Linked List: March 15, 2025

Saturday, 15 March 2025

WorkOS: Scalable, Secure Authentication ★

My thanks to WorkOS for sponsoring this week at DF. Modern authentication should be seamless and secure. WorkOS makes it easy to integrate features like MFA, SSO, and RBAC.

Whether you’re replacing passwords, stopping fraud, or adding enterprise auth, WorkOS can help you build frictionless auth that scales. Future-proof your authentication stack with the identity layer trusted by OpenAI, Cursor, Perplexity, and Vercel. Upgrade your auth today.

How to Generate a Report of Apple Intelligence Requests Sent to Private Cloud Compute ★

From Apple’s support documentation:

You can generate a report of requests your iPhone has sent to Private Cloud Compute.

Go to Settings, then tap Privacy & Security.

Tap Apple Intelligence Report, then choose a report duration for the last 15 minutes (default) or last 7 days. Choose off to disable the report.

Note: The report may be empty if there haven’t been any Private Cloud Compute requests since you changed the duration.

Tap Export Activity, choose a place to store the file, then tap Export.

The report is saved as a file named Apple_Intelligence_Report.json.

Open the file with a text reader.

These are the iOS instructions, but they’re exactly the same on MacOS 15 Sequoia. My first generated report was empty for the last 7 days, and it was empty again even after running the Writing Tools Proofread function on the text of my 4,000-word “Something Is Rotten in the State of Cupertino” article from this week. But when I ran the Writing Tools Summarize feature on the same text, I wound up with a long entry that was sent to Private Cloud Compute. So, at the moment, Summarize seems like a good way to invoke Private Cloud Compute, even from a relatively powerful Mac.

Here’s the summary Apple Intelligence generated. I have to say: it’s pretty good. It’s completely petty but also completely me to notice and object to the way it uses two spaces after periods — and worse, only some of the time. Also, the sentence “This raises concerns about the company’s ability to maintain its position as a leader in AI innovation” is, let’s say, off the mark.

Update: Howard Oakley wrote a post with a brief overview of the structure and contents of these reports back on October 29.

New RCS Spec From GSM Association Adds E2EE; Both Apple and Google to Support It ★

Jess Weatherbed, reporting for The Verge:

iPhone and Android users will be able to exchange end-to-end encrypted (E2EE) RCS messages in the near future thanks to newly updated RCS specifications. The GSM Association announced that the latest RCS standard includes E2EE based on the Messaging Layer Security (MLS) protocol, enabling interoperable encryption between different platform providers for the first time. [...]

“End-to-end encryption is a powerful privacy and security technology that iMessage has supported since the beginning, and now we are pleased to have helped lead a cross industry effort to bring end-to-end encryption to the RCS Universal Profile published by the GSMA,” said Apple spokesperson Shane Bauer. “We will add support for end-to-end encrypted RCS messages to iOS, iPadOS, macOS, and watchOS in future software updates.” [...]

“We’ve always been committed to providing a secure messaging experience, and Google Messages users have had end-to-end encrypted (E2EE) RCS messaging for years,” Google spokesperson Ed Fernandez told The Verge. “We’re excited to have this updated specification from GSMA and work as quickly as possible with the mobile ecosystem to implement and extend this important user protection to cross-platform RCS messaging.”

This is nothing but good news. But it’s wrong to frame this along Google’s lines, that they’ve been there waiting for Apple to support E2EE for RCS. They’ve been waiting for Apple to support RCS at all, yes, and Google has also implemented their own proprietary E2EE layer for RCS. But until now, there was no E2EE specification in the open RCS spec. Now there is. That’s why it’s not just Android ↔︎ iOS RCS messaging that wasn’t able to use E2EE, but even Android ↔︎ Android, unless both devices were using Google’s own Messages app.

I have also noticed recently that Google Messages and Apple Messages now do a pretty good job of supporting each other’s tapbacks. And that hasn’t done anything to really change the green/blue messaging dynamic. Both things are true: RCS makes cross-platform messaging way better and iMessage remains vastly superior to RCS.

What I’m most interested about with Apple’s implementation of RCS encryption is how they’ll indicate it visually in chats. It’s not going to be with blue bubbles. Blue means “iMessage”, not “encrypted” — it just happens to be that iMessage started as a protocol based on end-to-end encryption. There’s no such thing as a non-encrypted iMessage — it’s part of the protocol, and always has been. But what happens when new/updated Android phones support the new RCS encryption spec, and older devices don’t? A lock icon for the encrypted chats? If it were up to me, iOS would drop support for non-encrypted RCS — iOS should use RCS with E2EE for every device that supports it, and fall back to dumb old no-encryption-at-all SMS for all devices that do not.

