#216: Google I/O, Musk v. OpenAI Verdict, Andrej Karpathy Joins Anthropic & Meta Layoffs

0:00My belief is a lot of these AI leaders are going to be viewed as the villains, and it's hard to overcome that when you are literally building it and admitting it's going to reach the singularity

0:14but you don't have a plan for what that means. Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Reitzer. I'm the founder and CEO of SmarterX and Marketing AI Institute and I'm your host. Each week I'm joined by my co-host and SmarterX Chief Content Officer Mike Kaput as we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career. Join us as we accelerate

0:48AI literacy for all. Welcome to episode 216 of the Artificial Intelligence Show. I'm your host, Paul Reitzer. I'm with my co-host, Mike Kaput. We're recording a little early. We're doing Friday, May 22nd, about 10 a.m. Eastern time due to the Memorial Day holiday weekend. Our team is off on Monday. We didn't want to make Claire and Kathy have to work in the production side on their day off. We're recording a day early. If anything crazy happens on Friday and it's not in here, now you

1:22know why. All right. Today's episode is brought to us by AI Academy by SmarterX, which helps individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI-powered learning platform. New educational content is added weekly, so you're always up to date with the latest AI trends and technologies. The AI for Departments Collection features seven course series and professional certificates designed to jumpstart AI understanding and adoption. Those include marketing, sales, customer success, HR, finance, operations,

1:57and legal. The series are an ideal launch pad for organizations that want to level up their teams and accelerate AI adoption and impact. Individual and business account plans are available now. You can also buy single courses and series for one-time fees. As I've said on the show before, the cost of a single course series is actually, I mean, for a business plan, it costs more to buy an individual series than it does the entire year of programming. So we've tried to make the pricing as affordable as possible so that everybody can get in there and start taking the courses. So check out academy.smarterx.ai

2:33to learn more. And for individual plans, you can use pod 100 to take $100 off of that annual subscription. So again, that's academy.smarterx.ai and pod 100 on those individual plans. All right. Every week we go through an AI pulse survey. We give you the survey at the end of the show, and then we cover the results on the following week. So the pulse is informal polls of our audience to see how they're feeling about topics that we talk about on the podcast. Last week, we had two questions. How concerned are you about

3:07AI's impact on your own job over the next year? 36% not concerned at all. 33% not very concerned. 22% somewhat concerned and 9% very concerned. I got those right, Mike. Yes. That's good. Okay. Yeah. All right. And then are you personally ahead of your organization on AI adoption? This is interesting. So this mirrors, Mike, the state of the industry results. Exactly. Okay. So again,

3:41the question is, are you personally ahead of your organization on AI adoption? Keeping in mind, our audience is generally AI forward professionals and leaders. So it makes sense they would be ahead, but 60% answered way ahead of their organization. 24% somewhat ahead and 16% roughly on par. No one said their organization is ahead of them. So 60%, wow, that are way ahead. And again, that's not surprising

4:14actually for this audience. Yeah. Yeah. That's pretty validating for our state of AI for business research, too. Yes. Which we will talk more about, but you can also go grab that research if you miss it. What is it? State of business.ai, Mike? Is that right? That's it. Yep. Grab the report. Okay. So we had the Musk and OpenAI trial came to conclusion this past week, but we're going to kick off first with a big week from Google and their IO conference. Yes, Paul. So Google IO had wrapped this past week with what

4:46CEO Sundar Pichai framed as the start of the agentic Gemini era. He was basically signaling they are shifting from chatbots that answer questions to AI agents that take action on a user's behalf across Google's product stack. So a ton of announcements related to this and other Google initiatives. So first up Gemini 3.5 flash was launched and it's generally available May 19th, uh, 2026. It's now

5:16the default model in the Gemini app and Google's searches AI mode. Google says that flash outperforms last generations 3.1 pro on coding and agentic benchmarks. And it's roughly four times faster than other frontier models in output tokens per second. Gemini 3.5 pro is currently in internal use at Google. It is also rolling out next month to users. Google also introduced Gemini spark. This is a 24 seven personal agent running on a dedicated Google cloud virtual machine that is deeply wired

5:53into your Gmail docs slides and calendar spark begins beta next week for Google AI ultra subscribers in the US. So I believe that's their highest paid tier. It's a couple hundred bucks a month. And so if you have the cash and want to try this out next week, that might be a good, uh, spending of your money there. Google demoed spark doing things like parsing credit card statements to look for hidden subscriptions and monitoring school emails for deadlines with, uh, any high stakes action

6:23requiring user confirmation. Now, another thing that has everyone talking is Google announced what it called the biggest upgrade to the search box in 25 years. They are turning this into, in some cases, a dynamically expanding query field that accepts text images, files, videos, and Chrome tabs. Now AI mode currently now defaults to Gemini 3.5 flash and has also surpassed a billion monthly active users one year after launch. And they're also talking about these new information agents that are going

6:58to monitor the web 24 seven for you and look at user pick topics and actually report back on those. And those are rolling out this summer for AI pro and ultra subscribers. On the media side, Google launched Gemini Omni, which is an any input to any output model that has character consistency, physics accuracy, and conversational video editing. And that also, they talked about Google picks, which is a new image tool built on nano banana too. And what's interesting here is that Omni flash

7:32will be free inside YouTube shorts and YouTube creates starting this week. A couple of final things here. Google also demoed anti-gravity 2.0, the kind of second iteration of its agentic developer platform. It demoed that by building a functional core operating system core in roughly 12 hours using 93 parallel sub-agents for under $1,000 in compute. Closing the event, Google DeepMind co-founder and

8:03CEO Demis Hassabes told reporters that Google is, quote, at the foothills of the singularity. He cited rising machine autonomy, working agents, working code generation, and AI acceleration of science and math as evidence that the singularity era has already begun. So Paul, tons being announced here. We can unpack a couple of these. I thought it was interesting. Demis straight up saying we're kind of approaching the singularity. This is definitely an aggressive Google I.O. What'd you think?

8:34Yeah, Demis's messaging was very intentional. I'll kind of come back to that. We covered in April the Google Cloud Next event that I actually attended. And this certainly continued the messaging from there with this focus on agents. 3.5 Flash, as they describe it on the developer page, says it's a sustained frontier-level intelligence optimized for real-world tasks at higher speed and lower cost. And then as you alluded to, Mike, it's specifically designed for the agentic era and excels at sub-agent

9:06deployment, multi-stack workflows, and long horizon tasks at scale. There is an oddity about this model, though. Its knowledge cutoff is January 2025, which is peculiar because GPT 5.5's knowledge cutoff is December 1st, 2025, so full almost 12 months later. And then Sonnet 4.5 has a cutoff of around August 2025. So what that tells me is this is an old model. They stopped the training in January of 2025,

9:43and it sounds like maybe the harness is what is different about it. I don't know. It's just peculiar. That would be my guess because I think that's Gemini 3.1's cutoff, too. So I think it could be built on that, would be my guess. Just a more efficient model that's trained with the harness to do the agentic stuff, in essence. Yeah. There was a lot of talk online that 3.5 Pro was supposed to be announced this week, but it's actually being held up in review, potentially tied to what was supposed

10:14to be an executive order yesterday about the government's involvement in reviewing the latest models. So there's some talk that they planned to release something else and it just didn't get there. I've had very limited experience so far with Flash, but it was a lower quality, I would say, for the one use case I tried. It didn't know it could do something that it definitely has the ability to do, which is extract text from a PDF. And it was just like, I can't do it. And I'm like,

10:49no, you absolutely can. And I at one point told her, I was like, listen, chat GPT did it without a problem. Because I was going to get annoyed. I was like, I'm just totally straight up. So I actually flipped over to 3.1 Pro to do the same thing. And I had a really weird experience. Now this is not, this is just like a glitch and I'm just like noting because it's to kind of demonstrate how weird AI models can be. So when Flash wasn't working, I went into 3.1 Pro with the same

11:19personal project that I was working on that I needed to extract this text out of this thing. And this is what it said to me. So not, it didn't just say, I can't do it. It replied, here's the text extracted from your screenshot. And then it was like text messages, bro. So it's like, bro is saying this, we need to find out about his pension and social security. I don't know who to call for that. And then the next bubble was you. It says you, blue bubble. Okay. I'll figure it out. So instead of extracting the text,

