By now you’ve probably heard the promise of iOS 27: it’s a Snow Leopard-like year where Apple spent a lot of time not on big marquee features, but on smaller fixes and enhancements throughout the operating system. And that’s largely true—and largely to the benefit of the platform—but it’s also not the whole story. Because there is, of course, at least one major feature that comes to iOS 27 (and Apple’s other platforms) this year, and it’s a doozy: Siri AI.

The result is an OS that feels like a marked improvement over its predecessor, in everything from design to performance to capabilities. That’s not to say there isn’t still room for improvement, but this is the beta period, and improvements may be forthcoming.

Still, the promise of iOS 27 is that it’ll not only unlock new possibilities but also make your existing workflows smoother and quicker. That’s not a bad proposition for the device that many of us carry more than any other—as long as it can deliver.

As of today, iOS 27 is available to any and all who want to try it as a public beta. Some features are buggy, others aren’t yet finished, and every once in a while something may go haywire. But it’s all in the name of getting everything working for the final release in the fall.

Siri & AI: Two okay tastes that taste great together?

The call for Apple to do a major overhaul of Siri has existed almost as long as Siri as existed. Heck, I wrote a story about it back in 2015, more than a decade ago and only four years after the first version of Siri debuted. But even as competitors emerged and Siri moved to more Apple platforms, the voice assistant improved only in relatively small, targeted ways—a testament, perhaps, to the limitations of its underlying technology. With the recent rise of chatbots backed by large language models, it quickly became apparent how much Siri had fallen behind, and so drastic measures were necessary to bring it up to speed.

The good news is that Siri AI is capable: impressively so, at times. In the lowest of low bars, it certainly beats the pants off the previous incarnation of Siri, but its access to your own personal data and all of the app on your phone also unlocks a host of new possibilities.

But it’s also not without the pitfalls that underlie all LLM-powered chatbots: errors in understanding, hallucinations, and so on. Perhaps the biggest challenge of Siri AI will be getting people who have written Siri off as a limited feature, only to be used for trusted tasks—setting timers, checking the weather, playing music—to be more ambitious about what they ask the assistant.

New Siri is integrated throughout iOS, whether you access it via the traditional press-and-hold of the side button or by saying the wake word, or in its new home integrated with the Spotlight search when you swipe down from the middle of the top edge of the screen, a gesture that’s now available not just on the home screen but in any app. It’s got a new look too: a black liquid glass bubble with silver overtones. You can ask Siri questions by voice, or by simply typing to it.

There’s also a new standalone Siri app that stores all your conversations (with some exceptions). The app itself is pretty barebones, letting you view those conversations in List or Grid layouts, select multiple conversations to delete them, and pin or rename any single conversation. In Settings : Siri, you can choose to either open to the overview or to a new conversation and decide whether old conversations are kept for 30 days, 1 year, or forever. It’s a little strange that one-off queries (“What’s the average height of the Norwegian World Cup team?”) are stored alongside more involved conversations (“Can you help me build a specific RSS feed for this site?”); I also ended up with a few orphan blank “New conversations” which I deleted, but it’d be nice if you could filter to find those. Conversations are fully searchable and sync between devices via iCloud.

I’ve had a couple of real “wow” moments with the feature in the month-plus that I’ve spent with the beta. For example, a couple of weeks ago I was on a rare long solo car trip when I realized that I wanted to listen to a specific episode of a podcast. This wasn’t a podcast that I listen to regularly or even subscribe to, and to make matters more challenging, it was an older episode. Hard enough to find in the Podcasts app when you’re not driving down the highway, but something I would never have even attempted with old Siri. But hey, now’s the time to give it a shot, so I asked Siri AI, via CarPlay, to play the episode of “The Rest is History” about the American Revolution.

After a moment of chugging—like most of those AIs, the process isn’t exactly fast—Siri came back to report that it had found one episode of the podcast about the American Revolution. “Play that,” I said…and it did. No need to specify the precise title of the episode, which I would not have remembered from earlier in our conversation.

Moments like that make me feel like a Star-Trek-computer-like voice interface is here now. In large part because it does leverage the great strength of Apple’s AI implementation (and I’m not going to spend the time dithering about models and distillation and all that jazz, because the question to me is how do people use it and how does it work). But that strength is that the AI isn’t siloed off in its own website or app, but on your device with all the information that includes.

The same went for when I asked it about the date for a pool party my wife had texted me about, and who would be there. (It even correctly identified an attendee with an ambiguous name based on proximity to their partner’s name.) And when I was viewing a webpage with an upcoming flight that I asked it to add to my calendar. It was also able to identify a movie still in a brochure for a local movie theater, give me the performance of the Red Sox over their last ten games¹, and help me figure out which model of IKEA curtain rods I was dealing with. None of them things that I would have even thought about using Siri for prior to this incarnation.

