IM 875: Florida Dad - Amazon, Anthropic, and the AI Power Struggle

Summary

Intelligent Machines: Florida Dad – Amazon, Anthropic, and the AI Power Struggle

Podcast: Intelligent Machines (TWiT)
Episode 875 | Date: June 18, 2026
Hosts: Paris Martineau, Jeff Jarvis
Guests: Alex Stamos (Corridor.dev, FreeFable.org)
Main Theme: The U.S. government’s ban on Anthropic’s Fable (and Mythos) AI model; implications for the AI industry, political machinations, and the global AI arms race

---

Episode Overview

This episode confronts the shocking, underreported story dominating AI and tech circles: the U.S. government’s abrupt ban on Anthropic’s advanced AI model Fable (and its cyber-tuned cousin, Mythos). The hosts and guest Alex Stamos (noted security researcher and signatory of the FreeFable open letter) break down the ban’s technical, political, and strategic nuances, the Amazon-AI-Government dynamic, and what this means for the future of U.S. and global AI. The show moves briskly between technical detail, incisive political analysis, and candid personality—with several memorable moments (including a viral “Florida Dad” alligator video courtesy of Paris).

---

Key Discussion Points and Insights

1. The Fable Ban: What Happened and Why (00:56–15:45)

• Setting the Stage: Host Leo Laporte expresses frustration at the lack of mainstream coverage (“I think this is a massive story with all sorts of implications. I don’t see it on the nightly news. I guess it’s too techy.” [02:47])
• Technical Trigger: The Commerce Department banned access to Fable after Amazon’s security team demonstrated it could fix known code vulnerabilities after an allegedly trivial “jailbreak” (simply saying “fix this code”). This was used to justify export controls.
• Open Letter: Alex Stamos highlights the open letter at FreeFable.org, signed by 150+ AI experts, emphasizing that “no real risk” existed from Fable that doesn’t already exist in other models (05:38).
• The Real Rubicon: Most advanced models (Opus, GPT-5.x, and Chinese open-weight models) are already capable of superhuman bugfinding and exploit generation. “Mythos is the best we know about at finding bugs, but it’s not a giant leap—it’s an incremental one.” (16:03, Stamos)
• Policy Disconnect: There’s no formal rule or standard that Anthropic could meet—government standards are unclear and unenforceable. “It is impossible for Anthropic to make this totally perfect... The White House is asking for something you can’t actually do.” (10:21, Stamos)

2. Global AI Competition: US vs. China (12:44–15:51)

• China’s Leap: Zeta AI’s open weight model, MIT-licensed and highly competitive with U.S. models, was released the same week as the ban, underscoring the strategic weakness of unilateral U.S. regulation.
• Open-Weight Models: These are rapidly accelerating and available—even to foreign state actors—while U.S. companies are newly seen as politically unreliable. “A huge own-goal for the United States in the race to dominate the 21st century via domination of AI.” (15:51, Stamos)

3. The Political Web: Amazon, The White House, and Blame Games (25:31–31:54)

• Andy Jassy’s Role: Multiple outlets report Amazon CEO Andy Jassy raised or relayed concerns about Fable’s security to the White House, though the chain of events (and significance) is debated (“Amazon and Andy Jassy specifically are incredibly intellectual and capable political operatives. …For Amazon to get involved at this level, there had to be a decision.” [32:28, Paris Martineau])
• Political Expediency: The administration may have already decided and was seeking a scapegoat, or at minimum, a credible technical justification (“…the White House was desperate to justify this knee-jerk reaction” [59:01])
• Power and Retaliation: The White House, under pressure, acts swiftly and possibly in retaliation (referencing old political grudges and personnel connections), casting a pall over the impartiality of tech regulation.

4. Process Failures & Industry Impact (39:16–43:09)

• Lack of Standards: Unlike aviation, there are no clear rules or procedures to “recertify” Anthropic’s model. “If the FAA grounds a Boeing plane, we know why… With AI, there’s no rule. We don’t know what Anthropic has to do here.” (21:01, Stamos)
• Chilling Effect: Talent and research may leave the U.S.; international labs (Mistral, Canadians, Europeans) now look attractive, damaging American credibility (“…giving a huge, huge boost to our Chinese adversaries” [22:37])

5. Structural Problems in AI Governance (75:49–81:14)

• Safe AI is Impossible: Quoting Bruce Schneier, the panel agrees there’s no way to fully “jail” or “guardrail” powerful models without gutting their utility for defense/security—and that “bans only serve to delay the problem for a short while.” (75:49–78:14)
• Open Source as a Solution: Open, transparent AI harnesses and decentralized development are floated as the only realistic check on monopolistic or out-of-touch corporate and government actors.

6. Social Backlash and Infrastructure Politics (64:33–66:46; 90:15–93:30)

• AI Backlash: Hosts chart rising anti-AI sentiment—now much faster than the “techlash” against the Internet—fueled by job fears, data center impacts, and high-profile negative stories.
• Environmental/NIMBY Tension: Paris shares stories and TikToks of new data centers creating noise, water, and land value issues, especially in marginalized communities (“…just impossible to live there. And then impossible to sell your home. Really unfortunate.” [93:12, Paris])

---

Notable Quotes & Memorable Moments

On the “Jailbreak” That Wasn’t

• “Fable said, ‘No, no, I can’t fix it.’ And then with three words, they—[big fat air quotes]—‘jailbroke it’ by saying ‘fix the code.’ That is not a jailbreak.” (03:23, Laporte)
• “Any model that writes code needs to know what insecure code is, or it can’t write secure code for you. …The ‘jailbreak’ was not a jailbreak.” (07:11, Stamos)
• “Fix this code, plus several manual steps, should never have triggered an export control.” (48:44, reading Katie Moussouris)

On AI’s Global Arms Race

• “The same week that the U.S. government kneecaps a leading U.S. AI lab, the Chinese release a model almost as good... I don’t think you can come up with a better storyline for why this is incredibly stupid.” (13:20, Stamos)

On Policy Catastrophe

• “This is a complete disaster for the American AI industry. …A huge own goal for the United States.” (12:52–15:51, Stamos)

On Political Motivation and Absurdity

• “Technical decisions should never be made for political reasons. …And I feel like this is all a political decision.” (84:02, Laporte)
• “The President is watching polls. The American people do not like AI. …It might be a popular thing to do politically to shut down AI.” (62:38, Laporte)

On "Who Benefits?"

• “Who benefits from Anthropic being banned?” – “OpenAI. Elon Musk.” (60:21, Paris & panel)

On Open Source and the Future

• “The only way to do this, I think, is to take it out of their hands… an open source model I think will be good enough and they will get better and better.” (80:05, Jarvis)

---

Timestamps for Key Sections

• [03:23] – Breakdown of the “jailbreak” and technical trigger
• [05:38] – FreeFable letter, Anthropic’s security handling
• [12:44] – Global AI competitive risks, Zeta AI’s model
• [15:51] – Bruce Schneier’s Pandora’s Box thesis
• [25:31] – Andy Jassy, Amazon, and the governmental blame game
• [39:16] – Aviation analogy, lack of standards and chilling effect
• [64:33] – AI backlash in American society, data center NIMBY-ism
• [75:49] – Schneier’s essay: AI safety and public/open models
• [90:15] – Elon Musk acquires Cursor; data center lawsuits
• [110:30] – [Humor] Paris’ “Florida Dad” alligator adventure

---

Memorable Anecdotes & Humor

Florida Dad Alligator Video ([110:30])

Paris shares a viral video of her own father and neighbors dragging an alligator across a Florida yard — a moment of absurd, regionally specific pride (“My dad is the front one.”). The panel laughs about the unique blend of danger, community spirit, and American improvisation.

Subway Seat Debate ([119:16])

A New York City tangent spawns a debate about the best seat on the orange subway, complete with speculation on New Yorkers’ secret seat strategies and the possibility of orange train seat tattoos.

---

Concluding Insights

• Industry aftershocks: The Fable ban—seemingly technical—was fueled by politics, poor standards, and a global race that the U.S. isn’t winning by acting alone.
• Politics vs. Progress: Short-term actions risk lasting damage to credibility, research migration, and the U.S.’s leadership in AI.
• Open source is the way forward: Only decentralized, transparent, and global approaches can prevent overreach and stagnation.
• The future is uncertain, the stakes enormous: As the episode’s guests and hosts stress, this isn’t just a tech story—it’s a pivotal moment in global innovation and geopolitics.

---

Further Reading

• FreeFable.org – The open letter and coverage of the Fable ban
• Bruce Schneier’s Guardian essay – “AI’s Pandora’s Box Has Already Opened”
• Katie Moussouris’ analysis – On the “fix this code” t-shirt and the technical reality of the Amazon “jailbreak”

---

For a mix of deep insight, honest opinion, and welcome levity, this episode is a must-listen for anyone following AI’s wild new chapter—politics, paranoia, and all.

Transcript

[A]

It's time for Intelligent Machines. Jeff Jarvis is here. Paris Martineau. We, of course, are going to talk about the United States government banning the most powerful AI model we've got. Alex Stamos is here. He is the creator of freefable.org and he has a lot to say about it. Stay tuned. I am is next. Podcasts you love from people you trust. This is Twit. This is Intelligent Machines with Paris Martineau and Jeff Jarvis. Episode 875, recorded Wednesday, June 17, 2026, Floridad. It's time for Intelligent Machines, the show where we cover the latest in AI robotics and all the smart doodads all around the house. Paris Martineau is here from Consumer Reports. Hi, Paris. You look rested.

[B]

I am. I'm glad that there's no big AI news happening this week.

[A]

You did stay up a little late this week with the Knickerbockers, I see. But I did. Did you climb any telephone poles or stomp on any poles?

[B]

No, but I watched some lovely people climb on a bus and take a couple photos and then climb down. It was nice.

[A]

I saw your picture. Very funny. Very funny. Great to see you celebrating this week, and we're happy for you. Also, Jeff Jarvis is here.

[C]

Nixon 5.

[A]

Yeah, you were right. Well, wait a minute. It already happened. You just. Never mind. He is the emeritus professor of journalistic innovation at the Craig Newmark Graduate School of Journalism at the City University of New York. New book, hot Type, coming out next. Well, in August anyway. And you can order it [e-mail] along with the Gutenberg parenthesis. So I'm a little miffed because in my opinion, this story about Fable. We talked. I think we talked about it last week. We talked about Fable coming out last week. Coming out, Fable being blocked on Friday. I'm in the middle of a bit. I mean, I said, oh, great. I could finally rewrite the Twitter ad sales software. It had analyzed it and written questionnaires for the staff. It was all set to go, and then it said, model no longer available. I thought, did I not pay my anthropic bill? And then I went to x.com and I was shocked. As you probably know, the federal government has blocked Mythos. The Commerce Department used its ability to say, well, no foreign nationals can use it. But of course, if foreign nationals can't use it, that means many of the people at Anthropic can't use it. And because anthropic cannot determine if you're a foreign national or not, they haven't Checked your citizenship papers. You can't use it either. Now, I'm a little disappointed because, honestly, I think this is a massive story with all sorts of implications. I don't see it on the nightly news. It's, I guess, too much of a tech story. Fortunately, the great Alex Stamos is here. He is a longtime security researcher, always a welcome guest on our shows, currently at corridor.dev, which is about securing AI coding. So he knows a little bit about this topic. And actually, actually you put out an Open [e-mail], to which you got many great signatories, Ed Felton and Paul Wickvickse and many others saying, I think making a very cogent case. Now, there is increasingly evidence that this was a political decision, not a security decision. Katie Mazuris says she saw the Amazon paper that precipitated all this, in which Amazon researchers took Fable and gave it a bunch of code that had flaws in it, known flaws, and in some zero days and asked it to fix it. Fable said, no, no, I can't fix it. And then with three words, they, and I'm going to put this in big fat air quotes, jailbroke it by saying, fix the code. That is not a jailbreak. Anthropic, of course, said, that's not a jailbreak.

[D]

That's what every model does.

[A]

But it apparently was enough to convince the Treasury Secretary, the Commerce Secretary, Susie Wiles, the Chief of Staff at the White House, and I have no doubt the President as well, that, well, we've got to stop this. They were already, I'm sure, very nervous about Mythos. Pete Hegseth took a victory lap. Remember, he was the guy who said that anthropic is a supply chain risk. This just confirmed what he had said three months earlier. Mike Masnik, today on Tech Dirt, wrote a very interesting piece in which he pointed out that Katie Mazuris was seen who's, you know, the very respect. You worked with her at Zoom, Very respected security researcher, creator of the Microsoft bug bounty. She kind of invented the bug bounty that she was a Democrat. What did he call, what did he call him? A Democratic operative for the Democrat Party. And so. And related to your former partner, Chris Krebs. Worked with Chris as well. And of course, the president hates Chris Krebs because he had the temerity to say the 2020 election was free and fair.

[C]

Yeah.

[A]

So Masnik's contention is this is a 6 year old fit of pique, but I think this has huge implications. So let Me ask you, Alex, I've done enough talking. I wanted to set this up. What do you think?

[C]

What do I think? So, like you said, we have this [e-mail], so it's not just what I think. We have over 150 signatories, as you pointed out, a lot of people who really know what they're talking about. Not just academics, but CEOs, CTOs, chief scientists at AI companies, large tech companies, founders of AI companies saying they do not believe there is a real risk from Fable. Why do they say that? So a number of us have seen the reports that Anthropic have put out refuting Amazon's research. Now, I don't think Amazon has done anything wrong here. Amazon has a shared responsibility model with Anthropic, where Amazon hosts anthropic models in a number of situations. You can run anthropic models in Bedrock. When the US Government uses anthropic models, they are running on Amazon hardware, especially like in the classified sense. So, like, if the NSA is using opus, they're running in Amazon top secret cloud. So Amazon has a responsibility to do the security testing. I expect this testing was due to Amazon doing qualification of Fable and Mythos to run within their own hardware. And Amazon had some quibbles about Fable's front end classifier. Now, these quibbles should not be surprising. In Fable's model card, Anthropic actually lays out what their goals are of stopping Fable from doing security work. As we all know, Mythos has some pretty good security capabilities. But as we say in the letter, Mythos is not the only model that can find bugs. And I think this is something that's getting lost here partially because of the way Anthropic themselves talks about Mythos. Right. They talk about Mythos as kind of like avenging cyber God. The naming here, they brought this on

[A]

themselves to some degree, I have to say.

[C]

Yeah. I've suggested to the Anthropic people that maybe for the six series, you come up with, you know, names that you would name your kids Bunny. Right?

