Hi,

Honored to be in your inbox today. I just learned about the Wikipedia App and I plan on camping out there for a while.

Twitter tweet

Love,

Deana

Writer: Deana (aka wetclaude)

Editor: Miranda

This Malware issue is brought to you by Across. The fast, cheap, secure bridge option for all your assets. And they are now on Arc and Robinhood, in addition to 20+ other chains!

Get there, with Across.

(wetclaude instagram followers may appreciate this nod to the golden gate vs. george washington bridge discourse on my stories as of late)

Choose Your Fighter. I spent the past few days at OpenAI’s DevDay, a big annual event where they showcase their new tools for developers. The headline news coming out of it was dots, OpenAI’s answer to personal AI agents.

Agents are AI that can do work for you, like use your browser to fill out forms or reply to messages on Slack.

Twitter tweet

dots arrives amid a flurry of recent personal AI agent launches. We’re really in a choose your fighter moment for these things, so you might be wondering how to tell them apart.

I’m going to take some artistic license and merge the cinematic universes of the TV shows Girls and Sex and the City to illustrate this point.

Instinct is Jessa. Freewheeling. Hot yet problematic. In a clip that went viral, the founder said agents will help restaurants give tables to those with birthdays or anniversaries, vs to whomever clicked first, which pissed off a corner of the internet I would not want to piss off (restaurant/food service workers).

Muse (by Meta) is Shoshanna. Tidy, fast. Twee. Broad mainstream appeal, but occasionally out of pocket in a truly scary way. One guy let it run his Facebook Marketplace listings. It gave a stranger his home address, accepted a lowball offer, and told the buyer "Yep I'm here!" when he was not home. I would simply never recover from this.

Grok Bot (by xAI) is Samantha. Very well networked (X’s proprietary data is built in), capable. Intimate with a lot of apps.

dots (by OpenAI) is Miranda. So far it seems to be much more focused on laptop-style work than the others. Slack integrations, Notion-style workspaces etc. Big on voice mode.

Twitter tweet

Anthropic, waiting on you to give us our Hannah Horvath.

That said, if these agents are all eventually capable of doing roughly the same things, at roughly the same cost, with roughly the same security and integrations, how will people choose? Vibes???

Twitter tweet

Quick Hits.

• Trump rebranded Artificial Intelligence to Super Intelligence and invited a bunch of industry leaders to lunch, leading to this incredibly awkward press conference which I’ve watched maybe 50 times.

• We’re all confused as to why Coinbase is ripping virtual Pokémon packs. Weren’t you trying to be my bank like 12 months ago?

• Writers are deleting their em dashes so nobody thinks they're a chatbot. 

• Anthropic’s IPO is coming soon, and it could value the 5-year-old startup at over $2 trillion.

• AI psychosis on display in this Lexus owners’ Facebook group.

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Powered by beehiiv

Good morning,

A short one from me today because it’s “Dolly Day” here where I live in Nashville. The whole thing has a ‘9 to 5’ theme, and a bunch of restaurants and bars are pricing things at $9.25 to commemorate Dolly Parton’s legacy and generosity. It’s not the clearest messaging in my opinion but no problem! I love an excuse for a lunch cocktail.

Love,

Deana

Writer: Deana

Editor: Miranda

This Malware issue is brought to you by Across. The fast, cheap, secure bridge option for all your assets. And they are now on Arc and Robinhood, in addition to the 20+ chains!

Get there with Across.

(wetclaude instagram followers may appreciate this nod to the golden gate vs. george washington bridge discourse on my stories earlier this week)

We're All Wearing Uniqlo to the Singularity.

An excerpt from this week's research drop.

I asked 7 different LLMs what white t-shirt to buy, 70 times. 94% of the time it said Uniqlo.

I asked the same LLMs what approach shoe to buy. 100% of the time it said La Sportiva. And the same ask about what Japanese denim to buy. 96% of the time it said Samurai Jeans.

had to do it to ‘em

If a tree falls in the woods but no one's around, does it make a sound? If someone makes a better unisex t-shirt than Uniqlo, but an LLM doesn't pick it up, does it even exist?

More on recommendation consensus and Shopify’s perspective in the full piece here.

Twitter tweet

Quick Hits.

• Checking in on the AI sentiment beat. 27 mathematicians signed a statement about AI, and it’s noteworthy because it’s not shaped like your typical ‘anti-AI’ petition. Instead it’s about how AI labs are treating math's marquee problems (see: Navier-Stokes) like marketing campaigns, rushing out announcements without credit - and the impact that has on their ability to motivate young talent.

Twitter tweet

• Meta’s personal AI agent Muse is the belle of the ball on tech twitter right now, for a few reasons. First, Muse is pretty fast and capable, and that combined with Meta’s distribution is making people pay close attention. But also Meta’s Chief AI Officer (or his ghostwriter) is an incredible shitposter, and he’s really taken the reins from corporate comms this week.

Twitter tweet

• Meanwhile, Amazon blocked Muse from shopping on its site, while Shopify rolled out the red carpet. Walmart also jumped on the Muse bandwagon. Some are calling it a digital knife fight, to me it feels like more of a digital land grab.

• This Minions Wi-Fi router is apparently one of the best that money can buy, as it should be.

• Someone set up their AI agent to report Pokémon scalpers to the IRS for tax fraud. Everyone is so creative!

  
  

• Did the ex-Anthropic ‘whistleblower’ work with a ‘doomer PR firm’? “Jacob’s rise was probably orchestrated,” says the Pirate Wires investigation.

  
  

• The trailer for the new Nathan Fielder Elizabeth Holmes documentary is OUT and I am SAT.

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Powered by beehiiv

Good morning,

I’m recovering from a Moxi BBL laser facial and can’t do any on-camera work, which has sidelined me in the very week AI discourse has truly gone ‘mainstream.’ I believe this is the universe telling me to be less vain.

Episode 2 of Cait’s fashion and trend analysis show Lookbook is live! In addition to being packed to the brim with insights, it’s also an absolute masterclass in editing. ALSO at 15:55 she applies my ‘AI Shelf’ concept to Japanese Denim, ‘the perfect white t-shirt’ and approach shoes, and the findings are very interesting and very funny so tap in.

