ジェンスン・フアンは「解任するには大きすぎる」のか？

Has your podcast stalled? Or maybe you have a show you keep meaning to launch and you aren't sure where to even begin. Well, let's figure it out together. I'm Jason Howell, and I run a podcast consultancy called PodTuneUp. I've actually spent the last 20 years building shows at CNET, TWIT, and right here on my own network. And now it's time to help you with yours. Go to PodTuneUp.com and book a free 30-minute call with me. We'll take a look at your show or your idea, find out what's getting in the way, and map out your next move. That's podtuneup, one word, dot com. Humans will never be more intelligent than AI. There's going to be two types of companies. Those are great at AI and those that went out of business because they weren't. How do we build a future that is human-centered? I'm Rana Elkayubi. and on my podcast, Pioneers of AI, we answer that question and so many more. As an AI scientist, entrepreneur, and investor, I know what it takes to build AI that works for everyone.

Every week, I sit down with the pioneers shaping our future and we take you behind the scenes of the AI that's transforming our lives. Find Pioneers of AI wherever you tune in. What's going on, everybody? Welcome to another episode of AI Inside, the show where we take a look at the AI that is layered throughout the world of technology. I am only one of your hosts, Jason Howell. The other host of this show is Mr. Jeff Jarvis. How's it going, Jeff? Hey, boss. Good to see you. Good to see you, too. Boy, we got a lot of really big stuff to talk about right off the top But before we do, two things First of all, last week I met up for coffee with one of our executive producers Wonderful Oh, that's great Yeah, he was in the area and so he hit me up like a month ago and said

Hey, I'm passing through, you want to get coffee? And so we went to a local Pete's and chatted for close to an hour And it was awesome So Mark, thanks for hitting me up and let's do it again sometime It was a total blast. He did have some feedback that I could share with you. He says, you know, the show focuses a lot on Frontier. He says, you know, what about sharing some of that spotlight with non-Frontier? And I totally agree. A lot of times when I'm gathering news, so much of it is just OpenAI and Anthropik and Google. And today's sort of in that camp as well. I mean, it's hard not to because they steal the spotlight, right? It's hard to know what else is important. Yeah, and the nice people who have open-weight models don't tend to be spies, so, you know. Well, that's true. Maybe it's a little less dishy, the open-weight models. That we know of, anyway. It's a little tease ahead. Also, I want to highlight something that you put in the rundown that I thought was super cool that you posted on your Medium account.

Tell us about the glossary for AI cultists. You folks all know we're talking about Tescreal and all that. I just got frustrated. I saw a headline in Politico that said that the Doomers have won the election about AI and data centers. And no, it's the exact opposite. The Doomers lost, and the good guys won, and everybody's confused. So I just said, finally, everyone is confused about this language. So I did the glossary of their language versus ours. So when they say future, they mean the way future. We mean tomorrow, next week, next year. When they say human, they mean humans hooked up with machines. We mean us. When they say doom, we mean doom, but they have their own flavors of doom. And most importantly, when they say safety, they're not talking about the things that involve the environment and kids and disinformation. They're talking about the 10 to the 54th future human beings. So as I wrote this, it ended up being a tutorial on Tess Crayal because that's the basis of all of this stuff that they have. I didn't even intend that.

But it kind of added all up to TestGrow rather than trying to explain it from the beginning then. So I hope you find it useful. It's super useful. Absolutely. By the end of it, I finished writing it over the weekend, and then by the end I saw that Politico had changed its headline anyway. So somebody got to them and told them. I wonder if they did it. See if it's useful. Pass it around, folks. I'd be grateful. And if you have any other words I missed, by all means, let me know. Yeah, and a lot of this, these are themes that come up on the show regularly, like, for example, regulation, what we mean, government intervention in an industry or activity to ensure a mandated level of transparency and accountability of the public good. What the AI called this means, the so-called frontier AI companies say they favor regulation, but in reality, they want to write the regulations that favor themselves and regulatory capture and all that. That's just one example, but there's a lot of really great stuff in here, and I definitely wanted to highlight it because I think it's a good manual. I might pass this along to a couple of my friends because, yeah,

I think the election has brought some of this stuff and along with it, some of the misunderstanding around these things to the surface again. And so this is a really great. Right. And if all the politicians think of, when they think of AI as data centers, they're missing the real point. And I think to your feedback from coffee, the attention that's paid to the Frontier guys is the wrong kind of attention. They're listened to and we accept what they say. We don't press them for definitions. We don't press them for what they're saying. And we need to do that as journalists. So, thank you. Yes. Love it. Cool. All right. So, today we're going to start with the story that we basically dismissed on last week's show. Yeah, I've got to do a mea culpa here. because last week I said, oh, the Times are reporting on Jensen Wong and the Chinese thing. I can't believe he's too smart for this. I can't believe da-da-da-da-da. I dismissed it. My fault. Never underestimate the libido of men.

Yeah. For getting into a bad honeypot. Kind of seems where this is headed. Over the weekend, basically, Jensen Wong, CEO of NVIDIA, basically admitted to having an affair with Alice Wong, no relation to Jensen, of course. I sent a memo to employees at NVIDIA basically saying, like, yep, this happened. I'm sorry. You know, I'm sorry. All the standard things that happen that we've gotten very used to hearing when guys of significant power are caught doing things like this. They say they're mea culpa. They hope to move on. But moving on isn't so easy when it's Jensen Wong slash NVIDIA, and NVIDIA is the company that it is both in the AI realm, which is kind of interconnected with everything in big tech right now, and just the world and politics because NVIDIA and trade, international trade, trade with China, all this stuff.