11:55it like made up like a text chat between bro and me, apparently. And then I was like, I replied, I was like, that is completely incorrect. Like, look at the attached image. So it replied to me, I apologize for the mistake to be completely candid about how my AI systems are working in the current setup. I cannot actually see or process the image file you attached. Instead of accurately telling you that I lacked the ability to view the screenshot, I generate a completely fabricated response. If you are able to type out the text from the image or describe what it

12:31says, I am ready to help you navigate those specific details. So this is 3.1 pro asking me to type out the text that I'm asking it to extract. And then telling me it's just like made up an entire response. I was like, what the hell? That's crazy. It was just a weird. So totally just side note on how weird these models still are after all this time. So a few other notes, the significance of Omni, huge deal. It's a bit of a flex for them because OpenAI has had to pull back on their world model efforts and their

13:04video generation efforts because of the lack of compute. Google realistically is the only one right now that can really race forward on the world model side because of the infrastructure strength that they have. And Dem has talked about a lot about world models. We've talked a lot on the podcast about this. He sees this, as many researchers do, including Jan LeCun, as a necessary step for the next phase of AI advancement. The other note, and we were kind of joking internally about this yesterday, Mike, the product

13:34portfolio from Google, it's like you need an advanced AI model just to understand the product portfolio at Google. I have no idea how any of this fits together. It's so hard to navigate their offering and all of these features. We were actually working on trying to upgrade to Google Enterprise this week and our team went through hours of work just to try and even understand how to do that. Had to talk to a Google rep to even help us figure it out. And that's just a straight up like upgrading our standard

14:06platform to enterprise. And then when you get into all these other announcements and like, well, is it available or is it not? Is it built into Gemini app that I already have or is it not? Is it an ultra? Like it's dizzying to try and keep track of how this all fits in and what I'm actually supposed to be using? So the product people at Google, you know, it's tough. Like the innovation is moving so fast. It's hard to figure this out. And there's the innovations are coming from all these different areas within the company. And then like someone has to try and make sense of it all for the consumer.

14:38And so far it's tough. Like it's hard to figure out. And then the last notes I'll make is just this whole idea of the singularity. So we have talked at length about the singularity of concepts before. We've talked about Ray Kurzweil and, you know, his original book, The Singularity is Near, was I think 2005, Mike, if I remember correctly. Yeah. And then the singularity is Nearer, was 2024. And so the basic premise is it's this like hypothetical future point where AI becomes so advanced that it surpasses human intelligence and then just like

15:10hits escape velocity. Its ability to improve itself, like the recursive self-improvement we touched on a lot lately, starts to play a role. This is actually part of, we'll talk about Andre Karpathy in a couple of minutes. Like this is all connected that these AI can now start building smarter versions of themselves. And when that happens, you hit that escape velocity and it leads to what some believe is this future of abundance, but it also leads to potentially uncontrollable issues of something that's smarter than us. I guess a simple way to say it. And so

15:43the timeline Kurzweil has often given is AGI. So matching or exceeding human level intellect around 2029. He stayed pretty true to that. And as of right now, he's looking like he nailed it two decades in advance. And then anticipating the full realization of the singularity Kurzweil has said by 2045. So then why would Demis float this? Again, Demis doesn't usually say anything that isn't calculated. So in a semaphore interview, he actually said the decision at the end of the show was a very

16:15deliberate run. Quote, we debated it back and forth. Hassaba said in an interview a few hours after the event. He said, I was closing and I wanted to be authentic about what I'm thinking with AGI. The singularity, at least my interpretation of that word and that term means the era that we're in. So then it goes on to say this year, I really felt that it's the beginning. Agents are starting to work, becoming useful harness. Coding is starting to work properly. Areas of science and math are being accelerated. I've been reading, Mike, I don't know if you've read this one yet, The Infinity Machine by

16:49Sebastian Maliby. It's really good. It just came out last month, I think. I think I'm on chapter 14 now, maybe. If you want to understand Demis, DeepMind, the mission, how world models and recursive self-improvement and singularity play into it, I would highly recommend reading that book. And then the other couple of quick quotes was from an Axios article. He said he thinks AGI, when the machines are about as intelligent as humans, will arrive as soon as 2030, which he's been pretty consistent with that timeline lately. He also said the impact of AI is still underestimated, declaring

17:25that it will be 100 times as impactful as the industrial revolution. And that while it poses some risks, he said humans will harness the technology to solve many problems, especially in science and healthcare. So yeah, in synopsis, Google IO, an onslaught of updates with like dozens of features and products, continued confusion on the consumer side of what to do with all of these innovations and where they live and what they're available within and things like that. And then the grand vision

17:56that DeepMind is continuing to race forward and build this AGI and beyond future with all of us being like, what does this actually mean for everybody? You know, to that last point there, I did want to really quickly double click further into the search announcement because there's a lot of people like, what does this mean for us in the more very near term where, you know, I was doing a little research on what they actually announced and what they're kind of talking about here is, and I'm just kind of quoting from some articles from like Time and TechCrunch is like, instead of just that classic

18:31list of blue links and the typical AI mode, Google search will now also generate in some cases, not all cases, a custom page with an AI generated summary of what you're searching about, which is like AI mode, but then trigger a conversation with AI mode on the main page. So allowing you to basically just like chat conversationally with Gemini 3.5 flash at this point, which is really interesting. And again, Google actually posted a response to somebody on X saying like, hey, just to be really clear, basically,

19:04you will absolutely continue to see blue web links in search results because the question everyone's got is like, well, okay, AI mode is already eating into people clicking through to websites. But also, if you can then start chatting with follow-up questions and conversations right there in the search box, it's like, what is the point of a website anymore? And is your web content just feeding Google responses? So we'll see how that all plays out. I'm sure we'll talk about it more, but I think it's worth noting. And I just like can't escape this fact, Paul, that like long, long term, like not today and tomorrow, but longer term, I feel like

19:40I'm just operating under the assumption that 100% of the time, nobody will ever click through to your website from Google search again. And 100% of the time, someone will engage with your website through an AI agent. So like, I don't know how much that'll come to pass or how fast, but like, assume those things are true as a thought experiment. And a lot of things change. Definitely the organic traffic part, like, we've touched on this before, I've just been under that assumption for like two years. Yeah, when they started experimenting with AI mode, you could see it

20:11coming. And they were just doing this gradual experiment to get people used to it. And then you'll go from 10% of users seeing it to 20% to 50%. And then like how many times it shows up in search is the dominant thing. And so, you know, you can zoom out and just watch this happening. They're just trying to one, get the model efficient enough. And that's where 3.5 flash comes in, get an efficient enough model that they can serve up AI on every search without it bankrupting them, basically. Yeah. So that was one key is they had to get the models good enough and efficient enough to be able to do

20:46this. And now as the models keep becoming smarter and more efficient, it enables the user experience to be good. And then you deal with the ramifications of what does this actually mean to search and to media companies and publishing companies. And I think that it just has to work itself out over time, but they're obviously going in this direction and brands need to be preparing for this. Yeah. All right. Second big topic this week, you alluded to, Paul. So we've been covering the

21:17Elon Musk versus OpenAI trial the last several weeks. This past week, the jury came back and deliberated and the judge ended up giving a verdict influenced by the jury. That happened like literally right after our podcast episode, I think on Monday. But basically they deliberated for less than two hours and the judge rejected Elon Musk's $150 billion lawsuit against OpenAI and Sam Altman finding that Musk has failed to file his lawsuit within the legally required timeframe. So his claims are now

21:49dismissed after the jury's verdict. He had accused OpenAI as a reminder and Altman and President Greg Brockman of stealing a charity by attaching a commercial company to OpenAI, which was founded as a non-profit and taking more than $13 billion in investments from Microsoft between 2019 and 2023. Musk was asking for about $150 billion in damages. He wanted Altman removed from OpenAI's board and wanted OpenAI to unwind its move to a for-profit structure ahead of their potential IPO. Now, neither

22:22Musk nor Altman was in the courtroom when the verdict was read. Musk's lead lawyer, Mark Toberoff, gave a one-word comment outside. He just said, appeal. OpenAI's lead lawyer, William Savitt, said he was delighted with the verdict. OpenAI had basically argued throughout the trial that Musk's suit was a case of sour grapes from a competitor who could not beat OpenAI in the market. A few hours after the verdict, Musk started attacking the judge on social media, calling her a terrible activist Oakland judge, and saying the decision creates such a terrible precedent. He later deleted

