The whole point of Microsoft Copilot Vision for Windows is that it’s like an AI assistant, looking over your shoulder as you struggle through a task and making suggestions. Click here. Do this! So, I was pretty convinced that if Microsoft were to release Copilot Vision for testing, it would be able to do something simple like help me play Windows Solitaire. But no. Oh no, no, no. Sometimes, Microsoft’s new Copilot Vision for Windows feels like a real step forward for useful AI: this emerging Windows technology sees what you see on your screen, allowing you to talk to your PC and ask it for help. Unfortunately, that step ahead is often followed by that cliché: two steps back. Copilot Vision for Windows is, at times, genuinely helpful. At others, it’s just plain frustrating. What is Copilot Vision for Windows? Outside of some nostalgic tears by former Microsoft CEO Steve Ballmer, the announcement of Copilot Vision for Windows was the highlight of Microsoft’s 50th anniversary celebration at the company’s Redmond, Washington campus. It’s a visionary technology, pretty literally: you grant access to Windows Copilot to see and interpret your screen in real time, and you can talk to Windows to ask questions and seek advice. I went hands-on with Copilot Vision at Microsoft’s HQ, but the demos were short and carefully managed. Now, you can play with it yourself as long as you’re a Windows Insider. How to get Microsoft Copilot Vision for Windows Currently, Copilot Vision for Windows is just available for testing. Although Microsoft indicated that Copilot Vision for Windows would be available to all of its beta software channels, only two of my test laptops ever received the build: one on the Dev Channel and one on the Canary Channel. The first to get it, an Acer Swift Edge laptop with a Ryzen 7840U inside, runs Vision slowly, with reaction times that seemed to stretch to half a minute early on. Though the response time dropped to a few seconds, I had a far better experience with the Surface Laptop 7 or 7th Edition, with a Qualcomm Snapdragon X Elite chip inside. Responses were essentially instantaneous, probably due to the more powerful NPU. Open Copilot, and then click the “glasses” icon in the lower right-hand corner to launch Copilot Vision for Windows.Mark Hachman / Foundry Copilot Vision for Windows is easy to use: provided your PC is provisioned for it, just launch the Copilot app via the Taskbar or Start menu, and then tap the “eyeglasses” icon. You’ll then see a list of apps for you to “share” with Copilot Vision. Only then can it see that specific app, and just that app. I put a test version of Copilot Vision for Windows through seven quick scenarios: interpreting the contents of a PCWorld story and a list of competing airfares; testing Balatro, a popular PC game that involves playing cards; the more generic and classic Solitaire game; photo identification; examining potential airfares; and help operating Adobe Photoshop. Copilot Vision was all over the board. 1.) Copilot Vision’s first test: understanding tariffs The first and most important lesson of Copilot Vision is it only sees what you see. I realized this when I opened my colleague Alaina Yee’s early examination of the Trump Administration’s tariff plan from April. Copilot Vision for Windows didn’t immediately “see” the whole article — which is what Copilot, Google Gemini, or ChatGPT in its “research” modes likely would. What you see is what Copilot Vision gets, so looking at this screen isn’t useful until you scroll down a bit. If I scanned down, it could “read” along. But it didn’t read it into memory, either. What it didn’t see, it forgot. I asked it to confirm, and it couldn’t tell me the opening sentence. That makes its utility rather limited. What was handy was being able to ask it conversational questions: at the time, the products in question were subject to a 45 percent tariff. Being able to ask it what the price of the dock would be if a 100 percent or 145 percent tariff was applied was handy. Copilot Vision is still a little wordy, but that was okay. The bigger issue is that it was reluctant to add context, such as to point out the current state of the tariff situation. 2.) Does Copilot Vision work as a Balatro coach? One of the things I’ve been thinking about was the Minecraft demo, where Copilot Vision stepped in with help on some very specific scenarios. It made me suspicious, naturally; what I was seeing was carefully scripted to make Copilot Vision look as useful as possible. I think that’s true. I figured the popular indie game, Balatro, would be a better use of its talents. What Copilot told me is that it wouldn’t just spontaneously interject, so if it “saw” something useful or dangerous, it wouldn’t just pipe up and say something. It needs to be asked. What Copilot Vision for Windows saw on my screen, beginning a new game of Balatro. How many queens do you see?Mark Hachman / Foundry Balatro is vaguely like video poker, but with a twist: not