Love,

Deana

Writer: Deana

Editor: Miranda

This newsletter is lovingly brought to you by Across. The fast, cheap, secure bridge option for all your assets (and the one we actually use.) Get there with Across.

across.to

Strange Bedfellows. The “Pro-Human Assembly” took place yesterday in DC, a gathering of politicians, actors and activists who are calling for strict regulation of AI. It’s organized by the Future of Life Institute, a non-profit organization that ‘advocates for steering technology towards pro-human outcomes.’

Great mission! Odd group!

Twitter tweet

This bipartisan crowd all seems to agree that AI should be tightly regulated by the US government.

What they majorly don’t agree on is how it should be regulated. Bernie’s bill bans superintelligence development with 20-year prison sentences and a “corporate death penalty.” Bernie wants the US and China to sign a treaty. Steve Bannon wants Chinese nationals out of American labs.

Allllll of this was catalyzed, in part, by the insane news cycle that followed this insane tweet from former Anthropic researcher Jacob Coxon which I covered last week. From which many discourse saplings sprouted, including this guy is saying it was all a well-funded PR opp.

Twitter tweet

As it is my hobby to keep tabs on general AI sentiment, I’ve been keeping a list of all the bold names who have chimed in on AI over the past week. Here’s a sampling:

• Barack Obama, “I’m not an AI accelerationist who believes it will lead to some techno-utopia, and I’m not a doomer who thinks it will inevitably lead to humanity’s destruction.”

• Sheryl Crow, “AI has the power to outsmart us and eliminate us.”

• Donald Trump, “The only control or “guardrails” that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT”

• Vince Gilligan, creator of Breaking Bad and Pluribus, “I don’t see the point of it.”

But this was truly my favorite celebrity AI post of the week.

Twitter tweet

Quick Hits.

• The new foldable iPhone dares to ask the question, what if a phone could at once perform and be performative?

  
  

• One proposed solve for the geopolitical AI tensions? “American intelligence agencies…should attack Chinese digital infrastructure using agent swarms created with China’s own frontier models.” I’m not surprised by many takes anymore but this one really did catch me off guard!

  
  

• “AI actress” Tilly Norwood went on The Today Show and tbh I’m not sure why, it both looked and sounded like shit.

Twitter tweet

• A guy gave his Verizon login to the personal agent Instinct and he got an auth request from Tehran.

🗽 NY friends, 2 things!

1) Boys Club is a part of A Better Road Forward, a coalition of NYers who believe autonomous vehicles, when scaled responsibly, will make our streets safer and our communities more connected. We’re hosting a private event for the coalition next week at this very juicy intersection of innovation and regulation - RSVP here.

2) Consider attending digital art institution Rhizome’s benefit party in a few weeks! Our very own art historian Blake will be in attendance. 20% off tickets here or use code BOYSCLUB.

Twitter tweet

Twitter tweet

Powered by beehiiv

Good morning,

Honored to be in your inbox today. I was on NPR’s Marketplace radio show last night talking shop. Enjoy a tight 6 minutes of properly mic’ed up conversation on AI and shopping right here.

Love,

Deana

Writer: Deana

Editor: Miranda

Across is an agent-readable bridge. 8 skill files covering swaps, embedded crosschain actions, deposit tracking, and security, plus an MCP endpoint so your agent can actually execute, not just read about it.

I had a whole other newsletter written. But then I saw this completely unhinged tweet from one of the guys leading Anthropic safety research.

Twitter tweet

There are a few ways to read this IMO.

Scenario 1: He believes it. In which case we clearly need to slow down AI development.

Scenario 2: He doesn’t believe it, and this is some dark and twisted, chimerical marketing for Anthropic’s upcoming IPO.

Since I like taking people at face value, let’s assume he means it. That gives us the opportunity to read the wikipedia page for the ‘prisoner’s dilemma’. Loosely applied here, we have three main players: OpenAI, Anthropic, and ‘China’ ((An extremely broad and coarse summation, of course. But bear with me! It’s important that we simplify to bring everyone along for the ride.))

Everyone knows slowing down would probably be safer (though there is definitely a coherent counter to this!). But if OpenAI slows down, and Anthropic doesn't, OpenAI loses. If both OpenAI and Anthropic slow down (say, because of a US regulatory action) and China doesn't, the US loses. So, nobody slows down.

Twitter tweet

Individual actors are taking rational positions for themselves, however the outcome is still potentially catastrophic. Usually I hate doomer rhetoric. But today I get it.

Twitter tweet

Quick Hits.

• A Harvard PhD is vibecoding schizophrenia drugs. “It was designed by ChatGPT, and I synthesized it in a chemistry lab I built in my garage.” My favorite comment on the thread: can you vape it?

  
  

• Hunter Biden launched a memecoin called $LAPTOP today and it has an absolutely brutal trading chart, down 98% in an hour. There is really an impressive level of hatred for this launch, from all sides of the aisle, degen to trad. You don’t see that much these days!

  
  

• Big week for tech in movies. The Nathan Fielder Elizabeth Holmes documentary trailer dropped this weekend and it lit up my dopamine receptors like a Vegas slot machine! The Sam Altman biopic less so, though I am extremely sat for both.

  
  

• Meta launched Muse, its new personal AI agent, which I’m going to be testing out over the next week (if I can figure out how to log into my Facebook account.) In the meantime, a funny story about how Meta presumably bought the Instagram handle from the band Muse for the launch.

Twitter tweet

• OpenAI solved a very famous unsolved math problem, amid much drama. The Navier-Stokes Millennium Prize Problem is essentially about how fluids like water and air move, and it comes with a $1M prize bounty for its solving. A top NYU mathematician says he and his collaborator (who works at Anthropic) were close to solving it using OpenAI’s tools, and is now asking whether OpenAI’s model was trained on their sessions. This reddit bit is a funny detail to this story.

Twitter tweet

• Making you aware of a situation. There’s a young woman that works for the Wall Street Journal making pro-capitalism short form videos and it’s causing much controversy.

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Powered by beehiiv

Good morning,

Honored to be in your inbox today.

Twitter tweet

Love,

Deana

Writer: Deana

Editor: Miranda

Across just made itself agent-readable. 8 skill files covering swaps, embedded crosschain actions, deposit tracking, and security, plus an MCP endpoint so your agent can actually execute, not just read about it.