It's a whole combo also happening at a time, at least here in the U.S., where there was a major election. And as a result of that, a very significant transition of influence and power in politics here that's going to play a part here, too. So, yeah. So lots so much here. So, by the way, I just texted you. I put Justin Wong's apology through Pengram. and it said it was half AI written, which I think mainly makes fun of Pan Graham. Because some lines are just like, instead, I let them down and damaged the life we built together. I take full responsibility for what I did and the hurt I caused, and I accept the criticism that comes with it. Well, yeah, that's just boilerplate admission of an affair, right? Right. It's everywhere. So whether AI wrote it or whether the PR department wrote it or whoever. I think AI writing is based on human writing. Exactly.

Exactly. Yeah. So the memes, as I was saying, the memes are just delicious. No, we didn't mean that kind of Chinese model and putting Jensen in as Elizabeth Holmes with big eyes. Oh, yeah, that was a good one. That's a good one. The interesting angle to the story this morning, though, Jason, is I was listening to MS Now, and analysts there were saying that part of the story here is that Trump administration knew about this some time ago. Rubio wrote a memo warning of this some time ago. And now there's a lot of attention on the executive branch because, as it seems now, it's possible that they didn't alert Jensen Wong that he was having a relationship with a person who was apparently believed to be a Chinese spy. And by they didn't alert, you mean the government? The Trump administration. Yeah, the government. Whether that was the CIA or whether it was the FBI or whether that was other departments.

Okay. There are those who are saying that these various departments should have been alerted. They may not have been. Trump was told. And Trump met with Jensen Wong. Jensen Wong went to China with him. They sat right next to the premier between them. the question is to an extent you can say Jensen Wang one would hope would be cautious well he wasn't cautious obviously but knowing he's Chinese should have been at least aware of the possibilities here and I did say that last week is that everyone is on alert if you're dealing with someone who's from China whether they're the watcher or the watched but if the government knew and if the government didn't do what it was supposed to do, then the story backs up on them. And, of course, Jensen Wong is a major contributor to various Trump activities and has been around the president. And the government needs NVIDIA to succeed here.

So the government angle here, the political angle here, as the Democrats come into power in Congress, is going to be fascinating. Jensen Wong, he's going to have to appear in Congress in all kinds of embarrassing ways to deal with this. and he's going to be asked what he knew and what he didn't know about her and her activities, what did and didn't tell her, what is his security clearance, what information does he have, is he going to keep that security clearance if he has it. That's all fascinating stuff here to roil around. Right. As we figure out where this story goes. This is his position at NVIDIA, yeah, is his position at NVIDIA on shaky ground. He's too big to fire. And the board needs him and wants him, the company needs him. He doesn't have a succession plan. That's been an issue that some are worried about. I mean, we thought he was the good guy. And he's still one of the smart guys, and he's still one of the otherwise sane guys in the insane world of AI. But NVIDIA is the linchpin to so much, and he is the linchpin to NVIDIA.

And so I don't see any way that he can leave the company. Just how does the board have new oversight on him? Yeah, one of the many unanswered questions is, did he know that she was potentially... I mean, is she confirmed to be an asset? Or is it still just suspicion and passing? And again, it's China. So everyone... I'm not trying to say that... There's a sensitivity around that, a hyper-sensitivity around that. I mean, if you were East German at the time, it didn't come out to all your friends and neighbors who turned out to be informers on you until after the wall came down. And you didn't know. And in China, there are all kinds of pressure points on people. Yeah. And if you're in a certain position, then you're going to be asked questions. You're going to be expected to say things. Also, she bought, I think, was it 40,000 chips? And where did they end up? She was in a company providing offshore access.

One of the things about the Chinese not getting U.S. chips is all they need to do is hook into a cable to someplace in Singapore and use servers there. The Chinese could have access to NVIDIA chips by just not having them onshore. And her company was doing that offshore hosting. So that's her company, Megaspeed. That's what you're talking about. Megaspeed, yes. Yeah. So it's a complex story. If he was informed she was suspected as a spy and continued, that's one story. If he was not informed, that's another story. Doesn't let him off the hook for being dumb and following his other reasons, dumb to bad places. Yeah. But it becomes, again, a very different story. Now, should he have suspected? Yes. What did he tell her? Well, that's pillow talk. We're never going to know. We'll never know that. Right.

Yeah. So everybody's going to have fun on this one. Yeah. So apparently she, and she being Alice Wong, again, no relation to Jensen, same last name, but anyways, met at an NVIDIA event in early 2024. And then there have been photos of the two ever since. First photos of them at Computex in Taipei, June 2024. There's a video from May which shows her kissing him on the cheek. I mean, some of this feels really tabloidy, right? It does. And, you know, when I go to Europe, you do the kiss-kiss thing. Yeah, well, that's true. That's right. That doesn't necessarily mean anything. No. Although he is saying that, yes, I cheated on my wife and I'm super sorry. Well, and that's the other angle to this. I'm watching the show right now the HBO show War which is about a very high powered divorce it's really good interesting because what about

Jensen Wong's wife she and his kids work at the company his kids work at the company right in high level positions strategic positions and his wife has been a major basically advisor and supporter of his for the years and she has control, have control over shares that are in a trust and so on and so forth. If they got divorced, what happens to control the show? Of the company. Of the company, yeah. Just all these angles. And not to mention apparently, at least the reporting is there's no prenup in place so she would end up with half his fortune, which is somewhere in the region of like 150 billion worth? Is that right? I don't know what he's worth. That is just ridiculous that that's true. Well, it's very Amazon. So, Jason Wong's net worth is $148 to $155 billion.

