The Keynote

I loved the keynote’s “You want a great opening scene?” opening, which ran through a series of genre-inspired hook-you-from-the-first-shot pretend movie openings. Apple started with a title card that said “The following was shot on iPhone”, but the whole keynote — including the fun fake-movie openings — was shot on iPhone 18 Pros. Apple started shooting its keynote movies using iPhone Pros in October 2023, with the “Scary Fast” Mac event. That first one was shot on the then-just-released iPhone 15 Pro, the first iPhone Pro generation to support the necessary pro video features to enable Apple-level production quality (e.g. ProRes recording and Apple Log encoding).¹ For the next few years, Apple shot each of its keynotes using whatever the current iPhone Pro model was. But last year, with the September 2025 “Awe Dropping” event, they shot the keynote using the as-yet-unreleased iPhone 17 Pro itself — a bit of a risk given that it was running mid-summer iOS 26 beta software. They did the same this year, shooting the entire keynote, including the prelude fake movies, using iPhone 18 Pros. But, they couldn’t say that it was the 18 Pro on the title card because the 18 Pro wasn’t announced until later in the keynote.

There was at least one report from an ill-informed know-it-all who claimed, with odd certainty given how often he’s wrong, that Tim Cook would not appear in the keynote. But the last segment of the keynote’s cold opening gave us Cook, Tim Cook, framed in an obvious homage to the opening of 2012’s Skyfall:

I’ve stolen those comparison frames from this splendid blog post by Vaziri, Todd Vaziri, in which he points out a few other Bond homages in this keynote — including using the new variable aperture on the 18 Pro’s main camera as a reference to the Bond title-sequence gun barrel. Fun stuff all around, and a perfect way to commemorate the Cook-to-Ternus passing of the torch.² Also, the variety of visual styles in these mini movies really showed off the capabilities of the iPhone 18 Pro as a cinematic camera.

The little Cook / Bond homage that closes the opening segment starts with a voiceover from Cook himself (starting at 2m:07s):

Forget all that. Keep it classic. Open on an aerial shot, high above the California coast. A car drives down the highway. Then it disappears into a tunnel. Cut to an interior. Camera follows a mysterious figure, before revealing our main character...

No no no no no. Not me. That’s your guy. That’s your opener.

It’s a perfect segue, and Cook’s line reading is sublime. Cook’s on-screen persona improved a lot in the 15 years he served as CEO and keynote emcee, but even at his best he always came across as a bit, well, stiff. Voicing this introduction of John Ternus, on the other hand, he sounds cool as hell — relaxed, certain, and comfortable. I think the chairman role suits him better than CEO did. He’s a natural born éminence grise. And if this was Cook’s last appearance in a keynote, he went out on top.

Ternus’s Turn

For those of us invited to Apple Park, Ternus actually took the stage in person at 9:58 am, for a brief in-person greeting and introduction. What I recall isn’t anything he said but how he was greeted, which was with tremendous enthusiasm. The annual iPhone events always have a unique charge to them. WWDC is a far bigger event attendee-count-wise, but the September iPhone event is sharper, higher pressure. The exclusivity heightens the stakes — Apple only invites as many people as fit in Steve Jobs Theater: 1,000. The new era begins energy was palpable, if subtle. In the morning meet-and-greet coffee-and-apple-tarts mingling in and outside the Steve Jobs Theater upstairs atrium, attendees were genuinely unsure what to expect. A few were genuinely wondering if Ternus would bring back live-on-stage keynotes. (The apple tarts were supposedly made using apples grown on campus, and were quite good.)

In person, Ternus wore a dark t-shirt and jacket. In the keynote video, a burgundy shirt that clearly served as a reference to burgundy as the color of the year. (It’s hard to imagine Ternus in a cosmic orange shirt if he’d taken the top job a year prior.)

I think Ternus came across fine as emcee. Not great, but fine. He’s comfortable but I think the role still feels like a new pair of shoes he hasn’t yet broken in. Ternus’s opening line in the video was “Hello, and welcome to Apple Park”. I think that’s what he said on stage too. They’re leaving “Good morning!” as Cook’s signature, as they should.

Before any product announcements, Ternus laid out his theory of an “intelligent personal hub” for the AI era, making the case that if one were to design such a device from the ground up, starting today, you’d wind up with something exactly like an iPhone — with you all the time, with a great display, ubiquitous networking, great cameras and mics, and long battery life. The cynical take would be that this is an “If all you have is a hammer, everything looks like a nail” guiding doctrine. Apple is not a leading AI model company, and is completely sitting out the capex data center rat race, so of course they’re going to argue that the most popular product they make — the iPhone — is the central device of the AI era. But, well, I think it’s true. You need devices to run AI. I wrote an entire column about this back in May — “AI Is Technology, Not a Product” — responding to Steven Levy’s argument that Ternus “needs to launch a killer AI product”.

Levy, in Wired, wrote:

By the end of this decade, it’s unlikely that people will swipe on their phones to tap on Uber or Lyft. They will just tell their always-on AI agent to get them home. Or that agent will have already figured out where they need to go, and the car will be waiting without the friction of a request. “There’s an app for that,” may be replaced by “Let the agent do that.”

I wrote:

Actual products have to be real. Actual experiences have to rely on actual products. How exactly in Levy’s end-of-this-decade scenario will we tell our “always-on AI agent” to get us home? What microphone is listening to the command? What speaker is telling us the request was understood and acted upon? What screen do we look at to see how far away the hailed car is? I’d bet a pretty large sum of money that in 2030, when someone hails a ride-share vehicle to take them home, the most common product they’ll use to do that will be their phone. Whether they’re doing it via a verbal command issued to an “always-on AI agent” or good old tapping and swiping, it’ll be a phone.

Ternus’s “intelligent personal hub” doctrine seems very much in line with my retort to Levy’s not-entirely-thought-through fantasy. It also sounds an awful lot like a modern-day reimagining of a former Apple CEO’s “digital hub” strategy for the Mac. I highly recommend you take 8 minutes of your time, right now, and watch Steve Jobs’s introduction of the digital hub strategy from January 2001’s Macworld Expo. History doesn’t repeat but it does echo, and Ternus’s “intelligent personal hub” sure sounds exactly like Jobs’s “digital hub” echoing a quarter century later. If Ternus’s “intelligent personal hub” strategy works out like Jobs’s “digital hub” one did a quarter century ago, Apple’s going to be alright.

iPhone 18 Pro

Heretofore, to my recollection, the iPhone Pro always came last in the September keynotes (including the years when the higher-tier iPhones didn’t yet have the word “Pro” in their names, like the XS in 2018 and X in 2017). This year the 18 Pro came first. This worked both as the immediate product announcement after Ternus’s expression of Apple’s intelligent personal hub strategy, and to clear the final act of the keynote for the much-anticipated Duo. (It’s somewhat telling that the Air didn’t get that cleanup spot in the announcement batting order a year ago.³)

Thoughts on the colors, based on my hands-on experience:

• Black: Truly deep black. Reminds me a lot of the flat black iPhone 7 (not to be confused with the jet black iPhone 7 — what a year for black that was).
• Silver: Looks good. Not sure how it’s different from last year’s silver, because I had no way to compare side-by-side.
• Glacier: Nice. I think this will be popular. It’s definitely blue, not silver / gray (which you’d hope would be true in general, but especially alongside an explicit silver colorway).
• Burgundy: Last year the iPhone 17 Pro only came in three colors: silver, deep blue, and cosmic orange. I sort of thought of the strategy as “light”, “dark”, and “special”. But this year there are two specials, glacier and burgundy, but burgundy seems more special. Last year cosmic orange was clearly the color of the year. It was the most-used by far in Apple’s advertising, and I suspect it was the most popular sales-wise. Burgundy is that color this year. I don’t like it nearly as much, and I don’t think it will prove to be as popular or as iconic as cosmic orange. The cosmic orange iPhone 17 Pro seems set to go down as one of the most iconic iPhone colors ever made. I don’t think that will be true for the burgundy iPhone 18 Pro.