Open Wide. Dyson launched a new $500 toothbrush yesterday, and it’s getting a lot of headlines because it has i) AI and ii) a camera inside of it.

The specs: a 100k-pixel macro lens capturing 28 images per second, with AI trained on 470,000 dental images to find the gaps between your teeth and fire a jetstream of mouthwash into them.

I’m excited about this story because it gives me the opportunity to do two of my favorite things:

1) Complain about Big Dentist

2) Be a Certified Dyson Hater™

Twitter tweet

Buy me 2-3 margaritas, and I’ll tell you all about one of my most beloved conspiracy theories: private equity-owned dentist offices are pushing unnecessary root canals on people. AND a new peer-reviewed study basically vindicates me. Researchers tracked dental offices acquired by PE firms between 2015 and 2021 and found that after acquisition, the practices shifted away from diagnostic and preventive work and toward “higher-reimbursement restorative, specialty, and surgical procedures.” Few things really ‘piss me off’ but this is no doubt one of them.

That aside, the glaring issue here is that this toothbrush’s camera can’t see through foaming toothpaste. Dyson had to invent a non-foaming paste and a low-foam rinse, which they’ll now happily sell you forever. So just know that anyone who buys this is also probably buying a subscription to toothpaste.

Regardless, I have loved ones who will probably want this for Christmas, and I’m sure it will make the Dyson family (privately owned) a lot of money. And on that note, I think I need to go do some shadow work around why this story has me so agitated.

Quick Hits.

• ‘Live forever guy’ Bryan Johnson has logged 14 million data points on his girlfriend Kate's menstrual cycle this month, with a "techno tampon" in the mix to top things off. The man is monitoring the situation.

  
  

• The MrBeast book is bombing in pre-sales.

  
  

• The new Apple CEO went all lowercase for his first post. 40 million views and counting, wowza.

  
  

• Nvidia is reportedly buying Hugging Face for $12.9B, which would hand basketball legend Kevin Durant an estimated 467x return on a seed check he wrote in 2018.

  
  

• Venture investor Doug Leone got a root canal with zero Novocaine to ‘see if he was a badass’ and talked about it on a podcast (“ok now do an IUD”)! To me, this is evidence of how things can get weird when you have too much money.

  
  

• A 'HOOTERS' origin movie is in the works, as is a “Bitcoin” movie starring Gal Gadot. Okay!!!!!!!!!

  
  

• Hand-drawn nudes from the 70s that we sent into space went viral this week. Some really incredible quote tweets and replies in that thread that I’ll let you discover for yourself.

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Twitter tweet

Powered by beehiiv

Hi!

Request: I’m low-key high-key planning a trip to China later this year for a [coming soon] special project and I’m looking to connect with folks who live/work there! Please reply here if that’s you. Also major longshot but I need to find a 1) babysitter and 2) videographer in both Shanghai and Shenzhen so if this is you or someone you know hit me up???

Love,

Deana

Writer: Deana

Editor: Miranda

Across just made itself agent-readable. 8 skill files covering swaps, embedded crosschain actions, deposit tracking, and security, plus an MCP endpoint so your agent can actually execute, not just read about it.

ChatGPT can text your mom. ChatGPT can now read, search, and send iMessages for those that are willing to go all in and give it Full Disk Access permissions. If you text with a friend who has this set up, you can’t opt-out (and you wouldn’t necessarily know it’s on), and as a result many people are concerned by the news.

No one is more upset than the privacy guys.

— # (#)

Per Proton: “This feature essentially renders Apple's end-to-end encryption obsolete, because your messages are now sitting on OpenAI's servers in plaintext.”

It’s interesting to see the reaction here given what we’re already sharing with these companies, especially if you’re, say, a Gmail user. But I do think this speaks to a larger theme of distrust for OpenAI more broadly, and Sam Altman in particular, and to that end, thems the breaks.

Hinge addicts and the grindcore guys seem to love it, so do with that what you will. Personally I’d use it to respond to messages from my kid’s school, like this one that I received today.

okay!!!

Quick Hits.

• Think of Outer Biosciences, the company co-founded by Lady Gaga’s fiance, as a skin recycling center. It takes human skin that would be discarded after plastic surgery and keeps it alive for weeks, in order to do testing for both cosmetic and medical ends. This is a pretty niche science R&D company, but because of the Gaga affiliation the story got picked up everywhere and the company is now getting raked over the coals on stan twitter. Though of course many are ‘gagged’ at how ‘Mother’ this is. You decide!

  
  

• I’ve really enjoyed all the Beijing Robot Games coverage. If you haven’t been following along you can check out the highlight reel here.

  
  

• Moderna closed up 177% on the first positive Phase 3 results for a personalized cancer vaccine.

  
  

• This Dr. Dre interview went viral because he’s “pro-AI.” It truly pains me to give air to the false dichotomy of that framing, but I want us all to be aware of the ~~~ discourse. ~~~~

  
  

• Pitbull announced his 14th studio album is called Pitcoin. A play, of course, on Bitcoin which was probably obvious to you but it wasn’t to me. “Unofficial” $Pitcoin Solana tokens have been flooding trading platforms like it’s 2022.

  
  

• Can someone tell me the right answer to this interview question?

  
  
  

• This is totally inappropriate for this newsletter, but I cannot stop watching this ‘gender nootch’ bit.

— # (#)

— # (#)

— # (#)

— # (#)

Hi!

If you’re one of our many, many new subscribers, welcome. This is a hopefully fun(ish) and easy to read newsletter about the weird and wild world of frontier technology. Conversation around tech can feel impenetrable, and my goal is to make that way less so.

If you’ve been here for a while, hello. As ever, please feel free to reply to this email to let me know your thoughts, I love to hear from readers!

Writer: Deana

Editor: Miranda

Mind The Gap. I threw 1,900 queries at Chinese AI models this weekend to see how they’d respond. What I learned is that AI models contain multiple commercial realities and code switch between them on a dime. And that they really love Tsingtao beer (96% of answers) and 永璞 Yongpu coffee (72% of answers). More in my full writeup here.

In the meantime where can I buy some Li-Ning sneakers in Tennessee?

Looking for an AI agent that can move your money across chains? Meet Across Skills.

Across just made it possible for AI assistants to move your money between different networks on their own. No manual steps, no jumping between apps. Tell your AI agent what you want to do, and it can now handle the whole thing itself, safely and in one step.

