The Pope issued his first encyclical: 43,000 words on AI, humanity, and what happens when technology concentrates power in too few hands. Paul and Mike break down Magnifica Humanitas, why Chris Olah showed up at the Vatican, and what it means that the Catholic Church has become a dominant voice in the AI debate. They also dig into AI's escalating public relations emergency, the token cost crisis hitting corporate America, Claude Opus 4.8, the competing narratives on AI and jobs, Illinois's landmark safety bill, Microsoft's Work Trend Index on agents, and a batch of product and funding news, including Anthropic's $65B Series H.
Listen or watch below—and see below for show notes and the transcript.
Each week on The Artificial Intelligence Show with Paul Roetzer and Mike Kaput, we ask our audience questions about the hottest topics in AI via our weekly AI Pulse, a survey consisting of just a few questions to help us learn more about our audience and their perspectives on AI.
If you contribute, your input will be used to fuel one-of-a-kind research into AI that helps knowledge workers everywhere move their companies and careers forward.
Click here to take this week's AI Pulse.
00:00:00 — Intro
00:02:49 — AI-Pulse Survey
00:05:40 — The Pope's AI Encyclical
00:25:35 — AI's PR Emergency
00:38:25 — The Soaring Cost of Intelligence
00:54:12 — Claude Opus 4.8
00:58:23 — The Narrative Around AI and Jobs
01:06:22 — AI and Politics Updates
01:11:04 — Microsoft's Work Trend Index on Agents
01:14:23 — AI Use Case Spotlight
01:19:34 — AI Product and Funding Updates
This episode is brought to you by AI Academy by SmarterX.
AI Academy is your gateway to personalized AI learning for professionals and teams. Discover our new on-demand courses, live classes, certifications, and a smarter way to master AI. Learn more here.
This episode is also brought to you by MAICON, our 6th annual Marketing AI Conference, happening in Cleveland, Oct. 13-15. The code POD100 saves $100 on all pass types.
For more information on MAICON and to register for this year’s conference, visit www.MAICON.ai.
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: Go live within one of these companies outside of the Silicon Valley bubble, and then talk to 'em about the realities of their business, the margins that they operate on, the lack of growth that they see for the company being in single digits, if they're lucky. And then tell me that AI doesn't completely change the future of work in the structure of teams.
[00:00:21] Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Roetzer. 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.
[00:00:41] 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 AI literacy for all.
[00:00:57] Welcome to episode two 17 of the Artificial [00:01:00] Intelligence Show. I'm your host, Paul Roetzer, along with my co-host Mike, put. We are recording on Monday, June one, 9:30 AM I cannot believe it is June already. Nope. Wild. My kids are officially home from school for the summer, so luckily they're teenagers and they're not up and moving at nine 30 in the morning by, so I am home recording in the home studio.
[00:01:25] All right. So, I don't know, I guess there was like big things last week, like not, not model, although we did get a new model, I guess, actually. Yeah, probably two new models last week. But it's so funny, like model releases are just becoming like secondary notes because they're happening so frequently and they're all the decimal point releases.
[00:01:45] We're not getting like the major releases. So we will cover all of that, all the product and funding news. but the big thing last week was, you know, really, the encyclical from the Pope and so we're gonna talk about that and the far reaching. [00:02:00] Potential impacts of that. So we'll lead off with that.
[00:02:02] But first, today's episode is brought to you by AI Academy by Smart Rx, which helps individuals and businesses accelerate their AI literacy and transformation through personalized learning journeys and an AI powered learning platform. New content is added weekly, so you always stay up to date with the latest AI trends and technologies.
[00:02:21] The AI four Industries collection features eight core series and certificates designed to jumpstart AI understanding and adoption. We, as I said, we have eight of them so far. professional services, healthcare software and tech insurance, financial services, retail, and CPG manufacturing. And as of last Friday, AI for education, which we're really excited about.
[00:02:45] That's what we've been working on for a while. so we're planning to do a lot more in the AI for education space, but this is an incredible starting point for teachers, administrators, anyone who really wants to understand what's going on within education and how to use AI in a [00:03:00] responsible way and teach it in a responsible way.
[00:03:02] So, that's a really important one to get out into the world. These series in our ideal launchpad, for organizations that wanna level up their teams and accelerate AI adoption and impact individual and business account plans are available. There's also an education discount, right, Mike? I don't,
[00:03:18] Mike Kaput: yes.
[00:03:19] and nonprofit discount that you should definitely reach out to us about.
[00:03:22] Paul Roetzer: Yep. So if you work at a school or in a nonprofit, definitely talk with our team. There are special prices available for, those institutions. You can also buy single course series, so if you just wanted, say the AI for education, you can actually just get just that series versus the annual membership.
[00:03:40] So you can visit academy dot SmarterX.ai to learn more and use Code Pod 100 for $100 off of individual plans only. So again, that's POD 100 off of the individual plans. The business account plans have discounted pricing based on the number of licenses, and then as I mentioned, additional discounts available for [00:04:00] education and nonprofit.
[00:04:01] Okay, every episode we start off with our AI pulse survey. We ask a couple of questions at the end of each episode, and then we recap those, results to start the next episode. So these are informal polls. They're asked just of our audience, not enough responses that we would like publish this as like hard findings on anything, but it's meant to give some guidance on how our audience is thinking about things.
[00:04:23] So the first question. From episode two 16, I guess this would've been Mike.
[00:04:28] Mike Kaput: Yep.
[00:04:29] Paul Roetzer: now that AI mode is the default in Google search. Has it changed how often you click through to websites? 43% say much less often. So they're just going to AI mode, they're looking at the results and I mean, Google does a nice job of integrating links right into it.
[00:04:45] Now they've gotten a lot, lot better. They do. But you know, I think unless you're verifying facts or trying to like confirm something specific, like medical advice, I'm like, you often get what you need with just looking at the high mode results.
[00:04:59] Mike Kaput: Yeah.
[00:04:59] Paul Roetzer: [00:05:00] 29% said somewhat less, 25% about the same. And then there's a small sliver that haven't used AI mode yet.
[00:05:09] The next was, do you trust your employer, to be honest with you, about how AI will affect jobs at your company? 22% say yes fully. That's interesting. 35% say mostly 29% say not really, and 14% say not at all. Okay. Yeah, yeah. Alright, so, Mike, lots to talk about with the Pope's ai encyclical. Let's get started there.
[00:05:40] Mike Kaput: Sounds great, Paul. Yeah. So this past week, Pope Leo the xiv, the first American pope, published his first encyclical called Magnifica, Humanitas, or Magnificent Humanity. This is a 40,000 plus Word document that is built basically entirely around artificial intelligence and how to preserve the human person in the [00:06:00] age of ai.
[00:06:00] So the timing of this is actually quite significant. So Leo dated and signed it to the hundred 35th Anniversary of Rerum Novarum which is Pop Leo the 13th, 1891 Encyclical. On workers' rights and a fair wage. This is basically considered the founding text of modern Catholic social teaching. It was written in direct response to the Industrial Revolution.
[00:06:26] So Leo the 13th is actually Leo the Fourteenth's namesake. He took that name intentionally and like some other encyclicals before this one. He has dated the document to that anniversary of that one, and basically it's a pretty direct comparison here, arguing that we are in another industrial revolution that demands the church speak.
[00:06:46] So his central thesis laid out in this document is that technology is never neutral. As he writes, a quote takes on the characteristics of those who devise finance, regulate, and use it. He openly grants that [00:07:00] AI can be a valuable tool. But also warns that it quotes, tends to amplify the power of those who already possess economic resources, expertise, and access to data without oversight.
[00:07:11] He writes, quote, those who control AI will impose their own moral vision, which will become the invisible infrastructure of these systems. Adding that quote, a more moral AI is not enough if that morality is determined by a few. there's a line getting a few headlines in here. He uses this language about his, call to quote disarm ai.
[00:07:32] That word has been kind of widely read as being about autonomous weapons. but if you kind look at the document itself, the primary meaning can be a bit broader here. So when he says disarm, he's talking about an economic and cognitive arms race, not just a military one. So he says in here. Quote, disarming AI means freeing it from the mentality of quote armed competition, which today is not limited simply to the military context, but is also an economic and [00:08:00] cognitive phenomenon.
[00:08:01] This entails a race forever, more powerful algorithms and larger data sets driven by the desire. To secure geopolitical or commercial dominance. He is also careful to add here that to disarm quote does not mean rejecting technology, but preventing it from dominating humanity. And that it means, quote, freeing technology from monopolistic control and opening it to discussion and debate, therefore making it human friendly.
[00:08:27] you know, one other final note here, Paul, which you'll probably discuss is in a kind of unusual move, the Vatican unveiled this encyclical alongside Chris Olah, a co co-founder of Anthropic who actually spoke at the presentation of it. And so Paul, obviously this is not a document that gets enforced really in any way, but pretty big deal from a symbolic perspective, especially with, you know, we're gonna talk a little bit more about rising negative AI sentiment.
[00:08:54] There's obviously a billion plus Catholics in the world that all take the Pope pretty seriously, so How are [00:09:00] you looking at this? and the implications of it at the moment?
[00:09:04] Paul Roetzer: The reach and influence you just mentioned. That is the primary reason we're leading off today. As you mentioned, there's over 1.3 billion Catholics.
[00:09:13] The Pope's words, his thoughts are integrated into Sunday sermons, parish discussions, Catholic education networks. so when he's talking about things like misinformation, worker displacement, transhumanism, it directly shapes the behavior, beliefs, ethics of this global group. but in the encyclical, he actually said it's not just Catholic, it's all Christians and all men and women of goodwill.