Pricing:

Compared to the 17 Pro one year ago:⁴

So:

• Max remains $100 extra.
• 2 TB is now available on the regular-size Pro; last year it was exclusive to the Max.
• 256 and 512 GB storage options are only $100 more expensive this year.
• 1 TB costs $300 more this year.
• 2 TB costs $500 more.

256 and 512 GB are by far the most popular sizes (and I think 256 in particular far outsells 512 — the base size truly is large enough for the majority of users), so for most 18 Pro buyers, the RAM/SSD shortage is only resulting in a $100 year-over-year price increase. The higher storage tier prices aren’t too bad, all things considered.

It’s worth pointing out here that coincident with the announcement of the iPhones 18 Pro and Duo on Wednesday, Apple raised the prices of all existing iPhones in its lineup by the same amounts, depending on storage — $100 for models with 128, 256, or 512 GB storage; and $300 for the iPhone Air with 1 TB.

Technically, the iPhone 18 Pro mostly offers three new things compared to the 17 Pro: the A20 Pro SoC (faster cores, two Neural Engines instead of one, and 50% faster memory bandwidth); variable aperture on the 1× main camera, ranging from ƒ1.48 to ƒ4.0; and an all-new thermal management system, including a much larger vapor chamber. The A20 Pro means peak performance is faster; the new thermal system allows it to maintain peak performance longer.

In a curious decision, the regular-sized 18 Pro models ship with Apple’s C2 cellular modem worldwide, but the 18 Pro Max models use a Qualcomm modem in the U.S., and the C2 everywhere else. Apple’s C-series modems are power efficiency marvels, with (in my experience) just as good reception and speed. I’d be just ever so slightly bummed about this if I were a Pro Max fan in the US.

That said, both the 18 Pro and Pro Max models continue to get slightly longer battery life in the US models than worldwide models, because the US models are eSIM-only, but most (all?) models elsewhere still have a physical SIM tray. The US models use the space occupied by the SIM tray in international models to make the batteries physically larger. Going by Apple’s new “typical use” tech spec, the US Pro / Pro Max models are rated for 24 / 30 hours respectively; the international models with a physical SIM tray for 23 / 29. Not a big difference, but not nothing.

In line with my belief that modern iPhones are best thought of as cameras with phones, not phones with cameras, most of the 18 Pro’s year-over-year improvements (and the features that set the Pro models apart from all other iPhones, including the Duo) are related to photography — both still and video:

• Photographic Styles is now up to version 3, adding new controls for “texture”, film-look, and grain. One stated goal of this is to enable shooting images with a less processed look; on the other end of the aesthetic spectrum, though, I think it will also enable shooting with a more processed look, with the addition of “glow” and “soft skin” as texture alternatives alongside “film” (which is the only texture with the new grain option). The other, and default, texture option is “standard”, which effectively means no texture processing.

• Apple Reference Image is a new shooting mode (main 1× camera only, because it’s tied to new features on that sensor) that performs pixel-by-pixel signing of the sensor data and embeds that signature in the file. After capture, reference images can be sent to Private Cloud Compute servers for verification and comparison to the original image sensor data. To capture images in Apple Reference format, you must shoot them in that mode, and you must enable that shooting mode first. The embedded sensor data doesn’t make files that much larger, but the sensor data stored in the file does uniquely identify your iPhone. When you “develop” a reference image — something the user must choose to do (not something that happens automatically each time you capture one) — the raw sensor data, sensor identifier, and time window when the image was captured are sent to Private Cloud Compute. When you share a photo captured in Apple Reference Image format, the sensor identifier is not included in the developed image, and you have the option to share a plain version without the embedded verification data. Apple Reference Image capture is not available in the EU or China, the world’s two largest bastions of freedom, but according to an Apple footnote, users in the EU “with iOS 27, iPadOS 27, and macOS 27 will be able to develop and view reference images.”

• Cinematic Mode can now be applied after shooting a regular video; on previous iPhones the Cinematic Mode effect was only available via a specific shooting mode.

To the layman’s eye, though, the 18 Pro just looks like it comes in new colors. That’s not a complaint, just a fact. Oh, but the Face ID infrared camera is now underneath pixels on the display, so the Dynamic Island cutout is now smaller. So the part of the Dynamic Island that is always black is smaller, but because there are now more display pixels next to it, the 18 Pro Dynamic Island can now show three active Live Activities, up from two. Apple is getting closer and closer each year to achieving a no-notch / no-cutout display.

AirPods 5

Just like with AirPods 4, Apple continues to make two different products under the same name. For $130, the cheaper AirPods 5 now have active noise cancellation (previously reserved for the higher-priced tier of AirPods 4), including Transparency and Adaptive modes. For $150, you get AirPods 5 with a wireless charging case (using MagSafe, Qi, or Apple Watch pucks), volume control by swiping up / down on the earbud stems, and longer battery life. It’s a much better product for just $20 extra, but it must have worked out well for Apple with a similar split with AirPods 4. I honestly can’t see why anyone should buy the $130 models, but someone must be buying them.

Apple provided me with a pair (the $150 model) to review, with no embargo. I wore them on the cross-country plane ride home, and compared them to my own AirPods Pro 3 for noise cancellation. It’s hard to make a direct comparison of the overall acoustic quality, but solely in terms of reducing the noise of the airplane cabin while trying to work, I found the new AirPods 5 with full noise cancellation to be roughly equivalent to the AirPods Pro 3 in Adaptive mode. That’s pretty good, and makes the ambient noise level in an airplane much more pleasant. But the full noise cancellation mode of the AirPods Pro 3 delivers an entire step change in noise reduction that the open-ear AirPods 5 cannot match.

