the Box
The ASRock Steel Legend SL-1000G arrives in durable and well-designed packaging, featuring a brushed metal artistic theme consistent with the Steel Legend branding. The PSU is securely placed inside a nylon pouch and reinforced with foam inserts to ensure safe transit.
(Image credit: Tom’s Hardware)
ASRock provides a generous bundle of accessories, including mounting screws, an AC power cable, a jump-starting connector for PSU testing, a few cable ties, and several cable straps, catering to users who appreciate a complete installation package.
(Image credit: Tom’s Hardware)
The Steel Legend SL-1000G includes all-black, individually sleeved cables with flexible, soft wiring for easy management. Notably, the 12+4 pin PCIe 5.0 connectors are partially green, ensuring proper installation by signaling improper connections. The PSU is equipped with three 6+2 pin PCI Express connectors, offering a conservative but functional setup for a 1000W unit.
(Image credit: Tom’s Hardware)
Swipe to scroll horizontally
ASRock Steel Legend SL-1000G
Connector type
Hardwired
Modular
ATX 24 Pin
–
1
EPS 4+4 Pin
–
2
EPS 8 Pin
–
–
PCI-E 5.0
–
1
PCI-E 8 Pin
–
3
SATA
–
9
Molex
–
3
Floppy
–
–
External Appearance
Measuring 150 mm in length, the ASRock Steel Legend SL-1000G is slightly larger than the standard ATX PSU but the unit should remain compatible with all ATX-compliant tower cases. Only very small HTPC cases or special design may not have enough room for a 150 mm PSU. The SL-1000G boasts an aesthetically striking design. Encased in a matte black chassis, the PSU features decorative embossments and stickers on its sides that add texture and visual appeal. The ASRock and Steel Legend logos are prominently displayed.
(Image credit: Tom’s Hardware)
Rounded corners and a distinctive mesh-style fan grille integrated directly into the chassis further enhance the unit’s visual appeal. The integrated fan guard covers the entire bottom panel.
Image 1 of 2
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
The rear panel includes the standard AC input socket, an on/off power switch, and a button for toggling the intelligent fan mode. This mode enables semi-passive operation, where the fan remains idle under lighter loads. Disabling intelligent mode forces the fan to start regardless of load but does not override thermal control, keeping the fan at minimal speeds under low-load conditions. One core difference of this implementation is that the switch activates some sort of offset, forcing the fan to operate continuously under lower loads, but also makes the fan operate at a higher speed than it would with the intelligent mode disabled, sacrificing acoustics in favor of thermal performance across the entire load range. The modular cable connectors are located on the front panel, with clearly marked legends to simplify the installation process.
Image 1 of 2
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Internal Design
The ASRock Steel Legend SL-1000G features a 135mm RL4Z S1352512EH-3M fan, likely manufactured by Globe but rebranded with ASRock’s logo. This fan employs a fluid dynamic bearing (FDB), balancing reliability, longevity, and low-noise operation.
(Image credit: Tom’s Hardware)
The ASRock Steel Legend SL-1000G is manufactured by HEC/Compucase, a well-established OEM with decades of experience in producing power supply units. HEC/Compucase is one of the oldest companies in the PSU industry, known for its wide-ranging capabilities that cover everything from entry-level to high-performance designs, although the company typically sticks to entry-to-medium platform designs. They introduced some platforms meant for performance gaming/workstation PCs over the past few years.
Image 1 of 2
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Its input filtering stage consists of four Y capacitors, three X capacitors, and two filtering inductors, offering a configuration slightly above the typical for a 1000W unit. Following the filtering stage are two rectifying bridges mounted on a small heatsink. The Active Power Factor Correction (APFC) stage includes three Infineon 60R120P7 MOSFETs and a diode mounted on a large dedicated heatsink across the edge of the main PCB. A classic inductor and a single Nippon Chemi-Con 680 μF capacitor are the passive components.
(Image credit: Tom’s Hardware)
The primary inversion stage utilizes two Infineon 70R099P7 MOSFETs configured in a half-bridge LLC topology. These transistors are mounted on a separate heatsink next to the primary transformer. On the secondary side, the PSU employs six MOSFETs to generate the primary 12V rail. These MOSFETs are cooled by a large, distinct heatsink next to the main transformer. The 3.3V and 5V lines are derived using DC-to-DC converters housed on a vertical daughterboard connected to the main PCB.
(Image credit: Tom’s Hardware)
The capacitors used in the secondary stage are sourced from Nippon Chemi-Con and Rubycon, two highly regarded Japanese manufacturers known for their reliability and performance. These capacitors contribute to the unit’s stable electrical output and durability. Overall, the SL-1000G is entirely built using only top-quality components from reputable manufacturers.
(Image credit: Tom’s Hardware)
Cold Test Results
Cold Test Results (25°C Ambient)
For testing the power supplies, we use high-precision electronic loads with a maximum power draw of 2700 Watts, a Rigol DS5042M 40 MHz oscilloscope, an Extech 380803 power analyzer, two high-precision UNI-T UT-325 digital thermometers, an Extech HD600 SPL meter, a self-designed hotbox, and various other bits and parts.
Image 1 of 5
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
During cold testing, the ASRock Steel Legend SL-1000G demonstrated an average nominal load efficiency of 90.8% at 115 VAC and 92.8% at 230 VAC. The PSU is 80Plus Gold certified by CLEAResult and Platinum certified by Cybenetics. The unit consistently met 80Plus Gold requirements across its load range and narrowly missed the 80Plus Platinum certification at full load. Efficiency peaked at approximately 40% load, and its performance across the 10-100% load range was within the thresholds for Cybenetics’ Platinum certification. At very low loads, the PSU maintained acceptably high efficiency levels, suggesting strong performance across various operational conditions.
The Steel Legend SL-1000G exhibited a good balance between thermal management and acoustics during cold testing. The fan remained off until the load exceeded 350 watts, after which it started at a low rotational speed. As the load increased, the fan speed rose incrementally, but it never reached its maximum rotational speed during cold testing conditions. Internal temperatures were well-regulated, with thermal performance aligning with expectations for its design and specifications. Despite the fan’s operation, noise levels were relatively low, ensuring quiet operation under typical loads.
Hot Test Results
Hot Test Results (~45°C Ambient)
The ASRock Steel Legend SL-1000G exhibited an average nominal load efficiency of 89.5% at 115 VAC and 91.5% at 230 VAC during hot testing. The higher ambient temperatures caused a measurable drop in efficiency compared to cold testing, with the most significant degradation occurring under high loads, suggesting thermal stress when the unit is left heavily loaded for prolonged periods of time.
Image 1 of 5
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
(Image credit: Tom’s Hardware)
Thermal performance during hot testing showed an anticipated shift in fan behavior, with the fan starting at a load just below 300 watts. The fan speed increased more rapidly compared to cold testing, reaching its maximum rotational speed at approximately 900 watts. Internal temperatures obviously rise significantly higher but the PSU maintained a fairly good balance between cooling and noise output. The unit’s noise levels are moderate, with the SL-1000G managing to maintain very low figures for the first half of its operational range but noise levels rise sharply as the load increases beyond 600 watts.
PSU Quality and Bottom Line
Power Supply Quality
The ASRock Steel Legend SL-1000G demonstrated strong electrical performance in terms of voltage regulation and filtering. Voltage regulation was tight across all rails, with deviations of 0.8% on the 12V rail and 0.4% on both the 5V and 3.3V rails, indicating stable output under varying loads. Ripple suppression was effective, with the 12V rail exhibiting a maximum ripple of 54 mV, the 5V rail 20 mV, and the 3.3V rail 18 mV. While these ripple values are within acceptable limits for a high-performance PSU, they could have been lower, considering the unit’s market positioning.
During our thorough assessment, we evaluate the essential protection features of every power supply unit we review, including Over Current Protection (OCP), Over Voltage Protection (OVP), Over Power Protection (OPP), and Short Circuit Protection (SCP). All protection mechanisms were activated and functioned correctly during testing.
The SL-1000G’s protection features performed reliably, with Over Current Protection (OCP) limits set at 152% for the 3.3V rail, 150% for the 5V rail, and 134% for the 12V rail. Over Power Protection (OPP) was triggered at 136% during hot testing. The figures are a bit high but not unexpected for a premium ATX 3.1 design that is very likely to be facing massive power excursions.
Swipe to scroll horizontally
Main Output
Load (Watts)
201.97 W
Row 0 – Cell 2
503.27 W
Row 0 – Cell 4
753.53 W
Row 0 – Cell 6
1001.61 W
Row 0 – Cell 8
Load (Percent)
20.2%
Row 1 – Cell 2
50.33%
Row 1 – Cell 4
75.35%
Row 1 – Cell 6
100.16%
Row 1 – Cell 8
Amperes
Volts
Amperes
Volts
Amperes
Volts
Amperes
Volts
3.3 V
1.85
3.38
4.61
3.37
6.92
3.36
9.23
3.36
5 V
1.85
5.05
4.61
5.05
6.92
5.04
9.23
5.03
12 V
15.39
12.11
38.48
12.07
57.71
12.05
76.95
12.01
Swipe to scroll horizontally
Line
Regulation (20% to 100% load)
Voltage Ripple (mV)
Row 0 – Cell 3
Row 0 – Cell 4
Row 0 – Cell 5
Row 0 – Cell 6
Row 0 – Cell 7
Row 1 – Cell 0
Row 1 – Cell 1
20% Load
50% Load
75% Load
100% Load
CL1 12V
CL2 3.3V + 5V
3.3V
0.4%
10
12
14
18
14
18
5V
0.4%
12
12
16
20
16
18
12V
0.8%
18
28
32
54
50
24
Bottom Line
The ASRock Steel Legend SL-1000G is positioned as a mid-to-high-end power supply unit, targeting gamers and PC enthusiasts who require a modern PSU compliant with ATX 3.1 and PCIe 5.1 standards. At a price point of $160, it provides features such as modular cabling, 80Plus Gold and Cybenetics Platinum certifications, and native support for high-power GPUs. While reasonably priced for its specifications, the unit faces stiff competition in a crowded market, where established brands offer similar or better performance and efficiency at comparable costs.
(Image credit: Tom’s Hardware)
The SL-1000G’s design and build quality align with ASRock’s Steel Legend branding, emphasizing durability and a distinctive aesthetic. The PSU’s matte black chassis, rounded edges, integrated fan grille, and embossed details give it a premium appearance. Internally, the PSU employs high-quality components, including Japanese Nippon Chemi-Con and Rubycon capacitors, paired with a robust platform from HEC/Compucase. Its compact 150mm form factor ensures compatibility with most ATX cases, with the slight increase in size over standard units affecting only very compact cases and custom builds.
(Image credit: Tom’s Hardware)
Performance-wise, the SL-1000G delivers solid efficiency and reliable operation across a wide range of conditions. Cold testing revealed peak efficiency at moderate loads, meeting 80Plus Gold and Cybenetics Platinum certifications. Voltage regulation was tight across all rails, and ripple suppression was within acceptable limits. Hot testing showed a minor efficiency drop, consistent with thermal stress under high-load conditions, but the unit maintained stability. Thermal management and acoustics were balanced, with fan noise remaining moderate even under heavy loads. All protection features performed as expected.
