291 | Is a job apocalypse coming? Will Elon Musk take over the world? New model releases from open AI, and more important news for the week of May 8, 2026

Secure your spot for the MULTI-AGENT ORCHESTRATION AI COURSE: https://multiplai.ai/multi-agent-orchestration-course/

This week’s episode connects the dots between Anthropic’s explosive growth, Elon Musk’s surprising partnership moves, OpenAI’s enterprise expansion, and the growing debate around whether AI will create abundance — or a workforce crisis.

You’ll hear why the AI race is no longer just about better models. It’s now about compute, infrastructure, consulting, regulation, and who controls the future economy.

If you’re a business leader, this episode will help you understand what’s really happening beneath the headlines and what actions you should be taking now before the next wave hits.

In this session, you'll discover:

•  Why Anthropic’s Dev Day may signal a major shift in how AI agents operate 
•  What “Dreaming” means for long-term AI memory and autonomous improvement 
•  Why Anthropic’s 70–80X growth shocked even its own leadership 
•  The real story behind the Anthropic + SpaceX compute partnership 
•  Elon Musk’s larger strategy for AI infrastructure and space-based compute 
•  Why AI labs are rapidly moving into enterprise consulting services 
•  The growing debate around AI-driven layoffs and workforce disruption 
•  Key insights from Ezra Klein and Clara Shih on the future of jobs 
•  Why governments may begin regulating AI model releases 
•  New OpenAI releases that could reshape browser automation, voice AI, and testing workflows 

If you’re building a company, leading a team, or planning for the future of work, this is an episode you cannot afford to miss.

About Leveraging AI

• The Ultimate AI Course for Business People: https://multiplai.ai/ai-course/
• YouTube Full Episodes: https://www.youtube.com/@Multiplai_AI/
• Connect with Isar Meitis: https://www.linkedin.com/in/isarmeitis/ 
• Join our Live Sessions, AI Hangouts and newsletter: https://services.multiplai.ai/events

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Transcript

Speaker 2 0:00

Hello and welcome to a weekend news episode of the Leveraging AI podcast. The podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business and advance your career. This is Isar Matis, your host, and we have a completely jam-packed week this week. I probably could have recorded four full episodes of the things that are really important that I have to share with you this week. I will do my best to squeeze all of them into this one episode. Anthropic had their dev day this week. They've made also a huge announcement about a new interesting collaboration with SpaceX. We had multiple really interesting stories, opinions and data related to how AI will impact the economy or the workforce, if you want more specifically. We had the two big AI labs, both Anthropic and OpenAI, announce the fact they're going into the large scale consulting business, and they made the announcements on the same day. And so that tells you something on their ability to know what the other company is doing. And we have lots of really interesting and important rapid fire items, including the government potentially making the first step into nationalization of these AI models New model releases from multiple labs, including OpenAI and other really important stuff to talk about. So we have a lot to cover, so let's get started. This episode is brought to you by the Multi-Agent Orchestration course. It is a course that will take you from knowing nothing about how to maybe even use Claude, but definitely creating agents and scales and running them at scales in a way that can literally automate everything in your organization to your ability to do exactly that, all in four weeks, all online, two to two and a half hours a week with me as an instructor. It's not a self-paced course, so you can come in and ask questions. There are midweek meetings for you to make progress with asking me questions directly and getting your questions answered. It is the most transformative capability that existed ever. It literally changes the way you can run your business, whether you're a solopreneur or run a very large company. I've done this as a workshop to two different large scale organization, and it works there as good as it works for individuals. Actually even better because they have more resources and people can work together. So if this is something that you think can change your future, and I can guarantee you it can, go check out the link. We're now selling the third cohort. The first two sold out very, very quickly, and we have a few seats left on the third cohort that starts on June 22nd. So if you wanna know how to build agents at scale in a way that automate and connect to your existing tech stack and processes, don't miss out on this opportunity. The next course will probably be at the end of July, which is about a quarter away, and you probably don't wanna wait that long or you shouldn't wait that long. So click the link in the show notes and go and check it out. And now to the news. We will start with Anthropic's Dev Day. So Anthropic had their second ever Dev Day this week on May 6th, and they shared several different new releases that are interesting because first of all, they haven't made any announcements of any new models. They haven't shared any information or additional information about Mythos and where it stands and when and how it is going to be released to the public. What they did talk about almost completely is how they are improving the development and deployment of agents and how they are addressing the biggest gaps that exist today in the agent universe. Many things that people like me or the people that work with me or anybody else in the AI space has built on their own, they're starting to build into their product. So the flagship announcement they made is called Dreaming. Dreaming is now in research preview capability for Claude managed agent solution. And what it enables agents to learn from the past. Again, I've built mechanisms like these for myself and for my clients, but what it basically does is it allows Claude to not forget more or less anything, take notes about the things that are important and use these lessons learned in future sessions. If you want, this is almost like having an endless context window because every time Claude starts something new, it learns from everything it did in the past. Now, in addition, it has a proactive component where it can extract recurring patterns and then understand what mistakes were made or how it can be done better, and it can on its own, on the fly, generate new skills and new capabilities and new approaches to address these past patterns. For me, when I built this, that was a game-changing moment because everything I'm building now gets better and better over time on its own, and it is really fun and interesting and scary to watch. Now, they specifically said that doesn't retrain the model, and I understand that. It literally just takes notes on what happens in the past and tries to find patterns and then improve based on what wasn't working effectively in the past across multiple sessions. And while, yes, it does not generate a new