Ep 784: Google’s big I/O releases, Pope’s warning on AI, Meta’s dark AI cuts and more

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. As expected, the big AI companies grabbed plenty of headlines this week, as they always do. I mean, OpenAI came out with some pretty big Codex releases. Google announced like 100 AI updates at its I/O developer conference. Meta cut 8,000 jobs in a very dystopian way, and Microsoft and consulting giant EY formed a pretty noteworthy partnership. But it maybe was actually the Pope that grabbed the biggest AI headlines this week when he said that AI could warp human life. Yeah, there's lots. If you spend time every single day trying to keep up with the advancements and developments in AI, you shouldn't. It's too much. It's too time-consuming. Let me do that work for you on our weekly AI News That Matters show. Let's get into it. Welcome to Everyday AI and what's going on. My name is Jordan Wilson, and well, this thing's for you. Everyday AI is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me keep up with AI. And I tell you what matters, what doesn't, and how to use that information to grow your company and career. So this is the unedited, unscripted livestream, but to be the smartest person in AI at your company, make sure to check out our newsletter at youreverydayai.com. Every single day we recap the highlights from that day's show as well as all of the other AI happenings that you need to know. All right. And, you know, I, hey, I agree with Jackie here saying happy Tuesday, rough getting started after a long weekend in Chicago. Yeah, it is Tuesday. Normally we do our AI News That Matters on Monday, but with the holiday here in the US, our weekly Monday AI news recap is happening Tuesday, but we will be back to Monday next week. So yeah, whenever this happens, always get, you know, notes from our international audience like, Did, did this podcast stop? No, took a, took a break, right? Got a little sun. All right, but let's get into the actual AI news. So first, even though this seems now like a year ago, but this actually happened, uh, less than a week ago. So, uh, Google unveiled not just faster AI models but new agent tools and even a redesigned Gemini app and a lot more. So Google used its I/O developer conference to announce a broad upgrade to its AI lineup, led by a faster Gemini 3.5 Flash model, a new multimodal system called Gemini Omni, and a cloud-based personal agent named Gemini Spark. So according to Google CEO Sundar Pichai, Gemini 3.5 Flash, that is Google's newest model, beats the roughly 4-month-old Gemini 3.1 Pro on nearly every benchmark, with especially strong results on GDPVal, which measures economically relevant tasks. So Google says its new 3.5 Flash is also much faster than other frontier models and is optimized for running kind of longer running tasks and also is integrated into its newly upgraded Antigravity agent platform, where it claims it to be 12 times faster, with costs roughly one-third to one-half of comparable models. So Google said the new 3.5 Gemini series was built for agentic work, meaning it can handle those long autonomous sections, sessions, and complex coding pipelines. And also they said that they can even use it while building an operating system, which is something Google did in their demos. Google also introduced Gemini Omni as its next step beyond the video. So yeah, if you use AI video or if you're into AI creativity at all, you obviously know Google's Veo video platform. So next in that family is Gemini Omni. So Google described it as a multiple, multiple input, multiple output type model. I think they kind of said any to any that they've rebuilt from the ground up to handle any outputs, including text in the future, right? They said it's gonna be able to handle text, video, drawings, anything, right? So right now it's multiple inputs, but the output for Omni is still video, but it is a little bit different than their general VO, you know, AI inputs is, or sorry, AI outputs being video. Because right now it does have a little bit more of a multimodal reasoning aspect to it. I'll probably demo that tomorrow as we go hands-on with putting AI to work on Wednesdays. Probably show you a little bit of what's new across Gemini. So a lot of people are looking at the new Omni model, right? The anything in, anything out. And they're looking at just the video output, which I think is not really what it's intended to do. I think it's, you know, ultimately going to stand apart in the future when it can output more than just video, right? But right now, that's just what people are judging it on, and they're looking at Chinese models like Seeddance and saying this new, you know, Google Omni, you know, output, video output doesn't, you know, hold its weight against those. And well, it doesn't, but I think ultimately it is going