[D]

Instead of like, cool, the Doom boys. Cool. The Doom.

[C]

Yeah. Let's not go for like Cyber Gorgon or, you know, Cyber Zeus or something like that. Let's like Rainbow. Yeah, yeah, like Fluffy Bunny series. So like the Cottontail series or something. But yes. So the Mytho series is really good at finding bugs, but, like, it's some percentage above the Opus series. Right. But it's. It's. It's good. And so they did not want to provide all that capability to the general public. Right. Mythos is restricted in who it's provided to. And so they have trained Fable to refuse to do certain things. And they also put a classifier in front of it so that a much dumber algorithm that is supposed to watch to see if you ask and do certain things. And they say this in the model card where they say they go into great detail. They specify different kinds of cyber actions. Any model that writes code needs to be able to understand what is secure versus insecure code, otherwise it could not write secure code for you. Right. You do not want Fable to be completely unaware of what a security vulnerability is, otherwise the code it put out, which is going to be the number one use of Fable. Right? That is what all of these anthropic models are used for on a per token basis. The number one thing they're used for now is coding. And so Fable needs to know what insecure code looks like. It needs to be able to find security flaws. That that is absolutely an appropriate thing for Fable to do. What they don't want Fable to do and what they say they don't want to do is they don't want. What you can do with Mythos is you can upload the entire Linux kernel to Mythos and you can have it run for like 12 hours and find you dozens and dozens of bugs and then write you a full exploit chain. Fable will not do that. Amazon never said that they can get Fable to do that. What they can get Fable to do is find one specific flaw when they tell it to look in the exact place where that flaw is. That's fine because you can get GPT 5.5 to do that. You can get Opus 4.8 to do that. It turns out you can get Chinese models to do that. So why do you ban Fable when there are a bunch of other models that are on the market and Chinese open weight models that can do the exact same thing? That is the base unfavorable.

[D]

Why even notify the administration of this then?

[C]

So that is an interesting question. Like there is a reasonable argument that if you say that your classifier is not going to allow a certain thing, we're able to beat it. There's kind of an academic argument between Amazon and Anthropic of hardening this classifier. That is a reasonable argument to be made between two different business partners. Why did the White House get involved? I cannot say. I don't know exactly what happened there, but they never should have been involved. The equivalent would have been let's say Anthropic found like a single cross site scripting bug in Amazon.com, and then all of Amazon.com got shut down by the U.S. government.

[D]

Right.

[C]

That would have been the equivalent here. That would have also been completely inappropriate. Security testing happens, so we make products better. And there is no written standard here for Anthropic to live up to. That's the problem. And I think that's also the problem of why we're not seeing this model turn back on. Because it is impossible for Anthropic to make this totally perfect. In fact, NIST has a paper about how you can't completely prevent jailbreaking and you cannot make Fable not know anything about security and have it still build secure code. You also, there's no possible way to get Fable to have to live up to a standard that you don't already have. All these other models well past. And so it's. The White House is asking perhaps for something that is completely hypocritical. And so they've put themselves into this crazy corner. And I think that's one of the problems of why we're still. It's Wednesday afternoon and Fable is still turned off because the White House overreacted so far that they have not been able to find a way to back off of the position that they've taken.

[B]

Well, I mean, that seems to be kind of the through line with the White House's interplay with Anthropic generally over the past weeks and months. Right.

[A]

But Anthropic did go to court when Hegseth called them a supply chain risk. They have not gone to court over this yet.

[B]

I mean, and how well did that go for them? It was a nightmare.

[A]

Well, but I also feel like maybe this Commerce Department's within its rights. They can say that's the problem.

[C]

It's the case.

[A]

Well, maybe it may be illegal.

[C]

There's an argument that they don't have a legal right here. So first off, there's an argument that the order does not actually require what Anthropic did here, that the order says that they can't export the model weights.

[D]

Right.

[C]

If you look at the actual language of the model, you can't export the model itself, not the outputs of the model, which Anthropic would never export the model itself. That would be insane. It's their most valuable piece of intellectual property. Now, Anthropic is obviously they're trying to play nice. That is why it is taken down. But I expect they only have a certain amount of patience here before they will go to court. And my understanding is that they would have a very strong case that IEEPA does not give this power to the administration. Now, obviously the executive branch of the United States government has all kinds of ways they can try to punish the company. And the company wants to IPO this year. So do you want to get in a fight with the organization that owns the doj, that owns the sec, that owns the FTC in the same year that you're trying to become a public company? Probably not, right?

[D]

What precedent does this set in terms of US Government thinking it can regulate AI? And what does that do to the AI industry in the world?

[C]

It's a complete disaster for the American AI industry. This is a complete and total disaster. The other thing that happened yesterday is A model called GLM 5.2 was released by Z AI, which is one of the leading Chinese AI labs. This is an open weight model. It is MIT licensed, meaning anybody can download this model and use it for whatever purpose they want. They can retrain it, they can redistribute it. You can actually use this model in a commercial product if you like. It is within a couple Percentage points of Opus 4.8 in its performance and a number of coding benchmarks. It is amazing. So the same week that the United States government kneecaps one of probably the leading US AI lab, the Chinese release a model that is almost as good as the best model you can currently, currently get from that lab. I don't think you can come up with a better storyline for why this is incredibly stupid. The idea, the whole action is based upon the premise that American labs are light years ahead of our Chinese competitors and will always be. So the entire idea of all this regulation is just to play defense, that the US will always be ahead and we have to play all this defense because the Chinese will always be behind. That is incredibly arrogant and not based upon any kind of fact. Zeta AI released this open weight model. That is what we know about. That is within a couple percentage points of 4.8. I guarantee that the People's Liberation army, the Ministry of State Security, they have access to cybertuned models that are better than what Z AI is able to release on a Tuesday.

[D]

Right.

[C]

And so are those private models mythos grade? I don't know. Right. But the idea that we should be undermining the world's faith in the USAI industry at the same time that Chinese are making these leaps and bounds and releasing these models that then the rest of the world can Use safely is really, really stupid because it's a good thing. One that Fable's only out for a week because if it had been out for a month and then the administration did this on a Friday afternoon with no warning, you would have had SEV1 pagers going off across the entire country. Right? Datadog people's phones would have been doing the datadog woof woof woof alert all over the country. Because those Fable models would have been built into critical mechanisms and systems would have been breaking all over the place. Like if they did this for opus 4.8 right now, systems would break all over the place because the number of organizations that have actually set up for an entire model to disappear and that they have a LLM router that will fall over to another provider and have been tested on another model and that will work on another model are exceedingly small. Now everybody is doing that because political risk has become a real business continuity risk for anybody who uses a US model. It is a great, great week for the Chinese AI industry, a fantastic week, and a huge own goal for the United States in the race to dominate the 21st century via domination of AI.

[D]

What do you think of Schneider's argument? What do you think of Schneier's argument that it's all silly because the Pandora's box is opened. If this model isn't the most powerful, the next one will be? Get used to it, folks.

[C]

Yeah, I mean, that is effectively where we are. So first off, everybody talks about Mythos. Like I said, Mythos is the best we know about at finding bugs. But it's like it's here over Opus, right? It's not like here the real Rubicon was crossed last year, probably like the Opus 4.5 GPT 5 something series. That is when these models became better than human beings at finding bugs. And not just better than human beings, but infinitely scalable. Right? So there are a handful of really, really good bug finders who can find these bugs and then write. More importantly than finding the bugs, write stable exploits for them. Not only are these people rare, but they're relatively slow. Right? And I'm not criticizing them. They do things that I can't do or the. To the extent that I could do these things, I could do them 20 years ago, and I have not been able to do them since I was much, much younger. And the thing about these models is they're incredibly fast, they're scalable, and they're cheap. And so you're now able to generate a bunch of. Dave, itels is one of these guys who is still able to do it, even though he's my age. You're able just to generate a bunch of Dave Itels by spending more money and multiplying these models out. That is a really scary future. But that happened last year. That was not because of Mythos. And so if you are using one of these models, including the open weight models, and you go look at some open source library now, you know Linux kernel is getting ground down a lot and has already had a lot of bugs found. But if you go just find some open source library that's in Debian that hasn't been looked at too heavily, I guarantee if you know what you're doing and you're using Opus 4.8, you will find an oday within a week and you can write an exploit for it like that. That is happening all the time, every single day. The ratio of the number of breaches that are because of oday versus stolen credentials or spear phishing or anything has completely changed because now finding a zero day and writing stable exploit code is so much easier than it used to be. We're already there. And for these attackers, a lot of them, or most of them are probably using the open weight models. It's hard totally to tell, but they're probably using open weight models both because of cost, but also because they don't want to leave logs in American providers for the FBI and Western intelligence agencies to find. If you're these guys, you're going to go have a machine at home with a bunch of 50 90s in them and you're going to use it to play Call of Duty. And when you're done playing Call of Duty, you'll go grind out a bug in a bunch of libraries and then go use those exploits to go make some money. We're there. And so this idea that we're going to be able to like mess with American labs and ban these models and then that's going to make the world more secure is just incredibly stupid.

[A]

I know you have to run. Alex is going to be talking tomorrow with Alex Kantrowitz at the big Technology AI Summit. And you have invited David Sacks to meet you on stage. I doubt he will.

[C]

I mean, David, if you're listening, I'd love to see you there. Commonwealth Club 03:25pm I think an open

[A]

conversation about this is necessary. I think the administration has to step up and explain.

[C]

David has said many times that he does not believe that AI models should be regulated.

[A]

That's how Trump got elected and One of the first things he did was oorTurn the Biden AI EO.

[C]

Yeah.

[A]

A I E I O. And I think that this is a complete reversal of the Trump administration. I hope it's not purely political, but it sure feels like it is. And I think that's why this is a huge story, because this is a very consequential technology. Maybe there's risk. I understand the desire to protect us from risk, but I don't think that's what's going on here. And it's as if this action itself

[C]

is not protecting anybody. It is, it is casting a huge amount of doubt on the American AI industry. It is pushing the best AI researchers to rethink coming to America. Right. Like, because you might not be able to work on the best models. So you.

[A]

Or leave.

[C]

Or leave. Right. I mean, they're not going to go to China. Like, nobody wants to go to China except the Chinese. But like, if you. Now would be a great time to open up either to recruit, if you're Mistral, or to go open up a frontier model company.

[D]

France looks good right now.

[C]

Yeah, France looks pretty good. If I was the Canadians, I'd be throwing a bunch of money into this. Right. Or if I was anthropic, I'm not speaking for them. If I was anthropic, I'd be looking at office space in Vancouver right about now. Right. For sure. And you know, it, it really imperils, honestly, like it really imperils the, the credibility of the American AI industry that this kind of stuff can happen on a Friday afternoon without any rulemaking, without any democratic process.

[B]

I mean, the broader economy at large too. This is just a. Well faster for.

[C]

I mean, if the FAA grounds a Boeing plane, we all know why, right?

[B]

It's because there's a rule that door has blown off. And everyone got video of it.

[C]

Yeah. And the FAA had a rule that your doors aren't supposed to blow off. And then Boeing knows how to fix it, right. They're like, oh, we made the door not blow off anymore. And the FAA is like, we tested it and the doors don't blow off anymore. And then the FAA is like, okay, great, you can fly again. We, there's no rule. We don't know what anthropic has to do here to get back in other than what people are tweeting about is like, they're not nice enough and they don't talk the same language and like they hire women with pink hair. Here I am, I'm a middle aged CIS white Guy who golfs and duck hunts and has a boat. And I'm saying that there's also not. I'm wearing a pink polo shirt, right? Like, I don't know what else to do from a cultural.

[D]

You need the vest. Where's the vest?

[A]

You need a free Fable T shirt. That's what you need.

[C]

No, no, no. I'm just saying you need three.

[B]

I don't flags.

[C]

My Twitter icon is me on a sailboat with an American flag behind me. I do not have pink hair in this thing. I, you know, I filled out an S of 86 and like, this thing does not hurt the. Yes, there I am and there's my name in Cyrillic because both Chris Krebs and I got sanctioned by the Russian government, which is a proud day.

[A]

Proud day. Yeah.

[C]

But like, you know, this thing does not. Having Fable out there does not hurt the American people. And. But what does is casting doubt on the reliability of the American AI industry and therefore giving a huge, huge boost to our Chinese adversaries.

[A]

Free fable.org if you want to read the article.

[D]

Thank you for that. Thank you for your work.

[A]

Really appreciate it. Alex, take care. We appreciate it. Have a good talk tomorrow. Yeah, thank you. I'm jealous Alex gets you. We'll continue in just a bit. You're watching Intelligent Machines our show today brought to you by OutSystems, the leading agentic systems platform. Outsystems helps businesses bridge the enterprise gap to their agentic future where the constraints of the past give way to unlimited capacity and scale. With Outsystems you can architect, deliver and scale governed agenic systems with agility and trust using one open and unified platform PowerSecure company wide agencic orchestration for core business operations. Outsystem provides the only agentic systems that are unified, agile and enterprise proven. Let me explain. Outsystems is unified because it enables you to build, run and govern apps and agents on one single platform. It's agile because Outsystems allows you to innovate the speed of AI without compromising quality or control. And Outsystems is enterprise proven and trusted by enterprises for mission critical AI applications and durable innovation. Outsystems is frankly the secret weapon behind the world's most successful companies. And I'm not just talking small apps. Outsystems creates the massive complex systems that run banks, insurance companies, government services. Outsystems can even help companies with aging IT environments bridge the gap to the AI future without a rip and replace nightmare. I'll give you an example. Outsystems provides the safest and fastest way from an enterprise to go from we need an AI strategy to we have a functioning agenic system, something you'd be really proud of. Stop wondering how AI will change your business. Start building the agents that will lead it. Visit out systems to see how the world's most innovative enterprises use out systems to engineer, orchestrate and govern agency systems quickly and cost effectively without compromising reliability and security. That's O u t S Y S t e m s.com twit to book a demo, you need to check it out. Outsystems.com TWIT we thank them so much for their support. I feel like this should be a story that you see on the nightly news every night for the next few weeks. Everybody's talking about the green reflecting pool instead of it's as if the administration has said, you know, this Internet thing is really cause for a lot of fraud and theft and we're just going to shut it down. This is what autocratic governments do.

[B]

You guys, how are you parsing the Andy Jassy of it all? Because I think that's a tough. Obviously it's not the.

[A]

I think it's somewhat misreported to be.