better model, it definitely generates significantly better outcomes. The other thing that Anthropic released and announced this week is actually two features they announced in the past, but now they're going out into the public. One of them is called Outcomes, and the other is multi-agent orchestration. They're both going from research preview into a public beta, which means More or less everybody can use them right now, but they are still in beta mode. So Outcomes allows developers or humans, anybody who's creating this, to write a rubric that is describing what success looks like. And then a separate AI agent is evaluating the outputs of the process against the criteria as an independent evaluator, and then providing feedback to the process until it is actually done right. Again, I've built this for every one of my processes, and it works really, really well because instead of me having to be the first line of defense, the AI is doing it for me, and then the output that I have to verify or review is already much better than the first line of defense. So they're building this concept into the Claude infrastructure, and multi-agent orchestration just allows Claude lead agent to break down tasks into smaller tasks, assign them or delegate them if you want to specialize sub-agents who are going to be working in parallel through a shared file system, so they all know what's happening, and then give you the final output faster and more specialized. So this is now being shipped as a public beta, so everybody can use inside the Claude universe of managed agents. Now, early adopters are reporting dramatic results. So legal AI firm Harvey saw task completion rates increase roughly six X after implementing Dreaming. A medical document review company, Wise Docs, cut review times by 50% by using the Outcomes feature. So it tells you how significant this to actual real work. While these are really important and effective, and again, logical steps in the right direction, the two biggest and most surprising news were actually not the technical aspect. So first of all, what Dario Amadei shared is that Anthropic annualized revenue and usage grew between seventy and eighty X in Q1 of twenty twenty-six. They were planning very aggressively for a ten X yearly growth So a few statistics that Dario shared. API volume on the Claude platform is up nearly 70X year over year, and the average Claude code developer now spends 20 hours per week on the tool. I find this to actually be really low, but I guess there's a very long, long, long tail. I think if they would have looked at a median instead of an average, the number would have been much higher. Anthropic now counts over 1,000 enterprise customers spending more than one million per year on tokens. This is the explanation why they're financially growing in an exponential way that doesn't make any sense. It's because it doesn't add up 20 bucks a month licenses, which is what people are using. They have really large enterprise clients who are paying them for tokens, and the token consumption is going through the roof. And that doesn't matter how many people are using it. It matters how much each of the people are using it. And because people are using more and more tokens, running more and more sessions in parallel, this number grows very, very fast, significantly faster than the amount of people that they are onboarding as new individuals using Claude. But again, the most profound quote is that Dario said, and I'm quoting right now, "We try to plan very well for a world of 10X growth per year, and yet we saw 80X. And so that is the reason we have had difficulties with compute." This is Dario's way to apologizing for the several outages and issues that they had recently, and that very clearly explains why they do not have enough compute to serve the demand right now. If you remember a few weeks ago, I shared with you that I listened to an interview with Dario where he was talking about the bet on compute and why he's more conservative than OpenAI. And he basically said that even if they see a 9X growth, which is insane, and it's faster than anything at that scale ever grew in history, but he's betting on a 10X growth, so he's off by only 10%, he will go bankrupt because it's gonna be $100 billion short that maybe nobody will be able to give him. So he is going to be one of the most successful, fastest-growing companies in history, and he will go bankrupt because the numbers are so great, and hence why he was being more conservative in betting on compute. But now that puts them in a very big problem when they grew 70X instead of 10X. Which leads us to the most unexpected news of this week, which is a compute deal that was signed between Anthropic and SpaceX, giving Anthropic access to the full capacity of SpaceX Colossus-1 data center in Memphis that has over 300 megawatts of compute and more than 220,000 NVIDIA GPUs that are going to be available to Anthropic basically immediately. Now, this is both surprising and expected all at the same time and let me explain and provide my opinion of different aspects of this. So first of all, Musk previously called Anthropic misanthropic in an X post just in February of this year. He said that the company, and I'm quoting, "hates Western civilization." So you don't expect somebody like this to go and give an entire gigantic data center to somebody that that's what you think about them. But on the day of the announcement of this deal, Musk posted that he had spent time with senior Anthropic team members and was impressed, and he wrote, and I'm quoting, "Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector." So it sounds like Elon changed his mind about Anthropic and their goals, and Elon has always been a very clear supporter of let's use AI for humanity instead of as a way to destroy humanity. That's why he founded OpenAI. He wanted to be a countermeasure for Google because he did not believe that they will do the right thing with AI. So after they bought DeepMind, he wanted to have a group that will develop AI in a safe way for humanity. And despite everything he thought about Anthropic previously, and he was very clear on what he thinks about them, and he put them in the same bucket with OpenAI until basically this week, he changed his mind. But I think the story is a little more complex and nuanced than that. So let's look at a few other aspects of this. And I'm gonna change direction for a minute, but I promise you this will connect very well. I'm going to talk for a minute about hyperscalers and where they are at this moment. So if you look at Google, Microsoft, Amazon, Meta, and Oracle, they are collectively going to invest seven hundred and twenty-five billion dollars in AI infrastructure and power this year. Three-quarters of a trillion dollars that they're going to invest in new capabilities this year. Now, we, in parallel, they are looking at a backlog of orders from companies who wants to use these tools of over a trillion dollars. Now let's look at some specific examples. Google's Cloud backlog this week doubled from two hundred and forty billion to four hundred and sixty billion, driven by an Anthropic commitment to a two hundred billion over the next five years This deal, by the way, briefly propelled Google past