22:57that post, but he repeated to Forbes that the decision set a dangerous precedent. In a separate post, he wrote that the judge and jury never actually ruled on the merits of the case, just on a calendar technicality, because basically they're saying like you waited too long to file this suit after you had issues with them, you know, abandoning the non-profit approach. This basically removes the last major roadblock to OpenAI's IPO. One piece does remain unresolved antitrust claims that Musk made against OpenAI and Microsoft. The judge said she was skeptical of those claims given aggressive AI market competition,

23:33but has not yet ruled. So Paul, seems like at least the major first round of this battle is over in favor of OpenAI. Yeah, and I don't, you know, Musk obviously didn't get the outcome he was hoping for. I have to imagine he knew the statute of limitations was a likely barrier to this going anywhere, and maybe why he reached out to Greg Brockman a couple days before the trial to try and negotiate something. Who knows? But he obviously created plenty of chaos along the way. And I think the irony of all

24:07this is, you know, comes up throughout the trial about him creating a competing lab against OpenAI in this whole process. And then he basically pivots away from XAI being its own AI lab and trying to be a frontier lab in the middle of the trial and, you know, condenses it into SpaceX and makes it part of that, you know, forthcoming IPO. And then he goes and does this massive deal with Anthropics. So it's almost like, all right, fine. If I can't like ruin OpenAI in trial, I'll just go deal with, go do a deal with Anthropic and give them all the compute resources that I built for what was going to be my frontier

24:42lab. And so we'll touch on this, you know, in the end too, Mike, with the, you know, product and funding notes. But so OpenAI is now preparing for to file for their IPO as soon as today, May 22nd. So that's coming when the IPO is TBD, but they're going to file very soon. SpaceX has filed, they're racing toward a June IPO. But in the process, they disclosed that their deal with Anthropic for compute is going

25:15to pay, Anthropic is going to pay them $1.25 billion per month. Why this matters, SpaceX revenue for all of 2025 was $18.7 billion. So their deal with Anthropic worth about $15 billion per year is almost the equivalent of all their other revenue combined. So when we go back like two episodes ago, where we talked about Elon Musk two months ago was basically saying that Anthropic was the antichrist. And then

25:48like now he's given all his Colossus one compute and now apparently leasing out Colossus two compute to Anthropic. Why? Well, because he needs to IPO. He's becoming a cloud infrastructure company, basically an AI infrastructure company. And the future of SpaceX is an AI company. So even in their prospectus, they highlight that their total addressable market is $28.5 trillion, by far the largest total addressable market in human history, maybe ever in human history to come. And space only

26:20makes up $370 billion of that. Their Starlink connectivity satellites, only $1.6 trillion. So all the other, whatever that is, $27 trillion or so comes from them being an AI company. And so it all starts to make much more sense why he folded XAI in, why he basically just walked away from being a frontier lab, because to get the valuation of he's hoping to IPO it like a $2 trillion valuation, you have to be this. And you have to have that revenue coming in from Anthropics. So like a lot

26:51of things can be forgiven if someone's going to give you $15 billion a year to make it go away. So yeah, I just like the trial is the trial, whatever. It was a soap opera. It exposed a whole bunch of stuff that I'm sure nobody involved really wanted to have exposed. But I almost feel like Elon was playing a game of poker and just kind of like waiting to like call their bluff and open it. I was like, fine, let's go. Like, it's all out anyway. Like, let's just have this thing. And then Elon was stuck having to go to trial and all this other stuff came out that even he probably

27:23didn't even want out there is what it is. They all move on with their lives and, you know, go become trillionaires, I guess. I don't know. It's weird. But yeah, the numbers for the Anthropic thing I think is like the hidden lead here. It's like that's Elon's alternative was just screw them by doing this deal with Anthropic. Yeah. So speaking of Anthropic, our third big topic this past week, some big, big news. AI researcher Andre Karpathy announced that he is joining Anthropic. So as a reminder, we've talked about Karpathy quite a few times. He is a founding

27:58member of OpenAI, former director of AI at Tesla, where he led the full self-driving and autopilot programs. He also founded an education startup called Eureka Labs. He posted on X, I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. Very excited to join the team here and get back to R&D. So he's starting on Anthropics pre-training team under team lead Nick Joseph. This is basically the large scale training runs that give Claude its core knowledge and capabilities. It's the most expensive and compute intensive phase of building

28:30frontier models. An Anthropics spokesperson told TechCrunch that Karpathy will launch a new team focused on using Claude itself to accelerate pre-training research, interestingly, part of this kind of industry-wide race to automate parts of AI development. Just as a quick reminder here on kind of all the stuff we've talked about with him and what his trajectories looked like. So I said he is a founding member of OpenAI. He worked there from about 2016 to 2017. During this early stint, we talked about this quite a bit. It'll be relevant. He led a project called World of Bits where he

29:04was basically making this attempt to teach neural networks how to use a keyboard and mouse to take actions on the internet. That should sound familiar today, but at the time it was considered a little too early. In 2023, he returned to OpenAI. He basically thought the rise of large language models would actually start making World of Bits a possibility. Then he left OpenAI again in February 2024 and basically pursued personal projects. He was kind of serving as AI's most prominent independent educator

29:39for a bit here. He was launching Eureka Labs to build what he called an AI native school at the time. And then now he is coming back to, and he's coming to Anthropic for the first time here. And as a reminder, I was just thought this was interesting, Paul, then I'll kind of get your thoughts. Like, you know, I knew all this stuff in my head, but it said, you know, as I was looking through Notebook LM to figure out when we talked about him. In the last year alone, he's coined the term vibe coding, which is everywhere. He had this whole conversation about intelligence versus agency,

30:13which is a huge part of kind of how we start thinking about AI. And he built that auto researcher thing we talked about several, you know, several episodes ago, where he's basically using AI to conduct autonomous research experiments. Like this guy's been so busy even in his outside of the major labs, like what's going to happen when he goes to Anthropic? He's the AI influencer that all the AI influencers follow. Like, you know, it's like he moves markets in the AI industry just with like simple tweets. So it's a huge deal. I think I tweeted at the time, like, man, Anthropic's

30:47valuation just skyrocketed. Like he could have gone and started his own lab at a $5 to $10 billion valuation at the seed round, like easily. Like he could have just said, all right, I'm back in the game. I'm going to launch my own thing. I'm going to raise $2 billion or whatever. And he would get it. So everyone would have taken Andre. Like he could have gone to any lab he wanted. There's been, I mean, I'm going to say pressure. People have been tweeting, influencers have been tweeting at him like, man, you got to get back in the game. Like he's, you're missing this greatest

31:18opportunity in human history to build and be a part of it. You got to be at a frontier lab. Like, what are you doing out there? And so I think everyone kind of assumed at some point he would. Elon's tried to recruit him on X. He would have been welcomed back with open arms to open AI. Google would do anything to get, like everyone would take him. The last time, Mike, that I think we talked prominently about him, I went back to episode 189, which was sort of a pivotal episode for us. It was the first episode in 2026. And that was when we told the story of how something had changed over the

31:50holiday break and everything. And we tried to kind of piece together, like what exactly had happened? Well, it was Andre that happened. He tweeted that like something was different. So the couple of tweets I mentioned back then, I'll just flag them. So we had Igor Babishkin, who is an XAI founder and former DeepMind and OpenAI. So on December 26th, he had tweeted, Opus 4.5 is pretty good. And then that's when Andre tweeted, it's very good. People who aren't keeping up over the last 30 days

32:21already have a deprecated worldview on this topic. And so that was like, people were like, whoa, wait a second. If Andre's thinking that, what is going on? And then Ronan Anil tweeted on January 2nd, my current theory. So he's trying to explain like, what the hell just happened over this like week period? Why is everyone going crazy and thinking everything is now different? He said, my current theory is that everyone was secretly using Claude code to do real work and seeing improvements from Sonnet, then Opus. Then once Karpathy said he used it over the