Jeez. He's the 12th richest person. And that's only the 12th richest at that level. That just blows my mind. So the Jeff Bezos Mackenzie Scott was a settlement of billion This would be like double that Potentially. California. Potentially. Yeah, California, exactly. But the bigger question is then control and how secure is he? Yeah. And the board is filled with friends of his, which is not to be surprised. But he is critical to the vision of NVIDIA. You know, Sam Altman isn't an engineer. One could argue whether he can be replaced. He's really good at fundraising. Dario Amadei has his vision for Anthropic, but it's a bit of, in my view, crack putty, but it's still also a vision for the company. He could be replaced. Google's founders have been replaced.

Yeah. Bezos replaced himself. Right now, I don't think Justin Wong can easily be replaced. and it is so critical to the entire ecosystem of AI right now. Uh-uh. Yeah, crazy. So I put up this thing in the rundown of the comprehensive guide to the bank of NVIDIA. And I want to go to a great detail right now. All roads lead to NVIDIA. If you go through it, you see all the capital that is promised to suppliers, that is promised from companies that supplies. It has $99 billion in stakes in AI companies. It expects buybacks of $235 billion. It's given backstops of $164 billion to others, promised to suppliers $279 billion, advanced for parts and factories and all that.

it's a huge, huge economic crux to the entirety of AI right now and thus the entirety of the U.S. economy right now. All the growth we've had is AI. So you want to talk about too big to fail and too big to fire? They've got to find a way to put some controls on him. But even if you're the Democrats, I don't think you can afford to call for his head. the Republicans aren't going to because Trump has already said he's a good guy, he's important I don't think he can do anything and Trump needs him so it's been pretty vocal like no it's all good how does Justin Long manage going forward what does the next key don't look like I mean it's what a story and so once more folks I was wrong to dismiss this last week hey you know we don't know until we know hindsight is 20-20 all those things Meanwhile, Jason, get yourself some leather pants with a lock on them.

Oh, boy. Yeah, there's so much to learn from this story. The least of witches don't do that sort of stupid stuff. Don't, yeah. But anyways, yeah. And I feel bad for their family. I mean, this, you know, it doesn't get more high profile than this. This is the entire world knows about this intensely personal aspect of your life. And I just couldn't even imagine what that's like. You want to talk about a tough Thanksgiving dinner, man. Yeah. What exactly do we have to be thankful for right now, other than the fact that I'm worth $150 billion? Oh, boy. Yeah. So I guess we'll see. And I think it is interesting what you mentioned. And, you know, just because the transition of this moment of transition in American politics, it would be easy to think that, okay, so anything's possible. But you're right. NVIDIA is too important to so much of the economy, the tech economy, the AI economy, all of the above.

We talked about circular financing. And any time we did, I feel like NVIDIA is part of that story. If that changes. So what has been the impact on NVIDIA on Wall Street? Well, so the stock went down, I think, 5% the first day, 6% the second day. Today, it's up about 4%. So it's recovering a little bit. Yeah. So I think it's not going to go away. So if I had extra money, I'd probably say this is a dip to buy on. Never follow a buy. Don't follow our investment advice. Definitely. Well, wasn't there – there was also some short seller that saw clues of this that mentioned it like six months ago or something? Yeah. A Culper Research. Yeah, someone at Culper Research flagged the uncomfortably close relationship back in May. This was six months ago.

So, you know, a little while away. Yeah, I just – at the end of the day, I'm just really curious. It's like, how much did Jensen actually know going through this? And what did he say that was a value to her? So there's talk about this being a cyber security risk, but it's also just an industrial spying risk. Right. So what was more valuable to China? I'm not sure that it was security issues and national security issues. I think it was probably more industrial espionage. Well, so the U.S. suspects, reportedly anyways, that Megaspeed bought around $4.8 billion in NVIDIA hardware between June 24 and February 26, 2026, 2024. And then they say that the reason for that was to pass along those chips to Chinese companies that were barred from buying it.

But NVIDIA did an audit last year that found Megaspeed had about 50,000 fewer chips in its data centers than they had shipped, than had been shipped to it. So those two things together don't sound very good either. And we've talked about this the show often. Jensen Wong has been very vocal, saying that we're fools to cut off China from the chips. Yeah, I agree with it. And I think he's right, but it takes on a new nuance now. It does. Now it's kind of like, where's that coming from? Is that coming from a, like, under-the-radar coercion through this relationship? Or is that coming through a calculated kind of, you know, engagement through the relationship? It really is. This is going to be really interesting to find out what else about it. Was he just a dork? Exactly. Was he a pilot? Or was he going to use this to blackmail him? Yeah. Yeah. I'm thinking maybe there might be like a social reckoning

or a social dilemma or whatever that is. There's something here. There's something here coming up in the future, I imagine. It has all the makings of a good Hollywood experience. There are no charges, of course, at this point. There's nothing important to keep that in mind. This is still very much developing. and very much kind of speculation territory as far as like what. Yeah, I don't think there's going to be any criminal charges here. Because it would come from the Justice Department that's controlled by Trump. By the time we get two years from now, we're going to get a different Justice Department, maybe. We're going to be moving on to the next thing. Yeah. Well, speaking of moving on to the next thing, I, let's see here, I got a new machine, a new laptop. Oh, jealous, jealous, so jealous. I know, I know. I'm sorry, Jeff. but I got the Asus Google Book 14 delivered to me yesterday.