I feel bad for people who despise airplane cabin noise but whose ears aren’t comfortable with closed-ear fit earbuds like AirPods Pro. But the noise reduction quality of the AirPods 5 seems impossibly good given their open-ear design. I don’t find AirPods Pro uncomfortable, but I do find regular AirPods more comfortable, especially for long stretches like hours-long flights. I’ll take the small trade-off in comfort for the big improvement in audio quality and noise cancellation with the AirPods Pro, though. With AirPods Pro you can truly forget you’re on an airplane at all if you’re watching a movie with Vision Pro or even on a MacBook or iPad display in a dark cabin.⁵

Apple Watch Series 12 and Ultra 4

The big new health thing is a next-generation heart sensor that takes readings every 5 seconds — 60× as frequent as with Series 11 / Ultra 3 a year ago. Apple commissioned a study comparing their new watches against a slew of competing heart rate monitoring devices, and released a paper with the results, claiming that Apple Watch was the most accurate. That’s an Apple-commissioned, Apple-released study, so take it with a grain of salt, but it looks solid to me and I’m unaware of any other studies showing otherwise. (And if I’m reading the results correctly, rings like Oura’s are crummy heart rate monitors.)

Ultra 4 looks the same as Ultras 1–3 (although Ultra 1 only came in natural titanium, not black). Series 12 comes in some tweaked colors in matte aluminum ($400) and polished titanium ($700), and for the first time since Series 5 seven years ago, brings back ceramic, in two colors: pearl white and night blue ($900). Those prices are for the smaller 42mm size; add $50 to each of them for the larger 46mm size. As has been true for every Series model to date, only the higher-priced materials come with sapphire crystal displays; the entry-model aluminum models still use Ion-X glass. But now those glass displays offer Ceramic Shield 2, which treatment has proven to be a tremendous boon to scratch resistance on iPhones.

The S11 chip powering the new watches enables two very interesting, and instantly controversial Apple Intelligence features:

• Live Rewind is a feature that allows you to double-click the digital crown to bring up a transcript of the last 15 seconds of conversation around you. Apple claims, somewhat preposterously, that audio isn’t being continuously recorded, when it clearly is. What they mean is that audio is only being cached for 15 seconds, continuously. There is no way to set Live Rewind to transcribe more (or less, for that matter) than the last 15 seconds of audio.

• Siri Recap is a feature that produces Apple Intelligence-generated summaries of your conversations throughout the day, while the feature is enabled on your Apple Watch. These are not transcripts, just bullet-point style summaries. You can set Siri Recap to only be enabled during certain times (like work hours), and you can toggle it on and off in Control Center at any time. But when it’s on, that’s the only control you have over it. You can’t manually start / stop conversations to record, like you do when recording something with Voice Memos. You just turn Siri Recap on, and Siri generates summaries when it thinks you’ve had a conversation worth summarizing. It’s entirely and exclusively automatic, when the mode is enabled.

Neither of these features are available in WatchOS 27.0. They’re both described as coming “in beta later this year”. I asked Apple reps if that meant 27.1, and they wouldn’t answer other than to repeat “later this year”, but I thought I saw a smile. Both features are English-only for now.

Some people are objecting to the very existence of these features as privacy intrusive. I think they’re pretty reasonably thought out. There are surveillance cameras all over the place now. I was walking home from dinner here in Philadelphia a few weeks ago with a friend visiting from out of town, talking about this, and he stopped midway between the restaurant and my house and asked me to count how many cameras we were on at that spot. I think we counted seven or eight. Who knows how many we missed? I’m not saying that’s good or bad, it’s just the world today. So too it’s soon going to be with personal devices recording audio. This is inevitable. I’m glad to see Apple ship features like this early, with pretty limited scope, to provide good examples before other companies ship devices that record and save audio continuously throughout the day. I think it’s better for Apple to lead with privacy-minded versions of ambient transcripts and AI-generated conversation summaries than to follow after other products have hit the market with all-day recordings and whole-life transcripts.

Think back to Apple’s announcement of AirTags. A lot of people had a perfectly reasonable “Hey I don’t know if it’s a good idea for a product like this to even exist” initial reaction. But having Apple lead the way actually proved good. I suspect Live Rewind and Siri Recap will turn out like that.

It’s also interesting to think about why these two features are only on Apple Watch, not iPhone. I think it’s because once an Apple Watch is on your wrist and unlocked, it remains a “trusted device” until you take the watch off. An iPhone is only a trusted device while it’s unlocked. It’s locked, and thus untrusted, while it’s in your pocket or bag, or even just sitting on a desk. I leave my iPhone unattended in our kitchen all the time. I’m often not in the room for conversations that take place in the kitchen while my iPhone is there, sitting on the counter. I shouldn’t get summaries of those conversations. And, thinking more nefariously, I shouldn’t be able to hide my iPhone on or near someone’s desk, or in a conference room, or in a car, or anywhere else, to get Siri Recap summaries of conversations where I’m not present. Making this an Apple Watch exclusive feature avoids that. Thus, I think Apple Watch can be trusted to “listen” ambiently, all the time while it’s unlocked on the wearer’s wrist, in ways that even an iPhone can’t.

iPhone Duo

Ternus broke out the hallowed “But actually, there is one more thing...” to introduce the Duo, which line — to circle back to the keynote’s intro — is for Apple CEOs what “Vodka martini, shaken not stirred” is for actors playing James Bond.⁶

Pre-orders for the Duo don’t start until October 16, and it doesn’t ship until October 23. In my experience, Apple never seeds review units for any product until a few days before pre-orders open, so we’re not going to hear or see anything more about the Duo until those review embargoes lift on October 20 or 21, I bet.

Duo is available in only two colors: star white (polished silver) and night sky (a deep blue-black). As predicted, the Duo is expensive. I’ll put it alongside the 18 Pro models for comparison:

At every storage tier, it’s $800 more than an 18 Pro and $700 more than an 18 Pro Max.

Key facts:

• It has no Face ID sensor, either open or closed. The only authentication is an iPad Air-style Touch ID side button.

• It has a Camera Control button, but does not have an Action button.

• The interior display has a matte nano-texture display. It looks great and apparently this is a key element of Apple’s efforts to make the crease as unnoticeable as possible. By next year I expect all competing foldables to have matte interior displays and 1.4 : 1 aspect ratio displays.

• It is very satisfying to open and close.

• The outer glossy screen has a “Center Stage” 12 megapixel front-facing camera, that seems very similar to the 18 megapixel Center Stage front cameras in the iPhone 17 generation phones. Year-ago tech but still high quality. There’s a circular hole punch in the corner of the display for this camera.

• The inner matte screen has an under-the-display “FaceTime” camera for which Apple hasn’t even given a megapixel count. Clearly this is a lesser camera — the question is how much lesser. The idea presumably is that this camera is solely for use in Zoom or FaceTime apps, where you might want to use the bigger inner display to view the meeting, and don’t care as much about your own image quality. But there’s no reason ever to capture a self portrait with the camera on the larger inner display. If the Duo is opened, and you care about image quality for self portraits, you can shoot using the main camera and use the small display to frame yourself, turning the large display around to the back. When not using the inner FaceTime camera, there is no visible cutout showing where it goes. If you look closely at the display however, you can see where it is. Under-display cameras aren’t magic.