Unfortunately, not every moment is like that. Like any LLM-based chatbot, Siri AI sometimes misunderstands or gives confidently wrong answers. For example, when I asked it about when my neighbor said they’d be back from vacation, it found the relevant text message but said that no return date had been mentioned, despite the fact that it was right there in the message.

Of course, some of this is easily ascribable to this being beta software. But some of it also the new normal for dealing with computers in 2026. Is that better or worse than having old Siri’s frequent habit of shrugging at things it simply couldn’t do? I’m not sure that’s an easy qualitative call. In the end, it’s a mode shift: a different experience of how we interact with our voice assistants and chatbots, and we’ll have to see how people adapt.

All of that said, one thing I do appreciate about Siri AI’s implementation is that Apple has done a good job of, for the most part, making sure that the traditional things that Siri has always been good at—setting timers, playing media, adding reminders, controlling smart home accessories, and so on—are no worse than under old Siri. From what I can tell, Siri is mediating these requests and handling those things locally rather than handing them off to the cloud. That also means not every request to turn on lights or play your podcast gets logged into the Siri app as a separate conversation, which is I think the right approach: I certainly don’t need a dozen conversations where I’m setting my five-minute tea timer.

In general, I’m bullish on the future of AI, and fascinated to see what happens as it rolls out to more and more users. Will it be able to sway the habits of those who’ve been using Claude and ChatGPT? Perhaps not for advanced uses like coding, but for your run of the mill world knowledge queries? That seems eminently plausible. Especially because it’s already there, built into your phone.

One caveat, however. As always, the future is not evenly distributed: on the iPhone side, Siri AI’s only available on the iPhone 15 Pro and Pro Max, as well as the iPhone 16 or later. And, of course, due to Apple’s ongoing disagreement with the European Union, it won’t be available on that continent either.

All roads lead to Siri

While Apple Intelligence is still apparently a going concern, much of that brand has been subsumed by the “Siri” identity in iOS 27. For example, any time you put a blank cursor in a text field, you’re prompted with a Write with Siri button hovering aboard the keyboard. (Once you start typing, you get the usual word suggestions, and Siri is reduced to just an icon.) As someone who is perfectly happy writing their own text messages and emails, this prompt sometimes feels a little bit aggressive, and I wish you could disable it without turning off all of Siri.

There’s also a Siri mode in Camera, which replaces the previous Visual Intelligence interface. Now it’s just something you swipe to, like Video or Photos. In this mode, the Siri icon replaces the shutter button; snap a picture with that and it’ll attempt to tell you what you’re looking at; there are also image search and ask buttons which, respectively, query a search engine for image results and attach an image of what you’re looking at to a more targeted Siri query. But this mode essentially is just a front end for the Siri app—you can attach an image in a new conversation there too, including snapping one with the camera.

Liquid Glass half full

To call iOS 26’s Liquid Glass redesign contentious is an understatement on the proportion of saying that Lionel Messi is good at kicking a ball. In iOS 27, Apple has made two major moves: first, tweaking the overall look and feel of Liquid Glass, and second, providing a way to further fine-tune that look yourself.

In general, the iOS 27 incarnation of Liquid Glass is far easier on the eyes, dialing down the transparency in favor of legibility. Apple’s also imposed some distinction between user interface elements and content. For example, in list views like in Messages and Mail there are now much clearer “toolbar” areas at the top when you start scrolling, even if they still have some degree of translucency. App icons have also gotten a face lift; they’re sharper, and feel more vibrant.

In case you still aren’t happy with the default look, Apple also offers an option for adjusting the glass effect without resorting to the classic accessibility tweaks that can often have unintended design consequences. Settings: Appearance: Liquid Glass offers a slider that goes from More Clear to More Tinted, with a stop at the Default or anywhere in between. You can see right in the scrollable pane above the slider how your choice will look in a live interface, which is a nice touch.

It’s a rare concession from Apple that there is no one true way that the UI should appear, but given that the company has in recent years allowed far more interface choices (for example, Mail, Safari, and Phone all offering toggles between older styles and newer interfaces), this feels like par for the course. While the most stringent Liquid Glass detractors will likely never be satisfied, the improvements and ability to tweak further thus far feel as though they will satisfy most users.

Performance, no anxiety

Tempting as it might be to spend this entire preview piece talking about Siri AI, there are indeed other things in iOS 27 that are worth your time. Exactly which improvements they are depends on how you use your phone.

That’s because one of the odd things about looking at iOS 27 is that there’s just a ton of small stuff—those “quality of life” improvements that chip away at little frustrations and annoyances. Being able to send a text message while a photo or video attachment is still sending. Faster switching from Wi-Fi to cellular when you walk out of range of your house. The ability to change volume on your alarms independently from system volume.