Nvidia to the world's most valuable company. That lasted a very short amount of time in after-hours trading, but it doesn't matter. It pushed Google to be the most valuable company in the world, and now they're number two, still over five trillion dollars in valuation. These numbers were imaginary just two years ago. Microsoft Azure reported an eighty billion dollar in unfulfillable orders due to power constraints. Now, out of those one trillion dollars of backlog, six hundred billion plus come just from OpenAI. Which means there is one company in this entire ecosystem that if it fails for whatever reason, puts the entire ecosystem at risk because they're holding or they are promising sixty percent of the revenue that is going to be generated to that entire ecosystem, that the entire financing deals and whatever leverage that is happening right now at the highest scale ever is depending upon. Now, another point to remember, which I shared with you in the index, is an interview that Elon did as part of his plans to deploy AI-optimized satellites, basically data centers in space, that is going to generate more compute in space than on Earth within the next five years. Now, Elon said, and I'm quoting, "We think we can get the cost of space-based computes below terrestrial in about two to three years. The power and cooling are free," which is correct. Obviously, there's a lot of other really, really, really, really big challenges. But if anybody can pull this off, it's Elon and SpaceX. So now what are my thoughts of the current situation and why this deal is happening is actually, I think, four stories that are combined into one. Story number one is that there are three companies that are running towards an IPO this year. They're each trying to position themselves in the best situation possible in order to make their s- IPO the most successful as possible. Story number two is Elon's long rivalry with Sam Altman and OpenAI. Story number three is Elon's long-term goals and ambition to combine all his companies into, like, a grand unified solution. And story number four is a potential new vision of Elon for SpaceX as a company. So let's dive into each and every one of these stories. So as I mentioned, we have three companies. We have Anthropic, we have OpenAI, and we have SpaceX. All are most likely going to IPO this year. They're all trying to do everything they can in order to make their IPO more successful than the other ones, which is a problem because there's a finite amount of money, and we're talking about shitloads of money, that's a professional term, that is going to be poured into each and every one of these IPOs. So the big investors will have to decide how much money they want to put into each and every one of these buckets. Which means money that goes into one cannot go into the other, which means in addition to the fact that they're all running on their own thing, they are competing with one another. So positioning yourself as the most likely company to drive higher results and long-term ROI is gonna drive more investment in you versus the other company. Now, as we know, Elon, and I'll be very gentle, doesn't really like Sam Altman or OpenAI. To be more specific, they are in court right now, and Elon has done everything he can to put roadblocks in the path of OpenAI moving forward. Now, the other side is Dario doesn't really like Sam Altman. We've seen multiple examples of them, including not willing to shake hands or stand next to one another on stage and in big events and stuff like that. So on a personal level and at a company level, they are in rivalry. Now, Dario is, as we just mentioned, is seeing the most incredible growth any company has ever experienced in history. We're talking about a 70x growth in one year in billions of dollars. We're talking about a 4x growth in revenue in one quarter, from nine billion to now there's rumors that they're now at a pace of $44 billion a year pace in June, when they were in nine billion at the end of twenty twenty-five. The only reason he's not growing faster, which again, doesn't make any sense, but that's the real reason, is he's constrained by compute. Now, Elon, that owns xAI, which is now a part of Spac- SpaceX, he's seeing that he's not really competitive anymore with the AI labs on developing frontier models. He tried. He may try again. I'm not saying he gave up, but he tried. He- And Grok is a good model. I use it for several different things because from a value per money perspective, it's actually a very solid contender. Maybe not as strong as the Chinese, but it is a US-based model that gives you a very good value for money when it comes to compute. He had a lot of turmoil with a lot of leaders leaving xAI in the past few months, so he's not competitive right now, and he admitted that out loud multiple times. However, he is extremely good at building large-scale data centers faster and more effectively than anybody else. He built Colossus-1 in three months. He built Colossus-2 in less than six months, and he has lots and lots of extra capacity in compute right now because he built Colossus-2, which is much more advanced than Colossus-1, and he doesn't have the demand because his models are not that competitive. Now, in addition, there's the trial going on right now, right? In which Elon Musk is suing OpenAI and Microsoft and specific people to say that they are stealing his charitable donation to OpenAI that he helped establish. And his real battle is with OpenAI. So as much as he maybe doesn't like Dario and/or Anthropic, which I believe he doesn't like. I heard him talk about them many times before. I don't think that one meeting with them or even several meetings with them in this past week changed what he thinks about them. But the enemy of your enemy is your friend. Both of them, Dario and Elon really do not like Sam Altman, and they really do not like OpenAI, and so their ability to collaborate makes perfect sense right now. So here's the play I think Elon is making. He's helping OpenAI's competition, which by definition puts OpenAI at a disadvantage when they are already in a tough race for the IPO. Now, by the way, talking about the trial for a second, and I'm sorry for jumping back. This situation got me thinking of what happens if Elon wins the trial. Or forget about win. Let's say he just somehow delays or eliminates OpenAI's path to IPO As we discuss, the entire supply chain, entire companies are dependent right now on the ability of OpenAI to pay them six hundred billion and growing. What if they can't do that? What if they cannot go public and they cannot raise the funds that are required to do that? What happens to this entire industry? So there are two different options. One is that a lot of really big, really successful companies will either be in really bad situation or even go bankrupt. Or maybe, which again makes it even more interesting, can Anthropic, Google, and SpaceX takes over the huge commitment that OpenAI has that they can't fulfill and then grow even faster than they're planning to grow right now because there's gonna be extra compute cheaper because it's either going bankrupt or giving it to them. So very interesting situation that could evolve relatively fast. Now again, when I say relatively fast, the