• The regular cameras are a 1× / 2× main, and a 0.5× ultra wide. I suspect the main 1× camera is very similar optically to that of the iPhone Air, but the Duo has A20 Pro features like Photographic Styles 3, although it lacks Pro Controls, Apple Log, genlock, and other pro camera features. (The Air only has last year’s Photographic Styles 2.)

The initial introduction of the Duo was handled by two keynote newcomers: industrial design VP Molly Anderson, and human interface VP Steve Lemay. A hardware / software design duo (ahem), if you will. What made it worth having Anderson and Lemay present together is that the iPhone Duo exemplifies Apple’s integrated approach to product design. Are they a hardware company that writes their own software or a software company that makes their own hardware? The answer is yes.⁷ The defining aspect of the Duo isn’t that it folds open and closed — we’ve seen almost a decade of Android phones that do that — but its aspect ratio, roughly 1.4 : 1. Because that’s effectively the square root of 2, doubling the outer display’s shorter dimension when you unfold it produces an inner display with roughly the same aspect ratio. That’s elegant and satisfying conceptually. And I think in practice it will feel right, too.

Heretofore the iPhone has only had three display aspect ratios:

1. The original iPhone (2007) display was precisely 3 : 2 (480&NoBreak &NoBreak×&NoBreak &NoBreak320 pixels, and let’s call it 1.5 : 1 for comparison’s sake to the following ratios), and this remained true through the iPhone 4 and 4S (960&NoBreak &NoBreak×&NoBreak &NoBreak640 pixels, 2× retina).

2. Then the iPhone 5 (2012) display size went from 3.5 inches to 4 inches, but added all the extra pixels on the long dimension. Held vertically, it got taller, but not at all wider, changing the aspect ratio to 16 : 9 (roughly 1.77 : 1). This 16 : 9 aspect ratio held for all of the 5, 6, 7, and 8 generation iPhones, including the Plus models.

3. The iPhone X (2017) introduced the “all-screen” era. The actual device hardware didn’t change size or aspect ratio by much, but eliminating the chin (with Home button) and forehead and filling those areas with the screen, the display aspect ratio went to roughly 2.17 : 1, and every new model since then, including the new iPhone 18 Pro and Pro Max, has maintained that aspect ratio.⁸

Thus, over 20 years, iPhone displays (and thus all phones, since all the other phones mostly just copy iPhones) have only gotten more elongated: from 1.5 : 1 to 1.77 : 1 to 2.17 : 1. Now, though, the iPhone Duo goes squatter than the original iPhone for the first time, to 1.4 : 1 on both the outer and inner displays.⁹ That’s a noticeable change from the original iPhone, but it’s a downright striking change from every iPhone released in the last 10 years.

This aspect ratio is so different that it demands a different UI layout for the Duo to have any chance at being a good experience. That’s exactly what Apple has done, and what Steve Lemay was there in the keynote to unveil. A lot of UI chrome that previously was at the top or bottom of the display is now on the right — the dock on the home screen, the toolbars in apps, etc. And these elements are always on the right. You can’t turn the Duo upside down to move them to the left, nor is there an accessibility setting to move them to the left. This layout puts these controls on the right not because 90 percent of people are right-handed and Apple thinks lefties can get bent, but, I think (and so I’ve been told in conversations with Cupertinoites), because that’s the nature of a book-like shape. The arrow keys on keyboards always fall on the right, publishers don’t print left-handed books, and carmakers don’t make left-handed steering wheels. The right-side bias in the Duo iOS layout is like that.

I’ve only spent a few minutes, in total, playing with the Duo, so I really can’t say whether it succeeds or not. But I will say that it’s weird. Not saying it’s weird good or weird bad — just weird different. In some ways iOS on the Duo feels more different from iOS on all other post-iPhone-X phones than iOS on the original iPad felt from iOS on the iPhones of that time. The original iPad just felt like a much bigger iPhone. The Duo feels like a different layout entirely, in every single app that I had time to poke around with.

Is the Duo good? I do not know. Is it different? Very much.

Here’s the thing. If you take a 2 : 1 or even 2.2 : 1 aspect ratio phone display and turn that into a book-like foldable, the inner screen comes out square — or so close to square that we might as well just call it square. Nobody makes any device of any sort that just has a square display. Square is a terrible aspect ratio for a phone or a tablet or a laptop or a desktop display. The only square displays on the market are the inner displays for Android foldables from the last decade. Square being a shitty aspect ratio is half the reason almost no one has bought those things. The other half of the reason is that no developers truly adapted software to take advantage of the double-area inner displays. Google didn’t do it, Samsung didn’t do it, and since neither of them bothered to do it, of course third-party developers didn’t do it either. Nothing adapts to the inner display on Android foldables — it just expands to fill the space.

Apple’s insight to go with 1.4 : 1 on the outside and thus also 1.4 : 1 on the inside gives the Duo a chance. Maybe this will be a huge hit. But maybe not? There are no 5-inch phone displays with aspect ratios anywhere near as squat as the Duo’s. That’s a tablet aspect ratio. The current A17 Pro iPad Mini has a 2266&NoBreak &NoBreak×&NoBreak &NoBreak1488 pixel display: roughly 1.5 : 1. The 11-inch iPad Air is 1.4 : 1 and the 13-inch Air is 1.33 : 1. So I think the Duo’s inner display is going to feel just right.

But that outer display? I don’t know. Even in the best case, it’s going to be no better than a tolerable compromise. If it were actually a good idea there would be non-folding phones with 1.4 : 1-ish squat displays, and there aren’t any such phones. And even if the Duo proves to be a hit, I don’t think there will be. If you only have one display you don’t want it to be the size and shape of the Duo’s outer display. So the questions are: How tolerable has Apple’s UI redesign made it? And how will Duo users split their time between using it closed vs. using it open? Someone who spends more of their time using it open might absolutely love it with a “you’ll take it away only from my cold dead hands” fervor. Someone who winds up mostly using the outer display because so many of their iPhone interactions are quick take it out → use it for a bit to check something, jot something down, or snap a quick photo → put it back in pocket / bag situations, might wind up regretting it.

Given the design brief of creating a book-shaped foldable phone, I feel quite certain already that Apple has nailed it with the Duo. But having only seen the keynote spiel and poked around with one in hand for a few minutes Wednesday afternoon, I think it remains an open question whether a book-shaped foldable phone is a good idea in the first place. Folks from Apple who have actually spent time carrying the Duo seem genuinely enthused. I remain highly skeptical about its appeal for me personally, and somewhat skeptical about its appeal broadly. That’s OK if it’s not suited for me, but is suited for many (but not most) people. No one in the keynote described the iPhone Duo as the future of iPhone, like they did when introducing the iPhone X in 2017. I just wonder if the fundamental idea appeals to all that many people at all. We’ll find out starting in late October.