And, yeah, so I just received it last night. I unboxed it and then right away went right into Android Faithful Podcast. But I got to say, you know, you went to the event, so it's possible you saw this one in person while you were there. Yeah, this is, you know, pretty expensive. For the Google Books, it's right up at the top as far as the price of the others. It's $1,300 tied with the HP unit for most expensive of the first Google Books. It's really light. It's got a gritty metallic finish that's really nice. Looking at the display, it's matte, but it looks really nice in person. I've only barely set it up, and I think I need to retrain my fingerprint because every time I try and scan it, it doesn't work. Hey, it worked that time. So anyways, I haven't played around with any of the software here, which is kind of the reason that we would even talk about this on AI Inside, because it has the, let's see, if I jiggle the cursor, does it do the Gemini thing?

Or maybe I need to set it up. I haven't really set up anything. It's not doing it. Anyways, I'm excited to play around with this and kind of see what it's all about. But, yeah. So what's really interesting is a Acer. Hello. Hello. I've been using Chromebooks for 14 years. Hello. Yes. I went to the event. Interesting to watch the reception of this. Robbie Payne, who does Chrome Unboxed, does videos. They're great. They're really wonderful. And I hang on his word. He made a video that was very reasonable, very fair, and pretty harsh about the Google book and the things that it doesn't do. And not just like this versus that Google book, but Google. No, no. Everyone has raved about the hardware. I've heard no complaints about the hardware at all. Yeah. It's all software, which is all to say it's all Google's responsibility. Right. And they didn't think a lot of things through.

Some apps don't work well. Some things that should be an app or a PWA, whatever, that's nitpicky. There's just some stuff that they lost along the path here that doesn't work. they weren't ready to release. I think is basically what it is. So they came out with a new build, I think, five days ago or so that solved a bunch of problems. And I'm sure yours is on the latest build here. Maybe. I mean, I just unboxed it, so, you know, it probably needs to be updated. It probably does, yeah. And that fixed some things. It was using up too much RAM. Part of the advantage of this is because you're an Android god, when you're doing something on your Android phone, you should be able to just go right over and continue it on your laptop. It's really more of an Android machine than it is a Chrome OS machine. Yeah, but they lost some of the benefits of the Chrome browser along the way, and some of that doesn't work well.

Now, the other thing is that sometimes when you run something, it's running it on your phone, and that causes a lag, and that's an issue. So I would look at Robbie's video, Jason, and see how much you think was valid and how much has been fixed in the meantime. But the hardware, as I say, got great reviews. Now, the other thing that's happening is that these all have fans. What's the chip that I have in mind? The MediaTek one is not out yet. That's going to be a separable screen from keyboard, and a lot of folks are waiting eagerly for that. Oh, the detachable. Detachable, and a Mediatek chip, a lower power chip. So we'll see. I think some are waiting for that. Our friend Craig Newmark wanted that one, but it's not out yet, so he bought the one you have in your hand. Yeah, this one has an Intel, an Intel Core Ultra 5. Right. I'll be very eager to hear your review.

Yeah, I'm super curious to play around with it. I mean, at this point, I feel like I've heard and read so many negative reviews, like you said, of the OS experience. At the same time, I'm also kind of a little battle-worn when it comes to this stuff because I've been on Android for as long as I have, so I'm kind of used to things not working out of the box, so I might be a little bit more... I'm especially eager to hear you because you are such an Android expert and you live la vida Android. To see what advantages you see in it because you are so well-versed in Android is going to be really interesting to hear. I also want to hear about the AI stuff, the jiggling cursor. Babble is supposed to be on there. Oh, Babble. Rambler? Rambler. Rambler. There we go. Yes. I was like, Babble, that's like, I think that's like language for lessons. So it'll be interesting to hear how the AI works on there. So yeah. Yeah.

Yeah. Well, I will let you know I'm planning on spending the next two, three weeks with it to kind of see how it is both as a machine but also as like an AI computer and actually a little bit later we're going to talk about another AI computing story that could have really bumped up with this but we got to take a break. I do want to thank our patrons, patreon.com slash AI inside show. We do have a new patron to throw out the name here, Chad. Is it Stieb? Well in German it would be Stieb but I don't know what is Chad Steeb. That's better than my guessing. You at least have a point of comparison for it. When you join the Patreon, folks, if you have a chance to send Jason a note on the trance of your name, do that. It's helpful. And if you don't, I'm sorry if I mispronounce it. But it is nonetheless. We are very grateful to have you here, Chad. So thank you for joining us. You can join us as well. Patreon.com slash AI Inside Show.

Appreciate you all. I'm going to take a break. After the break, I'm going to give you a review of Gemini for Carbon because I've been using it for a few weeks. Jason gets all the goodies. Apparently, I get all the goodies. So we're going to talk about that after this break. The Wired Newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts. All right, so Google rolled out Gemini 4 Carbon just yesterday. Like, they had announced it a few weeks back. This is basically their next frontier model after Argon, which they had released only a couple of weeks before Carbon.

And I think that release kind of – it's confusing to me because when Argon came out, it was like, oh, finally, Google's stepping into the modern frontier. You know, they've been so behind. And then two weeks later, it's like, oh, no, actually, that wasn't it. This is it. This is the true frontier. That was just a step to it two weeks later. Anyways Carbon came out Google did the thing where it held it for a few weeks and was seeding it to cybersecurity professionals to battle test it and all that kind of stuff They also seeded it to some early access folks myself included So I've had Carbon for a few weeks now. And, you know, this is a model that you only get if you're in the $100 a month tier and up. So there might be a time where this comes down to the lower models. I'm sure it will at some point. Gemini app, AI mode and search, anti-gravity if you're of that tier. And Google is basically still recommending that Argon is for more of the everyday stuff.

