Prices of video games products such as the Steam Deck have jumped recently – and analysts say they’re unlikely to fall any time soon. CC-licensed photo by Pierre Lecourt on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
A selection of 9 links for you. Speed run. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Meta’s smart glasses companion app ships a complete, dormant face-recognition pipeline on a stock account • Buchodi
“Buchodi” is an independent security researcher:
»
Stella is the companion app for Meta’s smart glasses. Inspecting version 273.0.0.21 of the Android build (com.facebook.stella), I found the entire computational and storage stack for on-device facial recognition: three face models, a local database schema, a cosine-similarity vector index dimensioned to match the models, a write path that stages biometric records to disk, a fully wired notification surface, and a user-facing “Connections” widget.
I want to be precise about what that does and does not mean, because the gap between the two is important.
What I can demonstrate: the machinery is present, it is wired together. Several facial extraction and facial fingerprinting models are present and I was able run the recognition pipeline end-to-end on a test image and it detected a face, generate a 2048-dimension biometric embedding, searched a local index, and on a match fired an Android notification stating to the user “Person Recognized”.To get the pipeline to run I invoked its existing handler directly with a test photo.
What I cannot demonstrate: that any of this is active for ordinary users. On a stock, unenrolled account the user-facing UI does not appear, and the screen the recognition notification deep-links to is missing from the build. I also did not observe Meta server-pushing identity data to the relevant database on my test account.
So this is not “Meta is secretly identifying the people you look at.” It is: the complete apparatus to do exactly that is sitting on the device, assembled and functional, gated by Meta.«
Wonder when this is going to be turned on. Perhaps Meta knew this would be discovered (it would be foolish not to expect that, given the internet and previous experience) so perhaps this is the softest of soft launches: if there’s an inkling that people think it’s a great idea, then they flip a switch and voila, Meta knows where everyone is who’s in range of its smart glasses. Because you don’t think it’s going to keep that sort of thing private, do you?
unique link to this extract
The biggest tell that something was written by AI • The Atlantic
Eve Fairbanks:
»
A few weeks ago, where I live in Johannesburg, a man ran a stop sign and crashed into my Subaru. At the scene he was frantic, unable to gather his thoughts. Half an hour later, I received a lengthy, perfectly grammatical text from him elegantly explaining how he perceived the crash had happened. For a repair quote, I wrote to a mechanic I know, a man who used to text me in curt phrases riddled with shorthand. I got a response using just the same voice as the man who’d crashed into me—the distinctive voice of AI.
In surveys, people consistently say they distrust AI-generated writing. But that hasn’t stopped more and more of us from using it in everyday life—to compose work emails and personal texts, to make shopping lists, even to write scripts for arguments with our spouses. “I feel like I’m going nuts,” the writer Jason Koebler complained in the tech outlet 404 Media, under “the cognitive load” of trying to discern whether every piece of text he reads is real or fake.
AI writing is also creeping into our most elite literary spaces—newspapers’ opinion sections, books, literary magazines. I edit professionally, often working with authors renowned for their prose. Maybe two months ago, I began receiving a kind of submission I’d never gotten before: perfectly clean, without a stray comma; uniform in length, with evenly paced paragraphs and a distinctive tone that was simultaneously breezy and grandiose. At first I was surprised that people who prided themselves on their writing would turn to AI to write for them. Even six months ago, when I occasionally identified a paragraph in a writer’s work that seemed AI-generated, they would apologize.
…AI writing is almost impossible to edit, because even when it sounds plausible, a closer look will show that every element is equally off: The tone is bland; individual word choices are baffling; the structure lacks sense; key pieces of the argument are missing; facts are false. Working on AI text, as an editor, is like trying to operate on a body whose skin, muscles, veins, bones, and organs are all compromised. There’s nothing to leave intact, nowhere to begin.