to be about much more than that in the future. So Google said that Omni can take its own generated output and feed it back in for iterative editing in the first faster version, which is called Omni Flash. And that's been live since Google I/O is already available across Google's different platforms. so, uh, two, two notes on the first two big announcements, and those were definitely the headline grabbers in Gemini 3.5 Flash and Google Omni Flash. Well, they are the Flash or the smaller variants. So Google did say at least on the Gemini 3.5 Pro side, uh, we would be expecting the 3.5 Pro in June. Uh, and we'll see about Gemini, uh, Omni, uh, whether that's Omni Pro or an Omni regular, but they did say this is a family of models. That is meant to go beyond just the, the video output. All right, that's not all. Yeah, we're not going to go over all 100. Yeah, that Google literally said they announced more than 100 things. So we'll leave that list in our newsletter today. Not going to bore you by going over all 100. But the other big one, which is not live for most people, right? In typical Google fashion, you needed, I don't know, detective skills and a couple maps to realize what was released to who, where do you find it? When does it come out? When does rollout start, stop, right? Yeah. But anyways, they did launch Gemini Spark, which is, you can think of it essentially as like a cloud version of Open Claw, but that's what Google says is their personal agent that runs on dedicated virtual machines in Google Clouds. Google Cloud stays active 24/7 and can keep working in the background even when a user's computer is closed. But that is just rolling out in beta right now only for Ultra subscribers in the US. So that is obviously a very small majority of people who are using the Google Gemini products, right? Most people are using it as part of their maybe workspace organization. You might be on that $20 a month plan, but not many people are on the expensive plan. But on that, you know, note, to go along with the new always-on Gemini Spark agent, well, Google did at least bring down the cost of their more costly plans. So their Ultra plan that was previously $250 is now $200, which brings it in parity with the other top tier plans like the $200 Cloud Max and the $200 ChatGPT Pro plan. So they did knock down the price from $250 to $200. And just as OpenAI and Anthropic have offered, there is now a $100 a month Ultra plan. So yeah, if you do want to go test out, right, if you live inside of the, the Googleverse and, you know, everything you use is Google, Google Docs, Google Sheets, Google Drive, Gmail, etc., you know, it might be worth testing out that $100 a month Ultra plan, uh, just to get access to the Gemini Spark. I'll probably do that, uh, just for a month. I had downgraded, uh, my Ultra plan recently. I looked at my costs and I'm like, oh, this is bad, right? Uh, subscribing to max plans everywhere. So I'll probably, uh, myself re-upgrade to that $100 a month Ultra plan, uh, to go over the Spark agent. Let me know if you want to see anything on Spark, right? I do think this is going to be similar to Microsoft's Copilot Cowork that they introduced about a month and a half ago, right? Which is kind of built on top of Anthropic's technology, but it's just kind of like a virtual machine that's connected to all of your data. So it is more of this OpenClaw-esque, you know, agent that runs autonomously on your computer. For me, I have all that running in Codex, so I don't have a huge appetite to go out and use necessarily something like Spark. or even something like Copilot Cowork because I get a lot of that and most of that out of Codex. All right. Yeah, there's a lot more. We're going to go over just some of the highlights from Google, uh, I/O here. There's also just new AI search everywhere, right? The AI search experience has changed completely. So aside from it now being powered by Gemini 3.5 Flash, uh, there is a new AI search box and you can now search with text, images, files, videos, and Chrome tabs all in one place with AI helping the reasoning across them. There's also the smarter and more personal search as Google is adding information agents, generative UI, and personal intelligence. So then search can keep you updated, build custom answers based on your data, and connect with apps like Gmail and Google Photos if you opt into that. And then there's also the new Anti-Gravity 2.0, which is a standalone desktop app that serves as a central hub for agent interaction. So yeah, if you've used Claude Code, Claude Cowork, uh, Codex from OpenAI. This is kind of Google's version. I did go over this on our Friday show. Not a big fan of Antigravity, uh, 2.0, their first iteration of it. I think it's more super beta, not really ready for the limelight, but that's just my thought on it. Uh, Google also did announce