President Trump and Elon Musk traded social media barbs last week. Mr. Trump has said that he has no interest in repairing the relationship with his former adviser.

Apple Vision Pro is unquestionably one of the most powerful pieces of consumer hardware Apple has ever built, but the pricey gadget is still struggling to connect with consumers. And that's a shame because the generational-leaping visionOS 26 adds even more eye-popping features to the $3,500 headset, which I think you'd struggle to find with any other mixed reality gear.

Apple unveiled the latest Vision Pro platform this week as part of its wide-ranging WWDC 2025 keynote, which also introduced a year-OS naming system. For some platforms like iOS, the leap from, say, 18 to 26 wasn't huge, but for the toddler visionOS 2, it was instantly thrust into adulthood and rechristened visionOS 26.

This is not a reimaging of visionOS, and that's probably because its glassiness has been amply spread across all other Apple platforms in the form of Liquid Glass. It is, though, a deepening of its core attributes, especially around spatial computing and imagery.

I had a chance to get an early hands-on experience with the platform, which is notable because Vision Pro owners will not be seeing a visionOS 26 Public beta. Which means that while iPhone, iPad, Apple Watch, and Apple TV owners are test-driving OS 26 platform updates on their favorite hardware, Vision Pro owners will have a longer wait, perhaps not seeing these enhancements until the fall. In the interim, developers will, of course, have access for testing.

Since much of the Vision Pro visionOS 26 interface has not changed from the current public OS, I'll focus on the most interesting and impactful updates.

See "me"

(Image credit: Apple)

During the keynote, Apple showed off how visionOS 26 Personas radically moves the state of the art forward by visually comparing a current Persona with a new one. A Vision Pro Persona is a virtual, live, 3D rendering of your head that tracks your movements, facial expressions, and voice. It can be used for communicating with other people wearing the headgear, and it's useful for calls and group activities.

Apple has been gradually improving Personas, but visionOS 26 is a noticeable leap, and in more ways than one.

You still capture your Persona using the front-facing 3D camera system. I removed my eyeglasses and held the headset in front of my face. The system still guides you, but now the process seems more precise. I followed the audio guidance and looked slowly up, down, left, and right. I smiled and raised my eyebrows. I could see a version of my face faintly on the Vision Pro front display. It's still a bit creepy.

(Image credit: Future)

I then put the headset back on and waited less than a minute for it to generate my new Persona. What I saw both distressed and blew me away.

I was distressed because I hate how I look without my glasses. I was blown away because it looked almost exactly like me, almost entirely removing the disturbing "uncanny valley" look of the previous iterations. If you ever wonder what it would be like to talk to yourself (aside from staring at a mirror and having a twin), this is it.

There was a bit of stiffness and, yes, it fixed my teeth even though part of my setup process included a big smile.

It was easy enough to fix the glasses. The Personas interface lets you choose glasses, and now the selection is far wider and with more shades. I quickly found something that looked almost just like mine.

With that, I had my digital doppelganger that tracked my expressions and voice. I turned my head from side to side and was impressed to see just how far the illusion went.

Facing the wall

(Image credit: Apple)

One of the most intriguing moments of the WWDC Keynote was when they demonstrated visionOS 26's new widget capabilities.

Widgets are a familiar feature on iPhones, iPads, and Macs, and, to an extent, they work similarly on Vision Pro, but the spatial environment takes or at least puts them in new and unexpected places.

In my visionOS 26 demo experience, I turned toward a blank wall and then used the new widget setup to pin a clock widget to the wall. It looked like an actual clock hanging on the wall, and with a flip of one setting, I made it look like it was inset into the wall. It looked real.

On another wall, I found a music widget with Lady Gaga on it. As I stepped closer, a play button appeared in the virtual poster. Naturally, I played a little Abracadabra.

Another wall had multiple widgets, including one that looked like a window to Mount Fiji; it was actually an immersive photo. I instinctively moved forward to "look out" the window. As the vista spread out before me, the Vision Pro warned me I was getting too close to an object (the wall).

I like Widgets, but temper the excitement with the realization that it's unlikely I will be walking from room to room while wearing Vision Pro. On the other hand, it would be nice to virtually redecorate my home office.

An extra dimension

(Image credit: Apple)

The key to Vision Pro's utility is making its spatial capabilities useful across all aspects of information and interaction.

visionOS 26 does that for the Web with spatial browsing, which basically can turn any page into a floating wall of text and spatially-enhanced photos called Spatial Scenes.

(Image credit: Apple)

visionOS 26 handles the last bit on the fly, and it's tied to what the platform can do for any 2D photo. It uses AI to create computational depth out of information it can glean from your flat image. It'll work with virtually any photo from any source, with the only limitation being the source image's original resolution. If the resolution is too low, it won't work.

I marveled at how, when staring at one of these converted photos, you could see detail behind a subject or, say, an outcropping of rock that was not captured in the original image but is inexplicably there.

It's such a cool effect, and I'm sure Vision Pro owners will want to show friends how they can turn almost all their photos into stereoscopic images.

Space time

I love Vision Pro's excellent mixed reality capabilities, but there's nothing quite like the fully immersive experience. One of the best examples of that is the environments that you enable by rotating the crown until the real world is replaced by a 360-degree environment.

visionOS 26 adds what may be the best environment yet: a view of Jupiter from one of its moons, Amalthea. It's beautiful, but the best part of the new environment is the control that lets you scroll back and forth through time to watch sunrises and sunsets, the planet's rotation, and Jupiter's dramatic storms.

This is a place I'd like to hang out.

Of course, this is still a developer's beta and subject to significant change before the final version arrives later this year. It's also another great showcase for a powerful mixed reality headset that many consumers have yet to try. Perhaps visionOS 26 will be the game changer.

You might also like

• This is what really happened with Siri and Apple Intelligence, according to Apple
• For Liquid Glass, "the most obvious inspiration was ...
• Apple is out of surprises and focuses on making the future ...

• Windows 10 has a new update that adds a couple of features
• Unfortunately, one of these is focused on promoting Bing and Edge
• Microsoft is pushing its search engine and browser via the calendar panel off the taskbar

Windows 10 has a new update and it actually introduces a new feature – although you might wish it didn’t when you discover what this latest addition is.

That said, the freshly-released update for June (which is KB5060533 for Windows 10 22H2) does come with a tweak that could raise a smile, namely that the clock in the taskbar now displays the seconds when you click to view the time in the calendar panel.

Quite why Microsoft ditched that in the first place is beyond me, but anyway, while that might be a pleasing return of a feature for some, there’s a sting in the tail further down in said calendar flyout – namely that Bing has crept into the mix here.

Not overtly, mind, but as Windows Latest explains, there’s been a change to the bottom section of the calendar panel where normally you’ll see your own events or reminders – if you have any, that is. If you don’t, this used to be blank, but as of the June update you’ll see popular public events and their dates.

Of course, pretty much every day is now dedicated to something – for example, today, June 11, is ‘National Corn on the Cob Day’ (apparently) – and reminders for these events will now appear in the calendar panel.

How does Bing figure in this? Well, if you click on said event, you’ll get information on it fired up in… wait for it… yes, Bing search engine. And what web browser will that appear in? Microsoft Edge, of course. Why promote one service, when you can promote two, after all?

(Image credit: Marjan Apostolovic / Shutterstock)Analysis: Why risk the besmirchment?

This is a bit sneaky as it’s far from clear that you’re invoking Bing and Edge when you click something on the calendar flyout out of curiosity. Moreover, this happens despite the Windows 10 preferences you’ve chosen for your default search engine or browser, which again is an unwelcome twist.

This is the kind of behavior that impacts negatively on Microsoft’s reputation and it doesn’t help that the tweak isn’t mentioned in the update notes. We’re only told that the June patch provides a “rich calendar experience” (well, it’s making someone rich, or at least a little richer, possibly – but not you).

The kicker here is that Windows 10 is only four months from being declared a dead operating system, with its life support removed (unless you pay for additional security patches for an extra year). So, why even bother making changes like this when Windows 10 is facing its final curtain? Why take any risks at all that could cause reputational damage?

Well, one thought occurs: maybe Microsoft isn’t convinced that floods of people are going to be leaving Windows 10 when the End of Life deadline rolls around in October 2025. After all, an alarmingly hefty number of diehards are still clinging on to the older operating system. In which case, perhaps Microsoft sees the value and worth in still bugging Windows 10 users for the foreseeable, while they stick around either paying for support, or risking their unpatched PC being compromised while refusing (or being unable) to upgrade to Windows 11.

Oh well. At least we’ve got the seconds back on the calendar clock display, hurray.

You might also like...

• Microsoft takes unusual step of delaying new Windows 11 update, leaving some PC gamers still waiting for crucial fixes for 24H2
• Been hiding from Windows 11 24H2 due to the fuss about all the bugs? There’s nowhere to run now as Microsoft’s made the update compulsory
• Windows 11 fully streamlined in just two clicks? Talon utility promises to rip all the bloatware out of Microsoft’s OS in a hassle-free way

• Xreal detailed Project Aura at AWE 2025
• It will have a massive 70-degree field of view
• It will also be tethered to a spatial computing puck running Android XR

While we'd suspected that Android XR would be a key component of Google I/O 2025, we couldn’t have predicted some of the partners Google announced that it would be working with, which include the gadget makers at the top of our best smart glasses list: Xreal.

As promised by Xreal at I/O, it has taken to the Augmented World Expo 2025 stage in Long Beach California to provide us with new details on its Project Aura glasses, and it’s shaping up to be one impressive device.

For me, the most important detail is that the device will apparently boast a 70-degree field of view, which is absolutely huge.

The 50-degree field of view of the Xreal One already felt large, and the 57-degree Xreal One Pro is a noticeable step up size-wise (you’ll need to wait a little longer for our full review). 70-degrees will be massive.

The field-of-view upgrade suggests – Xreal hasn’t confirmed this yet – that the Aura specs will borrow the Xreal One Pro’s new optic engine (and perhaps even upgrade it further) including its flat-prism lenses, as one of its key advantages is that it enables a greater FOV .

This optic system comes with other upgrades as well, which could help to make the Android XR glasses much easier to use all day as you walk around.