All of these are just nice enhancements, sanding off rough edges and just generally making things work better. But because they’re so niche, lots of people are going to have different experiences with iOS 27, because it depends on what’s important to them. For example, is anybody going to be as excited as I am that Shortcuts now offers an “else if” option? No, probably not! But that’s okay, because I’m happy enough for all of us.

The good news is there are some places where things are just better across the board. In general, iOS 27 feels peppy. I know, we always talk about snappiness, but given the time Apple has spent on improving performance and removing bugs with this release, it’s reassuring to see that it’s borne out by my experience. Apps launch quickly, photos sync faster via iCloud, browsing my NAS via the Files app was downright speedy compared to earlier versions.

AirDrop transfers were also notably rapid, though I say was because in recent beta builds my iPhone has stopped sending files via that method, though it still receives them. (That’s betas, folks!)

Apple also touted improvements to search, particularly in Mail, an app that I use daily, which has always been a bit lackluster in that regard, especially when compared with Gmail. I do find it improved in iOS 27, even though I do have an inbox of more than 144,000 messages going back nearly 25 years. I have mostly been able to locate emails, even obscure and old ones, within a relatively short amount of time, which is definitely a nice improvement.

I’m most intrigued to see how this affects the experience of those using older iPhone models. Apple says that it’s brought its optimized CPU scheduler to devices all the way back to the iPhone 11, which ought to noticeably improve performance even for those who don’t have one of the latest and greatest iPhones, but the public beta experience will certainly bear out the truth of that claim.

Photo fit and finish

If there’s one app that always seems to get some love from Apple in its yearly updates, it’s Photos. No surprise there: our cameras document everything about our lives and we take pictures constantly.

The filter menu now lets you choose between sorting by when photos were added versus when photos were captured, as well as providing quick access to photos that were captured by you (as opposed to photos taken by others that you imported). I’m not exactly sure what the rubric there is—when I used that filter, it only showed me images back to 2017, when I have surely been capturing photos—and even just iPhone photos—for a decade before that.

There are some big improvements, particularly in Shared Albums, which now let you share pictures as full resolution, quickly download all photos from an album, and even respond with emoji other than just thumbs up. Keep in mind, though, that all members of a Shared Album will need to be upgraded to iOS/iPadOS/macOS 27 in order for anyone to use the new features.

Ad hoc slideshows have gotten improvements too. When you select a bunch of photos and play them as a slideshow, you can adjust timing and duration as well choose the transition.

For those who missed being able to assign star ratings and keywords to their pictures (shades of iPhoto!), good news, they’re both back. Keywords are accessible by swiping down in the detail view, whereas star ratings need to be specifically turned on in Settings: Photos: View Options, where you’ll find a Show Rating Controls options. To assign a rating, you’ll need to look for a tiny star icon at the top right while looking at a photo (if you have the spatialize feature on, it’s next to that).

There are some smaller niceties too. For example, you can now save out a frame of a video as an image, rather than jumping through hoops of screenshotting or exporting it to a Mac. And you can adjust the metadata on photos both individually and in bulk, if you need to correct the date, time, or location.

And, although Apple has touted that photos will upload to iCloud faster, there’s a new option for those of us for whom that still might not be fast enough. In Settings: Photos: Sync Immediately, you can tell Photos to prioritize syncing new photos over conserving battery life—for the day. Not unlike the AirDrop “Everyone for 10 minutes” option, it’ll turn the feature back off after the day is over. (In a nice touch, when you turn it on, it shows a counter of how many hours are left before it deactivates.)

But of course, Photos isn’t immune to the expansion of AI, either. Previously we got Clean Up, the tool that lets you remove unwanted elements from your photos. Apple says it’s improved that in iOS 27: you can now choose between Fast and High Quality modes, the latter slower but more powerful.

There are also a pair of new AI features: Extend and Reframe. The former is easier to understand: if you want to zoom out of a photo, whether because you’re trying to fit in a particular shape or simply to tweak the framing of a shot, Extend can generate a little more background, extrapolating from what’s already there. It works best on smaller adjustments with backgrounds that are uniform or predictable. Go too far and you’ll end up with totally invented elements, which can look strange or out of place.

Reframe is a bit harder to grok. Think of it as being able to move the camera after the picture’s been taken. Using the same features the company leverages for creating spatial scenes, it tries to separate out the foreground subjects and the background, and then adjust the perspective. Some pictures work better than others; it seems to help if there’s a clear delineation between foreground and background. Like Extend, it tends to work best for smaller adjustments; go too far and you can end up with some odd-looking artifacts. As tools, they have their utility, though it’s perhaps best used sparingly.