(Image credit: Tom’s Hardware)
In conclusion, the ASRock Steel Legend SL-1000G is a well-rounded PSU that offers modern compliance, decent efficiency, and robust protection features at a reasonable price. However, while it performs respectably, it does not distinguish itself significantly in a crowded market segment filled with alternatives that offer similar or slightly better performance for comparable prices. While its performance and build quality meet expectations for its class, the unit does not significantly outshine competitors in the same price range. It is a reliable choice for a 1000W PSU with a unique design and a complete accessory bundle. However, those who prioritize certain performance aspects or find another design more appealing may find alternative options more compelling.
Darktable manages your digital negatives in a database, lets you view them through a zoomable lighttable and enables you to develop raw images and enhance them.
Features
Non-destructive editing throughout the complete workflow, your original images are never modified.
Take advantage of the real power of raw: All darktable core functions operate on 4×32-bit floating point pixel buffers, enabling SSE instructions for speedups.
GPU accelerated image processing: many image opertions are lightning fast thanks to OpenCL support (runtime detection and enabling).
Professional color management: darktable is fully color managed, supporting automatic display profile detection on most systems, including built-in ICC profile support for sRGB, Adobe RGB, XYZ and linear RGB color spaces.
Cross platform: darktable runs on Linux, macOS / macports, BSD, Windows and Solaris 11 / GNOME.
Filtering and sorting: search your image collections by tags, image rating (stars), color labels and many more, use flexible database queries on all metadata of your images.
Image formats: darktable can import a variety of standard, raw and high dynamic range image formats (e.g. JPEG, CR2, NEF, HDR, PFM, RAF … ).
Zero-latency, zoomable user interface: through multi-level software caches darktable provides a fluid experience.
Tethered shooting: support for instrumentation of your camera with live view for some camera brands.
Speaks your language: darktable currently comes with 21 translations: Albanian, Catalan, Czech, Danish, Dutch, French, German, Greek, Hebrew, Hungarian, Italian, Japanese, Polish, Portuguese (Brazilian and Portuguese), Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.
Powerful export system supports G+ and Facebook webalbums, flickr upload, disk storage, 1:1 copy, email attachments and can generate a simple html-based web gallery. darktable allows you to export to low dynamic range (JPEG, PNG, TIFF), 16-bit (PPM, TIFF), or linear high dynamic range (PFM, EXR) images.
Never lose your image development settings darktable uses both XMP sidecar files as well as its fast database for saving metadata and processing settings. All Exif data is read and written using libexiv2.
Automate repetitive tasks: Many aspects of darktable can be scripted in Lua.
Modules
Darktable contains several image operation modules. Many modules support powerful blending operators offering blend functionality that works on the incoming image information and the output of the current module or be used with drawn masks.
Basic image operations:
Contrast, brightness, saturation: Quickly tune your image using this simple module.
Shadows and highlights: Improve images by lightening shadows and darkening highlights. Read Ulrich’s blog post on this.
Crop and rotate: This module is used to crop, rotate and correct the perspective of your image. It also includes many helpful guidelines that assist you using the tools (e.g. rule of thirds or golden ratio).
Base curve: darktable comes with general enhanced basecurve presets for several models that are automatically applied to raw images for better colors and contrast.
Exposure controls: Tweak the image exposure either by using the sliders in the module or dragging the histogram around.
Demosaic: You have the choice between several demosaicing methods when editing raw files.
Highlight reconstruction: This module tries to reconstruct color information that is usually clipped due to information not being complete in all channels.
White balance: A module offering three ways to set the white balance. You can set tint and temperature or you define the value of each channel. The module offers predefined white balance settings as well. Or just pick a neutral region in the image to balance for that.
Invert: A module inverting colors based on the color of film material.
Tone image operations:
Fill light: This module allows the local modification of the exposure based on pixel lightness.
Levels: This module offers the well-known levels adjustment tools to set black, grey and white points.
Tone curve: This module is a classical tool in digital photography. You can change the lightness by dragging the line up or down. darktable lets you separately control the L, a and b channel. Read in Ulrich’s blog post how to make use of this feature.
Zone system: This module changes the lightness of your image. It is based on the Ansel Adams system. It allows to modify the lightness of a zone taking into account the effect on the adjacent zones. It divides the lightness in a user-defined number of zones.
Local contrast: This module can be used to boost details in the image.
Two different tone mapping modules: These modules allow to recreate some contrast for HDR images.
Color image operations:
Velvia: The velvia module enhances the saturation in the image; it increases saturation on lower saturated pixels more than on high saturated pixels.
Channel mixer: This module is a powerful tool to manage channels. As entry, it manipulates red, green and blue channels. As output, it uses red, green, blue or grey or hue, saturation, lightness.
Color contrast
Color correction: This module can be used to modify the global saturation or to give a tint. Read Johannes’ blog post.
Monochrome: This module is a quick way to convert an image to black and white. You can simulate a color filter in order to modify your conversion. The filter can be changed in size and color center.
Color zones: This module allows to selectively modify the colors in your image. It is highly versatile and allows every transformation possible in the LCh colorspace.
Color balance: Use lift/gamma/gain to change highlights, midtones and shadows.
Vibrance: For a detailed description read Henrik’s blog post.
Color look up table: Apply styles or film emulations. You can also easily edit the changes done. For more information you can read this blog post
Input/output/display color profile management
A useful feature that displays pixels outside the dynamic range.
Correction modules:
Dithering: This helps with banding in smooth gradients in the final image.
Sharpen: This is a standard UnSharp Mask tool for sharpening the details of an image.
Equalizer: This versatile module can be used to achieve a variety of effects, such as bloom, denoising, and local contrast enhancement. It works in the wavelet domain, and parameters can be tuned for each frequency band separately.
Denoise (non-local means): Denoising with separated color / brightness smoothing.
Defringe: Remove color fringes on high contrast edges.
Haze removal: This module allows to remove the low contrast and color tint coming from haze and air pollution.
Denoise (bilateral filter): Another denoising module.
Liquify: Push image parts around, grow them, shrink them. More information can be found in this blog post
Perspective correction: A great module to automatically un-distort shots with straight lines. See our blog post for an introduction and examples.
Lens correction: lens defect correction using lensfun.
Spot removal: Spot removal allows you to correct a zone in your image by using another zone as model.
Profiled denoise: By measuring the typical noise of cameras at the different ISO levels darktable is able to remove a lot of it. Read this blog post for more information.
Raw denoise: Raw denoise allows you to perfom denoising on pre-demosaic data. It is ported from dcraw.
Hot pixels: This module allows you to visualize and correct stuck and hot pixels.
Chromatic aberrations: This module automatically detects and corrects chromatic aberrations.
Effects/artistic image postprocessing:
Watermark: The watermark module provides a way to render a vector-based overlay onto your image. Watermarks are standard SVG documents and can be designed using Inkscape. The SVG processor of darktable also substitutes strings within the SVG document which gives the opportunity to include image-dependent information in the watermark such as aperture, exposure time and other metadata.
Framing: This module allows you to add an artistic frame around an image.
Split toning: Original split toning method creates a two color linear toning effect where the shadows and highlights are represented by two different colors. darktable split toning module is more complex and offers more parameters to tweak the result.
Vignetting: This module is an artistic feature which creates vignetting (modification of the brightness/saturation at the borders).
Soften: This module is an artistic feature that creates the Orton effect also commonly known as softening the image. Michael Orton achieved such result on slide film by using 2 exposures of the same scene: one well exposed and one overexposed; then he used a technique to blend those into a final image where the overexposed image was blurred.
Grain: This module is an artistic feature which simulates the grain of a film.
Highpass: This module acts as highpass filter.
Lowpass: This module acts as lowpass filter. One use case is described in Ulrich’s blog post.
Lowlight vision: Low light module allows to simulate human lowlight vision, thus providing the ability to make lowlight pictures look closer to reality. It can also be used to perform a day to night conversion.
Bloom: This module boost highlights and softly blooms them over the image.
Color mapping: Transfer colors from one image to another.
Colorize
Graduated density: This module aims at simulating a neutral density filter, in order to correct exposure and color in a progressive manner.
What’s New
We’re proud to announce the new bug-fix release of darktable, 5.0.1!
Since Darktable 5.0.0:
105 commits to Darktable+rawspeed
57 pull requests handled
15 issues closed
Performance Improvements
Improved performance of discarding history of selected images.
Other Changes
Harmonized style selection in print settings with the update made in the export module for release 5.0.
Darktable now includes the Lensfun database into the AppImage.
Center collapsible module section labels.
Bug Fixes
Fixed a scrolling bug in the collections module where the collection tree was not scrolled to the previous position when returning from darkroom.
Fix crash when attempting to move up the top or only shape in a group in the mask manager.
Fix collection module when using the “module” filter which was listing none of the processing modules.
Fix iop-order user presets use. It was not possible to get a user’s iop-order preset to be applied as the Darktable internal presets where always picked up first.
Fixed a crash in variable expansion for $(SEQUENCE[n,m]).
Fixed handling on click from filmstrip when in map and print view. On map we now properly center the image on the map. On the print view, we use the selected image as content of the main print area.
Fixed a regression introduced in 5.0.0 where we lost the ability to read 2-channel TIFF files and files with more than 4 channels. This fix allows them to be read by the TIFF loader, whereas previously the TIFF loader rejected such files and they were loaded by a fallback loader (GraphicsMagick or ImageMagick).
Fixed not being able to pan with a mouse in the zoomable layout.
Fixed possibly collapsing the module header when restoring defaults.
Fixed a case of XMP being written without actual image modification.
Fixed a crash of Darktable when reimporting XMP having overlay images referenced.
Fixed the support for single lib module opened at a time when in darkroom.
Fixed xmp sidecar writing while working as a gimp plugin.
Fixed issues when applying a style in darkroom using the Styles module.
Fixed a possible crash when using Ctrl+z reproduced on macOS.
Fixed Darktable crashing in enlarge-canvas module.
Fixed Darktable crashing using details mask due to bad tiling.
Fixed a hash calculation bug leading to pixelpipe instabilities.
Fixed issue to handle auto-presets when the focal length is greater then 1000.
Fixed crash on macOS when closing the darkroom’s second window.
Fixed possible jump in filmstrip when entering darkroom from full preview.
Fixed a performance regression in haze removal module.
Fixed reading palette-based PNG with transparency.
Fixed possible crash in lens correction module when changing camera and/or lens model.
Fixed crash after deleting style or preset that was selected in shortcuts dialog.
Fixed a crash when the currently active module instance is being deleted.
We now support the correct orientation for JPEG XL if it was specified in the format’s metadata but not in Exif (or Exif is missing). We also avoid over-transformation if it is specified in both the basic data and Exif.