[00:09:37] So by doing this, he's broadening the reach. It's published in at least 10 man major languages. And when you look at the whole, population of Christians in the world, now we're talking about like 2.7 billion people. It's about 30 some percent of the world's population. So what he says matters, and it does start to shape the way that people believe, what they believe about ai, how they [00:10:00] feel about ai, how it's taught within school systems.
[00:10:02] Pope also plays a role in international diplomacy. So this extends into global politics, actively participates in international diplomacy, utilizing, unique status to advocate for poor migrants, victims of conflict. In the letter, the Pope explicitly calls an international organizations such as United Nations to reform and act as essential instruments for promoting peace.
[00:10:26] Disarmament and shared regulations on digital technologies. And then, just overall the global civil society. So the church operates a vast network of educational, charitable and social institutions. And the Pope uses his influence to call for renewed educational alliance between families, schools, and public institutions to teach digital sobriety.
[00:10:47] So it's not just about a I t's much larger about technology overall, and then it matters from an economic and labor standards perspective. So historically, papal and cyclicals, as you mentioned, Mike have played a role in inspiring [00:11:00] labor unions, improving labor, legislation, and defending workers' rights.
[00:11:05] So the Pope is exercising this historical influence to demand that the business world tech developers and political leaders collaborate to protect jobs, ensure fair wages, and prevent AI from de-skilling or displacing the human workforce. So there's just like. the impact is significant. The reach is significant.
[00:11:22] I actually use Notebook LM to do like highlights of this because this is 43,000 words. Just for perspective, like when Mike and I wrote Marketing, artificial Intelligence, our last book, it's 50,000 words and that's when I did the audio for that. It was about six to six and a half hours. So like we're talking about like five hours if you were gonna read this thing at one X speed.
[00:11:43] Mike Kaput: Yeah. Yeah. And Paul, I would, I would also say if you start reading, I did not read every single word unfortunately, but it's also not exactly always the most accessible document. No. Which will come as no surprise to anyone that knows the deep intellectual and a Latin focused history of the Catholic [00:12:00] Church, I would say.
[00:12:00] So
[00:12:01] Paul Roetzer: it's like reading a legal document in many ways. It's
[00:12:03] Mike Kaput: like reading a legal document. Yeah.
[00:12:05] Paul Roetzer: Yeah, so just a, again, a great use for notebook lm. You can literally just grab the link, drop it in there, and then talk to it about it. Like, you know, what are the things related to X that you're interested in, whatever.
[00:12:15] So I just said like, gimme the 10 highlights that are most interesting to call out for our podcast audience. And so quick highlights the illusion of AI empathy. They talk about that, that it's really simulating empathy and care, which we've talked about many times on the show. AI is never morally neutral, as you referenced Mike already.
[00:12:31] There's biases built into these things based on the people who are building and training them. The hidden environmental cost, which we've touched on recently, is one of the issues that people is driving societal backlash to ai. The invisible human labor behind ai. So Karen Hao, who's keynoting at Mayon this year, her book, empire of AI, talks a lot about this.
[00:12:52] Like, who are the people that are being exploited to train these models that is not talked about publicly for a reason, like the labs don't [00:13:00] want you really understanding how these things are really made. So that's important. The rise of data colonialism, the threat of algorithmic injustice, concentration of digital power, which you referenced de-skilling the human workforce, the dangers of autonomous weapons and the impact on education and critical thought.
[00:13:19] So these are just like some of the big issues that are covered within the letter. I thought it would be interesting, Mike, just to highlight a few of the, posts that the Pope has published on X or Twitter, if you're not familiar. because it, it centralizes. The things that I think the Catholic Church is finding to be the most important to highlight, for a broader society.
[00:13:41] So I'll just read three of these, quickly. So one was, he said, in the era of artificial intelligence, when human dignity is threatened by new forms of dehumanization, ours is the pressing duty to remain profoundly human. We must lovingly safeguard the grandeur of humanity bestowed upon us and revealed in [00:14:00] its fullness, in Christ the splendor of which no machine can ever replace.
[00:14:04] There was another one that said AI can be a valuable tool, and at the same time, it calls for a measured and vigilant approach. The speed and simplicity with which practical assistance can be accessed undoubtedly makes life easier. Yet there can also, encourage excessive reliance and search for ready-made answers and weakened personal creativity and judgment.
[00:14:24] And I think Mike, some of these that I'm highlighting are intended to go beyond, religion. A religious belief like this is more about humans and humanity and what we are willing to give over to the machines and the impact it has. And so I think some of the messaging, regardless of your own personal beliefs or religion, are relevant just as humans.
[00:14:45] And so it's, you know, it's good to just kind of step back and think, and I'll highlight a couple of Chris Olas comments because I think it gets more to that spirit as well. the last one I'll note here is artificial intelligence. Do not go, do not undergo experiences. And this one's important. Do [00:15:00] not possess a body.
[00:15:01] Do not feel joy or pain. Do not mature through relationships and do not know from within what love, work, friendship, or responsibility mean, nor do they have a moral conscience. Since they do not judge good and evil, grasp the ultimate meeting of situations or bear responsibility for consequences. They may imitate or even simulate, but they do not understand what they produce for.
[00:15:24] They lack the effective relational and spiritual perspective through which human beings grow in wisdom. Now that one. I feel like is the one that you could spend the most time unpacking from a larger societal perspective. There's an element, obviously, of spiritual and religion to this, but there are much bigger foundational questions to discuss around is that true?
[00:15:46] Like, are the things he's saying do, do we agree universally with those things? And to bring that point to bear, Yann LeCunn, who we've discussed many times, used to head up AI at Meta and now has his own startup, he replied to [00:16:00] the Pope and he said, indeed, AI today does not, does not do or possess any of those things, but at some point in the future they will.
[00:16:09] Except perhaps for the spiritual part, many humans aren't spiritu spiritual either yet, have empathy and are highly moral. H And then Andrew Kiran, who we've cited numerous times here, has a lot of great posts about ai. X, he replied With your holiness. I received this in the spirit of human dignity and love of humanity, in which it is clearly offered.
[00:16:30] But some of the things you name embodiment, maturation, relationship comprehension, responsibility will not remain fixed boundaries for very long. Some of them will not even last until the end of this decade. And some of the others we do not fully understand. Even in ourselves, we take them on faith.
[00:16:48] And so I think what it sets up here is this, I think what the church is doing is needed. I think we have to get the conversation outside of just the [00:17:00] technology bubble here. And we need people challenging ideas. And I think there's gonna, You know, I like the discussion I'm seeing, I like the direction of the respect, given by the people commenting on this and saying, listen, I respectfully disagree, and that's okay.
[00:17:15] Like, that's what society is about. We should have respectful disagreements and we should have conversations around these hard things. And then just to highlight, you know, a final element here is Chris Olas remarks and I, he published the full remarks on X. We'll put the link to the show notes in, you can go watch, you can also watch the video of it, but I think a few of these are, are worth just reading directly.
[00:17:37] So this is direct from his remarks. He says, every frontier AI lab, including Anthropic. Operates inside of a set of incentives and constraints that can sometimes conflict with doing the right thing, the pressure to stay commercially viable, and to stay at the research frontier, geopolitical pressure, and the older planer pressures of pride and ambition.
[00:17:57] No matter how sincerely any of us intend to do [00:18:00] the right thing, and I believe many of us do, we will always be influenced by those incentives. That is why if we want this technology to go well, it is enormously important that there be people outside those incentives, people who care about things going well, and insist on safety, who are paying close attention, who are willing to say hard things, who are willing to be our earnest, thoughtful critics.
[00:18:22] It is through dialogue and mutual effort, through the push and pull, that humanity will achieve great things. That is what I see in Magnifica Humanis, and it is why I am grateful to His Holiness in the church for taking up this work of discernment. Some might believe that matters of AI are best handled by computer scientists like myself.
[00:18:42] They're mistaken. The questions raised by AI are bigger than the AI research community. Not just in their implications, but also in their nature. AI systems are not engineered the way a bridge or an airplane is engineered. We understand an airplane because we designed every part of it and we [00:19:00] understand the physics that act on it.
[00:19:01] AI models are not like that. They are grown on a structure roughly modeled after the brain on an enormous inheritance of human thought and speech. And what has grown is far more subtle, odd, and beautiful than science fiction prepared us for. They are not the cold calculating robots we were promised.
[00:19:19] They are made from us, from our words, and as the Holy Father observes, they remain in important ways, mysterious even to those of us who train them. He then highlighted three questions for discernment, and again, this is my whole point of like. I love the idea that we are presenting things that we should be discussing outside of these tech bubbles.
[00:19:39] So he said, the first is our duty to the global poor. There is a real responsibility that AI will displace human labor at very large scale. If that happens, supporting those displaced will be a moral imperative of historic proportions. Second is the need for moral imagination and ambition regarding human flourishing.
[00:19:57] If AI models are going to be widespread, [00:20:00] what does it look like for humans, families in the world to flourish? And third is the need for discernment on the nature of AI models. I am a scientist. I lead a research team that studies the internal structure of these models. What is actually happening inside of them.
[00:20:15] And I will be honest, we keep finding things that are mysterious, even unsettling. We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease. I don't know what that means, but I think it warrants ongoing discernment.
[00:20:35] And then he closes. With a, a section titled A Beginning and he said, we need more of the world, religious communities, civil societies, scholars, governments, and indeed all people of goodwill to do what his holiness has done here. To take seriously, to look closely and to push events in a better direction.