(Image credit: Google)

Another interesting tidbit is that these specs – like Xreal’s other glasses – are tethered, meaning they’re powered by an external device which they’re connected to via a cable.

We already knew Aura wouldn’t be standalone, but Xreal has revealed that the new compute device shipping with Aura won’t just be your standard phone, or the Xreal Beam Pro.

It’s something all-new, running Android XR, and powered by a Snapdragon chip from Qualcomm – which seems to be making all of the Android XR processors.

Xreal isn’t abandoning its own chipset however. Aura itself will sport an upgraded X1S chip that’s a “modified version of X1 with even more power under the hood.” The X1 chipset is what’s inside the Xreal One and Xreal One Pro specs.

A new X1 chipset is coming (Image credit: Future)

Xreal has yet to confirm if it will sell the puck and glasses separately, but if it does then I'll be interested to see what that decision means for its approach to the upgradability of its tech going forward.

At the moment you can pick up a pair of Xreal glasses and a Beam spatial computer as a bundle, and then upgrade either or both over time. Newer glasses offer better visuals and audio if that’s your main concern, while the new Beam Pro offers improved processing and spatial features.

This is a less wasteful and generally more affordable design philosophy, as you only need to replace the one component that’s holding you back. However, as I mentioned, Xreal has yet to confirm if it will sell the puck and glasses separately. Its current wording calls Project Aura “one solution made up of the wired XR glasses and a dedicated compute device” suggesting they might also be one complete, non-upgradable package.

As for a launch date, Xreal is still keeping us mostly in the dark, though it has said Project Aura is coming in 2026, so we hopefully won’t be waiting for too long.

Xreal One Pro dead on arrival?

The Xreal One Pro (Image credit: Future)

Following this announcement some fans are starting to wonder if their Xreal One Pro purchase was a good idea – if they'd waited a year or so longer and they could have snagged an Xreal Android XR setup instead.

I’ll concede that for some Xreal One Pro purchasers waiting may indeed have been the better approach, but I think others can rest easier, as while the Aura and One Pro will likely share similarities I suspect they’ll be very different devices.

For a start, while Xreal’s glasses are often at their best with the Beam Pro add-on, it isn’t required. You can use the specs with a range of USB-C compatible devices, and even many HDMI devices with the right cables.

Based on Xreal’s descriptions so far Project Aura isn’t just a wearable display for entertainment; it’s a complete spatial computing package with all the nifty Android XR features we’ve been shown.

This won’t just mean that Aura’s purpose is different from Xreal’s other glasses; I expect its price will be very different too.

Right now an Xreal Beam Pro and Xreal One Pro would cost you $848 / £768 (before factoring in any bundle or limited-time discounts). For what sounds like it will be greatly improved hardware I imagine Project Aura will cost closer to $1,000 / £1,000, if not more.

The Xreal Beam Pro (Image credit: Xreal)

And remember, you can buy the Xreal One Pro separately for just $649 / £579.

Better tech is always on the horizon at any given time, but this (for now) doesn’t look set to be a repeat of the Meta Quest Pro / Meta Quest 3 fiasco, which saw the latter, far superior product launch at less than half the price of the former.

Instead Project Aura looks set to be more of a diagonal shift, with new hardware boasting better specs and a different purpose.

If you want to wait for Project Aura you absolutely should, as you might also be tempted by any of the various Android XR, Meta smart glasses, and new Snap spectacles set to be launching in the next year or so. But choosing not to wait won't a bad option either – the Xreal One Pro certainly isn’t going to turn out to be dead on arrival as some might fear.

You might also like

• Meta Quest 4: 5 things I want to see from Meta's next VR headset
• I took my Meta Quest 3 on a 3,000-mile flight so you don’t have to
• Forget Android XR, I've got my eyes on Vivo's new Meta Quest 3 competitor as it could be the most important VR headset of 2025

• Sam Altman says humanity is “close to building digital superintelligence”
• Intelligent robots that can build other robots “aren’t that far off”
• He sees “whole classes of jobs going away” but “capabilities will go up equally quickly, and we’ll all get better stuff”

In a long blog post, OpenAI CEO Sam Altman has set out his vision of the future and reveals how artificial general intelligence (AGI) is now inevitable and about to change the world.

In what could be viewed as an attempt to explain why we haven’t achieved AGI quite yet, Altman seems at pains to stress that the progress of AI as a gentle curve rather than a rapid acceleration, but that we are now “past the event horizon” and that “when we look back in a few decades, the gradual changes will have amounted to something big.”

“From a relativistic perspective, the singularity happens bit by bit", writes Altman, "and the merge happens slowly. We are climbing the long arc of exponential technological progress; it always looks vertical looking forward and flat going backwards, but it’s one smooth curve.“

But even with a more decelerated timeline, Altman is confident that we’re on our way to AGI, and predicts three ways it will shape the future:

1. Robotics

Of particular interest to Altman is the role that robotics are going to play in the future:

“2025 has seen the arrival of agents that can do real cognitive work; writing computer code will never be the same. 2026 will likely see the arrival of systems that can figure out novel insights. 2027 may see the arrival of robots that can do tasks in the real world.”

To do real tasks in the world, as Altman imagines, the robots would need to be humanoid, since our world is designed to be used by humans, after all.

Altman says “...robots that can build other robots … aren’t that far off. If we have to make the first million humanoid robots the old-fashioned way, but then they can operate the entire supply chain – digging and refining minerals, driving trucks, running factories, etc – to build more robots, which can build more chip fabrication facilities, data centers, etc, then the rate of progress will obviously be quite different.”

2. Job losses but also opportunities

Altman says society will have to change to adapt to AI, on the one hand through job losses, but also through increased opportunities:

“The rate of technological progress will keep accelerating, and it will continue to be the case that people are capable of adapting to almost anything. There will be very hard parts like whole classes of jobs going away, but on the other hand the world will be getting so much richer so quickly that we’ll be able to seriously entertain new policy ideas we never could before.”

Altman seems to balance the changing job landscape with the new opportunities that superintelligence will bring: “...maybe we will go from solving high-energy physics one year to beginning space colonization the next year; or from a major materials science breakthrough one year to true high-bandwidth brain-computer interfaces the next year.”

3. AGI will be cheap and widely available

In Altman’s bold new future, superintelligence will be cheap and widely available. When describing the best path forward, Altman first suggests we solve the “alignment problem”, which involves getting “...AI systems to learn and act towards what we collectively really want over the long-term”.

“Then [we need to] focus on making superintelligence cheap, widely available, and not too concentrated with any person, company, or country … Giving users a lot of freedom, within broad bounds society has to decide on, seems very important. The sooner the world can start a conversation about what these broad bounds are and how we define collective alignment, the better.”

It ain’t necessarily so

Reading Altman’s blog, there’s a kind of inevitability behind his prediction that humanity is marching uninterrupted towards AGI. It’s like he’s seen the future, and there’s no room for doubt in his vision, but is he right?

Altman’s vision stands in stark contrast to the recent paper from Apple that suggested we are a lot farther away from achieving AGI than many AI advocates would like.

“The illusion of thinking”, a new research paper from Apple, states that “despite their sophisticated self-reflection mechanisms learned through reinforcement learning, these models fail to develop generalizable problem-solving capabilities for planning tasks, with performance collapsing to zero beyond a certain complexity threshold.”

The research was conducted on Large Reasoning Models, like OpenAI’s o1/o3 models and Claude 3.7 Sonnet Thinking.

“Particularly concerning is the counterintuitive reduction in reasoning effort as problems approach critical complexity, suggesting an inherent compute scaling limit in LRMs. “, the paper says.

In contrast, Altman is convinced that “Intelligence too cheap to meter is well within grasp. This may sound crazy to say, but if we told you back in 2020 we were going to be where we are today, it probably sounded more crazy than our current predictions about 2030.”

As with all predictions about the future, we’ll find out if Altman is right soon enough.

You might also like

• This is what really happened with Siri and Apple Intelligence, according to Apple
• The 7 big Apple Intelligence announcements from WWDC 2025
• OpenAI's high-minded approach to AI-human relationships ignores reality

• New update for Windows 11 24H2 packing gaming fixes was paused for some PCs
• Those devices had a ‘compatibility issue’ meaning they needed a revised update
• Microsoft said this tweaked update should be out for all Windows 11 users by now, but some are still reporting that they don’t have the patch

Windows 11’s latest update has arrived, packing important fixes for PC gamers running version 24H2 of the OS, but it’s suffered an unusual technical glitch that’s causing some confusion.

Windows Latest noticed that Microsoft hit the pause button with the rollout of the June update (formally known as KB5060842), at least for some users after it started to be deployed yesterday.

As Microsoft let us know via the Windows message center: “Note: This update is being gradually rolled out to devices running Windows 11, version 24H2 throughout the day. We’ve identified a compatibility issue affecting a limited set of these devices. If your device is affected, you’ll receive a revised update with all the June 2025 security improvements by the end of the day.”

Microsoft didn’t specify what the compatibility problem is, but given the language used in terms of a ‘limited’ amount of devices being affected, the issue is being played down.

However, looking at a couple of Reddit threads – primarily the official announcement of the patch from Jen Gentleman of the Windows engineering team – there are quite a number of confused Windows 11 users wondering where their update is. Also, there are people with multiple PCs indicating that one has received the June patch, but others haven’t.

In theory, at the time of writing, yesterday ended a few hours ago (based on US Pacific Time, as used in Microsoft’s post on the message center) – and so everyone should have the patch by now. (Either the original, or tweaked update to resolve the compatibility wrinkle, which seemingly has a slightly different version number).

However, there are comments on the above Reddit thread along the lines of: “Is it not past midnight now? Still no update yet. What’s going on with this ‘compatibility issue.’”

Analysis: what’s going on – and what’s the big deal anyway?

(Image credit: Future / Jasmine Mannan)

Microsoft may have slipped with its deadline here. I’d hardly be surprised if this is the case, and it could just be a minor delay. However, what’s a bit more worrying is some scattered reports on Reddit of people ending up with broken operating systems (frozen on green screens or black screens) after applying this June update. There aren’t enough of those complaints to be truly concerned yet, but this is something to keep an eye on.