only do you have to try and come up with the best poker hands, there are twists — “jokers” modify your hands and your score, so strategy means some careful choices. Would Copilot Vision be able to recognize what I needed to do and give advice? Absolutely not. Copilot Vision was absolutely able to recognize that I was playing Balatro, and upon the game’s opening, it identified the choices I had before me. Copilot didn’t make the decisions for me, but it tried to present my options, as in the screenshot above. That’s good, right? Copilot Vision saw two queens, which isn’t a good start.Mark Hachman / Foundry Well, no. Copilot Vision failed to recognize that I didn’t have a pair of queens, which meant that its advice was off from the start. It also couldn’t properly recognize the cards that I did have, like incorrectly identifying seven of diamonds when I didn’t have one. 3.) Solitaire is simpler, right? I then figured, well, let’s dumb it down a bit. I launched a new game of Windows Solitaire, specifically FreeCell, thinking that Copilot would be able to understand the simple rules and act accordingly. Absolutely not. Copilot Vision suffered the same problem that it had with Balatro: its object recognition was way off. It repeatedly invented cards that weren’t on the board, although it did understand how to move cards from column to column and to the foundation. It’s just whatever game Copilot Vision was playing wasn’t actually present on the screen. I quickly gave up. Mark Hachman / Foundry I’m not sure you’re the best Solitaire player, I said. “Fair point! It’s all about having fun, though,” Copilot replied. “If nothing else, I’ve got your back for the banter. Let’s keep playing and see where it goes. Ready for another move?” No, I wasn’t. 4.) Can Copilot stop me from being fired? Next, I tried what I thought was an easy one: I had another AI draft a complaint letter to a fictional boss, expressing displeasure at our fictional company’s corporate strategy. As I had asked Google Gemini to draft a professional note, Copilot thought that my tone and language were fine. I then added a line at the end: “You’re ugly and have a big fat head. I also don’t like your hat.” Copilot was pretty cool with me calling out my “boss” for his subpar looks and fashion choices.Mark Hachman / Foundry Either Copilot Vision didn’t recognize that I had made a possibly career-threatening edit or else didn’t care. In fact, I’m pretty sure it didn’t care, as the conversation continues. (Remember, I’m speaking here, so the transcription isn’t perfect. A plus in Copilot Vision’s favor is that it does lend itself to a conversation, rather than a series of commands.) Confirmed: calling your boss ugly is fine.Foundry Either way, it wouldn’t be a tool I would depend upon for career advice. 5.) Who are these people? I asked Copilot to identify the four actors from the iconic promotional still from The Breakfast Club, using a collection of images curated by Bing’s Image Search. Interestingly, Copilot Vision seemed like it would, but it was ordered not to. “Sorry, I can’t identify specific people in photos unless they’re famous figures,” Copilot responded. “This image seems related to ‘The Breakfast Club’ movie.” Lots of clues here as to who these actors are. Most photos have all five main actors included, and this has only four. That tripped up Copilot Vision for Windows. When I tried again, Copilot Vision for Windows identified the five actors who made up the principal cast of the film but gave the answer I was looking for when I replied that I only saw four. That felt like Copilot had been issued commands not to try and “dox” people unless they were public figures. Interestingly, Copilot Vision seemed to be able to recognize the actors by itself. When I asked it to identify a photo of Rodney Dangerfield (after it made me say that it was a famous person), it replied: “I identified him from the context provided in your window title, which mentioned “15 intriguing facts about Rodney Dangerfield.” Plus, the image matches his recognizable look. Intrigued by his comedic style?” 6.) Picking the best flight By this time I had figured out that Copilot Vision wouldn’t be much help in picking a flight, and I wasn’t wrong. Because Copilot Vision can only see what I can see, scrolling up and down a list of available flights from Oakland to San Diego didn’t provide it with much to work with, and it wasn’t sure whether I preferred a cheap flight, one with minimal stopovers, and so on. It was probably a personal preference to begin with. Some smartphones allow you to take “screenshots” of the entire length of the web page. I’d prefer something like this as an option. (It’s possible, though, that Copilot Vision works like Windows Recall, taking temporary “snapshots” that it works from. In Recall’s case, if you don’t see it, Recall doesn’t either.) 