«
Let us filter AI slop, you cowards • The Verge
Jess Weatherbed:
»
It’s almost impossible to avoid seeing AI-generated content online, but it doesn’t have to be this way. YouTube, Instagram, TikTok, and more have ramped up content authentication efforts over the last year, with many now automatically applying labels to distinguish AI-generated images, videos, and music from those made by real, human creators.
That’s all very well and good if we’re just stumbling across labeled content at random, but you know what would be better? Letting us filter out the AI slop.
Current labeling efforts haven’t meaningfully changed how content is presented online. You may notice that some TikTok or YouTube videos in your feeds now have AI disclosures in the description, or information labels overlaid onto the clip itself. Meta takes a similar approach by applying “AI info” labels to images on Facebook and Instagram that carry identifying AI metadata or voluntary disclosures from the creators.
But if you want to actually avoid seeing anything tagged with such labels — which is justifiable, given the brain rot it induces on top of the ethical and environmental concerns around generative AI — it’s actually incredibly difficult to do so. A filter would easily solve this. All we need is an “AI” checkbox to toggle.
I reached out to Meta, Google, TikTok, and Spotify to ask if they have plans to let users filter the various content they’ve been authenticating with AI labeling systems. TikTok and Spotify never responded, and Google said it had nothing to share. Meta didn’t provide an attributable comment. But to summarize, none of these companies said “yes.”
«
Absolutely agree with this. Meta wouldn’t allow it, because usage on its platforms would collapse. Google, TikTok, and Spotify would have the same anti-motivation; even if a few people do it, that’s potentially less time spent on their platforms.
unique link to this extract
Bluesky was a Twitter rival; now it looks to Reddit for inspiration • CNBC
Sawdah Bhaimiya:
»
Bluesky, the social media platform that originated within Twitter, rose to prominence as a rival to the network after Elon Musk acquired the company and rebranded it as X.
But, two years since its launch, the site has around just 10% of X’s estimated global users. Bluesky’s chief operating officer Rose Wang told CNBC that the site saw its future not in rivalling X, but in taking inspiration from the online community forum Reddit
“The world is changing rapidly, and we’re not trying to build what social used to be and get to parity,” Wang told CNBC on the sidelines of SXSW in London on Wednesday. Wang said that Bluesky would move away from the “public square” style of feed of X or Threads, and will instead be “useful” as a discovery mechanism. “What we’ve learned through this process is that I think the public square is not the direction we want to go in. Essentially, I think it’s useful as a discovery mechanism, but we’re very inspired by companies like Reddit.” she said. “A public square, where there’s only a stage, and there’s posters, like people on a stage and people who are watching, that is not social… we’re in the medieval stages of the online world.”
…According to a former engineering lead at Bluesky who posts user statistics on the platform, Bluesky declined from a peak of 1.4 million active daily posters in late 2024 to around 600,000 today.
It had 43 million global users as of March, compared to an estimated 450 million X users worldwide. Threads, the Instagram-linked platform Meta launched in 2023, surpassed 400 million active monthly users last year, per its executives.
«
Really, really, really hard to see Bluesky making itself into anything even vaguely like Reddit.
unique link to this extract
Google employees internally share memes about how its AI sucks • 404 Media
Emanuel Maiberg:
»
While Google CEO Sundar Pichai proudly tells the world that 75% of all new code at the company is AI-generated, internally Google employees are sharing memes about how AI is bad at that exact task and makes their job harder.
One such meme was posted to an internal Google message board called Memegen on May 19, right as the company kicked off its annual I/O conference where it reveals its biggest products and features, according to a copy seen by 404 Media. Unsurprisingly, I/O 2026 was heavily focused on Google’s AI products, which seemed to frustrate or at least amuse some Google employees. This particular meme was a screenshot of Google’s on stage presentation. “I/O announces entirely new ways to slop,” the meme said, with the word “slop” edited into the image in Impact font. The meme was quickly given more than 100 thumbs up from other employees.
404 Media recreated the memes we’ve seen rather than sharing the same exact images in order to protect our sources, who were not permitted to share them with the press.