Regarding some specific Duo details:

• The lack of Face ID is a serious sad-trombone bummer, and in my opinion could sink the first-generation model. Side-button Touch ID is OK, but OK is not good, and even good is not insanely great. Face ID is insanely great. For whatever reason, the only iPads with Face ID are the iPad Pro models. Every time I review an iPad Air or iPad Mini, my biggest gripe is the lack of Face ID. I complained about it reviewing the iPad Air in 2022 and the iPad Mini in 2021, and in the five years since then my annoyance has only grown. With Face ID, so long as you’re looking at the screen as you unlock the device, it’s like not having it locked at all. It just unlocks, seemingly automatically. With Touch ID you must think about putting one of your registered fingers on the button every single time you unlock it. It never ever feels automatic. I feel quite certain Apple tried to fit a Face ID sensor on the outer Duo display and could not make the engineering work.

• I’m much more concerned about convenience than security, but Face ID is 20× more secure than Touch ID. That’s according to Apple’s own Apple Platform Security whitepaper, updated just last month, which states:

  The probability that a random person in the population could unlock a user’s iPad, iPhone, or Apple Vision Pro is less than 1 in 1 million with Optic ID or Face ID — including when Face ID with a mask is turned on. For a user’s iPad, iPhone, Mac models with Touch ID, and those paired with a Magic Keyboard with Touch ID, it’s less than 1 in 50,000. This probability increases with multiple enrolled fingerprints (up to 1 in 10,000 with five fingerprints) or appearances for Face ID (up to 1 in 500,000 with two appearances).

• As a sign of how inconvenient the lack of Face ID is, Apple is promoting the fact that the iPhone Duo can be unlocked by your Apple Watch, like a Mac. No other iPhone needs such a crutch.¹⁰ Based on my experience with Touch ID iPads, wearing an Apple Watch all the time to unlock the Duo will feel essential. I don’t wear an Apple Watch all the time, and don’t want to start wearing an Apple Watch all the time, so I suspect this alone will keep me from switching to an iPhone Duo personally. On the other hand, for people who do wear an Apple Watch all the time and are happy to do so, the Duo’s lack of Face ID might prove to be only an occasional inconvenience rather than a constant annoyance.¹¹

• Contrary to some speculation, the Duo does support MagSafe. (Samsung’s latest, the Galaxy Z Fold8 and Fold8 Ultra, support inductive Qi charging but don’t have magnets, so you need to use a case or something to get chargers to snap into place.)

• Closed, the Duo has — per Molly Anderson’s dulcet keynote description — “a distinctive asymmetric shape, with two rounded corners, that invites you to open it.” Expect this to be copied by every other foldable phone in 2027, and then for Android dorks to argue that it’s a totally obvious design. (Same for the matte interior display.)

• Apple rates the Duo as IP68 for dust and water resistance, with exactly the same description as the iPhones 18 Pro (and all other iPhone models currently in the lineup): “Rated IP68 (maximum depth of 6 meters up to 30 minutes) under IEC standard 60529”. Samsung’s new Z Fold8 is only IP48. With these IEC ratings the second numeral is for water. The first numeral is for “solid foreign agents” (read: dirt and dust). iPhone Duo’s IP6x rating means “dust-tight”. The Z Fold8’s IP4x rating isn’t even “dust-protected” (that would be IP5x) but instead only protection against particles 1mm or larger. 1mm is pretty big for a piece of grit. That doesn’t mean it’s susceptible to ingress from every particle smaller than 1mm, but it’s not good. Pocket lint is a real thing and people put their phones in their pockets.

I’m Trying So Hard Not to Be Insufferable About Predicting the ‘Duo’ Name That I’ve Put This Entire Section 6,500 Words Deep Into the Column

Yours truly, back on April 14:

I have no inside knowledge about what Apple plans to name this device, but I’ll eat my proverbial hat if they name it “iPhone Fold”. That name is so dumb it’s what Samsung calls their foldables. You don’t name a device for what it does, you name it for what it connotes. A good name conveys feeling, not just function. “iPhone Ultra” or “iPhone Max” would both work, and Ultra sounds more luxe. So while unsurprising, that’s probably the best bet, even without the reliable word of Mr. Digital Chat Station.

But if you want my take on a wildcard name, one with some history, how about “iPhone Duo”?

My thinking re: “Duo” seemingly jibed with that of Apple’s crack marketing team:

• It’s two devices in one.
• The early 1990s PowerBook Duo (and corresponding Duo Dock) were extraordinary Macs that were ahead of their time.
• It’s not ultra, if we take ultra to mean “better than what’s less than ultra in every way”. The iPhone Duo is different — quite different — from the iPhones Pro but, to name one conspicuous example, the iPhones Pro have better camera systems. (The Duo is ultra-priced, but one can also say it’s double-priced.)
• Duo looks and sounds so good alongside Neo.

When I suggested “Duo” in April and again the day before the event, the number one objection was that Microsoft had used the name “Duo” for one of its many forgettable Surface products. The half dozen or so readers who argued thus might be the only people alive who remember this product. No one cares. Microsoft also had a Surface Neo, and no one cares about that one either. Even if Apple hadn’t offered a line of (successful, memorable) PowerBook Duos from 30 years ago, it wouldn’t matter that Microsoft had used “Duo” in 2020. For chrissakes Cisco had the rights to the name “iPhone” in 2007 and had been making products with the name since 1998.

I love the name Duo for this iPhone. It just feels right. And, just five days in, I already find myself saying and writing just “Duo”, without “iPhone”, when referring to it.

Miscellaneous

• Too many cooks spoil the stew. It was very fun and a little touching to have the Cook hand things off to Ternus at the start, but I think there were too many presenters in the keynote. Duo was effectively introduced three or four separate times. First by the Anderson / Lemay (I’m going to pun it again) duo, then Joz, then Craig Federighi (covering much of the same software ground Lemay did), then Johny Srouji (covering much of the same ground Anderson did). Maybe the thinking was that everyone wanted a part in Ternus’s debut keynote. Maybe this is just the way of the future — many small segments, rather than one continuous on-stage live monologue. It didn’t make the keynote run overly long (running time was just 1h:17m), so maybe I’m all wet here. But I think what I wanted was to see Ternus do more of the explaining — not just the what but the why.

• Apple’s YouTube video of the entire keynote has over 49 million views, and that doesn’t count the views from Apple’s own hosted version, presented on Apple.com and Apple TV. Apple has 12 other shorter keynote-related videos on YouTube, including “Introducing the new iPhone 18 Pro” with 18 million views, and “Introducing the new iPhone Duo” with 28 million views. For comparison, this year’s WWDC keynote on YouTube has 8.9 million views. 9 million views is a hell of a view count for the introduction of operating systems, but the iPhone remains king — not just for Apple, but the industry. It remains the Super Bowl of the annual tech calendar.

• For an even better comparison, in the full version of that 2001 Macworld Expo keynote that I suggested you watch earlier in this column — the one where Steve Jobs introduced the Mac-as-digital-hub strategy — when Jobs takes the stage, he’s beaming with pride when he says there are “tens of thousands” of people watching the stream live. Both streaming and Apple have come a long way in 25 years.