current court is supposed to be making a decision by the end of May. But whichever way it goes, I'm sure it will go to different levels of appeals, so it's not gonna happen, very, very quickly. From Elon's perspective, that's not a bad thing because again, before this gets settled, OpenAI cannot go public because if they're not a for-profit company, they obviously cannot be a publicly traded company. So there is a very interesting- And again, I'm not a lawyer, and I don't know how this may or may not evolve. But from my personal point of view, This could turn into an entire miniseries on Netflix just on that aspect of it. But the other point that I mentioned earlier is that Elon is potentially positioning SpaceX as the next hyperscaler. So he is currently selling compute already. He just committed Colossus-1 to Anthropic. And he's planning to build data centers in space. It is the only company in the world that can do that. It is the only company in the world that can do that because they are the largest launcher of payload to space by a very, very big spread. They're building significantly larger launch capacity with Starship that he's planning to build hundreds of ships from so he can launch to space and do his Mars mission, which means he is going to have the ability to launch a huge amount of payloads to space. SpaceX is also the only company that launches compute to space on regular basis because of Starlink. So they have the experience, the knowledge, and the ability to generate a compute, launch it to space, and operate it in space in connection to Earth because they're doing it right now. Not even nearly the same scale that he's talking about. But again, if there's one company that can do that, that is SpaceX. So let's say Elon is overestimating his ability to launch to space, which is not going to be the first time he's overestimating something. More or less everything he estimated, he overestimated, whether it's how many Teslas they're going to sell, how many self-driving cars there's gonna be and when, and so on and so forth. But he always eventually got there. But let's say in the next five years, he can launch twenty percent of the extraterrestrial compute capacity that he's talking about. He might become the largest compute provider to companies on Earth, despite the fact that the compute is not going to be on Earth. Combine that with the fact that he already has a fast internet network in space to enable all of this to talk between the different computers, between the different satellites, and back to Earth. And again, he's the only company that has that. That is positioning SpaceX in a completely different light for the IPO. So going back to connecting the dots to what I said before, Google, for a short brief of time, became the most valuable company on the planet because they announced the growth not in search and not in paid ads, but in a commitment for compute, which is the direction that is Google has been pushing, and obviously Amazon has been pushing, and now Meta is pushing as well. If SpaceX can become that, they will not only be the largest company when it comes to shipping payload to space, they're also going to be one of the largest high scalers in the world all rolled into one company Now combine that with the fact that Tesla cars are a huge distributed network of AI compute that they're already using right now. Combine that with the fact that they're planning to ship more Optimus robots, which are additional AI distributed compute, and there's supposed to be a lot more of them than there are Teslas, and you have the perfect storm from Elon's grand vision to control a huge amount of the AI ecosystem and deliver that across multiple layers of the AI universe in a very, very successful way. So if I have to imagine why Elon is suddenly giving a data center, he's not giving, he's selling the capacity of his crowned jewel Colossus I to Anthropic almost overnight is that vision. And if along the way he can hurt Sam Altman and OpenAI, that would be a really nice bonus. So these are my thoughts on the current situation. But the good news for all of us Claude users is that Anthropic is planning to double the five-hour and daily rates of tokens, which means I might be able to downgrade to the 100 hour a month plan from the 200 hour a month plan and maybe not hit the wall. I'm going to test that out and see if it's actually going to work. So good news for Claude users doubling the rates of tokens that you're allowed to use in different time frames and very, very interesting news for the rest of the industry and the impact on the rest of the world. Now, the next topic that I want to talk about is the potential impact of AI on jobs. And you heard me talk about this many, many times, but we got lots of really interesting viewpoints as well as data points this week that I worked very hard to try to combine into a single story. So let me tell you the story, and I hope it will educate you as well. So first of all, a few pieces of news of announcement that happened in this past week or in recent time. The first one comes from Coinbase, who announced on May 5th that they are going to let go of 14% of their workforce, which is about 700 employees. And CEO Brian Armstrong is calling it, and I'm quoting, "AI-native restructuring." So they're not just cutting headcount, they're actually rebuilding their organization. They're cutting what he calls pure managers, and they're adding instead player coaches, which are basically managers who also do individual contribution work by managing entire agents. They're building what they call an AI-native pods, which is possibly one person or a small team that is running AI agents that cover the work of engineers, designers, and product managers. And as part of the announcement, Armstrong said, and I'm quoting, "Mass layoffs are coming to every company." That if that sounds familiar, if you've been listening to this podcast, we heard Jack Dorsey, the CEO of Block, saying something very similar. So if you remember, at the end of April, they cut 40% of their workforce, which is four thousand employees, and he said, "Your company is next." So that sounds very, very similar direction. In addition, we had Freshworks this week lay off more than 10% of their staff, also citing AI as the main driver. In parallel, we had Shopify announce additional layoffs. They laid off, about 53 people, additional layoffs in, their partnership division and a bigger number In their operational division and that is combining to over five hundred people laid off in twenty twenty-five. Now, this announcement comes almost exactly a year after Shopify's CEO, Tobi Lütke published his AI-first memo that became a template that then several other CEOs have shared. You can go back to these episodes and find that and listen, exactly what was discussed. But the memo basically said that if you want to stay in the company, you have to start using AI, and that you cannot hire any new employees unless you can prove that AI cannot do the work. So they have transformed how Shopify works. But despite that, what he said in the current layoffs was the following, and I'm quoting: "What you see right now is not AI layoffs." And he continued, and he called AI, and I'm quoting: "The perfect scapegoat, a