Carbon, which is the latest one, is the heavy, right? It's the one that you want to use for real kind of like structural, deep research, coding, like real deep coding stuff. That's why it's really, you know, they're touting it as being so good at cyber defense and cybersecurity and stuff. It actually tops artificial analysis intelligence index right now, which is good for cyber defenders, that sort of stuff. So Carbon is out. With these things, it's like, okay, I have access to this amazingly powerful model. What the heck do I do with it? I don't know. I imagine you'd probably find yourself in the same position, Jeff, where I'm like scratching my head. I'm like, I don't even know how to test this. Like what can I do that somebody else can't do better, right? A lot of people out there, like I read their roundups and they're like, I built this. It's this hugely architectured thing that's really impressive and everything. I'm like, I just run my business is what I do. So I ended up – I showed off the design of the AI Inside website last week.

but this was made entirely top to bottom by Carbon. Little did you know. Little did you know. So this was done in a matter of basically two afternoons where I basically went into Carbon and I said, I've got a show page for the podcast. I don't feel like it represents the personality of the show. It doesn't match the new podcast art. I want a website that feels dynamic, feels modern, brings in the design notes of the podcast art into the design and everything. And I made sure to tell it, don't put anything – don't publish anything. This is all staging. And so the back end of our website is a service called Beanly, which used to be called podcastpages, I think, .io. And so that's kind of like our CMS or whatever. And so I basically just like gave it the keys. Like I logged in and I said, all right, I want you to stage a new website.

And from there, it was just two afternoons of going back and forth with it. You know, here's where I'm at. And then I'd go top to bottom. All right, and desktop. Oh, there's a weird little thing here. You know, I just do voice memo or whatever. Spend 10 minutes like telling it all of my feedback and then hit send. And then it would go to work for another hour or whatever. And it would send out all these, you know, it had these swarms of agents that it was spinning out in different directions to tackle different things so that it was faster. And in a matter of two days, I ended up with a website that I asked. Yeah, it was really great. I have to say, like, it really did everything so well, and it's such an upgrade from the previous version. How many back and forth, how many iterations would you say there were? How many iterations? If I had to guess. Yeah, it is hard to say. I mean, if I had to, oh, God, how would I even, you know, represent that in numbers? I mean, you know, there were probably like six, eight different kind of versions.

It wasn't like 30 versions or something. No, and I think that was actually, and that might even be more than there was. It was probably about six to eight. I could almost always find a little something that was weird or something that was like justified strangely or whatever. But for the most part, I mean, the first time I ran it and then it came back with the website, it looked very, very close to what you see on the landing page because I basically focused on the landing page first. But there were little details that it missed, like little hover things like this would pop out. The join the Patreon button would pop out if I hovered, but these buttons might not pop out or whatever. And so I'm like, OK, I want all those kinds of buttons to be uniform and to match their behavior, you know. And so I'd send it along with a bunch of other instructions. And it would just do it. And then it would come back. And I'm like, gosh, well, I'm really having to search high and low to try and find something that's wrong with this. There was, like, a static button here next to the new episodes every Wednesday. I was like, wouldn't it be cool if that was, like, a recording flashing indicator?

And it did that, you know. And then it, like, put that on every single page has the little flashing indicator, you know, no matter what page you go to. It's, like, consistent. This little breakout thing that highlights the important part of the title for every page. At a certain point, I was like, okay, but the fourth wall, which is where our merchandise page is, looks old and weird. It's a different site. It's not embedded in my site. I was like, I want the fourth wall page for our shop to somewhat match the design. And so it completely matched it. You don't even know that you're at a fourth wall page anymore. It looks exactly like the site, except it adds in the little payment. It's pretty freaking amazing. And this was a one-shot. This was like, just make it match. And I went over when it was done, and it was good. I was like, I have no changes. This thing looks awesome. So, you know, there are probably much better ways to test carbon. And actually, I'll talk about one more here in a second that I did. But I was incredibly impressed with how easy this was, how good it looks,

like how genuinely satisfied I am with it. Like, it's exactly what I was hoping for. So, I don't know. I consider that a huge win. Yeah. Yeah. I heard about how strong these models are for web coding, and maybe some of this stuff is stuff that other ones could do as well, but it has this swarm feature that you can activate, slash swarm, and that will break all of the tasks down into these individual things. So, in other words, it makes everything go faster, but it also kind of focuses the agent on that specific work instead of everything. And that's helpful. Yeah, and I think when we talk about this in present terms, we talk about swarms of agents. Yeah, that sounds so odd. In a way, it's subroutines. Yeah, totally. The thing is, it makes it as it goes. Oh, I need something to do this. Okay, I'll make a subroutine that goes and does that and then finishes it. And meanwhile, I'm doing something else, right? Yep. Now I realize suddenly, oh, in my work, I realize this would make this better.

This would make this more useful. I don't even know that you have to explicitly call for the swarm, but I did because I wanted to test it. And sure enough, you know, it did the trick. One other thing that I can talk about, but I can't really show it and you'll understand why, is part of my wife and I's kind of retirement plan is investment properties. We have a handful of properties throughout the country. And one thing I wanted to do is I'm like, I've got all of these disparate, disconnected data stores of information related to the properties. And we've got this one financial site that we use to kind of keep track of our expenses. And if we're doing our job right, lining up like property management reports and breaking them out for different repairs and everything and putting that into the whole log. And I knew that it had fallen really out of date. Like we try to keep up on it, but it's hard to catch all the details and everything. So I basically pointed carbon to all of those data points. There were probably like seven or eight different directions, you know, Google Drive, a folder on my computer, all these different things.