@scottjla on Twitter in reply to my pelican riding a bicycle benchmark:

I feel like we need to stack these tests now

I checked to confirm that the model (ChatGPT Images 2.0) added the "WHY ARE YOU LIKE THIS" sign of its own accord and it did - the prompt Scott used was:

Create an image of a horse riding an astronaut, where the astronaut is riding a pelican that is riding a bicycle. It looks very chaotic but they all just manage to balance on top of each other

Tags: text-to-image, pelican-riding-a-bicycle, ai, generative-ai, slop, chatgpt

I was a guest on Lenny Rachitsky's podcast, in a new episode titled An AI state of the union: We've passed the inflection point, dark factories are coming, and automation timelines. It's available on YouTube, Spotify, and Apple Podcasts. Here are my highlights from our conversation, with relevant links.

• The November inflection point
• Software engineers as bellwethers for other information workers
• Writing code on my phone
• Responsible vibe coding
• Dark Factories and StrongDM
• The bottleneck has moved to testing
• This stuff is exhausting
• Interruptions cost a lot less now
• My ability to estimate software is broken
• It's tough for people in the middle
• It's harder to evaluate software
• The misconception that AI tools are easy
• Coding agents are useful for security research now
• OpenClaw
• Journalists are good at dealing with unreliable sources
• The pelican benchmark
• And finally, some good news about parrots
• YouTube chapters

The November inflection point

4:19 - The end result of these two labs throwing everything they had at making their models better at code is that in November we had what I call the inflection point where GPT 5.1 and Claude Opus 4.5 came along.

They were both incrementally better than the previous models, but in a way that crossed a threshold where previously the code would mostly work, but you had to pay very close attention to it. And suddenly we went from that to... almost all of the time it does what you told it to do, which makes all of the difference in the world.

Now you can spin up a coding agent and say, build me a Mac application that does this thing, and you'll get something back which won't just be a buggy pile of rubbish that doesn't do anything.

Software engineers as bellwethers for other information workers

5:49 - I can churn out 10,000 lines of code in a day. And most of it works. Is that good? Like, how do we get from most of it works to all of it works? There are so many new questions that we're facing, which I think makes us a bellwether for other information workers.

Code is easier than almost every other problem that you pose these agents because code is obviously right or wrong - either it works or it doesn't work. There might be a few subtle hidden bugs, but generally you can tell if the thing actually works.

If it writes you an essay, if it prepares a lawsuit for you, it's so much harder to derive if it's actually done a good job, and to figure out if it got things right or wrong. But it's happening to us as software engineers. It came for us first.

And we're figuring out, OK, what do our careers look like? How do we work as teams when part of what we did that used to take most of the time doesn't take most of the time anymore? What does that look like? And it's going to be very interesting seeing how this rolls out to other information work in the future.

Lawyers are falling for this really badly. The AI hallucination cases database is up to 1,228 cases now!

Plus this bit from the cold open at the start:

It used to be you'd ask ChatGPT for some code, and it would spit out some code, and you'd have to run it and test it. The coding agents take that step for you now. And an open question for me is how many other knowledge work fields are actually prone to these agent loops?

Writing code on my phone

8:19 - I write so much of my code on my phone. It's wild. I can get good work done walking the dog along the beach, which is delightful.

I mainly use the Claude iPhone app for this, both with a regular Claude chat session (which can execute code now) or using it to control Claude Code for web.

Responsible vibe coding

9:55 If you're vibe coding something for yourself, where the only person who gets hurt if it has bugs is you, go wild. That's completely fine. The moment you ship your vibe coding code for other people to use, where your bugs might actually harm somebody else, that's when you need to take a step back.

See also When is it OK to vibe code?

Dark Factories and StrongDM

12:49 The reason it's called the dark factory is there's this idea in factory automation that if your factory is so automated that you don't need any people there, you can turn the lights off. Like the machines can operate in complete darkness if you don't need people on the factory floor. What does that look like for software? [...]

So there's this policy that nobody writes any code: you cannot type code into a computer. And honestly, six months ago, I thought that was crazy. And today, probably 95% of the code that I produce, I didn't type myself. That world is practical already because the latest models are good enough that you can tell them to rename that variable and refactor and add this line there... and they'll just do it - it's faster than you typing on the keyboard yourself.

The next rule though, is nobody reads the code. And this is the thing which StrongDM started doing last year.

I wrote a lot more about StrongDM's dark factory explorations back in February.

The bottleneck has moved to testing

21:27 - It used to be, you'd come up with a spec and you hand it to your engineering team. And three weeks later, if you're lucky, they'd come back with an implementation. And now that maybe takes three hours, depending on how well the coding agents are established for that kind of thing. So now what, right? Now, where else are the bottlenecks?

Anyone who's done any product work knows that your initial ideas are always wrong. What matters is proving them, and testing them.

We can test things so much faster now because we can build workable prototypes so much quicker. So there's an interesting thing I've been doing in my own work where any feature that I want to design, I'll often prototype three different ways it could work because that takes very little time.

I've always loved prototyping things, and prototyping is even more valuable now.

22:40 - A UI prototype is free now. ChatGPT and Claude will just build you a very convincing UI for anything that you describe. And that's how you should be working. I think anyone who's doing product design and isn't vibe coding little prototypes is missing out on the most powerful boost that we get in that step.

But then what do you do? Given your three options that you have instead of one option, how do you prove to yourself which one of those is the best? I don't have a confident answer to that. I expect this is where the good old fashioned usability testing comes in.

More on prototyping later on:

46:35 - Throughout my entire career, my superpower has been prototyping. I've been very quick at knocking out working prototypes of things. I'm the person who can show up at a meeting and say, look, here's how it could work. And that was kind of my unique selling point. And that's gone. Anyone can do what I could do.

This stuff is exhausting

26:25 - I'm finding that using coding agents well is taking every inch of my 25 years of experience as a software engineer, and it is mentally exhausting. I can fire up four agents in parallel and have them work on four different problems. And by like 11 AM, I am wiped out for the day. [...]

There's a personal skill we have to learn in finding our new limits - what's a responsible way for us not to burn out.

I've talked to a lot of people who are losing sleep because they're like, my coding agents could be doing work for me. I'm just going to stay up an extra half hour and set off a bunch of extra things... and then waking up at four in the morning. That's obviously unsustainable. [...]

There's an element of sort of gambling and addiction to how we're using some of these tools.

Interruptions cost a lot less now

45:16 - People talk about how important it is not to interrupt your coders. Your coders need to have solid two to four hour blocks of uninterrupted work so they can spin up their mental model and churn out the code. That's changed completely. My programming work, I need two minutes every now and then to prompt my agent about what to do next. And then I can do the other stuff and I can go back. I'm much more interruptible than I used to be.

My ability to estimate software is broken

28:19 - I've got 25 years of experience in how long it takes to build something. And that's all completely gone - it doesn't work anymore because I can look at a problem and say that this is going to take two weeks, so it's not worth it. And now it's like... maybe it's going to take 20 minutes because the reason it would have taken two weeks was all of the sort of crufty coding things that the AI is now covering for us.