32:53holidays, everyone else followed. My entire timeline is Claude code heaven. Someone replied, even Google engineer, which Ronan was a Google engineer. And he replied, I used to be a Google engineer too, leveled all the way up and feel if I had agentic coding and particularly Opus, I would have saved myself first six years of my work compressed into a few months. So that sort of set the stage for this like chaos that has ensued since January, where agents just all of a sudden started being more reliable, more autonomous. Claude code started getting all this

33:24love. And then everyone started chasing Anthropic. Anthropic's valuation went from like, I don't know, 200 billion or something in December to 900 billion. They'll IPO at 1.5 to 2 trillion. All since December when Karpathy basically tweeted like everything changed. So yeah, he's a wildly important person in the deep learning era, let's say. So since 2011, if you made a list of the 10 most influential people, he's probably on the top 10 list of those people. It's interesting to just connect some dots

34:01with him working on pre-training, potentially using Claude to pre-train and also his work with auto researcher. He's certainly not the only person to invent trying to use AI to do research, but you have to wonder if Andre is the one that cracks like a recursive self-improvement basically. Mike, my guess is he thinks he knows the path and he needs the unlimited compute that Anthropic can bring to the table to do it. Yeah. Because why else are you going and reporting like you're three steps away from the CEO? You're not the number three guy. You're not coming in as the guy,

34:35but he's like, he doesn't even care. He seems like a very humble genius influence. He's amazing to listen to. Yeah. And I really feel like it's just, it took him seeing, I feel like he did with World of Bits. He went back to open AI in 23 because he thought World of Bits was close. So for him, it's all about the pursuit of unlocking the intelligence. Yeah. And if he looked and he's like, Claude code's amazing. I get unlimited tokens. I'm in. Like, I don't even care what you're paying me or what my title is. Like I'm back. So it'll be really fascinating to see what he builds.

35:10No kidding. All right, Paul, before we get into rapid fire, this week's episode is also brought to you by Macon, our marketing AI conference for marketing and business leaders. This is happening October 13th to the 15th here in our home base of Cleveland, Ohio. Big news this past week, we are thrilled to announce that Kevin Roos, the award-winning technology columnist for the New York Times, has joined our 2026 speaker lineup. Kevin writes The Shift, which is the Times' column on the intersection of tech, business, and culture. He's the co-host of Hard Fork, the paper's weekly tech

35:43podcast. He is the bestselling author as well of three books, including Future Proof. And he's currently writing a new one, documenting the race to build AGI. His keynote at Macon is called the AGI Chronicles. He will share the inside story of that race and how to navigate what comes next. Now, what's really cool is Kevin joins a stellar 2026 lineup that also includes Karen Howe, the award-winning AI journalist and author, Andrew Yang, former presidential candidate, and tech and economic futurist. Dan Slagan, SVP of marketing at Zapier. Paul, you, of course. And we

36:20also have a ton of other people within the marketing and business and AI worlds who are going to be speaking and more speakers soon to be announced. So this event is just the thing we look forward to most every year. It's three days of keynote sessions, workshops, conversations built specifically for marketing and business leaders who are actively figuring out how to adopt and scale AI across their organizations. So one really important note when you listen to this this week, ticket prices go up at the end of the week on May 30th. So if you register before then, you get the best price possible.

36:54You can also use the code POD100 at checkout and save an additional hundred dollars on top of the current rate. So to do that, go to Macon, M-A-I-C-O-N dot A-I. That's Macon, M-A-I-C-O-N dot A-I to register.

37:14Okay, Paul, let's dive into some rapid fire topics this week. So first up, unfortunately, this past week, Meta announced it was reassigning 7,000 employees into four new AI-focused organizations. Then it went ahead and laid off roughly 8,000 workers, which is about 10% of its workforce two days later. That has been kind of rumored for quite a while, and we finally saw it come to pass. So Meta's chief people officer, Janelle Gale, described the new organizations that are these kind of AI-focused

37:44organizations in internal memo as using AI-native design structures with fewer managers per employee than other parts of the company. For some context here, Meta employed more than 78,000 people at the end of 2025. The company had also told workers in April that it would be closing 6,000 open roles. Employees in the US received 16 weeks of severance plus two additional weeks where every year worked if they were part of those layoffs. Zuckerberg has basically bet the future of the company on AI.

38:14In January, on an earnings call, he said Meta planned to spend 115 to 135 billion dollars this year much of it on AI development and data centers. The company has also dialed back its Metaverse work and started including AI use in employee performance reviews. Like we talked about on a previous episode, in the weeks before these layoffs, Meta employees were already a bit on edge because the company had disclosed in April they had a new internal policy authorizing Meta to record keystrokes and mouse

38:45activity from company devices in order to train its AI models. Unfortunately, a leaked audio clip of Zuckerberg defending the practice surfaced the exact same day the layoffs began. So Paul, I've both read online and actually heard from people I know that this was a very rough round of layoffs. It sounds like morale at Meta is truly awful at the moment. It does not sound good. There's been some people

39:15who have been let go who have posted about their experience there and kind of being happy to to be out of there. I'm sure there's pockets of it that are good like it's you never know like we're not there we don't know the full story but from the things that are public it's not great. The leaked audio you mentioned I'll just read quick excerpt said this is Zuckerberg speaking. The AI models learn from watching really smart people do things the average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks. So if

39:46we're trying to teach the models coding for example then having people internally build tools or solve tasks that help teach the model how to code we think is going to dramatically increase our models coding ability faster than what others in the industries have the ability to do who don't have thousands and thousands of extremely strong engineers at their company. So it when you're told that you're training the AI to replace you it generally isn't going to go super great I would think yeah it doesn't give people the greatest motivation to teach it properly and then the memo that was sent by Zuckerberg on May

40:2020th to employees I'll just read three quick paragraphs it said this is the most dynamic I have seen our industry I'm optimistic about everything we're building to give billions of people the power to express themselves and connect with people they care about I'm also optimistic about delivering personal super intelligence to everyone we've always focused on putting power in people's hands this is how we believe progress is made in the world these values are what makes us different and they are why Meta has been successful but success isn't given AI is the most consequential

40:51technology of our lifetimes the companies that lead the way will define the generation we're transforming our company to make sure we'll always be the best place for talented people to have the greatest impact so I guess if you're still there and you're one of those people maybe you have some confidence I don't know but then the one thing I saw Mike that just sort of took off last week DD is the the Twitter handle he's a guy I think based in San Francisco tapped into the world I don't know if he's an engineer or not but plenty of AI people responded to this tweet and it was 13 million

41:23views as of this morning so I think this maybe captures those of us who don't live in the Silicon Valley bubble what is going on there because again it's a very different world than most of us experience so here's what he tweeted the vibes in San Francisco feel pretty frenetic right now the divide in outcomes is the worst I've ever seen over the last five years a group of about 10 000 people employees at anthropic open ai xai nvidia meta founders have hit retirement wealth of well above 20 million back of

41:56the envelope ai estimation he's saying everyone outside of that group feels like they can work their well-paying but less than 500 000 job for their whole life and never get there worse yet layoffs are in full swing many software engineers feel like their live skill is no longer useful the day-to-day role of most jobs has changed overnight with ai as a result number one the corporate ladder looks like the wrong building to climb everyone's trying to align with a new set of career paths should I be a founder

42:28is it too late to join anthropic or open ai should I get into ai what company stock will 10x people are demanding higher salaries and switching jobs more and more number two there's a deep malaise about work and its future why even work for all the peanuts quote unquote will my job even exist in a few years many feel helpless you hear the permanent underclass conversation a lot and that's in quotes and that actually got retweeted a lot that whole permanent underclass idea especially from young people it's

42:58hard to focus on doing good work when you think man if I joined anthropic two years ago I could retire number three the mid to late middle managers feel paralyzed many have families and don't feel like they have the energy or network to just quote unquote start a company they don't particularly have any ai skills they see the writing on the wall middle management is being hollowed out in many companies number four the rich are particular aren't particularly happy either no one is shedding tears for them and rightfully so but those who have quote made it experience a profound lack of purpose too some have gone from

43:34less than 150k a year to 50 million in a few years with no ramp it flips your life plans upside down for some comparison is the thief of joy for some they escaped to new york to quote live life for others still they start companies just cause often to win status points they never imagined that by age 30 they'd be set i once asked a post economic founder from someone who'd made it i guess friend why they didn't just sell the company and they said and do what right now everyone wants to talk to