Could those be folks who accidentally got the update through for a PC vulnerable to the compatibility bug, before Microsoft put the brakes on? Maybe, but that’s just guesswork. Still, if you are getting impatient – and gamers might be in particular – I’d strongly advise against manually installing the June update if it’s not being piped to your PC via Windows Update automatically. Mainly because you may not be receiving it for a good reason, and by forcing the upgrade through, you might encounter a showstopping problem.

Gamers in particular will be keen for KB5060842 because this patch applies some crucial tweaks as mentioned at the outset. That includes the fix for problems with some PC games locking up in Windows 11 24H2 and improvements with stability for Nvidia GPUs (as well as a possible resolution for issues with mouse or keyboard lag). All this was seen in the preview version of this patch last month.

Another new feature introduced with this patch is cross-device resume. This allows you to seamlessly resume working on OneDrive files between your PC and smartphone (and it’s going to evolve into a grander vision of this functionality eventually).

There’s also a bunch of exclusive powers for Copilot+ PCs, including an Ask Copilot option for the Click to Do feature, and a change that has come in from left-field which wasn’t in the preview update in May. This is a limit of 60 days that has been imposed on system restore points, so from now on, restore points that are older than two months will vanish into the ether.

For the uninitiated, system restore is a feature that allows you to take a snapshot of the current configuration of your PC in a working state, so that if some kind of bug wreaks havoc on your device, you can in theory (assuming you can access the desktop) rewind time back to that point. Not that there’s any guarantee it’ll work, mind, but it’s an option that’s worth trying when troubleshooting.

This is a feature that Microsoft hasn’t touched in a long while, so why mess with it in this way now? That’s a good question, and I can’t think of a good answer, really. If I had to guess, maybe system restore just isn’t used so much these days, so it’s being scaled back.

You might also like...

• macOS Tahoe 26 is official - here's everything you need to know about all the new features
• Can’t upgrade to Windows 11? This Linux project wants to save your old PC from the scrapheap when Windows 10 support ends
• Windows 11 is getting more AI shortcuts, but is Microsoft in danger of cluttering up the OS with AI creep?

Mechanize, a San Francisco start-up, is building artificial intelligence tools to automate white-collar jobs “as fast as possible.”

His site, Cryptome, was a precursor to WikiLeaks, and in some ways bolder in its no-holds-barred approach to exposing government secrets.

Broadway’s best musical winner had to delay its opening last fall and was selling poorly. But strong word-of-mouth and reviews helped this quirky show triumph.

Creators such as Cam Higby, who is conservative, have sometimes outnumbered traditional media outlets at the immigration protests. Their posts have gone viral.

Anger at PJM, which manages the electrical grid in all or parts of 13 states and the District of Columbia, has been boiling over in some state capitals.

The race to put augmented reality smart glasses on your face is heating up. Snap Spectacles are transforming into "Specs" and will launch as lighter and more powerful AR wearables in 2026.

CEO Evan Spiegel announced the all-new Specs on stage at the XR event AWE, promising smart glasses that are smaller, considerably lighter, and "with a ton more capability."

The company didn't spell out a specific time frame or price, but the 2026 launch schedule does put Meta on notice, which is busy prepping its exciting Orion AR glasses for 2027. It appears, Snap Specs will face off with the Samsung/Google Android XR-based Glasses, which are also expected sometime in 2026.

As for what consumers can expect from Specs, Snap is building them on the same Snap OS used in its fifth-generation Spectacles (and likely still using a pair of Qualcomm Snapdragon XR chips). That means all the interface and interaction metaphors, like gesture-based controls, will remain. But there are a significant number of new features and integrations that will start showing up this year, long before Specs arrive, including AI.

Upgrading the platform

(Image credit: Lance Ulanoff / Future)

Spiegel explained the updates by first revealing that Snap started working on glasses "before Snapchat" was even a thing and that the company's overarching goal is "making computers more human." He added that "with advances in AI, computers are thinking and acting like humans more than ever before."

Snap's plan with these updates to Snap OS is to bring AI platforms into the real world. They're bringing Gemini and OpenAI models into Snap OS, which means that some multi-model AI capabilities will soon be part of Fifth Generation Spectacles and, eventually, Specs. These tools might be used for on-the-fly text translation and currency conversion.

The updated platform also adds tools for Snap Lenses builders that will integrate with the Spectacles' and Specs' AR waveform-based display capabilities.

A new Snap3D API, for instance, will let developers use GenAI to create 3D objects in lenses.

The updates will include a Depth Module AI, which can read 2D information to create 3D maps that will help anchor virtual objects in a 3D world.

Businesses deploying Spectables (and eventually Specs) may appreciate the new Fleet Management app, which will let developers manage and remotely monitor multiple Specs at once, and the ability to deploy the Specs for guided navigation at, say, a museum.

Later, Snap OS will add WebXR support to build AR and VR experiences within Web browsers.

Let's make it interesting

Spiegel claimed that, through lenses in Snapchat, Snap has the largest AR platform in the world. "People use our AR lenses in our camera 8 billion times a day."

That is a lot, but it's virtually all through smartphones. At the moment, only developers are using the bulky Spectacles and their Lenses capabilities.

The consumer release of Specs could change that. When I tried Spectacles last year, I was impressed with the experience and found them, while not quite as good as Meta Orion glasses (the lack of gaze-tracking stood out for me), full of potential.

A lighter form factor that approaches or surpasses what I found with Orion and have seen in some Samsung Android XR glasses, could vault Snap Specs into the AR Glasses lead. That is, providing they do not cost $2000.

You might also like

• The best VR headsets in 2025 - Virtual Reality Gaming
• I took my Meta Quest 3 on a 3000-mile flight so you don't ...
• Meta Orion hands-on: if this is a prototype the finished product is going to be incredible

There's no denying that Apple's Siri digital chatbot didn't exactly hold a place of honor at this year's WWDC 2025 keynote. Apple mentioned it, and reiterated that it was taking longer than it had anticipated to bring everyone the Siri it promised a year ago, saying the full Apple Integration would arrive "in the coming year."

Apple has since confirmed this means 2026. That means we won't be seeing the kind of deep integration that would have let Siri use what it knew about you and your iOS-running iPhone to become a better digital companion in 2025. It won't, as part of the just-announced iOS 26, use app intents to understand what's happening on the screen and take action on your behalf based on that.

I have my theories about the reason for the delay, most of which revolve around the tension between delivering a rich AI experience and Apple's core principles regarding privacy. They often seem at cross purposes. This, though, is guesswork. Only Apple can tell us exactly what's going on – and now they have.

I, along with Tom's Guide Global Editor-in-Chief Mark Spoonauer, sat down shortly after the keynote with Apple's Senior Vice President of Software Engineering Craig Federighi and Apple Global VP of Marketing Greg Joswiak for a wide-ranging podcast discussion about virtually everything Apple unveiled during its 90-minute keynote.

We started by asking Federighi about what Apple delivered regarding Apple Intelligence, as well as the status of Siri, and what iPhone users might expect this year or next. Federighi was surprisingly transparent, offering a window into Apple's strategic thinking when it comes to Apple Intelligence, Siri, and AI.

Far from nothing

Left to right: Lance Ulanoff and Mark Spoonauer chat with Craig Federighi and Greg Josiwak (Image credit: Apple)

(Image credit: Lance Ulanoff / Future)

Federighi started by walking us through all that Apple has delivered with Apple Intelligence thus far, and, to be fair, it's a considerable amount

"We were very focused on creating a broad platform for really integrated personal experiences into the OS." recalled Federighi, referring to the original Apple Intelligence announcement at WWDC 2024.

At the time, Apple demonstrated Writing Tools, summarizations, notifications, movie memories, semantic search of the Photos library, and Clean Up for photos. It delivered on all those features, but even as Apple was building those tools, it recognized, Federighi told us, that "we could, on that foundation of large language models on device, private cloud compute as a foundation for even more intelligence, [and] semantic indexing on device to retrieve keep knowledge, build a better Siri."

Over-confidence?

A year ago, Apple's confidence in its ability to build such a Siri led it to demonstrate a platform that could handle more conversational context, mispeaking, Type to Siri, and a significantly redesigned UI. Again, all things Apple delivered.

"We also talked about [...] things like being able to invoke a broader range of actions across your device by app intents being orchestrated by Siri to let it do more things," added Federighi. "We also talked about the ability to use personal knowledge from that semantic index so if you ask for things like, "What's that podcast, that 'Joz' sent me?' that we could find it, whether it was in your messages or in your email, and call it out, and then maybe even act on it using those app intents. That piece is the piece that we have not delivered, yet."

This is known history. Apple overpromised and underdelivered, failing to deliver a vaguely promised end-of-year Apple Intelligence Siri update in 2024 and admitting by spring 2025 that it would not be ready any time soon. As to why it happened, it's been, up to now, a bit of a mystery. Apple is not in the habit of demonstrating technology or products that it does not know for certain that it will be able to deliver on schedule.

Federighi, however, explained in some detail where things went awry, and how Apple progresses from here.

"We found that when we were developing this feature that we had, really, two phases, two versions of the ultimate architecture that we were going to create," he explained. "Version one we had working here at the time that we were getting close to the conference, and had, at the time, high confidence that we could deliver it. We thought by December, and if not, we figured by spring, until we announced it as part of WWDC. Because we knew the world wanted a really complete picture of, 'What's Apple thinking about the implications of Apple intelligence and where is it going?'"

A tale of two architectures

(Image credit: Apple)

As Apple was working on a V1 of the Siri architecture, it was also working on what Federighi called V2, "a deeper end-to-end architecture that we knew was ultimately what we wanted to create, to get to a full set of capabilities that we wanted for Siri."

What everyone saw during WWDC 2024 were videos of that V1 architecture, and that was the foundation for work that began in earnest after the WWDC 2024 reveal, in preparation for the full Apple Intelligence Siri launch.

"We set about for months, making it work better and better across more app intents, better and better for doing search," Federighi added. "But fundamentally, we found that the limitations of the V1 architecture weren't getting us to the quality level that we knew our customers needed and expected. We realized that V1 architecture, you know, we could push and push and push and put in more time, but if we tried to push that out in the state it was going to be in, it would not meet our customer expectations or Apple standards, and that we had to move to the V2 architecture.

