At WWDC, the corporate insists that its AI transformation is properly underway — however the proof feels somewhat skinny

This is this week’s free version of Platformer: a have a look at Apple’s high-stakes Worldwide Builders Convention. After a bruising yr, the corporate is looking for to reassure buyers that higher days are forward — however the story feels somewhat skinny.
Leaving Substack and Twitter means we’ve got fewer locations we are able to promote our work. Wish to kick in a couple of dollars to assist us? In that case, take into account upgrading your subscription today. We’ll e-mail you all our scoops first, like our current one about Meta’s new hate-speech guidelines. Plus you can talk about every immediately’s version with us in our chatty Discord server, and we’ll ship you a hyperlink to learn subscriber-only columns within the RSS reader of your selection.
This column touches on AI. My boyfriend works at Anthropic. See my full ethics disclosure here.
It has now been a yr since Apple executives stood on stage and heralded a revolution to their platform that will be powered by synthetic intelligence. Or, as the corporate sought to model it, Apple Intelligence — a model of AI that was each extremely personalised to you and your units, and extra protecting of your privateness than its rivals. Apple Intelligence would make the most of the corporate’s deep data about its customers, scouring their emails, texts, and contacts to route requests across the working system and take actions in your behalf.
It was an interesting imaginative and prescient. However it never shipped. In March, amid mounting questions over when a bunch of options on account of ship in 2024 would possibly truly arrive, Apple lastly squeaked out an update: they might now arrive “within the coming yr,” a lot later than deliberate, and lengthy after Apple had aired tv adverts selling them.
It was extra unhealthy information for Apple in a yr that has seen loads of it, from a failed tug-of-war with President Trump over tariffs to being excoriated by a federal judge over its malicious compliance in an antitrust case, who referred the corporate to prosecutors for a prison contempt investigation.
All of which set the stage immediately for an unusually high-stakes Worldwide Builders Convention. Apple stays one of many world’s greatest firms, in fact, and its merchandise largely proceed to be beloved. However for the primary time in a very long time, there is a way that Apple has begun to lose the plot — too slow to acknowledge the potential of AI, and never succesful sufficient to understand it.
In some methods, Monday’s keynote presentation appeared to verify a few of these suspicions. Lots of the bulletins have been nearly comically modest: a brand new naming scheme for the corporate’s working methods! A Telephone app for the desktop! Home windows on the iPad you can resize!
There was additionally, entrance and heart, a streak of defiance. After CEO Tim Cook dinner welcomed individuals to WWDC by selling an upcoming Apple film about race automobiles, software program engineering chief Craig Federighi started his remarks with the AI options that Apple did ship final yr: writing instruments, a ChatGPT integration, and 15 different issues that I largely needed to look as much as bear in mind what they did. (Do you know what the Image Wand is?)
There have been additionally guarantees of latest AI options to come back. One replace will allow you to ask ChatGPT about anything you’re seeing on your screen. One other will give third-party developers access to on-device massive language fashions. And its Xcode growth device will get code completion help from OpenAI, although builders also can herald their very own fashions utilizing API keys.
That is all advantageous, although given what we’re seeing elsewhere — LLMs working unsupervised for seven hours to refactor code bases, for instance — it hardly looks like the cutting-edge. Extra disappointingly, I feel, the numerous Apple bulletins on Monday did not cohere round any explicit imaginative and prescient for its expertise other than “use it extra.”
At different occasions, Apple has sought to encourage with visions of how expertise can uniquely advance human progress. However the Apple of “Suppose Completely different” and computer systems as bicycles for the thoughts was exhausting to see on Monday. As a substitute, we noticed a protracted procession of information employees engaged within the tedious meta-work of contemporary life: rearranging home windows on an iPad; polling the group chat about occasions to satisfy; establishing a digital clock.
Distinction that with Google, which used its developer conference this year to indicate an agent that browses the online in your behalf, an AI filmmaking device already being utilized in manufacturing, and Gmail writing in your voice. A few of these options are presently vaporware. However there is a clear imaginative and prescient there.
In fact, Apple has its personal little productiveness boosts to supply. My favourite factor Apple confirmed off on Monday was an replace to Highlight — the built-in launcher on macOS. Whereas it was initially a novel approach to seek for information in your Mac, Highlight has lengthy been changed on my computer systems with extra full-featured launchers like Alfred and Raycast.
The subsequent model of Focus will take a step in Raycast’s course — evolving from a easy launcher to a type of command heart to your laptop. Along with file search, you can take actions, reminiscent of sending messages. You will additionally be capable of run shortcuts. There’s nothing there that will trigger me to uninstall Raycast, however this is the type of stuff I wish to see Apple do: make critical however easy-to-use options for individuals who wish to get stuff executed somewhat extra effectively.
On the opposite finish of the spectrum, for me anyway, was “Liquid Glass” — a redesign of Apple’s working methods designed to make every part translucent and borderline unreadable. I have never seen the design up shut and know that reflexive dislike of redesigns is so widespread as to be cliché. And but I am struck by, regardless of all of the minutes spent on stage discussing Liquid Glass, how little of that point was spent explaining what it was meant to do.
We heard endlessly about how fairly it is, how it displays the sunshine, and the way it shifts effortlessly into darkish mode. However we heard nearly nothing about what the design is meant to do.
On the danger of lapsing into one other cliché — quoting Steve Jobs — that looks like an oversight. In 2003, Jobs had this to say to the New York Occasions, about the design of the iPod: ”Most individuals make the error of considering design is what it appears to be like like. ‘Individuals suppose it’s this veneer – that the designers are handed this field and advised, ‘Make it look good!’ That is not what we expect design is. It is not simply what it appears to be like like and appears like. Design is how it works.”
Liquid glass appears to be like just like the product of a design course of led by somebody saying “Make it look good.” I am certain it “works” advantageous. However I am unsure what it is meant to do this its predecessor didn’t.
How a lot does any of this matter? Even on this unsure footing, Apple stays a monolith. Most individuals not often change gadget ecosystems, and even Google has but to determine an AI function highly effective sufficient to spur massive teams of iPhone homeowners to commerce them in for Pixels.
Apple nonetheless has loads of new widgets it can exhibit. However the imaginative and prescient behind all that exercise feels somewhat muddy — and the corporate’s rivals, together with Google and OpenAI, appear to have a a lot stronger thought of what they’re doing. Apple can nonetheless do nice issues — however for now, it appears fixated on the low-hanging fruit.