44:05me if i sell i will only have money i understand that many reading this scoff at the champagne problems of the valley society is warped in this tech bubble what it is often well off anywhere else in the world is bang average here unlike many other places tenure intelligence and hard work can be loosely correlated with outcomes in the bay living through a societal transform transformative gold rush in that environment can be paralyzing am i in the right place should i move is there time still left am i gonna make it it's psychologically torments many who have moved here in search of

44:39success ironically a frequent side effect of this torment is to spin up the very products making everyone rich in hopes that you can vibe code your path to economic enlightenment so i just wanted to read the whole thing because i think it's one it's a view of the valley two i think there's elements of that that you can extract out the extravagance and the crazy salaries and you can actually start to see some similar ideas for the for the the non-silicon valley knowledge worker like yeah is this corporate ladder real like is this stable can i go get that job and stay at a company for 10 15 20 years um what

45:16the hell is the future of work like am i even prepared for it i'm a middle manager like what am i going to be doing in three years do they need middle management with all this ai stuff going on like so maybe not like hey i went from making 150 000 to having 50 million maybe that's not like relatable to people but i feel like everything else in here is somewhat relatable where we're all sort of in this like what is what is really going to happen in like three to five years what am i supposed to do what are my kids supposed to do just remove that extravagant um financial outcomes from the equation it it

45:48actually has a ring of truth to it yeah it's super interesting it's it really gives you also some perspective into just what different lives people are living yeah yeah and again if you're not out there and you don't have friends in those circles yes it doesn't even register like i i run in some circles with people like in you know back we spent a lot of time in boston there was a lot of people who made a lot of money and like early in their careers being at hubspot or places like that or i've spent a lot of time with people out in the valley and you just it's a different life but there

46:21is a lot of what he said is so true like if you know someone who's made it and they're in their 30s it's like whoa okay like i got all the money now but like what like no one cares what i have to say anymore yeah i don't know what i'm gonna do with my life so no one's gonna like everybody's got their own stuff going on no one is gonna like really feel bad for someone who's sitting on like 50 million dollars at age 32 but in the valley that is a real psychological issue yeah all right so next up

46:53somewhat related here this past week cloudflare ceo matthew prince published a wall street journal op-ed explaining after he had sent out an employee memo why he laid off more than 20 of his workforce earlier this month even as cloudflare posted more than 30 revenue growth and strong free cash flow prince actually said he had not found another example in u.s business history of a public company growing more than 30 percent laying off more than 20 of its workforce but he predicted what we did is

47:24likely going to become the norm over the next year so he references in this op-ed a framework from peter drucker the management consultant's 1954 book the practice of management and he goes into this and says it divides every business into three roles builders who create products sellers who sell them and measurers which prince defines as everyone else finance auditing legal compliance middle management operations etc ai prince argues is not coming for builders or sellers but it is coming for

47:55measurers he said that the vast majority of the people he laid off were measurers so cloudflare cut middle managers across the organization reduced its marketing team trimmed back office finance ops other quote unquote measurer roles and in his original email him and his co-founder said that you know ai is the reason that this is happening they are restructuring around a world where builders and sellers are still very very important but ai is kind of increasingly filling that measurer role um he actually said cloudflare's

48:30own internal ai usage had increased more than 600 percent in the previous three months alone now paul this kind of happens this was getting a lot of attention but then something more recent happens that is getting probably just as much or even more which is at the same time we had click ups founder and ceo announce a 22 percent headcount reduction despite the business being the strongest it's ever been because and this is very similar to cloudflare ai is fundamentally changing both what is possible in a business and how

49:02organizations need to be structured so this really does not sound like these ai washing claims that people have made in the past these are really healthy businesses pivoting proactively towards what they perceive to be the new normal of like how a business needs to operate in the agentic era yeah the the people who are stuck on this we're just solving for over hiring during the pandemic are going to need to find a new talking point because that is not what any of this is um so zeb evans is the click up one you mentioned

49:32so clip click up as a project management software company um i'll i'll just read a few excerpts from his because i again i this one blew up it was at five million two hours ago it's at six million now as i'm like looking at it before as we're reading this so he talked about this reduction of headcount by 22 but the business is as strong as ever and he's going to own this he said this isn't about cutting costs most savings from this change will flow directly back into the people who stay we'll be introducing million million dollar salary bands if you create outsize impact using ai you'll be paid

50:08outside of traditional bands he goes on to say i only see two options wait for this to play out gradually in the market or be honest about what i'm seeing and act proactively under a header the 100x organization he said the primary change is that we're restructuring on what i call a 100x org the goal is to 100x output the roles required to build at the highest levels are fundamentally different than they were a year ago he goes on to say the common narrative is that ai makes everyone more productive it doesn't many of the workflows of today if left on change create bottlenecks in ai systems

50:43the roles are going to evolve but waiting for that to happen naturally means falling behind in the 100x org he says that actually is heavily dependent on people infinitely more people than today this is only possible with 10x people that have embraced and adopted the new ways of working he then goes on to talk about what he calls these 10x engineers under a header the builders agent managers and the front liners so based on what i was saying last week about the architect the orchestrator and the apprentice i'm really intrigued by these like ways people are thinking about

51:14structure and everything so he said here's what we validated recently at click up the great engineers the ones who can orchestrate architect and review are becoming a 100x engineers they're not writing code they're directing agents that write code the skill is judgment um he goes on to say the new world is about enabling your 10x engineers to become 100x he talks about product management and design how they're kind of merging how designers that have customer focus become more like product managers and product managers that have intuition for user experience become more like designers because

51:48everyone can use ai to do everything he said everything outside of managing systems orchestrating ai and reviewing outputs becomes a bottleneck then he goes on to say ironically the people that automate their jobs with ai will always have a job they become owners of the ai systems or agent managers we have many examples of these people at click up he said the underlying systems in which we operate are absolutely critical to get right i think most companies are delusional to think they can iterate on existing systems and compete in this world uh he said that in a world that becomes saturated with ai

52:22with ai communication the human touch will matter more than anything to customers one-on-one meeting time with customers is something that shouldn't be automated the systems around the meeting should be so that frontliners spend 100 of their time with customers this is actually exactly how i think about our customer success team with ai academy it's like i'm going to automate absolutely every piece of the customer success success experience i can so the more human element is that they're literally spending every hour of the day talking to people and helping them in person

52:52on calls like that's what success to me should be is like the human to human interaction and so like totally like automate everything else you can that doesn't devalue the experience to the user um to the customer so and then he talks about this whole idea of like rewarding them with these crazy salary bands which sound crazy today but they're not like again i i think about my own um vision for smarter x it's like you know why can't you have one to two million in revenue per employee or even higher

53:23where traditionally it might be like three to four hundred thousand would be exceptional it's like why that's not exceptional anymore like and then but you take that the the profit gains from that their operating margin that comes with the increase in revenue per employee and you give that back to your employees you cap what your profit is like it's like hey we don't need to be more than 30 percent operating margin whatever that number is and beyond that we're giving this back to our people like and so that's i think the kind of companies where people are going to want to work that have this it is not profit at all costs we are going to run an extremely profitable company but you will all

53:58benefit and hopefully have an amazing culture in the process not live under the threat of i'm just using you to record how to do your job so i can replace you in six months um and then i saw i saw an interesting tweet as a follow-up retweeting zebs which was howard lehrman who is the founder of yext he said everyone is obsessed with ai making a 10x engineer a thousand x engineer the recent reductions at cloud fair and click up have me realize the plot is equally about the inverse ai amplifies the negative impacts of poor performers and this is really important if a person with poor taste

54:33who makes mediocre judgment calls and doesn't properly build things customers love is able to produce 10x more work does the company even want that hell no he said productivity isn't about as many people as possible token maxing ai is a double-edged sword especially when it's used to produce net new work if you give a bad artist a pen that can draw 100 times as fast you're going to pile up with a lot of junky artwork very quickly and since it happens so quickly leaders are now able to see who is picasso