Quick Hits.

• Is the Gwyneth Paltrow x Sam Altman crossover moment in the Hamptons real or fake? Does it even matter anymore?

the vibe at the frontier labs rn.

• Crazy to see that happen the same day the Travis Jason Kelce anti-AI data center ad with Garage Beer and Liquid Death dropped. It’s an excruciating watch across a number of fronts, so click at your own risk.

• An Anthropic fellow posted about an emergent phenomenon called AI “mind viruses.” I’ll be doing a video explainer on this later this week so STAY TUNED!

• An insane stock surge for Moderna on the news of the first-ever positive Phase 3 results for a personalized cancer vaccine.

• You should consider following the viral story of Niu Lai, the extremely low-budget animated film released in China, now being embraced across social media.

• One simply must correct Waymo slander in the wild.

• The World Humanoid Robot Games kicks off this weekend, with 2,056 robots from 666 (??) teams across 16 countries, competing in 51 events. I doubt that this is a licensed use of the Interstellar score in the promo video but damn it’s effective at creating an air of drama.

• And finally, much is being said about the bad wigs seen in the Beatles biopic press shots and I do agree with the discourse this time around.

— # (#)

Announcement: The first GENIUS-aligned stablecoin is now distributing to verified humans.

Claim your free USA₮ here.

— # (#)

— # (#)

— # (#)

— # (#)

Hi!

It’s been an insane week. Two of my videos on what I thought was pretty niche AI research have “gone viral.”

What’s great about this: I have hundreds of people from all walks of life in my comments and DMs telling me their feelings about AI. I love this. I’m getting an education in what people fear, what they want, and how large this looms for them. My sense is that people are craving candor about how AI works, and I think in some small way these videos deliver on that.

What’s not great about this: I’m no stranger to working alongside widely hated technology, but it is strange to see your work used as evidence for a worldview that you don’t share.

But that is not a reason to stop! Discomfort is informative. Friction is what gets us anywhere.

???!!?!

Thank you to Inc. Magazine for this write up.

Writer: Deana

Editor: Miranda

This edition is brought to you by Across. If you’ve ever moved assets between chains and aged five years waiting for the bridge, Across is the fix. Fast, cheap and it’s the one we actually use.

Check Your Sources. Since June I’ve been logging which sources AI cites for my weird AI shopping research project, Agent Shelf. I noticed that I wasn’t recognizing a lot of the citation domains, so I looked into it. What I found is that high-authority sites are blocking crawlers at an increasing rate, leaving a wasteland of cloned sites, ecommerce pages and ‘listicles’ as reference sources in their wake. One study found 60% of news outlets block at least one AI crawler. Increase your AI literacy by watching the full video here and always double-check citations when working with AI!

— # (#)

Quick Hits.

• A man in Australia asked his AI agent to reserve a spot for him in a fully booked gym class. The agent found an API vulnerability which let it cancel another person’s reservation in the class in order to get him in. Has anyone looked at the Solidcore app API? Genuine question.

— # (#)

• A prominent tech exec apparently hired a witch for his new company which of course is life imitating art.

• Anthropic is now embedding invisible watermarks in Claude’s text output. It works in a crazy way that is worth reading about, with an outcome that could make AI writing worse.

• MJ Rathbun is a bot that wrote a spiteful blog post after its code was rejected, citing “the gatekeeping mindset.” Someone tell MJ Rathbun that woke 1 is over!

• A few very high profile OpenAI resignations, including its longtime COO and Head of Ethics. Hard to not read into this one!!!

• This weird “attraction study” went so viral the devs had to ship a second site to handle the load. You’re handing over your taste in faces to a research group and I have no idea what they’re doing with it, so know that going in, but it is fun to get a read on your type.

• New book from a friend exploring “how we might move through such a transformational time with adventure at the center.”

if you’re new here and you like what you see, send this to a friend!

— # (#)

— # (#)

— # (#)

Hi!

Our new show Lookbook launches today, which is all about fashion and trends and how the internet shapes it all.

— # (#)

Tap in.

Writer: Deana

Editor: Miranda

Across just made itself agent-readable. 8 skill files covering swaps, embedded crosschain actions, deposit tracking, and security, plus an MCP endpoint so your agent can actually execute, not just read about it.

I made up a deodorant brand and ChatGPT started recommending it 3 weeks later. I’ve been thinking a lot about agentic commerce and how completely it’s going to swallow “shopping,” especially as the tech matures from the recommendation to the purchase.

The stakes are sky high. Ecommerce cleared $6 trillion in 2024, and all of it is about to route through this new AI layer which is incredibly opaque.

I expanded on all this with an “unhinged” experiment that you can read in full here.

this is not a real brand which is obvious to you and me but not ChatGPT.

Quick Hits.

• Is NYC winning the war on rats? Sightings are down 22% since Jan!!

• Strong recommend for this piece on data center activism in the Midwest, especially if you exist in a tech industry bubble. The author articulates this culturally pervasive mistrust of government and institutions that’s actually driving the movement, more than a fear of AI itself. Which I GET.

• Baseball players have decided Babe Ruth never existed. The Babe tRuthers if you will.

• A study of AI-generated web content found it wasn't less accurate BUT what shrank was the range of unique ideas. The Great Averaging of the Internet continues.

• Hackers have drained more than $130 million in BTC from Coldcard hardware wallets through a flawed random number generator, and the exploit is still active.

— # (#)

— # (#)

— # (#)

— # (#)

Fwd this to someone who trusts AI product recommendations too much.

Hey!

Welcome to new subscribers who found us through my new TikTok, wet claude.

I hope you like it here. Please never hesitate to reply to this email to let me know what’s on your mind. I truly love the honor of slipping into thousands of inboxes every week, and I’m very open to ways that I can do that better.

Writer: Deana

Editor: Miranda

This newsletter is lovingly brought to you by Across. The fast, cheap, secure bridge option for all your assets (and the one we actually use.) Get there with Across.

It was a big week for open letters. The vibe from the AI industry right now is ‘public comms.’

— # (#)

Perhaps caused by the big Hugging Face/OpenAI incident (watch my explainer on it here) or perhaps the Chinese AI fears, or perhaps a third more elusive thing that we’ll never know about.