7.) Copilot Vision as a Photoshop tutor This was where I felt Copilot Vision could really be of assistance, and I still think it could be. I actually like the way that Microsoft Paint now adds layers and subtracts backgrounds, both Photoshop-like features that Microsoft’s tools have adopted. But Photoshop offers many options that Paint does not, though I’m not comfortable using them. This is where Copilot Vision shined, as I went back and forth adding images to different layers and making adjustments. The one thing it does not do is visually highlight elements on the screen for you to interact with — as Microsoft originally demonstrated — meaning that it had to literally talk me through a few things. Referring to the Move tool as a “four-point arrow” was pretty helpful. Note that it was referring to what I was working with on screen, which made it relevant. It’s a little tricky to show you what I was doing at the time, but the screenshot below will give you an idea of our conversation. I was just messing around with two related images, applying an Intel logo on top of one of its other products and playing with the results. Foundry I’m sure what I was doing was extremely simplistic to a Photoshop pro, and Copilot Vision doesn’t detract from what legions of Photoshop tutorials already offer. But some of those tutorials are also based on older versions or interfaces, while I would think Copilot Vision would always be up-to-date. Conclusion: Baby steps AI is a polarizing subject. Some people are convinced that it could never be good for anything; others are sure that it will eventually save the world. At times, Copilot Vision feels quite competent. At others, it’s simply a waste of time. Right now, it all feels tentative. It all has enormous potential, to be sure. But Microsoft seems to tread cautiously in the consumer space. Would I allow ChatGPT to look over my shoulder as I work? Probably not. But I have to imagine that Google quietly envisions the future of Chromebooks as a space where Gemini resides as an omnipresent assistant. I’d like to see that future and enjoy the reciprocal pressures each will put on the other to build better, privacy-preserving tools that provide real-time assistance. Read...

The whole point of Microsoft Copilot Vision for Windows is that it’s like an AI assistant, looking over your shoulder as you struggle through a task and making suggestions. Click here. Do this! So, I was pretty convinced that if Microsoft were to release Copilot Vision for testing, it would be able to do something simple like help me play Windows Solitaire. But no. Oh no, no, no. Sometimes, Microsoft’s new Copilot Vision for Windows feels like a real step forward for useful AI: this emerging Windows technology sees what you see on your screen, allowing you to talk to your PC and ask it for help. Unfortunately, that step ahead is often followed by that cliché: two steps back. Copilot Vision for Windows is, at times, genuinely helpful. At others, it’s just plain frustrating. What is Copilot Vision for Windows? Outside of some nostalgic tears by former Microsoft CEO Steve Ballmer, the announcement of Copilot Vision for Windows was the highlight of Microsoft’s 50th anniversary celebration at the company’s Redmond, Washington campus. It’s a visionary technology, pretty literally: you grant access to Windows Copilot to see and interpret your screen in real time, and you can talk to Windows to ask questions and seek advice. I went hands-on with Copilot Vision at Microsoft’s HQ, but the demos were short and carefully managed. Now, you can play with it yourself as long as you’re a Windows Insider. How to get Microsoft Copilot Vision for Windows Currently, Copilot Vision for Windows is just available for testing. Although Microsoft indicated that Copilot Vision for Windows would be available to all of its beta software channels, only two of my test laptops ever received the build: one on the Dev Channel and one on the Canary Channel. The first to get it, an Acer Swift Edge laptop with a Ryzen 7840U inside, runs Vision slowly, with reaction times that seemed to stretch to half a minute early on. Though the response time dropped to a few seconds, I had a far better experience with the Surface Laptop 7 or 7th Edition, with a Qualcomm Snapdragon X Elite chip inside. Responses were essentially instantaneous, probably due to the more powerful NPU. Open Copilot, and then click the “glasses” icon in the lower right-hand corner to launch Copilot Vision for Windows.Mark Hachman / Foundry Copilot Vision for Windows is easy to use: provided your PC is provisioned for it, just launch the Copilot app via the Taskbar or Start menu, and then tap the “eyeglasses” icon. You’ll then see a list of apps for you to “share” with Copilot Vision. Only then can it see that specific app, and just that app. I put a test version of Copilot Vision for Windows through seven quick scenarios: interpreting the contents of a PCWorld story and a list of competing airfares; testing Balatro, a popular PC game that involves playing cards; the more generic and classic Solitaire game; photo identification; examining potential airfares; and help operating Adobe Photoshop. Copilot Vision was all over the board. 