Apple Did Demo Swift Assist at WWDC Last Year, and Has Shown It, Under NDA, Since Then ★

In an item earlier this week observing that Swift Assist, the most ambitious Xcode-related Apple Intelligence feature shown at WWDC last year, not only hasn’t yet shipped but still is not in beta, I wondered whether Apple actually demoed it live last year. John Voorhees, writing for MacStories from WWDC last June, reports that they did:

Earlier today, I got the very first live demo of Swift Assist, one of the many developer tools introduced today by Apple. I also saw code completion in action. It was an impressive demo, and although the tools seem like magic and will undoubtedly be valuable to developers, they do have their limitations, which are worth exploring. [...]

The code completion demo also included a live demo of Swift Assist. Unlike code completion, Swift Assist requires an Internet connection because requests are sent to the cloud. As a result, it takes several seconds for Swift Assist to return results. The delay was noticeable compared to the speed of code completion, but it wasn’t a painfully long wait either.

I heard this week from a third-party developer who was invited to Apple for a one-day hands-on session with Swift Assist late last year. Swift Assist was definitely working, but seemingly not working too well. From that source: “The UI is very much complete (just like Siri), but the results the LLM produces were not very good. It could make very basic demo apps with a prompt like ‘make an app that takes the NASA satellite JSON and shows the current satellites traveling overhead right now’, but not too much more than that. It fell apart on more complex tasks.”

I remember the remote-inference-only aspect of the Swift Assist presentation from my Xcode briefing at WWDC: that because of its complexity, Swift Assist would not execute locally and would only run via Private Cloud Compute. My own notes on this from WWDC were mostly related to the privacy and security implications. That developers should feel safe using Swift Assist even with confidential code and projects because Private Cloud Compute would be guaranteed private. I also remember thinking, at the time, that I should be more skeptical about Apple’s claims about Apple Intelligence features that would execute locally, on-device, rather than the ones that would execute remotely, via Private Cloud Compute, because the way almost all “AI” features from other companies over the previous two years worked was entirely in the cloud. Apple’s statements that Apple Intelligence will perform much inference locally, on-device, seemed like the stretch goal.

But now in March 2025 I’m beginning to think it’s the other way around. What features and aspects of Apple Intelligence run in Private Cloud Compute, today, in March 2025? Do any? I’ve been poking around for a few days and I don’t have any answers. Is Private Cloud Compute running in production yet? How would we know? If you know, let me know.

Update: “How to Generate a Report of Apple Intelligence Requests Sent to Private Cloud Compute”.

Imagine How Powerful Meta Might Be Today If Their PR Wasn’t Run by Sycophantic Morons ★

Katie Notopoulos, writing at Business Insider (Apple News+ link):

It’s possible that this strident defense is backfiring — creating a “Streisand Effect” that’s publicizing the book even more.

Notopoulos isn’t one to pull punches or hedge, normally, but there’s no question that Meta has Streisand Effected the hell out of Wynn-Williams’s Careless People. I’m on my way to buy a copy this afternoon and I hadn’t even heard of its existence a day ago.

Sarah Wynn-Williams worked at Facebook from 2011 to 2017 — and her book, “Careless People,” details what she said were a bunch of bad things the company did. It also contains allegations that Joel Kaplan — who is now Meta’s chief global affairs officer — sexually harassed her. (Meta said this week that Kaplan had been cleared of the harassment allegations in 2017 after it investigated Wynn-Williams’ complaint.)

The book was released with hardly any pre-publishing fanfare: It was announced by its publisher, Flatiron Books, an imprint of Macmillan, only a few days before it came out Tuesday. That’s an unusually short timeline.

Meta’s reaction has been emphatic and multi-pronged. The company created a page on its website that detailed a series of the claims in the book. It said some of these — like claims about its ambitions in China or its alleged failure to act in Myanmar — have already been reported in the news media as far back as 2017 and have been acknowledged by the company publicly. The company lists seven “new” claims from the book and then offers links to past coverage and company responses — calling the claims “old news.”

Meta’s strongest move was on Wednesday when it won a ruling in arbitration that said Wynn-Williams could no longer promote the book because of a non-disparagement clause in a contract she signed as an employee. The ruling was granted on an emergency basis and is temporary pending the completion of the full arbitration process, The Washington Post reported.