And I said, I want this online database to be brought completely up to date and verified for the past four years. And it spent and it put out a swarm of like 10 different agents. And it was cool to kind of check in on each swarm and be like, what are you focused on? Oh, you're pulling the property management reports from that property management website. Oh, you're searching my email for this type of data or correspondence or whatever. And it just did it. And it actually saved me some money. Like it found a duplicate entry, which actually saved some money. And it was just really impressive. It took a number of hours for that one to complete. But once it did and I verified and went through and checked, I was like, oh, see, this is always stressful at tax season. And now I can just be like, go. Carbon, can you find me some more money, please? Yes, I would love for you to find me even more money. So anyways, Carbon, huge fan. I think the kind of big takeaway for me is, again, and I think we said this with Argon,

but definitely feeling this with Carbon, is that Google feels more and more like they're kind of back in the front of the race along with the other frontier leaders. Yeah, you never call them out. They deserve to be there with this. Yeah, and I think this is also my first experience working deeply with Antigravity, which is, you know, kind of Google's standalone software for coding specifically, but you can use it for things that aren't coding, and it just approaches it differently, and I actually really like Antigravity a lot. So it was a ton of fun. Yeah, good stuff. And, you know, if you've got an account of a certain level, you have access to it now. I imagine that access is going to come down over time. Jason takes you inside. I try. I try. So that is Gemini 4 Carbon. Let's see here. What else do we have? We have Fable 5.2 as well as Mythos 5.2.

I think this is just kind of along the lines of agents and new models. Anthropic released their new versions of their top tier. And what's interesting about this, for me anyways, is Remote Labor Index, which is a way of – it's a benchmark basically for AI that tests if and how complete it can actually carry through on freelance projects. So the things that you might, unfortunately, hire someone for on like Upwork or Fiber, this is how they're testing it for that, right? So they threw 240 projects, things like video editing, CAD, game development, data analysis, all that kind of stuff. Stuff that I think a lot of these companies are saying, oh, this is going to do all that stuff someday. But then when you actually try and do it, you're like, yeah, but it doesn't do it nearly as well as a human does. And I think the kind of general stat to know here is that they basically, with Fable 5.2, they doubled what it was a year ago.

Well, actually, more than doubled. A year ago, the top score was 2.5%. Now, Fable 5.2 is at 41.3%. So nowhere near perfect. And they did a human – they had human analysis of the results to kind of check for things like, is the job complete? Is it comprehensive? Is it just, you know, is it saying, hey, I did the thing you asked me to, but it's incredibly flawed when you actually look at it and not polished? And it just kind of seems like these things are getting better and better at that. They are. They are. For better and for worse for those people on Fiverr who are doing some of these things. Unless you can do more because of it. Unless you can go to the details. Unless you can take charge of it. Well, yeah. Like, I think about my own use. So, on the Android Faithful podcast, I am the post-producer. So, I still edit that video. I don't edit AI Inside, except for today. But I generally hire an editor for AI Inside.

And I work with an editor to do post-production and publishing and all that, which is great. But for Android Faithful, I do it. And I have now created a system inside of Chachapiti. I actually want to test it with Carbon, where when I'm done editing the main show, I fire off the skill and I say, these are the three stories I really liked from this thing. Break out the clips into their own video files. Pull out thumbnail screenshots that could be integrated into a thumbnail. create the thumbnail in Photoshop, and then give me all the pieces so I can upload it to YouTube Studio. And I've done that the last couple of weeks on the AI Inside YouTube channel. And it's not perfect, but it does, I'd say, like 75% of the work. Hell, that's good. And then I sit down and I polish it. And it's such a better use of my time because I'm doing a million things all the time. So that's what this reminded me of. It's like, no, we're getting closer to that actually being an incredibly polished experience instead of just mostly.

Yep. Yeah. That's where these things are headed. Do we need to mention anything about the IPO? No. The founders are each worth $18 billion if they end up with the current price. It's going to be anywhere from $1.3 to $1.5. They're pushing it. you know and Dario was on some video this week around the socials predicting all these diseases Alzheimer's and cancer will be cured within 10 years because of his technology which is just irritating all these scientists doing all this work and they come along with the hubris to think oh we can do all that we'll see yeah we'll see about that keep your head down you can say all the things you want but how the heck do you know So you're just making a guess like everyone else, and who the heck knows. Yep. Yeah. Interesting stuff there. I feel like every time I open TechMeme, there's something about Anthropic IPO, and at this point I'm just like, okay, I get it.

I get it. Oh, yeah. No, we just kind of get to the point where we move on from it. But I guess we've got to wade through it. OpenAI had numbers out, and their losses were bigger than people thought, so they're really hanging back on the IPO. So at least we'll have a break, I think. Yeah. Yeah, hopefully so. Do you remember Cyanogen on the Android? Yeah, it brought in memory cells that I hadn't heard that name in a long time. Cyanogen was like an alternative operating system for, well, based on Android. It was an alternative version of Android that attempted, I would say, to right some of their perceived wrongs of Google's version of Android and to give users the ability to do things that Google might not allow them to do, right? A lot of people in the modding world, the ROM flashers and everything in the Android space, love Cyanogen.