that they have more than 900 million monthly users, up from 400 million a year ago, while daily queries have risen sevenfold. And the company is also adding a desktop app for macOS, new agent tools for developers through Antigravity 2.0, and tighter links to search, shopping, and improved watermark detection through SynthID. So, my gosh, so much going on at Google I/O. Like I said, I'll probably dedicate tomorrow's show to going over some of the highlights since the last couple of weeks we've been going hands-on with Codex. All right. Speaking of Codex, I don't know what the OpenAI team is going to announce this Thursday as they've kind of pivoted to a weekly Thursday release cadence. And my gosh, this week's releases were absolutely bonkers. So let's just quickly go over some of what's new inside of OpenAI's Codex. And if you did miss it, the last two Wednesdays, we've given the beginner's guide to Codex. And if you don't know what Codex is, Essentially think of it as ChatGPT for work, right? That's really what I think they're ultimately going to be branding it as. You know, it's a desktop application. But here is what's new in Codex, and it is the fastest growing kind of AI segment in the world right now, at least in terms of new Codex users. So they introduced App Shots that's available in the Codex app on macOS that lets users press both Command keys to send the frontmost app window to Codex with a screenshot and available text. This one is huge, especially for someone like me. I am screenshotting constantly and copying and pasting in, uh, content into Codex, and this does both. So it's more than just a screenshot. It automatically brings whatever screen you're working on into Codex as well as any available text. So it is like a two-for-one bringing all that content in there. Uh, also, Goal Mode is now fully available and this is big. So this is the first time we've seen a true goal mode in a kind of user-friendly desktop interface. Both Codex and Cloud Code have integrated this into the CLI or command line interface tools that are more for developers. But this is the first goal mode that's kind of available for the everyday knowledge worker from one of the big players. So goal mode essentially It just means you can give Codex a goal as, as long as it's well-informed, it is going to work on tasks for hours or maybe even days. You're going to want to make sure that you use this kind of— unless you're on the $200 plan like I am, you're going to want to use this wisely and make sure you're monitoring it because, yeah, whether you're doing something goal mode in Claude Code or Codex or anywhere, it can go through your credits fairly quickly, especially if you're only on that $20 a month plan in Codex. And yes, all you need to do for Codex is just use your ChatGPT plan. It's the same thing. So yeah, goal mode, absolutely incredible. If you remember things like Ralph Loops or, you know, kind of, you know, think of something like a heartbeat, you know, file in kind of the Open Clause of the world. It's kind of like this. You just give it an open-ended goal and it's going to continue to work. It's just essentially not going to stop until that goal is accomplished, which is really cool. Next, there is remote computer use. So this is different than the essentially mobile app or ChatGPT remote controller that you can do now by controlling your computer in by using your mobile device. This is actually when your computer is locked. So this is big. This is no more having to carry around a, you know, half a jar laptop, which is what I was doing back in February. So now you can actually control a locked computer. Inside Codex. So you might be saying like, okay, why? I don't understand this, right? So for me, as an example, I'm on my Mac Studio right now. My new MacBook Pro is closed, but I have certain projects that I've been running locally on that Codex instance. So if I wanted to, I could essentially control my computer even though it's closed and locked, which opens up some new capabilities. And I think it does kind of turn this, you know, kind of local machine aspect from all these autonomous AI instances to almost like cloud, right? So I think the, the lines right between local, you know, quote unquote on-prem machines and what you can access via anywhere is starting to blur. Yeah, there's a lot more. Also, there's plugin sharing, which is huge. I think this is one of the bigger enterprise, you know, roadblocks that a lot of big companies have run into. But you can now share codecs plugins across organizations. Yeah. And there was a whole lot more. But those are just some of the big updates from Codex. And make sure whether you want to know more about those or any of the other big AI features this week, make sure you go back and listen to episode 783 from Friday. That is our Friday Features show where we go