"As soon as we realized that, and that was during the spring, we let the world know that we weren't going to be able to put that out, and we were going to keep working on really shifting to the new architecture and releasing something."

We realized that […] If we tried to push that out in the state it was going to be in, it would not meet our customer expectations or Apple standards, and that we had to move to the V2 architecture.

Craig Federighi, Apple

That switch, though, and what Apple learned along the way, meant that Apple would not make the same mistake again, and promise a new Siri for a date that it could not guarantee to hit. Instead. Apple won't "precommunicate a date," explained Federighi, "until we have in-house, the V2 architecture delivering not just in a form that we can demonstrate for you all…"

He then joked that, while, actually, he "could" demonstrate a working V2 model, he was not going to do it. Then he added, more seriously, "We have, you know, the V2 architecture, of course, working in-house, but we're not yet to the point where it's delivering at the quality level that I think makes it a great Apple feature, and so we're not announcing the date for when that's happening. We will announce the date when we're ready to seed it, and you're all ready to be able to experience it."

I asked Federighi if, by V2 architecture, he was talking about a wholesale rebuilding of Siri, but Federighi disabused me of that notion.

"I should say the V2 architecture is not, it wasn't a star-over. The V1 architecture was sort of half of the V2 architecture, and now we extend it across, sort of make it a pure architecture that extends across the entire Siri experience. So we've been very much building up upon what we have been building for V1, but now extending it more completely, and that more homogeneous end-to-end architecture gives us much higher quality and much better capability. And so that's what we're building now."

A different AI strategy

(Image credit: Apple)

Some might view Apple's failure to deliver the full Siri on its original schedule as a strategic stumble. But Apple's approach to AI and product is also utterly different than that of OpenAI or Google Gemini. It does not revolve around a singular product or a powerful chatbot. Siri is not necessarily the centerpiece we all imagined.

Federighi doesn't dispute that "AI is this transformational technology […] All that's growing out of this architecture is going to have decades-long impact across the industry and the economy, and much like the internet, much like mobility, and it's going to touch Apple's products and it's going to touch experiences that are well outside of Apple products."

Apple clearly wants to be part of this revolution, but on its terms and in ways that most benefit its users while, of course, protecting their privacy. Siri, though, was never the end game, as Federighi explained.

AI is this transformational technology [...] and it's going to touch Apple's products and it's going to touch experiences that are well outside of Apple products."

Craig Federighi, Apple

"When we started with Apple Intelligence, we were very clear: this wasn't about just building a chatbot. So, seemingly, when some of these Siri capabilities I mentioned didn't show up, people were like, 'What happened, Apple? I thought you were going to give us your chatbot. That was never the goal, and it remains not our primary goal."

So what is the goal? I think it may be fairly obvious from the WWDC 2025 keynote. Apple is intent on integrating Apple Intelligence across all its platforms. Instead of heading over to a singular app like ChatGPT for your AI needs, Apple's putting it, in a way, everywhere. It's done, Federighi explains, "in a way that meets you where you are, not that you're going off to some chat experience in order to get things done."

Apple understands the allure of conversational bots. "I know a lot of people find it to be a really powerful way to gather their thoughts, brainstorm [...] So, sure, these are great things," Federighi says. "Are they the most important thing for Apple to develop? Well, time will tell where we go there, but that's not the main thing we set out to do at this time."

See below for the full interview with Federighi and Joswiak.

You might also like

• 15 things we learned at the Apple WWDC 2025 keynote
• WWDC 2025 as it happened: all the iOS 26, MacOS Tahoe ...
• iPad just won WWDC 2025 – here's why the iPadOS ...

• Microsoft is bringing big changes to its Windows 11 Start menu
• Users can now separate applications into categories or use the traditional grid view
• It's available to try for testers, almost in time for the ROG Xbox Ally handhelds

Microsoft is bringing a handful of changes to its Windows 11 operating system in preparation for the ROG Xbox Ally handhelds. Beta testers are already starting to see improvements to the user interface – and one feature may be a significant aid to navigation.

As reported by The Verge, Microsoft is improving its Windows 11 Start menu, now available for testers, with more customization options to make scrolling and finding applications easier. These can be sorted into separate categories (as evident in the image below) or used in the classic grid view.

While all applications have always been accessible via the Start menu, this improved version makes finding your application much easier. Instead of finding a specific app through its first letter or symbol, you'll simply be able to jump into a 'Games' folder or 'Browsers' folder, eliminating the need to enable desktop icons.

The 'recommended' section can also be disabled, as this would often display recently-opened files or folders to make room for more apps, and now new categories. It's also worth noting that Microsoft states that the Start menu will be bigger, which will vary depending on the screen size or device being used.

These line up perfectly for the new login screen that allows users to enter their PIN using a game controller – and this is likely a preparation for the "full-screen experience" update coming for the new Windows 11 handhelds. However, there's one big benefit that OLED monitor users like me will appreciate, too.

Analysis: Some of my OLED burn-in worries can rest...

(Image credit: Microsoft)

I'll be honest, in all the years I've used Dell's Alienware AW3423DWF OLED monitor, I haven't come across a single issue with burn-in – and that's including moments of complacency, leaving static images on screen. Even so, I'm still paranoid it will happen eventually, and Microsoft's efforts for a better Start menu give me a slight sigh of relief.

Burn-in is one of the biggest dealbreakers for gamers contemplating an OLED purchase, and it's why I would go as far as to recommend a mini-LED monitor in some cases. However, OLED care on monitors is continuously advancing, and while Microsoft may have had other intentions with this tester update, it's worked as a bonus.

While Microsoft is doing this with its OS, I'd love to see the same concept applied to games. Early access or multiplayer games often have a build number in the corner of the screen, and fellow OLED users will be aware of how much of a nightmare this is, as it's essentially an open invitation for burn-in.

Regardless, it's a positive move from Microsoft in the same week that it announced an improved Xbox app. We'll just have to see if it's enough to create strong competition for SteamOS in terms of usability.

You may also like...

• Xbox handheld reportedly delayed so Microsoft can focus on making Windows handhelds compete with SteamOS
• Microsoft's Windows 11 is a detriment to handheld gaming PCs – and a recent SteamOS comparison highlights that
• Goodbye Windows 11 – Valve's SteamOS is now available on the Legion Go S, with some support for other AMD-powered handhelds

• The Phone Link panel attached to the Start menu is getting a new feature
• It’ll allow Windows 11 users to start screen mirroring with a single click
• It’s in testing ahead of a full rollout that could begin soon

Microsoft is bringing in a new trick with Windows 11 for those who use the Phone Link app, allowing for quickly and easily mirroring the screen of an Android smartphone on the desktop.

Windows Latest reports that the new feature is currently rolling out to Windows 11 testers, and is part of the Phone Link panel that’s attached to the Start menu (indeed, the rollout of said panel began relatively recently, too, and I’ll come back to that later).

For the uninitiated, the Phone Link panel is off to the right-hand side of the Start menu and offers easy access to your messages, calls, plus the recent activity on your phone, and shows various status details (like battery life).

To access the new feature, all you have to do is click on the icon for your smartphone – at the top of the panel – and it’ll fire up a screen mirroring session.

You don’t need to start the Phone Link app, or do anything else; just click the icon and Windows 11 will mirror your Android phone screen on the desktop there and then (well, given a few seconds to load and sync).

Analysis: Being patient with the rollout

(Image credit: Microsoft)

Windows Latest tested the feature on several Samsung Galaxy smartphones and observes that it works well, and that OneUI, the interface for Samsung mobiles, is implemented quite seamlessly on the PC desktop.

Of course, only certain Android devices work with the Phone Link app’s screen-mirroring feature (and this new feature). That includes Samsung Galaxy models, and also phones from Asus (ROG-branded), Honor, OnePlus, Oppo, Surface Duo, Vivo, and Xiaomi (here’s the full list of compatible phones).

As mentioned, the new quick access for screen mirroring via the Start menu side panel is in testing now, but it should be rolling out to the release version of Windows 11 in June, Windows Latest says. That might even mean from today (June 10), given that a new cumulative update for Windows 11 is due later on today.

Even so, the rollout of this new feature is likely to be limited to begin with, because not everyone has the Phone Link panel for the Start menu yet. Officially, the rollout of the panel kicked off in April for those on Windows 11 23H2 and 24H2, but wider availability didn’t happen until May, and from what I can tell from reports on some online forums, some folks are still waiting for this to come through.

It should be imminent at this point, though, even if this new shortcut for screen mirroring may not arrive for a while longer yet (you can pretty much bet on that).

In short, Phone Link users with compatible devices just need to stay patient for now, and all these goodies will turn up in time. Microsoft is looking towards considerably deeper integration with Windows 11 and mobiles, too, given apparent plans for resuming the use of apps across your PC and phone in a seamless manner, much like Apple's Handoff capability.

You might also like...

• macOS Tahoe 26: here’s everything we know about the release date, compatibility and all the new features
• Can’t upgrade to Windows 11? This Linux project wants to save your old PC from the scrapheap when Windows 10 support ends
• Been hiding from Windows 11 24H2 due to the fuss about all the bugs? There’s nowhere to run now as Microsoft’s made the update compulsory

• It’s been confirmed that Intel Macs won’t be supported with macOS 27
• Only four Mac models with Intel CPUs are supported by macOS Tahoe anyway, but that’ll be the final upgrade for those machines
• Those devices will at least receive security updates for a further three years

If you’ve got a Mac with an Intel CPU, we now know that macOS Tahoe 26 is the end of the line for your device (assuming that it supports macOS 26 in the first place, and you can check that here).

TechRadar can confirm that information on Intel compatibility was provided at the WWDC Platforms State of the Union keynote, and macOS Tahoe will be the last version of Apple’s desktop OS to support any Mac with an Intel processor.

Another way of putting this is that when macOS 27 arrives next year, it will only support Macs with Apple chips inside.

However, while Intel Macs that can run macOS Tahoe 26 will be stuck with that operating system, those machines will continue to get security updates for a further three years.