Some of the phones that I have on the cabinet behind me in my shot certainly have some sort of Cyanogen OS flashed on it from a decade ago. Cyanogen was the most well name for an alternative OS version of Android on Android devices And Kurt McMaster was one of the guys leading it a co and he's basically saying that he's taking the lessons that he learned from Cyanogen, which has been gone for a number of years at this point. We haven't heard from Cyanogen because it folded. But he's taking it to go after Windows now with a company called Quora Systems. Which is essentially Quora is a Linux-based desktop operating system with AI agents built into it from the core, which is, I don't know, I guess the trend that we're going to see more and more of is this idea of why bolt AI into your product afterwards?

Why not just build it in from the core? Actually, to a certain degree, it's kind of like what the Google book is promising. He says, hey, look, this is a laptop OS with Gemini built in at its core. Just think of all the things you can do with a device like that. That's kind of similar to what Kurt McMaster and Quora are promising here. So the idea is that those agents, because they're embedded into the operating system at the core, They can work across the apps more seamlessly, access all the files, keep all that context in between the tasks that it runs. It can wall off the workspace for each of those runs, all that kind of stuff, or combine the context. And, yeah, it runs locally when it can, and then it can reach out to cloud models when it needs heavier lifting, which I think is another trend that we're going to see a lot more of as we go forward. So, I don't know. It's an interesting idea. Do we need a laptop OS like that or a computing OS like that?

It's hard for me to visualize exactly what it is, but I do think it makes sense that if we think that AI is going to utterly make us rethink applications, who needs an application? Tell it I wanted to do this, and it goes and does it. The same, I think, may be true of an operating system, that it doesn't matter what it is as long as it works with all these AIs and does what you want it to do, fine. Yeah, I mean, it really does call into question this continual question around, like, an app is designed for humans. App isn't necessarily designed for agents. The agent can use an app. Right. But if an agent is the next kind of phase of computing of how we use our devices, then why not build around the agent instead of around the human? although I can hear a lot of people screaming right now don't do that don't not build it for me just goes counter to what we've been using well I talked to a news executive at lunch yesterday

and I said there's going to be a parallel web that's built for AI and agents and you as a publisher should build your presence on that how was that received they didn't get it they were like no still very resistant yeah It still feels like it's way out there, but I don't know that it is. It feels like it's coming faster and faster. So this is going to launch in next year at some point, free to use. The money would come from subscriptions bundling those AI models and agent services. And, again, the idea here is that it wouldn't be – this wouldn't be a computing platform based on one singular AI model. This would be a computer platform that's designed to reach out to the right model for the right task or the ones that you've set up and then to know when to use that one versus this one. I think there it's important for the – if it's going to do that, for it to have a real strong sense of that because a lot of users,

although a lot of the users of this are probably going to be pretty nerdy and AI-focused and AI-driven, but a lot of people aren't going to know why you use that versus this for this or that task. Yep. You know, it's kind of like you've always talked about having a really easy-to-install capability, like installing an app, but for AI models for the people that just don't know. Yeah, right now the hardware requirement is too big. AMD, you know, the new box to compete with the Spark, that's $1,000 cheaper. It's $39.99. same kind of size, same concept but that's still expensive and I think what's going to happen is that Suleiman was talking about the training of apps is so inefficient as such redundant data that that's going to get boiled down and it's going to make much smaller models and then those models are going to be able to run on different hardware and then I'm going to see something that's game changing yeah very interesting

Okay You've got to get to a parade now I'm just realizing Yes, for those who aren't aware Here in the U.S. it is Veterans Day today Which is a holiday for many people There is a big Veterans Day parade that happens every year in downtown Petaluma where I live And my older daughter asked me today She said, hey, you want to go to the parade with me? I was like, oh, my daughter's asking me to do something with her. Okay, let's do this. Priority thing, yeah. Yeah, so I'm going to edit this. I don't think there's anything really important in the lightning round. You don't? No. I think you can go to the three. I know, I know, but we do have to take a break. As fast as I do. Okay, take a break. We do have to take a break. And then we will come back and we will lightning round. The lightning round is a lightning round. All right. Stand by. We'll be back in a moment. This episode of the AI Inside Podcast is brought to you by OutShift by Cisco. That's Cisco's incubation engine for frontier technology.

Here's where multi-agent AI actually breaks down. And it's not where most people think. It's not the individual agents. Those are getting really good. The problem is what happens between them. Agents pass messages back and forth, but they can't actually think together. And when they don't, you get cognitive failures, misreadings, unverified claims, false consensus, quiet underperformance that your conventional monitoring just isn't even going to flag. Well, OutShift by Cisco is building the solution, the Internet of Cognition. It's an open source foundation for building multi-agent systems from design all the way to production with shared context, shared memory, and the guardrails to actually get the results you expect from AI, not the results you have to debug later. If you're building multi-agent systems, this is where to start. Read the paper, experience the demo, grab the code at OutShift.com. That's O-U-T-S-H-I-F-T dot com. And we thank them for their support of the AI Inside podcast.

The Wired Newsroom is known for award-winning reporting on how technology shapes our world. On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday, wherever you get your podcasts. All right, the lightningest lightning round you've ever seen, although it doesn't have to be that crazy. I'm doing all right. I'm doing all right. Last week, we talked about the data center backlash because of the election results. I think a day later or two days later, OpenAI basically said, hey, we've got a national community compact. We're going to pay full electricity and grid infrastructure costs at data centers that we own and operate. There's not going to be any more NDAs with public officials.