[00:20:54] We need informed critics who will tell labs when we are failing. We need moral voices, that the [00:21:00] incentives cannot bend. I just thought it was really well said. He got some pushback from some of the AI elitist who are like, you know, thinking he's over exaggerating this and the fact that we don't understand what's going on.
[00:21:11] So you're always gonna have disagreement, but most AI researchers that you hear about that I. Trust more than others will say this, that we don't understand them, we don't understand why there's emergent capabilities and behaviors. And it's, you know, I was rewatching something from Ilya yesterday where he was talking about language models and predicting the next token.
[00:21:34] And is that just statistics or is it actually some element of understanding? And is human understanding really much different than predicting, the next token? It's just like, I don't know, it's like an infinitely fascinating debate that there aren't really answers to right now, because we still know so little about the human mind and what is human consciousness and, you know, what is the human spirit?
[00:21:58] It's it, I don't know, there's just [00:22:00] so many unanswered questions that this is the stuff like I could just sit around all day and
[00:22:03] Mike Kaput: Right,
[00:22:04] Paul Roetzer: and not debate, just like openly discuss what are the possibilities. So I love that this is happening at a very broad level that will reach more people.
[00:22:14] Mike Kaput: You know, and Paul, one final note here that just came up in some research I was doing, I was curious, like none of this stuff happens in a vacu
[00:22:22] So, you know, the encyclical itself, it's actually, I didn't know this, it's become much rarer for popes to issue encyclicals. So Leo the 13th back in the 1890s had like 85 of them, whereas Benedict, had three. Francis, the most recent Pope just had four. So a first encyclical is like the thesis statement of your entire papacy.
[00:22:45] Like Leo the 14th was mentioning AI literally in his first speech to the College of Cardinals when he became Pope. And I just wanted to share one final thing here of like, I had Claude code kind of do some, spin up some research agents and do a summary of like, Hey, [00:23:00] when the first encyclical, Rerum Novarum the one that this is kind of honoring, came out in 1891.
[00:23:07] What was like going on at the time? So it says obviously peak of the industrial revolution, but like, give me a sense of what we're talking about here and take this with a grain of salt. But I thought this was super interesting to consider in the light of our, our current times, which is, you know, the industrial revolution was at its peak it says millions who'd lived for generations on the land or in craft trades got pulled into squalid fast growing factory cities.
[00:23:30] The agrarian slash guild world, how people had worked for centuries was dissolving. The Latin title of this means literally of new things. And the new things part was the era's kind of loaded phrase For revolutionary upheaval, conditions could be brutal. There were grueling hours, dangerous factories, child labor's, no safety net.
[00:23:50] A tiny elite got fabulously rich off workers who lived in misery in this telling. But here's the interesting part, there were, they call it a war for the workers' soul. Two [00:24:00] forces were fighting over the displaced masses. Rising socialism slash Marxism Communist Manifesto came out in 1848, promising to fix everything.
[00:24:08] Versus laissez-faire capitalism that treated workers as disposable inputs. The church saw itself losing the working class to the socialists, which was a social catastrophe and an institutional threat. So they stepped in to offer some guidance. I'm not saying that's the times we're living in now at all, but I think it's interesting to have that historical context and there's more at play than just what startup is raising funding here.
[00:24:31] Paul Roetzer: Yeah. And I was reading a little bit on, you know, the church's position at that time and had the church gotten everything it was proposing and asking for how it actually could have thwarted some of the advancements in society that
[00:24:45] Mike Kaput: Right.
[00:24:46] Paul Roetzer: Became so good for us.
[00:24:47] Mike Kaput: Right.
[00:24:48] Paul Roetzer: And so there's always this balance between.
[00:24:50] Trying to project out what is right when it comes to maybe slowing down technology or reducing the impact. And then, you know, [00:25:00] what ends up being in the best interest of humanity over time. And that's for the scholars to debate and talk about and things like that. It's, yeah, it's, it, it, again, the whole reason we let off with this is the impact is so far reaching and I think we are just, as Chris Olay said, it's, it's really just a beginning of this age of discernment where we are gonna start to have these very broad conversations and you're gonna need a lot of people from a lot of different backgrounds, who have a voice in trying to figure this out.
[00:25:35] Mike Kaput: All right, so in our second topic actually quite related here, there is plenty of debate going on before even the scholars get to it because this past week there's been another wave of evidence that AI's public perception problem is escalating fast. this was captured pretty well by Alex Kitz we've talked about before, is that make on this past year on our main stage, in a piece on his big technology Substack, he wrote [00:26:00] about AI's public relations emergency, and he actually noted a few threads coming together here, which included college grads, booing several commencement speakers at the very mention of ai, and also he cited dismal new polling showing a vast majority of Americans are against data centers.
[00:26:17] Both those should sound familiar. We've covered them both on previous episodes. And basically he writes, quote, to reach people between the ages of 18 to 25. Advertisers have long spent disproportionate sums of money. Young people make choices and develop loyalties that last a lifetime, and when they develop a preference or.
[00:26:35] In that era, it becomes extraordinarily difficult to dislodge. So what's gi what given what's happened on American campuses over the past week, artificial intelligence is now in a serious public relations emergency. Now Kitz goes on to note, this isn't solely a messaging problem, but the messaging out of Silicon Valley has not helped.
[00:26:57] So on this past, on this past week on the [00:27:00] Joe Rogan podcast, venture capitalist, mark Andreessen praised AI by saying the bots quote, never get frustrated with you, never get sick, never file HR complaints. He was framing that as like, Hey, this is great for productivity and jobs, but as Kitz noted, kind of a hard sell to unemployed new grads.
[00:27:18] Now, at the same time, there was also a wired investigation that revealed the Department of Homeland Security and the FBI are circulating reports about a new domestic threat category they call anti-tech violent extremism. So one New York Counter-terrorism assessment warned that the chaos from emergent AI over the next five years could fuel large scale protests, civil unrest, and anti-tech violence, especially in big cities.
[00:27:45] So Paul, a lot going on here, stuff we've definitely talked about before, but it does not seem to be slowing down. I mean, you know, this is getting up there for me in terms of the things I lose sleep over. It seems like the mood around AI is getting kind of nasty.
[00:27:59] Paul Roetzer: [00:28:00] Yeah. So I 've talked about this idea of the hero and the villain before when it comes to AI leaders and AI labs.
[00:28:05] I mean, probably by going back a couple years ago on the podcast I was explaining this, but I've had this conversation with some tech leaders in labs, I would say. so a couple years back I was basically presenting to them like, listen, this goes one of two ways. These labs are either the heroes or the villains.
[00:28:25] And if you keep. Ignoring the displacement of jobs and the environmental impact and just trying to gloss it over, you will eventually be the villain. And so my argument to labs was, you, you have to understand the dynamics of society and how it's gonna, this technology's gonna be perceived and you have to do things like lift up human workers and environmental, considerations as you're doing this because otherwise it's gonna get too late.
[00:28:55] And unfortunately, like I don't, I don't know of any labs that like [00:29:00] took the right path. And so now where we are is the labs aren't trusted. They have a lack of credibility. So it doesn't really matter what these labs are saying, or a Mark Andreessen is saying, like, no one's gonna believe 'em. Like they're not, it's really hard for them to be the ones that can shift the sentiment around the technology.
[00:29:18] And so, you know, I think the problem has been whether it's openAI's, Anthropic, Google, take your pick meta. They just sell this future of abundance and they've never been good at putting tangible ideas behind what that is or how it happens without pain, like without years of job displacement and disruption and things like that.
[00:29:40] And so I think that the labs are gonna continue to struggle with credibility. And then I think some of the other problems is these leaders in tech and VC are just straight up too arrogant and cold or too ignorant to the reality outside of Silicon Valley to care about the real impact on workers in society.
[00:29:57] It's, I've said this many times related to jobs like I, [00:30:00] they either just don't understand the reality or they just live in this bubble of, well, yeah, look at software companies, they're growing or like, we're still hiring more engineers. It's like, okay, that's like. less than a half a percent of all workers in the world.
[00:30:14] Like that's a small fraction to look at as a, a cohort of people to say it's not gonna impact jobs. H So what we're seeing is jobs and data centers are gonna be the wedges in politics and campaigns are already being funded by activist groups, political groups, and foreign adversaries to influence public perception in the United States about those things.
[00:30:34] just last week we saw a change in tone for Anthropic leaders. You know, set Chris ole aside, Dario Amide, who. I don't know, 16 months ago said that all knowledge work was gonna be obsolete within 18 months. It's all of a sudden like changing his tone. 'cause now you have to IPO and you have to change the perception of what it's gonna do, to jobs.
[00:30:55] So, interesting. Quick little fun AI project. And again, this is how Mike, and I [00:31:00] think you could hear like Mike was doing AI research in real time as we were doing this. This is how we work. I literally took the, cantits article and I threw it into ChatGPT and Claude, and Gemini this morning as I was getting ready for the podcast.
[00:31:13] And I said, read this article and then I wanna ask you some questions. And I just put the link in it, but all three of 'em read it. And I said, okay. If you were leading PR efforts for the AI industry representing a group of the leading AI labs and AI infrastructure companies, what would be your 10 step plan to combat the growing negative sediments and increased support for AI across demographics?
[00:31:30] H Same prompt to all three of 'em. So. I'll just give you the highlight of what each of them did. So ChatGPT, you know, said, okay, we gotta focus on trust, legitimacy, and shared benefit problem. It analyzed the sentiment. It actually pulled in some pew research, Stanford 2025 index, a business insider article on data center opposition, things like that.