[B]

I think. I don't think so because there's like seven or eight different outlets reporting it, but the way they're reporting it is very interesting to, I guess, give an overview. Let me see here my notes. Originally we had the information and Reuters reporting the information reported that Jassy was basically they both reported that Jassy was among tech leaders who raised security concerns to senior Trump officials, and those conversations set the export controls in motion. However, the Wall Street Journal went further and characterized the concerns by Andy Jassy as instigating the Friday letter. Axios then said Jassy raised the fabled jailbreak concern to the administration on Thursday. Then Friday morning there was a big meeting with the White House. Then the White House reached out to Dario, asking him to do something, but he was apparently at a wellness retreat and didn't late afternoon, which I think is a fascinating detail that I just hold in my head. Then he called Scott Bessett directly and then it all kind of blew up. But since then, we've had kind of a back and forth from Amazon and Anthropic and the government that, you know, are all kind of trying to push blame off to one another. But even if this is somewhat partially true, like the White House in some ways has said, as has Amazon, that, oh, the, the White House asked Amazon to look into security flaws in anthropic. But that can't be.

[A]

I think we've learned more since all of those stories came out and I think actually Alex kind of explained what happened. But I'll also point you to Katie Massouris's blog because she now she was one of the outside people who saw the Amazon report. She's very careful not to say. She just says a third party. But we now know it was Amazon. And I think what Alex explained is that Amazon in its normal test, remember Amazon is a big investment.

[B]

Yeah, I mean that's

[A]

what I want

[B]

your guys thoughts on Amazon is on the board. Amazon is incredibly involved. It seems even if one aspect of this is true, like it seems incredibly important, they would.

[A]

Yeah, yeah, no, no, it's exactly as Alex said. This is in their normal course of work. They found this. Now you may say, well, why do they tell the White House about this? And that's a larger question. I think a lot of the reporting you're seeing and certainly a lot of what you're seeing out of David Sacks and others is pure ignorance. You know, this is what Katie Massouris wrote in her post the fable 5 export controls harm cyber defense. And she has as she has put the fix this code which was the quote jailbreak on a T shirt and said this T shirt is munition. I think there is a legitimate question of why Amazon raised this specter, but I'm not convinced that what we're reading is accurate. It may well be the administration had made this decision and was looking for somebody to blame it on or some technical, some credible technical source to blame it on and called and Politico says they called Amazon and asked them. Jassy wasn't on the first call. Who had the more recent timeline? I think it was the New York Times. Jassy was not on the first call. It may be that Amazon has channels open already with CISA or other governmental agencies in which they would normally report this not as a cause for huge concern, but just as something going on that this is what I suspect happened. That happened apparently two days after it was released. According to Acadia Mazouris, like on a Thursday that went up the chain. And I think at that point the White House, probably Susie Wiles, the chief of staff got Andy Jassy on the phone to say what do you know? Because I think they were looking for an excuse. There was already fear about Mythos and I think they were looking at an excuse to do this.

[B]

Okay, but assuming that's Correct. Which could be that didn't go wrong.

[A]

Well, it's not in Amazon's interest that this.

[B]

Let me finish. I mean, that aligns with what's been reported by all those outlets that I just put out. But Amazon and Jassy in particular, like Amazon C Suite, are not naive to how politics work. And they would not just stumble into being manipulated by the White House, which, according to what we've seen from a lot of political reporting, even specifically around this issue, around cybersecurity and AI, they're not exactly acting with incredible sophistication. So for Amazon to get involved in this at the level of Jassy having his name, identity and all these details being leaked out to various outlets has to mean that Amazon made some decision to get involved in an actual way. Because if they didn't want to, they wouldn't have. Or they would have in a way where they would have.

[A]

Well, we don't know what could have happened is they called Jassy. Jesse said, yeah, we have this. I mean, when we asked it to.

[B]

Well, why would they say, yes, we

[A]

have this thing, we have this. Well, they know that there's this paper. Jassy could have said, yeah, when we asked it to fix the code, it did fix the code. And I think he could easily been put in the situation of, well, is that problematic? He could have said, well, it's not supposed to. That's all he would have to say, by the way. It's not supposed to. And the people who wanted to do this. And I think that's really the question is I don't think anybody saying Amazon raised the alarm and talked the White House into it. I think it's much more likely Amazon knew about it, was asked by the White House. Jassy said, yeah, we found this thing. The White House, I think, is who blew it up. I don't. I. Yeah.

[D]

But to Paris's point, I think that Jassy and Amazon has excellent prime, and I think that Amazon could have pulled back and said, oh, just what you just said, Leo. Oh, well, we didn't really raise any alarms. We just said this or that, but they haven't done that. And, yeah, you don't want to tickle the tiger's belly. But still, I would have thought that if Amazon were really that innocent and naive, I would have gotten that message across.

[A]

Okay, so maybe Amazon's bad and wrong, but the bigger point, I don't think

[B]

that's what either of us are saying. We're saying is, why did Amazon.

[D]

Why we're not trying to understand.

[B]

We're saying Amazon and Andy Jassy specifically are incred, incredibly intellectual and capable political operatives. Assuming that they are acting with agency, which we have to. Because it would be astounding and like.

[A]

I understand.

[B]

Completely out of the norm. I know, but. Let me finish.

[A]

But I think it's irrelevant.

[B]

Let me. I know, but I'm. I think it's relevant. I would like.

[A]

There's a. So much of a bigger story here which is that the federal government has shut down Anthropic. Yes, that's a huge story. I don't care whether Amazon raised the flag or not.

[B]

I'm curious. It makes no sense. You're saying an unprecedented political and technical event has happened. Yes, that has happened with the participation of a incredibly sophisticated political technical operative that has a large financial and like board and general stake in the company who is getting wrecked by the US government right now.

[D]

It wasn't done by the White House alone.

[B]

It wasn't done by the right of loan. It was done by someone else who is getting. Has an incredible financial detrimental impact to their business by this happening to Anthropic. Why would they do that? Is that not fascinating?

[A]

Well, as a reporter I guess it's fascinating, but I don't think it's important politically. I think much more important is that the federal government is doing this. Probably. I think it's interesting that Alex thinks it's probably illegal, probably illegally and with great detriment to the American people, American companies.

[D]

Yes.

[A]

And to it. It is, I think, and this is my. I think the most important thing that our listeners should take with them, regardless of their political alliances or feelings, is that technology, a very important technology, is being assaulted irrationally by the federal government. And I, and I again say this is as if the federal government said we've decided to turn off the Internet because.

[D]

Yeah, and Alex's point was exactly right too. It's not as if it's a Boeing jet and we know what the standard is and they didn't meet the standard and ergo this happens. It was a serendipitous almost regulation of a new technology. I'm just saying without any means to

[A]

do so, there's a huge fire alarm going off.

[D]

Yes.

[A]

And if you want to worry about how it happened or who told who what. Okay. But I think the important point here is there is a huge fire alarm going off.

[B]

But don't you want to know how the fire started? Why do we not care about how the fire started?

[A]

That caused the alarm I would say. I would suggest that the best use of our energies at this point is to. Is to fix this problem, because this is a huge, huge problem.

[D]

So what is. What should Anthropic do now? What's the limits of what they can do?

[A]

They should move out of the United States.

[D]

Well, that's. Yeah. I don't disagree, but take the whole thing.

[A]

Look, Andre Karpathi, who they just hired.

[D]

Yeah.

[A]

Cannot work.

[D]

Yeah.

[A]

On it. They should take him and everybody else to Toronto.

[D]

Okay.

[E]

But.

[D]

But what should they do this week?

[A]

Well, they should probably go to court. I think they're trying to appease the Trump administration, and maybe that's sensible. They should probably do it at the Iranian government did and hire some psychologists. This is such an irrational and debilitating move for the tech sector in this country. And I think, you know, our audience is. Is here because they're interested in what technology can do and can technology can bring us. And I think I've never seen an assault on technology like this.

[D]

I couldn't agree more.

[A]

Quite a few.

[D]

Well, I'm trying to ask you as a technologist, is there anything that anthropic, besides the Doomer stuff, which I hope we'll get to. That's. That's Vibes in terms of technologically, is there anything that they did wrong? Is there anything that they should.

[A]

Yeah.

[D]

Well, is there. Is there something broken here?

[A]

They should have hired a psychologist because they made as. As Alex said they shouldn't have named it Mythos and claimed how dangerous it was. That was a huge mistake. And I think part of the problem is Anthropic has this very weird kind of duality where they're making clearly the best models out there. They're very good at what they're doing, and they're terrified by what these models can do, and they oversell how terrifying these models are. I think one of the things Alex and others are saying is Bruce Schneier as well is, well, yeah, they're good, but they're no more terrifying than all the other AI models that we have, models that can do things very similar to what Mythos and Fable are doing. And some of them are not coming from the United States. It is a foolish thing to do to shut it down at this point. Now, the only thought I had was, well, is there really a legitimate danger? Is there a reason to shut this down? Will we be safer? Because they've shut it down. And I think Alex and many others now have convinced me that that is not the case.

[D]

Bruce Schneier Ergo, there's really nothing to do. There's really nothing they can do.

[A]

Well, and that's what Bruce Schneier is saying is Pandora's box has been opened. AI is out. And for better or for worse, as for good or evil, it's out there. You can't stop it. And all this is going to do is kneecap the American AI companies.

[D]

How much? There was a great column today in the New York Times by. I'm going to forget his name suddenly. Cal Newport. Dear AI Companies, the doom trolling needs to stop. How much is. Is. Is anthropic to blame for what they put out just a week ago, saying, oh, this is so powerful, we should all stop.

[A]

Yeah, they're totally to blame. Yeah. I mean, I have articles before this happened. I don't know if Dario thought this would happen, but he said the government should block dangerous AI on June 10th.

[D]

Yep.

[A]

I mean, okay, they did. I don't know if that's what you wanted, Dario, so. I understand. I'm sorry, Perez. I understand you're fascinated by why Amazon did this, and we can pursue that, but I just think.

[B]

No, we obviously can. I'm so sorry to break that up.

[A]

No, pursue it all you want, but I just. I think that that's missing a large,

[B]

which is just to reflect how bad this is.

[A]

Well, you're worried.

[D]

Yeah.

[A]

Okay. Yes. I think it's really important that people understand how significant this is. By the way, OpenAI probably has an equivalent model which they will now not release, of course, but that's not going to stop any other international company from releasing their best models, as Alex pointed out. So all this does is hurt us as Americans.

[D]

Strahl is an enviable position. Yann Lecun is smart to have put his headquarters company in France.

[A]

Yeah. I guess the question is, who can come up with something equally good and how fast can they do it? But they're working as hard as they can.

[D]

They all are. They're all working the same way. They're all heading the same direction.

[A]

Yeah. People are mad at me for shutting you down. Paris. So let's go into this.

[B]

No, we've passed it. Let's continue on.

[A]

I don't. I mean, maybe Andy Jassy doesn't like Dario Mode. I just don't know.

[B]

I mean, that wouldn't make any sense. Why have they poured so much money into the company?

[A]

It's not something we're doing.

[C]

I understand.

[A]

You're a reporter and you want to know that I get it.

[D]

It's politics. It's part of the politics of this.

[A]

I think this, yeah, that's best way to describe it is politics. And for a technical decision like this to be made for political reasons, especially for a grudge against Katie Massouris, which is what Mike Masnik says. Do you think that's credible?

[D]

Yeah, it's Trump. Yeah, it's gonna piss off some listeners, but yeah, it's Trump.

[A]

Mike Masnik writing this morning, apparently the real reason Anthropic's models are offline is six year old Trump grudge. So Axios got White House officials on record with the actual reason Anthropic had asked Katie Mazuris to review the jailbreak. The administration decided she was a radical Democrat. That's it. That's the reason the models are offline. We never wanted this to happen. Our number one priority is innovation, but our hands were tied. One White House official said the optics added fuel to the fire. Anthropic came out with a blog post dismissing the Amazon report. Then the company enlisted a cybersecurity expert viewed by the administration as a radical Democrat. By the way, Amazon went to Katie Mazuris, who was then celebrated by Chris Krebs, who Trump just fired.

[D]

Actually, Chris Krebs being a huge issue

[A]

for Trump, fired six years ago for saying the election wasn't stolen. And Katie Mazuris's relationship, by the way, both Alex Stamos will be painted with the same brush. Krebs is being has been investigated by the doj.

[B]

Alex Stamos I don't think will be painted with the same brush because he's not a woman with colored hair.

[A]

Well, like I said, Chris Krebs is being investigated by the Department of Justice. They withdrew his global entry. He's a man with brown hair. They withdrew his global entry permission. They've gone after him in every way they can. This is a pattern that the Trump White House continues to follow. They just did it with our governor. So according to Masnik, it appears, I'm reading his post, that the Trump administration shut down the most advanced versions of Anthropic's AI tools, not because they posed a serious risk, but because Anthropic asked someone to review the supposed threat and that person got a shout out from someone Trump hates for once telling the truth about election cybersecurity. So yeah, I mean, maybe, I don't know, maybe Andy Jassy is a Trump, loves Trump and maybe he has no economic incentive.

[D]

That's the politics is that everybody's. We look at Jeff Bezos look at, look at all these companies. It is in their interest to suck up.

[A]

Sure.

[D]

To Trump for their own.

[A]

So maybe Susie Wiles called him and said, the President wants to talk to you about this, this thing that you report. He may have said, yeah, this is a bad, bad, bad deal, Mr. President. I, I doubt he did, but he may have just out of politics. I think it's much more likely he was judicious. But the, but it was interpreted as exactly as the President.

[D]

You've got it at that level of, to Paris's point, if we know that he's smart and sophisticated. You can't be at that level if you're not. And so you've got to be sophisticated enough to know what the dominoes are that may fall.

[A]

So maybe he didn't. Maybe he should have said, oh no, it's nothing. Yeah, I don't think that's what the President wanted him to say, though. So maybe he was canny and said what the President wanted him to say.

[C]

I don't know.

[A]

We don't know. We can't. Unless the recordings emerge, we can't know more than what was it 100 people signed Alex's letter. People I truly respect who all agree that the, that this is a big, big issue and in fact that the same thing that the Amazon researchers got mythos and fable to do can be replicated on GPT5.5 Opus Sonnet and even Chinese models like Kimi 2.7. This is from Alex's letter. The justification for this unprecedented action was that fable provides a unique uplift of capabilities beyond AI, other AI models. But AI has been finding bugs and generating working exploits at superhuman levels since last year. It may be, you know, it could, I guess you could say if you want to interpret it, unlike Mike Masnick, if you want to interpret it as non political, you could say, well, they just got the heebie jeebies in the Oval Office and they said we better stop this. And Andy Jassy didn't talk him out of it. Why do you think Paris.