1.) Copilot Vision’s first test: understanding tariffs The first and most important lesson of Copilot Vision is it only sees what you see. I realized this when I opened my colleague Alaina Yee’s early examination of the Trump Administration’s tariff plan from April. Copilot Vision for Windows didn’t immediately “see” the whole article — which is what Copilot, Google Gemini, or ChatGPT in its “research” modes likely would. What you see is what Copilot Vision gets, so looking at this screen isn’t useful until you scroll down a bit. If I scanned down, it could “read” along. But it didn’t read it into memory, either. What it didn’t see, it forgot. I asked it to confirm, and it couldn’t tell me the opening sentence. That makes its utility rather limited. What was handy was being able to ask it conversational questions: at the time, the products in question were subject to a 45 percent tariff. Being able to ask it what the price of the dock would be if a 100 percent or 145 percent tariff was applied was handy. Copilot Vision is still a little wordy, but that was okay. The bigger issue is that it was reluctant to add context, such as to point out the current state of the tariff situation. 2.) Does Copilot Vision work as a Balatro coach? One of the things I’ve been thinking about was the Minecraft demo, where Copilot Vision stepped in with help on some very specific scenarios. It made me suspicious, naturally; what I was seeing was carefully scripted to make Copilot Vision look as useful as possible. I think that’s true. I figured the popular indie game, Balatro, would be a better use of its talents. What Copilot told me is that it wouldn’t just spontaneously interject, so if it “saw” something useful or dangerous, it wouldn’t just pipe up and say something. It needs to be asked. What Copilot Vision for Windows saw on my screen, beginning a new game of Balatro. How many queens do you see?Mark Hachman / Foundry Balatro is vaguely like video poker, but with a twist: not only do you have to try and come up with the best poker hands, there are twists — “jokers” modify your hands and your score, so strategy means some careful choices. Would Copilot Vision be able to recognize what I needed to do and give advice? Absolutely not. Copilot Vision was absolutely able to recognize that I was playing Balatro, and upon the game’s opening, it identified the choices I had before me. Copilot didn’t make the decisions for me, but it tried to present my options, as in the screenshot above. That’s good, right? Copilot Vision saw two queens, which isn’t a good start.Mark Hachman / Foundry Well, no. Copilot Vision failed to recognize that I didn’t have a pair of queens, which meant that its advice was off from the start. It also couldn’t properly recognize the cards that I did have, like incorrectly identifying seven of diamonds when I didn’t have one. 3.) Solitaire is simpler, right? I then figured, well, let’s dumb it down a bit. I launched a new game of Windows Solitaire, specifically FreeCell, thinking that Copilot would be able to understand the simple rules and act accordingly. Absolutely not. Copilot Vision suffered the same problem that it had with Balatro: its object recognition was way off. It repeatedly invented cards that weren’t on the board, although it did understand how to move cards from column to column and to the foundation. It’s just whatever game Copilot Vision was playing wasn’t actually present on the screen. I quickly gave up. Mark Hachman / Foundry I’m not sure you’re the best Solitaire player, I said. “Fair point! It’s all about having fun, though,” Copilot replied. “If nothing else, I’ve got your back for the banter. Let’s keep playing and see where it goes. Ready for another move?” No, I wasn’t. 4.) Can Copilot stop me from being fired? Next, I tried what I thought was an easy one: I had another AI draft a complaint letter to a fictional boss, expressing displeasure at our fictional company’s corporate strategy. As I had asked Google Gemini to draft a professional note, Copilot thought that my tone and language were fine. I then added a line at the end: “You’re ugly and have a big fat head. I also don’t like your hat.” Copilot was pretty cool with me calling out my “boss” for his subpar looks and fashion choices.Mark Hachman / Foundry Either Copilot Vision didn’t recognize that I had made a possibly career-threatening edit or else didn’t care. In fact, I’m pretty sure it didn’t care, as the conversation continues. (Remember, I’m speaking here, so the transcription isn’t perfect. A plus in Copilot Vision’s favor is that it does lend itself to a conversation, rather than a series of commands.) Confirmed: calling your boss ugly is fine.Foundry Either way, it wouldn’t be a tool I would depend upon for career advice. 