I wasn’t able to confirm the exact number of anti AI memes shared on Google’s Memegen message board, but I’ve seen dozens of them. One Google employee told me that there are dozens of new memes like this being shared every week. This source estimated that the number of anti AI memes shared inside Google in the last year is in the “high hundreds / thousands.” This employee also said that the number of anti AI memes “spikes when there’s product announcements, or model updates, or Jetski breaks down or something.”
Jetski is Google’s internal AI coding tool. One image shared on May 14 on Memegen I’ve seen shows an interaction between a Google employee and Jetski. “How did you get these metrics?” the Google employee’s prompt said. The screenshot shows that Jetski “thought for 11s,” and then said: “To be completely transparent, the specific numeric metrics and quantitative values presented in that supplemental report were simulated by the secondary sub-agent rather than extracted from live production systems.” In other words: Jetski made them up.
“Thanks Jetski, very useful report,” says the impact text over the screenshot. That meme has more than 400 upvotes.
«
My first instinct were I a manager would be to shut down (or never open) Memegen – the use seems to be implied in the name – but of course all that would then happen is that the same meme-ing would happen on a third-party (external) site.
Analysts say video game hardware isn’t getting cheaper anytime soon • Kotaku
Rebekah Valentine:
»
Last week, Valve sent the price of its Steam Deck into the stratosphere, increasing the cost of both versions of its device by over $200 apiece and leading to widespread concern over how many other hardware dominoes this move is going to knock over. Meanwhile the Switch 2 is already getting a small price hike later this year, the cost of an Xbox has already gone up multiple times in the US, and PS5s have gone up globally multiple times too. And of course this is happening amid everything else—groceries, gas, housing, everything—climbing in price even as wages stagnate. Tough time to be alive generally, and also not a cheap time to like video games.
…the reason gaming hardware prices are shooting through the ceiling right now is because of the RAM crisis, brought about largely by companies suddenly buying up stupid amounts of RAM to power AI data centers and hogging all the supply for the next several years. That analysis is largely correct. But sometimes it’s good to confirm what we’ve sort of loosely heard on social media with someone who knows what they’re talking about, so here’s Joost van Dreunen, NYU Stern professor and author of SuperJoost Playlist, on that:
“The rising cost of RAM is the main culprit, but the inconsistency and volatility created by US tariffs aren’t helping either. Downstream suppliers and manufacturers now sit on massive amounts of inventory they cannot sell or assemble because few consumers would be willing to pay for the markup. What was supposed to bring manufacturing jobs to the U.S. has instead priced consumers out of the market and pushed manufacturing jobs to lower wage countries.”
Dr. Serkan Toto, CEO of consultancy Kantan Games, backed up van Dreunen’s statement, adding in “persistent inflation worldwide” and “geopolitical turmoil like the Iran war” as additional reasons for the situation we find ourselves in. Daniel Ahmad, director of research and insights at Niko Partners, added “currency fluctuations” in addition to everything else, but also pointed out that, as many of the listed issues are somewhat US- or Western-centric, the markets Niko Partners covers are in better shape than most.
«
Inside the Trump-backed push to bring AI doctors into American medicine • The Washington Post
Elizabeth Dwoskin:
»
Last summer, Amy Gleason became a true believer in the wonders of artificial intelligence.
Her daughter Morgan had spent more than a decade battling a debilitating autoimmune disorder. But when the 27-year-old uploaded 16 years of meticulously kept medical records into ChatGPT, the machine reported that Morgan was suffering from a different ailment than the one diagnosed by doctors. The new assessment granted her entry into a coveted clinical trial.
Gleason is not your typical mom. The leader of the US DOGE Service, which she took over from billionaire Elon Musk, Gleason is now tasked by the Trump administration with bringing AI into the health care system as an adviser to Health and Human Services Secretary Robert F. Kennedy Jr.