I constantly throw tasks at AI that I don't think it'll be able to do because every now and then it does it. And when it doesn't do it, you learn, right? But when it does do something, especially something that the previous models couldn't do, that's actually cutting edge AI research.

And a related anecdote:

36:56 - A lot of my friends have been talking about how they have this backlog of side projects, right? For the last 10, 15 years, they've got projects they never quite finished. And some of them are like, well, I've done them all now. Last couple of months, I just went through and every evening I'm like, let's take that project and finish it. And they almost feel a sort of sense of loss at the end where they're like, well, okay, my backlog's gone. Now what am I going to build?

It's tough for people in the middle

29:29 - So ThoughtWorks, the big IT consultancy, did an offsite about a month ago, and they got a whole bunch of engineering VPs in from different companies to talk about this stuff. And one of the interesting theories they came up with is they think this stuff is really good for experienced engineers, like it amplifies their skills. It's really good for new engineers because it solves so many of those onboarding problems. The problem is the people in the middle. If you're mid-career, if you haven't made it to sort of super senior engineer yet, but you're not sort of new either, that's the group which is probably in the most trouble right now.

I mentioned Cloudflare hiring 1,000 interns, and Shopify too.

Lenny asked for my advice for people stuck in that middle:

31:21 - That's a big responsibility you're putting on me there! I think the way forward is to lean into this stuff and figure out how do I help this make me better?

A lot of people worry about skill atrophy: if the AI is doing it for you, you're not learning anything. I think if you're worried about that, you push back at it. You have to be mindful about how you're applying the technology and think, okay, I've been given this thing that can answer any question and often gets it right. How can I use this to amplify my own skills, to learn new things, to take on much more ambitious projects? [...]

33:05 - Everything is changing so fast right now. The only universal skill is being able to roll with the changes. That's the thing that we all need.

The term that comes up most in these conversations about how you can be great with AI is agency. I think agents have no agency at all. I would argue that the one thing AI can never have is agency because it doesn't have human motivations.

So I'd say that's the thing is to invest in your own agency and invest in how to use this technology to get better at what you do and to do new things.

It's harder to evaluate software

The fact that it's so easy to create software with detailed documentation and robust tests means it's harder to figure out what's a credible project.

37:47 Sometimes I'll have an idea for a piece of software, Python library or whatever, and I can knock it out in like an hour and get to a point where it's got documentation and tests and all of those things, and it looks like the kind of software that previously I'd have spent several weeks on - and I can stick it up on GitHub

And yet... I don't believe in it. And the reason I don't believe in it is that I got to rush through all of those things... I think the quality is probably good, but I haven't spent enough time with it to feel confident in that quality. Most importantly, I haven't used it yet.

It turns out when I'm using somebody else's software, the thing I care most about is I want them to have used it for months.

I've got some very cool software that I built that I've never used. It was quicker to build it than to actually try and use it!

The misconception that AI tools are easy

41:31 - Everyone's like, oh, it must be easy. It's just a chat bot. It's not easy. That's one of the great misconceptions in AI is that using these tools effectively is easy. It takes a lot of practice and it takes a lot of trying things that didn't work and trying things that did work.

Coding agents are useful for security research now

19:04 - In the past sort of three to six months, they've started being credible as security researchers, which is sending shockwaves through the security research industry.

See Thomas Ptacek: Vulnerability Research Is Cooked.

At the same time, open source projects are being bombarded with junk security reports:

20:05 - There are these people who don't know what they're doing, who are asking ChatGPT to find a security hole and then reporting it to the maintainer. And the report looks good. ChatGPT can produce a very well formatted report of a vulnerability. It's a total waste of time. It's not actually verified as being a real problem.

A good example of the right way to do this is Anthropic's collaboration with Firefox, where Anthropic's security team verified every security problem before passing them to Mozilla.

OpenClaw

Of course we had to talk about OpenClaw! Lenny had his running on a Mac Mini.

1:29:23 - OpenClaw demonstrates that people want a personal digital assistant so much that they are willing to not just overlook the security side of things, but also getting the thing running is not easy. You've got to create API keys and tokens and install stuff. It's not trivial to get set up and hundreds of thousands of people got it set up. [...]

The first line of code for OpenClaw was written on November the 25th. And then in the Super Bowl, there was an ad for AI.com, which was effectively a vaporware white labeled OpenClaw hosting provider. So we went from first line of code in November to Super Bowl ad in what? Three and a half months.

I continue to love Drew Breunig's description of OpenClaw as a digital pet:

A friend of mine said that OpenClaw is basically a Tamagotchi. It's a digital pet and you buy the Mac Mini as an aquarium.

Journalists are good at dealing with unreliable sources

In talking about my explorations of AI for data journalism through Datasette:

1:34:58 - You would have thought that AI is a very bad fit for journalism where the whole idea is to find the truth. But the flip side is journalists deal with untrustworthy sources all the time. The art of journalism is you talk to a bunch of people and some of them lie to you and you figure out what's true. So as long as the journalist treats the AI as yet another unreliable source, they're actually better equipped to work with AI than most other professions are.

The pelican benchmark

Obviously we talked about pelicans riding bicycles:

56:10 - There appears to be a very strong correlation between how good their drawing of a pelican riding a bicycle is and how good they are at everything else. And nobody can explain to me why that is. [...]

People kept on asking me, what if labs cheat on the benchmark? And my answer has always been, really, all I want from life is a really good picture of a pelican riding a bicycle. And if I can trick every AI lab in the world into cheating on benchmarks to get it, then that just achieves my goal.

59:56 - I think something people often miss is that this space is inherently funny. The fact that we have these incredibly expensive, power hungry, supposedly the most advanced computers of all time. And if you ask them to draw a pelican on a bicycle, it looks like a five-year-old drew it. That's really funny to me.

And finally, some good news about parrots

Lenny asked if I had anything else I wanted to leave listeners with to wrap up the show, so I went with the best piece of news in the world right now.

1:38:10 - There is a rare parrot in New Zealand called the Kākāpō. There are only 250 of these parrots left in the world. They are flightless nocturnal parrots - beautiful green dumpy looking things. And the good news is they're having a fantastic breeding season in 2026,

They only breed when the Rimu trees in New Zealand have a mass fruiting season, and the Rimu trees haven't done that since 2022 - so there has not been a single baby kākāpō born in four years.

This year, the Rimu trees are in fruit. The kākāpō are breeding. There have been dozens of new chicks born. It's a really, really good time. It's great news for rare New Zealand parrots and you should look them up because they're delightful.

Everyone should watch the live stream of Rakiura on her nest with two chicks!