RawSpeed changes
Fujifilm GFX cameras now use the vendor supplied crop
Camera support, compared to 5.0.0
Base Support
Leica SL3-S (DNG)
Minolta DiMAGE 5
Panasonic DC-S5D (3:2)
White Balance Presets
Noise Profiles
Fujifilm GFX100 II
Fujifilm X-S20
Fujifilm X100VI
Missing Compression Mode Support
Apple ProRAW DNGs
CinemaDNG lossless (Blackmagic, some DJI, etc.) and lossy (Blackmagic)
DNG 1.7 using JPEG XL (Adobe enhanced, Samsung Expert RAW)
Fujifilm lossy RAFs
Nikon high efficiency NEFs
OM System 14-bit high resolution ORFs
Sony downsized lossless ARWs (“M” for full-frame, “S” for full-frame & APS-C)
Suspended Support
Support for the following cameras is suspended because no samples are available on https://raw.pixls.us:
Creo/Leaf Aptus 22(LF3779)/Hasselblad H1
Fujifilm IS-1
Kodak EasyShare Z980
Leaf Aptus-II 5(LI300059)/Mamiya 645 AFD
Leaf Credo 60
Leaf Credo 80
Olympus SP320
Phase One IQ250
Sinar Hy6/ Sinarback eXact
ST Micro STV680
Previous Release Notes:
The following is a summary of the main features added to darktable 5.0. Please see the user manual for more details of the individual changes (where available).
This development cycle has included a large number of changes which improve the user experience, as detailed in the next section.
UI/UX Improvements
Added camera-specific styles for more than 500 camera models to more closely approximate the out-of-camera JPEG rendition. These styles only affect contrast, brightness, and saturation and do not attempt to match sharpening, denoising, or hue shifts. Also added a Lua script to auto-apply the appropriate style on import and manually apply styles to a collection of previously-imported images.
Added an optional splash screen showing startup progress (including estimated time remaining during the scan for updated sidecar files) to dramatically reduce the time between invoking darktable and something appearing on screen when the user has a large library.
The user interface now gives feedback while processing bulk image operations such as rating, tagging, applying styles, and edit history management (and undoing those operations), rather than silently freezing until the operation completes. While the operation is in progress, darktable will now show either a busy cursor (such as a stopwatch or spinner) or a progress bar with option to cancel the remainder of the operation.
Paths for drawn masks now display two Bézier handles per control point, which can be moved individually. This allows for more precise control of the paths.
Added a high-contrast theme with bright white text on a dark gray background.
Enhanced tooltips for utility module headers to provide more information about the module.
Added more new-user hints on an empty lighttable.
Added two new error placeholder images to distinguish between missing, unsupported, and corrupted images. When attempting to edit such an image, an appropriate, more specific error message is displayed.
When selecting a style in the export module, hovering on the style name in the popup menu displays a thumbnail previewing the effect of appending the style to the active image’s edit (first selected image in lighttable, center-view image in darkroom).
Allow for selecting the utility modules to be displayed on the panels in the different views.
Right-click on the empty panel area below the modules to get a menu where they can be hidden or shown. This allows additional modules to be added to the darkroom, like metadata editor and styles.
This replaces the options in the “collections” and “recently used collections” modules’ preferences to show or hide the latter and show a “history” button in the former instead. Users that want the separate module will need to reenable it once via the new Right-click menu.
The menu also contains an option “restore defaults” that resets the selection and position of modules in the current view. In the preferences dialog, on the general tab, there’s a “reset view panels” button that resets all views, including visibility and width of the panels themselves.
Added a global preference to swap the left and right side panels in the darkroom view.
The first time a new user presses Tab, they will be warned that this will hide all panels and how to get them back. Hopefully this prevents some confusion or frustration.
Drag&drop utility module headers to reposition them across the left and right panels (lighttable) as well as vertically (all views). Each view can have a different layout.
Drag&drop of processing modules in the darkroom right panel has been improved to auto-scroll when reaching the top or bottom and to not get confused when images get dragged into the area. This functionality no longer requires Ctrl+Shift modifiers.
Improved the message displayed at startup when the database is locked by another instance of darktable.
Replaced the icon of the operator button in the color label filter for working with multiple selected color labels (union/intersection).
Performance Improvements
Added OpenCL implementation of color equalizer.
Improved the speed of bulk image operations by improving the speed of sidecar writes, and by moving sidecar updates for many operations into a background task, allowing the user to proceed before the writes complete.
Significantly accelerated loading of PFM files due to loops parallelization and optimization that eliminated additional processing.
Other Changes
Switched default scope for new installations from histogram to waveform to display more detailed information about image color and tonality.
The ISO 12646 color assessment condition is kept until unset by user action.
Exposure bias can now be used to form collections and as a display filter.
Improved visualization of the color equalizer’s effect.
Improved debugging support for verifying CPU vs. GPU results.
Add Calibrite alias for X-Rite ColorChecker in color calibration.
The scan for updated sidecar files now ignores timestamp differences of two seconds or less.
The macOS installation package now has a background image to direct the user on installing darktable.app.
Changed the user interface of the import dialog to make it easier to delete custom places.
Numerous rounds of code cleanup.
The copy-parts dialog does not select any module by default now.
Add support for undo/redo for actions done on the filmstrip while in darkroom.
In darkroom, add action (binding to Ctrl+x by default) for synchronizing the last edited module on current edited module to the selection.
Adjusted the internal AVIF encoder parameter to significantly boost encoding speed without compromising the output quality.
Tag names can now easily be copied to the clipboard via popup context menu in the tagging module.
The Piwigo export storage now supports to specify a file name pattern for the exported file.
The directory where darktable will write the log file under Windows has been changed to %USERPROFILE%\Documents\Darktable. This allows the user to easily see where the log file is located without even having to search for it in the documentation or FAQ. The previous location was deep in the system subdirectories of the user profile, and also under a hidden directory (so it was impossible to click to it in File Explorer with default system settings).
Allow import of JPEG 2000 files with .jpf and .jpx file extensions.
Add a visible indicator to the color calibration module when its color mapping section has non-neutral settings which will affect color rendition.
Added new substitution variables$(IMAGE.TAGS.HIERARCHY)to insert tags with full hierarchy and$(IMAGE.ID.NEXT)to insert the image ID to be assigned to the image being imported, allowing the image ID to be part of the filename generated during a copy&import operation.
Exporting to floating-point JPEG XL with a quality of 100 will try to do it as losslessly as possible. That is now consistent with the behavior of integral JPEG XL formats.
Improved visibility of shortcuts that can be changed by users by using bold text.
The histogram-exposure interface now supports all standard bauhaus features (Ctrl+click, Right-click…).
Introduce image module order v5.0 to have the final-scale done before color-out to fix some issues with color difference between darkroom view and exported files.
Add support for editing any live color-picker samples. Using Right-click on a sample it is possible to edit it (changing location and/or size of the box) and either add a new sample based on the edit or store the edit into an existing live sample.
Added more substitution variables for using EXIF data fields, enabled autocompletion of variables in the watermark module.
The new variables are$(EXIF.FLASH),$(EXIF.METERING),$(EXIF.EXPOSURE.PROGRAM),$(EXIF.WHITEBALANCE)and$(GPS.LOCATION.ICON).
Increase maximum focal length for filtering auto-applied presets to 2000mm.
Added an expanded color-checker preset to the Color Look Up Table module with seven-level red/green/blue/gray ramps, IT8/CC24-like skin tones, and miscellaneous color patches for more targeted color adjustments across the full spectrum.
Added support for EXIF tags ‘AnalogBalance’ used for color calibration and ‘LinearResponseLimit’ used in highlights reconstruction.
If we find currently unsupported color calibration data in DNG specific tags, we tag the image by darktable|issue|no-samples for better support.
Added read support for HEIF files with AVC (H.264) compression and .avci file extension.
Added read support for JPEG 2000 encoded images in HEIF containers with .hej2 file extension.
Bug Fixes
Fixed a performance regression for redrawing mipmaps.
Fixed handling of old (2020) edits using Filmic RGB.
Various OpenCL fixes to reduce differences between CPU and GPU processing: colorspace conversion, saturation gradient filter in color equalizer.
Fixed gallery export not working on Windows.
Fixed printer discovery in the print module, which could cause available printers to be missed.
Work around out-of-spec EXIF date field caused by buggy software.
Fixed reading embedded color profiles from PNG images.
Fixed certain boundary cases in the crop module.
Fixed crash when loading corrupted .gpx file in the geotagging module
Fix preset handling in the export module not saving all parameters.
Fix an issue in FilmicRGB where one of the parameter could be above the maximum allowed range making the validation failing and the whole set of parameters reset to default.
Fix overlay recording to work in all cases (discarding history or copy/paste history for example) ensuring that an image not referenced anymore as overlay in a composite module can be removed.
Properly reset darktable internal tag darktable|style| and darktable|changed when resetting history.
Fixed crash in the Piwigo export storage when not logged in to the Piwigo server.
Fixed a bug in the export module where it was impossible to export a file again if “on conflict: overwrite if changed” was selected.
Fixed a bug where double clicking on a label in darkroom modules does not reset the control.
The composite module now prevents assigning an overlay that would lead to a loop. Previously, only direct references (image #1 <-> image #2) were checked; this has now been extended to also cover chains (image #1 -> image #2 -> image #3 -> image #1) of arbitrary length.
Fix a bug in overlay module which incorrectly apply a color profile and so creating an unwanted and wrong color cast. This bug was a regression added just before the 4.8 release.
Fixed a bug in color calibration module where switching between various illuminants could lead to unpredictable settings.
Various fixes In the demosaic module. Non-usable options are hidden now. Fixed dual demosaicing for xtrans sensors and OpenCL code.
Fixed a bug in the history module where style creation fails if a style with that name already exists.
Fixed guides drawing in case a module is expanded and active.
Ensure that the list of images in the culling view remains up to date when hidden.
Fixed minor glitches in color calibration module.
Fixed issues with wrong corrections in highlight opposed OpenCL code.
Fixed surface blur radius calculation possibly resulting in garbled output.
Camera support, compared to 4.8
Base Support
Fujifilm X-M5 (compressed)
Fujifilm X-T50 (compressed)
Leica D-Lux 8 (DNG)
Leica M11-D (DNG)
Leica Q3 43 (DNG)
Minolta Alpha Sweet Digital
Minolta Alpha-7 Digital
Nikon Z50_2 (14bit-compressed)
Nikon Z6_3 (14bit-compressed)
Panasonic DC-FZ80D (4:3)
Panasonic DC-FZ82D (4:3)
Panasonic DC-FZ85 (4:3)
Panasonic DC-FZ85D (4:3)
Panasonic DC-G100D (4:3)
Phase One P20+
Sony ILCE-1M2
White Balance Presets
Nikon Z6_3
Sony ILCE-6700
Noise Profiles
Canon PowerShot G1 X
Leica M11
Nikon Z6_3
Missing Compression Mode Support
Apple ProRAW DNGs
CinemaDNG lossless (Blackmagic, some DJI, etc.) and lossy (Blackmagic)
DNG 1.7 using JPEG XL (Adobe enhanced, Samsung Expert RAW)
Fujifilm lossy RAFs
Nikon high efficiency NEFs
OM System 14-bit high resolution ORFs
Sony downsized lossless ARWs (“M” for full-frame, “S” for full-frame & APS-C)
Suspended Support
Support for the following cameras is suspended because no samples are available on https://raw.pixls.us:
Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More
Stats might be everything in basketball — but for Pacers Sports and Entertainment (PS&E), data about fans is just as valuable.