target that explains away difficult decisions without fighting back." Now, if you want the sector-wide news, TechRadar's reported that the tech industry is nearing a hundred thousand layoffs so far in twenty twenty-six. We are just in the beginning of May. Now, before we continue about the other parts of this, I wanna say two things about Jack Dorsey and about Coinbase. So if you remember when Elon Musk took over Twitter, then turned it into X from Jack Dorsey, he fired seventy-five percent of the employees in the company, which may tell you that maybe Jack Dorsey doesn't run the most efficient shop possible. So his ability to lay off forty percent of the people now that AI is available, and again, Elon did that before that was an option, maybe he just doesn't run very lean when he grows businesses. On the Coinbase front, Coinbase had a horrible recent run. That's not unique to them. That is the entire crypto economy. And as a CEO of a publicly traded company He had two options. He had to choose between saying his entire industry is not heading in the right direction or say that they're restructuring based on the latest capabilities and technology, and the choice is very, very clear. You cannot say as a publicly traded company that your entire industry is going to shit because it's not gonna turn very well for your stock. But if you say that you're restructuring and you're building new AI capabilities in order to make your company more efficient, it sounds really, really awesome. I'm not saying that's the entire story. I'm just giving you another way to look at this. But we also had two very interesting things published this week, one by Ezra Klein, who is a contributor to The New York Times. He wrote an opinion piece to them. other from Clara Shih, who is the founder and CEO of Hearsay Systems. She's the former CEO of Salesforce AI, and most recently she was the head of business AI at Meta. So she knows one of two things about AI. Each and every one of them wrote a very interesting paper, and I'm going to start with some of the facts that they're quoting in these two papers. Because it's a perfect segue to the previous points that we just discussed with the layoffs. So the first one is a Quinnipiac poll of March twenty twenty-six that's found that seventy percent of Americans believe that AI will lead to fewer job opportunities, up from fifty-six percent a year ago. Thirty percent are worried about their own job. By the way, that's by itself is a very interesting, gap when seventy percent think it's gonna hurt jobs, but only thirty percent are thinking that it's gonna hurt their own job, which tells you a lot of it may not be completely justified fear, but that's still the statistics. Stanford, in their twenty twenty-six AI index, found that seventy-three percent of AI expert feel positive about AI long-term effect on jobs versus twenty-three percent of the general public that thinks that. 23% of people think that AI will have a positive impact on jobs. Seventy-three percent of AI experts feel the same way. They obviously have a vested interest in saying that gallup research in March 2026 shared that Gen Z excited about AI fell from 36% to 22% in this past year. So Gen Zs are not excited about AI. That's in a very gentle term. The flip side is Gen Zs that are angry about AI rose from 22% to 31%. So more or less the exact flip of the process. Now, in the canary in the coal mine, from Stanford Digital Economy Lab, they shared that even after controlling for ZIRP, which stands for zero interest rates period, and macro factors, AI is a primary statistically significant driver of cuts to entry-level jobs. Now, both Klein and Xi cite this paper. Bloomberg on April 13 of 2026, 43% of recent US graduates are underemployed. Fifty-two percent of the class of 2023 landed in roles not requiring their degree. That's half. Half the students from the 2023 class are doing something that has nothing to do with what they went to school for and have probably pretty big debt to cover from these jobs that is not what they were planning to work at. The Bureau of Labor Statistics projected and Experian has shared that the US household debt is now eighteen trillion dollars and growing, and that over half of Americans live paycheck to paycheck right now. So with that in mind, everything I talked about, the recent layoffs and the different opinions from the different CEOs, two that are very aggressively saying it's an AI, thing, and one that is saying, "No, it's not an AI thing. We are using AI, but we're not laying off people because of AI." So with that, let's start with Ezra Klein. Ezra Klein wrote a paper to The New York Times that is called "Why the AI Job Apocalypse Probably Won't Happen." So Klein starts with the poll that I mentioned in the beginning, that seventy percent of Americans be- believe that AI will lead to fewer jobs. And he's saying that the fear is rational given the rhetoric that comes from the AI lab leaders. He's noting that the loudest voices that are predicting white-collar wipeout has been coming from the CEO of the labs themselves. So if they're saying that, why won't the public believe that? Now, Klein leans heavily in his arguments on Alex Imas, who is an economist in the University of Chicago, that published a paper that is called "What Will Be Scarce?" And what he summarizes is that every honest answer on what does AI do to the economy has to begin with identifying what becomes scarce next. So what is Imas' paper talking about? What he's talking about is the fact that everything in history, every big change that happened in history for humans happened from what was scarce at that time. So he's framing it this way: when calories were scarce, most people farmed to generate more calories. When agriculture revolution solved that, goods became scarce, most people became manufacturers. When industrialization solved that, technical knowledge became scarce. Doctors, lawyers, software engineers got paid for the rarity of knowledge, right? So the question that he's asking is if AI commoditizes technical knowledge, what becomes scarce next? Now, what Imas is betting on, which Klein is endorsing, is that what becomes next, what becomes scarce, is work where humans actively want humans doing it, where you want judgment, taste, accountability, trust, care. And Klein is quoting Imas saying, "People are looking at an economy as it exists and asking which tasks AI can do. They should be asking which jobs people won't want AI doing or which services AI will make us want more of." Now, the second thing that Ezra Klein did is he interviewed several economists, and most of these economists he interviewed were quite skeptical that mass joblessness is on the horizon. What they're saying, and as he's summarizing it, is that AI is more of an automation wave in a long historical sequence of automations, and all of them generated fears of mass displacement, and none of them actually produced the same level of displacement. Now, Klein himself does not completely dismiss the displacement option. He mentions that AI may be a different kind of technology from previous tools because previous tools were complementing what humans do, and now AI has the flexibility and capability to actually substitute what humans do. I