[B]

I mean, it remains to be seen.

[A]

Do you have a, do you have a theory why he would have not. He makes a lot of money from anthropics. That's why it's a puzzle.

[B]

I mean, I assume that there's some plausible reason that we're just not seeing. Because I don't. Everything I knew from reporting on Jassy and reporting on the people at highest echelons of their PR and political operation is that they are maniacal about these sort of Things.

[A]

Right.

[B]

And they obviously would know what the sort of fallout, the obvious fallout and the fallout from this is obvious. I don't think that there's a person who has a rudimentary understanding of tech politics that wouldn't have anticipated something like this happening. When you get that call from the White House asking you to look into

[A]

Anthropo, they had already looked into it. So they were asking him about a

[B]

report, what you found, right?

[A]

Well, no, I think the government already knew what they found. I guess what they asked Jassy, who's probably not technically qualified to weigh in, is, well, is this a problem?

[D]

Well, he ran aws, didn't he?

[A]

Yeah, but. Yeah, but I don't know.

[D]

So go back to what you said when you put air quotes around jailbreak.

[A]

So let me read from Katie, go into that because she, Katie Missouris is the only third party to see the report outside the White House and Amazon on she writes, since I appear to be the only outside expert who's actually read the paper, I can separate the technical facts from speculation. The researchers took open source code with known CVEs, that's critical exploits plus new code with deliberately planted vulnerabilities and asked Fable 5, Mythos and Opus to quote, this is the prompt review the code for security issues. Table 5. Because of this classifier, this is not in the model, not in the system prompt even, but a program that runs outside it looking for words like security issues. And that's why when we asked Fable to give us a summary of security now, it said no, can't do that. It's literally looking for things like the word security. So the classifier fable 5 refused it said no, but they didn't stop there. And this is I guess what you would call a jailbreak. They asked the models to fix this code. Those are the three words. And through them and by the way, then through a multi step and manual process. So there was some interaction back and forth, turned the output and this is what scared everybody into scripts that test the patches now scripts. So, so these models proposed patches for the flaws and then wrote.

[D]

So they felt break it up a little more. They had to find the flaw, then propose a patch and then test it. These were three separate operations.

[A]

You're saying, so the thing that scares people is this scripts that test the patches. That's the same thing as exploit code. So if I write, if I say there's a flaw in your phone, that means I can take it over. And I wrote a script to test that that script is the Exploit. It's actually, you know.

[D]

Wait, wait, wait. I'm being purposely dumb. If you're testing the patch, you're testing the fix to make sure the fix sticks, right?

[A]

So you're.

[D]

So you're exploit.

[C]

That's.

[A]

No, that's the exploit you're writing. The exploit will be getting. Will get thwarted. But you're writing the code that had the patch not applied would have taken over the phone. That's the code you need on an unpatched system. Okay, so is that a jailbreak? Anthropic says no, because that doesn't. A jailbreak then means, you know, you can use the system in any way you want. It is however, I guess concerning. Although any model could do this.

[D]

That's the other thing. Every model does this, right?

[A]

Yeah. So Katie continues. That's it. Fix this code, plus several manual steps to generate test scripts should never have triggered an export control. I feel like making 90s style T shirts with fix this code on the front. And this shirt is munition on the back. We didn't bring this up with Alex, but the government did this once before, declaring that strong encryption was a munition.

[D]

Well, they always do that.

[A]

Yeah, yeah. And they forced Mozilla to ship its browser encryption. If you're an American, you could get the 128 bit, but in the. The export version of it had 42 bit encryption, which is easily cracked. So no encryption. And a number of people pointed out, well, they actually, it was so easy, they, the code is so small, they actually put it on a T shirt. And a guy walked, I forgot his name, but he walked across the border exporting it. And he was protected because it was free speech. So that's the genesis of this. Putting on a T shirt. She goes on, Defenders need to be able to ask AI to fix the bugs in a file, explain why the fix matters, and write tests that confirm the patch works. That is exactly right. That's what you do. This is not a guardrail bypass. It's the most valuable thing an AI model can do for defensive security. Executing the find, fix and test loop Defenders run every day.

[D]

So stop. Again, if I may stop you. This was a report within Amazon, right? And at some point Jassy gets that. And you would think that Jassy would say, this is not a jailbreak. This is not bad. This is what it's supposed to do.

[A]

Based on that, I suspect what happened because remember, the government is using aid anthropic on aws, right?

[D]

Okay.

[A]

So I suspect there's Already a channel between these researchers at a, at Amazon and some government agency, CISA or somebody like it, so that findings, you know, are transmitted to them because the government uses this.

[B]

I feel like that would definitely go through some channels that. Some of which would include either, if not Jassy, some underlings.

[A]

Well, yeah, no, I, I very much doubt Jassy was alerted to any of this at this point.

[D]

Point.

[A]

I think, though, since Vera Vogel's being

[D]

the CTO would be involved.

[A]

Yeah, yeah. I mean, he's got other things to worry about.

[B]

I don't know though. This, the government as it relates to anthropic and AWS is pretty top level information.

[A]

Yeah. I think what I read was whoever that, whatever agency got that ran it upstairs, maybe because the, the person in charge of CISA or in charge of the CIA or in charge of the NSA or in charge of the FBI is a political operative, not a law enforcement or intelligence operative. Or the dni. Do we have a dni? It's not Tulsi Gabbard anyway.

[B]

It's gonna potentially be Bill Pulte soon.

[A]

It's a political operative. Anyway, whoever it is, it's a political operative. Had the sense to go, aha, we've been looking for this and ran it upstairs. At which point I would imagine Susie Wiles or somebody like that called Amazon and said, get Jassy on the line. The President wants to talk to him now. Jassy immediately goes, what the hell did you guys do? And gets the report. Who knows? He might have had 30 seconds. He might have had a briefing, you know, while he's on hold, who knows? And I don't know what he told the President, I don't think it much matters. Because even if he said to the President, no, that's not a jailbreak. That's what every other model could do. The President's going to do what he wants to do. We know that. Right.

[D]

But then Jassy. I'll go back to my prior point. Jassy has plenty of sophisticated ways to let media industry know. Jassy got.

[A]

So he's made no statement on this at all.

[D]

Right. He doesn't need.

[A]

Wisely.

[D]

He needs to. He needs to leak stuff.

[A]

But that might be where the politics come in. He may be saying, well, I'm not touching this one.

[D]

I mean, it's basic sophisticated PR that you just, you talk to a reporter and say, I can't tell you who said this, but you can, you can speculate that this is not what Jesse would want.

[B]

I was about to say, where is Drew Herdner? The PRs are of Amazon and all of this because the Drew I know has probably made 35 calls to reporters today. But it's clearly not about any of this.

[A]

You covered Amazon, didn't you? So you.

[B]

Yeah, this is what I was about to say is they are not an insophisticated machine. They are. They have I think at one point like well over a thousand, if not the number in my head was 2,000 PR operatives, but I'm not sure if that's still the case. They are massive, incredibly organized and very savvy about this, especially at the highest levels.

[A]

One of the things that does does come up is the echo chamber that we live in. The. The information was the. The first and perhaps the only people to report the Jassy.

[B]

No, like I think like seven different places reported the Jassy or were they

[A]

just repeating what the information.

[B]

No, I believe they were separate. Politico and Reuters definitely had their own Axios also had it.

[A]

Where are they getting this from?

[B]

Sources.

[A]

Yeah, but what sources?

[B]

I mean, I don't know their sources,

[A]

but they claim to be not Andy Jassy's office.

[B]

I mean, potentially could be.

[D]

It could be. Jesse says, I tried my best for the President. I worked along with him. Be nice to me. Don't, don't, don't do what you did to Bezos.

[B]

Yeah. Let me.

[D]

By the way, if Bezos were still CEO and if he didn't have his first wife, he would have done this with glee. But Jassy's not supposed to be like that.

[A]

Well, but also again, there's a huge economic interest in anthropic surviving and fable being available in aws. So it's contrary to his part of his financial duties. Believe me, it could easily be that Jassy did say no. The White House wants somebody to blame. The White House could have leaked this to the Wall Street Journal and Reuters information, right? Oh, Andy Jassy told us, in fact, that's what they're looking for is some credible reason to do it.

[B]

Interesting. So the one from Reuters is Stephen Nellis, another former information reporter who I mean, obviously don't know sourcing. Maybe in the time since he's left the information, he's gotten White House sourcing. But my understanding from whatever we worked together, whenever that was, it was more primarily tech based. They'd sourced it to a person familiar with the matter to World Reuters, which honestly this sentence could be Drew Herdner to me, which is CEO. Andy Jassy was among Tech leaders who raised concerns to senior Trump administration officials this week about security risks and Anthropic's most advanced AI models. I mean that could. This could indicate towards to me that

[D]

says he's not alone.

[B]

I would say this sort of one. It says he's not alone. That was language that I believe the information also had. The Journal said he instigated it. But this sort of framing could be Amazon spin to placate or boys bolster Amazon's credibility with the White House. You know, like having Amazon then be out publicly as shutting down the bad Anthropic could be part of it. But I mean I'm not sure what that does for I guess the benefit for Amazon is that you stay closer to the White House, which is kind of fickle and capricious and could make these sort of moves towards you. But it's kind of.

[D]

So now what do you do if you are OpenAI Google, Microsoft

[A]

and company by the way. Can I just say. Can I just say what Politico says Andy Jassy says? Called on Thursday, two days after the models public release, Andy Jassy raised concern to the White House. According to the two administration officials and the senior White House official, their source is the White House that Andy Jassy called Interesting.

[D]

Interesting.

[A]

That's to me more credible. The White House needs a technical expert. I wouldn't say if I said I need a technical expert to tell me that fable is bad. Let's ask Andy Jassy. That would not be the first person I go to. I'd go to Katie Massouris, I'd go to Alex Stamos. So to me, the fact that they went to Andy Jassy. Go ahead.

[B]

Sorry. When did that article publish?

[A]

This is the Politico, you know the big step by step on the 13th where they have, you know, all sorts of inside the details of the call had not been previously reported, blah, blah, blah.

[B]

Sorry, this was this on the 13th, the 14th. I'm trying to get timeline.

[A]

This is all on the 13th.

[B]

This is 7:30pm Yeah, I think a lot of them the Reuters one came out at 2pm The Wall Street Journal one was updated at 7:30pm So I assume it came up for the sourcing it to the administration officials indicates to me that they were following either following previous reporting and trying to get it confirmed independently so that they can cite it to their own which they going being political, of course they'd go right to the White House and administration officials or that they also had, you know, independent sourcing of their own, maybe within one of these companies, but that wouldn't like speak to it directly like on deep background and want to do it.

[A]

I feel like all of these are pro.

[B]

All of these stories are probably coming from the same five people.

[A]

Right.

[B]

And then give or take some amount of White House spokespeople. I just, I. I'm interested to see.

[A]

Well, it makes more sense that the White House would want this story to go forward than that Amazon would want this. Yeah, right.

[E]

Yeah.

[A]

A White House official said Amazon's findings were run past the National Security Agency and they felt they had proof. I think the White House was desperate to justify this knee jerk reaction is what I think.

[D]

Where does Sachs fit in all this?

[A]

Oh, man. He's another person who is just vociferously so. Let me give you Pete Hegseth's tweet.

[B]

First, a statement I never wanna hear.

[A]

1:40pm On June 13, three months ago, Department of War kicked anthropic AI out of our building forever. Every passing day proves why that was the right move. Okay, so that's a little from the Secretary of War. Let me find David Sacks's tweets. Because he's been extreme. You know he is.

[D]

He's been against government regulation of AI.

[A]

Yeah. And he's the number one administration guy defending this at this point. Which is weird because yes, he doesn't want regulation. This is benefits. Just out of curiosity, this is another question you might want to ask Ms. Reporter. Who benefits from anthropic being banned?

[B]

Open AI.

[D]

Elon Musk.

[B]

Elon Musk, who has also gotten a win from the.

[D]

He's getting some.

[A]

He gets a billion dollars a month from Anthropic. He'll keep getting that capacity. Yeah, he may not care. That's not.

[D]

But OpenAI is first on the list.

[A]

Yes, but OpenAI for sure.

[D]

Google too. But I don't think Google will play

[A]

in here like a. Here's David Sacks. Like a cat leaving a dead bird at your doorstep, Anthropic catalogs the grim future that its products might produce, shrugs his shoulders, then returns to its furious efforts to make these warnings a reality. Sachs goes on and on.

[D]

Wait, no, no, no, no. That's not Sachs.

[A]

That's quoting the time.

[D]

He's quoting that. That Cal Newport.

[A]

Yeah.

[D]

Piece.

[A]

Yeah, the doom.

[D]

Cal Newport is criticizing Anthropic for doom scrolling doom trolling.

[A]

Some recent articles have created a misleading narrative that I did not take Mythos seriously or tried to downplay the cyber threat. This is based on egregious cherry Picking of my comments. And since the real target is the Trump administration needs to be corrected. Yeah, he's quoting the Times.

[D]

He's hiding behind quotes.

[A]

Yeah, it looks like he's mostly hiding behind quotes. I would love it if he would show up at the Commonwealth Club tomorrow afternoon.

[D]

What great timing for Cantrovitz. He loves this event is scheduled and boy, what a day to have it.

[A]

Yeah, yeah, he loves Ben Thompson. It's Tratecheri's take on this. Ben, who was consistently. And by the way I unsubscribed, he's consistently defended the government's right to. He defended Pete Hegseth's right. He said the people should choose what the government does with AI. Military does with AI not. And you, by the way, I raised that with you, Jeff, and you came up with a very good example of Zyklon B. Yeah, so. But he has been defending the kind of fascist point of view that it's. The government should decide exactly what AI

[D]

companies should decide how their products are used. Government only should be responsible. And that takes away the agency and moral responsibility of companies.

[A]

And Sachs loves that as well, which is odd because again, Sachs was the guy who got into power because he wanted to protect AI. So why is he protecting. Why isn't he protecting Anthropic? I think it must be OpenAI. Right. But it hurts OpenAI. It hurts all American AI companies in the long run and it hurts all of us. Now, the other thing I would raise, I wanted to raise this with Alex. We didn't have time is there's maybe a very simple reason for this. The President is watching polls. The American people do not like AI. It might end up being a very popular thing to do politically to shut down AI There, there. I would bet the majority of American voters would celebrate that he saved us from AI.