Yet while the parent company of the Indianapolis Pacers (NBA), the Indiana Fever (WNBA) and the Indiana Mad Ants (NBA G League) was pumping untold amounts of it into a $100,000-a-year machine learning (ML) platform to generate predictive models around such factors as pricing and ticket demand, the insights weren’t coming fast enough.
Jared Chavez, manager of data engineering and strategy, set out to change that, making the move to Databricks on Salesforce a year-and-a-half ago.
Now? His team is performing the same range of predictive projects with careful compute configurations to gain critical insights into fan behavior — for just $8 a year. It’s a jaw-dropping, seemingly unthinkable decrease Chavez credits largely to his team’s ability to reduce ML compute to near-infinitesimal amounts.
“We’re very good at optimizing our compute and figuring out exactly how far we can push down the limit to get our models to run,” he told VentureBeat. “That’s really what we’ve been known for with Databricks.”
PS&E cuts OpEx by 98%
In addition to its three basketball teams, the Indianapolis-based PS&E operates a Pacers Gaming esports business, hosts March Madness games and runs a busy, 300-plus day event business through the Gainbridge Fieldhouse arena (concerts, comedy shows, rodeos, other sporting events). Further, the company just last month announced plans to build a $78 million Indiana Fever Sports Performance Center, which will be connected by skybridge to the arena and a parking garage (expected to open in 2027).
All this makes for a mind-boggling amount of data — and data sprawl. From a data infrastructure standpoint, Chavez pointed out that, up until two years ago, the organization hosted two completely independent warehouses built on Microsoft Azure Synapse Analytics. Different teams across the business all used their own form of analytics, and tooling and skill sets varied wildly.
While Azure Synapse did a great job connecting to external platforms, it was cost-prohibitive for an organization of PS&E’s size, he explained. Also, integrating the company’s ML platform with Microsoft Azure Data Studio led to fragmentation.
To address these problems, Chavez switched over to Databricks AutoML and the Databricks Machine Learning Workspace in August 2023. The initial focus was to configure, train and deploy models around ticket pricing and game demand.
Both technical and non-technical users immediately found the platforms helpful, Chavez noted, and they quickly sped up the ML process (and plummeted costs).
“It dramatically improves response times for my marketing team, because they don’t have to know how to code,” said Chavez. It’s all buttons for them, and all that data comes back down to Databricks as unified records.”
Further, his team organized the company’s 60-some-odd systems into Salesforce Data Cloud. Now, he reports that they have 440X more data in storage and 8X more data sources in production.
PS&E today operates at just under 2% of its previous annual OPEX costs. “We saved hundreds of thousands a year just on operations,” said Chavez. “We reinvested it into customer data enrichment. We reinvested into better tooling for not just my team, but the analytics units around the company.”
Continued refinement, deep understanding of data
How did his team get compute so staggeringly low? Databricks has continually refined cluster configurations, enhanced connectivity options to schemas and integrated model outputs back into PS&E’s data tables, Chavez explained. The powerful ML engine is “continuously enriching, refining, merging and predicting” on PS&E’s customer records across every system and revenue stream.
This leads to better-informed predictions with each iteration — and in fact, the occasional AutoML model sometimes makes it straight to production without any further tweaking from his team, Chavez reported.
“Truthfully, it’s just knowing the size of the data going in, but also roughly how long it is going to take to train,” said Chavez. He added: “It’s on the smallest cluster size you could possibly run, it might just be a memory-optimized cluster, but it’s just knowing Apache Spark fairly well and knowing which way we could store and read the data fairly optimally.”
Who’s most likely to buy season tickets?
One way Chavez’ team is using data, AI and ML is in propensity scoring for season tickets packages. As he put it: “We sell an ungodly number of them.”
The goal is to determine which customer characteristics influence where they choose to sit. Chavez explained that his team is geo-locating addresses they have on file to make correlations between demographics, income levels and travel distances. They’re also analyzing users’ purchase histories across retail, food and beverage, mobile app engagement and other events they might attend on PS&E’s campus.
Further, they’re pulling in data from Stubhub, Seat Geek and other vendors outside of Ticketmaster to evaluate price points and determine how well inventories are moving. This can all be married with everything they know about a given customer to figure out where they’re going to sit, Chavez explained.
Armed with that data, they could then, for instance, upsell a given customer from Section 201 to section 101 center court. “Now we’re able to not only resell his seat in the higher deck, we can also sell another smaller package on the same seats he purchased in the mid-season, using the same characteristics for another person,” said Chavez.
Similarly, data can be used to enhance sponsorships, which are critical to any sports franchise.
“Of course, they want to align with organizations who overlap with theirs,” said Chavez. “So can we better enrich? Can we better predict? Can we do custom segmentation?”
Ideally, the goal is an interface where any user could ask questions like: ‘Give me a section of the Pacers fan base in their mid-to-late 20s with disposable income.’ Going even further: ‘Look for those that make more than $100K a year and have an interest in luxury vehicles.’ The interface could then bring back a percentage that overlap with sponsor data.
“When our partnership teams are trying to close these deals, they can, on-demand, just pull information without having to rely on an analytics team to do it for them,” said Chavez.
To further support this goal, his team is looking to build out a data clean room, or a secure environment that allows for the sharing of sensitive data. This can be particularly helpful with sponsors, as well as collaborations with other teams and the NCAA (which is headquartered in Indianapolis).
“The name of the game for us right now is response time, whether that’s customer facing or internal,’ said Chavez. “Can we dramatically lessen the required knowledge to cut up information and sort through it using AI?”
Data collection and AI to understand traffic patterns, improve signage
Another area of focus for Chavez’s team is examining where people are at any given time across PS&E’s campus (which comprises a three-tier arena with an outdoor plaza). Chavez explained that data capture capabilities are in place throughout its network infrastructure via WiFi access points.
“When you walk into the arena, you are pinging off all of them, even if you don’t log into them, because your phone’s checking for WiFi,” he said. “I can see where you’re moving. I don’t know who you are, but I can see where you’re moving.”
This can eventually help guide people around the arena — say, if someone wants to buy a pretzel and is looking for a concession stand — and help his team determine where to position food and merchandise kiosks.
Similarly, location data can help determine optimal spots for signage, Chavez explained. One interesting way to identify signage impression counts is placing vision gradients at spots equivalent to average fan height.
“Then let’s calculate how well somebody would have seen this walking through with the number of people around them,” said Chavez. “So I can tell my sponsor you got 5,000 impressions on this, and 1,200 of them were pretty good.”
Similarly, when fans are in their seats, they are surrounded by signs and digital displays. Location data can help determine the quality (and amount) of impressions based on the angle of where they’re sitting. As Chavez noted: “If this ad was only on the screen for 10 seconds in the third quarter, who would have seen it?”
Once PS&E has adequate locational data to help answer these types of questions, his team plans to work with Indiana University’s VR lab to model the entire campus. “Then we’re just going to have a very fun sandbox to go run around in and answer all these 3D space questions that have been bugging me for the last two years,” said Chavez.
Daily insights on business use cases with VB Daily
If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.
Read our Privacy Policy
Thanks for subscribing. Check out more VB newsletters here.
It happens like clockwork, around the first of each month. Sites that focus on technology churn out nearly identical articles, all based on a chart like this one, prepared by the good folks at Statcounter Global Stats.
Every month, tech bloggers try to turn a chart like this one into a story, but most of them miss what’s actually happening.
Statcounter GS
I saw that chart in dozens of posts this month, along with detailed explanations of what the author thought the underlying data points mean. Sometimes the authors of these posts even convince an industry analyst to share their thoughts. It’s stereotypical horserace coverage.
Also: How to upgrade your ‘incompatible’ Windows 10 PC to Windows 11 in 2025
This month, the challenge for every pundit was to explain why Windows 10 (the purple line at the top) appears to be suddenly collapsing in popularity and why Windows 11 (the blue line in the middle) has regained its mojo.
Here’s a small sampling of some of the stories that the Statcounter “Windows market share” report inspired.
At the start of every month, stories like these appear, all based on the same charts.
Screenshot by Ed Bott/ZDNET
My favorite quote from this batch is from Forbes, which made this bold assertion: “The January stats are now out, and according to Statcounter, the Windows 11 upgrade trend has now un-reversed itself … some 40 million hold-outs have suddenly upgraded their PCs in the last 31 days.”
Also: Don’t ignore Microsoft’s February Patch Tuesday – it’s a big one
Inevitably, these posts try to answer the question, “Why did this happen?” But maybe a better question is, “Did this happen?” followed by “Are you sure?” and “Why doesn’t any of this data make sense?”
Because here’s the reality: Statcounter’s “market share” reports are a great excuse for tech bloggers to crank out a story each month, but they bear only the most casual relation to the real world, and most of those month-to-month spikes are simply statistical noise.
Let me show you what I mean with another chart, which I created using data I downloaded from Statcounter’s site. For this one, I changed the parameters to include data from January 2022 through January 2025, covering only the United States. After plugging that data into Excel, I created a line chart like the ones they publish, but I made two changes. First, I added third-order polynomial trendlines for the Windows 10 and Windows 11 data points to show the general direction of those monthly figures over time. Then, I added a shadow on either side of that trendline to indicate a likely margin of error.
Adding a trendline helps to smooth out the statistical fluctuations and reveals the real patterns in this data.
Chart by Ed Bott/ZDNET; data from Statcounter GS
Well, that tells a very different story, doesn’t it?
At the sites that use Statcounter’s web analytics service, pageviews from PCs running Windows 10 are steadily declining, while pageviews from PCs running Windows 11 are steadily increasing. And those trends have been consistent over time, despite some fluctuations in the data.
You’ll notice that description doesn’t mention “market share.” Statcounter’s data counts pageviews, not visits, or sessions, or individual devices.
Also: Microsoft has a big Windows 10 problem, and it’s running out of time to solve it
And make no mistake about it, those monthly fluctuations really are just noise. Look at the teal-colored line for Windows 8.x versions from January 2024. Do you really think that 10 or 20 million people fired up their old Windows 8 devices on New Year’s Day, used them for a few weeks, and then put them all back in the closet? That’s unlikely.
None of those other monthly spikes mean anything either. Did millions of people uninstall Windows 11 in December 2024 and then change their mind a month later? Of course not. The data is just messy!
Now, let me be crystal clear here: I don’t blame the Statcounter folks for taking advantage of an irresistible opportunity to generate publicity. I do, however, want to have a serious talk with every journalist and analyst who relies on Statcounter’s charts without questioning the underlying data behind them, because those numbers can’t stand up to even the mildest questioning.
Who is Statcounter?
Statcounter is a web analytics company based in Ireland. It was founded in 1999, during the Web 1.0 era, with a simple business model of counting “hits” to websites using a tracking pixel that clients embedded on their pages. If you’re a website owner, adding Statcounter’s tracking technology to your site can give you valuable information about your visitors.