55:06and who is not and adjust accordingly so i just thought that was kind of like a cool perspective of like why what i said last week on the show was if you don't become ai forward you have no job prospects in three years in any industry like whether you like ai don't like ai whatever your feelings are toward it i i i like i said then i totally empathize with people fearing the environmental impact thinking it's stealing creativity and you know our our ideas and like i get all of that and i'm a writer by trade my wife's a painter by trade like i understand it but i'm just telling you that like you won't have a

55:41job if you don't embrace it and learn how to use it responsibly and that's what this demonstrates it's like every company i talk to right now is looking for the ai forward professionals in their current staff and the ones who aren't won't be there in a year yeah and they will replace them with people who are who can do 10x the work of the people who wouldn't embrace ai that is again like i've there's a few times where i've stressed on this podcast through the years like what i think an inevitability is and i have a very high confidence level based on the conversations i'm having every week that that

56:15is what the future looks like across every industry you are either ai forward or you don't have a job and that goes back to what i said three years ago about what does every company look like it's ai native built smarter from the ground up it's ai emergent it's a legacy company that figures all this out and evolves or it's obsolete and the obsolescence doesn't happen in like two months six months 12 months necessarily but you wake up one day you're like oh our business is cooked like we have nothing now and i think that's how a lot of people's careers are going to be if they don't figure this out and like find a way to balance their own beliefs about ai but still

56:53leverage it to to be an impactful person within an organization and that might mean going and working at a company that's taking the most responsible approach possible to it you will not find a company three years from now that isn't doing ai because they will be out of business yeah and one just quick final note here especially in light of the meta layoffs being the previous topic i actually highlighted this in the cloudflare ceo's message to the company there was something really interesting and i realized like layoffs are really fraught and there's a lot of pr out there so take it with

57:25a grain of salt but he wrote at one point if we are asking our team to be world-class we have a reciprocal obligation to be world-class in how we treat them we are pairing the directness of these measures meaning the layoffs with severance packages that lead the industry the packages for departing employees will include the equivalent of their full base pay through the end of 2026 and then there's a bunch of other stuff in here only point being like if you are making these announcements while your business is doing really well this seems like the appropriate messaging to pair with it and

57:58reading his letter whether you agree with it or not compared to the letter we read from mark zuckerberg is night and day and really really interesting of like i think there are ways to maybe do this in a more human centered way and i think that along those lines mike of the the bio package and stuff uh jeff bezos made a lot of headlines this week saying that he thinks that we should be eliminating federal income tax for the bottom half of u.s earners that the people making

58:28all the money should just bear the burden of paying all the taxes and that he was actually going to lobby trump personally on this idea and so i i feel like we're entering a phase where you're going to start hearing a lot of ideas about how do we find a balance in society when the rich are going to get significantly richer yeah and and the people in the middle are going to stay in the middle or or really struggle um and so i just i feel like we're going to start every week like someone one prominent

59:00is going to have some new idea about like i think mark cuban a couple weeks ago was like taxing tokens i think it was his idea like yeah so you're going to hear all these theories it's part of the reason why we put andrew yang on stage at macon it's like i want to i want people who have ideas to like present those ideas to people they might not be the solution but like let's at least start thinking about all these different ways we can do this and i believe organizations have a responsibility to be doing what you're saying where you're trying especially if you're benefiting from that reduction you better

59:30be finding a some balance there yeah okay next up a gallup poll released this past month for the first time they have measured public opinion on ai data center construction and they found that 71 of americans oppose having a data center built in their local area with 48 strongly opposed only about a quarter favorite gallup set the data center opposition explicitly against something that's also pretty polarizing which is nuclear power so in this survey 53 of americans said they oppose a nuclear plant

1:00:06being built in their area so less than the data center wow um the high point for nuclear opposition since gallup began tracking it in 2001 was apparently 63 so data centers are already more pop unpopular than nuclear has ever been at its worst um among the opponents environmental concerns are really the big concern so half of them cite excessive resources with water and energy each named by 18 as a concern

1:00:36another 16 mentioned pollution there's quality of life concerns like traffic economic concerns like higher utility bills and those account for kind of the remainder of the big concerns here um among the people that do support them two-thirds say they have economic benefits 55 specifically named jobs and interestingly and we'll talk about this in the next segment too what how this gets complicated this crosses every major democratic group demographic group but democrats are much more strongly opposed

1:01:07than republicans 56 to 39 women more strongly opposed than men 55 to 43 gallup's conclusion here is that overcoming local opposition is now a major hurdle in the expansion of ai computing and ai infrastructure is on track to become an important campaign issue in state and local elections this year so paul i i can't say this is like surprising to me based on the anecdotal stuff we've been talking about but it's really interesting to see hard data on this and it's something we're just going to keep digging into

1:01:39because last week when we talked about the um eric schmidt getting booed at these commencement but he wasn't alone there's multiple instances of commencement speakers getting booed talking about ai a lot of the feedback i got and including some people you know professors in college said that they don't really hear as much about the kids complaining about the impact on jobs it's it's almost always environmental and like data centers in particular is a hot button issue i found that intriguing to where like you and i mike were just like okay we got to start digging more and more into what's going on

1:02:09here in cleveland we're seeing this there's an effort right now in our hometown um to put data center in particular near the airport i think is the one that's getting some issues um and there's some public backlash to that i've seen some things for in ohio we have a lot of farmland in ohio it's pretty yeah if you've never been to ohio it's pretty flat and there's a lot of farmland um and so there's a lot of spots to drop data centers there's water sources and you're seeing this this public backlash um so yeah it's it's a challenging issue the nuclear one is wild to me right man um but so you know again

1:02:48like i don't want to get into conspiracy theory stuff on this podcast we always try and like keep this pretty straightforward in terms of how we talk about these things i would just note that there's a couple of things at play here so one is you have to find the wedges and political campaigns that move people if you see data that 71 of people don't like data centers it will 100 become a major part of a political campaign the other thing you have to be aware of is the geopolitics of foreign actors influencing public perception in adversarial countries the u.s does this to other countries

1:03:25and other countries do it to us if you see an issue that causes division within the public or if you see an opportunity to slow down ai progress in the united states the number one thing you could do would be stop data centers getting built and slow down energy infrastructure so if foreign actors see that the u.s population is starting to divide over a topic and that same topic may actually have a byproduct of slowing down u.s progress then you may start to see things on

1:03:58social media from accounts that aren't created in the united states pushing negative connotations of a thing that is not to say that it doesn't have negative impacts on the environment it's not to say that concerns aren't real it's to say that when an opportunity emerges foreign actors use that opportunity and so like i'll just kind of leave it at that for right now but this is one that will that will have multiple layers of complexity including very real concerns about these things

1:04:29being built in local communities yeah it's really funny paul i say this like mostly jokingly but i couldn't help but think of like an analogy to i know you like this as well the three body problem where it's like you know like the aliens try to kill our science as a species to hold us back while they try to get ahead essentially i'm like well that's kind of could be happening here too i don't know yeah i mean if you like conspiracy theories you're welcome like go ahead and start searching for the gist of what i was just explaining and i can almost guarantee you there's all kinds of

1:05:02information i haven't gone down the rabbit hole no i'm just telling you how i know this stuff works having studied misinformation campaigns in the past and um what foreign actors will do so yeah it's just something to keep an eye on and something to maybe educate people in your family about if they're not aware of how social media works and one final note here paul you had mentioned the commencement speeches i actually pulled some like deep research we had talked about kind of looking into you know there are these people booing at commencement speeches that we talked about last week and we

1:05:35were kind of curious with the emails we were getting people saying it was because students don't like the environment uh environmental impact of ai um but we wanted to look into what was going on so i ran a few of these and just again take everything with a grain of salt but i have vetted these stats against the the credible sources we found and you know i kind of teed it up for deep research said like hey here was the segment here's what we're trying to look into we're just curious especially younger americans where is the animosity coming from towards ai and just give me like gallup polls pew research

1:06:09harvard quinniapec polls stuff like that so i'm just going to read a few quick stats so interestingly according to the polling the biggest thing is still jobs because 59 of 18 to 29 year olds view ai as a threat to their future job prospects that's from the harvard kennedy school institute of politics they did a fall 2025 poll 81 of gen z believe advancements in ai will lead to a decrease in overall job opportunities uh quinnipiac university did a poll on that in march uh just this past march that's