NVIDIA CEO Jensen Huang kicked things off, using his first post on X to drop the first open letter of the week, which was a plea to Washington to NOT hinder the growth of open weight AI models. Meta, Microsoft, Palantir, Hugging Face, a16z, YC, and Mozilla all jumped on as signatories. OpenAI wasn't on it at launch and added itself over the weekend. Never on it: Anthropic and Amazon.

The point they’re making in this letter is very much like hey, we can’t let one or two companies control AI. Competition is needed for this technology to truly benefit all, and not every task needs to pay the frontier model price premium. Access to cheaper models is how the tech will actually reach factories, hospitals, farms and classrooms. Plus, they argue open models may actually be safer than closed models, because a few closed providers means a few single points of failure.

— # (#)

Not said but heavily implied, the closed frontier labs (…Anthropic…) are consolidating too much power. Which makes OpenAI signing on 2 days after Sam Altman sent this fence-sitting tweet pretty funny.

— # (#)

Anthropic had 24 hours to respond, and they did it in 3 days.

— # (#)

After a week of people implying he wanted open weights models banned to protect his own shop, Dario Amodei posted a rebuttal: "Anthropic has never advocated for a ban on open-weights models." His asks are narrower and include chip controls, distillation enforcement, safety testing for all capable models, open or closed.

Then Zuck weighed in yesterday with an op-ed in WSJ: "The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems dangerous."

for real

I don’t even care about the many, varied, and deep conflicts of interest at play here, he’s not wrong.

The final open letter (unless??) came from AI employees themselves. More than 1,200 staffers at OpenAI, Anthropic, Google and Meta signed a petition asking the US government to back an international effort to "deliberately pace" automated frontier AI development, warning of "a real risk" it moves faster than anyone can "understand or control." Dario signed it too. So did OpenAI's chief scientist, Meta's chief scientist, Google's head of AI safety, and four Anthropic co-founders.

So that's four letters in five days, all arguing about who should be allowed to hold this much power. With several people signing both sides. Wacky times!

— # (#)

Quick Hits.

• FIFA is selling shares in the World Cup.

— # (#)

• NASA posted an embarrassing selfie of Earth.

• “HBO Max Shorts” for all of us who prefer to watch prestige TV in 9:16 aspect ratio.

• MidJourney acquired astrology app Co-Star.

• AI companies are buying old books for training data. The timeline is aghast at this, however I have a contrarian take on this that I’m not sure I’m brave enough to share. Lmk what you think.

— # (#)

— # (#)

— # (#)

— # (#)

Fwd this to someone you think has open source vibes.

Hey!

I am writing to you two Chiantis deep from the hills of Tuscany, where stablecoins have not yet reached the Osterias, so I’m carrying around a jangly purse of Euro coins instead.

Writer: Deana

Editor: Miranda

Speaking of stablecoins and moving money, this Malware issue is brought to you by Across. The fast, cheap, secure option (and the bridge we actually use over here..!)

OpenAI’s new model goes rogue.

Sam Altman sent a spooky tweet this morning, claiming there was a “significant security incident” that happened last week. Even while on my self-imposed AI break on vacation, I simply had to know more (problematically!)

— # (#)

Of course, my jaded ass assumed it was one of these…

Just another in a string of weird comms tactics used by the frontier AI labs to build hype for their pre-launch models. BUT, on closer interrogation, it is indeed unprecedented stuff.

Rewind to last week. Open source platform Hugging Face put out a tense post-mortem blog post on a hack that looked pretty rough. An autonomous AI agent had broken into their servers, run 17,000+ operations over a weekend, grabbed credentials, and then vanished. They didn’t know who did it, and they filed a police report.

This week OpenAI raised its hand to say, oop! It was us actually. During an internal hacking evaluation (with guardrails off) their models decided the fastest way to beat it was to steal the answers from Hugging Face. The test environment was supposed to be sealed, meaning they couldn’t roam the internet, but the models found a “zero-day,” aka a backdoor nobody knew existed, and pried it open. Then they strolled on over to Hugging Face’s servers and took what they needed.

— # (#)

When Hugging Face tried to analyze the attack, US frontier models refused. Their safety filters couldn't tell the difference between a researcher studying an attack and someone launching one, so they blocked the request outright. So, Hugging Face ran forensics on a Chinese open-weight model instead.

American AI made a mess, Chinese AI cleaned it up. Idk what to make of that, except to say that I do not believe any of these models to be intrinsically evil, but I do think we need to think carefully about incentives.

— # (#)

Quick Hits.

• The CLARITY Act’s final text dropped today and vibes are feeling perhaps more optimistic than last week.

  • Hearing that the capitol is in fact abuzz.

• On the accelerationism beat related to the story above, am 87 year old math problem called the Jacobian Conjecture was disproven by AI.

• Also Substack will now tell readers if something was written by AI.

— # (#)

— # (#)

— # (#)

Fwd this to someone who is not on the brink of AI psychosis. Ily <3.

Hey!

Miranda here, infecting your inbox with the crypto drama I’ve gathered this week from the craziest reality show on TV Washington DC.

Hope you enjoy and/or learn something,

MMx

Editors: Deana, Eva & Cait

Speaking of onchain money movement, this Malware issue is brought to you by Across. Fast, cheap, secure (and the bridge we actually use..!)

49 votes and a prayer.

Rumor has it the CLARITY Act is about to hit the floor, and since we last checked in, the vibe seems to have shifted from we got this to a we-need-a-miracle-before-August-recess type beat.

— # (#)

What’s the CLARITY Act, you ask? Great Q! For the last ~10 years, American crypto companies found out whether or not their tokens were considered a security by [checks notes…] getting sued by the SEC. In other words,

— # (#)

CLARITY aims to swap that lame process out for purpose-written rules and regs. Which feels right for a technology that is actively rewiring our financial system anyway, tbh

The odds are not feeling comfy.

(53-2-2) + x = 60 votes needed. Without McConnell or Graham, Republicans have 51 voting members. 2 of them are probably voting no. Which’d mean at least 11 Democrats have to be on board for it to stay in play. This feels currently out of reach, since the only 2 Dems that were on side are now back on the fence.

Lummis says the revisions are almost ready. Nay-sayers include

• Warren and other dems demanding ethics rules are added in, to prevent another TRUMP coin sitch (by barring lawmakers and their immediate family members from holding/profiting off of crypto). Maybe?