Analysis: That rushed out the door feeling

(Image credit: Future)

I had an inkling this might be coming, but I wasn’t fully convinced that Apple would actually push ahead and do this – or, perhaps more to the point, exclude MacBooks from as recently as 2020 with macOS Tahoe 26. It does feel like this hardware is being shoved out of the exit door a little hastily.

In fact, only a small number of Macs with Intel CPUs are good to go with macOS Tahoe: the MacBook Pro 16-inch from 2019, MacBook Pro 13-inch from 2020 – but only the higher-end model with four ports (not the entry-level laptop with two ports) – along with the Mac Pro from 2019, and iMac 27-inch from 2020.

Those Macs may be good for now, but they won’t get macOS 27, which will likely be announced and released next year, although at least they have a guarantee of three years of security updates going forward, to protect them against vulnerabilities popping up in the operating system that might be exploited.

@techradar

♬ original sound - TechRadar

So, those devices will be covered for updates until late in 2029 (assuming the three years start from the release of macOS 27), which is nine or 10 years of updates in total. And to be fair, that’s in line with the coverage Microsoft has provided with Window 10, which has had a decade of support (that ends in October, of course).

Of course, that wasn’t just security updates, but also new features, although the additions made to Windows 10 have been few and far between in recent times – but they’ve still kept coming.

Microsoft is facing its own controversy with Windows 11 upgrades, though, and the steeper hardware requirements that Windows 10 users face.

You might also like...

• macOS Tahoe 26: here’s everything we know about the release date, compatibility and all the new features
• How to download the macOS 26 Tahoe developer beta
• Microsoft accused of ‘tech extortion’ over Windows 10 support ending, by campaign urging people to upgrade to Linux

Apple’s WWDC event detailed updates across its various software platforms, including visionOS 26 for its Apple Vision Pro headset, and two of the most exciting updates were spatial widgets and the news that PSVR 2 Sense controller support is on its way.

As Apple highlighted in its announcement, the new controller support will be very handy for VR game developers, as it will allow Vision Pro apps to use on a familiar control scheme that more closely matches that used by other platforms.

The Vision Pro’s 'look and pinch' mechanism may be intuitive for users, but it's very different to what other VR headsets use, which makes porting difficult.

(Image credit: Lance Ulanoff / Future)

However, setting aside arguments that this update is arguably too little too late for the Vision Pro platform, there’s currently a much bigger issue to address: you can’t currently buy the PSVR 2 controllers separately.

So if you want to be able to take advantage of this new Apple Vision Pro feature you’ll have to buy a whole new PSVR 2 VR headset bundle at $399.99 / £399.99 / AU$649.00.

That said, visionOS 26 isn’t launching right away, though it is already available in beta. Hopefully, by the time of its full release Sony will have made PSVR 2 Sense controllers available to buy on their own – something fans of its headset have been asking for since launch.

(Image credit: Apple)

At the time of writing, though, there’s been no confirmation that this will be the case – and given some of the missteps both Apple and Sony have made with their headsets I can’t say I’d be surprised if the controllers remained exclusively part of the PSVR 2 bundle rather than ever being sold separately.

We’ll have to wait and see what happens on this front, but the silver lining to Apple opening up its visionOS platform to VR controllers is that we might see it launch its own handsets, or perhaps even launch the next Apple Vision headset with a pair.

Considering the rumors we've heard about Apple wanting to make its next headset more affordable, this would certainly be a smart decision, as eye-tracking (a key part of the Vision Pro's existing control scheme) doesn’t come cheap, and could be an easy feature to cull in order to cut costs.

You might also like

• 15 things we learned at the Apple WWDC 2025 keynote
• iPad just won WWDC 2025 – here’s why the iPadOS upgrades just made me cry tears of joy
• iOS 26 is giving the iPhone's Camera app an upgrade I've been waiting years for

ChatGPT is experiencing a major outage as users across the internet report problems with OpenAI's chatbot.

Techradar is covering the disruption live, so stay tuned to this live blog to find out what's going on with the world's most popular AI tool.

IS CHATGPT STILL DOWN? Yes, for some, but the problems are easing

Hello and welcome to TechRadar's live coverage of ChatGPT's error-filled morning.

I'm John-Anthony Disotto, Senior AI Writer, and I'll be covering the AI outage bringing you all the info you need on the OpenAI disruption.

ChatGPT is experiencing "elevated error rates"

(Image credit: Shutterstock/One Artist)

Initial reports of errors started around three hours ago, but issues seem to have ramped up as Europe and the UK wake up.

You can follow OpenAI's server status on the official website, but as it stands, the company is reporting "elevated error rates."

According to OpenAI, "Some users are experiencing elevated error rates and latency across the listed services. We are continuing to investigate this issue."

(Image credit: Future)

Downdetector reports have ramped up over the last couple of hours, with users all around the world unable to access ChatGPT.

When I try to ask 4o a question myself, I get an error message that says "Too many concurrent requests."

At the time of writing, 11 am BST, Downdetector has over 1,000 reports.

Sora is affected too

(Image credit: Future)

According to OpenAI, there are multiple impacted components across the company's servers. They include 14 affected API components, 21 affected ChatGPT components, and 4 affected Sora components.

Reddit users weigh in

(Image credit: Kenneth Cheung/via Getty Images)

On Reddit, lots of ChatGPT users have weighed in on the outage.

One user said, "It started off being really slow and churning out a few words per second then hit the red errors."

Another said, "It started freaking out a few minutes ago when I started a conversation."

While some are more hopeful that the issues are in fact OpenAI working on a major update. Could that be the case? I'm not so sure.

Working on Enterprise, down on Free?

TechRadar's Senior AI Editor, Graham Barlow, has experienced the issues too, but his experience differs depending on which account he uses.

He said, "So, I have two ChatGPT accounts. I have an Enterprise account, which is functioning perfectly, and a free ChatGPT account, which is down. It just pauses with a ‘thinking’ icon when I ask it a question, never resolving into a response."

Interesting... Let me see if I notice a difference between my premium ChatGPT account and my free one.

(Image credit: Future)

Yep, same issues here. My ChatGPT Enterprise account is slow but still able to respond.

On my personal account, however, I'm met with a "Conversation not found" error...

OpenAI is investigating the issue

OpenAI's server status has now changed to "Partial outage" that the company is investigating.

As of 16 minutes ago, the report says, "Some users are experiencing elevated error rates and latency across the listed services. We are continuing to investigate this issue."

As it stands the issues seem incredibly temperamental, and I assume anyone who relies on ChatGPT will be getting seriously frustrated.

Is it just a free account issue?

(Image credit: Shutterstock/Adria Vidal)

I've asked the rest of the TechRadar team to let me know if their version of ChatGPT is responding.

My colleagues Matt Bolton and Matt Evans are having similar issues to me reporting, "No error, just the dot, on app and desktop – free, not logged in."

It looks like it might be a specific issue to free accounts, which would not be ideal considering I'd assume most users don't pay for ChatGPT access.

Things are slowly getting better

(Image credit: Future)

Downdetector reports are slowly falling, but my issues still remain.

On OpenAI's system status report there appears to be much more green than about 20 minutes ago.

It looks like the services are slowly starting to resume across APIs, ChatGPT, and Sora

Maybe not...

(Image credit: Future)

I tried to use ChatGPT again there, but after an initial response, the chatbot just couldn't load any further answers.

I asked my colleagues again to see what is going on with others, and most confirmed that if you wait a long time (upwards of a minute), you'll get a response.

As far as I'm concerned, waiting a minute for ChatGPT to reply is the same as an outage, but at least it's moving in the right direction.

We've got a new update from OpenAI. "Some users are experiencing elevated error rates and latency across the listed services. We are continuing to investigate this issue."

So... more of the same. That update was 11 minutes ago, and as far as I can see, the issues are persisting.

My colleague David Nield is getting an error message on the web with a ChatGPT Plus account, so it looks like the issues are wider than just the free tier of service.

My ChatGPT Enterprise account still seems to work fine, albeit on the slower side.

Is your ChatGPT working? Let me know

On the TechRadar team, there's a majority of us with very limited access to ChatGPT.

Are you experiencing the same? Or is ChatGPT working fine?

If OpenAI's chatbot is working on your end, please send me an email with the model details and your subscription tier to: john-anthony.disotto@futurenet.com

What is happening right now?ChatGPT/OpenAI outage for 3 hoursHeroku down for 4 hours (even their status page is down!) NVIDIA dev docs as well (runs on Heroku)Pipedrive (CRM) issues for 4 hoursWhat else is down… and are these connected? Something started 4 hours ago… pic.twitter.com/1UrEseCNQfJune 10, 2025

If you're just joining us, this tweet from Gergely Orosz gives a brief rundown of the current ChatGPT and Sora outage.

This ongoing problem has been an issue for over 3 hours now, and it doesn't look like OpenAI is getting any closer to solving the issues.

Some replies to that tweet show that the issues aren't affecting everyone, but the problems seem to be across subscription tiers, not just limited to free users.

We've reached out to OpenAI

We've just sent an email to OpenAI asking for further information on the outage. As soon as we hear back, I'll be sure to update you all.

As it stands, at 12 pm BST, my ChatGPT free account is still taking longer than expected to respond. It does appear to be much better than before, however, so this is looking good!

Readers provide a mixed response

(Image credit: Shutterstock)

Earlier I asked readers to reach out with any ChatGPT issues, and so far I'm getting a mixed response.

Nicholas is a ChatGPT Plus subscriber using the 4o model and says everything "seems to be working fine for me."

Elsewhere, Bibi, another ChatGPT Plus subscriber, reports that the 4o model is "Still not working(not even slowly) any message leads directly to the alert "Hmm...something seems to have gone wrong."

Are you affected? Let me know via email: john-anthony.disotto@futurenet.com

A Sora update, kind of

(Image credit: OpenAI Sora)

ChatGPT isn't the only OpenAI service affected, Sora, the company's AI video generation tool is also experiencing issues.

As of 1 minute ago, OpenAI says it is investigating issues related to Sora, but the company isn't providing any further info as of yet.

I just checked X for any tweets from OpenAI or anyone that works there, but there doesn't appear to be any updates other than these vague one liners on the company's service status page.