Other features

Apple’s made a big push on child safety across its product lines for the 27 updates. Some small improvements ought to really make the process more usable, such as the ability to convert an Apple account to a child account, default settings for time limits developed in conjunction with the American Academy of Pediatrics, and a much-requested option to lock the Screen Time section of Settings on a device. Taking advantage of some of these features will require you to update all the compatible devices in your family, but you can still use the on-device improvements.

Image Playground now has improved image generation and the ability to generate photo realistic images. While some of those results are better, many of them are…unsettling. Even after a couple years, I personally find this feature misguided and off-putting.

Passwords has an agentic AI feature that will scan your list of compromised passwords and then present you with a list of the ones it believes it can update for you. Even on that subset though, my experience was uneven, with just 5 of 18 updated. It’s a very cool idea, and should hopefully see improvements as the technology progresses, but right now it’s hit or miss.

If you’ve ever been on a customer service phone call and needed to look up salient details like a flight confirmation number, the Phone app will now surface relevant info during the call. I haven’t been able to test this yet, but it seems to be a real potential timesaver. And fun fact: if you call someone on their birthday (assuming you have it in Calendar or Contacts), you’ll get a little animation and note about it.

In a “better late than never” improvement, Calendar now has natural language input when creating new events. Third party apps like Fantastical and Google Calendar have done this for years. Unfortunately, my experience with this was a little disappointing, with it not correctly assigning the event to the right day and ignoring my attempt to provide timezone information; nor did modifying the event with natural language after the fact seem to work. I hope that gets ironed out before the end of the beta period.

For those who love endlessly looking at 3D imagery of cities, Maps gets an improved flyover view with better 3D modeling of buildings and landmarks. In my testing, this isn’t fully rolled out everywhere yet: Boston’s satellite maps were actually two-dimensional compared to iOS 26’s, but if I flew down to New York, I could see far more detail on a building like One World Trade Center.

FaceTime now lets you show the person you’re talking to your front and back cameras simultaneously. No more endlessly flipping back and forth!

Apple’s improved Shortcuts with a bunch of new actions and streamlined automations by including them as part of the shortcut itself. There’s also an impressive new AI “describe your shortcut” mode. It works pretty well for simpler shortcuts, but can start to get a little off the rails the more complex you get. Unfortunately, Shortcuts has been among the buggiest features during the beta period; I’ve had a lot of them stop working, and run into some perplexing issues.

Safari likewise has a few new features, including being notified when a page content changes (as frequently as on the hour, if you’re watching for price changes, for example), as well as a “describe your extension” feature not unlike the Shortcuts functionality. I haven’t had as much of a chance to put this through its paces, but it’s a fascinating addition and will be interesting to see if lots of people who never would have tried their hand at this kind of thing will end up making extensions.

You can also now put an extra large widget—i.e. one that takes up an entire screen—on your home screen. I tried this with Weather for a while and it was…interesting, but ultimately not for me. Still, it feels like the age of the home screen that’s just app icons is slowly dying off.

Conclusion

Despite being put forth as a year of small updates, there’s quite a lot in iOS 27, even without including the massive overhaul that is Siri AI. Apple had found itself on its back foot after its first attempt to integrate Apple Intelligence, but my experience is that Siri AI largely delivers on what the company promised—albeit a couple of years later than intended.

But I’m also very bullish on the enhancements and quality of life updates throughout the operating system. The improvements may be small, but they add up. Every time you don’t have to go searching for an email while you’re on a call or a photo shows up faster in your library or you find exactly what you’re looking for in search, well, I’m not going to say an angel gets its wings, but you get less frustrated. And that’s not nothing. In fact, that’s everything. The goal of technology should be to work for us and make our lives easier.

As for the shortcomings, it’s worth reminding ourselves that this is a beta period. Proceed with caution. Sometimes things just don’t work, and you may have to live with it until the next build comes along. But sometimes things also improve build-to-build, and that’s the fun of being at the cutting edge. For the next couple months you get your chance to live in the future.

Let’s just get this out of the way: I liked macOS 26 Tahoe. I liked it because it marked the first time in years where Apple seemed to embrace the fact that Mac users deserve productivity improvements, not just syncing up with whatever (often useful, yes) new features were being introduced to iOS.

I didn’t like the interface changes, especially the way the redesign handled toolbars and sidebars. It actually felt like macOS dodged a bullet by being treated as an afterthought, because pouring more Liquid Glass atop macOS would have made the situation even worse. But over the last summer, I got used to the interface offenses and instead focused on all the improvements.

Turns out that at least some of my friends and colleagues decided that Tahoe’s interface offenses were too great to bear, and despite all those tasty new productivity boosts, they would sit this one out and hope that Apple came to its senses with macOS 27.