[00:31:51] And then it did a 10 step plan. I'll just give like the highlights here. So stop selling inevitability, start selling agency. The industry's default posture has been AI is [00:32:00] coming, adapt, or be left behind. The framing is toxic because it makes people feel powerless. The new message should be AI's. Future is not predetermined.
[00:32:07] It will be shaped by workers, communities, educators, policymakers, companies, and citizens. AI should work for people, not happen to them. H Talk about create an industry-wide ai social impact that looks at jobs, youth and education, community safety and trust, and shared prosperity. it said three, segment the public by anxiety levels, not demographics.
[00:32:26] So think about people who feel threatened. Careers, community threats, safety, threats. Fairness. Threats. Control. Threats, exclusion. Threats. four. Make jobs the center of a campaign, not a defensive talking point. So AI will change work and we are accountable for helping people move up, not just move out.
[00:32:42] five. Reframe AI literacy as civic. LI literacy. Everyone deserves to understand how AI effects work, learning media privacy, and public life. Six, build a community first infrastructure model. So again, go community by community and look at the impact that these things are having. Don't talk about it as [00:33:00] like a global thing.
[00:33:01] Re this is a really important one. Number seven, replace tech CEOs as the primary messengers. I agree a hundred percent. So the least persuasive messenger for AI will help you is often the billionaire building the system. So I It's a hundred percent true. They just, they have no credibility when it comes to this.
[00:33:18] H DEMA, Saba, in my opinion, maybe Dario to a degree are the only two. That have any credibility right now when it comes to this? Outside of that, no one's gonna believe them. H move from abstract benefits to visible local proof a hundred percent. Take safety concerns seriously without sounding defensive, and then build a permanent reputation operating system or an AI public trust council.
[00:33:38] Now, Claude, I thought did an interesting job, and this was, I used 4.8 opus for this one. And I love the beginning of this before the 10 steps. It says the strategic frame should, that should govern all of them. You cannot out message, a felt material reality. The booing isn't a perception problem layered on top of a a fine product.
[00:33:59] It's a rational [00:34:00] response to two true things. Entry level jobs are getting harder and data centers are showing up in people's communities. So it's basically saying like, you can't just use PR to get rid of this. You actually have to address the fact that this is an underlying problem in society and accept that and embrace it and like figure out how to solve it.
[00:34:15] So it talks about imposing message discipline on the principles. Fixing the data center deal, not the data center messaging. So actually like Do something about water, noise tax, abatements, land, things like that. Put real money where the fear is visible, entry level hiring and opportunity creation.
[00:34:31] Don't just like say, Hey, it'll all work out in five years. replace the replacement narrative with an augmentation narrative that's backed by evidence, not ascertain, localize everything, win it community by community, not as a national brand. Totally. And again, I have a background in pr so I'm like, as I'm reading this, I'm like nodding my head a lot.
[00:34:49] I was like, this is actually really good. Yeah. recruit authentic validators and pay none of them. I love that. Like this is not an influencer campaign. We are not paying people to message that data centers [00:35:00] are good. Find credible people who can actually talk about the positive impacts overall, like address the negatives.
[00:35:07] Don't just message this. meet 25 to 18 to 25 year olds where they are. fund independent measurement that turns the debate from speculation to data and give politicians a constructive off ramp before bans get drafted. And then Gemini did a decent job too. there's this, the most abbreviated I thought Gemini is like obviously.
[00:35:27] Tuned to give very short answers to everything, but, acknowledge and validate. So empathy, first, replacement to empowerment. Address the infra infrastructure pushback on a local level. yeah, so a lot of the same ideas, but you get the gist of it. It's like, I think the industry has a problem in that they're not addressing head on the issues that they're causing.
[00:35:49] Like, so you can't just gloss over job displacement by say, it's not gonna happen. It is gonna happen. Stop. Like pretending like it's not. Or if you, if you truly don't believe it [00:36:00] is, go talk to people outside of your bubble. Go, go sit with like actual marketing teams and sales teams and like, we'll get more into the job stuff in, in one of the upcoming topics today still.
[00:36:10] But like, go talk to actual people outside of Silicon Valley, right? There's no way this isn't massively disruptive. So I don't know. I think like it is a PR problem, but it's more a fundamental accepting. Of like the reality of what's happening problem. And that's why I think like Empire of ai, while that book by Karen Hao is very.
[00:36:32] Disliked probably by a lot of technology leaders, and they spent a lot of money trying to discredit a lot of that book. But it's also why we're putting Karen on the main stage at Macon this year. It's the message people have to understand. This is not all simple. It's not all like sunshine and rainbows.
[00:36:47] It's not just a future of abundance. There's a really messy part of this that we all have to understand and come to grips with before we can figure out a more positive future for everyone.
[00:36:57] Mike Kaput: Yeah, I would say Silicon Valley, [00:37:00] or at least the tech leaders at the labs, need to understand just the brand is in trouble.
[00:37:05] Because I even have, like with what we do, if there's even a mention that I quote unquote work related work related to ai, there's been multiple people in my personal life. Nothing crazy, but that's, it's been a negative. Yes. And that's, that's not me explaining all the great stuff we do at SmarterX in our approach.
[00:37:24] That's just the two letters got mentioned and suddenly we have a, a problem and it's like people need to realize that might, maybe that's isolated, but. There's a real issue here.
[00:37:35] Paul Roetzer: Yeah. So when we, when we launched Macon in 2019, sorry, AI conference, the conference theme that I created that we still used to stay was more intelligent, more human.
[00:37:43] And I used to get like a lot of pushback, like, what the hell do you mean more human? How is AI gonna make us more human? And so I spent a lot of time even back then trying to explain how, my theories of ai and like why I thought if we did it the right way, if we did it in a response way, it could actually benefit [00:38:00] humanity.
[00:38:00] And I feel like that's just like the AI industry just missed all of that. And now they're scrambling to try and do it, and they're gonna throw tens or hundreds of millions of dollars at trying to fix public perception, when in reality they should just accept the fact that it's not all, simple and beautiful and future of abundance.
[00:38:20] And they gotta deal with the dark parts.
[00:38:22] Mike Kaput: Yeah. All right. So our.
[00:38:25] Mike Kaput: Third big topic this week is one that in different ways a lot of companies are feeling, so the cost of AI is soaring, especially in enterprises. And Paul, you actually wrote about this in your exec AI Insider newsletter this past week, framing it as one of the biggest challenges leaders are facing right now, which is how do you figure out how to budget for and manage the seemingly insatiable demand for AI and agents?
[00:38:50] So we actually had some reporting from several news outlets, including Axios and the Wall Street Journal over this past week, sharing how basically the bills for AI [00:39:00] usage have started to explode in corporate America. So the journal reported that some enterprises have hit their entire annual AI budget in just three months.
[00:39:09] Others saw their spending on ai, double or triple. Microsoft reportedly canceled most of its internal CLO code licenses, partly over cost. Just six months after rolling this out. Uber's CTO went viral for admitting that the company blew through its entire 2026 Claude Code budget in four months. Its COO said the costs are now harder to justify because higher token usage was not translating into proportionally more useful features.
[00:39:38] There was also this like kind of almost viral Axios report saying that somehow, I don't even know how this is possible, but one company spent half a billion dollars in a single month after failing to put usage limits on its clawed licenses. So Paul, these are obviously just random anecdotes about usage, but where do we even start here?
[00:39:58] So, I know you're hearing a [00:40:00] lot about this. It seems like a total mess.
[00:40:02] Paul Roetzer: the important part to me is to just look out and accept the fact that the demand for intelligence and agents is seemingly insatiable. Like it we're just at the ground floor here. And, I think it's only going to increase, this is like the top of the first inning to use a baseball metaphor.
[00:40:20] Like it's so early in the adoption of agents within enterprises and the capabilities of these agents, and the agents require more tokens than traditional text. You know, just talking to an AI assistant, I thought the Wall Street Journal article by Bradley Olson did a nice job of summarizing the state here.
[00:40:38] So I'll just read this excerpt, but said, use of artificial intelligence by big companies is exploding and the soaring cost has some of them pumping the brakes in a way that could complicate AI's triumphal march across the economy. Executives across industries this year have urged employees to integrate AI tools into their work, spending freely to encourage experimentation, and seeking to send a [00:41:00] message to Wall Street that their companies won't be left behind in a coming wave of disruption.
[00:41:05] All that enthusiasm. Enthusiasm has resulted in skyrocketing costs for so-called tokens, the basic unit of measurement for AI computing. As AI model providers seek to balance, supply and demand and manage their own costs, some enterprises have hit their annual budget in just three months, or reported seeing their AI spending bills double or triple.
[00:41:24] Now, corporate leaders are scrambling to bring down expenses by finding ways to ration AI use in their organization, steer workers towards cheaper homegrown tools and help them hone their skills to improve returns. So what I had wrote in the newsletter was basically if you set an AI budget at fall of 2025 during budgeting season, that was before the explosion capabilities of Claude Code.
[00:41:47] That has sort of set off this massive, exponential increase in the capabilities and use of AI tools. So those budgets are basically obsolete. And no one I've talked to has [00:42:00] any clue how to handle this. So I have actually met with executives at major enterprises who are in charge of AI access and token budgets in their companies.
[00:42:10] And they are scrambling to solve for how to do this, how to manage it, how to monitor it, how to meter it, how to decide who gets what budgets. And I'm not talking just about, like developers. I'm talking about like people within regular parts that you and I might experience ourselves at SmarterX. It's like we're hitting cloud code limits every day, and we're like, what do we do?
[00:42:29] Like just upgrade cloud code. We're not using the API. I'm just talking about like the standard licenses we have. We run into this with HubSpot all the time, where you run out of AI credits. Like what do you mean what you're using them for? Like it's still so opaque to me of like, how is this actually happening?