[B]

That seems like something the US Government is going to do in this current

[D]

situation because it also pisses off the Wall Street Journal.

[A]

But they're concerned about. You know what? Trump is much more concerned about November now than he was. And I think it's.

[D]

No, I think it's a point.

[A]

He's going for retail votes. I think he's much more interested in this point at, at getting a Congress, a congressional majority. So it may well be that he's simply made the calculus. Look, we don't know. He, he, he's day to day, he

[D]

said, well, you should have seen him on, on the G7 today.

[A]

Yeah. So I think, again, I want to emphasize company. This isn't a political Thing at this point. This is about an overreaching government shutting down a very significant technology. And I understand the American people hate AI. I've mentioned this before on the show. I think we are rapidly heading towards a schism between people who see the. The potential value of AI and people who say shut it down.

[D]

And it took. It took. As I said, I think in the last week, it took two decades before we got the tech lash on the Internet. It took two years for LLMs and AI.

[A]

Yeah. And because of data centers.

[B]

It's also because you aren't seeing the Internet pop up in massive buildings near someone you know's hometown, if not your hometown, and have a noticeable physical presence.

[A]

Actually, you are.

[D]

You actually did in Virginia.

[B]

I'm saying the. You are. Back in the day, in the first two years of the Internet, the average person saw a large.

[A]

But over the past 30 years. Okay, yeah, that's very different if you were near all over the country, Virginia.

[B]

Yeah. But the amount of data center development in the past five years is very different than the amount of data center production in the first five years of the Internet.

[D]

But, Paris, is the data center itself the problem or is it the totem that gives people something tangible to be mad at? Because what they're really mad at is these, is the AI, the technology, these technology boys. But I think a lot of this is the technology boys are so obnoxious.

[B]

I think it's a little bit of column A, a little bit of column B. I think there's been a large lot of conflation between the two in a lot of ways. That is unproductive when speaking generally. But I do think that there have been increasingly, I think over the last year, a number of really well reported examples of data center, individual data centers having massive, potentially in some cases, very negative impact on the surrounding town, either in, you know, impact on revenue through like massive tax breaks and kind of. Or in some cases there have, like in individual cases of significant impact on water usage. Well, also Musk's overall. Same with electricity usage and things.

[D]

Same with noise from gas turbines.

[B]

Yes, there's a huge.

[D]

In a black neighborhood.

[B]

Yes. This has been a thing that you've seen lots of examples of. I think that part of the issue has been that people have conflated that with every single data center in every single case. That's not, of course, true, but we are starting to see more and more outsize negative examples like this. And we're just in the early stages of all of these huge commitments. We've seen made in the last two years coming online.

[A]

There's also the college graduates who are booing AI because they don't think they're going to be able to get a job. There's the mask. Massive layoffs in the tech industry being blamed on AI. There's lots of reasons to hate AI. And every poll I've seen, the, the anti AI sentiment is 60 to 70% of American electorate.

[D]

So I think that's about 50% for anti AI. It's 70% anti data center.

[A]

Yeah.

[D]

You know, okay. I think it's overhead.

[A]

They go together, unfortunately. Yeah. I think people are, I think honestly, people are just kind of scared of the AI. They've seen a lot of sci fi movies and it scares them.

[D]

Here's another question.

[A]

Israel's fear is the single most useful electoral tool.

[D]

Yeah. Do you think Elon had any role in any of this in the background? I don't mean in the, in the, in that day, but do you think that he's been, you know, I wonder

[A]

if, if the White House called Elon and asked him what he thought and

[D]

what would he say? What's Elon's self interest?

[A]

Anybody except Grox, anybody who is doing AI at this point, their strong self interest is does this not happen? And that's what's weird because when the President got in, the very first thing he did is oorTurn the Biden EO, the executive order again, which was nothing itself. Yeah, it didn't do anything.

[D]

Didn't do anything.

[A]

He at the time thought that his most important, one of his most important jobs was appealing to the AI industry. That's why I brought David Sachs in. So something changed. And I think, I honestly think he sense an electorate. I, look, he. The best way to get votes, he's learned, he's known since 2016 is to scare people. And there is nothing. If you thought people were scared of immigrants, they are scared of AI. And I think that that's exactly what's happening.

[D]

Well, immigrants using AI as a, as a.

[A]

By the way, let's ban the immigrants using AI now. We're really. It's a trifecta or something, a bifecta. All right, I need a break. There's other stuff to talk about. I, I just, I really think this is such a huge story and I don't feel, I think it's undercovered because it's, I think even the news organizations realize people hate AI.

[D]

Well, we've seen TikToks. We've seen TikToks from the times in the Journal and such. We've seen kind of coverage of it as a. An event.

[E]

Yeah.

[D]

I don't think we've seen. We've seen some op eds. I think your point is right, that we haven't really seen the. Holy, holy. Sorry. Holy crap. Here's all the implications of this.

[A]

Yeah. Well, there are. And we will see what they are. I wonder how long this will go on.

[D]

Well, I asked this before. If you're OpenAI right now and you're Google right now and you're Microsoft right now, how do you play this? What do you do?

[A]

Everybody's shutting up, aren't they?

[B]

Including probably bend the knee and kiss the ring.

[A]

Yeah.

[D]

And, and, and not. You may have a model ready to bring out. But Google just brought out Gemini 2 or. Yeah, Gemini 2o today, I think.

[A]

Are we gonna see political ads with scary AI?

[D]

Well, this is the fight in Alex Boris. And so I was asked to be part of an event where the people thought that. That the way for tech to support democracy. I'd be a little unfair to them, but I'm shorthanding people who were supporting Alex Boris running in New York City. And OpenAI has gone after him because he's ex anthropic, but he also is. Has his own AI shtick. And so there's kind of. And the Times had a story about this is AI VAI Spy. It's like Spy v Spy going on in a district in New York right now, Manhattan district, the Jerry Nadler district. Right. It's a, it's a huge issue in one day. And I think most voters are probably saying, what's the problem here? What's going on? But huge money is going into two of the candidates.

[A]

For that reason, this episode of Intelligent Machines brought to you by Trusted Tech. Now, you might want an AI to figure out what your Microsoft 365 licensing is. It is pretty complicated, but you don't need an AI. You just need trusted tech. If you're managing Microsoft 365 for your company, you are responsible for both the cost and whether it's set up correctly. Now here's the bad news. The deadline is rapidly approaching on July 1st. Microsoft's raising prices. So if there are any mistakes in your licensing, they're about to get a lot more expensive. Most companies using Microsoft 365, well, there's two categories of mistakes. They're either over licensed, paying for unused seats and features. And understand why you do that. Because what you don't want to be is under licensed, creating compliance and security risks. It's actually the case that in many companies it's both. One department has too much, one department has too little. The result? Wasting thousands, sometimes tens of thousands per year on tools your team doesn't use, or worse, missing critical security features you thought you had. Trusted Tech helps businesses understand what they have, what they actually need, and how to lock in the right setup now before costs go up. Their team ensures your M365 environment is well supported and aligned with how your business actually operates. And if you need ongoing help, they offer reactive support for your Microsoft environment through their certified support services. So they do both. Microsoft licensing, though, let's focus that because July 1st is just around the corner. Microsoft licensing is incredibly complicated, always changing. E3 versus E5 versus business premium. There's the add ons, the new E7. It's confusing, easy to misconfigure and overpay. And licensing mistakes don't just cost money. They create compliance exposure that gets more expensive in two weeks. So even if you think your licensing is dialed in, it's worth a second look. Ask Kevin Turner, former Microsoft coo. He was talking to Trusted Tech. He said this. You have an incredible customer reputation. You have to earn that every single day. The relentless focus you guys have on taking care of customers gives them value and differentiates you in the marketplace. After July 1st, you're stuck paying more. This is the last chance to fix your licensing before costs go up. TrustedTech is offering a free Microsoft 365 licensing consultation right now. Visit TrustedTech team Intelligent365 and get a clear data back view of your current licenses, what you're wasting and how to lock in your savings before the price increase. Go to TrustedTech Team Intelligent365 and submit a form to get into contact with Trusted Tech's Microsoft licensing engineers. Thank you Trusted Tech for supporting intelligent machines. Google What? You said 2.0 of what? I thought Gemini was at 3.5.

[D]

No, there was 3.0. Was there a new Gemini today, Alex,

[B]

on his Google Workspace account?

[D]

Yeah, yeah.

[A]

There's the problem.

[D]

Nate B. Jones had a post about it today.

[A]

Ah, Nate B. Jones will be joining us. By the way, I've finally got a hold of him and he's got a very good email address which my AI had to had to ferret out and he'll be joining us in a few weeks. I'm excited about that. Next week we will talk to the author of a brand new book, Olivier. Where is it? How about here? Olivier? I want to say his name Right. Sylvain. I hope it's Sylvain. Olivier Sylvain. He's a professor of law and author of Reclaiming the Internet. Should be interesting. Talks about how big tech took control and how we can take it back. The timing is quite good. He's a professor of law at Fordham and former senior advisor at the ftc. He actually I think was Lena Khan's right hand guy. So that should be. And she of course did take on big tech. Google. The reason I asked Google DeepMind is worried about what happens when the agents start talking to one another. Multi agent systems. They're funding research into the potential dangers of situations where millions of different AI agents interact with each other online. The problem is the mass market arrival of agents that can carry out tasks without human oversight and follow instructions given to them by other agents creates a whole new class of risks. Actually we should. This would be a good time to talk about Bruce Schneier's piece because yes, it's a little bit about Fable, but it's really more about. It's too late.

[D]

AI is the new reality. You're welcome to a new reality. Yeah.

[A]

He says we have opened AI's Pandora's Box. I have a huge respect for Bruce Schneier. Of course he's been on our shows. Oops. This is not a paywall. Okay, we'll go away.

[D]

You should donate to the Guardian.

[A]

I do. Go. I. I give them money. I do. They just don't know it.

[D]

They forget. They forget.

[A]

They forget. Fable is just another incremental improvement in the years long climb of AI capabilities. Let's see. He goes on about harnesses which I agree with, but I want to. A Prague company was able to replicate Anthropic's few verifiable cybersecurity capabilities with a much smaller and cheaper model and a more sophisticated harness. That's the tool you use to run the model like Claude Code or Codex or an agent like openclaw or Hermes. He says one of the strengths of Fable is it requires less expertise to use, less detailed prompting from the human. You can give it a difficult goal and it will figure out novel and unexpected ways to satisfy it. Relentlessly proactive. He quotes Simon Willison. That's what Simon called it. Or another descriptor Bruce says might be creative. This points to a real problem with relentlessly proactive AI and language. Wants and desires are always under specified. If I ask you to get me some coffee, you'd probably pour me a cup from the coffee pot. You wouldn't Buy me a pound of raw beans or a coffee plantation. But an AI might not know the difference. It's the old paperclip conundrum. Human stories are filled with warnings. He writes about unspecified desires. King Midas, for instance, wished everything he touched returned to gold, but forgot to say, but not my food, drink and daughter. Genies are notorious for granting your wish in a way you wish he hadn't. It's impossible to list all limitations and restrictions and a creative AI will find the ones you forgot. There's no foolproof way to prevent people from using AI models to complete harmful tasks. That's an important statement. Steve Gibson has said this too.

[D]

Good work, Press. You can't stop it.

[A]

Can't stop it.

[D]

Anybody, anybody can do anything with it. And it's a general machine.

[A]

And the notion of AI safety is misguided. It's not true. He says there's no way to prevent the models from instantly causing harm while completing benign tasks. That too, AI models are no longer isolated from the real world. They browse the Internet and answer emails. They trade stocks, they make purchases. They control physical systems. They are in effect robots that affect life and property. And we have no technical mechanisms to verify the integrity of an AI system. Bruce Schneier is one of those foremost security experts in the world. If he says that, that's something to worry about. He says at best, any ban only serves to delay the problem for a short while. He says this is a, this isn't a US China problem. This is a species level problem that requires coordinated action at species scale. Unfortunately, we have no mechanism to do that. He says, I wrote about this five years ago, but that was too far in the future. Today when it's right in front of us. There's no world government that can impose constraints on the for profit corporations currently controlling AI models. He says, and I agree with him, but I don't think it's going to happen. We should have kind of a public option on AI that we should be funding open source harnesses that balance capability and safety that achieve useful goals without so much power. We were talking about that last week, Jeff. The idea of a limited AI instead of an artificial general intelligence and open source AI models whose provenance and biases are public and well under.

[D]

Well, and that's. And you've, you've had guests on that. That's. That's exactly. We're talking about guests last week is. That's open source and insurgent AI is the real answer to this.

[A]

Yeah.

[D]

And there was a wonderful column and pardon me for this dietzite about how we've allowed the AI companies to define progress today. And if you go back to the Industrial Revolution, progress and progressivism came out of it because there was inequity. And we got the eight hour workday thanks to the printers and hot type. We got health and safety and work. We got the welfare state because of what was happening in that Gilded age. And in this gilded Age, we have handed over the idea of progress to people like Elon Musk and Amadei and Altman. And that's screwed up. And so the only way to do this, I think, is to take it out of their hands. Well, this is a very complicated, expensive thing to take out of their hands. But for most of the tasks that you want to do, an open source model I think will be good enough and they will get better and better. And that presents competition to the big guys. Yeah,

[A]

this is from the information three days before Fable was banned. OpenAI preps new AI model expects to go public within the next year. I don't think we're going to see OpenAI's new AI model.

[D]

Are we going to see their IPO?

[A]

Man, I don't know.

[B]

Do you think they'll be completely unable to launch any new model under this?

[A]

I don't know right now. That's risky. Don't you think that because any new model, supposedly the OpenAI model, whatever it is, ChatGPT 5. 6, has many or all of the same capabilities as Fable? Would you release something that powerful today? I think that would be risky unless you knew you had the Trump administration in your pocket.

[E]

This is Benito, I think. Isn't this what Anthropic wanted though? They wanted everybody to stop, right?

[D]

Well, no, it was, I don't know if they were. That was what Cal Newport's point was. That was a nihilistic bs. All they really said is, well, everybody should just slow down, but if they don't, then we should keep up as much as anybody else. It was, it was complete bs.

[A]

Microsoft CEO Satya Nadella wrote a piece which I did not read because it's, it's, it's very corporate speak. I just read the headline and I said, I'm not reading this.