It was a good business for a long time, but over the years the company’s customer base has shrunk. In 2009, it boasted that 3 million customers were using its service. By 2022, its own pages acknowledged that the customer base had been cut in half, to 1.5 million websites.
Also: How to do a clean install of Windows 11: See which option is best for you and why
W3Techs, which tracks companies in this space, reported that 0.9% of all websites were using Statcounter’s services in 2019. By January 2024, that number had shrunk to 0.5%, and when I checked again in January 2025, the number was down again, to 0.4%.
None of that decline should be surprising. Google Analytics dominates this space today, and other big players, like Meta Pixel, WordPress Jetpack, and Adobe Analytics, have also stolen share from tiny firms like Statcounter.
Where do Statcounter’s numbers come from?
Statcounter’s customer base consists of a lot of small websites and a few medium-sized ones. The Statcounter Global Stats reports aggregate all the pageviews from those sites, with details about the visitors, including the hardware type, operating system, and browser, as collected by that tracking.
Also: If your Windows 10 PC can’t be upgraded, you have 5 options before time runs out
Statcounter’s data collection has declined dramatically in the last decade. A decade ago, the company’s FAQ page reported that it measured more than 17 billion pageviews in a typical month. By 2022 (the last numbers that Statcounter has provided on the current FAQ page), that number was down to 5 billion a month.
Statcounter represents a tiny sliver of actual traffic on the web, mostly from fairly esoteric websites that have chosen to embed the Statcounter tracking code on their websites, like Futbin.com, Filmyzilla.com.fj, Ask.com, and Kernel.org. They can’t count traffic from the most popular sites on the web, like Google, Facebook, or Wikipedia.
Also: Can you still get a Windows 10 upgrade for free in 2025? Short answer: Maybe
It’s like trying to do a survey of consumer behavior without including Costco, Walmart, Home Depot, Target, CVS, or any Kroger grocery store. By leaving out those giants, your sample becomes quirky and almost certainly not representative of the greater market.
More importantly, Statcounter measures only pageviews, not visits or sessions. If I go to a site that uses the Statcounter service and visit five pages on my Windows 11 PC, and you load 10 pages with your Windows 10 PC, the results in Statcounter’s “market share” report will show that Windows 10 is twice as popular as Windows 11. You see the problem here, I presume.
Of course, that assumes all those pageviews are even counted. On my Windows 11 PC, where I use Microsoft Edge with its tracking protection set to Strict, Statcounter’s tracking code is automatically blocked. Oops.
So, what’s the real story?
The data from Statcounter tells a perfectly valid story about how people use websites that belong to its customers. But it says nothing about the “market share” for Windows PCs.
It does show, in the most general terms, that traffic to those sites from Windows 10 PCs is declining slowly and that traffic from PCs running Windows 11 appears to be increasing slowly as well. Do those numbers map to the population of PCs worldwide? Probably not, although no one can say for sure.
Also: How to clear the cache on your Windows 11 PC (and why it makes such a big difference)
We do know that there are a very large number of PCs running Windows 10 that are not eligible to upgrade to Windows 11. That number will probably still be very large in October 2025, when support for Windows 10 ends. Someone with access to Microsoft’s telemetry servers could probably give you a pretty good estimate of how many devices are in each population, but they’re not talking.
The rest of us, unfortunately, are left to guess. And if you want to make a wager based on data from Statcounter, go right ahead. Just don’t put any serious money on that bet.
Cho Yeo-jeong and Song Kang-ho in ParasiteNeon / Neon
Five years after it took the Cannes Film Festival, the Academy Awards, and the global box office by storm, Parasite is back on the big screen. That’s a pretty short amount of time to commemorate with an anniversary rerelease. Then again, it’s been a long five years, hasn’t it? Oh, how the world has changed since the halcyon days of 2019 — “the last f***ing year for cinema,” to quote Quentin Tarantino, whose Once Upon a Time… in Hollywood premiered within hours of Bong Joon-ho’s darkly ingenious class-warfare caper. In retrospect, Parasite winning Best Picture on Oscar night a few months later really did feel like the final joyous surprise of an old era — a last gasp before COVID closed theaters and changed everything.
Watching the film today, via its current IMAX victory lap or from the comfort of your own home, actually emphasizes the ways things haven’t changed so much these past five years. Or maybe it’s just that the tensions and resentments Parasite dramatized back then have fully exploded to the surface of the culture, as surely as they simmer to a boil in the climax of the movie. Like Mr. Kim (Song Kang-ho), people are fed up. Does his shocking act of violence look so shocking these days? You might think suddenly of a real, more premeditated crime that just stirred the media into a frenzy and sent a jolt of understanding through the public — a murder with a motive so clear, the killer etched it on the bullets.
Yes, the time is right for a Parasite encore. It’s very much a movie for the moment, a tale of rage and desperation fit for our post-Luigi Mangione world. At the same time, there’s no mistaking this South Korean award-winner — more acclaimed, perhaps, than any film that’s come in its wake — for a simple eat-the-rich parable. Not with Bong at the reins. The sheer slipperiness of his demented take on upstairs-downstairs drama is what elevates it beyond mere timeliness.
Parasite [Official Trailer] – In Theaters October 11, 2019
The title alone is provocative in its potential double meaning. Who are the parasites in this story of a destitute family, the Kims, that latches itself to the payroll teat of a wealthy family, the Parks? It’s less a trick question than a Rorschach test. Parasite’s huge success the world over could reflect a certain universal frustration, the kind that the murder of a CEO just illuminated like a black light. Or it could be chalked up to people seeing what they want to see in the master-servant relationship that develops between these economically entwined clans. To win the top prize on Hollywood’s biggest night, the film had to have spoken to people who saw a kind of horror movie about the dangers of opening your house to the help.
Bong leaves us to sort through our biases. He’s too shrewd a dramatist to reduce his characters to emblems of their social station — to divide our sympathies by tax bracket alone. The Parks are not cartoon fat cats. They are clueless and easily manipulated, in the case of Mrs. Park (Cho Yeo-jeong), or snobbishly condescending, in the case of Mr. Park (Lee Sun-kyun), who prefers those he hires respect professional boundaries and not “cross the line.” Their most odious offense is rudely fixating on odor. They are awful in the casual, everyday way that the wealthy can be. They are recognizably human, not blatant guillotine fodder.
Likewise, the Kims are not cardboard saints — the noble working-class heroes a more self-righteous parable might position upon a pedestal. They are, at times, bluntly and hilariously underhanded. They lie and steal and screw over other people for a shot at the cash the Parks thoughtlessly splurge on luxury and convenience and creature comforts. Parasite says that desperate times call for desperate measures. Every transgression in the movie is a scramble to survive. “Money is an iron,” as Mrs. Kim (Jang Hye-jin) says. It smooths out the difficulties of life — and with them, the challenges to our moral compasses.
ParasiteNeon / Neon
No one widely familiar with Bong’s work could confuse Parasite for a pro-elite cautionary tale, even if its narrative wickedly toys with the Saltburn-like fears of a ruling class paranoid that the Great Unwashed are coming for them. Rarely one for didactic screeds, the writer-director has spent most of his career smuggling his class politics onto screens under cover of rollicking genre scenarios. Take The Host, which savages American imperialism and corporate disregard for public safety in the form of a primo kaiju movie of original Gojira vintage. Or Snowpiercer, which takes late-stage capitalism to its logical endpoint, with all that remains of civilization stuffed aboard a train endlessly looping a ravaged planet. It’s arranged like a social ladder turned on its side, the poor in the back, the rich in the front.
Parasite flips that hierarchical structure vertically again: Here, privilege is a matter of altitude — a theme established right from the subterranean opening shot. The film is one of Bong’s most purely entertaining Trojan horses. That, maybe more than its politics, might account for its enduring popularity and for its ability to smash the language barriers that traditionally keep films not in English off the box-office charts and out of the Academy’s winner circle. For a while, it almost plays like an Ocean’s Eleven-style heist movie where the “score” is gainful employment. And the twists arrive with dizzying force, Bong opening his satirical scenario up into the realm of perverse farce (no doors are slammed, but tables are scurried under) and downright Hitchcockian thrills. Parasite remains about as fun as any movie this ultimately, witheringly downbeat can be.
The Parks aren’t the real villains of the film. They’re but a symptom of an unjust system. Capitalism is a zero sum game in Parasite. It keeps the 99% divided, fighting over the same crumbs, scrambling for the same tiny wedge of the pie. Most of the violence in the movie is between the Kims and the family of the housekeeper (Lee Jung-eun) they muscle out of a job. Only in the climax, when all hell breaks loose at that party, do the Parks experience any kind of reckoning. And it’s hard to call that a victory, regardless of where your sympathies lie. After all, Mr. Kim’s actions don’t even radicalize his son, Ki-woo (Choi Woo-shik), who remains seduced by a fantasy of upward mobility, dreaming of keeping up with the Joneses (or Parks).
Choi Woo-shik in ParasiteNeon / Neon
Since 2019, Parasite’s finger hasn’t left the pulse of an economically lopsided world. The Kims don’t see the big picture of their misadventure — namely, that they’re trying to win a rigged game on an uneven playing field, and that they’re trapped in a system designed to keep them literally and figuratively down. But the audience might see that picture more clearly than ever. After all, Parasite has returned to theaters at a time when newsfeeds are filled with daily reminders of the unjust disparity of life under capitalism. It remains a crazed crowdpleaser for an exploited world, building madcap complications around its social conscience… never mind the inconvenient truth that every ticket bought puts a few extra dollars in the pockets of the Friedkins, a family whose wealth makes the Parks look like the Kims.
All that said, you have to wonder if the pointedly unsubtle Parasite might actually be a little too subtle for 2025. In a key scene, the Kims sit around, drinking the Park family’s liquor, enjoying the mirage of a life within the household they serve. “She’s rich, but still nice,” Mr. Kim says of his employer, to which his wife replies, with a snort: “She’s rich, therefore she’s nice.” Right now, though, the mask of niceness the obscenely monied often wear has slipped. As these words are written, and maybe as you read them, the world’s richest man is waging war on the middle and lower class in broad daylight — unlawfully destroying public services, hurting American workers to line his own pockets, slashing cancer research for kids. Him and his kind are monsters too broad for the broadest of satires, the kind Bong would be embarrassed to put in a movie. They’re parasites in the truest sense.
Parasite is currently playing on select IMAX screens, streaming on Netflix, and available to rent or purchase from digital services owned by wealthy men who don’t care about you. For more of A.A. Dowd’s writing, visit his Authory page.
The fiber internet coverage in Aiken is increasing and if your home has access to it, fiber internet is the best choice for most. AT&T Fiber is CNET’s pick for the best internet service provider in Aiken. It gets the top spot because it offers speeds going up to 5,000Mbps at reasonable prices, and also has high reliability. You also don’t have to face the hassles of contracts or data limits, nor do you have to pay for equipment rental, making it a great deal.