YouTube chapters

Here's the full list of chapters Lenny's team defined for the YouTube video:

• 00:00: Introduction to Simon Willison
• 02:40: The November 2025 inflection point
• 08:01: What's possible now with AI coding
• 10:42: Vibe coding vs. agentic engineering
• 13:57: The dark-factory pattern
• 20:41: Where bottlenecks have shifted
• 23:36: Where human brains will continue to be valuable
• 25:32: Defending of software engineers
• 29:12: Why experienced engineers get better results
• 30:48: Advice for avoiding the permanent underclass
• 33:52: Leaning into AI to amplify your skills
• 35:12: Why Simon says he's working harder than ever
• 37:23: The market for pre-2022 human-written code
• 40:01: Prediction: 50% of engineers writing 95% AI code by the end of 2026
• 44:34: The impact of cheap code
• 48:27: Simon's AI stack
• 54:08: Using AI for research
• 55:12: The pelican-riding-a-bicycle benchmark
• 59:01: The inherent ridiculousness of AI
• 1:00:52: Hoarding things you know how to do
• 1:08:21: Red/green TDD pattern for better AI code
• 1:14:43: Starting projects with good templates
• 1:16:31: The lethal trifecta and prompt injection
• 1:21:53: Why 97% effectiveness is a failing grade
• 1:25:19: The normalization of deviance
• 1:28:32: OpenClaw: the security nightmare everyone is looking past
• 1:34:22: What's next for Simon
• 1:36:47: Zero-deliverable consulting
• 1:38:05: Good news about Kakapo parrots

Tags: ai, kakapo, generative-ai, llms, podcast-appearances, coding-agents, agentic-engineering

The Good News First

Just this week I wrote about a hidden defaults preference you can set to turn off most of the insipid menu item icons in most of Apple’s first-party apps in MacOS 26 Tahoe. I bemoaned the fact that Safari — generally an exemplar of what makes a great Mac app a great Mac app — generally ignored this setting, leaving most of its menu item icons in place. I am delighted to report that that’s fixed in MacOS 26.4. With the preference set to hide these icons, Safari now only shows a handful.

Here’s a link to the screenshot of the old before/after, taken on MacOS 26.3.2. Boo hiss. Here’s the new before/after, taken on MacOS 26.4:

In Tahoe 26.3 (and presumably, earlier versions of Tahoe), 16 of 19 menu items in Safari’s File menu still showed an icon with this setting enabled. In 26.4, only 5 of 19 do.¹ The rest of Safari’s other menus have been updated similarly, and look so much better for it.

It’s interesting to me that Safari was updated to support this hidden preference in 26.4. I take it as a sign that there’s a contingent within Apple (or least within the Safari team) that dislikes these menu item icons enough to notice that Safari wasn’t previously recognizing this preference setting. (And I further take it as a sign that within Apple’s engineering ranks, the existence of this defaults setting is widely known.) Keep hope alive.

Now the Bad News

Another recent Tahoe-related tip I’ve been writing about was using a device management profile to block the prompts in System Settings → General → Software Update to “upgrade” from MacOS 15 Sequoia to 26 Tahoe. I first wrote about it a month ago, linking to a post from Rob Griffiths. I then wrote about it again, just this week, linking to a YouTube video from Mr. Macintosh.

Ever since this technique started making the rounds, there was widespread commentary that it was taking advantage of a bug, not a feature, in MacOS 15 Sequoia. The 90-day “deferral” period to block the Tahoe update prompts was supposed to be from the date the Tahoe major release (26.0) was released, not from the most recent minor release. Welp, with this week’s release of MacOS 15.7.5, this bug is fixed, and Tahoe shows up in the Software Update panel in System Settings even if you have one of these device management profiles installed. Alas.

All is not lost, however. The same video from Mr. Macintosh shows a second, slightly less elegant way to banish all signs of Tahoe in Software Update (just after the 9:00 mark). The trick is to register your Mac for the MacOS Sequoia Public Beta updates (or the developer betas). This blocks all signs of Tahoe. You don’t actually have to install any future betas of Sequoia (at the moment, there are none available). Just make sure you have Automatic Updates disabled too. I’d rather risk inadvertently installing a public beta of 15.8 Sequoia than inadvertently “upgrading” to Tahoe.

Not every website has an RSS feed. Some never did. Some had one years ago and quietly removed it. And some sites have content that updates regularly but was never structured as a feed in the first place: job boards, product listings, event calendars, changelog pages. Until now, if a site didn’t offer RSS, you were out of luck.

Web Feeds is a new feature that creates RSS feeds from any website. Point it at a URL, and NewsBlur analyzes the page structure, identifies the repeating content patterns, and generates extraction rules that turn the page into a live feed. It works on news sites, blogs, job boards, product pages, or really anything with a list of items that changes over time.

This is a huge feature and has been requested for years. I’m so thrilled to finally be able to offer it in a way that I feel comfortable with. Other solutions including having you select story titles on a re-hosted version of the page, but it was clumsy and error-prone. This way, we use LLMs to figure out what the story titles are likely to be, present the variations to you, and then let you decide what’s right. So much better!

How it works

Open the Add + Discover Sites page and click the Web Feed tab. Paste a URL and click Analyze. NewsBlur fetches the page, strips out navigation and boilerplate, and analyzes the HTML structure. Within a few seconds, you’ll see multiple extraction variants, each representing a different content pattern found on the page.

Progress updates stream in real-time while the analysis runs. NewsBlur typically finds 3-5 different extraction patterns on a page. The first variant is usually the main content (article list, blog posts, product grid), but sometimes the page has multiple distinct sections worth subscribing to. Each variant shows a label, a description of what it captures, and a preview of 3 extracted stories so you can see exactly what you’d get.

Select the variant that matches what you want to follow, pick a folder, and subscribe. NewsBlur will re-fetch and re-extract the page on a regular schedule, just like any other feed.

Story hints

Sometimes the initial best guess isn’t what you’re looking for. Maybe the page has a blog section and a job listings section, and you want the jobs. Click the Refine button and type a hint like “I’m looking for the job postings.” NewsBlur re-analyzes the page with your hint in mind and reorders the variants to prioritize what you described.

What gets extracted

For each story, NewsBlur extracts whatever it can find: title, link, content snippet, image, author, and date. Not every field will be available on every site, and that’s fine. At minimum you’ll get titles and links. The extraction uses XPath expressions, which means it’s precise and consistent across page refreshes as long as the site’s HTML structure stays the same.

When things change

Websites redesign. HTML structures shift. When NewsBlur detects that the extraction rules have stopped working (after 3 consecutive failures), the feed is flagged as needing re-analysis. You’ll see a feed exception indicator, and you can re-analyze the page with one click to generate updated extraction rules.

Use cases

Some examples of sites that work well with Web Feeds:

• Company blogs without RSS — Many corporate blogs dropped their RSS feeds years ago. Web Feeds brings them back.
• Job boards — Track new postings on a company’s careers page.
• Government sites — Follow press releases, meeting agendas, or public notices.
• Changelog pages — Monitor when a tool or service ships updates.
• Event listings — Keep tabs on upcoming concerts, conferences, or local events.
• Product pages — Watch for new arrivals or restocks on stores that don’t offer feeds.

Availability

Web Feeds are available to Premium Archive and Premium Pro subscribers. The ongoing feed fetching and extraction runs on NewsBlur’s servers like any other feed.

If you have feedback or ideas for improvements, please share them on the NewsBlur forum.