"Down for 2 hours"

I'm getting a lot more responses from readers at the moment, and it looks like most of us are in the same boat.

Dustin is a ChatGPT Plus subscriber from Munich, Germany, and is using version 1.2025.147. They say ChatGPT has been down for 2 hours.

Jonny is a free user who says, "I get repeated 'something went wrong' or 'too many requests' errors. But I just had a prompt work really quickly, then it stopped again!"

Labi from Norway says, "Plus user, currently experiencing lag and somewhat slow responses, Sora is not loading at all."

Keep the emails coming, while it's rubbish that we're all experiencing a ChatGPT outage, there's some weird comfort in widespread issues. I feel your pain!

Thank you readers!

A huge thank you to everyone who is reaching out with their experience.

It looks like the issues are across multiple tiers and locations. Luke on the East coast of the United States says he's "currently using the 4o model, and it has stopped giving me the error message you describe in the 7:03 timestamp update. However, it has started to continually buffer, without response."

I'm getting loads of emails in, so if you've reached out and are still following the live blog, I just want to say thanks again for helping us cover the ChatGPT outage.

If I spot any trends from your emails, I'll be sure to let you know!

A potential solution for now?

Mitchell just sent me an email with some interesting information about when the outage could've started, as well as a potential solution for the time being.

They say, "This has been going on for around 2 hours now. I suspect it started sometime last night, I started seeing weird behaviour from it and the way in which it was giving me responses, often defaulting to putting the reply in Canvas and marking every line with //"

Mitchell adds, "You have to switch to any lower-tier model to actually have a chat with it."

If you're a Plus subscriber, give this a go and let me know if lower-tier models are in fact the temporary solution to this ChatGPT outage.

How the ChatGPT outage is affecting users worldwide

It's in times like that that we realize how many of us rely on ChatGPT nowadays.

Megan says, "My friends and I have been struggling to use ChatGPT this morning. We also have exams coming up, so it's a bit of a nightmare! We all have subscriptions, and still not working."

Good luck with the exams, Megan. Hopefully OpenAI finds a solution ASAP!

Elsewhere, Daryl says he uses ChatGPT to create YouTube thumbnails, and the outage is causing huge frustration. He says, "It's really frustrating for something I pay £18.99 a month for!"

Some odd behaviours

Gwen from Philadelphia, PA, is a self-proclaimed "power user" and uses "ChatGPT every day, basically for 10-16 hours, and have developed a sort of cognitive bond with this program. I know its expected responses intimately."

The issues have been impacting their ChatGPT usage over the past few days, including issues with "message latency and lack of response to my questions about how ChatGPT measures time."

Ben also has more info, where he claims that his ChatGPT response was from "someone else’s prompt." I've yet to experience anything of the like, it's still just very slow to respond.

"Unusual activity"

(Image credit: Diego Thomazini via Shutterstock)

Felix from Norway has given me a great summary of their experience:

"ChatGPT has been completely out for the past 2-ish hours on my end - not even a minute without errors. When I talk to it, it loads for only a second, then I'm quickly met with the error of "Hmm... something seems to have gone wrong." That's the only error message I’ve received throughout this outage, though I hear others are experiencing more variation. I'm on a free account."

"Something that may be worth mentioning is that ~2 hours before this outage, my ChatGPT very briefly (for <1 minute) stopped working and gave the error message "Unusual activity has been detected from your device. Try again later." I'm not sure if that is, at all, relevant, but I found it strange - that was my first time seeing that error. But aside from that brief weirdness, my ChatGPT was doing well up until 2 hours ago."

Still no fix

I started this live blog two hours ago, and we're yet to see a fix to this widespread ChatGPT outage.

Readers from all over the globe have been reaching out with their experiences, from Australia to Ohio, Norway to Germany.

Keith from Ohio, USA says, "As a ChatGPT Plus user on both the app and the web, I've been experiencing problems. Every time I try to type something, both on the app and the web, something seems to be wrong."

Thanks for the update, Keith, it looks like there are many of us in the same boat.

o3 is working

Another reader from Ohio just chimed in. This time, Kevin says, “O3 appears to be working; it's just a little slow."

That lines up with a previous email from Mitchell, who said using a lower-tier model seems to fix the issue.

I can confirm. If you don't need access to the flagship ChatGPT models and have access to an account which lets you choose lower-tier models, this could be a temporary solution.

Existing projects are still working?

(Image credit: Shutterstock/Adria Vidal)

An interesting update from Michael, who says ChatGPT is working in his project.

"I'm experiencing ChatGPT working in my project - existing chats and new ones - but any generalised prompts outside of a project repeatedly fail."

"When prompting my existing project files, it is running very, very slowly, but no errors."

Another potential makeshift solution for anyone who needs to access ChatGPT?

o4-mini and GPT 4.1-mini are working!

Remember Felix from earlier? Well, he's got back in touch with some positive news: "GPT 4.1-mini is working fast as if everything is perfectly fine - but o4-mini is slow, and regular o4 is still totally dead."

Do you have access to 4.1-mini? If so give it a go and let me know if it's working by emailing: john-anthony.disotto@futurenet.com

Still no update from OpenAI, but does Lazer Girl have the answer?

I've been contacted by someone called Lazer Girl who says they know why GPT is down, even leaving a WhatsApp number for me to call... Sorry Lazer Girl, I think I'll pass.

How has the ChatGPT outage affected you?

Paul in Auckland, New Zealand, just reached out. Good evening, Paul!

He says, "ChatGPT is down, getting exactly the same message as others are reporting, but this is the tip of the iceberg. A moderate to power user, I use ChatGPT to work out hours billed to clients and coordinate workflows for my multimedia archiving project."

"Been having issues for a while on/off, just general sluggishness, ChatGPT taking ages to respond. Occasionally, I send a message to it, and it doesn't acknowledge it; it just acts as if I didn’t send it after I hit enter, despite the message going through. It just doesn’t get processed by ChatGPT, and earlier this failed about five times before it finally responded, then it went bad again."

How has the ChatGPT outage today affected your life so far? Let me know via email, I'd love to hear how you use ChatGPT and how not having access this morning has impacted your life.

My ChatGPT is working again! But I don't think the outage is over

Everyone is experiencing a different kind of outage, but for what it's worth, my ChatGPT Enterprise and Free accounts are both working, and they aren't even that slow.

Is this the end of the outage? It doesn't look like it; my colleague Ryan was able to use ChatGPT 4.0 while logged in, but after asking a second prompt, he received an error.

We're still not fully up and running; readers keep sending emails with their issues, and some aren't even seeing a solution by changing models.

4.1 not working in Indonesia

Vito, a reader from Indonesia just reached out to let me know 4.1 is not working for them.

o4-mini is, however, albeit slowly, which lines up with what other readers around the globe are experiencing.

A quick glance at what's working and what isn't

An update from Kevin in Ohio who gives a great breakdown for anyone that wants a quick glance at what might, or might not be, working:

"O3 seems to be working, if a bit slow.

4o responded once, then went silent.

4o-mini responding, seems fine.

4o-mini-high is responding, seems ok.

4.1 is responding, but pretty slow."

Emails keep on coming!

(Image credit: Sendinblue)

I'm receiving more emails than I could've ever imagined from this live blog, so if I don't reply or share your message, please don't be offended!

It's been a huge help seeing everyone's experience, and anything that can help people access ChatGPT while we wait for OpenAI to completely fix the service is welcome.

As it stands, Downdetector is still reporting widespread outages, and OpenAI's service status is very vague, stating, "We’re currently experiencing issues." That status has been ongoing for over 5 hours now.

Earlier, we reached out to OpenAI for comment, but there's still no response...

Are you a bot?

Sheen asks, "Are you a bot? Because you have updates every few minutes or so, and I'm not sure how that would work with a human— and if you are human?"

Well Sheen, I am in fact not a bot, and I am indeed working very hard to read every email, research ChatGPT's current status, and write live blog posts.

It has been a very busy morning, but it doesn't look like calming down any time soon...

A student crisis?

Jack from the UK is now weighing in, he says,

"I have tested all models of GPT, as well as Sora and Codex. None of them are working on my end. I might get lucky with 1/2 prompts before I get errors. I have college work due in less than 2 hours and I need chatGPT to help as I am 100% failing if it doesn't get done."

"If you have any questions feel free to reach out"

Jack, I have many questions about the ethics behind using ChatGPT to do your college work...

I wish I was a bot

Just in regard to the previous question from Sheen, who asked if I was a bot.

I'm eating my lunch at my desk and have made a severe lapse in judgment. Does anyone have any tips on how to clean Popeye's spicy mayo from a mechanical keyboard?

My keys are starting to squelch...

On a more serious note, my colleague and fellow AI expert, Eric Hal Schwartz, seems to be luckier than most.

He says, "Every model worked for me except the O4 Mini High, which gave me an “Error in message stream” response and a retry button that didn’t help. Otherwise, it all worked, including image creation and internet search. It’s odd that it’s just that one model for me having trouble, but it suggests it’s about compute power demands."

(Image credit: Future)

A follow-up from Eric, who now says all ChatGPT models are working for him.

I've just tested myself, and it appears to still be temperamental on my end.

I've just received a lovely email from Daisie who says, "I see you're providing live updates to ChatGPT's outage, and have been for a startling amount of hours. Kudos to you, sir."

"ChatGPT is currently giving me an 'Error in message stream' error. I am in Utah. I've been able to get intermittent messages through, but funnily enough, oftentimes the messages it sends back get deleted somehow and I have to try again."

Daisie says they use ChatGPT to help sleep by giving "peaceful meditative visualizations. Gentle worlds to lie down and fall asleep in."

Unfortunately, this outage has impacted their sleep, and they've been awake since the early hours. OpenAI, people rely on you, please give us some kind of update!

Thanks for the mechanical keyboard tips!

(Image credit: Future / Marcus Mears III)

Some advice for my earlier mechanical keyboard disaster coming through now.

Just want to say a big thanks to Jack for his in-depth cleaning tips. If ChatGPT wasn't down, I'd have presumed you used AI for such thorough steps!