Good news for them: While they’ve spent a year without Apple’s first native clipboard history, Shortcuts automation and AI access, and an improved Spotlight, their upgrade to macOS 27 Golden Gate will net them last year’s features as well as a host of other improvements, all wrapped in an interface that is an unequivocal improvement over what Tahoe wrought.

To those folks, welcome back to the current OS version. To those of us who spent a year floating on the glassy surface of Tahoe, welcome to Golden Gate. Available today as a Public Beta and arriving on every Apple silicon-based Mac this fall, it’s an update I’m not going to have to spend a year apologizing for liking.

Design takes a step forward

The fact that there are more design changes in macOS than any of Apple’s other 27 updates speaks volumes about how broken many of the Tahoe design decisions were. Apple describes the design as “even more refined,” because in Apple marketing-speak, broken features are never remedied or fixed—past features are just improved. The truth is, Apple has rolled back numerous missteps and made changes to others to mitigate the issues they caused.

Floating glass sidebars inset from window edges—a choice that made sense metaphorically, but looked awful—are gone, essentially reverted to their previous design, sitting at the left edge of windows, with a darker background for contrast and colored icons to boot.

Toolbars, one of the most important interface conventions of macOS, were flattened and ghostly in Tahoe. In Golden Gate, they pick up better shading and are more legible, though (strangely) the toolbar in inactive windows offers more contrast and legibility than in active ones! In Photos, toolbars are actual bars, not horizontal areas of floating glassy buttons—which means the system no longer has to attempt to add shading behind those areas in order to improve legibility. They’re just legible.

In Golden Gate, windows generally just feel… a bit more square. Apple has made window corners more consistent, and not having the curved corners of the sidebar right inside the curved corners of the window edge also helps make it feel a little less like you’re under the tyrannical rule of the Circle King.

Icons look better across the board in Golden Gate, probably because of a new layered icon compositing approach Apple is taking across its platforms. Unfortunately, apps that dare stick any of their extremities out of the mandated squircle shape are still put in a featureless gray box of shame, which makes it harder to identify icons, especially in appearance modes like Clear and Tinted that remove color as an identifier.

To hear Apple tell it, users are split on the transparency effect of Liquid Glass: some people despise it, and some people can’t get enough of it. As a result, all of Apple’s OS 27 updates introduce a slider—it’s in the Appearances pane of Settings on macOS—that lets you customize the transparency from sort-of-mostly-transparent to sort-of-mostly-opaque. Given that this update rolls back a lot of Apple’s decisions from a year ago, I’m inclined to believe that there really are two camps when it comes to Liquid Glass. Giving users an option to adjust the setting is probably fine, though I’ve currently got the slider right in the center. We’ll see how that goes.

Siri AI takes the Spotlight

Siri has been a part of the Mac experience for a decade now, arriving on the platform five years after it debuted on the iPhone. While I’m sure that many Mac users have embraced it over the years, it has always been an afterthought at best for me. When I’m sitting at my Mac, I’ve got a keyboard and trackpad to get my job done, so I don’t generally feel the need to talk to my computer. And while you’ve been able to type to Siri for a while now… well, the other problem with Siri has always been that Siri’s results are rarely as useful as the results of typing those words into a search engine or some other tool.

In macOS Golden Gate, as with all the other OS 27-era releases, Apple has upgraded Siri to become Siri AI, with LLM-derived features that promise much better results. On top of that, Apple has finally done what it should have done long ago and merged Siri and Spotlight, so there are no longer two separate “type to find things” interfaces on the platform.

The result is… still a work in progress, at least so far. Siri AI is a beta feature inside a beta operating system. Betas within betas. First off, it’s a struggle to remember that Siri is actually useful and I should take advantage of it. Years of training myself never to consider Siri have to be untrained, and that’s going to take some time.

Thus far, I’ve found Siri AI’s responses to general-knowledge queries to be very good. But part of the appeal of Siri AI is its integration with your personal on-device data. This is where things start to fall apart, though this is not surprising—it’s an early beta, and many third-party apps will need to update to better integrate with this system. I don’t use Apple’s default calendar or email apps, which puts me at a disadvantage. It frequently failed to answer fairly basic questions about files on my system—”where’s that PDF that contains my company’s tax ID number?”—that I’d hope that a smart assistant with access to my entire Mac should be able to find.

There’s also the question of Siri/Spotlight choosing the right context. When you type Command-Space, Spotlight comes to the fore—and that’s good, because if you want to use it as an app launcher, you want instant responsiveness. When I typed a question that’s more appropriate for Siri, however, I found I needed to wait a few seconds for “Ask Siri” to appear so I could ask my query.