Sponsored
NewsGuard Audit:
AI Fashions Fail on Information Subjects 20% of the Time

Get the audit to see why the AI fashions aren’t trusted on information matters. Our public audits don’t present the person scores of the chatbots.
However, in the event you work at an AI firm, we’ll disclose your fail charge to you.

The Musk-Trump blow-up
Elon Musk relaxed his marketing campaign in opposition to President Trump over the weekend, turning the eye of his X feed to an curiosity he nonetheless shares with the administration: whipping up anti-immigrant sentiment. On Sunday he reposted certainly one of Trump’s Fact Social posts, seeming to sign some type of intention to make up.
Trump, for his half, has performed down the importance of Musk’s departure. However on Saturday he told NBC News that “there could be ‘critical penalties’ if tech mogul Elon Musk funds Democratic candidates to run in opposition to Republicans who vote in favor of the GOP’s sweeping finances invoice.” And the administration is involved about the security implications of putting in Musk’s Starlink on the White Home.
Elsewhere:
The retreat from content material moderation
The Occasions reviews that YouTube quietly relaxed its community guidelines in December in order that now as much as half of a video can violate the platform’s pointers and stay on the platform beneath an exemption for discussions of topics within the public curiosity. No matter what you make of the adjustments right here, it’s extremely cowardly of YouTube to not have made any of this public. If you need individuals to observe your guidelines, first you need to inform them what they’re.