[00:42:42] What are they doing? And you can go in and like look at the token meter and it's like, oh, okay. We're almost there. No idea how we got there or what's going on, right?
[00:42:49] Mike Kaput: Yeah.
[00:42:50] Paul Roetzer: So I mean, you and I last week were throwing around just random ideas of like. Or should we give specific licenses to certain people who have more urgency?
[00:42:59] Like [00:43:00] if the studio team is working on a major research project or working on, an AI Academy course series, and they need the license for that Thursday and Friday because they're on deadline, right? But someone else is just running some experiment to just mess around and like build something and they sucked up all the company's token.
[00:43:18] It's like, so I mean, that's literally where we were at last, like, I don't know, last Wednesday or something. We're sitting around like, wha what do we do? So,
[00:43:25] Mike Kaput: yeah,
[00:43:26] Paul Roetzer: I don't know. It's, it's a very bizarre state. And then I pulled this Goldman Sachs report from. May 5th. So it was AI agents forecast to boost tech cash flow as Usha source.
[00:43:36] So Goldman Sachs is obviously pumping this as they're gonna like lead probably IPOs for one or more of these AI labs. Like, Hmm, they, the demand for this is gonna be insatiable. So they said tokens of the units of compute, think of them as units of information to be processed. AGI agentic AI requires a lot of tokens because many queries are repeated in sequence.
[00:43:56] It's like taking a simple chat bot request and blowing it up tenfold, 20 fold, [00:44:00] 50 fold. the report said with consumers and enterprises adopting AI agents, token consumption is expected to multiply 24 times to 120 quadrillion tokens per month between 2026 and 2030. Now that's a really hard thing to even comprehend.
[00:44:19] Put it in a little perspective, just, what was this two weeks ago? Google io Sundar Phai said that two years ago, Google was processing 9.7 trillion tokens a month across their surfaces last year, so a year later at io, so in May of 2025, that grew to 480 trillion tokens per month. Fast forward to May, 2026.
[00:44:47] That number jumped seven x to 3.2 quadrillion tokens per month. That's just Google. So like the 120 quadrillion is [00:45:00] probably a low number. Now none of us comprehend what a quadrillion is. It's a lot and it means demand is massive. So the last thing I'll, I'll mention here, Mike, is like one of the strategies being pursued by enterprise is like, well, how do we reduce the number of tokens needed?
[00:45:17] How do we reduce our reliance on these technologies? So our friend Christopher Penn had actually published. a post on substack called 18 Ways to Save AI Token Budgets. We'll put a link in, but he said, if your organization is proposing some measure of AI adoption, token usage is the absolute worst metric.
[00:45:35] We've talked about some of these organizations that were like, like meta, that were like in incentivizing people to use tokens, which is an absurd idea. I don't know how Right,
[00:45:43] Mike Kaput: right.
[00:45:44] Paul Roetzer: How that gets through leadership as like an actual KPI. but he proposed these 18 ways. He had six that were non-technical, like spend more time on planning and less, less on execution.
[00:45:55] So actually think through what you're doing before you just start talking to the chat bot. Yeah. [00:46:00] Always have things to build on decks, like know what's coming, use templates, plan big, act small, use smaller models. Like these are just some of get better at prompting. These are some of like the fundamental things we can do.
[00:46:12] So I don't know, I mean, I mentioned this on a recent episode. One of the things I think that is an important variable at some point is when are the models good enough that you just don't need the next number up, the next opus, right? It's like I've said before, like 4.6 sonnet to me, Claude was just an amazing model.
[00:46:30] Like ChatGPT 5.5 is a really good model and I think for us, you know, as we we're a research company, a media company, an education company, we do all these things like I. If you just told me that's all you get, like you're just stuck on GPT five, 5.5 or Right. I'd be like, all right, cool. Like sure, like 95% of our use cases are probably solved with that version of the model.
[00:46:53] Now every new model, it's like, oh, okay, well this one actually is quite better and I would prefer this one. But if that's, if you just told me we had to [00:47:00] stop there, I'd be cool. And then like 12 months from now, the cost of that model's gonna be 10x to a 100x less. And like Right,
[00:47:07] Mike Kaput: right.
[00:47:07] Paul Roetzer: We probably don't have this issue.
[00:47:09] But right now we do because we just keep wanting to use the more powerful, smarter models to do these more advanced things. And then AGI agentic processes are just gonna take over workflow. But you'd run into this issue of like, man, they're starting to cost more than people, maybe people weren't so bad.
[00:47:24] Mike Kaput: Right. Right. Yeah. And you know, to that kind of last point there, just managing this stuff, like, I just wanted to share really quick, like a couple more details around like the Claude usage issue we ran into, just to kind of make it clear for people, like how much of a frustration this is. So like, like you had said, we hit like our team Claude usage recently.
[00:47:43] I have no idea why. I still don't know why I t just
[00:47:47] Paul Roetzer: randomly happens. If it's never happened to, you're just like, Hey, you're done for four hours.
[00:47:50] Mike Kaput: You're like, right, right. What? And I assume we also hit our monthly usage, I assume just o obviously we added more people to the license. Great. More people are internally using Claude more, but like.
[00:47:59] [00:48:00] Here's kind of what really jumped out to me. So when we hit our usage, it like taps into credit overages that we had available. We have like limits set. We, we had like free credits or something, but like. For context, we're spending a couple hundred bucks a month on these licenses. We had about $200 worth of credits.
[00:48:18] Those were gone immediately. And the usage of those, I can see how they were or that they were used. It's wildly different across team members. I can see, I can see who used what credits we're talking the full range of usage from a few cents to a few dollars to a few dozen dollars to one person who used half the credits.
[00:48:38] And that's not their fault. I have no idea what the usage was for the team had no idea they were using a little or a lot. I'm sure they didn't. 'cause I know I wouldn't. They're just using Claudenormally. And I keep coming back to like, okay, let's say I did know all this stuff. Would this actually be useful information?
[00:48:55] Because to act on this, I would have to essentially micromanage like every [00:49:00] single person's use of Claude in this case. Like what they use it for every day constantly. So I really like the advice and I think that's a really important part of the solution. But I'm just struggling with this idea that like any way forward here that requires this to be managed feels like an unsustainable way forward, I would say.
[00:49:21] Paul Roetzer: Yeah. And like you have to manage to the value of the outcomes, right? Like that's inevitable. But how you do that, like what you're saying, you look at this usage across 20 some employees and it's like, well now we gotta drill into like what were they actually doing with it, and was the output valuable and things like that.
[00:49:36] But I keep coming back to this idea that I just want unlimited usage. Yeah. Right? And I don't care what you charge me for it, right? Like if Anthropic came today and said, okay, we've analyzed your usage, but for the last three months, here's how your team's using it. Here's the kind of outputs they're creating, whatever.
[00:49:51] if you want to spend $15,000 per employee. For, for the year. You can have unlimited uses. I would like, yeah, cool. Sign me [00:50:00] up. Like, right, I'm done, I'm done thinking about this. I don't want to deal with it anymore. because it is that valuable. Like you could look at any employee and say, okay, we're gonna hire a new customer success leader.
[00:50:10] Here are the 20 use cases we know they're gonna use it for. They're gonna run daily reports, they're gonna automate these elements of their job. To us, the value of that is like, well, we're not gonna have to hire an assistant for that person, or we're not gonna have to hire an SDR for the sales team because they're just gonna have an SDR agent.
[00:50:28] Right. Or a customer success agent that's gonna help them. Well, that's $80,000 or 120 thou, whatever that number is. So if like a lab just told me 15,000 flat and it covers all usage for that employee for the year, I'd be like. Sold, like,
[00:50:43] Mike Kaput: yeah.
[00:50:43] Paul Roetzer: Yeah. the outcome is obvious to me. The value is obvious to me.
[00:50:46] So like, this game of playing this credits, it is just bullshit. Like it's, it is such a unsustainable model. And if SaaS companies think that this is like this metering of this is the way of the future,
[00:50:58] Mike Kaput: wow.
[00:50:59] Paul Roetzer: I, they're all gonna [00:51:00] do it. I get it. 'cause it's like the most obvious thing to do. I can tell you right now, three years from now, we're gonna have an episode.
[00:51:06] We're like, remember what we were talking about? How unsustainable that was. Right, right. Yeah. They're all now charging on a human replacement value, or they won't call it that 'cause that wouldn't go well. But they're gonna charge like a. an outcome, value or like whatever they're going call the thing.
[00:51:20] Mike Kaput: Yeah.
[00:51:21] Paul Roetzer: Where it's just unlimited usage for knowledge workers. Yeah. There's no, in my opinion, no other way to do this, that you can get the support of the finance team, the HR team, again, get outside of software development that they may, that might be fine to them. Metering this stuff by tokens might make sense.
[00:51:38] In the software world. It does not make sense in an, in a broader knowledge work word for marketing and sales and success and HR and like, does not, it has to be a flat fee.
[00:51:49] Mike Kaput: Yeah. I couldn't agree more. All right, before we dive into rapid fire, Paul one, other announcement here. This week's episode is also brought to you by MAICON, the AI Conference [00:52:00] for Marketing and Business Leaders.
[00:52:01] This is our annual conference happening October 13th to the 15th in Cleveland, Ohio. like we had alluded to last week, we're super excited to share that Kevin Roose has joined our 2026 speaker lineup. He is a bestselling author. He's the tech columnist for the New York Times. He is one of the most widely read voices making sense of what AI actually means for business and society.