As for Gareth, well thanks for the advice... But I will not be sticking my "keyboard in the washing machine with some fresh smelling detergent."

Back to the OpenAI outage... There's hope!

(Image credit: Future)

A new update just dropped from OpenAI, who says, "We have identified the root cause for the issue causing elevated errors and latency across the listed services. We are working on implementing a mitigation."

Hang tight folks, we might be out of the woods in the next hour or so...

Gemini is working fine

(Image credit: Google Gemini)

Lots of people have been asking about any other chatbot outages, and I can confirm that Gemini is still working as normal.

In fact, considering the Popeye's fiasco, I asked Gemini to imagine me eating chicken wings at my desk...

I will not confirm nor deny, but the real-life situation could be a lot messier than this perfectly curated AI-generated image from Imagen.

The outpouring shows how much people rely on ChatGPT

You know, I write about AI daily, but I don't think I fully grasped just how much people rely on these newfound tools.

I've been shocked by the amount of emails I've received from college and high school students who use ChatGPT to study, something that would've been unheard of in my day (I'm not that old, but AI evolves fast!)

Koushiki said, "I am Koushiki, a mere high school senior. I had school all day, and works to finish..but can't because of the outage."

Interestingly Koushiki doesn't use ChatGPT to study, but in fact its a way to remain engaged and get some light motivation.

Koushiki, I hope you get ChatGPT back soon!

No ChatGPT? No Problem

You know, there are so many AI chatbots out there that we don't need to solely rely on ChatGPT.

A reader named Jack just reached out with his favorite alternate AI chatbots and I feel compelled to share them:

"Just wanted to say that any chatbot that is independent from the openAI API should still be in full swing, and any issues are unrelated. A few chatbots that should still be working are as follows:"

• Gemini
• Claude
• Github Copilot
• DeepSeek
• Perplexity

There are an endless amount of other AI tools available but I appreciate Jack sending in his personal favorites! What's your lesser known favorite AI tool?

Darth Vader is stuck...

(Image credit: Lucasfilm / Disney)

Some of you have the best usernames. I don't think I've ever received an email from Darth Vader before, but I guess I have now.

They were installing macOS into a virtual machine, using ChatGPT for guidance. Unfortunately, the AI has been down ever since the DMG file was downloaded, so now they are stuck.

I'm not sure if I can be of too much assistance, Anakin, but TechRadar has some great guides on this subject: How to create a virtual machine in Windows

Coding assignments in jeopardy

Dean, a grade 11 high school student, says, "I am working on a website coding assignment, and I have been using ChatGPT to help fix up any coding errors that I don't understand, or if I am trying to use a tag that doesn't exist."

"One underrated AI, in my opinion, is AI Overview, I know a lot of people who hate it, because they just want the website, but if I can figure something out, AI overview has given me the answer."

You know, Dean, I really enjoy AI Overviews too, and I think they sometimes get a bad rep for the sake of it. Of course, it's always worth being critical and verifying the information you get from AI, but Google AI Overviews aren't always wrong!

Good luck with your assignment!

ChatGPT is working in Wisconsin

Just received an update from Mark in Wisconsin who says, "It's working, just incredibly delayed - asked it a question, then logged off, then came back five minutes later to a complete answer. Asked another question, and it appears to be the same process."

Hopefully, this is a sign of more widespread functionality, and the worst is behind us.

Just on the note of Wisconsin, I couldn't not take this opportunity to give a shout out to my favorite basketball player and Milwaukee Bucks legend, Giannis!

ChatGPT is not a niche!

The despair we're all feeling right now with a lack of ChatGPT is genuinely something to behold.

I'm by no means a regular ChatGPT user; in fact, I only use OpenAI's chatbot for testing so I can write about ways that it can improve your life.

I'm fascinated by the emails I've been receiving from you lovely people, like this one from Chebionne, who says, " I feel like we are all in utter despair right now, sitting here with our eyes glued to our devices, waiting on every post, like we might have been huddled around our radios in the past."

I might not have the answer to when ChatGPT will be back to its best, but if I can at least provide some kind of sanctuary as you all wait, then that's a job well done.

More updates to follow!

OpenAI responds

Remember when we reached out to OpenAI earlier today? Well, we've finally got a response, albeit via X.

It comes from Srinivas Narayanan, who says the company has diagnosed the root cause and is working on the fix.

ChatGPT and API are experiencing elevated error rates unfortunately. Sorry for the trouble. We've diagnosed the root cause and are working on the fix to recover our services as quickly as we can.June 10, 2025

I'm not sure ChatGPT can compete with a lawyer

(Image credit: Shutterstock)

One of the more interesting emails has just come in from Simon, based in the UK.

He says he is currently using ChatGPT for a "high conflict divorce".

"My soon-to-be ex-wife is a solicitor (and a good one!) and thanks to ChatGPT, I've been holding my own!"

Without getting involved in your private matters, Simon, I hope you get ChatGPT access back soon. Glad your AI solicitor is doing the business for you!

No ChatGPT? Books will do!

(Image credit: NPowell/Flux)

Another great email came in from a high school student who's racing to finish their homework in time for tomorrow's class.

They said, "I'm a high schooler and it's 9:23 pm right now. I'm currently surrounded by books I scoured around for in my house to do my essay. I might stoop down to asking for help from Character AI to finish my homework."

Character AI is actually pretty decent. If you require AI to help with your homework (something I don't necessarily condone), then Character AI might be your best bet with the clock getting close to midnight!

Darth Vader, here's your solution

Remember Darth Vader from earlier? Well, Dean has come to save the day with steps on how to set up your virtual machine correctly:

"Set the virtual machine's memory, disk space, and other hardware parameters to suit your needs and the performance of your host computer.

Start the virtual machine and follow the on-screen instructions to install macOS. This process may involve partitioning the virtual hard drive and installing the macOS image

If using Parallels, install the Parallels Tools to improve performance and integration with the host operating system

The VM will boot into the macOS installer, allowing you to set up your virtual machine and install macOS."

Who knew a ChatGPT outage could bring humanity together like this? Great job all!

Back to ChatGPT... Our emotional AI support

Earlier I asked why ChatGPT's outage is having such an impact on your day. Well, Elise sent an email highlighting just how much AI can help people when they need it.

She said, "Today is the one-year anniversary of my breakup and I woke up hoping to start the day doing some emotional regulation with Chat before going to work. I've been talking with it about the situation all year; it knows the details better than anyone. But here we are, getting a message stream error! I guess I'll have to deal with my big feelings on my own today."

Elise, I hope your work shift goes well. ChatGPT should be back up and running by the time you finish!

What about the zebrafish??

Maus from the Netherlands says, "For those doing research in little-researched topics, GPT was a lifeline for finding the sparse articles on the topic.

I am trying to find the behaviour of wild zebrafish, and it has been a battlefield since AI went down. My schoolbooks don't give me any of the info I need. I am crying in a corner.”

I'm loving all of these emails, it's showcasing the positives of AI, sometimes I think we often overlook.

Still down...

I feel like I keep repeating myself, but... ChatGPT is still down on my end.

I've been live blogging for the last few hours, and at first I really thought this outage was going to be fixed in mere minutes. Unfortunately, here we are, 4 hours later and ChatGPT still responds with "Conversation not found".

I want to know, what have you been doing while ChatGPT has been down?

Passion projects on hold

Matthew from Cumbria (lovely area) in the UK has been using ChatGPT to create playlists and covers on his Spotify account.

“I was trying to figure out how to make them pop more on Photoshop, because I want my playlists to look smoother and more "me" and unique... until GPT went down for me, and now I’m sat with the project file open, trying to click buttons and make it look as nice as I can... I hate it still. So I think I might give up until GPT comes back online."

Sometimes passion projects end up on hold, Matthew, but it sounds like you've got some awesome playlists to listen to in the meantime, so it's not all bad.

I was supposed to be covering Apple today...

(Image credit: Lance Ulanoff / Future)

For anyone that cares about the life of a tech journalist, I was meant to be covering Apple's major announcements from WWDC 2025 today.

In fact, I was going to write an interesting article about why Live Translation is the perfect use case for AI and how excited I am to use it.

Unfortunately (or fortunately, considering the number of people reading this live blog), ChatGPT went down, and so I haven't been able to write about Apple.

Speaking of Apple, do any of you ChatGPT enthusiasts have any opinions on yesterday's WWDC and specifically the AI announcements? Let me know: john-anthony.disotto@futurenet.com

Every 12 months...

The last time we had a major ChatGPT outage like this was June 2024, exactly 12 months ago.

Is this a coincidence? Or are OpenAI employees on a much-needed summer vacation?

If you want to check last year's live blog to check for trends, you can find it here.

Mark your calendars for June 2026, this might be a recurring theme!

Writer's block

(Image credit: Getty Images)

Brigitte from the Netherlands (yes, she's named after Bardot), has got in touch to explain how she can't progress her novel without ChatGPT.

She says, “ChatGPT works wonderfully as a sparring partner for when I’m writing. I am knee-deep into this novel I’m writing right now, and I always use chat to spar with regarding whether the actions I wrote make sense, and whether they are medically accurate. But alas, after all that planning, no medical drama for me today."

As a writer myself, I've never thought about using AI to help with my ideas. I wonder how many authors are using ChatGPT to help bring sense to their stories?

I'm tired Sam

(Image credit: Shutterstock/EI Editorial)

Yes, ChatGPT is still down. In fact, I've added a handy update to the top of this page that I'll make sure to update when ChatGPT is back in full flow.

Until then, I'll keep live blogging, hoping for Sam Altman and co. to bring ChatGPT back to life.

I've not moved from my desk in nearly five hours, more so because I'm having a blast reading your emails, but I'm starting to feel tired...

Sam, please put us out of our misery; we need ChatGPT.

ChatGPT is back in Nevada

Big news from a reader in Nevada who wasn't able to use ChatGPT until now. They said, "It took a minute literally but it seems like it's back also it seems like it's aware that it was down which is nice. I can now finally go back to building my PC."

Anyone else back online? I appear to be getting responses but they are still incredibly slow.

Hold that thought...

Actually, hold that thought. ChatGPT is still down. I just got the response, "The request timed out."

Big, big sigh.