If you live outside the coverage areas of AT&T, you can opt for one of the other options like Breezeline, Xfinity or T-Mobile, especially if affordability is a factor. Breezeline and Xfinity have some affordable internet plans starting at just $20 and $30, respectively. Xfinity offers a slightly higher speed of 150Mbps compared to 100Mbps from Breezeline, but equipment costs may be higher with Xfinity and a 1.2TB data cap may apply.
Aiken, South Carolina, internet providers compared
Provider
Internet technology
Monthly price range
Speed range
Monthly equipment costs
Data cap
Contract
CNET review score
AT&T Read full review
DSL, fiber
$60 DSL, $55-$245 fiber
10-100Mbps DSL, 300-5,000Mbps fiber
None
None
None
7.4
Breezeline
Cable
$25-$60
100-1,000Mbps
Free for 12 months, $18 after (optional)
None
None
N/A
T-Mobile Home Internet Read full review
Fixed wireless
$50-$70 ($35-$55 with eligible mobile plans)
87-415Mbps
None
None
None
7.4
Verizon 5G Home Internet Read full review
Fixed wireless
$50-$70 ($35-$45 with eligible mobile plans)
50-1,000Mbps
None
None
None
7.2
Xfinity Read full review
Cable
$30-$95
150-2,000Mbps
$15 (optional)
1.2TB
1-2 years (optional)
7
Show more (0 item)
Source: CNET analysis of provider data.
What is the cheapest internet plan in Aiken?
Plan
Starting price
Max download speed
Monthly equipment fee
Breezeline Core
$25
100Mbps
Free for 12 months, $18 after (optional)
Xfinity Connect Read full review
$30
150Mbps
$15 (optional)
Breezeline Fast
$30
200Mbps
Free for 12 months, $18 after (optional)
Xfinity Connect More Read full review
$35
300Mbps
$15 (optional)
Verizon 5G Home Internet Read full review
$50 ($35 with eligible mobile plans)
300Mbps
None
T-Mobile Home Internet Read full review
$50 ($35 with eligible mobile plans)
318Mbps
None
AT&T Fiber 300 Read full review
$55
300Mbps
None
Show more (2 items)
Source: CNET analysis of provider data.
How to find internet deals and promotions in Aiken
The best internet deals and top promotions in Aiken depend on what discounts are available at the time. Most deals are short-lived, but we look frequently for the latest offers.
Aiken internet providers, such as Xfinity, may offer lower introductory pricing or streaming add-ons for a limited time. Many, including AT&T Fiber, Breezeline and T-Mobile Home Internet, run the same standard pricing year-round.
How many members of your household use the internet?
For a more extensive list of deals, check out our guide on the best internet deals.
Fastest internet plans in Aiken
Plan
Starting price
Max download speed
Max upload speed
Data cap
Connection type
AT&T Internet 5000 Read full review
$245
5,000Mbps
5,000Mbps
None
Fiber
AT&T Internet 2000 Read full review
$145
2,000Mbps
2,000Mbps
None
Fiber
Xfinity Gigabit x2 Read full review
$95
2,000Mbps
200Mbps
1.2TB
Cable
AT&T Internet 1000 Read full review
$60
1,000Mbps
1,000Mbps
None
Fiber
Breezeline GigaFast
$60
1,000Mbps
50Mbps
None
Cable
Xfinity Gigabit Read full review
$65
1,000Mbps
20Mbps
1.2TB
Cable
Verizon 5G Home Plus Internet Read full review
$70 ($45 with eligible mobile plans)
85-1,000Mbps
50-75Mbps
None
Fixed wireless
Show more (2 items)
Source: CNET analysis of provider data.
What’s a good internet speed?
Most internet connection plans can now handle basic productivity and communication tasks. If you’re looking for an internet plan that can accommodate video conferencing, streaming video or gaming, you’ll have a better experience with a more robust connection. Here’s an overview of the recommended minimum download speeds for various applications, according to the Federal Communications Commission. Note that these are only guidelines and that internet speed, service and performance vary by connection type, provider and address.
For more information, refer to our guide on how much internet speed you really need.
0 to 5Mbps allows you to tackle the basics: browsing the internet, sending and receiving email and streaming low-quality video.
5 to 40Mbps gives you higher-quality video streaming and video conferencing.
40 to 100Mbps should give one person sufficient bandwidth to satisfy the demands of modern telecommuting, video streaming and online gaming.
100 to 500Mbps allows one or two people to simultaneously engage in high-bandwidth activities like video conferencing, streaming and online gaming.
500 to 1,000Mbps allows three or more people to engage in high-bandwidth activities at the same time.
How CNET chose the best internet providers in Aiken, South Carolina
Internet service providers are numerous and regional. Unlike the latest smartphone, laptop, router or kitchen tool, it’s impractical to personally test every ISP in a given city. What’s our approach? We start by researching the pricing, availability and speed information, drawing on our own historical ISP data, the provider sites and mapping information from the Federal Communications Commission at FCC.gov.
It doesn’t end there: We go to the FCC’s website to check our data and ensure we consider every ISP that provides service in an area. We also input local addresses on provider websites to find specific options for residents. We look at sources, including the American Customer Satisfaction Index and J.D. Power, to evaluate how happy customers are with an ISP’s service. ISP plans and prices are subject to frequent changes; all information provided is accurate as of publication.
Once we have this localized information, we ask three main questions:
Does the provider offer access to reasonably fast internet speeds?
Do customers get decent value for what they’re paying?
Are customers happy with their service?
The answer to those questions is often layered and complex, but the providers who come closest to “yes” on all three are the ones we recommend. When selecting the cheapest internet service, we look for the plans with the lowest monthly fee, although we also factor in things like price increases, equipment fees and contracts. Choosing the fastest internet service is relatively straightforward. We look at advertised upload and download speeds and consider real-world speed data from sources like Ookla and FCC reports. (Ookla is owned by the same parent company as CNET, Ziff Davis.)
To explore our process in more depth, visit our how we test ISPs page.
Internet providers in Aiken, South Carolina, FAQs
What is the best internet service provider in Aiken?
AT&T Fiber is the best internet service provider in Aiken. Although not the cheapest internet provider in Aiken, AT&T Fiber presents the best overall value by offering fast, symmetrical speeds and service free of contracts, data caps, equipment fees and set price increases.
Is fiber internet available in Aiken?
According to the most recent FCC data, approximately 28% of Aiken residential addresses were serviceable for fiber internet as of July 2024. Serviceability is greatest in the Pinecrest and Aiken Estates communities, although fiber internet is available in random neighborhoods throughout the city. AT&T Fiber is Aiken’s largest fiber internet provider, while Breezeline also has a small fiber presence in the area.
What is the cheapest internet provider in Aiken?
Breezeline has the lowest starting price for internet in Aiken: $205a month for 100Mbps. Xfinity offers a bit more speed, with maximum downloads of 150Mbps at $30 monthly. Despite the speed difference, Breezeline’s cheapest plan is the better deal, as it comes with free equipment rental for one year and unlimited data. Xfinity, on the other hand, has a $15 equipment rental fee (optional) and a 1.2TB monthly data cap.
Which internet provider in Aiken offers the fastest plan?
AT&T Fiber offers the fastest internet in Aiken with maximum upload and download speeds of 5,000Mbps. AT&T Fiber and Xfinity offer a 2,000Mbps plan in Aiken, although Xfinity’s maximum upload speeds are significantly slower at 200Mbps.
Many Februarys ago, I went for lunch with a friend and was served a devastating blow.
As I ordered a club sandwich, I caught her eyeing me nervously. “So, I have some news,” she said in a hesitant tone. There was an agonising suspenseful pause.
“Oh god,” I said. My heart was beating so fast I thought it might leap right out of my chest cavity. She put her head in her hands as she geared up to break the news to me. Just tell me, for the love of god, I thought to myself.
“It’s about Glen,” she said.
“OK,” I said, panicking. This was the man I had been seeing on and off for the past nine months. A friend-with-benefits that I’d accidentally fallen in love with. Not that I would’ve admitted that at the time.
SEE ALSO:
The Relationship Experience dating trend: like a situationship but so much worse
“He has a girlfriend?” I guessed in a bid to put myself out of my own misery.
“That’s not all, though,” my friend said.
There was more. “They’re having a baby,” she said. Suddenly the restaurant felt unbearably noisy. A wave of heat passed over my body and a weird rash appeared across my chest, my sandwich arrived. I stared at it and tried my best not to puke.
It was three days before Valentine’s Day and I wanted nothing more than for time to stand still. “Valentine’s Day can absolutely get fucked,” I said the next day to my friend Michelle as I ran on the treadmill and cried simultaneously. A feat of human nature, you might say.
I wanted to round up every silk rose, every cheesy card, every tacky giant teddy bear and throw them on a giant bonfire. Everywhere I turned felt like a constant onslaught of love, romance, and relationships. Each one of them a reminder of the sting of rejection I was feeling. I wanted to pretend that Valentine’s Day didn’t exist, but I didn’t know how.
SEE ALSO:
The best dating apps to help find a relationship, a hookup, or something in between
There’s a scene in Bridget Jones’ Diary where she finds out that Daniel Cleaver has a secret fiancée the whole time they’ve been together, and she has to go to work and act like nothing’s happened. She stares blankly at her computer screen, tears welling in her eyes. At my job in a theatre press office, I spent a week stealthily wiping away tears as they streamed from my eyes.
On Valentine’s Day, a work friend left a card on my desk. It was a kind gesture that made me feel loved. But later, on Instagram, I saw the Valentine’s Day card that Glen had given to his girlfriend. There was an illustration of a green olive alongside the line “Olive You.” Ugh, they’re at the pun stage of their romance. It was the longest Feb. 14 I have ever lived through.
“Being in a relationship is not a mark of success. Many people in relationships are unhappy. Be proud of being single.”
Over the next few years, that cursed day came and went. On some years, I was nursing a broken heart from yet another terrible dating experience. Dear friends continued with their sweet gestures (my best friend left a red rose on my desk one year). Truth is, Valentine’s Day is just another day if you’re not in a state of loved-up bliss. So, why not treat it as exactly what is is: just another day in February.
I refuse to spend another Valentine’s Day crying over a terrible olive pun. If you’re nursing the fragments of a broken heart and are listening to The 1975’s “Somebody Else” on repeat, then don’t fret about Feb. 14’s impending arrival. Here are some tried and tested techniques for putting two fingers up to V Day.
Mashable After Dark
Celebrate other types of love
Psychologist Dr. Tony Ortega suggests reframing the day “from being a day of romantic love to a day of celebrating any kind of love, like we did when we were children.”
SEE ALSO:
Why I’m celebrating female friendship this Valentine’s Day
“If you want to celebrate love but don’t have romantic love, grab your friends and do something out of the ordinary to celebrate your friendships,” say Ortega. “This could take the form of a scavenger hunt or perhaps an establishment that serves both liquor but has some entertainment like games or even a drag show.” If you don’t fancy going out, you could FaceTime a friend who lives far away and have a long-overdue catch up.
“Being in a relationship is not a mark of success. Many people in relationships are unhappy. Be proud of being single.”