There’s a new Ask Siri entry that’s been added to the top of every contextual menu, and while I don’t mind the idea that Siri can act on just about any context you can think of on the Mac, I don’t love how it’s implemented. The contextual menu used to open with all of its contents down and to the right of the pointer (unless you were too far to the right or bottom of the screen); now it starts shifted down a little bit, as if the old context menu was being drawn and then an Ask Siri item was being overlaid right at the top of the menu, sticking up over it. It’s darker than the rest of the menu. If you click on it, it floats forward, the rest of the menu disappears, and you can type in the resulting box. It’s disconcerting every time I right-click on something.

The most useful actions tend to be placed at the top, and I’m not convinced that even a fully functional Ask Siri would always be the most useful command in a context. This feels more like a weird add-on to existing context menus, and all to promote the new Siri features. Apple, if a feature is good, users will use it. Distorting the geography of context menus on the Mac to promote them feels aggressive and almost desperate.

Touchscreen Macs, at last?

One of the most remarkable features of macOS Golden Gate does not really make sense if you take it at face value: Apple has given a major upgrade to Sidecar, the feature introduced in 2019 that allows Macs to use iPads as external Mac displays. As someone who always travels with a 13-inch iPad Pro, that’s a nice feature, though since the 2022 introduction of Universal Control, which lets me use my Mac’s keyboard and mouse on an iPad still running iPadOS, I’ve found very little use for Sidecar.

The big update is that Sidecar, which previously only supported the Apple Pencil, now works with the iPad’s touchscreen. Apple has updated its own apps and supplied third-party apps with the appropriate tools to support full touchscreen support. You can now scroll Mac browser windows with a finger and drag windows around on screen… but only on that iPad you’re using as an extra display.

It’s not hard to read between the lines: As numerous reports have suggested, Apple is planning on releasing its first touch-enabled MacBooks late this year or early next year. Using Sidecar as a way to demo the technology and allow developers to test their apps in advance is really clever.

Beyond that, this feature is delightfully boring. It just does more or less what you’d expect. Adding touch to macOS isn’t going to be revolutionary—it’s been on Windows for ages, and iPads have had optional trackpads and keyboards for six years now. But it’ll be a nice incremental benefit. I still use touch gestures when I’m working on an iPad in laptop configuration, because sometimes it just makes ergonomic sense. If my hands aren’t on the keyboard and I need to scroll something or tap something, a quick poke with my finger does the job. It feels natural—and once you’ve spent some time with that configuration, it’s jarring to use a Mac laptop and realize that you’re poking the screen to no effect.

After a moment’s glee in being able to scroll a Finder window with my finger, using Sidecar with touch to drive the Mac interface just became… normal. Apple has done a good job implementing the feature. Menus drop down reliably. Buttons are tappable. Scrolling works. There’s just not a lot to report, and that’s just fine. It works.

Shortcuts, reduced and enhanced

The Shortcuts app, the cornerstone of Apple’s cross-platform automation technologies, gets a notable upgrade during this cycle. Most of it is probably due to the needs of Describe a Shortcut, a remarkable new feature that lets you input some text and then watch as your device converts it into a Shortcut. As I wrote in June:

This is as close as we’ve come, across 40 years, to the original dream of putting computer power in the hands of everyone. You can literally tell your device what you want it to do, and when—”every morning show me my to-dos and calendar events for the day”—and it will generate a program to do it and a schedule to run it…

But when it works, which is most of the time, it’s magical. And perhaps its best feature is that you can iterate on your Shortcuts. If there’s something it doesn’t do quite right the first time, you can specify changes you’d like to see—”only show me items from my Personal calendar”—and it’ll rewrite the Shortcut to take those changes into account. In that way, it’s emulating the back-and-forth conversations that can make vibe coding so remarkable.

Describe a Shortcut is limited, but it really works quite well. I hope it continues to grow and improve, because it lowers the bar for user automation across Apple’s platforms in a delightful way.

For those of us still building Shortcuts with our own fingers like animals, some of the other improvements will help that work get done. There’s a new Otherwise If statement that can be added to If-Else blocks (you get to it by control-clicking on the Else block and then choosing Add “Otherwise If”), making the structuring of branching shortcuts logic much, much simpler.

There’s also support for persistent data storage, which is kind of huge. You can save data, which syncs via iCloud and is accessible across your devices, and Shortcuts can both access it and update it! That will make Shortcuts capable of behaving more like other apps, all without having to rig up other ways of storing data in iCloud Drive or elsewhere. (Yes, this canonizes Simon Støvring’s Data Jar in Apple’s OS, and that’s a good thing.)