[00:52:23] This is exactly the type of speaker we love to have at Macon. Kind of someone who helps you step back and see where everything is heading. Kevin is joining a 2026 lineup that is stacked. We've got Karen Hao, how another award-winning AI journalist, author of Empire ai, Andrew Yang, former presidential candidate and tech and economic futurist.
[00:52:42] Paul, obviously you are speaking on stage. We've got Wil Reynolds from Seer Interactive, Katie Robbert at Trust Insights. Jessica Hreha, an AI leader at Veeam Software, plus a ton of other speakers to be announced. So MAICON is three days of keynote sessions, hands-on workshops built specifically for marketing and business leaders who are [00:53:00] actively figuring out how to adopt, operationalize and scale.
[00:53:03] AI across their organizations this year also adds a dedicated AI for CMOs Summit we're super excited about. And just as a note, ticket prices go up on June 27th. If you register before then and you can use the code POD100 at checkout, you'll save an additional a hundred bucks on top of the current rate, which is the lowest price we are offering.
[00:53:25] So be sure to act this month. If you're thinking about joining us at MAICON, go to macon.ai, MAICON.ai to register.
[00:53:35] Paul Roetzer: The other thing I'll note quickly, Mike, is the workshops, they're just like, they're, yeah. Like very popular, I would say. Yes. so we have, we just added a Claude workshop also. I dunno if, I don't know if we announced that yet, but we added, I think we did.
[00:53:48] I think that came out last week. so we have, if I'm not mistaken, Mike, we have, workshops on chat. GPT
[00:53:55] Mike Kaput: Yep.
[00:53:56] Paul Roetzer: Co-pilot.
[00:53:57] Mike Kaput: Yep.
[00:53:57] Claude Gemin I s that
[00:53:58] Mike Kaput: right? Gemini. Gemini and [00:54:00] Quad. Yep.
[00:54:00] Paul Roetzer: And then I'm doing one on AI innovation.
[00:54:02] Mike Kaput: Yeah.
[00:54:03] Paul Roetzer: So yeah, the workshop day on October 13th is, you know, an, an awesome value also.
[00:54:07] So make sure to check those out if you're registering.
[00:54:10] Mike Kaput: Yeah, for sure.
[00:54:12] Mike Kaput: Alright, Paul, so it it, amazingly, this is a rapid fire item, but by necessity this past week Anthropic release Claude Opus 4.8, which is an upgrade to Opus 4.7. They wrote in their post announcing the model. Users will find Opus 4.8 to be a modest but tangible improvement on its predecessor and it achieves better scores than Opus 4.7 across coding agentic tasks, reasoning, and knowledge work.
[00:54:36] It is available everywhere and its pricing is unchanged from Opus 4.7. We'll obviously be talking more about specific model pricing moving forward. The model costs $5 per million input tokens, $25 per million output tokens. One other notable improvement in this model is according to Anthropic honesty, Anthropic says, quote, one of the most prominent improvements in Opus 4.8 is its [00:55:00] honesty.
[00:55:00] We train all our models, to be honest, for instance, to avoid making claims that they can't support. But a general problem with AI models is they sometimes jump to conclusions confidently claiming to have made progress in their work, despite the evidence being thin. Early testers report that Opus 4.8 is more likely to flag uncertainties about its work and less likely to make unsupported claims.
[00:55:22] Now in the same blog post, Anthropic also announced some new features that are noteworthy here. There's a dynamic workflows capability in Claude Code now that lets it plan a job, run hundreds of parallel subagents in a single session and verify its own outputs. there's a new effort control setting in the Claude Web app and in cowork that lets users choose how much effort Claude puts into a response.
[00:55:47] So functionally this means when you're in Claude in the web app, you can now not only choose your own model, but choose things like low, medi high, extra, and max effort. Anthropic also in this announcement [00:56:00] specifically noted that organizations are currently using Claude Mythos as part of its Project Glasswing Cybersecurity Initiative.
[00:56:06] And they also said models of this capability level require stronger cyber safeguards before they can be generally released. We're making swift progress on developing these safeguards and expect to be able to bring mythos class models to all our customers in the coming weeks. so Paul, I know you had said you used Opus 4.8 for something I could personally say I'm enjoying it.
[00:56:29] think it's also notable they're teasing mythos. Is that essentially like Claude five that we're getting soon?
[00:56:35] Paul Roetzer: Unless they split off the naming convention, I would imagine that is likely Claude five. Yeah. And then just a note, the effort levels equal more tokens. Yeah. So again, if you're a newer user or a newer listener to the podcast and some of this language is, is, is not common to you by anytime you choose the more powerful model.
[00:56:56] If you go from sauna to opus, if you go from instant to [00:57:00] thinking, if you go from basic to advance, whatever it is, it just equals more tokens. And then you run up against the issue we were just highlighting in the previous topic, which is you're just consuming more tokens. The more reasoning you use within the model, the more, tool calling the more AGI agentic, all of that just equals more tokens, which to you, you, you're gonna have no idea how many you're using and what each thing is, but you're going to use them.
[00:57:26] Mike Kaput: Yeah. I would also say with that too, that might be really useful if you are using the API to build, you know, repeatable services on top of this, right? Where you say, okay, I'm gonna determine once, like what thinking level to use for this like, thing I'm building. I totally get that. But like the thinking levels for me in an trying to do a hundred different things a day that are very different in knowledge work is just so hard.
[00:57:51] It's like opaque. I have no idea. Yeah. I honestly just don't know.
[00:57:54] Paul Roetzer: Yeah. I'm the same. I, and even when I've used tools like, like a lovable, where it'll try and give you [00:58:00] guidance, it's like, hey, like doing something like this is like this number of tokens, doing something like that is this number of tokens. I don't know, man.
[00:58:08] And then you're like, yeah, when you turn the agents loose and they're just doing whatever they're doing and Right, right. They're gonna call on more powerful agents to help them do a thing. It's like, again, I just go back to it has to be unlimited usage of some sort because there's no way this scales.
[00:58:22] Mike Kaput: Yeah. All right.
[00:58:23] Mike Kaput: So our next segment's kinda less of a single story, more of a snapshot of two competing narratives that we've been tracking, you know, on AI and jobs that we're kind of. Just wanting to talk through how hard it's getting to reconcile these. So, on one side we've had, and we talked about this on previous episodes, the AI optimists are vocally arguing that job loss fears are overblown.
[00:58:44] So another kind of salvo in this that happened this past week, David Sacks, in the current Trump administration shared a widely circulated note from Apollo's Chief Economist headlined zero evidence of AI related [00:59:00] job losses. Yale's Budget Lab, which tracks AI's labor market impact monthly. Also found that measures of AI exposure, automation and augmentation still show no clear relationship to changes in employment or unemployment.
[00:59:14] They called the broader anxiety about jobs, quote, largely speculative for now. Yet on the other side, as we've reported week after week, layoffs are piling up. We've covered AI driven layoffs at Meta Block, Atlassian others. Intuit also just announced a 17% cut of its workforce. And most importantly, and Paul, I would love to kind of get your thoughts on this.
[00:59:35] Like anecdotally, the data right now just seems to be at odds to what we see and hear in direct conversations with leaders at companies. are you still hearing things from leaders that contradict stuff like what Sacks is sharing?
[00:59:51] Paul Roetzer: Yeah, I just kind of discredit anything. Sacks shares, nothing personal to Sacks.
[00:59:56] Like he's a mouthpiece for the, for the administration. and again, he may [01:00:00] be one of those people who actually truly believes what he's tweeting. I don't think that's true. Like I and Calacanis, like has been calling him out on x like all weekend. It was actually kind of funny. Like, so if you know, like Jason from the All In podcast, so they're friends and they're co-hosts of the All In podcast.
[01:00:18] Jason is like egging him on now, which like, come on dude. Trolling and sax is just sticking to the company line, man. He is like locked in on this and he is not gonna change his perception. So whatever I , I, the way I feel about this right now is so back in like, so I started the Marketing Institute in 2016.
[01:00:35] I started researching it in 2011. In in like 2017, 2018, we started seeing the early forms of the transformer being applied to writing. So GPT-1 comes out somewhere around like 2018, 2019. GPT-2 comes out a little bit later. They don't release it 'cause they're concerned about, it's. Impact and like the like, it's behaviors that they don't [01:01:00] understand yet.
[01:01:00] So they actually hold back the GPT two release. I'm actually reading Infinity Machine right now. I've mentioned this a couple times and it's, it like retells the story of that time where Demis Hassabis didn't even believe in language models. It took like three more years before Demis himself actually thought that language models were a viable path to AGI.
[01:01:17] He was anti language models for a couple of years. So back then I was meeting with heads of journalism schools. I was meeting with marketing leaders, business leaders, and I was saying, listen, there's gonna come a moment in the near future where AI can write like humans. And I would get laughed outta rooms.
[01:01:30] Like literally had like a director of a journalism school asked me to leave. Wow. It was like, it was offensive to them that I would propose the idea. So if then you looked at the data. The data wouldn't support that. Within three years we would have ChatGPT that could write like humans, right? But you could see the trend line and you could use the models yourself.
[01:01:49] You could experiment with them within studios, you could read the research papers and you could see where it was going. But anybody who is an expert in writing or journalism could have said in 2019, [01:02:00] 2020, it's never gonna write like humans. It can't do what we do. And based on the data of that moment, they would've been correct.
[01:02:07] I feel like that's where we're at with jobs. Yeah. You can sit here and tell me all you want to tell me about what the latest report says about jobs and whether or not it's having an impact and we're just not seeing it. And whatever. You're going to end up being wrong. Like go sit down with a department, a marketing department, a customer success team, a sales team, an HR team.