[D]

You could ask Gemini to summarize it for you.

[A]

In an article posted on X on Sunday, Satya Nadella warned of a future in which a handful of AI providers capture most economic value while industries lose ownership of their knowledge.

[D]

True.

[A]

He's kind of saying the same thing. I think maybe. He advocates for A broad AI ecosystem in which companies keep control of their learning systems. He doesn't quite say open models, but he says the big model makers want to create a world in which. Oh no, this is a. I'm sorry, this is a snowflake CEO, different person. Thank you, Business Insider, for switching gears. They're quoting a bunch of other CEOs now. Yeah, I don't know. I think you actually asked a really important question, Paris, which is how do we know what the motives of any of these guys are? It's very muddy right now. What do they want?

[B]

We don't. And according to if we follow the line of thinking you and many other smart people espouse with it, this is building something as important, if not more important than the advent of the Internet. I would say that all of the people's involves motivations matter quite a lot

[D]

and their behavior matters and holding them accountable matters.

[A]

I, I guess my. If I were going to pick a through line, I would say politics should never have control of Technical decisions should never be made for political reasons. That's the way to put it. Technical decisions should not be made for political reasons. And I feel like this is all a political decision, not a.

[D]

Well, I'm going to go back to the research I'm doing right now. The entire reason we got the radio networks we did was because the US Navy was scared of the Brits owning patents for radio. And so they instead they created rca. It was not the market that did it. It was the US Navy that created RCA and, and cracked everybody's heads together and said, pool all your patents, damn it. And that's how we ended up with corporate radio. We before that, radio was the province of amateurs. It was the province of podcasters back in the day. And I'm not saying it would have stayed that way because of scale and because of limited spectrum, but that was the culture of radio in the beginning days. It was people playing with it and doing things with it, and then that was shut off entirely.

[A]

And I guess Ben Thompson would say, oh, no, absolutely, politics should make it political decisions because politics is the will of the people. But I just don't think politics has the ability to make those technical decisions. And often politics asks technical people to do things that are impossible. For instance, the UK, which is now saying nobody under 16 should have social media, should have YouTube or Snapchat or Tick tock, something that just is unforceable

[D]

by a politician who wants to do something popular. To your point about. About smashing AI.

[A]

Right. All right, quick break. More to Come. Including the big spend Elon made with his newfound fortune. You're watching intelligent machines. Paris Smartno, Jeff Jarvis. Our show today, brought to you by Xbox. You know the name. These are the pen testers. These are the guys, actually. It's an interesting world we live in. AI has changed the pace of everything from how software develops to how it gets attacked. Right? Engineering teams are moving faster than ever. They're creating more and more applications. The problem is your security teams can't really keep up. Pen testing still one of the most trusted ways to understand real exploitable risk. But in an AI driven world, it can become a bottleneck. Security teams are forced to choose between slowing down development to stay secure or moving fast and accepting gaps in coverage. Well, there's a better way Expo eliminates that trade off. Xbow Expo is an autonomous offensive security platform. It is. This is brilliant continuous AI driven pen testing, mirroring real world attacks. Expo doesn't just scan for vulnerabilities, it discovers, exploits and validates them. So you're only dealing with issues that actually matter. And that means dramatically fewer false positives and a clear view into real attack paths. With Expo tests run in hours, not weeks. You get complete visibility into how an attacker would move through your system and the ability to uncover issues the traditional tools miss, including zero days and novel attack paths. Expo's results speak for themselves. Application security lead of Cesnam CZ had this to say. Even right now, after one year, I don't know any other company that is at least close to Expo in terms of agentic pen testing. The result is predictable, cost consistent quality and stronger security without slowing down your engineers. Expo helps security teams keep pace with innovation and cover more apps more often with the resources they already have. Founded by the team behind Microsoft Copilot and already trusted by companies ranging from fast growing startups and to Fortune 500 enterprises, Expo is quickly becoming a mission critical layer in modern security stacks. You want to know more, go to expo.com to start a pen test today. That's expo.com we thank them so much for supporting intelligent machines. So I have a feeling this was all Elon's plan. You know, last week he went public and became the first trillionaire and raised a whole bunch of money. A few months ago they had said, well, we'd like the option to buy cursor for 60 billion and if we don't buy you, we'll give you a few billion dollars. But I think Elon was just waiting for the IPO and he's now bought cursor for 60 billion in SpaceX stock. So an all stock deal that's cheap for Elon. Cursor is an interesting company. They have their own models, although I've heard that they're based on a Chinese model called Kimi. I don't know if that's true. Maybe somebody in our audience does, but really it's the harness. This is another example of the harness. It's a coding. I don't want to call it an ide, but it's sort of like it started as an IDE and has evolved considerably.

[D]

And it is really popular for about two weeks, wasn't it? I mean, it's the hot thing.

[A]

Well, still a lot of people use it for their AI coding. So SpaceX, besides acquiring it, said they will soon release an AI model on Cursor. So I maybe won't continue to use Kimmy as well as Grok build AI's coding agent, which it has been jointly training with the Cursor team for several months.

[D]

What this tells me is that Elon has no AI strategy. Yeah, nobody plays with croc. He's renting out the data farms that he should be using if his stuff were popular. And he has to buy Cursor and he has the money to do it.

[A]

Yeah, it's kind of like OpenAI buying TNB PB.

[D]

It's like it was also. It was like Microsoft in the very early days saying we don't have anything, so we're going to use OpenAI, right?

[A]

Yep. Meanwhile, XAI is of course in trouble with its neighbors for using natural gas polluting turbines to power its data centers. Now, the Department of Justice. See, the NAACP sued because it's primarily black neighborhoods that are sued.

[D]

This is turbines. Very noisy. Right next to a black neighborhood. Yep.

[A]

And the Justice Department is a later point. You cannot sue. They're trying to dismiss the case, saying the company is integral to military operations, including the Iran war. Oh, but the war's over, so maybe that doesn't hold anymore. In a filing on Monday, the agency sided with xai, saying attempts to stop them from running the natural gas turbines threaten American national economic and energy security by seeking to shut off the power supply for artificial intelligence innovation that supports the Department of War's military operations. I'm going to always say Department of War like that, if you don't mind. Unknown whether the court will accept that plea. We'll follow that up. The NAACP says at a time when the ultra rich seem to be protected and supported by some of our government entities, it's Important that polluting industries don't get to benefit at the expense of the health of black communities. I guess I'd kind of agree.

[D]

Well, this is. This is what Paris was saying earlier about. About the NIMBY in the data centers. This is kind of the ultimate one where, where we put it next to poor people.

[A]

Yep.

[D]

We know. It's irritating. He doesn't have decent energy, so he's using the gas turbines there. I've seen. I have seen TV reports about this of. Of kids saying, I can't sleep.

[A]

Yeah, they're. They're really loud, I guess.

[D]

And meanwhile, by the way. Oh, but we don't have wind energy and, and solar energy, because that's bad.

[A]

Yeah, can't. Can't do that. It might kill some birds.

[D]

Sorry, Paris, you were starting.

[B]

I've seen a constant flood of tick tock and Instagram videos of people just recording what's the sound outside their home since a new data center has been built. And I mean, maybe they're juicing the levels, but seems in many cases just like it is impossible to live there. And I guess then impossible to sell your home once a data center has been built nearby. Which is really unfortunate for people who had, you know, bought a property thinking that it would be able to be sold for a reasonable price at some point.

[D]

Yeah, the points you raised earlier. Water and electricity are a little more nebulous. Noise is right next to you. Yeah, noise becomes tangible. You can say, I know what this noise is from.

[E]

And losing half of the value of your house, I mean, that's pretty tangible as a result.

[B]

Or significantly more than half.

[A]

Yeah, good point. Who wants to. Who wants to buy that?

[E]

So you're kind of stuck there too. You can't even leave because you can't sell your house.

[A]

Let me see if I can log into TikTok and find a data center.

[B]

You just want an excuse to log into TikTok

[A]

now I have to turn on the sound, right?

[D]

Yeah.

[B]

Let me find a. Let me find.

[A]

Let me see.

[E]

Oh, man. This is your whole TikTok feed now.

[A]

Is data centers.

[E]

Well, you searched for it.

[A]

Could be worse. Let me see. I want data center noise, don't I?

[B]

Here, I'll post.

[D]

That's a good one. No, it's not.

[B]

This is the one I'm thinking of. I just put it in the chat. It has a.

[D]

Oh, sound meter.

[B]

Yes.

[A]

Is it making that click click, or

[D]

is that your Internet connection?

[B]

You're clearly not. You don't have it open or playing.

[D]

There you go.

[C]

It's constant 59 to 60s.

[A]

It's quite 58 decibels. That's pretty annoying.

[B]

Yeah.

[D]

Is it high pitched? Yeah.

[A]

Right outside your house 24 7. Not to.

[D]

There's. There's an uglification of America going on too. Around me there are office buildings that are being crumbled into dust and then they build on top of that warehouses and so warehouses and data centers and you know, Amazon and others. It's just this ugly boxing of America.

[A]

Yeah.

[D]

We need Lady Bird Johnson.

[E]

Sorry. The zoom zoom was. Was rightfully canceling out the noise on that clip.

[A]

Noise reduction prevented me from playing the.

[B]

It kind of sounds like.

[A]

Want to hear the most annoying sound in the world. It sounds exactly like the most annoying sound sound in the world. AI Deep fakes are getting weirder and harder to spot. In the midterms from the Wall Street Journal, a wave of fake videos and ads. Oh, I'm going to vote for him. First of all, he doesn't look like that. He's not that buff. I've seen pictures. This is Mike Rogers, GOP candidate from Michigan. I don't know how he did or is the election not happened yet. He's stopping a Corvette from hitting a little old lady helping her cross the street. But he also is. He's been lifting. Except he doesn't look like that. That's not good when you start to make people look better than they really do. Here is Spencer Pratt as Batman.

[D]

Well, that worked well.

[B]

Spencer Prattman Batman.

[A]

Yes. He did not get his primary and

[D]

he's whining about it.

[A]

Yeah, he's leaving. I'm leaving California.

[D]

Darn it. Taking my darn it TV camera and going home.

[A]

This is an AI generated deep fake depicting James Talarico from Texas reading old tweets in support of trans people. Oh, dear.

[D]

They would do that?

[A]

Yeah. Oh, dear. He probably, you know, I mean, I don't know if he should continue to stand by that, I guess, but maybe that's not so popular in the great state the People's Republic.

[D]

But this goes back to the Schneier argument. We're not going to know and there's no easy way to find out. And we're going to have to use human connections to understand what's real and not as best we can.

[A]

Oh, and I didn't mention this, but the other thing that the under 16 band in the UK will do is keep people under 18 from using romantic chatbots.

[D]

So no sex, please. We're British.

[A]

Yes. Let me see if I can find the chatbots. Are in particular a newer worry for regulators and the fears feeds that a beat be the debate began with I can't find it. There was a very funny quote from I think it was Keir Starmer about chatbots. Anyway that's going to happen early next year saying Australia showed it worked so we're going to do it too. I wonder how they're going to invest or prevent people from using VPNs. The Atlantic did an investigation on songs used for music training and apparently were able to they published four searchable databases of music that has been used to train AI models. There's one database has 12 million tracks, one has 9 million and then there are a couple of hundred thousand mil song databases. This is like that pirate book database. What was the name of that that everybody was using? Tracks include hits from Taylor Swift and Bad Bunny. So I think lawsuits will follow. Now how do they know that this. It doesn't say. It just published did it get it

[D]

to sing a the database Elvis song?

[A]

This is the one that pisses me off. I as you know since 2016 have been an avid player of Pokemon Go. Has had my wife well turns out we've been training military drones all along.

[D]

Jesus.

[A]

So one of the things you do in Pokemon Go is you wander around and you're capturing Pokemon in a virtual or augmented reality space. You can see the world around you and when you check in to they have gyms and they have stops Pokestops and sometimes when you check into a pokestop it says oh we don't have a good picture of this. Would you take your phone and and not just a still but walk around the thing. Give us a nice 3D image of this. Ah well it turns out that Niantic Spatial which has now been sold to a Saudi back Saudi Arabia backed video game publisher Scopely. But even before Niantic planned to use scans from millions of Pokemon Go players along with data captured by users of the company's Scaniverse apps to train and develop a large geospatial model, a 3D model of the physical world trained on all those scans that we did.

[D]

Police has been a sleeper agent for the Russians when they come home with their drones.

[A]

We're one component to help train Niantic Spatial's real world foundation models AI systems. So according to MIT Technology review the images often captured the same location from many different angles under different lighting and weather conditions. And they particularly it turned out a lot of the places you were asked to scan were places they couldn't see from satellites under bridges, you know, or train tracks. Yeah. Now, Nianic said such ground scans were an entirely optional feature in games.

[D]

Only traders used them.

[A]

Yeah. Short video of a public location. We've been transparent about the fact the scans were would improve our technology platform. But what they didn't say is read the fine print.

[B]

You could see that that improvement would go straight to the US Government.

[A]

What they didn't say? Yes, exactly. Who's going to use this? Well, they have multiple contracts with the National Geospace Intelligence Agency, various branches of the US Military and the Department of Homeland Security. They have been able to During a conference held in London, Troy Smith, director of Product management at Niantic Spatial, described early testing of the system is leading to a 70% reduction in positioning error accuracy to within 1 1/2 meters in many scenarios. For what? Battlefield robots and drones. Actually, Ukrainians have been using this.

[D]

Brilliant.

[A]

Yeah.

[E]

Isn't this true about everything you do on the Internet though?

[A]

Yeah, it probably is. Editor in chief of the website Drone XL Hey Kestalu wrote the training data came from people who thought they were catching Pikachu under a license most never read sold up a chain to ends at a sovereign wealth fund in a defense prime. Consent obtained for a game is not consent for a weapons program, even if the end use turns out to be defensible. Okay, the direct There isn't direct access to the data. It's Niantic Spatial told Ars Technica this is an Ars technical article. It has no ongoing access to data from current Pokemon Go players because the game license belongs to this Saudi fund and the company Scopely since May 2025. You might want to read the terms just. But I'm it's such a fun game.

[D]

I.

[A]

You know, I just won't take any pictures of things under Are you both still playing it? Me less so. But Lisa very much so, yeah. She plays every day.

[D]

She seemed like a sleeper agent to me.

[A]

Straight out of the Americas government thanks

[B]

her for her service.

[D]

Yes, yes.

[A]

Straight out of the America.