1:06:45actually maybe they're all answering our state of ai for business report yeah 71 yeah um environment was still a big thing but in it kind of stack ranked all the factors and was like four out of five interestingly and number two was actually uh what what deep research categorized as cognitive atrophy skill erosion loss of meaning and work 80 of gen z believe that relying on ai to complete tasks faster will make future learning more difficult that's from gallup in april uh third was distrust of ai

1:07:16outputs and critically i think the institutions deploying them this one's pretty wild this is not just young people but only five percent of americans feel that ai development is being led by people or organizations that actually represent their interests which is quite cynical um one final note here according to edelman 70 of us respondents again not just young people believe business leaders it's more likely that business leaders are not being fully honest with employees about ai related job

1:07:47cuts that's part of that kind of institutional mistrust and then of course there is the environmental um depletion a 72 percent of americans express concern about the environmental impact of ai so not to say it's not a huge issue but there's more going on here too it sounds like with younger generations yeah one note i'll add to this mike i actually um there was a lady on twitter ashley mayer um and she had tweeted about the issue with eric schmidt and i thought it was it was a pretty good observation she said i'm seeing people trying to wordsmith eric schmidt's commencement address

1:08:22and sure there was a lot of room for improvement but my hot take is that there is no way that he or any leader in tech could deliver a broadly inspiring pitch for ai in this moment these speeches are not happening in a vacuum and anyone who is seen as an architect of this technology revolution has too much to gain to be a credible messenger and so i actually replied that this is a major challenge for the ai labs that they are the ones building it so it's really hard for them to be trusted to say

1:08:52the the positive things if people have this negative perception and no matter what you say they're just not going to trust you and that's why like neutral voices are so important it's part of why we do our very very best on this show to be as objective as humanly possible and tell the good and the bad because my belief is a lot of these ai leaders are going to be viewed as the villains and

1:09:16it's hard to overcome that when you are literally building it and admitting it's going to reach the singularity and like but you don't have a plan for what that means like it it's a slippery slope for the ai labs right now and the pr and comms is going to be very very challenging yeah no kidding all right next up kind of related just on the social and political level there's two stories this past week about kind of how fractured a the ai political debate has become um so first there's

1:09:46the new ai industry pack leading the future which we've talked about before has actually poured more than a million dollars into the democratic primary to defeat state assemblyman alex boris it's a lawmaker who championed new york's recent ai safety law this pack is backed by andreessen horowitz open ai president greg brockman and palantir co-founder joe lonsdale but interestingly enough boris has also raised more than 364 000 from employees at top ai labs and 420 000 from workers at ai safety organizations

1:10:22anthropic employees led the corporate giving with like 168 000 but even google open ai palantir microsoft and meta employees also chipped in now second we had a bunch of stuff come out this past week about the trump white house preparing an executive order that would establish a voluntary pre-release review process for frontier models that basically they were proposing under this framework labs would share new models with the government at least 90 days before public release and grant

1:10:52access to certain critical infrastructure providers um it looks like though paul right before we kind of went on air that last minute lobbying by the tech industry may have killed this fully so i kind of just want to get your sense of like what's going on here we've got literally i feel like people in ai are both divided but also we don't know where this is not a the us or them issue there's people on all sides that are lobbying for and against and everywhere in between yeah i was kind of shocked that the executive order was going to happen because um there is so much backlash even within the republican

1:11:27party to this and so on wednesday it gets leaked that they've invited like a dozen or more executives including amaday and altman and musk and all these people for the next day to the signing of this executive order yeah and i was like one that's a really weird list there's no way all those people are showing up on like 24 hours notice yeah um but then uh yeah it comes out that david sachs calls trump because he gets read into the executive order earlier this week and calls him direct and he's like you can't do this and it's basically going to destroy our efforts to take on china

1:12:00and whatever else and then there was a couple articles because it's all happening so fast it's like oh elon musk also called well musk tweeted he's like no i didn't like i didn't even know about it like i still don't know what's in the executive order i didn't call him and like tell him not to do it so yeah it's like some of the labs want like more governance some don't some republicans want more governance some don't some democrat it's so wild and then the only executive order that actually happened this week ends up being governor gavin newsom in california yeah so his is

1:12:30about um to explore an overhaul of labor policies to deal with potential mass job displacement from ai so wild and like it doesn't i don't know the executive order doesn't get i mean it's it's washington this shit could change by the time we finish recording this but like it sounded like sacks talking points are the same talking points he's always had around this i don't know why it would make a difference like morning of or night before that you got a call and said the same thing and now all of a sudden it like means something different i don't know the whole thing seems like a

1:13:02cluster i have no idea what they're doing um but i think that that that executive order anticipation of it is what was holding up potentially holding up the 3.5 pro from google and sam altman has said publicly now that they have another model internally at open ai that is doing crazy things and so there's rumor it's 5.6 like gpd 5.6 um but that that also is being held up by what was going to be an executive or around voluntary vetting of these models so who knows yeah we'll we'll see how it all plays out

1:13:36but i want i do just speculate and wonder how much of the societal or narrative backlash is starting to affect the administration because there are cracks appearing in the facade because previously it was just go go go yep yeah they're gonna have to do something especially yeah if the data centers and i don't i mean i don't i don't know i don't even write they would even say this um

1:14:00there's a lot of like rural areas where data centers would go yeah and so you politically you have to like understand the demographics of where these would be built and what voting base that is yep and so yes i would imagine there might be polls that are coming out that are saying whoa we gotta be really careful here that would push for regulations yeah and then there's where all the money's coming from that drives the campaigns uh that don't want the regular i don't know man

1:14:33it's so bizarre it's like so many levels of complexity to this one too well buckle up because you know speaking of complexity in politics here we also have kind of a resurgence or at least the continuation of this anthropic versus the department of war debacle it's been happening so on may 19th the dc circuit court of appeals held heard nearly two hours of oral arguments in anthropics appeal against the department of war which we have talked about at length um the may 19th panel of three circuit judges

1:15:04appeared divided on what they heard so one of the judges was the most pointed against the department of war they called this designation of a supply chain risk spectacular overreach uh this judge said they saw no evidence the department of war had supported its claim that anthropic is a bad actor um the other two judges were more skeptical of anthropic one of them noted that this is the first time the underlying 2018 statute is being interpreted by a court and stressing that this panel needs to be clear about

1:15:35what is even reviewable by them so the court essentially ends up issuing a written opinion here um and once until that happens this is kind of still in limbo and i'm curious paul like if anthropic like loses this appeal like they're they're they're still in the doghouse right they're still designated a supply chain risk i don't know mike all i can picture is that the executive order actually was signed on thursday and dario amade who was invited to be there was there in the in the oval office right while the government is considering them a supply chain risk right right the optics of

1:16:09that would have been amazing yeah um i have no idea man i i like again the way the administration functions is you can be their their their villain and their hero like in the same day like we've seen it with elon musk like they they just like hated each other and they're trashing each other and like the worst things you could say and then like a week later it's like sitting at a game together i don't know right it just it literally is just which way the wind blows i'm sure they're talking

1:16:41to anthropic daily across all like nsa like all these other departments are in conversation with them because of mythos and where the max models are going and and things like that and and then like simultaneously they're like oh yeah yeah sorry about that lawsuit but like we'll just keep going we'll figure that out like blah let's do deals in the meantime and i don't know it's just so bizarre it's it's kind of like silicon valley without living there is really hard to process the that that's reality yeah i feel that way about washington all the time i'm like unless you're there and even when

1:17:12you're there it probably doesn't feel like reality a lot of times to people on both sides of the aisle it's probably just totally every day you go through you're like this is like not real what we do every day right it's like more of comedy or drama than anything you see on tv yeah it's like yeah it's like yeah like tv shows and you're just like every day you think you shake your head like i can't even believe that that was my day today yeah yeah okay so next up we have our ai use case spotlight so every week we're giving you a quick look under the hood at the real ai use cases we are exploring

1:17:44building or deploying in our work at smarterx so paul i'm going to share a quick one and i think you've got one too to share and we can talk through that sounds good so for me this one is not actually like a single tool and use case um this is more like around workflows and like methodology that i've kind of started to experiment with um and rebuild kind of my entire work day around what ai enables now i'm not remotely close to the final vision of this but i wanted to kind of share it in case it's helpful so you know i kind of like sit down at my desk every morning and start thinking about