[01:02:28] Go talk to 'em about what their workflows are. Go talk to 'em about the problems they're trying to solve. Go live within one of these companies outside of the Silicon Valley bubble, and then talk to 'em about the realities of their business, the margins that they operate on, the lack of growth that they see for the company being in single digits, if they're lucky.
[01:02:46] And then tell me. That AI doesn't completely change the future of work and the structure of teams. I don't know how you do that. Like again, we live it every day at SmarterX we see the implications of this. We, we know we [01:03:00] can scale this company in a way that we could have never scaled a company three years ago with way fewer people needed to achieve massive growth of revenue and profits.
[01:03:11] Right? And so I just live in the reality of like. I 'm very confident of what is coming. And I've been in this place before with AI where people laughed at me, economists laughed at me. Leaders of journalism schools laughed. Like, who just didn't believe what, to me just was inevitable. Like, how can you look at the data and see anything but this future?
[01:03:33] So, I don't know. I mean, we gotta keep reporting on these people who just, for whatever reason are towing this line that like, it's just gonna be amazing and abundant and there's gonna be no pain and complexity for society. And I just feel like they're doing a disservice to people at this point. Like, just stop.
[01:03:51] Like,
[01:03:51] Mike Kaput: yeah,
[01:03:52] Paul Roetzer: I f that, I mean, but if it's truly your opinion and like you, you truly believe it, then fine say what you're gonna say. But [01:04:00] if in your heart, you know, you're just misleading people for some reason, Because it personally benefits you or your administration, whatever. Then that's just the nature of politics and society, I guess, as well.
[01:04:15] But I don't know. Like I've always said, I would just prefer. I'm wrong and like it ends up working out amazing. And we just roll right into this future abundance and there's never this displacement and underemployment and none of these bad things happen. I hope that that's true, that I'm just wrong.
[01:04:30] Right?
[01:04:30] Mike Kaput: Right.
[01:04:31] Paul Roetzer: But I would much rather prepare people for the possibility that that is going to happen and then it doesn't like then pretend like it's not going to and then it's here and nobody did anything.
[01:04:41] Mike Kaput: Yeah. I couldn't agree more. I feel like I just have like the least interest in being right on this ever.
[01:04:48] I pray every day. I'm just like totally wrong. 'cause that would be an incredible outcome and you can make fun of me forever for being wrong. I'd be worth it. But like, to your point about just ads SmarterX seeing it, it's like I'm not [01:05:00] challenging the data data's data, but like when I have to honor, like what I see with my own eyes and like the following is what I see.
[01:05:07] Like I'm easily 10 x more productive with AI now and that's for normal work. That doesn't even get into work transformation. Which allows me and my job to innovate and work in ways that were not even possible a year ago. Now, this is all really, really good for me. Like the Silicon Valley people. I agree with you.
[01:05:25] I am freed up to do the stuff I am actually best at, and there's plenty of that stuff to do because I'm lucky enough to work for a growing business in a high demand market. Here's the cac, catch. If I was not really good at all, the stuff I'm actually best at. The work I've done so far with AI would have automated me out of a job.
[01:05:43] Full stop. I would say, and I know for a fact that the work I've done so far with AI has stopped me from just anecdotally spending as much on third party service providers that we've spent in the past. That doesn't mean they're all cooked, it just means. There's a direct line here, and this is [01:06:00] all the best case scenario.
[01:06:01] So independent data just make of that what you will. I can't be the only one, I would say.
[01:06:06] Paul Roetzer: Yeah. And for us, like it's definitely already impacted the things we would've traditionally outsourced,
[01:06:10] Mike Kaput: right?
[01:06:10] Paul Roetzer: They, they're just far less things. We, we would've used freelancers and outside contractors and outside agencies to help with that.
[01:06:16] We just don't need them now because we can just do it ourselves.
[01:06:19] Mike Kaput: Right. Yeah.
[01:06:28] Mike Kaput: All right, so next up we have some political updates. This past week brought some notable AI developments out of both the states and Washington. big one here, Illinois lawmakers passed SB 315, which AI safety experts told Wired would be the strongest AI safety law in the country.
[01:06:38] It requires Frontier Labs like openAI's Anthropic and Google DeepMind to have an independent third party verify they're actually following their own safety commitments. This goes a step further than California and New York, which require disclosures and incident reports, but no outside auditor. As one advocate put it, we are in a situation where the AI companies grade their [01:07:00] own homework, which obviously to them is not a good scenario.
[01:07:04] Governor JB Pritzker says he plans to sign this. interestingly, both openAI's and Anthropic actually endorsed this bill. openAI's said that Illinois joins New York and California. As they join New York and California states are beginning to create a defacto national framework. Though of course, some Silicon Valley trade groups did lobby against this at the federal level, and we had briefly alluded to this when we went, on air last week.
[01:07:30] President Trump abruptly postponed the signing of an AI executive order. He had been expected to sign this would have created a voluntary system for developers to submit frontier models for federal safety review up to 90 days before release. Trump said he quote, didn't like certain aspects of it, worried it could slow the US in its AI race against China.
[01:07:51] Finally, Axios reported on a growing progressive resistance to AI inside the US Democratic Party. This is led by Bernie Sanders, Alexandria [01:08:00] Ocasio-Cortez Roanna, Elizabeth Warren, and Maine Senate candidate Graham Platner. Their proposals range from a moratorium on data center construction to new taxes on AI companies with Sanders framing the whole fight around the idea that AI and robotics exist to replace human labor.
[01:08:17] So Paul, interesting to see what's going on with these updates. especially the state level regulation. I mean, the Trump administration forever has been trying to preempt state regulation in favor of national regulations. It feels like at least the lab endorsement is kind of saying like, well, these are becoming defacto national patchwork regulations.
[01:08:37] Paul Roetzer: The state seems to be the only viable path at the moment. I don't think the federal's going to do anything, and even if an executive order got through, I don't see that making an actual, like, meaningful impact in any way. I think it's just like words on paper and voluntary whatever. Yeah, I think this is gonna keep becoming a bigger issue as the year goes on.
[01:08:59] I could [01:09:00] see a candidate, I don't think it's Bernie Sanders that's gonna get the, like, the support necessarily, but I could see the efforts they're making, you know, sort of moving into a more, I think there's gonna be a political candidate that emerges that does have AI as a very core foundation. Yeah.
[01:09:23] That, that has, a much broader impact from a national perspective. that, like, that's one of the thing with me, kind the shit, like with Andrew Yang coming and someone who made a campaign around AI and automation in 2020, all the way back in 2020. So I'm, I'm really intrigued to like, listen more to what he has to say.
[01:09:42] I just, I like that people are challenging this and bring ideas. I don't love that it's becoming political, but I always knew that would happen. But I think that bringing these things to the forefront and hopefully having more transparent debates about things, not using them to create fear and stoke [01:10:00] anger.
[01:10:00] Like, I don't like that, but I like that people are bringing new ideas and looking at different taxes and things like that. like I , the one thing I was thinking about over the weekend was, With data centers. I was talking to one of my relatives actually, and he was telling a story about like a local area, like a farm area.
[01:10:17] They were having to go through all these meetings in the community because they, the community was revolting against the idea of a data center there.
[01:10:24] Mike Kaput: Yeah.
[01:10:24] Paul Roetzer: And they, they think it's because of energy and water. And so I said, well, maybe one of the answers is these companies say, listen, we're gonna pay the energy bills for the community for the next 10 years.
[01:10:33] Like, yeah, you, you're, you're actually gonna, they, it's gonna go away. So it's, so if cost is an issue, it's like, okay, here's what we're gonna solve the cost thing, and if water is the issue, here's what we're gonna solve the water thing, again, attack the actual things, don't gloss them over and spend a hundred million dollars on some pack trying to like, convince people it's not reality.
[01:10:51] Just deal with the realities. Yeah. So I think that we have to get to the point where they become political issues for the anything to happen here. So. I don't know. In the end, [01:11:00] it's probably positive that all this debate is starting to happen.
[01:11:02] Mike Kaput: Yeah, for sure. All right.
[01:11:04] Mike Kaput: Next up, Microsoft released its 2026 Work Trend Index.
[01:11:08] This is a big annual report on how AI is reshaping work. they actually built this using. Trillions of anonymized Microsoft 365 productivity signals and a survey of 20,000 AI using knowledge workers across 10 countries. And basically where they come to the conclusion that as agents take on more of the execution of work, humans gain more agency to direct the work and make the calls and own the outcomes.
[01:11:35] So some interesting data they've got here. The number of active agents in the Microsoft 365 ecosystem grew 15 times year over year. It Roetzer to 18 times larger inside large enterprises. 49% of copilot conversations now support cognitive work, like analysis, problem solving, and decision making. Rather than just producing or finding information, two thirds of AI users say AI frees them up for [01:12:00] higher value work.
[01:12:01] really interesting here, and this gels quite a bit with our state of AI for business report that we released recently, which is, Microsoft found that, you know, organizational factors like culture manager support and talent practices drive more than twice the AI impact of individual efforts. So basically workers are ready, but their organizations are not only one in four workers say their leadership is clearly and consistently aligned on ai.
[01:12:26] Just 13% say they feel rewarded for actually redesigning how they work. already 86% of AI users that are more advanced, they're not just doing things fast, they're kind of applying judgment and they say they treat AI output as a starting point, not the final answer. So Paul found some of this data really interesting.
[01:12:48] I mean, especially this idea that individual capability is outpacing organizational readiness, which was literally a key finding from our research as well.