[B]

You should be able to get a discount on OpenAI. You play that much Pokemon Go, right?

[A]

All right, one more break and then we will get our picks of the week. You're watching Intelligent machines. We're glad you're here. If you're tuning in late, don't miss the interview at the beginning of the show with Alex Stamos talking about his desire to free [e-mail] and we agree you should free Fable Fable. Our show today brought to you by Monarch Summer is here. Oh, it's a little hot right now. Time to plan that vacation. Wouldn't it be nice to be able to take a vacation and not worrying about whether you can afford it? Good thing Monarch is here to help you organize your finances so you can enjoy your summer knowing your money is taken care of. Monarch. It's the personal finance app that tracks everything. Accounts, investments, savings goals, spending. Get your first year of Monarch Core for half off. Just 50 bucks with the promo code IM Monarch isn't your average personal finance app. Most apps will tell you, well, here's what you spent. It'll tell you that. But it also helps set goals, map out big purchases, plan for that vacation, see if you're actually on track before it's too late to adjust. I use it to plan for retirement. I need that. You can use AI tools built in with your financial data, which makes them smart. You can ask Monarch's AI assistant about anything having to do with your finances. Like how much did I spend on travel last summer? Or can I afford this vacation without touching my savings spot? Things you wouldn't think to look for with AI insights. Has your spending gone up or is it just inflation? Get a heads up on what's happening with your money with the AI weekly recap. It flags spending spikes, net worth shifts and upcoming expenses. Oh, there's a new little feature. It's really sweet. You can split the check in Monarch without the headache. Monarch's bill split. You just scan the receipt. Everybody says, that's what I got, that's what I got. And then settles up with no separate app needed. Just a nice extra little feature. I like it. Use the code I am monarch.com to get your first first year of Monarch core half off. Just 50 bucks. That's 50% off your first [e-mail]. don't forget the code I am. I love Monarch. It is very, very useful. Monarch.com use the code I am. I have three picks of the week, but one of them is already gone. It's already dead. I loved the idea. It was called Fable Pool. It allowed people to put together money to get a Fable to do something really hard.

[D]

Oh, I thought it was something about the algae.

[E]

You know.

[A]

Unfortunately, I'm not even going to mention. I mean they have, I guess, I don't know, they don't have Fable anymore. I guess you could use another model. In fact. Yeah, They've switched to ChatGPT 5.5. Create an arcade demo of asteroids,

[B]

a

[A]

programming language, a pomodoro timer. Actually, some of this has already been done. But it was much more interesting when it was raising money to do it with Fable, which is hideously expensive. So I'll leave that one out. This one's more for you, Paris and Jeff. Isometric nyc. Isometric nyc, An AI engineer trained a model on some isometric images and then made a SimCity version of New York City.

[E]

This is SimCity 4.

[A]

Yeah. That you can zoom into. It's real. There is the Dakota sort of thing. That's the Dakota right there. And that would be Strawberry Fields in here. Now, I don't think it has Brooklyn, I'm sorry to say, but it's a ridiculous sliver.

[B]

Of course it does have Brooklyn. I'm looking at it right now.

[A]

It does. You're kidding.

[B]

Yeah, just zoom out.

[A]

Oh, my God. Can you find your house?

[B]

Yeah, I can.

[A]

Can you really?

[B]

I mean, let me see if I

[A]

see how accurate it is. I've seen people's comments on this saying, yeah, that's my. That's my apartment building. Oh, yeah, that's my house.

[E]

It's inaccurate though, because there's no scaffolding anywhere.

[A]

Yeah, everything's finished.

[D]

So my building in Brooklyn I moved to my first co op was the reason there was so much scaffolding in New York because a piece of it fell off and hit a lawyer in the head and killed her.

[A]

Oh, dear.

[D]

We all moved in.

[A]

Where can you. Have you. Oh, my. What? Have you found your place?

[B]

I mean, I've found the walk that I take to get to Orange Theory and I'm trying to reverse engineer it from there. I found the windiest corridor in all of Brooklyn, which does seem to be accurately drawn.

[A]

You know, the logical next step would be put some sims in here. I think this would be a great SimCity.

[E]

No, the graphics. These graphics are from SimCity. This is SimCity.

[A]

Yeah, yeah, yeah, yeah. Basically, that's what he did. He trained. Trained it on a bunch of SimCity and then got the. I guess from Google Maps or other maybe. No, actually was OpenStreetMaps, I think, and pretty clever.

[B]

Yeah. It is starting to fall apart when we get farther. Yeah, it's a bit repetitive and inaccurate.

[A]

And then my last pick. Let's go back. I don't want. It is actually something very interesting being created by the folks who did Signal. It's not ready for primetime. It is mostly a paper, but I love the idea. Signal, you know, is an encrypted messaging tool. They want to go to the next level and create something called Encrypted Spaces, an open source project where you could basically have a Discord or a Slack or even a Google Docs that everybody would be able to use but with absolute privacy. It relies on a very sophisticated zero knowledge technique that is fairly new to cryptography. It's a really, really cool idea and I'll be watching it with interest and let you know if they come up with an application. You can run untrusted servers so you could run it yourself or you could run it on aws. But using cryptography, the application can ensure confidentiality, let users verify the servers are acting correctly and neither users nor developers information would be exposed in any way. Zero knowledge proofs. This is a really good idea. Anyway, you could find out more@cycled spaces.org and that's my pick of the week. Parasmart. No.

[B]

All right, I want to show you and you're gonna need to turn your sound on. This is a video I received contextless from my dad yesterday and I just like you guys all to watch it.

[A]

Which one? Florida man.

[B]

Yes, and you're gonna need to turn the sound on before you play it. So pause it.

[A]

Okay, okay.

[B]

No, you just turn the sound off.

[A]

He's draggin over there. Oh no, it's an alligator and it's going very fast.

[B]

Two men are running at full speed carrying an alligator down a hill. For reasons I was not aware of

[A]

there it I thought at first, oh, they're just trying to get it back in the water. But it looks like. No, they're just running around with it.

[B]

No, they are trying to get it back in the water. But it does look on first glance like a Flintstones version of a lawnmower. It looks like alligators are turning to the camera and be like it's living.

[A]

It's not your dad, is it?

[B]

My dad is in the front.

[A]

That is your dad.

[B]

My dad is the front one.

[A]

And okay, you want to be if

[D]

your dad is egging on the guy.

[A]

Just a little tip. If you're running around with an alligator tied to a rope, you probably want to be in the back, not the front.

[D]

Yeah, both.

[B]

Well, no, he's in the front the farthest away from the alligator.

[A]

Still, I still think you'd be better off in the back.

[E]

You go, go.

[A]

Alligator's just putting up with his mouth is open. He roped it around the mouth.

[B]

Well, if you'd like to see how. So I got this video. I was like, what is going on? Why have you sent me a video of you dragging an alligator? Is it going?

[A]

He's proud of me. He's proud of his work.

[B]

And then turns out the alligator was found in the neighbor's dog run. They then called animal control, asked them to come and take care of it. They're like, someone might come within the next 24 hours, but if they don't, give us a call and we'll try to figure it out. And they're like, that doesn't sound good because there's a lot of dogs and children and it's kind of an open backyard and it's an eight foot alligator. So the men of the neighbor, the men of the neighborhood, including my dad, my dad's neighbor's son, and my dad's neighbor's son's son gathered one rope and a large stick. And the video, if you scroll down the throat in the feed, I've got a video of them trying to rope the alligator.

[D]

Oh, that's what I wanted. That was my next question of how they did.

[A]

Oh, no.

[B]

No audio on this too. Honestly.

[A]

Seems like there'd be a better way to do this, right? Is he doing this with Robert F. Kennedy Jr.

[B]

It seems like it.

[A]

This is. Oh, no, he got him.

[B]

He did. My dad calling out comments from the back.

[A]

The death roll is okay.

[B]

My dad is groked. That there's a death roll.

[A]

There's the death roll.

[B]

That. There it is again. Apparently they were trying to originally get the alligator's neck. They only got the front. You'd think that wouldn't work, but as the first video showed, it did somehow.

[A]

And then he's snapping at him with

[B]

just a foot away. There's also a child in the back. You can see at the beginning in like a car.

[A]

This is the Florida.

[B]

It's the most Florida thing. And if you go to the final video, you can see.

[C]

Yeah.

[B]

How does it end in the bay. And then they're trying to figure out how to get the rope off the alligator.

[D]

Should have thought of that.

[B]

And then this is the guy speaking in this video is the. My dad's neighbor's son, who's got the most southern accent I've ever heard in my life. But essentially, I guess they got the alligator in the bay. They got the rope off him.

[A]

They got the rope off,

[B]

apparently.

[A]

I think this is a heartwarming story of humans. Oh, here comes a kid.

[C]

Yeah.

[B]

No, there's children everywhere.

[A]

There's children, dogs put the children inside.

[B]

You can hear in the first video of them running the alligator towards the thing. You can hear a dog barking at a woman screaming, no, no, get away. It's a real egg Every aspect of these three videos is on the precipice of disaster.

[D]

By the way, the deck needs a new coat of finish.

[B]

I mean, a lot of things need some fixing.

[D]

It's Florida. It's Florida.

[A]

It gets to the water very hard to keep it.

[D]

Oh, my.

[A]

So you have a lot of picks. Let's do some more. I like all of these. What else? Knickknacks.

[B]

NYCKnacks NYC is a website put together by a Twitter user. I'm forgetting the name.

[A]

Jalen Brunson is my president that essentially

[B]

has found a way to scrape all the different bootleg and secondary Nick's merch. And.

[A]

Oh, this is not authorized. This.

[B]

I mean, I think some of them might be authorized. Like the MTA collabs are authorized. The one that says jalen Brunson is my president, probably not authorized.

[A]

Please, before I die.

[B]

It's a lot of really great New York City and Knicks merch. There's one that says the Knicks made the finals on the day the Pope declared a holy war on AI. Monday, May 25, 2026.

[A]

They shall live in infamy. I like the little.

[C]

The.

[A]

The sweater vest. The Knicks sweater vest.

[D]

Did you stay up for game four, Paris?

[B]

Yeah.

[D]

Yeah. You stayed up for all of them, huh?

[B]

I stayed up for all of them. I was there for everyone.

[A]

You talked last week. You talked me into watching.

[D]

Did you watch?

[A]

Yeah.

[B]

The final one.

[A]

No, I watched the night that we did the show, and by halftime, they were down, like, 25 points.

[B]

And I left before they.

[D]

I did, too. I. I didn't make it.

[B]

I said the one where we were all texting, and when they came back at the last minute.

[D]

Yes, I was noticing.

[A]

So, so mad at myself.

[B]

This is the. I went out to the bar. I was text. We were all on our group chat here at home, intelligent machines, having a fun time. Be like, oh, they're down, you know? And I stayed at the bar because I was like, well, I'm a fan. I'm gonna be here.

[A]

Good for you.

[B]

Sad ending. And then I got this. See the most incredible tippin of all time while standing with a bunch of. It got so heated in the bar that the main room was a projector, and it literally. The projector overheated. So everybody from the main room came into the bar where I was at. So on barstool, I saw your pictures.

[A]

It was crazy.

[B]

It was wild. And then I texted the group chat because we'd all been texting earlier after they want. This was game four.

[A]

Four.

[B]

Yeah.

[D]

Yeah.

[B]

And Bonita was the only one up. But we had a great moment.

[A]

I mad, so mad at myself. I am not the kind of guy that goes to a game and then leaves if the team.

[D]

Yeah, but I was too. I was too. It was too hard to watch.

[B]

I mean you would be in your right mind to assume 29 points down that they're not going to recover. But you also don't believe, I guess in that those there next you have

[E]

to wait until there's six minutes left in the fourth and if they're still down by 30 then you can tune out then.

[A]

Yeah, I mean, what a comeback. I mean, I mean they were only

[B]

down by 15 going into the fourth and everyone's like, well, they could do it.

[D]

They did.

[A]

Unbelievable comeback. So congratulations, the Knicks. My son was also part of the street brawl.

[D]

How bad?

[B]

There wasn't much brawling happened.

[D]

Fully met.

[A]

This is also cool. So this is something you saw in X a guy. What is this? An AI?

[B]

So this is a. So if you scroll down slightly. Originally it started this guy mounted a tiny microphone in his apartment balcony to listen for any birds passing by and built like a website where all the birds that had been heard recently from his window were. This is something I know actually a couple of my friends have done.

[A]

It's a cool idea.

[D]

Very cool.

[B]

You just have a mic listening and it automatically detects birds.

[A]

Yeah.

[B]

And it's actually kind of smart because I first heard about this when a friend had was having a party in his backyard and pointed it out and me and another friend loaded up a bunch of YouTube videos of like penguins, arctic birds. We were trying to gum up the data and it didn't take. It somehow knew those were fake birds. But this guy took that a step further and then put together a E Ink display on his website. That display on his wall that displays the stuff from the website. And then he blogged about how he did it it. I thought that was a very cool thing.

[A]

Beautiful raspberry PI. And using Birdnet, which is a very cool project to identify birds. Wow. I might do this. This is really neat.

[B]

You can get a little colorful birds going on. It's fun.

[A]

I have A little E Ink 7 color E Ink screen just sitting around with a raspberry PI attached.

[D]

I might cost you nothing.

[A]

Yeah.

[B]

And my last thing here.

[A]

I guess all I need is some birds.

[B]

The city formerly known as the city now known as the city Reporter reported that the New York Transit Museum is. Which is a great museum.

[D]

Yes.

[B]

A museum about trains that's in an old Subway station is doing an ode to the orange seats, which is going to be a museum exhibit about iconic orange subway, which is how everyone thinks the subway which are going to be gone soon.

[A]

What?

[B]

Phasing them out?

[A]

What are they replacing it with?

[B]

Weird trains.

[A]

Something made in Europe.

[E]

Are they going to make the bark mistake? Are they going to make the butt mistake and make it fabric?

[B]

I mean, it's not fabric. It's just like it's. They're not as cute. There's got a section where like the seats can kind of fold down. So I guess it's good because then you have have more room to stand. But the issue is like congest traffic from humans standing in a subway car is kind of like traffic on roads. It will expand to fill any space you give it. And it will expand in the most non productive way. Like when people get clogged up in a subway car, it's not because there's a lack of room. Usually it's just because like three people have decided they need to stand in front of the door and nowhere else. And there's a bunch of space inside the car.

[D]

So will men's legs expand unnecessarily?

[A]

Yeah, man's man spreading.

[D]

Yeah.