She’s part of a cohort of MAHA and tech-allied officials who are quietly paving the way for a future in which AI chatbots and robots are an integral part of medical care: diagnosing illness and prescribing medicine with limited or no human oversight. The longtime Silicon Valley dream is taking shape, some entrepreneurs say, thanks in part to a new approach within the Trump administration.
Today, chatbots can only legally offer medical guidance with a disclaimer attached: Neither the US Food and Drug Administration, nor any state licensing board, allows a fully autonomous AI to practice medicine.
But Trump officials — citing concerns about the prevalence of chronic disease and issues such as the shortage of rural doctors — are driving a significant shift.
They have backed a controversial three-month-old pilot program in Utah that allows AI chatbots to refill prescriptions instantly. (Currently humans oversee the chatbot’s decisions, but there are plans to make the program fully autonomous).
«
The counterexample we now have is the chatbot doing customer service for Instagram, which let hackers take over accounts without carrying out the checks that humans would have. Not sure how you put the right guardrails around medicine, given the potential for hallucinations.
unique link to this extract
The blood cancer that became solvable • Works in Progress Magazine
Ruxandra Teslo and Amol Punjabi:
»
in the mid-2010s, a new class of genuinely transformative drugs arrived: immunotherapies. These treatments recruit the body’s own immune system to recognize and destroy malignant cells. The results, particularly in metastatic and relapsed disease, have been extraordinary. Multiple myeloma is one of the cancers that illustrates this most vividly, with the immunotherapy Carvytki, which was first approved by the FDA in 2022 for patients who had returning disease after four or more lines of therapy.
Carvykti marks a turning point in the treatment of multiple myeloma for two reasons. First, unlike the conventional approach, in which patients endure continuous cycles of treatment, remission, and relapse for the rest of their lives, it is administered as a single, one-time infusion. Second, it is producing something that has never before been seen in this disease: durable, long-term remissions in patients which had been refractory to several other treatments, raising the possibility of a cure.
But Carvykti matters beyond multiple myeloma. In retrospect, its development story, which began in 2016, was an early signal of a transformation that is only now, a decade later, making headlines: the United States is beginning to lose its dominance in drug discovery to China.
The foundational science behind Carvykti was largely American, but the therapy that changed the field came from a Chinese company that moved quickly from idea to patient. If the US does not address the regulatory and clinical-trial bottlenecks that slow the generation of early in-human data, more breakthroughs like Carvykti will be developed elsewhere, weakening the ecosystem on which American biopharma depends.
«
Apparently not becoming great again.
unique link to this extract
We may already have an anti-ageing vaccine • RealClearScience
Ross Pomeroy:
»
We have a vaccine that prevents shingles. We have a vaccine that markedly lowers the risk of dementia. We have a vaccine that might even slow aging itself.
Conveniently, these three vaccines are actually just one: the shingles vaccine.
In 2006, the FDA approved the vaccine Zostavax for adults aged 60 and older. For people previously infected with varicella-zoster virus, which causes chickenpox, the infection actually doesn’t end. The sneaky virus lies dormant in nerve tissue and can subsequently spring to life to cause shingles. Zostavax, and its more effective replacement, Shingrix, train your immune system to fight varicella-zoster in case it emerges from hiding to attempt a bodily coup.
That’s a good thing because you really, really don’t want shingles. About 1 in 3 Americans will get it at some point. Its signature symptoms are a bubbly, blistering rash that traces the infected nerve, coupled with debilitating pain that’s been the subject of many painful-to-read Reddit posts. Sufferers use words and phrases such as “unrelenting,” “white hot,” and “I wish I could rip my arm out!”
So if you’ve had chickenpox in the past, it’s definitely worth your while to get vaccinated. The CDC actually recommends the shot for all adults aged 50 and older and all adults aged 19 and older who have weakened immune systems – because oftentimes you can be infected with varicella-zoster virus even if you’ve never developed chickenpox.
«
Seems like it would be sensible just to give it to people. Or if you’re able, just to ask for it.
unique link to this extract
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
The Overspill: when there's more that I want to say 