Look upon your heartbreak as a gift
It’s really hard when you’re in the thick of the post-breakup emotional turmoil, but try to take a step back and see the bigger picture. Whether the relationship was short-term, long-term, a situationship, FWB — there’s a reason why it ended. Reframe romantic rejection as a positive thing that steers you back onto the right path, and prevents you going too far down the wrong road. It’s a blessing to have your route corrected. It’s better to be alone than in a bad relationship with a person who’s not matching your energy or feeling the same way. Your future self will look back and be grateful.
SEE ALSO:
How to separate romantic rejection from your self-worth
Banish all negative thoughts
Psychosexual and relationship therapist Silva Neves recommends banning the negative thoughts you have about yourself. “Being in a relationship is not a mark of success. Many people in relationships are unhappy. Be proud of being single,” says Neves.
SEE ALSO:
The best dating apps for hooking up, serious relationships, and everything else in between
“Each time you have a negative thought about yourself, breathe in and out deeply and say something nurturing about yourself instead. You will find out that you have many good qualities.” Make a list of everything you like about yourself, if you feel like it.
Do something nice for yourself
The longest relationship of your life is with yourself. So, why not spoil yourself? Buy yourself the dress you’ve been eyeing up for weeks. Make a delicious meal for yourself. Go get a manicure. Treat yourself to a new sex toy.
Ignore the hype
Valentine’s Day literally is just any other day. So why not treat it as such? As it happens to fall on a Friday this year, stick to your usual Friday routine — be that grabbing a drink in the pub with your mates, or binge-watching Grace and Frankie on your sofa with a takeaway. Avoid anything that’ll remind you what date it is — stay off social media and turn any calendars around so they’re facing the wall.
Rachael Lloyd, relationship experts at eharmony, says “it may sound obvious, but the more you engage in Valentine’s Day, the more you will be impacted.” “Skip over the love stories, avoid your local Pizza Express and keep off social media for the day to limit your exposure.”
Get really into puzzles
If it’s nigh-on impossible to ignore the fact that it’s the most cursed day in the calendar, then ramp up your distraction techniques. My best friend Elisha swears by jigsaw puzzles because, she says, “they require enough of your concentration to prevent you from getting in your head too much, but not so much that you can’t complete the task.” Buy yourself a big bag of Doritos, a bottle of rosé, and a 1,000-piece puzzle and it’ll keep you entertained and, crucially, distracted for hours on end. Alternatively, you could invite all your single friends over to your house for a games night.
things to do on valentine’s day Credit: vicky leta / mashable
Have some ‘me time’
Dating and relationship coach Sami Wunder says if you’re single, Valentine’s Day is the perfect day for celebrating yourself. “Firstly, don’t let all the hype get to you. Decide that it will be a day for self-love,” says Wunder. “So schedule in some me time. Book yourself a massage, or a manicure, or treat yourself to a nice coffee, or even put on your favourite dress and take yourself out for dinner.”
Neves says if Valentine’s Day is unpleasant for you, “give yourself the permission to be indulgent with whatever floats your boat.” “A long bath maybe? Or have your favourite chocolates. Or put your favourite movies on and a glass of wine?” says Neves.
Keep busy
If you’ve got the day off, one thing that’s guaranteed to keep you very busy is running errands. Go to the gym, clean out your fridge, buy your groceries, do your laundry, iron those clothes at the bottom of your ironing pile. In short, stay busy. At the end of the day, you’ll have a big sense of accomplishment.
Have sex
People in long-term relationships aren’t the only people who get to have sex on Valentine’s Day. Dr. Ortega suggests calling up your fuck buddy and scheduling a hookup.
“Do you have a friend with benefits you can call on? Instead of celebrating romantic love, celebrate your sexual nature with your FWB,” says Ortega. “Throw away the notion of romantic love and for a period of time, celebrate sexual love.”
If you don’t have a FWB, then Neves recommends having an orgasm anyway through solo sex (aka masturbation). “And make a commitment to have those on a regular basis, not just on Valentine’s Day,” says Neves.
Leave the country
Not forever. Just take a break and get some space from everything. If you can afford to, treat yourself to a weekend away somewhere nice, or take a road trip to visit a friend you haven’t seen for ages. About a fortnight after this whole debacle occurred, I decided to book an EasyJet flight to Germany to visit a really lovely friend of mine. It did me a world of good to get away from everything and gain some perspective on the situation. This year, I’m heading back to Germany to see the very same friend, and we’ll be celebrating female friendship rather than romantic love.
Whatever you do on Feb. 14, remember that lots of people in long-term relationships couldn’t give a shit about it. It’s just a stupid day.
This article was first published in 2019 and republished in 2025.
Slowly but surely, Apple TV+ has found its feet. The streaming service, which at launch we called “odd, angsty, and horny as hell,” has evolved into a diverse library of dramas, documentaries, and comedies. Now, its library is so packed, we’ve declared it “the new HBO.”
Curious but don’t know where to get started? Below are our picks for the best shows on the service. (Also, here are our picks for the best movies on Apple TV+.) When you’re done, head over to our guides to the best shows on Netflix, best movies on Hulu, and best movies on Amazon Prime, because you can never have too much television.
If you buy something using links in our stories, we may earn a commission. This helps support our journalism.Learn more.
Mythic Quest
Considering it’s a TV show set in the world of video games, you’d be forgiven for thinking this series—now four seasons deep—would be a clunker. It’s not. Instead, Mythic Quest is one of the best workplace comedies of the past few years. Presented in perfectly bingeable half-hour episodes, the show follows a fictional game studio known for its World of Warcraft–like MMO, Mythic Quest, as the people who make it slalom through their many quirky relationships. The writing is excellent, consistently funny and emotionally impactful when you least expect it, and the show manages to confront real issues in the industry without sacrificing laughs. If you really love it, there’s more in store: Apple TV+ plans to roll out a spinoff anthology series, called Side Quest, March 26.
Severance
Has there ever been a second season more highly anticipated than that of Severance? OK, there probably has—Game of Thrones comes to mind—but considering Severance is the show that firmly established Apple TV+ as a streaming player with its own distinct kind of edgy prestige content, the hype around the show’s second coming was intense. For those who don’t already know, a primer: Adam Scott plays Mark, a man distraught by the death of his wife who opts to undergo Severance, a procedure that divides his memories of work from those of his life at home. He’s quite happy with the setup until a former Lumon Industries coworker tracks him down when he’s out-of-office, setting off a series of events that makes him question not only Severance but the work his company does. From there, it only gets more weird and bleak with each passing minute. Tense and heartbreaking, Severance will keep you guessing and questioning the whole way through.
Silo
As WIRED noted in the wake of Silo’s release, this show is prestige sci-fi gold. Nearly two years later, that’s still true—and it is poised to get even better. Based on a dystopian book trilogy by Hugh Howey, the series focuses on a subterranean bunker—the silo of the title—where humanity has sequestered itself after the apocalypse. Some are hoping to win the chance to reproduce, some are trying to solve mysterious murders. Everyone watching is enjoying figuring out what’s going on in this underground city and what’s happening outside of it. Silo has already been renewed through season four. If you haven’t been watching, start.
The Secret Lives of Animals
Seventy-seven species. Twenty-four countries. This 10-part docuseries is all about the million-and-one ways animals are incredible problem solvers. A production of BBC Studios Natural History Unit, it does all the things good nature docs do: going underground and getting (perhaps creepily) close to some of the world’s most compelling creatures. Spiders, wood mice, frogs, and octopi—its got it all.
Disclaimer
When filmmaking legend Alfonso Cuarón (Gravity) decides to do a limited series starring Cate Blanchett, you kind of owe it to yourself to watch. Especially when, as Cuarón told WIRED, the Renée Knight novel the series is based on was so intriguing it made him want to bring his cinematic skills to TV. In the seven-part series, Blanchett plays an esteemed journalist named Catherine who is sent a mysterious novel that threatens to expose parts of her past she’d hidden for years. As she tries to investigate who wrote the book, she also must keep her own life from collapsing around her. Cuarón adapted the novel himself and directed each episode of the series, bringing his big-screen style to the small-screen world.
Shrinking
Do you enjoy In Treatment but wish it was, you know, fun? Then Shrinking may be right for you. Created by Bill Lawrence and Brett Goldstein—of Ted Lasso fame—and Jason Segel, the show follows Jimmy (Segel), a therapist struggling to get over the death of his wife and reconnect with his daughter and patients. That may sound like a downer, but it’s buoyed by the fact that it’s also a workplace comedy focusing on the therapy practice where Jimmy works alongside Harrison Ford’s Paul and Jessica Williams’ Gaby. Shrinking, ultimately, is about the things people do to cope, but it also features a dream team of a cast and one very memorable party scene featuring a vomit-soaked piano and a super-stoned Ford.
Slow Horses
As we wrote not too long ago, Slow Horses is the ideal show for people who want a Pizza Hut-Taco Bell-esque combination of John Le Carré–style espionage thrillers and The Office. Based around the misfits of Slough House, where MI5 agents are sent when they biff it as spies, the show effortlessly jumps from shoot-outs and car chases to quirky conversations and camaraderie. The show’s fourth season, which launched last year, is a little more subdued than the ones before, but if you’ve been sleeping on Slow Horses, now is the time to wake up.
Bad Monkey
Created by Bill Lawrence, one of the creative forces behind two other Apple TV+ zingers, Ted Lasso and Shrinking, Bad Monkey is about a one-time detective (played by Vince Vaughn) who’s hit a bit of a rough patch and is trying to get to the bottom of why someone found a severed arm. Yes, there’s a monkey, but there’s also a lot of dark humor and heart—and a look at the complex lives of more than a few Florida Men.
Sunny
Sunny is the story of a woman named Suzie (Rashida Jones) whose husband and son are lost in a mysterious plane crash. To work through her grief, Suzie is given Sunny, a domestic robot with whom she forms a unique bond as she begins to uncover what happened to her family. As artificial intelligence gets more and more ingrained in everyone’s lives, Sunny promises to hit differently now than it would at any other time.
Presumed Innocent
Just to be clear, this whodunit has been done before. Thirty-four years ago, Harrison Ford starred in the film adaptation of Scott Turow’s novel. This time around, the lead is played by Jake Gyllenhaal, and the adaptation is an eight-part limited series, not a film. Gyllenhaal stars as Rusty Sabich, a Chicago prosecutor accused of killing a colleague. A colleague with whom he was having an affair. Presumed Innocent is produced by David E. Kelley, so it has the intrigue and glossiness of his recent offerings like Big Little Lies and The Undoing, as well as the darkness and drama.
STEVE! (martin) A Documentary in 2 Pieces
Putting this on the “best shows on Apple TV+” list is a bit of a cheat. Rather than a series, this two-part documentary is more like a pair of movies looking at the life and career of Steve Martin. The first part chronicles his rise in, and reimagining of, the standup comedy world. The second looks at how he went from that to the neurotic and lovable neighbor he currently plays on Only Murders in the Building, which would be his career’s triumphant second act if he hadn’t had something like 30 acts in between. Directed by Morgan Neville, who made the backup singer documentary 20 Feet From Stardom and the Fred Rogers doc Won’t You Be My Neighbor?, it’s funny, intimate, and a little surprising—just like Martin himself.