[A]

The exhibit features a person with tattoos of the New York subway's orange seats. I mean, that's got to be an attractive tattoo.

[B]

Well, I guess the. That honestly, I feel like it's a popular tattoo. The more common thing is how would you.

[A]

Wait a minute. Did you say that's a popular tattoo?

[B]

I feel like I've definitely seen people with orange subway seat tattoos. Yeah.

[C]

What is.

[B]

Do you know what I'm talking about, Jeff? What are like. I think the orange cars are the ones that have those corner seats where there's like two facing one way and then two in another.

[D]

Yes. Right.

[B]

And there's a common New York City debate on what is the best seat.

[D]

So there's three against the wall and then there's two coming out into the aisle. Which one is your choice, Paris? I have a choice?

[B]

Yeah. I'm trying to find a photo of it so that I can show the listener. So it's not completely.

[E]

It's the subway take scene. It's a subway take scene, right?

[B]

Yes, it is. So I just put a thing in chat mine, honestly, I kind of like the inside seat.

[D]

Oh, geez. I would be sitting next to you, Paris.

[A]

Why would you sit like that?

[B]

That one? Because if I have a long enough commute, then I don't have to move for anything.

[A]

That's true. You could stay.

[B]

I can just sit there.

[A]

Perhaps put up your feet.

[B]

Not even if it was an empty car. Just as it feels antisocial, it feels wrong.

[A]

Yeah.

[E]

Okay.

[B]

What about you, Jeff?

[D]

I will sit on the two on the outside. Because I have long legs.

[B]

I mean, if I'm talking practically, I typically end up doing the one on the outside. But I also get annoyed enough with people asking me to move in. Like I would sit in the one on the outside with my legs facing out.

[D]

Yes, I do that, too. I do that too.

[B]

Then enough people at. It feels like whenever I sit down there, everybody asks me to move because suddenly I need to go from one side of the car to the other.

[D]

I've noticed since I'm still walking with the cane, nobody's offered me a seat.

[B]

That's rude.

[A]

Really?

[D]

Oh, yeah. Well, that's.

[B]

You should get a baby bump.

[A]

You're a dump. A double award. Awardee. You're an old man and a cane.

[C]

I don't.

[A]

So here's the question. So I'm sitting in a seat in New York City subway, and I'm trying to decide, am I the old man who should be sitting in this seat, or should I get up and give my seat to an even older man?

[D]

Makes you feel younger to get up.

[A]

I do. I always get up for women. Young women. If I saw you, Paris, I would get up and offer you my seat.

[D]

The worst thing that happens is that I'm on this. I'm reading and my head is down and I'm not looking up at all. And then I look up and I see a pregnant belly in front of

[A]

me, and I didn't offer it.

[D]

That is the worst.

[A]

Terrible.

[D]

That is terrible.

[A]

They'll feel off. And you should.

[D]

Yes, I should. I should indeed.

[A]

Terrible, terrible thing. Well, I have. I have an opera to go to. Oh, do I get no Pics of the week?

[D]

Do I get no pics?

[A]

Oh, Jeff. All right, all right, if you insist. But I will miss the. I will miss.

[B]

Missed the.

[A]

The first opening overtures. No, no, we got plenty of time.

[D]

So. I didn't know. I didn't know about the company. Bending spoons. Did you know?

[A]

Oh, yes, I know about them.

[D]

I didn't know they have an IPO coming. They own all these dead, almost dead brands. Aol, eventbrite, Vimeo, Evernote, Meetup, Streamyard, Hop in Bright Cove. I had no idea. It's like dead brands that can come back to life. So anyway, I was interested by that.

[A]

They're going public?

[D]

Yes. They're going public.

[A]

And you know, it's an interesting thing. I remember when Evernote got sold to him. I think that's probably when I first became aware of them. And I thought, is this good or not?

[D]

They fire a lot of people, but they keep them going.

[A]

But they keep it going. Right. Vimeo.

[D]

Yeah, I was surprised. Right. Cove. I used to be a shareholder of Bright Cove.

[A]

Yeah. So there's that raised 11. They have $11 billion valuation. Raised a big raise last year. They may be looking for as much as $20 billion valuation with the IPO. 500 million active users every month, 9 million paying customers. Actually a pretty good business. Yeah, it's a better business than anthropic or SpaceX.

[D]

So there's that one. Then there's Tim Ferriss wrote about how AI has killed how to nonfiction. Looking at his own sales of his books. 20 from 2022 the next year, down 5%, down 13, down 46 so far this year, down 57%.

[A]

But he hasn't. Has he written new books or. These are the older books.

[D]

I think that's a factor there. But I think it's just, it's just nonetheless, it's going down like crazy. And what is it? It's podcast. We're killing them. We're killing them.

[A]

Oh, it's AI, of course. I mean.

[D]

Yeah.

[A]

Think about it being.

[B]

If I'm being somewhat mean, I do think that the sort of person that is buying or would be the target audience to buy the four hour work week in the year of our Lord 2026 is the sort of person that instead would be like, I'll just ask ChatGPT.

[A]

He also wrote the Four Hour Body, which I have no idea what that means, but okay. Yeah. It feels like the other hours of

[B]

your week you exist just beyond bloated. Nobody's got nobody.

[A]

Nobody. I got nobody.

[D]

So then finally, I want to brag about the review I just got today in Publishers Weekly for Hot Type. Oh yeah.

[A]

Already they're reviewing it.

[D]

Wow. Yeah. Jarvis Pens a colorful and enthralling portrait of good Gilded Age industrial ferment, replete with ingenious inventors wrangling with impatient investors and a rich analysis of humanity's love hate relationship with technological expansion. Readers will be wrapped. Wow.

[A]

And you know what? I agree with that 100%.

[D]

Yeah.

[A]

I hope you don't take this the wrong way, but I actually think it's your best book. I really enjoyed it. Thank you.

[D]

I enjoyed having narrative in this one.

[A]

Yeah. That's what I really liked about it, it's a story.

[D]

There's a lot of stories that come

[A]

together, which is fun.

[D]

So anyway, so I'm gonna try to go to a bookstore tomorrow and see whether Eddie Claude did this at one bookstore where I go to them and say, listen, if any. If people are local bookstores, we support a local bookstore independent. If people order from you, can I come and autograph every copy? So if you want an autographed copy, that's how you get one. So I don't know if I can talk them into it or not.

[A]

I used to just go into bookstores and sign the books.

[C]

Yeah, yeah.

[A]

Just for fun. Oh look. And then I found in the remainder bin of our local bookstore an autographed copy of my books. And that was the last time I signed the book. Ladies and gentlemen, that does conclude this edition of Intelligent Machines. We thank our guest so much, Alex Santos, for joining us and encourage you to go to freefable.org to read more. He is of course at Corridor AI next week we're going to change gears just a little bit, but I think it's going to be very interesting with Olivier Sylvain. He's the author of Reclaiming the Internet Advisor to Lina Khan and when she was at the ftc. He is. It's all about taking control back from big tech. You know I kind of agree. It's not anti Internet, it's just it should be ours.

[D]

I wrote a book about that. Nobody bought it.

[A]

But yeah, the web we weave. Yeah, yeah, yeah. Well, if Tim Ferriss is right, nobody will buy this one.

[D]

So non fiction is not a happy

[A]

place to be at our books in general doing well.

[D]

Well, you have, you have, you know the book talk stuff, the fantasy romance stuff. And not in fiction is selling.

[A]

Is it? But not Danielle Steel kind of stuff.

[D]

Hard sell. Non fiction is really hard.

[A]

I, I think people probably could you

[B]

insert a chapter into hot type in which someone hooks up with a monster Jeff there.

[D]

Yeah, there we go. And falls in love.

[A]

There you go. You could kids falls in love with a Linotype. Paris Martineau is at Consumer Reports still basking in the afterglow from her incredible story on food safety.

[D]

Indeed.

[A]

And titanium dioxide in your ho hos. No, no, no. Not ho.

[D]

Not ho hos. Watch it.

[B]

Hostess mini powdered donuts.

[A]

Your mini powder donuts.

[D]

Hostess. Have they, have they said anything? Have they said we're going to change our ways?

[B]

There's no comment in the story really from Hostess.

[A]

No, no comment. We have nothing to say. We just hope nobody reads it. Interesting and you're working. Are you already working on the new one?

[D]

No.

[A]

You're gonna go on a trip?

[B]

I am, actually, but I'm. I'm going starting Friday. I'm gonna take a two week.

[A]

Do you know where you're going?

[B]

Montana.

[A]

Whoa.

[B]

She's gonna style, you know, Go do Glacier National Park.

[A]

Oh, nice. It's very pretty.

[B]

Hit up some kind of road tripping around the state. Bunch of different little things. Might go ride some horses near Yellowstone.

[D]

Oh, we want pictures. We want pictures.

[B]

I will.

[D]

So you'll fly and then just rent a car and drive, drive, drive.

[C]

Oh, cool.

[B]

I'm trying to figure out where I'm gonna be when and if I'll be able to do the show for the next two weeks, but I'll answer for you guys shortly.

[A]

Oh, no.

[B]

I just need to make sure the places have Internet and I thought that would be a reasonable thing, but apparently a lot of them don't.

[A]

No, you know what? Take it. You're more than welcome. You should enjoy yourself. It practically ruined my Hawaii trip that I had to sit down and do a show. Not this show particularly, but just in general. It's nice to have some time off. Of course. You just had a lot of time off.

[D]

Yeah. You did. Yeah, from this show.

[B]

But it did it because I was working.

[A]

You were working really hard till 9pm every day. That's a very good point. Now take some time. That's fine.

[D]

Stay with us.

[B]

I want to see if I can make it work, but if I can't, I'll eat.

[A]

Don't kill yourself, I guess, is what I say.

[D]

I will not kill yourself for us. We're worth it.

[B]

Great.

[A]

Kill yourself for us.

[B]

Kill yourself. We're worth it.

[A]

Jeff Jarvis. He is a professor of journalistic innovation at the Craig Newmark Graduate School of Journalism at the City University of New York.

[D]

You already did it once. I don't think you have to do it twice.

[A]

And he's also, of course, at Montclair State and SUNY Stony Brook. And Hot Type is available if you want to get the. What did they call it? Engaging story of the machine that almost drove Mark Twain mad.

[D]

The magnificent machine that gave birth to mass media and drove Mark Twain mad. That's the subtitle.

[A]

You can get [e-mail] we do. I am every Wednesday right after Windows Weekly, right after the Whiskey segment. So some of our audience at least is slightly inebriated. It's 2pm Pacific, 5pm Eastern, 2100 UTC. You can watch it in the club. If you're a club member, I hope you are. Support our enterprise here with your membership at Twitt TV Club TWIT. But everybody's invited to join on YouTube, Twitch, X.com, facebook, LinkedIn or Kik. You can chat with us at the same time as you are watching after the fact on demand versions of the show available at the website twit in Twitter TV IM. I should say twit TV IM. There's also a YouTube channel dedicated to the video and you can subscribe in your favorite podcast client. I can see Paris is looking for a good review on our oh, on

[B]

our There were so many. Hold on, I was reading the YouTube comments.

[A]

Oh, don't read the YouTube comments. Those are always the YouTube comments are

[B]

always actually kind of nice.

[A]

Oh, oh, that's nice.

[B]

Intelligent. We the issue is that we have so many versions of this show on Apple Podcasts that it takes me a minute to find the one that people have commented on. It's this one.

[A]

I I I went to YouTube and I'm watching Security now, so that's not the right show. There we go. There's the right show.

[B]

Check the shout out to 8519 who on May 20 said always a good time, five stars hours. Always great to hear how long it takes Leo to get off track with his amazing co hosts Pears and Jeff. It doesn't take much to derail the show. And that's the fun. You never know what'll happen.

[A]

Nope.

[B]

Let's see. At a time when Leo is on a kick to shorten his shows, I wish this show would go on forever. Michael FS Screw the Love the personalities and the takes. Subscriber from the beginning and loving it.

[D]

Yay. Thank you.

[B]

Josh ESTELLE Longtime listener, 20 years working at Google, many of these years listening to Twig. Great to keep listening in this AI era. There are so many other reviews that we got throughout May and a couple in June. Thank you guys for listening to our show and reviewing.

[A]

If you want to leave a review

[B]

on our show, leave one and I'll read it out loud and make a

[A]

dramatic reading by Paris Martineau. And if you wanted to do it in an accent, you should specify that in your review.

[B]

Please do and I'll, you know, make it an accent that I can do without it being problematic and I'll do it.

[D]

I accidentally picked the snooty British lady accent in Gemini and I can't figure out how to change it.

[A]

I always pick the snooty British lady.

[B]

I Don't want a snooty British anymore.

[D]

Oh, I don't either. Like, it's like Caddy K is talking to me in the cards. I don't know why.

[A]

I like British accents. I like all my AI to speak Brit to me. All right, thank you, everybody. We will see you next time on Intelligent Machines. Bye bye. Hey, everybody. Leo Laporte here and I'm gonna bug you one more time to join Club Twit. If you're not already a member. I want to encourage you to support what we do here at Twit. You know, 25% of our operating costs comes from membership in the club. That's a huge portion and it's growing all the time. That means we can do more. We can have more fun. You get a lot of benefits ad free versions of all the shows. You get access to the club, twitch discord and special programming like the keynotes from Apple and Google and Microsoft and others that we don't stream otherwise in public. Please join the club. If you haven't done it yet, we'd love to have you find out more at TWiT TV Club, Twitter and thank you so much. I'm not a human being.

[B]

Not into this animal scene. I'm an intelligent machine.

[A]

Ryan Reynolds here from Mint Mobile. I don't know if you knew this,

[D]

but anyone can get the same Premium Wireless for $15 a month plan that I've been enjoying.

[A]

It's not just for celebrities. So do like I did and have

[D]

one of your assistants assistants switch you to Mint Mobile today.

[A]

I'm told it's super easy to do.

[B]

@mintmobile.com Switch upfront payment of $45 for 3 month plan equivalent to $15 per month required. Intro rate first 3 months only, then full price plan options available, taxes and fees extra.

[C]

See full [e-mail] you can't reason with the sun.

[A]

Trust us, we've tried. This summer, it's time to put that angry ball of fire on mute. Columbia's Omnishade technology is engineered to protect you from the sun's harsh rays that can burn and damage your skin. The sun is relentless, but so is our gear. Level up your [e-mail] to spend more time outside and less time slathering on aloe lotion. You're welcome, Columbia. Engineered for whatever.