1:18:16what should ai be doing for me or could be doing for me while i'm working on kind of my most one most important thing for the day because like i personally think even in the age of ai like single tasking is more important than ever in certain contexts because there's things i like need to put my full attention on with like sustained maximum cognition and deep work and not be interrupted and even if i'm using ai to do that it's got to be like one thing at a time i think for me to really create as much value as possible so it's really hard to just say let's fire a swarm of

1:18:50agents and go do your work while they work because then you're getting pinged every like three or five or ten or fifteen minutes your context switching you have to switch complete completely how you're thinking about problems there's no way to do deep work this way so i've been trying to figure out like what is the balance there so i kind of just started experimenting with how to restructure to draw some type of balance and where i've kind of landed on is like every day my work day starts by sorting work into two piles pile one is everything that can be like essentially handed to an agent and i'll

1:19:25come back to that in a second i fire that off in the background first close that window away like put it out of sight pile two is like my sequential list of like here's the most important thing i need to focus on and complete then the next thing then the next thing go do that with your full focus while these agents run um the agents then come back with the results and i review those when i'm ready to take a break from you know my single tasking and all of that but the key here is like you have to be essentially iterating constantly because the agents can only really go do stuff they've been taught to

1:19:58do if there's any requirement for review refining going back and forth i need to actually stop everything and build a skill for clawed code for instance to go do that thing before this can work at all so i've kind of got this almost like iterative cycle it's like hey do we have skills for it great go do those things while i work on this important thing if we don't have skills for it i got to drop everything build one and so on and so forth so i kind of try to do this every day it's working reasonably well but it's definitely like a weird thing to get used to i would say now is this

1:20:31primarily in co-work or where what are you yeah i'd say claw like a range of different things but mostly like either clawed the web app or clawed code just to sometimes build skills faster but yeah um but yeah i'd say combo those two but you could you could transpose this to basically any type of workflow i think yeah that's cool yeah um yeah my i would just i'll give a brief one so i've been working on a very high level longer term project i'll say that deals with finance legal operations uh accounting

1:21:07all sort of wrapped into one desired outcome i guess i'll just kind of leave it at that um and so it requires me to quickly learn advanced topics and then be able to have very educated discussions with my accountants and my attorneys and so i will have a meeting i know we're going to talk about like five topics that i am not an expert in and so i will prep for that it's like what questions do i need to ask i will understand topics that we're going to cover in real time in the meetings i will have a

1:21:40thread open and it's like okay i don't understand this concept explain this to me in more simple terms or because it exists within a project within claude or chat gpt and this specific example using claude and so i have a project open it's trained on all the knowledge base it knows all the previous chats around it so it has all the context it needs and it's able to help me in real time like analyze things and ask the right questions so i would just say the you know the dominant use i often have for ai is as a thought partner and advisor and this is like hardcore relying on it heavily in that role

1:22:14and then follow on documents and analysis of the meeting notes things like that where i'm just like over the next three months it'll be my sidekick for everything i do with this project and so the continual use of the memory the context windows the knowledge base becomes essential for me to to execute the thing i'm working on yeah that's so cool i feel like people underrate how valuable that context is over time it's not just about getting like an answer or support support on a decision

1:22:45in the moment it's about a month from now you have this built up context that helps you make a better decision yeah and peace of mind and there's artifacts that need to get created along the way so like you know as a result of meaning i have to now go go do things and i'm not it's not something i'm involving the team with i mean some of the teams knows what i'm working on that they will know what i'm talking about as i'm talking about it but like there's pieces of things that have to get created that i have to be the one to to design and and bring to the finish line and so me and claude are doing that together like whenever i need to create something an interactive html file a deck things like

1:23:19that it's all just me and claude pretty much that's awesome yeah all right paul that's our final segment ai product and funding updates i'm going to run through these very quickly as we wrap up here so first up open ai is preparing as we talked about to confidentially file for an ipo as soon as this week uh targeting potentially a september public day view at evaluation that could exceed one trillion dollars open ai also said an internal general purpose reasoning model disproved an 80

1:23:50year old paul erdos conjecture in discrete geometries are very very famous math problems and basically this proof was validated by fields medalist tim gowers and other external mathematicians so tldr is that ai was capable of solving some of the world's hardest math problems now open ai also launched something called guaranteed capacity a new enterprise offering that lets customers lock in reserved open ai compute through one two or three year commitments in exchange for discounts that increase with annual spend

1:24:24available on a first-come first-served basis until the reserve capacity sells out uh open ai has also announced it's become a c2pa conforming generator and is adding google deepmind synth id invisible watermark to images generated through chat gpt codecs and the open ai api alongside a public preview verification tool that lets users check whether an uploaded image carries provenance signals anthropic disclosed a roughly 45 billion dollar compute deal with spacex like we talked about

1:24:56and spacex publicly filed its s1 with the sec for what would be the largest ipo in history right now um they're looking to raise up to 75 billion at a roughly 1.75 trillion valuation with a roadshow scheduled for june 4th pricing on june 11th and trading as early as june 12th the ai risk assessment nonprofit meter me metr published its first frontier risk report based on a february to march pilot exercise inside anthropic google meta and open ai concluding that the lab's most capable internal ai agents

1:25:31could plausibly initiate small unauthorized deployments deceive monitoring humans and bypass security measures no big deal no big deal at all that alone could be a whole episode i would imagine

1:25:46uh cloudflare in addition to the other news we talked about published a defensive evaluation of anthropics unreleased mythos preview model this is under the project glasswing access program we've talked about they find the model could chain previously low severity bugs across more than 50 internal cloudflare code bases into single high severity exploits and write working proof of concept code uh cloudflare warned that the model's organic refusals were not consistent enough to serve as a complete safety boundary again no big deal how does say how does cyber security experts sleep at night

1:26:21maybe they don't i get asked that sometimes with ai but it's like man i look at that world and i just like i mean you just have to accept like something is always going to go wrong i guess yes and there's like a thousand things going wrong that you don't even know about yet yeah i wonder if you're a cyber security person chime in i'd love to know and also how we can help because it's not a it's a stressful time man nvidia reported some earnings they reported revenue of eight eighty one point six billion for q1 fy 27 up 85 percent over year over year with data center revenue of 75.2 billion

1:26:56up 92 percent uh they raised actually their quarterly dividend from one cent to 25 cents per share it's a 25-fold increase and finally mckinsey announced something called the acorn plan an overhaul of partner compensation that reduces the cash portion of partner profit shares uh to from 95 to 90 percent and shifts more compensation into equity and the important part is this is framed by the firm as a response to ai and outcome-based pricing making consulting revenues more volatile both points that

1:27:32we have mentioned time and time again how would you like to be an nvidia executive announce that your revenue is up 85 percent year over year when you're already on a historical like stretch and your stock goes down one percent the next day yeah that's like that's just offensive i mean if you're one of the i'd just buy more stock i guess yeah i mean all they care about is like well yeah but three years from now is the demand going to be so high it's like geez oh man you just can't win i mean can't no but i don't again i don't feel sorry for anyone in nvidia who have been minted hundred dollar 100 millionaires

1:28:06like 100 million or more right um because there's a whole bunch of multi-millionaires in that nvidia staff but tough go when you're putting up those numbers and people are like yeah yeah man better luck next time yeah and one final announcement here paul our ai pulse survey like we talked about go to smarterx.ai forward slash pulse and this week we're going to be asking about a little bit about your own search behavior given google's changes to search and also we're asking about how you think

1:28:38about your employer's honesty about how ai will affect jobs at your company we'd love to hear your input takes a couple minutes to fill out maybe 30 seconds honestly so if you go to smarterx.ai forward slash pulse we would love to hear from you paul we we you know made it to the end of another week i don't feel as depressed as i did after the last 100 this is a lot better it was a little bit more balanced yeah i'm feeling a little bit more upbeat about ai today so hopefully you are all listening are also a little more positive today all right thanks mike thanks paul thanks for listening

1:29:13to the artificial intelligence show visit smarterx.ai to continue on your ai learning journey and join more than 100 000 professionals and business leaders who have subscribed to our weekly newsletters downloaded ai blueprints attended virtual and in-person events taken online ai courses and earned professional certificates from our ai academy and engaged in the smarterx slack community until next time stay curious and explore ai