over the big AI features of the week. All right, this one now feels like it happened like 3 months ago. It didn't, right? So normally we do our AI News That Matter shows on Monday, and this happened just hours after last week's show, but it is still one of the bigger AI news stories of the week. So have to tell you about the update in the Elon Musk versus OpenAI trial. So a federal jury in Oakland found that Elon Musk waited too long to sue Sam Altman, Greg Brockman, and OpenAI, ending the high-profile case that could have affected the future of AI. So essentially, Elon Musk's long battle against OpenAI is over for now as he officially lost this case. More on that here in a second. So a unanimous jury ruled that Musk's claims were filed after the legal deadlines had already passed. So the court rejected Musk's claim against Altman, Brockman, OpenAI, and Microsoft on the grounds of statute of limitations without reaching the core merits of the dispute. So Musk's team said it plans to appeal, arguing the judge and jury decided the case on a calendar technicality rather than on the underlying allegations. So the case centered on claims that Altman and Brockman unlawfully enriched themselves and breached duties owed to OpenAI when the company moved toward a for-profit structure. So the jury deliberated for less than 2 hours. Yeah. Uh, after a 3-week trial that drew major attention across the tech world. So time limits were a major issue in the case, with deadlines of 3 years for the charitable trust claim and 2 years for the unlawful enrichment claim. Obviously, Musk very much missed those time limits, as OpenAI co-founders had discussed the for-profit shift as early as 2017, and OpenAI created a for-profit arm in 2019. But Musk did not sue until 2024, therefore completely missing, uh, kind of the statute of limitations that were the crux of Musk's claims. So Musk claimed that he waited so long because he believed Altman's, uh, reassurances, but he later said he became alarmed after Microsoft invested $10 billion in OpenAI's for-profit arm in 2023. Microsoft called the verdict a confirmation that the claims were untimely, while OpenAI said the ruling preserves the company's current structure and investor backing. So we'll see what happens here with the, you know, going through appeals from Musk's team. I said this back in 2024, and I stick to this. This is theater. This is nothing but theater. This is my opinion on this, right? This is theater from Musk's side. If you look at the facts, there hasn't been an actual case, right? I'm no legal expert, but I can read. I said this back in 2024 when the lawsuit was first filed. I said, this is sour grapes, this is technicalities, and this is theater, nothing else. I continue to feel that way, and I will continue to feel that way even as more facts get unveiled in the case. And ultimately, the reason for this theater, people are like, well, what's the, what's the big deal? Well, hey, both companies are going public. So, uh, SpaceX, uh, or SpaceX AI or XAI SpaceX, whatever, uh, the company is called nowadays, is going public in about a month with a reported $2 trillion plus valuation, uh, despite low revenues. Uh, and OpenAI and Anthropic are reportedly going, uh, public this year as well. So I think if nothing else, uh, Elon Musk is using this for leverage on his IPO, to, you know, kind of create a driving storyline and to keep his company in the headlines. That is my thought. But pretty big loss, at least for Elon Musk in this first round. All right. Next, there was supposed to be a noteworthy AI order from the White House, but it hit an unexpected yet kind of quiet roadblock that a not, not a lot of people are talking about. And this one, at least when it comes to AI's impact on your business, could have been and still might be pretty big. So the White House, like I said, had planned a huge executive order on AI that was supposed to be released on Friday. However, that was paused after venture capitalist and Trump adviser David Sachs objected directly to President to Trump just before the rollout, according to reports. So this reversal matters because the order was meant to shape how the government handles frontier AI models, which could affect how fast major companies can release new tools and how much oversight they face. So according to reports, a senior White House official said Sachs had, had earlier taken part in a review of the executive AI order that was supposed to be released and believed it to be too broadly supportive before changing his stance on Wednesday night. So the official said Sachs raised concerns that a voluntary agreement could eventually turn into a mandatory requirement and then called Trump Thursday morning without even telling most of his own staff. That call reportedly derailed the plan and blindsided White House staff, who had been working