will add to that from my own thoughts that previous revolutions automated manual labor. Every previous revolution automated a specific kind of manual labor, and humans switched more and more to doing brain work, also known today as white-collar jobs. And now we're taking that away. And then the question I'm asking is what is left after we take that away? The only thing that is left is emotions, and I'm not sure how to drive an entire economy based on emotions. That, by the way, AI can mimic pretty well. The second topic that some of these economists are mentioning is that productivity mains might be an illusion And he's quoting that studies disagree whether AI is actually making people and companies more productive or merely giving them and their managers the illusion of productivity. Again, from my own perspective as someone who's using AI every single day for most of the hours of the day, and I'm growing two and a half businesses with a team of two, and I'm helping many other organizations implement AI in ways that is really transforming what they're doing and how they're doing it, I can guarantee you it is not an illusion. Those who are saying it's an illusion, like some of the people he has interviewed, just do not understand what is possible right now. And again, I'm not speculating on what will be possible a year from now or five years from now. I'm talking about what's possible right now and how dramatic its impact is on the work, on digital work as it is right now. I would argue that you can automate 80% or more of digital work with AI capabilities right now, and the only reason it's not happening is friction in the way it is being implemented, and risks, and data analysis, and data cleaning lists, and stuff like that. It's not a magic wand, but from a capability to do this. The capability exists today. The final argument that he's making, which is very interesting, is that he thinks that a world where eight millions are being displaced is actually worse than a world in which 80 millions are being displaced. And the reason he's saying that is that 80 million is a COVID scale event, and society will mobilize for that as a whole. Governments, industries, everybody will work together to figure it out. An eight million displacement is only 5% of the economy, and nobody will really care until it starts hurting very, very much, and it's gonna be a painful process until we get there. And he summarized it beautifully, and I will quote: "When I'm feeling optimistic about the world AI might make possible, I imagine a world in which we are richer than we are today and are encouraged to live more fundamentally human lives doing more fundamentally human things. When I'm feeling pessimistic, I imagine something like that same world, but the wealth will be hoarded, and we will value a depth of human connections that we no longer know how to provide." So his point of view is that there is no huge job displacement, but that there might be displacement that will be under the emergency level, and hence not enough actions are going to be taken, and hence there's going to be bad times for some people. The other opinion comes from Clara Shih, that I told you before who she is, and she wrote an opinion paper on LinkedIn that she called "The Six Confronting Myths We Tell Ourselves About AI and Jobs: A Call to Action." let me tell you how she opened her paper, and that's a direct quote: "The AI ship has sailed. We're not putting the genie back in the bottle, and we shouldn't want to, given the geopolitical stakes. But also doing nothing is not going to automatically result in AGI abundance for all. And she continues to say that in the current course and speed, AI will actually make the inequality and polarization of Western society even worse. And then she says, and I'm quoting again, "Everyone talks about the need for new energy infrastructure for the AI era. Shouldn't we also be talking about a new social infrastructure?" So then she lays out the six myths that we have right now that is helping us feel better about AI, either in individuals or as a society. So myth number one is that layoffs are just a correction from the ZIRP era. Again, zero interest rate and she's saying about this the following, and I'm quoting: "It is not helpful and actively harmful to tell people that everything is fine and will return back to what it was before after this correction." Again, if you connect this to what we talked about in the beginning of this section about the huge amount of layoffs that are happening right now, whether they're happening because of AI or people are using AI as an excuse doesn't really matter. And she's also quoting the Stanford Digital Economy Lab Canaries in the Coal Mine paper that we talked about before, saying that even after controlling for zirp, the AI is still a primary statistically significant driver of job cuts at entry-level jobs. The second myth is Jivon's paradox. So the argument that she is fighting is that The cheaper AI lowers the cost of tasks like coding, legal, research, et cetera, unlocking talent demand and creating more jobs like everything Jevons impact does. However, what she's saying is that history shows us that even when Jevons effects occur, supply growth often outpaces demand growth, resulting in lower wages. Similar to past disruptive technologies, AI slashes the skill floor for the once premium jobs, flooring labor supply and compressing wages. So what she's saying is that, yes, Jevons paradox works, meaning when things gets cheaper, we use more of it and not less of it. By the way, Jevons did it on coal usage back in the UK many years ago, but that effect stands today, right? When you make something cheaper, people use more of it because now more things that were not possible from an economical perspective now become cost-effective and drive a positive ROI. But what she's saying is that at the same time, if the stars do not align, it also generates a significant drop in the wages that are paid for these jobs because the supply and demand are not aligned. And she gave several different examples, including the London cab story on how GPS and Uber has completely slashed the pay that taxi drivers make in London. Her third myth that she talks about is the AGI timeline debate, and I'm quoting, "Whether AGI arrives in twenty thirty or twenty forty-five, the workforce displacements happening right now is real, measurable, and accelerating. Waiting for consensus on a timeline before taking action is like waiting for hurricane forecasters to agree on the exact landfall time before boarding up windows and evacuating people to safety." Now, you heard me say this almost exactly the same way multiple times. AI today is already good enough to automate Most digital labor. What happens next is not really relevant. It doesn't matter when and where we're gonna hit AGI, if it's even clear what that means. The impact on the economy and society is already here. Now, if you've been listening to this podcast for a while, you know that the opening of this podcast talks about a tsunami that happened that is already coming toward shore, and people on shore are still drinking margaritas versus getting ready. It's not the hurricane comparison, but it's the same exact idea. Myth number four that she's