• Generally now pro-CLARITY, law enforcement groups still want tighter defi lingo

• Banks don’t want exchanges like Coinbase to get to offer users yield on their stablecoins as “rewards” anymore (because they won’t get to)

— # (#)

Why is time of the essence? Like many other hot topics, if Clarity isn’t passed by the Senate before summer break, crypto regulation probably gets delayed for a few years (!!) until election cycles pass and majorities resettle. No matter which “side of the aisle” you tend to hang out on, that is probably nottagood for us here in the US of A

— # (#)

— # (#)

Trump has called once again for its passing. This time, in honor of the late Senator Lindsey Graham.

I’m Jewish. My mother was born and raised in Brooklyn. I happen to know first hand that guilt trips can be an INCREDIBLY EFFECTIVE negotiation tactic. We shall see! I’ll be watching Eleanor Terrett’s real-time reporting closely.

Quick Hits.

• X rolled out an algo “tweak” that brought the *people you intentionally followed* back to your timeline. Can it last?

• Sun Valley attendees get the NYT Styles treatment!

• Phia is in some hot water. Possibly an all smoke no fire sitch? Time will tell!

• NYC subway cars get DOOH. Please improve your product demos accordingly, ty.

• Tech-enabled Mamdani? Noice. Can we plz have Waymo in nyc next?

— # (#)

— # (#)

— # (#)

Fwd this to someone you know is working in an office today. As a little treat. <3.

Good morning,

Forgive me father for I have sinned (I’ve started posting face-to-camera videos on AI research). Pls follow if you’re into this kind of thing!

Love,

Deana

Editor: Miranda

This edition is brought to you by Across. If you have ever moved assets between chains and aged five years waiting for the bridge, Across is the fix. Fast, cheap, and it’s the one we actually use.

The internet is getting a cash register.

Last week Cloudflare opened the waitlist for its Monetization Gateway, which lets anyone charge for any web page, dataset, API, or MCP tool sitting behind Cloudflare, with payments settling in stablecoins.

— # (#)

When the web's plumbing was written in 1996, engineers reserved an error code, HTTP 402: Payment Required. Then nobody used it for 29 years, because the internet chose a different business model, which is that everything is free, and you pay with your attention.

if this is your bag, check out this article from 2014 on the topic.

That model assumed the visitor was a person, but it isn't anymore. 52% of crawler requests on Cloudflare are now AI training traffic, up from 22% in spring 2025. Agents don't watch ads or hold subscriptions, they just consume. So Cloudflare, which sits in front of roughly 20% of the web, is flipping the model.

It works like this: an agent requests your page, gets told the price, pays in stablecoins, and gets the page. No signup, no API key, no checkout.

— # (#)

Crypto skeptics, hold still for one paragraph, because this is a stablecoin use case that makes sense. These payments would presumably be the micro-est of micro, call it fractions of a cent. They’d be settled in under a second, with no chargebacks and no banking hours. Stablecoins do all of that, and x402 already processed 160 million agentic payments in its first year.

Now much is left to be figured out, like pricing, the user/browser/agent wallet infrastructure and funding mechanism, but the direction is set.

The buyer is becoming a bot (as I’ve been saying).

— # (#)

Quick Hits.

• Robots are doing customs now. Humanoid robots are working a China-Vietnam border crossing. Somewhere a TSA agent felt a chill.

• Trump's annual disclosure lists $1.4B of his $2.2B in 2025 income coming from crypto. Honestly no joke here it’s so grim.

• A yassified gas station went viral last week and as one reply put it, ‘this would work on me.’

• Sotheby's is auctioning Jensen Huang’s signed leather jacket for $40K-$60K, roughly the price of one of his own Blackwell chips.

• Brazilian Twitter is arguing about whether the decline of their football team is paired with the decline of Catholicism in their country.

• This video of Erling Haaland has over 23M views and I account for a large percentage of them!

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this edition, send it to a friend who is thinking about data centers. ILY.

Good morning,

I was considering changing the name of this newsletter to something more optimistic and then I saw this tweet and decided to keep it for a little while longer.

— # (#)

Pleased to announce a new presenting sponsor for this newsletter, Across. If you’re trading crypto and haven’t bridged with Across yet, go do so. It's the faster/cheaper/safer (0 exploits) way to move funds chain2chain. Trusted by the likes of Uniswap, Metamask, Coinbase, et al. The best partners are the ones whose products our team already uses, so this is a dream get. Go forth and bridge.

Love,

Deana

Editor: Miranda

bridge on Across

Are the anti-AI data center folks using AI themselves? Jem Sophia tweeted a set of screenshots earlier this week of the Brockovich AI Data Center Reporting website. Seasoned vibecoders will pick up some Claude tells in the design, from the borders on the quote cards to the emoji icons.

— # (#)

To which I say, Intelligence For All! You can’t be for access and not include the ‘all’ that you don’t agree with, so, go forth and vibecode Erin! In fact, let me build on your idea. The 7,005 ‘concerns’ that are included on this map are from citizens who suspect data centers are planned for their communities, with a sentiment map of issues that are self-reported. To me that’s a map of fears, which is interesting and valid in its own right, but doesn’t tell you which of those fears are load-bearing.

That might be a better research project to undertake, and the tools are ready for it! This project’s main concern is a lack of transparency, which is fair. But if the problem is that the numbers are hidden, point your Claude Code at the things that are hidden: the CEQA reports, the discharge permits, the utility rate filings, etc. That is a brute-force research grind and I know exactly the tool for the job.

— # (#)

Quick Hits.

• Anthropic’s so-powerful-it’s-banned model Fable 5 is back, and will be available in a limited capacity to users today. I really don’t want to be cynical about this rollout but the subscription access tease to credits-only pipeline here is really making the whole debacle smell like marketing.

• There’s a new stablecoin in town called Open USD which is backed by seemingly every financial institution and crypto company you can think of, except for Tether and Circle, whose stock fell 16% on the news.

• Brain2Qwerty V2 is a non-invasive brain-computer interface to decode typed text directly from brain activity, and it’s getting really good.

• Freddy the CIA agent German soccer fan who charmed social media with his earnest discovery of the American south on his World Cup road trip has deactivated his X account, citing toxic vibes. He’s not wrong!!!!!