on the release. So, yeah, reportedly, you know, all of the big AI CEOs were supposed to be at the White House for this unveiling on Friday, and then it was derailed. I think we got word of this late Thursday night. So 4 people familiar with the matter said industry leaders also pushed back on a proposed voluntary review process for cutting-edge AI systems. So companies had wanted the White House to shorten the window for sharing new models from 90 days to 14 days, showing how quickly the tech industry wants clear rules and faster approval paths. So one of the key, you know, marketed benefits of this, well, was the White House was going to, you know, essentially approve or inspect these models 90 days before release to ensure their safety, which at the time I'm like, 90 days? Like, the AI companies aren't working that far in advance. Right? I've been told both on the record and off the record that, you know, big AI labs aren't working 90 days out. So when I saw this, I'm like, I don't think this is going to work even if it's voluntary, mainly because companies don't even know their roadmaps 90 days out because they're, you know, realizing whether it's through, you know, recursive self-improvement or, you know, new fine-tuning techniques, new technologies, right? They don't know always what's going to be happening 90 days out. So, uh, now we're seeing that companies were kind of pushing back. And ultimately, it seems like this was David Sachs calling President Trump and wanting a short— shorter window. So we'll see if there is an updated kind of amendment to this agreement or if it's kind of dead on arrival. So a readout, though, of the draft was kind of leaked and shared. And it said that the National Security Agency, or NSA, would have final say on which systems counted as covered frontier models, with other agencies helping build a classified benchmarking process within 60 days. So the draft, again, that was kind of scrapped for now, also stressed that the review process would remain voluntary and would not create mandatory licensing or preclearance, a detail that matters. For companies racing to build and sell advanced AI products. So, you know, I think some of this smoke kind of started with Claude Mythos, right? Not fully releasing it to the public, but reportedly they were working with the Trump White House on it. So, you know, as these models become more and more capable, more and more autonomous, there has been this kind of gray area of like, okay, what happens when and if a model is fully released to the public? and companies are not ready for it, right? Because one of the, um, obvious use cases for these, one of the ill-advised or one of the bad use cases is for cybersecurity, right? In the bad way, cyberattacks, right? So when these, uh, when these, uh, models, and we're seeing this now, when their cyber capabilities are better than, well, any humans, that can mean very bad things when released to the general public, right? In theory, If a certain, you know, antiquated banking system or something like that has security loopholes, hackers could use publicly available off-the-shelf models to bring down entire systems. It's obviously a key concern, which is why some of these original talks about the, the White House kind of, quote unquote, clearing AI models before their release. This is why some of the talks even started. But it does seem like this kind of last-minute phone call, kind of at least for now skirted some of these plans. I do think that we'll see an updated version of this cleared. Maybe one of the things is being that 90-day, you know, limit down to 2 weeks or down to 30 days. We'll see. But we'll continue to cover it as it comes up. AI moves too fast to follow, but you're expected to keep up. Otherwise, your career or company might lag behind while AI-native competitors leap ahead. But you don't have 10 hours a day to understand it all. That's what I do for you. But after 700+ episodes of Everyday AI, the most common questions I get is, where do I start? That's why we created the Start Here series, an ongoing podcast series of more than a dozen episodes you can listen to in order. It covers the AI basics for beginners and sharpens the skills of AI champions pushing their companies forward. In the ongoing series, we explain complex trends in simple language that you can turn into action. There's 3 ways to jump in. Number 1, go scroll back to the first one in episode 691. Number 2, tap the link in your show notes at any time for the Start Here series, or you can just go to starthereseries.com, which also gives you free access to our inner circle community where you can connect with other business leaders doing the the same. The Start Here series will slow down the pace of AI so you can get ahead. All right, here's nothing dystopian about this. Yeah, it is. So, uh, the Meta AI layoffs