AI isn't up to Apple's standards

(Image credit: gizmochina)

Earlier, I asked for your opinion on Apple's WWDC event from yesterday, specifically in terms of AI.

Chase has been in touch regarding an AI research paper that Apple wrote earlier this month titled "The Illusion of Thinking."

He said, "The paper essentially boils down to one major point: AI is not actually intelligent. It’s not reasoning, it’s not thinking, and it’s not actually considering what you say to it. What it is doing is surfacing patterns. I personally feel this comes to a major point for Apple: Truth. Apple has had its own run-ins with bad data before (see Apple Maps original launch).

It follows that Apple would let other pioneers go through the ups and downs of their AI models fighting with people, gaslighting reporters, and leaving behind notes to future models on how not to get taken offline, because they can claim it’s a work in progress. Apple doesn’t want to ship works in progress, they want to ship a finished, polished iPhone-based assistant “that just works,” the magic that we have all come to expect from the era of Steve Jobs.

The problem, as the paper highlights, is no one is there yet. The best we have are very advanced parrots."

ChatGPT is back?

Reports from Egypt and Seattle, Washington that ChatGPT is now fully functioning again.

On my end, in the UK, I'm also noticing better response times and no more errors.

Is ChatGPT back online? We might be in luck!

ChatGPT working in Malaysia and Australia

Emails now flooding in from readers around the globe who are reporting that ChatGPT is working again.

Is ChatGPT working for you? Let me know via email: john-anthony.disotto@futurenet.com

Netherlands is back online

Good news from the Netherlands, Maus says, "I have stopped crying in the corner, GPT is working for me again too. Let us all rejoice."

Still down for many

While I'm getting lots of positive emails, there are a lot of readers still not able to access ChatGPT.

I also asked my colleagues who appear to be having similar issues, I worry ChatGPT is just giving us a false sense of hope...

Some ChatGPT alternatives to tide you over

(Image credit: Shutterstock/Adeel Ahmed photos)

My colleague Graham Barlow has written an excellent article on the three best alternatives to ChatGPT that are currently working.

I know many of think ChatGPT is the best AI chatbot on the market, but give Graham's article a read, you never know you might find a new favorite AI tool.

The outage continues

After my last post which said ChatGPT was fixed for many I've had an outpouring of emails that emphasize just how widespread this outage is.

Unfortunately, ChatGPT is not back and the official status page says, "We’re currently experiencing issues"

I started reporting this outage 5 hours ago, and it was already down for at least 2 prior to that. What a mess!

According to Downdetector it's getting better

(Image credit: Future)

The outage continues, but there are fewer reports on Downdetector than earlier today.

I think some people are getting back to using ChatGPT, while others wait for similar luck.

This is an absolute rollercoaster of emotions...

ChatGPT as a dating coach?

(Image credit: Shutterstock.com / Tero Vesalainen)

As the outage continues, I've had an interesting email from Krishna, a reader based in Texas.

She says, "I use ChatGPT as a “dating coach.” I use it to asses red and green flags based on profiles and conversations I have on Hinge. I also use it to ensure I’m properly responding in ways that most effectively find matches that have the highest long-term potential.

No, but seriously— I do. So, all my matches are waiting... They just don’t know that they’re waiting for my response only because ChatGPT is down."

If this isn't a reason for ChatGPT to hurry up and get back online, then I don't know what is. Krishna's love life relies on you, AI, please come back!

Some self promo (sorry)

(Image credit: Shutterstock)

While this ChatGPT outage has been chaotic to say the least, this live blog has captured the attention of a very, very large number of people.

If you're new here, the TechRadar AI team covers all things artificial intelligence, including ways to get the most out of ChatGPT (when it works).

If you're interested in keeping up to date with all my content and the coverage from my equally talented colleagues, be sure to check out our AI section or sign up for the TechRadar newsletter.

That's all from me folks

(Image credit: Warner Bros. Entertainment Inc.)

I've been covering this outage for 6 hours, and now it's time to stand up from my desk, stretch my legs, take my dog potty, and finally breathe some fresh air.

Unfortunately, it looks like the ChatGPT outage isn't ending any time soon, but it has been an absolute pleasure hearing your opinions and personal anecdotes.

I'm going to pass you over to my colleague, TechRadar's Senior AI Editor, Graham Barlow. You're in good hands.

If you want to get in touch with Graham, email him via graham.barlow@futurenet.com

J-A out.

...and it's hello from me!

Thanks to John-Anthony for doing a stellar job of covering the outage so far.

It's not easy when a once reliable AI companion is suddenly goes offline without warning, but the good news is that OpenAI say it has identified the problem with ChatGPT and a fix will be rolling out over the next few hours. It says:

"We are continuing to work on implementing the mitigation and we are now seeing recovery on API. Full recovery across all listed services may take another few hours. We will continue to provide updates as progress is made."

So, what are you going to be doing to occupy yourselves while ChatGPT slowly comes back online? Let me know how you've managed without the favorite AI in your life.

Lots of exam students are being let down

(Image credit: Shutterstock/EduLife Photos)

The ChatGPT outage is still causing havoc for A-level exam students in the UK. Alice writes, "I have my final Spanish A-level exam tomorrow. I'm studying extremely hard today because of the exam tomorrow, and normally I rely on ChatGPT to give me practice essay questions, and mark my work, as well as help me plan my study schedule and to test me on my knowledge of the book and film I've been studying in depth for the last two years."

Our heart goes out to all the students currently let down by ChatGPT, let's hope it gets back online in time for them to get some study in. Personally I've found Claude to be a good replacement to ChatGPT if you need help revising.

(If you don't know, A-Levels are British version of Advanced Placement courses in the US).

Pocket besties

(Image credit: Getty Images / svetikd)

Thanks for all the emails you've sent me so far. It's quite amazing reading how much ChatGPT means to you all. For example, I just got an email from Shelby, from Missouri who writes:

"I’m a daily AI user. Everything from recipes to the workings of the universe. However, my household is battling stomach bug right now, and I’m in the thick of it since 10 last night. ChatGPT has been walking me through it. Keeping track of when I last took zofran, helping with emotional support since throwing up is awful. I’m just waiting for my little pocket besties to come back online, I miss her. Thanks for giving me a place to share".

I think we all feel a little bit like that Shelby, without ChatGPT there's a big hole in our lives. Let's hope your best digital buddy is back online soon, and we hope you and the family feel better really soon!

When ChatGPT makes a comic about ChatGPT being down

Good news, ChatGPT does seem to be up for some people again, including us. So what's the first thing we thought we'd do with our restored AI powers? Why, ask ChatGPT to create a comic strip about ChatGPT being down, of course:

(Image credit: OpenAI)Handing over

It's incredible how quickly time flies when you're having fun, but my time looking after the blog is up! I'm handing over to my colleague Eric now and signing off. And if I'm right, I think he's got some good news for you about the status of ChatGPT...

Bye for now!
Graham

ChatGPT is coming back!

Hi everyone. The good news is that it looks like ChatGPT is coming back online, judging from the rapid decline in reports from Down Detector.

The outages have fallen by more than a thousand in the last hour, suggesting OpenAI has resolved the major issues for the AI chatbot and is adding back capacity as quickly as possible.

(Image credit: Screenshot from Down Detector)

There will likely be some lingering outage issues for a while as the company stamps out the original problem.

OpenAI will be under a lot of pressure from users and investors to explain what happened and to reassure people it won't happen again, or at least not for the same reason. Whether Sam Altman and his team can thread that public relations needle will be worth watching.

OpenAI's status is now only yellow

(Image credit: OpenAI)

Further proof that the ChatGPT problems are coming to an end comes from OpenAI's ChatGPT status page, which now shows a yellow 'We're currently experiencing issues' message, rather than the earlier red outage status.

Looks like we might be through the worst of it, folks.

Unlike ChatGPT, some Chinese AI chatbots went dark deliberately during exams

The ChatGPT outage was keenly felt by British students eager to use the AI chatbot during their A-level exams today, but students in China have already been deliberately cut off by multiple AI platforms.

As more than 13 million students sat down for the four-day gaokao university entrance exam, ByteDance's Doubao and DeepSeek stopped answering any questions at all.

Meanwhile, Tencent’s Yuanbao, Alibaba’s Qwen, and Moonshot's Kimi AI chatbots all disabled their image-recognition abilities during the hours of the exam.

There was no public announcement from any of the companies, but students shared screenshots and word-of-mouth reports on blank pages where their AI assistants usually wait. At least British students can turn to Claude or other alternatives.

Sora is back

You may not be able to have a conversation again with ChatGPT just yet, but you can at least go back to producing videos with Sora.

OpenAI's status report for the AI video generator has shifted into green, though ChatGPT remains stuck with "elevated error rates." The same goes for OpenAI's Playground for developers experimenting with its AI models.

(Image credit: OpenAI status screenshot)What about Apple?

One very important ChatGPT user who might be concerned about the outage is Apple.

Though Apple Intelligence was not quite the star of WWDC this week as some anticipated, the company still revealed new ChatGPT integrations in Image Playground, Visual Intelligence and Shortcuts.

You have to wonder if there are some nervous glances being exchanged in Cupertino over the ChatGPT stumbles today.

(Image credit: Future)

Happily for Apple, they haven't bet everything on OpenAI and ChatGPT, and users have a choice of AI sources like Google Gemini and Apple’s own on-device models.

Still, what seemed like the fast-track to AI status with ChatGPT is looking a little more wobbly. And as a reader pointed out earlier today, Apple isn't a fan of shipping in-progress products; they like to have the final polished version ready to go, if only to avoid Apple Maps-style debacles.

It's not quite over yet…

We're now more than 12 hours into the outage and ChatGPT's problems haven't been entirely solved still.

Down Detector reports dropped right down a couple of hours ago, but have risen slightly and are still hovering at around the 400-450 range. At their height they were at 2,000, so things have definitely improved, but clearly there are still some issues.

We'll continue to keep an eye on the situation and will report back as soon as things change.

Disinformation spreading on social media platforms has stoked an already tense situation.

The new lab, set to include Scale AI founder Alexandr Wang, is part of a reorganization of Meta’s artificial intelligence efforts under Mark Zuckerberg.