saying is that new grad unemployment is improving, and that is correct. The recent polls are showing that the unemployment rate of new grads is coming down. But she's saying, and I'm quoting, "Unemployment rate is a blunt instrument. It counts anyone working one hour a week as, quote-unquote, employed. What it misses is underemployment." The CS grad is doing data entry. The finance major is waiting tables. The accounting grad with a hundred thousand dollars in student loans now driving for DoorDash. And again, the recent statistics shows that very, very clearly, and I mentioned before, fifty percent, fifty percent of grads of the class of twenty twenty-three are not being employed in jobs that requiring what they actually went to school for. Myth number five: Just tell them to go to trade schools. So I'm going back to quoting her again. "This one has a bipartisan appeal and actual contains a kernel of truth. Skilled trades are genuinely undersupplied, physically hard to automate, and pay a living wage. Electricians average sixty-two thousand dollars a year, plumbers even more. There are durable, dignified careers. But the math doesn't scale. The Bureau of Labor Statistics projects roughly thirty-eight thousand new trade jobs per year nationally over the next decade. We currently have two point three million underemployed recent grads, and that's before accounting for the ongoing wave of mid-career displacements in finance, legal, marketing, and administrative roles." Myth six: AGI will bring great abundance to society. And I'm quoting her again. "Today, there's over eighteen trillion in US household debt and over half of Americans live paycheck to paycheck. So what exactly is the path from here to there? Productivity gains accrue to the owners of the productive assets. This is a policy of choice, and historically, it is the default choice when no one intervenes. The Industrial Revolution generated extraordinary abundance. It also generated child labor, 60-hour workweeks, and urban poverty on a scale that required decades of labor organizations, progressive taxation, and social insurance to address. The abundance was real, but its distribution across society was not automatic. And then she summarizes this way. "These six myths are dangerous not because they're entirely false, but because they're partially true enough to feel like an excuse to wait and see and do nothing in the meantime. The AI ship has sailed. I'm very bullish on the productivity gains that are coming. AI will bring abundance, but so far, nothing points to the abundance being shared with all of society." So here's what I think about all of this. There's a very big gap between what I believe or hope versus what I actually know as a fact. So fact number one is that AI generates huge opportunity in the short term because the adoption rate is very different from one company to the other. With the same resources, you can now do dramatically more and win significant market share in your industry right now. This is one of the reasons I've been pushing so aggressively on AI education and training because it provides an unfair advantage in a way that was not possible any time previously in history. Within a couple of weeks of the right training, your company can be significantly more productive and be able to be significantly more competitive in a specific niche. For those of you who care about this quick reminder, go and check the link in the show notes about the courses that we are providing. But the second fact that nobody's talking about is that there is finite demand. Once enough companies in a specific industry or specific niche learn how to use AI effectively, and they all will, they cannot grow anymore. At that point, you can do the same level of work to supply the needs that products, the services of your clients with X percent less employees. And I'm not gonna debate whether that X percent is five, 10, 20, or 50, but less employees. And if you want your company to survive, your only way to keep the people that can survive is to let the other people go. Otherwise, you will go bankrupt because you won't be competitive, and then you're not serving any of your employees. Fact number three, the level of unemployment in the Great Recession we had after the 2008 financial collapse peaked at 14%. That's it. The top unemployment touched 14%. It was less than 10 for most of that time. The Great Depression 100 years ago, the worst economical time in US history, peaked at 20% unemployment. so if this AI wave drives unemployment in the teens, but very different than previous unemployment waves, these are people who potentially make $100,000 to $300,000 a year, the economy just stops. When the economy stops, there are more layoffs, which is going to generate a very vicious cycle moving forward. Now, what I hope and assume is that AI will lead to abundance. However, I believe that these graphs don't align. These are like sine and cosine graphs, and the impacts of the first one before the good wave hits are going to be very dramatic, which means we're going to experience very turbulent times in the near term in order to maybe get to a future of abundance. Now, I obviously don't have a crystal ball. I don't know anything you guys don't know, but this is what I'm seeing right now, what I'm feeling right now, and what I'm thinking right now is, and I really, really hope I'm wrong. My third deep dive story was supposed to be the AI labs turning into consulting companies. But because this is already a long episode, I'm gonna turn it in real time into a rapid-fire item, and we're gonna go through a few rapid-fire items in just a few minutes, and you're gonna learn a lot from what happened this week in the next few minutes. So the two large labs, OpenAI and Anthropic, each launched their own private equity-backed enterprise AI services venture, both of them on May 4th and just hours from one another Both of these companies are going to do the same thing. They're going to provide experts into businesses to help them implement AI from the relevant lab, right? So OpenAI's people will help companies implement OpenAI's tools, and the same thing with Anthropic. And they're doing this in partnership with private equities because it makes perfect sense. Now, the reason this is a brilliant move for all involved parties, from the lab's perspective, that's very obvious. That gives them an immediate path to the companies the private equities own and manage, right? So it, they don't have to invest in marketing or sales and so on. They get immediate access to hundreds of mid to large scale companies that they can implement their tools in. The other aspect of this is the private equity companies. They are using it both as a defensive and offensive move. The defensive part is basically ensuring that their portfolio companies stay competitive in the AI era, and the offensive play is that they may turn them into the next McKinsey and PWCs of the world, giving them a stake at a multi-trillion dollar consulting economy. But this, again, move makes perfect sense and continues a step they did before. So they started collaborating with the top leading consulting companies previously, but now they own a stake in the actual process, which makes a lot more sense from the labs, as