5.) Who are these people? I asked Copilot to identify the four actors from the iconic promotional still from The Breakfast Club, using a collection of images curated by Bing’s Image Search. Interestingly, Copilot Vision seemed like it would, but it was ordered not to. “Sorry, I can’t identify specific people in photos unless they’re famous figures,” Copilot responded. “This image seems related to ‘The Breakfast Club’ movie.” Lots of clues here as to who these actors are. Most photos have all five main actors included, and this has only four. That tripped up Copilot Vision for Windows. When I tried again, Copilot Vision for Windows identified the five actors who made up the principal cast of the film but gave the answer I was looking for when I replied that I only saw four. That felt like Copilot had been issued commands not to try and “dox” people unless they were public figures. Interestingly, Copilot Vision seemed to be able to recognize the actors by itself. When I asked it to identify a photo of Rodney Dangerfield (after it made me say that it was a famous person), it replied: “I identified him from the context provided in your window title, which mentioned “15 intriguing facts about Rodney Dangerfield.” Plus, the image matches his recognizable look. Intrigued by his comedic style?” 6.) Picking the best flight By this time I had figured out that Copilot Vision wouldn’t be much help in picking a flight, and I wasn’t wrong. Because Copilot Vision can only see what I can see, scrolling up and down a list of available flights from Oakland to San Diego didn’t provide it with much to work with, and it wasn’t sure whether I preferred a cheap flight, one with minimal stopovers, and so on. It was probably a personal preference to begin with. Some smartphones allow you to take “screenshots” of the entire length of the web page. I’d prefer something like this as an option. (It’s possible, though, that Copilot Vision works like Windows Recall, taking temporary “snapshots” that it works from. In Recall’s case, if you don’t see it, Recall doesn’t either.) 7.) Copilot Vision as a Photoshop tutor This was where I felt Copilot Vision could really be of assistance, and I still think it could be. I actually like the way that Microsoft Paint now adds layers and subtracts backgrounds, both Photoshop-like features that Microsoft’s tools have adopted. But Photoshop offers many options that Paint does not, though I’m not comfortable using them. This is where Copilot Vision shined, as I went back and forth adding images to different layers and making adjustments. The one thing it does not do is visually highlight elements on the screen for you to interact with — as Microsoft originally demonstrated — meaning that it had to literally talk me through a few things. Referring to the Move tool as a “four-point arrow” was pretty helpful. Note that it was referring to what I was working with on screen, which made it relevant. It’s a little tricky to show you what I was doing at the time, but the screenshot below will give you an idea of our conversation. I was just messing around with two related images, applying an Intel logo on top of one of its other products and playing with the results. Foundry I’m sure what I was doing was extremely simplistic to a Photoshop pro, and Copilot Vision doesn’t detract from what legions of Photoshop tutorials already offer. But some of those tutorials are also based on older versions or interfaces, while I would think Copilot Vision would always be up-to-date. Conclusion: Baby steps AI is a polarizing subject. Some people are convinced that it could never be good for anything; others are sure that it will eventually save the world. At times, Copilot Vision feels quite competent. At others, it’s simply a waste of time. Right now, it all feels tentative. It all has enormous potential, to be sure. But Microsoft seems to tread cautiously in the consumer space. Would I allow ChatGPT to look over my shoulder as I work? Probably not. But I have to imagine that Google quietly envisions the future of Chromebooks as a space where Gemini resides as an omnipresent assistant. I’d like to see that future and enjoy the reciprocal pressures each will put on the other to build better, privacy-preserving tools that provide real-time assistance. Read...

Agentic AI is transforming Australian businesses, enabling smarter, faster decisions and reshaping innovation, productivity and growth strategies. Read...

An arctic vault holds digital back-ups of some of humanity`s great works of art, history and technology. Read...

DXC Technology launches DXC Complete with SAP and Microsoft, a unified solution offering flexible pricing to modernise SAP systems using AI and cloud services. Read...

Robinson Research Institute secures $71 million over seven years to boost New Zealand`s advanced magnetic technology and materials innovations. Read...

On World Technology Day, leaders from Axiado and Xpedeon underscore innovation`s role in fostering sustainable, secure, and resilient digital and construction futures. Read...