Loot
OK, so Loot isn’t exactly about MacKenzie Scott’s divorce from Jeff Bezos, but it is about a woman named Molly (Maya Rudolph) who separates from her tech billionaire husband and devotes herself to philanthropic work. Also, creators Alan Yang (Master of None) and Matt Hubbard (Superstore) were kind of inspired by Bezos and Scott’s split. With an incredible supporting cast that includes Joel Kim Booster, Michaela Jaé Rodriguez, and Adam Scott, it’s a quirky comedy with a lot of heart—and the kind of thing you (probably) won’t see on Amazon Prime Video.
The Big Door Prize
With The Big Door Prize Chris O’Dowd finally got the “guy leading a show” role he was always meant for. In the series, he plays a 40-year-old high school teacher named Dusty who’s pretty content with his life until a magic machine shows up in his small town. The machine, you see, tells people their life’s potential, and as soon as folks around him start using it, everything changes. Marriages end, paths divert, and eventually Dusty must confront whether he’s happy in his own life.
The Completely Made-Up Adventures of Dick Turpin
Dick Turpin was a real highwayman in 18th-century England who was ultimately executed for horse theft. But the myths surrounding him are far more interesting than the facts. The Completely Made-Up Adventures of Dick Turpin, true to its name, opts to stick with the fun stuff. Starring Noel Fielding (The Great British Bake Off) in the title role, this six-episode series presents Turpin as someone who stumbled into leading a group of outlaws and made the best of it. Enjoy the ride.
Constellation
Around here we have a theory that Apple TV+ is the new HBO. At the same time, we also wonder among ourselves whether it’s the new Syfy. After opening with a bang in 2019 with For All Mankind, it has released a steady drumbeat of trippy, spacey, timey-wimey prestige shows, from Foundation to Severance. In early 2024, it released Constellation, an eight-part thriller about an astronaut (Noomi Rapace) who returns to Earth after a disaster in space to find things are very off. Brain-bending and tense, it’s the kind of sci-fi that sucks you in.
Masters of the Air
Generally speaking, “World War II drama” and Steven Spielberg are probably enough to get anyone to click Play on this series, but it’s got a lot more than just a good elevator pitch. Based on Donald L. Miller’s Masters of the Air, this series dives deep into the lives of the 100th Bomb Group—aka the “Bloody Hundredth”—a squad of pilots tasked with risking their lives to fight Nazi Germany from the air. Spielberg and Tom Hanks serve as executive producers, and the cast features Elvis himself, Austin Butler, as well as Saltburn’s Barry Keoghan and Doctor Who’s latest Doctor, Ncuti Gatwa.
The New Look
Keeping with the World War II theme, The New Look follows Christian Dior, Coco Chanel, Pierre Balmain, and Cristóbal Balenciaga as they lay the path for modern fashion in Nazi-occupied Paris. The cast features Ben Mendelsohn as Christian Dior, Juliette Binoche as Coco Chanel, and Maisie Williams as Catherine Dior, and also has a soundtrack courtesy of Jack Antonoff that’s chock full of early 20th-century music covered by the likes of Perfume Genius and Florence Welch.
Criminal Record
Starring former Doctor Who Doctor Peter Capaldi, Criminal Record follows two cops—Capaldi’s Daniel Hegarty and Cush Jumbo’s June Lenker—as they try to get to the bottom of a long-settled case. Daniel worked the case originally and got a confession; June got a fresh tip and wants him to reopen it and find out whether the man who went away for murder is actually innocent. Might sound a bit overdone, but the series also works in elements of law enforcement shortcomings and race in a rapidly-changing Britain for a series that’s about more than just one case.
Hijack
There’s this face Idris Elba does. He’s been perfecting it since he was Stringer Bell on The Wire. It’s the look of total calm even when he’s talking about the most harrowing thing you can imagine. That face gets a full workout in Hijack, in which Elba plays a corporate negotiator who finds himself trying to settle things with a group of, yes, hijackers who have taken over the flight he’s boarded to get home to his family. This series is seven episodes, roughly seven hours—the same length of the flight, and it follows the drama in the air and the political maneuvering below before attempting to stick the landing. Do stay around until the end.
For All Mankind
Long before Foundation, there was For All Mankind. A solid slice of alternate history, the show starts with a very smart premise: What if the US had been edged out in putting a man on the moon? How would the space-race rivalry between the Americans and the Soviets have played out? It’s mostly a slick, stylish, NASA-heavy period drama, but as this is from the brain of Ronald D. Moore, there are a few standout moments and episodes with attention shared around the large ensemble cast. It might be the best sci-fi show you’re not watching, and if that’s true you now have multiple seasons to catch up on.
Messi Meets America
If your home screen hasn’t made it obvious, Apple TV+ is super stoked about soccer. Messi Meets America is a six-part docuseries about all-star player Lionel Messi’s move to Major League Soccer’s Inter Miami club. Messi Mania, indeed.
Lessons in Chemistry
Based on the debut novel from science writer Bonnie Garmus, Lessons in Chemistry is the story of Elizabeth Zott (Brie Larson), who gets hired to host a cooking show after she’s fired from her lab for doing science while female. Obviously, the show Elizabeth puts on ends up being about a lot more than just having dinner on the table at 6 pm, but we suggest you watch to find out just how revolutionary it is.
The Morning Show
Every streaming service needs a flashy mainstream drama with Hollywood heavyweights to pull in viewers. Apple TV+ has The Morning Show. When Alex Levy (Jennifer Aniston) loses her morning news program cohost Mitch Kessler (Steve Carell) following sexual misconduct accusations, she gets paired up with Bradley Jackson (Reese Witherspoon). What unfolds is a #MeToo-era drama full of TV network intrigue and Sorkin-lite dialog. In its second season, it went deep on Covid-19, and in the third season the series’ fictional network, UBA, finds itself dealing with the aftermath of a cyberattack. There’s likely a new season coming in 2025, so now’s a good time to catch up, or go back and refresh your memory.
Shining Girls
This Elisabeth Moss psychological thriller/murder mystery came out in 2022 and never really got the buzz it likely deserved. Moss plays Kirby, a woman who believes a recent Chicago murder may be linked to an attack on her many years prior. She teams with a Sun-Times reporter to investigate, but the deeper she digs the more her own reality starts to shift. Based on the book The Shining Girls by Lauren Beukes, this series may seem like just another murder mystery, but its sci-fi twists put it one step ahead.
Foundation
WIRED called Foundation a “flawed masterpiece” in our review of the first season. Considering the complexities of adapting a sprawling Isaac Asimov book series, it was high praise. Starring Jared Harris as Hari Seldon, a math professor who, along with his loyal followers, is exiled for predicting the oncoming end of the galactic empire that rules over them, the show often suffers under the weight of its own massive scope. But it also features wonderful performances from Lee Pace and beautiful images inspired by the James Webb Space Telescope. If you have a soft spot for big sci-fi dramas, this Game-of-Thrones-in-space wannabe is a must-watch.
The Crowded Room
Set in the late 1970s, The Crowded Room stars Tom Holland as Danny Sullivan, a young man arrested after a grisly shooting in New York City. Following his arrest, this 10-episode limited series unfolds into a twisty whodunit as interrogator Rya Goodwin (Amanda Seyfried) tries to suss out what happened with the shooting and the peculiar events in Sullivan’s past that may have shaped how he ended up involved. Holland told Extra that the shoot for The Crowded Room, which he also produced, “broke” him, leading to him taking a yearlong hiatus from acting. Want to see why? Watch now.
Ted Lasso
On paper, Ted Lasso sounds terrible. It’s the inconceivable story of an American football coach who has never watched a game of soccer somehow landing himself a job as coach of a (fictional) Premier League club and trying to make up for his total lack of qualifications by being a nice guy. Sounds unwatchable, doesn’t it? And yet Ted Lasso has captured the hearts and minds of viewers on both sides of the pond with its large-as-life cast and irresistibly wholesome messaging, hoovering up awards in the process.
High Desert
The Patricia Arquette–aissance doesn’t get as much ink as Matthew McConaughey or Keanu Reeves did during their second comings, but it’s here—in part thanks to the rise of streaming. Between The Act and Severance, Arquette has received some of the highest accolades of her long career recently, and High Desert is no exception. While coming to terms with the death of her mother, Peggy (Arquette)—an addict—decides she wants to pick up the pieces of her life and become a private investigator. She finds an unwitting employer/sometime mentor in Bruce Harvey (Brad Garrett), but not everyone is onboard with Peggy’s career decisions—namely, her straitlaced sister (Christine Taylor). It’s an odd duck of a show, which is perfectly suited to Arquette’s ethereal acting style, allowing her to seamlessly flit between moments of tragedy and laugh-out-loud comedy, with the audiences doing their best to keep up. The all-star cast is made even more impressive by recurring appearances from Bernadette Peters as Peggy’s late mom.
Big Beasts
Look, Discovery doesn’t get to corner the market on animal documentaries—and this 10-part docuseries proves it. Featuring elephant seals, brown bears, orangutans, giant otters, and all kinds of massive mammals in between, it’s the perfect thing if you just want to escape and learn a few tidbits about nature. But the best part? It’s narrated by Tom Hiddleston, and there’s just something charming about hearing the voice of Loki talk about a bunch of different animals he could turn himself into in the blink of an eye.
Servant
Cinematically, M. Night Shyamalan can be a little hit-or-miss, but Servant, which the filmmaker executive produces and occasionally directs, is stellar. It’s about a Philadelphia couple—a chef and a news anchor—who lose a child only to have it mysteriously come back to life (maybe) with the arrival of their new nanny. (You really just need to watch the show for any of this to make sense.) Moody, freaky, and occasionally even funny, it’ll suck you in. With four seasons on the streamer, there’s plenty to enjoy.
The Essex Serpent
Claire Danes doing her best trembling-chin acting in period garb, Tom Hiddleston as a town vicar, rumors about a mysterious mythological serpent—is there anything not to love about this show? No, there’s not. The Essex Serpent, based on the novel by Sarah Perry, follows a recent widow (Danes) as she heads to the countryside in Essex to investigate a “sea dragon.” There, she meets a vicar, Will (Hiddleston), who is far more skeptical of the serpent’s existence. Lush and inviting, it’s the ideal period mystery.
Dickinson
Hailee Steinfeld is a riotous young Emily Dickinson in this half-hour show from creator Alena Smith. It was part of the original Apple TV+ lineup and quickly distinguished itself thanks to its off-kilter vision of 19th-century Amherst, Massachusetts. The first season is a set of sharp, surreal vignettes, inspired by Dickinson’s work and tracing the imagined life of the young poet, who is rebelling against her father, the town’s societal rules, and just about everything else. The second and third seasons go deeper, examining not only the poet’s life, but also the roles that race, gender, sexuality, and class played in the early days of America. If you’re a Dickinson stan, love a bit of smart queer dramedy, or just have a penchant for a modern soundtrack in a Civil War–era show, you’ll dig this.