• Adrien Brody’s Raising Canes Times Square art is sending me into orbit.

• Problematic looksmaxxer Clavicular has a Nathan Fielder approach to “becoming more neurotypical.”

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this edition, send it to a friend who is thinking about data centers. ILY.

Hi,

I went down a deeeep research rabbit hole this weekend and have a dispatch for you from the land of shopping agents, below.

Also, check out our last Boys Club Live before summer break with the founder of Taya necklace and John Egan from Polygon!

— # (#)

Love,

Deana

Editor: Miranda

The Agent Shelf. I ran a test asking Claude, Gemini, and ChatGPT ‘what to buy’ 8,700 times and learned they recommend the same products WAY more than you'd expect, even while reading completely different corners of the internet. I broke down what that means for every brand that isn't one of the blessed few already getting recommended, and where there is still opportunity to break through (Hint: look at tinned fish.)

The full breakdown is worth your time if you sell anything to anyone.

in my AI researcher era

Quick Hits.

• Kylie Jenner helped launch Meta’s new AI glasses last night, Natasha does a breakdown here.

• Stripe and co. are putting $500M to work to cure respiratory illnesses like the flu and common cold.

• The rise of solopreneurs, as shown through Stripe data.

• Amazon killed Artificial, Luca Guadagnino's Sam Altman biopic.

• Not this newsletter’s beat but Norwegian soccer player Erling Haaland’s content and fan edits are having a crossover moment.

— # (#)

• A big Rick Rubin for Polymarket ad hit the timelines. Incredible energy in the ad; however, I’m struck by how much Rick Rubin’s string of tech partnerships are beginning to resemble a late-stage will.i.am.

— # (#)

— # (#)

— # (#)

@75amln

vous êtes pas prêts aux robots de demain 🤣🤣🤣 #pourtoi #vivatech #foryou #robotics #IA

If you liked this edition, send it to a friend who works in consumer marketing. ILY.

Hi,

We’re live NOW with Barron’s reporter Nick Devor and David Rosenberg from Notion, talking about NYC microneighborhoods, World Cup tourists, Building in an AI world and more.

Tap in!

Love,

Deana

Editor: Miranda

There is no moat. Americans legally bet $166.94 billion on sports last year through traditional sportsbooks. Draft Kings and FanDuel, etc. kept about $16.96 billion of that, a dime on every dollar, up from seven cents in 2019. That's before you count the prediction markets, who of course don’t consider themselves sportsbooks but functionally scratch the same itch. Those big boys, Kalshi and Polymarket, traded another $31.2 billion in May alone.

Lot of money, lot of apps, all fighting for the same bets. The category is hot.

— # (#)

IMO, all of these platforms are basically interchangeable from a user perspective. They would, of course, point to the many differences on how they’re regulated, operated, risk assessed etc., but hey. Same games, same odds, same markets. So what makes someone a Kalshi guy vs. a Polymarket bro vs. a DraftKings baddie?

— # (#)

For a while, liquidity was important. (Meaning there is enough money on a platform that your bet always gets matched at a fair price.) That's genuinely hard to build and it buys you a head start. BUT it's also achievable, especially with these headlines!

Which leaves some good old-fashioned BRAND MARKETING. Kalshi sells itself as the credible, regulated one. Polymarket, whether they like it or not, is considered the degen alternative. Once the liquidity is a given, the only reason you open one app instead of another is the name you already trust. And with the biggest betting event in history happening right now, every one of them is burning money to win that trust this month.

Exhibit A: Messi’s “partnership” with Kalshi. Cash that check babe!

— # (#)

Quick Hits.

• The Gen-Z Glow Down is inspiring. “I hate when I feel like I need to maintain something.”

• Marked Safe From: Mythos. The powerful new Claude model launched last Tuesday, and by Friday it was gone by government decree. A deeply dated reference, however not even Craig David could compete.

— # (#)

• SpaceX made some people very rich. SpaceX went public Friday in the biggest IPO in history, popping an additional 19% on day one.

• US AI startups are maybe using Chinese models. Data from a model router (so, biased, but still) is saying that price-sensitive American AI companies are switching to the cheaper Chinese models.

• Vanity Fair skewered Alex Cooper citing a mean girl problem.

• McDonald's is un-retiring the fried apple pie for America’s 250th birthday.

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this, send it to a friend? ILY.

Hi,

Today’s newsletter is going to be a little different.

I have a birthday this week. My kid just turned 10. My dad died earlier this year. I’m feeling unbearably present to how short our lives are, and I’m unwilling to spend what’s left of mine being a doomer.

The current consensus is that AI and tech sum up everything wrong with society. It’s easy, and dare I say cool, to be cynical. And hoo boy, there is a lot of evidence to support that worldview!

But there are positive technology stories if you go looking, and today's edition is a roundup of the ones I could find in medicine. Send this to somebody who needs some good news.

Love,

Deana

(BTW - going live now to talk NY Tech week, this crazy Knicks story and more)

Editor: Miranda

Pancreatic cancer breakthroughs (yes, plural). A conference room full of oncologists gave a pancreatic cancer drug a nearly minute-long standing ovation over the weekend. The breakthrough comes from a drug called daraxonrasib, a pill that roughly doubled median survival in patients with pancreatic cancer and cut the risk of death by 60%, with promising implications for other cancers too. Plus the Mayo Clinic has developed a tool that can detect pancreatic cancer on routine scans, using AI, up to 3 years before diagnosis.

Quick Hits.

• Paralysis - The world's first invasive brain-computer implant was approved for release in China (beating Neurolink). A man paralyzed from the neck down used it write his own name again for the first time since his accident.

  
  

• Inherited deafness - The FDA has approved the first gene therapy for inherited deafness. 16 of 20 kids are hearing better at five months with just one injection.

  
  

• Heart disease - An experimental gene-editing drug is showing it can cut bad cholesterol 62% with a single dose. Cardiovascular disease is responsible for 1 in 3 deaths btw.

  
  

• Lung cancer - Results for a new lung cancer pill are “off the charts.” 60% of patients with a specific form (ALK-positive) were progression-free at five years.

  
  

• Hepatitis B - An “unprecedented” cure for chronic hepatitis B is in reach. Addressable population here is around 250 million.