have finally happened. So according to Reuters, Meta has cut about 8,000 US employees, or roughly 10% of its workforce, as the company shifts more money toward AI infrastructure and cost control. So reports said the layoffs come as Meta plans about $145 billion in spending this year on data centers and hardware chips for its AI push. The company has also reportedly canceled about 6,000 open roles and is simplifying some job functions into a broader AI builder title, signaling a major internal restructuring at the company. An additional 7,000 employees have reportedly been moved into AI-focused work, showing how aggressively Meta is redirecting staff toward its machine learning strategy. So yeah, not only was there reported 8,000 employees who just lost their jobs late last week, but a lot— thousands of internal jobs that were not around AI are now AI roles. But that is not the part— as sad or as strange as that sounds, that's not even the part that's grabbing the most headlines. It's actually kind of the method that Meta is taking in the interim to get to this. Because according to reports, Meta is facing a ton of backlash over a new internal computer monitoring program that logs employees' mouse movements clicks, dropdown menus, and keystroke shortcuts on company laptops. So the stated purpose of all of the tracking is to collect training data so AI systems can learn how employees complete routine digital tasks, according to internal messaging described in the reports. So employees have been told there is no opt-out on the company-issued devices, which has intensified concerns about privacy and workspace surveillance. So The New York Times and Wired report that more than 1,500 employees signed a petition against the monitoring, while some workers have posted anti-surveillance flyers in Meta offices. So this is not the first and this is not the last that we will hear of this type of technology being used before mass AI layoffs. This also isn't the first time that we're going to see a large enterprise company essentially start to take all non-AI roles and to tell workers who have nothing to do with working in or around AI that, oh, hey, now if you want to keep your job, you work in and around AI now. So, yes, this has been described as dystopian by a lot of people. And I would kind of agree. Right. Because essentially for the last couple of months, Meta employees have known, right? They've known that these cuts were coming. They've been reportedly, uh, I, I wouldn't say they've been using this software. Uh, it's more like they've been subjected to this, uh, kind of surveillance-esque software that monitors literally every keystroke, every mouse movement. So, uh, imagine, right? And this is why you've seen, uh, kind of reports recently of the morale at Meta at a literally industry low, uh, and how people just don't want to work there anymore, uh, because of this. So I'm guessing what Meta is doing, they're gonna take a PR hit in the short term, uh, in hopes that they become more and more profitable in the long term. Uh, their model that came out, Muse Spark, was actually okay, right? It actually did fairly well on benchmarks, and they did say that they're gonna have a bigger version. So, uh, I'm guessing that Meta is going to be one of the first big companies that comes out and says, yeah, we don't really care anymore. We are restructuring the entire company AI first. We're going to take some, uh, some PR black eyes. We're going to be monitoring literally every single thing you do because employees have essentially said that they felt like they've been training their AI replacements for months with these systems learning everything they do and then jobs just disappearing. Uh, so, uh, kind of, yeah, kind of a weird one to even talk about and think about, but, uh, unfortunately, yeah, I think this might be the future, uh, of work at some big enterprise companies where this just becomes the status quo, right? Talk about dark and gloomy potential days ahead. Not saying this is the good way, but I do think a lot of companies are gonna go this way. I would love to think that there's, uh, huge enterprise companies that are gonna find the more optimistic way, uh, in the more upbeat way that, you know, prioritizes, uh, human intelligence. But in the short term, I mean, we'll see how Meta's stock, you know, goes over the past couple of weeks or over the next couple of weeks, because I think that will, for better or worse, potentially serve as a blueprint for others in the tech industry. All right. Another recent trend. This one maybe a little better than others. You know, yeah, companies, big tech companies now, you know, those that build AI models are now becoming consultants. Uh, so according to reports, EY and Microsoft have expanded their partnership with a commitment of more than $1 billion over 5 years to help