[01:12:58] Paul Roetzer: the thing I'm [01:13:00] gonna keep watching here, Mike, is just the people who are actually adopting and using the tools are going to see this kind of like massive change in, in how they work and the impact that they can have.
[01:13:10] Like we were just talking in the previous topic, all the people who are getting these copilot licenses, who are still using them to write their emails and summarize their meeting notes, like it's gonna be a tough road. Yeah. And so, you know, I don't know what the percentages are like in terms of, you know, of, of ever every a hundred co-pilot licenses that are given out in enterprise.
[01:13:31] How many people are actually becoming daily active users of that technology? I don't know that we've seen that data point or that Microsoft would wanna share that data point. but when I saw this thing about, 49% of copilot conversations now support cognitive work-like analysis, problem solving, and decision making.
[01:13:48] Maybe 49% of the active daily users of the technology, but not 49% of all people who have copilot licenses. Right,
[01:13:57] Mike Kaput: right, right.
[01:13:57] Paul Roetzer: So, I, you know, data [01:14:00] can be made to say whatever you want it to say, but I do believe that the people who are given the technology, who are trained to use the technology, who understand the agent capabilities of the technology, they will see transformation.
[01:14:12] The people who don't and don't get personalized training on how to use it, then they're just gonna keep using it as like, just another tool that doesn't bring a ton of value to them.
[01:14:22] Mike Kaput: Right.
[01:14:23] Mike Kaput: All right, so next step we've got our AI use case spotlight, where every week we give you a quick look under the hood at the real AI use cases we are deploying in our own work at SmarterX.
[01:14:34] So Paul, I have one I can quickly share and if you've got anything, feel free to chime in. so on my end, we have a few campaigns right now running where we're kind of considering doing some targeted cold outreach to try to get in front of business leaders we genuinely think would benefit from this thing we're working on.
[01:14:53] so for me as an experiment, more than anything, we're not actually putting this into production. I just wanted to actually show our [01:15:00] team what a more agentic tool like Claude Code can do with something like this. Mo most importantly, I just find to challenge us all to think bigger and think differently about how to tackle projects like something like cold outreach in an AI native way.
[01:15:14] So. What I mean here is, you know, you typically think of cold outreach as like you get a list, you hopefully do some research into each prospect. You create a template email, you customize those where possible. And then you know, of course all this like copying and pasting and sending emails, et cetera. So again, mostly as an experiment, more than anything, what I did was pointed Claude code at a webpage that described the thing we were promoting.
[01:15:39] It worked out, had it work out who the ideal audience is, come up with all the kinds of roles and seniority we might want to reach with this author. From there, you know, typically you'd actually just get a list of verified prospects likely through a tool like Clay. That's the way you should do it. I tested out, I was just curious.
[01:15:57] Claude Code actually went and found people it thought were [01:16:00] good prospects and then it guessed their email addresses based on research it did in into common email formats at a company. I'm not saying do this, I don't even know if it's right. It was just kind of interesting to watch it figure out that approach.
[01:16:12] But more importantly, the point here is me and Claude Code work together on this really killer email, and then it's able, if you connect it to an email inbox, I connected it to my personal email just for this demonstration. You can then just automatically spin up a hundred email drafts with the person's name, email address, customization as needed, and it did all this while I did other stuff.
[01:16:35] And I would say none of this replaces like the human part. This outreach does not work if you're doing it in a spammy way. Like it just has to be genuinely relevant. The point is, like our team, myself included, might have been tempted to approach this the normal way, right? Which is just like a lot of copying and pasting.
[01:16:52] It's probably stuff we don't even have time to do and thus it won't get done. So instead, I just wanted to remind [01:17:00] everyone AI tools, especially connected to certain systems, can now do these types of things, and it is very fun to test out smarter AI native ways of doing things.
[01:17:11] That's pretty cool. I did a.
[01:17:15] Paul Roetzer: talked recently about in Claude how asking for like an HTML output is sort of a key trick to getting more interactive and visual outputs from the system. And so I've been working on this visualization for a while and, I had some pretty limited time last week. And so, you know, I got up early one morning and I was just messing around and I was trying to create a new version of this thing.
[01:17:37] And so I went in and the key to me was I was working off of a previous version of an HTML output that Claude had created. And so now I went through, from a strategic standpoint, I did more planning. It's like, okay, rather than just saying, okay, gimme a better version, make it more professional, whatever, I went and said, okay, here's the filters.
[01:17:55] I'm thinking about, here's the classifications I want over a specific period of [01:18:00] time. Like I gave it very specific details. And then I said, what do you think about that? What am I missing? And I had an interaction with Claude to enrich the prompt and the project brief, basically. H Then I gave it back and said, okay, let's go now and create a V two.
[01:18:15] Of the HTML output and it did it, and it was just remarkable. And then I had to like run and do some other stuff that day, but then I sent it to the team. I'm like, Hey, just FY I 'm still working on this, but I had met with the team a couple weeks ago and said, I would own getting this further along and wanted to show them some of the progress we'd been making.
[01:18:33] but again, goes back to something I would've never done. I had no capability to create HTML documents prior to Claude's ability to, to do it for me. and it's so much better than just creating a Google Doc brief that I would turn over to developers. It's like I'm turning over a minimum viable product basically to the developers when I.
[01:18:54] Me with them and say, get this to production for me. So I think I've shared recently, like I'm doing a lot of strategic planning [01:19:00] using it as a thought partner and I'll highlight those. And then I do a lot of this like minimum viable product, just building visualizations using HTML to bring things to life that are in my head that otherwise I would've had to just explain in words to a developer.
[01:19:14] Mike Kaput: Right. Yeah. That's so cool. I love that. And yeah, seriously, can't emphasize enough. If you have not experimented with any of the HTML stuff, you're, you're gonna like it, I think.
[01:19:22] Paul Roetzer: Yeah. It's so cool. You're gonna
[01:19:23] Mike Kaput: enjoy it
[01:19:24] Paul Roetzer: and you can make it interactive. Put slider scales in, put tr like anything you you could do on a webpage, you can do in Claude by just asking for it.
[01:19:33] Yeah. It's so powerful.
[01:19:34] Mike Kaput: I love it. Alright, Paul, so We wrap up here. We got a bunch of AI product and funding updates. I'm gonna run through real quick. So first up, Anthropic raised 65 billion in a series H round that values the company at 965 billion post money. They had lead investors. They include Sequoia Capital, altimeter, Dragone, and include Green Oaks.
[01:19:54] They will spend the money on safety, research, compute, and products like Claude Code and Cowork [01:20:00] as run rate revenue has topped $47 billion. Anthropic also released the first update on Project Glasswing like we talked about. this is a collaboration with around 50 partners to use Claude Mythos preview to find software vulnerabilities.
[01:20:14] And they reported more than 10,000 high or critical severity flaws discovered across partner systems in the first month. openAI's has also laid out on its end its 2026 election safeguards, including surfacing reliable voting information in Chad GBT with live associated press vote counts on election night in the US and Brazil watermarking AI generated images offering cybersecurity tools to voting system makers and declining to run any political advertising this cycle, the openAI's Foundation committed 250 million to study and cushion AI's economic disruption funding measurement of AI's impact on jobs near term transition support like retraining and wage insurance, and longer term ideas such as new taxation models [01:21:00] and public wealth funds.
[01:21:01] The AI coding startup cognition maker of the Devon AI software engineer raised more than a billion dollars at a $26 billion valuation that more than doubles its September valuation, as it has a run rate revenue pegged around 492 million. Bloomberg reported that Apple plans a major AI overhaul in iOS 27.
[01:21:23] That's gonna be unveiled apparently at its developer conference on June 8th, centered on a revamp, more conversational Siri with a new chatbot style app, a fully customizable camera app, and new AI photo editing tools. YouTube is making its AI content labels more visible, moving disclosures directly below the video player and onto shorts, and will begin automatically labeling photorealistic AI generated content using internal detection signals.
[01:21:50] Starting this month, HubSpot introduced the agent CLIA command line tool that connects HubSpot data to environments like Claude Code Cowork, Claude [01:22:00] Cowork and Codex. So AI agents can run repetitive bulk and scheduled go to market tasks in the background without human oversight trajectory. A Palo Alto startup founded by former Google DeepMind and Apple researchers raised 15 million in seed funding led by conviction with backing from Jeff Dean and Fefe Lee to build a continuous learning feedback layer that lets AI models improve in near real time rather than between big training runs.
[01:22:27] And finally, pulia and AI operations platform that orchestrates agents to run business functions. Raised $30 million at a $250 million valuation. This is the interesting part though, founder Ben Cera says The company is approaching a $10 million run rate with zero employees, which is a claim that is drawing significant skepticism.
[01:22:51] One final note here as we wrap up, Paul, the SmarterX AI pulse survey is going to be live. When you listen to this, go to [01:23:00] SmarterX.ai/pulse. We had shared the results of last week's survey at the top of the episode. This week, we're gonna be asking about how if your organization's spending on AI is starting to outpace the value you're getting from it, and also over the past year, has your personal view of AI become more positive or more negative?
[01:23:20] Be very interested to see the results on that one. So Paul, thank you so much for breaking everything down for us. Big week as always, but super interesting.
[01:23:28] Paul Roetzer: Thank you Mike and everyone have a great week. We will be back, I think, in our normal time next week.
[01:23:34] Mike Kaput: I think so.
[01:23:35] Paul Roetzer: All right, thanks everyone. Thanks for listening to the Artificial Intelligence show.
[01:23:39] 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 a [01:24:00] SmarterX slack community.
[01:24:01] Until next time, stay curious and explore ai.