I mentioned before. Switching gears to something that may be a beginning of a really weird and not necessarily positive era is that the Trump administration is reportedly in early discussions regarding potentially putting out an executive order that would establish government's review process for AI models before public release. Basically, if you're a US company, you will not be able to release a model without a review and approval by the government. This is not a nationalization step yet. The reason I'm saying this is because AI will have an impact on more or less everything the government controls, whether it's infrastructure, national defense healthcare, education, et cetera. AI will be embedded into all of that. If the government wants to continue to control what it is controlling right now, it needs to control AI as well. And so I assume, I know nothing. I assume over time we will see more and more aspects of nationalization happening in AI around the world and the US in order to allow the government to keep doing what it's doing right now in a way that will be done with AI, because everything will be done with AI. So apparently, White House staff has already briefed leaders from Anthropic, Google, and OpenAI on these plans, with Google, Microsoft, and xAI reportedly already agreeing to allow the US government to test their models before public release Now, is that good or bad? It depends on who you ask. I believe that there needs to be an international body that will include governments, that will include industry, and that will include research from universities to figure out how to move AI forward safely for the sake of humanity. I think what the US government is doing right now doesn't make any sense. It doesn't make any sense because they're only going to stop or control what's coming from US labs. They have zero control on what's coming from the open source world and/or from the Chinese side of the world or European or any other players in the industry. And yes, they might be a little behind right now, but what does it matter? They will be either ahead or at where we are right now in just a couple of months. And so what the US government is doing right now is not gonna save the world. It may slow down the US companies. Now, again, I don't think that's the right argument. I think the right argument is to somehow, and I don't see the political situation right now being supportive of that, but somehow get to a situation just like we control nuclear weapons, or chemical weapons, we will control the development and deployment of AI across the globe. I really, really hope that's gonna happen faster sooner than later, and also take care of all the things that were mentioned by clara Shih that will take care of how the benefits are being distributed in a fair way that will really benefit humanity versus just a few individuals. By the way, this is a very interesting change of direction for the Trump administration after they were doing everything to deregulate AI, including canceling Biden's era Center for AI Standards and Innovation, and including JD Vance saying in front of the entire world, "The future of AI wouldn't be won through safety concerns, but by building." And now a few very quick notes on recent releases. OpenAI just released a Codex plugin for Chrome that can do more or less everything and maybe even more than what Claude plugin does in Chrome. So it runs natively in Chrome. It can use multiple tabs. It can test your applications. It can do research, and it can do a lot of really cool things because it understands exactly what's happening in your browser. Again, that's not new. We had this from Google, and we had this from Claude, and now we have this from Codex as well. It is new and very powerful if you are in the OpenAI universe. They are claiming that from a testing perspective, it is the best testing environment to be able to run tests on new user interface that you're developing. And they're claiming that it is way more effective because it's not built on taking screenshots and analyzing them. And that is interesting to test, and I'm going to test it in the next couple of weeks and report back to you. OpenAI also launched three new real-time voice models which are built on a reasoning model, meaning it's not just a voice communication, it actually does a lot more, and it has agentic capabilities built into the audio models themselves, which I think is really attractive and will be very, very powerful. These models are GPT Realtime 2, GPT Realtime Translate, and GPT Realtime Whisper, and their names kind of give away what they do. Realtime 2 is the full model. Realtime Translate translates in near real time to over 70 languages, And the demo that they did is freaking incredible. By the way, another entire industry that gets wiped out of people who do real-time translation of different things, such as in the UN and other places. And GPT Whisper, which does transcription from real-time voice in those languages. OpenAI also released GPT 5.5 Instant, which is their latest small and fast model that is now the default model in everything OpenAI unless you pick something else, and it is also the only model that is available on the free version. It is the first of the instant models that gets classified as a high capability in both cybersecurity and biological and chemical preparedness categories in their evaluation criteria. They still haven't defined it as critical risk, but it is the first one that gets the designation of high capability for what they call the instant smaller models By the way, independent people who checked it saying that this instant model is better or in par with the top models we had at the end of 2025 from the leading labs. So a very capable model that runs faster and cheaper than the main model. It is the similar thing we've seen from every single one of these labs previously. You release the full model, then you distill it to a smaller model that works almost as good as the main model, but at a much higher efficiency level. So another great model from OpenAI. And that is it for this week. There's a lot of other stuff that happened this week that I would like for you to know, including a cloud power AI coding tool that deleted an entire software company's database plus backups in nine seconds. AWS launching an MCP server. Etsy launching a native ChatGPT app, and many other things. But if you wanna learn more about them, go and check out our newsletter that you have the link in the show notes to do that. So open your application, And you can find the link to the newsletter in the links in the show notes. You can also join our Friday AI Hangouts, where we have a large group of amazing individuals who share what they're doing with AI, how they're learning, what they're learning, what's happened this week, and having really interesting discussions. That's every Friday at 1:00 PM Eastern Time, and you are welcome to join that. It's completely free. And as I mentioned in the beginning, you can also find links to our next course, which runs on June 22nd. That's it for this week. We'll be back on Tuesday with another amazing how-to episode that will teach you how to implement AI effectively in your business. And until then, have an amazing rest of your weekend.