organizations scale AI across the enterprise. So yeah, this new FDE role we've been hearing about a lot recently, the Forward Deployed Engineers. So that's what Microsoft is rolling out here in conjunction with EY. So their FDEs will work directly with consulting giant EY's industry teams using Microsoft Microsoft's AI-native Hyper-Velocity Engineering model to speed up implementation inside client change and delivery structures. So EY said the expansion is an important milestone in its longstanding alliance with Microsoft across tax assurance consulting and EY Parthenon. Uh, the companies say they want to help clients become what they call frontier firms, meaning organizations that use AI at scale, train workers for new tools, and keep improving AI-driven processes over time. So this is not new, and Microsoft is technically the last of the Big Four to officially make this pivot into becoming AI consulting firms within the clients that use their products. That's because in a 6-week stretch, OpenAI, Anthropic, and Google have all made the same move. Uh, while they're spending maybe a little less of their focus on just selling models, and instead they're starting to deploy their people inside of Fortune 500 companies to better use their models. Yeah. Addressing the capability gap that I've been talking about for now years. So OpenAI launched a $4 billion endeavor called the Deployment Company or DeployCo on May 11th. So about 2 weeks ago with 19 different private equity and consulting partners like TBG, Bain, McKinsey, Capgemini, etc. And they're embedding FDEs or those forward deploy engineers directly inside of client workflows. Same thing. Google Anthropic also followed right after that with a $1.5 billion joint venture alongside Blackstone, Goldman Sachs, etc., targeting mid-market companies through private equity portfolios. And then Google took the platform route, committing $750 million at their Cloud Next conference to 120,000 partners such as Accenture, Deloitte, PwC to build Gemini agents. All right. And our last big AI story of the week, like I said, I think the one that is going to be talked about most this week, maybe it's not the Meta layoffs, maybe it's not Google I/O, maybe it's not the White House delayed AI, uh, project there. I think you're going to be hearing a lot about Pope Leo, his recent warning on how AI could warp human life. So Pope Leo has issued his first encyclical called Magnificat Humanitas, a 43,000-word warning that artificial intelligence could reshape society in dangerous ways if it is not guided by human values. So the Vatican released the document last week and Leo signed it as well. Exactly 135 years after the previous Pope Leo XIII's, I believe, Rerum Novarum, which was the landmark Catholic teaching on the industrial era labor and social justice movement. So Leo says AI is useful but is not neutral because it reflects the priorities of the people and institutions institutions that build, fund, and train and deploy it. So Pope Leo warned that AI could weaken human judgment by giving people instant answers that reduce creativity, discernment, and the patience needed to search for the truth. He also says AI can imitate care without real relationship, which could leave vulnerable people to mistake chatbots or digital assistants for genuine human connection. Another concern is inequality, since computing power, data, and influence are concentrated in the hands of a small number of companies and actors. Pope Leo says that AI could also distort democracy by accelerating disinformation and making it harder to tell fact from fiction, which matters for voters, workers, and businesses trying to make informed decisions. So he warned that AI could make war easier by speeding up lethal decisions and distancing people from responsibility. And he says that no algorithm can make war morally acceptable. Experts say the documents could matter well beyond Catholic circles, especially as governments debate AI rules for hiring, credit, public services, and safety. So, yeah, if you didn't read the 43,000 word, encyclical. I didn't yet either, although I probably will at some point here this week. But I think it's an important document to keep an eye on, even if you are not Catholic. I think this is going to really set a— set a tone. I'll just say this, right? Especially coming up before US elections here in a couple of months. I do think as AI models, right, as the, you know, super strong and super capable models, right, like we talked about cyber concerns, I think with disinformation and misinformation and AI being actively used as a tool to sow division, I think is an important topic to keep an eye on. So whether you're thinking about how you consume information on a day-to-day basis, what you believe, what you don't, how you're making decisions in your company. Just know if you haven't come