The arrival of Describe a Shortcut has unfortunately distorted the Shortcuts interface in some unfortunate ways. On the Mac, there’s a new three-paned interface, where the third pane is a giant preview of a shortcut’s color, icon, and (this is fun) an AI-generated summary of what the shortcut does. The pane exists because of the text box at the bottom, which lets you type input to Describe a Shortcut to make changes to whatever shortcut you’ve got selected. Which is fine, except it’s an enormous amount of screen space devoted to something that may be entirely irrelevant. Somehow, Shortcuts won’t let you hide this pane.

New horizons for Photos

During WWDC week, I wrote about the changes to Photos this year, but just to reprise:

Apple has finally updated its Shared Albums feature, which had been compromised since it launched because it didn’t offer the ability to share full-resolution images. Over the years, Apple added other methods of sharing groups of photos via iCloud, and those could include full-resolution images, but this one prominent feature felt stuck in the past.

Now it’s getting a proper upgrade, with support for full-resolution images and allowing for full collaboration with people on other platforms so that everyone can contribute to a shared photo album. I’m relieved that I will soon have to stop explaining the differences among the various ways to share items in Photos and warning people away from Shared Albums.

As for the three AI-powered features in Photos, they’re a mixed bag. I have high hopes for a much improved Clean Up, which was already okay but could be a lot better. The new version is much more adept at artfully clipping unwanted items out of an image and filling those areas with in-context imagery, though of course it’s not perfect.

The other two features, Extend and Spatial Reframe, require the use of an advanced diffusion model that’s only available via Private Cloud Compute, and as a result, they take time to execute, since Photos will need to upload your photo, wait for a result, and then download the result.

Extend feels like a good feature, since there are plenty of scenarios where your image needs just a little more headroom or width. It’s also going to be great for straightening images, since Extend can fill in the slivers of unknown image that are exposed when you rotate, which otherwise require that your image be cropped as you rotate.

However, every pixel you expand the selection increases the jeopardy that what’s going to be generated is weird or fake. Everyone’s mileage may vary, but I found that I was much more comfortable expanding a photo a little bit to gain some headroom than doing it a lot, forcing the AI system to invent more objects or scenes. I was occasionally frustrated that I couldn’t slightly extend the bottom of some images, where subjects’ legs or feet were cut off. Also, you have to be aware of incredible fabrications—extending a graduation photo added a gigantic floating tassel and an invented piece of furniture, and extending a stadium shot added a roof where there isn’t one. Judicious use, carefully analyzed, is my advice.

Then there’s Spatial Reframe, which brings together a load of existing Apple technologies, including the spatial scanning algorithm it used to create spatial photos on Vision Pro. This feature works by scanning your photo locally using that algorithm, inferring a depth map that is then used to build a 3-D version of the image that you can pivot a bit, up and down, and left and right. This is the effect that allows you, on the Vision Pro, to feel like you can move your head and see parallaxes shifting, even though, if you look closely, the exposed content behind a subject is just a simple generative fill. It all happens so quickly, and in service of a live 3-D effect, that it’s often not noticeable, and even when it is, it’s not that big a deal.

The bar is a lot higher for a fixed, 2-D photo at full resolution. So after you use Spatial Reframe to slightly move the perspective of a shot, all the data is sent up to Private Cloud Compute, where a new version of your shot is rendered—including much more advanced generation of all of those pixels that are revealed by parallax or at the edges of the frame.

Some results are more synthetic than others. Unlike Extend, Spatial Reframe changes the entire perspective of the picture, which requires everything visible to be re-rendered at full quality. The result sometimes feels artificial. Again, my first impression is for people to use this feature with care and caution.

There are a handful of other nice, new Photos features, including new Captured By Me and Identity Documents sections (quick, find that photo of my passport!), and a triumphant return of star ratings for individual photos.

Wait, there’s more…

There are so many small enhancements in Golden Gate that it’s hard to mention them all, and every time I turn around, I keep bumping into more of them. Keeping in mind that we’re entering the long cruise to the fall release of the final version, here are some other highlights so far:

• Apple has added some light menu-bar management, specifically to deal with items that overflow the notch on some MacBook models. I’m glad that it’s taken this tentative step, though I wish it were more—it’s okay, Apple, you can sherlock all those menu-bar-management utilities out there¹—and of course, it should’ve happened back when the notch first appeared (checks notes) nearly five years ago. Better late than never, I guess.

• Safari’s now got an AI-driven tab organizer, so if you’re someone who keeps a zillion tabs open, you can have Safari group them into smaller, more navigable clusters. I’m not a zillion-tabs person, but I noticed that this approach is also being applied to the Start Page, where recently viewed pages also now appear clustered by topic.