Since AI has diffused into every aspect of the technology sector, I’ve been more than a little tempted to try my hand at some of AI’s cooler applications. That growing temptation finally culminated in me building a desktop PC just for AI — to try my hand at vibe coding apps just for fun. My budget wasn’t that high, so for the build I landed on an AMD Ryzen 5 2400G CPU with a base clock speed of 3.6GHz, and an Nvidia RTX 3090 video card. That combination was validated by my fellow PC builders online as entirely suitable for AI, so I felt confident I was onto a good thing. And they weren’t wrong! My new PC worked well for my newest hobby, allowing me to dabble in making simple apps in DeepAgent. But with the gift of hindsight, I now realize that I made a big mistake with my build, and I deeply regret it. The issue was and still is that I had built a PC suitable for one use case only, and in doing so it has since become obsolete to my life. I arrived at my rig’s configuration by unknowingly breaking one of life’s less spoken-about rules: “Know thyself.” By that I mean two things: The first is that (with the exception of my work PCs that are just for work) in my life I have Buckley’s chance of compartmentalizing my personal devices for just one use — I’m just too busy. My phone is the classic example of that –- it’s my mobile calendar, my repository for holiday snaps, and my communication’s hub, all in one. Secondly, I had overlooked a personality trait that I sometimes exhibit, one that I share with the character Toad from the classic children’s story The Wind in the Willows, and that is the tendency to get all worked up over a new hobby that can last for several months but then I lose interest quickly and stop it abruptly. Of course, I should have known that AI was the most recent of these temporary hobbies, soon to be replaced by something else. These two oversights would have been no problem at all had I selected more versatile hardware for my AI PC. But I had chosen potato-like components, suitable for running LLMs but not much more. That aha! moment came after a lengthy coding session when I decided to give the neurons a rest and loaded up a game of CounterStrike: Go. My spiffy AI PC, which until then I had otherwise been chuffed about, at that very moment became an insubordinate nuisance. Yes, I had a decent graphics card, but my PC was severely bottlenecked at the CPU and no number of settings changes were going to improve that. Like most gamers, I have a minimum frame rate that I simply can’t bear to play beneath — that’s 60FPS. The best I could achieve on my AI PC, even in games with cartoony graphics, was a measly 40FPS — so it was of little use to me for any kind of serious gaming. Building a PC with limited capability was not a great idea.  Pixabay: Andreas160578 Sure, I could have just changed out the CPU for a more powerful one, but I had other bottlenecks too; Like the slow 8GB RAM onboard, and a PSU so underpowered it was only really suitable for running the lights in a chicken coop. If I had to find a moral in all of this, like in a family talk session at the end of a cheesy 1990s sitcom, it would be this: Don’t build a PC for just one application; reach higher with your hardware specs right from the get-go and build a more powerful and versatile rig capable of a whole bunch of stuff. If you do, you will save yourself a ton of hassle and possibly a ton of money too. And, no matter what your next hobby is (composing music? VR gaming? Heck, even snapping time lapse videos of your cat!) you will have a powerful enough rig to cover your needs! Further reading: I regret buying into the RGB hype for my gaming PC Read...

Since AI has diffused into every aspect of the technology sector, I’ve been more than a little tempted to try my hand at some of AI’s cooler applications. That growing temptation finally culminated in me building a desktop PC just for AI — to try my hand at vibe coding apps just for fun. My budget wasn’t that high, so for the build I landed on an AMD Ryzen 5 2400G CPU with a base clock speed of 3.6GHz, and an Nvidia RTX 3090 video card. That combination was validated by my fellow PC builders online as entirely suitable for AI, so I felt confident I was onto a good thing. And they weren’t wrong! My new PC worked well for my newest hobby, allowing me to dabble in making simple apps in DeepAgent. But with the gift of hindsight, I now realize that I made a big mistake with my build, and I deeply regret it. The issue was and still is that I had built a PC suitable for one use case only, and in doing so it has since become obsolete to my life. I arrived at my rig’s configuration by unknowingly breaking one of life’s less spoken-about rules: “Know thyself.” By that I mean two things: The first is that in my life I have Buckley’s chance of compartmentalizing my devices for just one use — I’m just too busy and inevitably end up using them for everything. My phone is the classic example of that –- it’s my mobile notetaker, my repository for holiday snaps, and my communication’s hub, all in