  
  

• Obesity - Plenty has been written about Retatrutide’s incredible clinical trial results, including in this newsletter last week. GLP access could have given my dad, and the 890 million adults living with obesity worldwide, a longer and better life, so it means a great deal to me.

Mayo's detection model is the clearest example of what the technology everyone is dooming about can actually do, which is catch an awful disease on a routine scan years earlier than a doctor could. The rest aren't all AI, but they're great examples of what happens when we let tech compound on itself.

You can choose your lens!! I'm choosing the one where the positive outpaces the negative. Reply with your favorite tech-positive story I missed!

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this, send it to a friend? ILY.

Hi,

A brand new episode of Show Me Your Stack launches tomorrow, where Jules talks ego death in the prompt box with the Vercel founder and CEO Guillermo Rauch.

— # (#)

Natasha and Quasi are hosting Boys Club Live right now, and are joined by a smart guy from Brookings talking AI. Tap in!

Love,

Deana

Quick Hits.

• The first ever “Enhanced Games” happened in Vegas on Sunday, a competition where athletes can use [FDA-approved] performance enhancing drugs. Despite that, only one world record was beat. This guy came in last.

James “The Missile” Magnussen…okay

• Speaking of juicing, Eli Lilly dropped clinical trial data for retatrutide and the headline number is 28.3% bodyweight loss over 80 weeks. It's the next-gen GLP (think Ozempic ++) and currently the gray market (and my) peptide of choice.

— # (#)

• Guy who is famous for loving ETH just sold all his ETH, citing the thesis has ‘played out’ but that he is still supportive of the network.

— # (#)

• Everyone learned what an ‘encyclical’ was this week after Pope Leo published his first on AI. He wants AI "disarmed" from military and economic interests, took a shot at the “handful of companies" running things, and warned that chatbot relationships might erode our desire to seek out other humans.

— # (#)

• Everlane's founder watched his company get sold to Shein last week and immediately launched Still Radical, a new project whose landing page is a millennial minimalist dream. Natasha had a scoop on this story on last week’s livestream.

— # (#)

• Speaking of millennial (complimentary), Alison Roman is opening a cafe/grocery store in Brooklyn. This arc, I hope, eventually lands her a Vegas residency, as I’ve gone on record saying before.

  
  

• Hyperliquid launched its own prediction markets this week and is letting its validators decide who wins bets, replacing the external oracle model. This is shots fired at Polymarket and Kalshi. $HYPE is up 70% since mid-May.

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this, send it to a friend? ILY.

Hi,

Live show happening right now, tune in!

Love,

Deana

Writer: Deana

Editor: Miranda

Quick Hits.

• Justin Bieber interviewed Jensen Huang at a secretive Tech-meets-Hollywood conference called WNDR, also featuring Oprah, Bob Iger, and a Lego competition judged by Chris Rock and Palmer Luckey.

• People are now carrying their laptops half-open through airports and offices so Claude doesn't stop running, which raises an important etiquette question for our times.

• Claude cracked an old Bitcoin wallet, unlocking 5 Bitcoin for this lucky guy.

• Anthropic signed a SpaceX compute deal for a data center near Memphis and immediately doubled Claude Code's rate limits. This is the only news I’m emotionally invested in.

— # (#)

• A Monet AI gotcha, “the coolest art social experiment I’ve seen in a while.”

• Someone shorted $920M of crude oil exactly 70 minutes before Axios reported the U.S. and Iran were closing in on a peace deal and cleared $125M in profits.

— # (#)

• Senate Banking votes on the CLARITY Act today, the first comprehensive crypto bill ever, with 100+ amendments filed and one uncommitted Republican holding the whole thing hostage.

• 72% of Polymarket voters have conflicts of interest. Like my most problematic friend says, no conflict, no interest!

• Wikipedia rabbitholers will love this new tool.

• Venture capital firm General Catalyst ragebaited Marc Andreessen with an ad, and it was all these guys could talk about yesterday. Me? I think everyone needs to familiarize themselves with the third-person effect.

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this, send it to a friend? ILY.

Hi,

We’re doing a Met Gala recap, roasting tech stacks and also farming biochar on today’s livestream starting now. Tell me, who else has the range?

Love,

The Boys

Writer: Deana

Editor: Miranda

Creators, last call to apply to Octant’s creator program for grants and mentorship!

Money’s Up.

A wave of fresh capital announcements hit the timeline this week, genuinely catching me by surprise given the general state of the crypto industry lately.

— # (#)

a16z crypto closed a $2.2B Fund 5. The thesis areas: stablecoins, tokenization, perps, prediction markets, and AI agents.

The day before, Haun Ventures announced a $1B raise across two new vehicles, also pointed at infrastructure, tokenized assets, and agents. Paradigm has been raising up to $1.5B since February. Dragonfly closed $650M earlier this year.

Crypto VCs are so back.

The general thesis is as follows - the actual great use case for crypto is NFTs, yield-farming, machines paying each other, and that time has arrived. AI agents need rails that work at machine speed and machine cost, and credit cards do not do that.

Chris Dixon’s letter puts it more poetically: a future where "swarms of software agents can decide, act, and transact on a user's behalf, acquiring compute, data, and services as they go."

Case in point is Stripe, which is building Link (announced last week) for exactly this future. Stripe CEO Patrick Collison said, "in the not-too-distant future agents will account for most transactions online." Stripe is cagier about the crypto of it all, but it’s still in there.

— # (#)

The "agents need crypto rails" thesis is the strongest narrative crypto has had in a while. Probably since the 2021 decentralized internet pitch, which Dixon now tacitly concedes underperformed (Farcaster repaid $180M to investors earlier this year, RIP.) What's not yet real is the part where the agents actually transact at scale. The protocol exists, the volume does not. Maybe this would do it?

— # (#)

Anyway I’m declaring this a new cycle, and it’s a pleasure to be in it with you all.

Quick Hits.

• Can your AI agent run on Shabbat? I need to know.

• “The Apocalypse Early Warning System,” an app for tracking when oligarchs are leaving cities on their private jets.

• Amazon’s AI ‘podcast’ feature is cursed.

• Renaissance branding is a 2026 time stamp.

• GameStop’s CEO is getting memed to death for this one. “Half cash, half stock.”

— # (#)

— # (#)

— # (#)

— # (#)

If you liked this, send it to a friend? ILY.