• The Passwords app gets an injection of agentic AI through a new feature that offers to change your compromised or easy-to-guess passwords for you, navigating various websites and generating new, high-security passwords automatically. It’s a clever idea, but the most unfinished of any feature in the beta. It managed to change three of the 39 passwords it attempted. If you were trying to break into my Ikea account, well, you’re going to need to order more Kallax shelves on someone else’s dime now.

• Both Finder and the document save option in the menu bar will scan the contents of a file and offer up naming suggestions. I didn’t love the suggestions it offered, but your mileage may vary. I’d like to see this feature applied to auto-generated filenames like screenshots, though!

A note about stability

This summer, the vibe has been that Apple’s 27 releases are very stable for betas, which is not surprising since improved stability and reliability have been key Apple goals. And I would agree, I’ve been running the developer beta on my main Mac since the week after WWDC, and it’s been mostly fine.

But don’t be fooled: These are betas, and they will do weird stuff. I have to occasionally reboot my Mac to get things working normally. Shortcuts just stop working, and then start working again, willy-nilly. Third-party apps frequently quit unexpectedly, or refuse to quit at all and hang out in my Dock forever! And as I mentioned above, my menu bar manager is busted, though the developers are working on it.

If you’re willing to put up with the quirks in exchange for a glimpse at the future of the Mac, jump in. If you refused Tahoe entirely, you are in for an extra treat. And if you’d rather just wait… relax. Fall will be here before you know it.

Since the very beginning, it’s been clear that computers provide incredible power to those who know how to use them to get work done. The challenge has always been how small the group of “those who know how to use them” has been.

The Apple II included support for BASIC, a simplified programming language that was intended to let new computer users write programs. “In my opinion, the real thing [the Apple II] is doing right now is to teach people how to program the computer,” a young Steve Jobs told the New Yorker in 1977. The first computer programs I wrote were in BASIC.

There was also the Logo language, which helped introduce beginners to programming concepts through the manipulation of a virtual robot called a “turtle.” In the ’80s, HyperCard tried to broaden the programming community through its use of HyperTalk, another language designed to speak to beginners.

The (still!) current examples of this on the Mac are AppleScript, which used a format based on English-language sentence structure; and Automator and Shortcuts, two automation tools designed to create flow-chart-based programs.

The goal is always the same: To give regular people the ability to harness the power of computers. And we’ve never, ever been closer to the goal than we are today.

No, I’m not talking about vibe coding apps, mostly because that requires a level of focus and detail that most “regular people” are not going to want to provide. Show a civilian Xcode and watch how their eyes instantly glaze over.

I’m talking about Shortcuts—specifically, Apple’s Describe a Shortcut feature in the forthcoming macOS 27, iOS 27, and iPadOS 27.

This is as close as we’ve come, across 40 years, to the original dream of putting computer power in the hands of everyone. You can literally tell your device what you want it to do, and when—”every morning show me my to-dos and calendar events for the day”—and it will generate a program to do it and a schedule to run it.

Yes, there are lots of limitations. Describe a Shortcut doesn’t work with third-party apps, only Apple’s own stuff (at least, for now). It can sometimes get confused, especially with complex queries. And it has an interesting tendency to kick things to the Use Model action—why am I not surprised that an AI model likes to build Shortcuts that themselves use AI models?

But when it works, which is most of the time, it’s magical. And perhaps its best feature is that you can iterate on your Shortcuts. If there’s something it doesn’t do quite right the first time, you can specify changes you’d like to see—”only show me items from my Personal calendar”—and it’ll rewrite the Shortcut to take those changes into account. In that way, it’s emulating the back-and-forth conversations that can make vibe coding so remarkable.

I was also impressed with the fact that Shortcuts created with this method are… just Shortcuts. With one click, you’re looking at the actual Shortcut blocks that the feature has assembled for you. You can edit them as you see fit. They’re not special in any way, other than that they were built with an actual English-language sentence, not a line of code that resembles one superficially.

I’m sure that future developments will make it even easier for us to tell our devices what to do. But this particular advance feels especially big to me, like we’ve crossed some invisible line of demarcation. We are entering the era where the computers program themselves—and, for the most part, are able to understand what we’re asking them to do.

Adam Engst at TidBITS took the time to extract every item on the WWDC 2026 keynote slide that captured our attention last week. Even better, Adam grouped the changes into operating systems, apps, and other categories. While the 27 OSes dominate the list with improvements and revisions, you can also note how many changes Apple is making to Photos and Shared Albums, Find My, Music, and a number of other across-the-ecosystem tools.

Then Marcin Wichary used Adam’s post to slice the data in a different way, counting the number of times the slide used words like fast (59 times) and reliable (22 times). Marcin reasonably argues that speed is the easier problem to attack, while reliability is fuzzier and much harder to pin down.

Go to the linked site.

Read on Six Colors.