“Read the book that Meta not only doesn’t want you to read, but wants to prevent its author from even talking about” is a hell of a marketing angle. What an own goal.

‘Ted Lasso’ Renewed for a Fourth Season ★

Rick Porter, The Hollywood Reporter:

After more than a year of speculation, Apple TV+ has ordered a fourth season of its Emmy-winning comedy Ted Lasso. The pickup comes after co-creator and star Jason Sudeikis closed a deal to reprise his role as the title character, the manager of the fictional AFC Richmond soccer club in London.

I dig Ted Lasso but this is like 0.1 percent as interesting as When are we going to get Severance season 3?! Gimme gimme gimme. And we haven’t even gotten to next week’s season 2 finale yet.

Best vs. First ★

CNBC story from 2016:

At a recent public appearance at the Utah Tech Tour, in a conversation moderated by Utah’s Senator Orrin Hatch, Apple CEO Tim Cook pointed out that Microsoft had tablets on the market decades before Apple.

Cook emphasized his company’s timing coming to market with new products to underscore the idea that it’s nearly impossible for a company to be the best, the first and to make the most of a given product. [...]

“It doesn’t bother us that we are second, third, fourth or fifth if we still have the best. We don’t feel embarrassed because it took us longer to get it right,” says Cook.

“For Apple, being the best is the most important and trumps the other two by far.”

This has been one of Apple’s guiding mantras for decades, and it has served the company very well. But it stops holding water when they promise to be first, but then aren’t first and aren’t the best.

If you only ever promise A, B, and C — and never mention X, Y, or Z — even when competitors ship their versions of X, Y, and Z first, your silence speaks for itself. Either you don’t think X, Y, and Z are important, or, you think it’s worth taking more time to get them right. But if you promise A, B, C, X, Y, and Z, and then only ship A, B, and C, you just look lost when competitors ship X, Y, and Z.

Jay Peters, last month for The Verge:

Google’s next I/O developer conference will take place on May 20th and May 21st, the company announced today. The event will be “open to everyone online” and will include “livestreamed keynotes and sessions,” according to an FAQ. Like past years, there will also be an in-person component at the Shoreline Amphitheatre in Mountain View, California.

That’s still two months off but I got interested in the dates for I/O this week. After we recorded Friday’s Dithering (on which we talked about my “Something Is Rotten in the State of Cupertino” piece and, more so, the reaction to it, the resonance it seemed to strike), Ben Thompson and I were spitballing, and it popped into my head that Apple’s “more personalized Siri” delay is a marketing gift to Google.

What I would do if I worked at Google is prepare a live demo of Google Gemini on a Pixel phone doing exactly what Apple showed in last year’s announcement at WWDC, and then again in the Bella Ramsey TV commercial that Apple pulled from YouTube. Something like this:

Presenter: This is a live demo, on my Pixel 9. I need to pick my mom up at the airport and she sent me an email with her flight information. [Invokes Gemini on phone in hand...] Gemini, when is my mom’s flight landing?

Gemini: Your mom’s flight is on time, and arriving at SFO at 11:30.

Presenter: I don’t always remember to add things to my calendar, and so I love that Gemini can help me keep track of plans that I’ve made in casual conversation, like this lunch reservation my mom mentioned in a text. [Invokes Gemini...] What’s our lunch plan?

Gemini: You’re having lunch at Waterbar at 12:30.

Presenter: How long will it take us to get there from the airport?

Gemini presents a Google Maps driving directions popup window showing it will take 21 minutes.

Then do another live demo with the “What’s the name of the guy I had a meeting with a couple of months ago at Cafe Grenel?” example from Apple’s now-pulled Apple Intelligence commercial. The exact same demos, but real: live and on stage. These would be great demos even if Apple had never promised to deliver them. But given that Apple did promise them for this year, and has now delayed them until “the coming year”, they’re devastating if Google can show them actually working on Apple’s own original timeline.

Stick to what Gemini can actually do, on actual Pixel phones running the new beta software — which by some accounts already includes the delayed personalized features of Siri. The point should be showing Google AI technology, accessing personal data from Google cloud services, using Google devices, delivering on the promises made by Apple a year prior. But there’s no need for Google’s presenters to mention Apple or Siri, or even mention unnamed “competitors”. Everyone watching the I/O keynote will recognize those demo prompts and draw the competitive conclusions for themselves. Nothing comes across as confident like not even acknowledging, let alone naming, your competition. And nothing serves as proof like a live working demo.