We're going to publish our water, our power, our noise, and local hiring numbers per site. See, our data centers are friendly. Don't worry about it. We got you. Kind of, yeah. Yeah. Yeah. Oh, this is going to be the most lightning, lightning round ever. I'm going to shut up. You don't need to. I'm always the one who rescinds it. It's okay. No, the lightning round isn't meant to just be here's a bunch of headlines. It's all good. If you got anything, throw it in there. I put this in because of my meeting with Mark Starcher. Tencent says its new Hi-4 model puts it in the top tier of open models. It gets a full release. Well, it got a full release a few days ago. Weights, system cards are all coming in a few days. Artificial analysis is scoring at 35, which would be the second best from a Chinese lab behind Moonshot's Kimi K3.1 Pro, which is at 46.

So top tier, not the absolute tier. Tencent is definitely someone who we don't talk about a whole lot. So I wanted to, you know, give them a shout. Mark, if you've got thoughts on Tencent and their AI, send us a message on Discord and I'll include it next time. That's something we can do. I do think this is super smart. Reddit is turning its conversations into product reviews called Reddit Reviews. It's rolling out to all U.S. users throughout its mobile and web properties. basically organizes conversations about products into what people are saying about those products because, I mean, we've talked about it plenty of times on this show, that Reddit is such a valuable resource of information for how people talk about everything. And one big reason that people go to Reddit is because it's a great place to find how people feel about their experience with certain products and everything.

But you kind of got to go searching. And so this is Reddit's solution to say, hey, we've got all this valuable information out there about how people use and what they think about these products. Why don't we give them a more synthesized offering of what those reviews actually say on the site? And, yeah, it seems like a pretty valuable resource to me. Yeah, I'm curious what Redditors will say because they've been disintermediated here from having folks. It's like a little mini media empire where folks are going to read the summary of what they all have to say together rather than their individual opinions. It may be more efficient for the reader. It's not an unwise thing to do. But what's the benefit to the people who bring that content to Reddit? We'll see what the reaction is. Well, yeah. Does it disincentivize people from going deeper and going into the conversations and still continuing to share and everything about these things? I don't know. I kind of wonder if they're two different people because when I'm looking for a product and I go into Reddit to find the review, I'm not necessarily going there most of the time to participate.

I just want to see the thing and move on, which is different from the people who are in these threads. Yes, exactly. There's two different things about them. Use cases, yes. So I think they're two different users potentially. Yeah, I agree. Yeah, I will be curious to see, though, if it ruffles some. Because without those users, Reddit's nothing. Yeah, true. True. True. If this reduces that feedback loop, then that's no good for Reddit. But Reddit recognizes the value of its information for AI, and now it's kind of putting it to work for itself. And so that's the smart move on Reddit's part. Meta's Muse agent on Apple Watch and Wear OS. So you can look down at your watch and see that cute little cuddly bear or whatever the heck that is smiling at you and fire off. Following your orders. following your orders with a smile on its face, slurping your data with every last drop. And Netta also released, what is it, Ember, Muse Ember is what it is. It's lightest, cheapest model, 10 cents per million input tokens.

It's built for kind of the grunt work of agents, so tool calls, routing, data extraction, synthesizing long documents, that sort of stuff. Which again makes sense because I think a lot of that's got work can be done by lower level models. Oh, for sure. You separate that out. I think people are going to use those. So that's wise. Completely agree. And that is the lightningest lightning round we've had to date probably. Go to JeffJarvis.com for everything that Jeff is on the parade. It's all good. Jeff Jarvis, go there. He's awesome. Go to PodTuneUp. If you've got a podcast or you're thinking about podcasting, here, I'll open it up, and you want to talk with someone who knows a lot about podcasts, hey, that's me, podtuneup.com. AIinside.show with our brand-new redesigned website courtesy of Gemini for Carbon has everything you need to know about this show. Patreon.com slash AIinside.show for everything you need to know about what we do here.

We got our daily podcast five days a week. I get a steady stream of free subscribers, and I'm wondering if they're coming in from the one free episode that we do every week on Friday. Now the trick is converting into a five-day-a-week membership. So think about it. Patreon.com slash AI Inside Show. If you are an executive producer like Dr. Do, Jeffrey Maricini, Radio Asheville 103.7, Dante St. James, Bono DeRick, Jason Neifer, Jason Brady, Anthony Downs, Mark Starcher, a.k.a. my coffee buddy, Karsten Samachki, Evan Muziel. If you are one of them, then you get your name read out just like that. You also get, what do you get, a free t-shirt, all sorts of things. So, patrons, we really appreciate you. And our deepest gratitude and appreciation. And a huge thank you to Daniel Croft for some video support behind the scenes. Normally Anthony Lemos, but he's out this week. Happy Veterans Day, Anthony,

who is a veteran. And I will edit the show for Amos. So I'm going to get right to that right now. Thank you, Jeff. It was a pleasure. Always is. And we'll see you next time on AI Inside. Take care, everybody. Bye. This is Bill Bredesen, former Democratic governor of the state of Tennessee. If you love digging into how history shapes U.S. policy, check out the new season of You Might Be Right. This is Bill Haslam, former Republican governor. This season we're asking the questions that has defined every pivotal moment in American political history. Is democracy really under threat, or is that just a story as old as the republic itself? We're tackling the biggest questions shaping the 2026 midterms, so join us. Follow You Might Be Right wherever you get your podcasts. The Wired Newsroom is known for award-winning reporting on how technology shapes our world.

On Wired's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building, challenging, and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday. wherever you get your podcasts.