to this realization already, you should probably assume, right, that most of all information you consume is either AI-generated or influenced by AI. So you really just have to start, you know, thinking more with your brain, right? It's something I'm trying to do more and more is, yes, I consume so much you know, content and outlines and research that is done by AI. But, I always try and be vigilant about thinking with my own brain, right? And ultimately being the one who makes the decisions. So, all right, that is it for the big AI news stories for this week. But let's quickly go over the bullet points of what's new and what's next. So some of these smaller stories, kind of our roundup, some rumors, some leaks, et cetera. So OpenAI says their new internal, one of their internal models that they've not yet released has solved an 80-year-old math problem. Cursor released Composer 2.5, offering near Opus 4-7 levels at a fraction of the cost. Anthropic is reportedly going to release a preview of Mythos-1 for Claude Code and Claude Security. So yeah, we might be closer to getting the mysterious mythos out in the wild. OpenAI began selling multi-year capacity contracts to lock in AI compute demand from big enterprise companies. OpenAI co-founder Andrej Karpathy jumped to Anthropic as he announced that he joined Anthropic. Right. I just like when he tweeted that, I don't know why I had like the same vibes as when LeBron James said he was taking his talents to South Beach. Right. Played for Miami Heat. I don't know why that's what I thought of. Reports say that Deepseek is nearing its final stretch of a $10 billion fundraising round. Uh, Hark raised a $700 million round for its still-secret AI interface. Something to keep an eye on, right? Uh, kind of led by Figure AI's Brett Adcock. A new report says OpenAI may file their IPO or their initial their initial public offering paperwork in coming weeks, but yet may still hold off on their actual IPO, waiting for the timing. OpenAI and Dell team up for a Codex partnership to help enterprises deploy it on-premise. Intuit plan to lay off over 3,000 employees while refocusing on AI. ClickUp did something similar with a big percentage of their employees. OpenAI launched ChatGPT for PowerPoint in beta. It's actually really good. Blackstone and Google launched $25 billion AI data center push in the US. Anthropic announced that users under 18 cannot use their AI services. Deepseek is making its version of Codex with a desktop-first AI agent, right? Yeah, the, the flock to local desktop versions continues. Anthropic released Claude Managed Agents Beta. With self-hosted sandboxes in private MCP tunnels. The Pentagon is reportedly testing AI models that rival Anthropic after cutting ties with Anthropic's Claude. Cohere open-sourced their model Command A+, a sovereign mixture of experts model for on-prem multimodal work. OpenAI used— oh, already talked about that. Spotify and UMG, one of the major music groups, announced announced paid AI covers and remixes for premium users. I'm actually really looking forward to that one. Uh, the NSA released security guidance for AI, AI-driven autonomous systems. Amazon turned Alexa Plus into an AI podcast maker for on-demand conversations. We covered that one in our Friday Features show as well if you want to know more. xAI expanded access to Grok Build, a genetic terminal coding assistant, to all X Premium Plus subscribers. I don't even understand all of the X tiers. It seems like there's 20 of them, but their version of, you know, Claude Code, Claude Coworker, Codex is now rolling out to more people. President Donald Trump posted AI-generated images of shackled immigrants, drawing immediate backlash. Stability AI launched Stable Audio 3. For generating music with AI. Google introduced shopping cart updates tied to its AI commerce push, and Anthropic, uh, acquired Stainless to improve developer tools and API integrations. All right, a lot happening this week in AI. As always, I hope this recap was helpful. If it was, remember, we normally do this every single Monday, bringing you the AI news that matters. On Wednesdays, we go hands-on with demos at our AI at Work on Wednesdays. On Fridays, we do Friday Features. That's really just focusing on just new AI features that you can use that are available now. So a little bit different than our Monday News Recap. And then Tuesdays and Thursdays, we usually bring a variety of different shows. So if you're new here, thank you for tuning in. Please take a minute to subscribe to the podcast if you haven't already. then make sure to go to youreverydayai.com to sign up for the free daily newsletter to become the smartest person in AI in your company. Thanks for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'all. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers, and we'll see you next time!