one. Secondly, I had overlooked a personality trait that I sometimes exhibit, one that I share with the character Toad from the classic children’s story The Wind in the Willows, and that is the tendency to get all worked up over a new hobby that can last for several months but then I lose interest quickly and stop it abruptly. Of course, I should have known that AI was the most recent of these temporary hobbies, soon to be replaced by something else. These two oversights would have been no problem at all had I selected more versatile hardware for my AI PC. But I had chosen potato-like components, suitable for running LLMs but not much more. That aha! moment came after a lengthy coding session when I decided to give the neurons a rest and loaded up a game of CounterStrike: Go. My spiffy AI PC, which until then I had otherwise been chuffed about, at that very moment became an insubordinate nuisance. Yes, I had a decent graphics card, but my PC was severely bottlenecked at the CPU and no number of settings changes were going to improve that. Like most gamers, I have a minimum frame rate that I simply can’t bear to play beneath — that’s 60FPS. The best I could achieve on my AI PC, even in games with cartoony graphics, was a measly 40FPS — so it was of little use to me for any kind of serious gaming. Building a PC with limited capability was not a great idea.  Pixabay: Andreas160578 Sure, I could have just changed out the CPU for a more powerful one, but I had other bottlenecks too; Like the slow 8GB RAM onboard, and a PSU so underpowered it was only really suitable for running the lights in a chicken coop. If I had to find a moral in all of this, like in a family talk session at the end of a cheesy 1990s sitcom, it would be this: Don’t build a PC for just one application; reach higher with your hardware specs right from the get-go and build a more powerful and versatile rig capable of a whole bunch of stuff. If you do, you will save yourself a ton of hassle and possibly a ton of money too. And, no matter what your next hobby is (composing music? VR gaming? Heck, even snapping time lapse videos of your cat!) you will have a powerful enough rig to cover your needs! Further reading: 8 truly useful AI tools that make your life and work easier Read...

Intel said Wednesday that the company is adding its XeSS 2 technology to even more games, including top-tier titles like Diablo IV and Assassin’s Creed Shadows — pushing them into the “definitely playable” category by increasing framerates up to a whopping four times higher. As part of the announcement, Intel also revealed that some sort of XeSS support now exists in over 200 PC games. XeSS 2 was added in conjunction with the $250 Intel Arc B580 (“Battlemage”) architecture at the end of 2024, a “budget” graphics card that certainly helped alleviate some of the sticker shock that accompanied cards from AMD and especially Nvidia. It combines upscaling, frame generation, and latency in reduction, and doesn’t necessarily require an Intel GPU, either. The most important thing about XeSS is that it applies additional frames where it’s really needed — at the low end. Increasing a game’s frame rate from 260 fps to 360 fps looks great on paper, but where XeSS shines is pushing a game’s framreate up from below 60 fps, for example, to 90 fps and above. That swaps what could be a bumpy, frustrating experience into something that’s truly fun. Now, Intel says XeSS 2 has been added to 10 more games for a total of 19, following Intel’s decision to release the XeSS 2 SDK. Today, Intel is releasing estimated gaming performance on games like Marvel Rivals, where XeSS accelerates frame rates from 43 fps (native) to 138 fps using XeSS 2. That’s an enormous bump! Ditto for Rise of the Ronin: pushing frame rates up from 37 fps to 100 fps means an enjoyable, playable game. Those numbers are for the discrete Arc B580 card, of course. But Intel’s XeSS 2 is equally important in its latest Intel Core Ultra 9 285H processors, too. In my “Arrow Lake-H” review of the MSI Prestige 16 AI Evo (B2HMG), the problem I discovered is that at the time of the review only two games supported XeSS 2 (F1 2024 and Marvel Rivals) limiting the chip’s appeal as a gaming option. Now, with more games supporting XeSS 2, and hundreds supporting XeSS, you’re seeing a number of games suddenly become playable on the Intel Core Ultra 200 H-series platform. Intel is also saying that with the XeLL (low latency) technology built into the XeSS 2 technology, display latencies in supported games can drop by about half, improving the overall smoothness of the the game itself. In all, what you should expect with XeSS and its increased support overall is just a better gaming experience. One of Intel’s strengths has always been in engaging and influencing software developers, so it’s likely that this trend will continue. Combine that with the $250 Arc B580 — which my colleague Brad Chacos considers the best GPU under $400 — and Intel’s progress in notebook CPUs, even low-end hardware is gaining the ability to play today’s top-tier games. Read...