Governing
- An investigation into how a DOGE staffer with no medical experience used outdated and inexpensive AI models that made mistakes to identify which Veterans Affairs contracts to terminate. (Brandon Roberts, Vernal Coleman and Eric Umansky / ProPublica)
- A profile of Dominick McGee, a right-wing influencer who posts conspiracy theories to his X account and one of the members of new media invited to press briefings at the White House. Regardless of fervent pro-Trump posting, he nonetheless struggles to pay his payments. (Stuart A. Thompson / New York Occasions)
- Trump is reportedly expected to sign an executive order extending the TikTok deadline again, as a mid-June deadline approaches. When you’ve determined to disregard the regulation you possibly can simply prolong issues so long as you need. (Alex Leary / Wall Road Journal)
- The Trump administration is reportedly seeking a potential consent decree in an FTC deal that could force Meta to pay restitution to conservative users and businesses they claim were harmed by content moderation. This is full nonsense, however anticipate Meta to pay it anyway. (Lydia Moynihan / New York Submit)
- A Trump crypto wallet and trading app has been shelved, but World Liberty Financial will buy a “substantial position” in the Trump memecoin, Eric Trump said. (Jason Shubnell / The Block)
- Trump signed an executive order reversing many of Biden’s cybersecurity initiatives, including ones that would have prioritized AI testing for cyber defense, required software vendors to comply with new federal standards, and accelerated the rollout of quantum-proof encryption. (Eric Geller / Cybersecurity Dive)
- Trump reportedly authorized Treasury Secretary Scott Bessent’s team to negotiate away recent restrictions on the sale of technology and other products to China. (Lingling Wei / Wall Road Journal)
- A slew of new apps aimed at providing immigrants support and legal information is emerging amid Trump’s attacks. (Gabriela Martínez Córdova and Alicia Fàbregas / Remainder of World)
- Meta hosted hundreds of ads promoting “nudify” apps on its platforms, this investigation found. The corporate has since eliminated numerous them, it stated. (Emmet Lyons and Leigh Kiniry / CBS Information)
- Criminals are increasingly using “residential proxy” services to hide and mask IP addresses among everyday online activity, researchers warn. (Lily Hay Newman / Wired)
- Two suspects in last month’s bombing of a fertility clinic in Palm Springs, Calif. used an unspecified generative AI program to help plan the attack, the FBI said. (Annie Palmer / CNBC)
- A High Court in the UK warned lawyers that could face criminal prosecution for presenting false material generated by AI. And perhaps they need to? (Lizzie Dearden / New York Occasions)
- AI regulation proposals in the UK are being delayed by at least a year as ministers look to introduce a more comprehensive AI bill to address safety and copyright. (Eleni Courea and Kiran Stacey / The Guardian)
- Getty Images said its landmark copyright lawsuit in the UK against Stability AI does not pose a threat to the generative AI industry as Stability AI suggested. (Sam Tobin / Reuters)
- China’s most popular AI chatbots like Alibaba’s Qwen and apps like Tencent’s Yuanbao and Moonshot’s Kimi temporarily suspended its functions and photo recognition services to prevent students from cheating during the country’s annual college entrance exams. (Luz Ding / Bloomberg)
- Starlink has reportedly secured its license to roll out its satellite internet services in India. (Santosh Kumar and Shwetha Sunil / Bloomberg)

Trade
- OpenAI has hit $10 billion in annual recurring revenue, it said, which excludes licensing revenue from Microsoft and large one-time deals. (Ashley Capoot and Kate Rooney / CNBC)
- A look inside a secret math meeting, where leading mathematicians were surprised by how the o4-mini AI chatbot could solve the complex problems they devised. A captivating case of some AI skeptics coming round. (Lyndie Chiou / Scientific American)
- ByteDance has reportedly replaced US staff with managers connected to China as its TikTok Shop sales missed its previous goal of $17.5 billion in transaction volume last year. Greater than 100 TikTok Store workers within the US have been fired. (Bloomberg)
- X is partnering with Polymarket, naming it the platform’s official prediction market partner. (Aisha Malik / TechCrunch)
- Usage of X’s Community Notes has plummeted this year, this analysis found. One cause that you do not wish to base your content material moderation system on volunteer efforts is that the more severe your platform will get, the extra that folks will cease volunteering. (Joe Murphy / NBC Information)
- Meta is reportedly in talks to make an investment in AI startup Scale AI that could exceed $10 billion in value. (David Carnevali, Shirin Ghaffary, Katie Roof and Kurt Wagner / Bloomberg)
- Google is rolling out “scheduled actions” for Gemini AI Pro and AI Ultra subscribers, which will let users ask the AI assistant to perform tasks at specific times. (Emma Roth / The Verge)
- YouTube is rolling out a new feature that will let creators upload different thumbnails for videos dubbed into a different language. (Dan Whateley / Enterprise Insider)
- Microsoft will soon add a “safety” category to its leaderboard of AI models and rank them to build trust with cloud customers. (Rafe Uddin and Cristina Criddle / Monetary Occasions)
- Amazon plans to invest at least $20 billion in Pennsylvania to expand data center infrastructure. (Reuters)
- Snapchat now has an app for the Apple Watch, which will let users preview incoming messages and reply to them, but will not let users send photos. (Andrew Liszewski / The Verge)
- Duolingo CEO Luis von Ahn said he didn’t expect the amount of blowback he got for going “AI-first,” and says it was mischaracterized – AI is not going to replace the company’s workforce it said, just the jobs of a “very small number of hourly contractors.” (Emma Jacobs / Monetary Occasions)
- This year’s virtual FediForum conference introduces new apps, including Bonfire Social, Channel.org and Bounce, which let users have new ways to move around the open web. (Emma Roth / The Verge)
- AI research organization EleutherAI said it released one of the largest collections of licensed and open-domain text for AI training. (Kyle Wiggers / TechCrunch)

These good posts
For extra good posts day-after-day, follow Casey’s Instagram stories.

(Link)

(Link)

(Link)

Discuss to us
Ship us ideas, feedback, questions, and Liquid Glass designs: casey@platformer.news. Learn our ethics policy here.