Everyday AI Podcast – An AI and ChatGPT Podcast - EP 571: Google’s AI makes phone calls for you, ChatGPT Agents and more AI News That Matters

Starting point is 00:00:00 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. Meet Firefly AI Assistant, now live and Adobe Firefly, the all-in-one creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. Google just rolled out an AI that will call local businesses for us so we don't have to.

Starting point is 00:00:53 U.S. President Trump has some pretty intense AI plans. There's a bunch of Nvidia GPU drama and chat GPT agents are here and they're pretty impressive so far. As always, another extremely busy and meaningful. meaningful week in AI news. So don't spend hours every single day trying to keep up and wondering what it means for your career or your company. Instead, spend every Monday with us here on Everyday AI as we bring you the AI news that matters. All right. I'm excited to get into it. Hope you are too. What's going on to you? My name is Jordan Moulson and I'm the host of Everyday AI. This is your daily live stream podcast and free daily newsletter helping everyday business leaders

Starting point is 00:01:39 like you and me, not just keep up with all these AI developments, but how we can leverage this information to grow our companies and our careers. So it starts here with the daily, unedited, unscripted, live streaming podcast. But where you really get the edge is by going to our website at Your EverydayAI.com. So there, you can sign up for our free daily newsletter. We're going to be recapping as we do every single Monday through Friday. The highlights from that day's podcast as well as keeping you update with everything else that you need to know in the world of AI. And it's literally a free generative AI university. You can go sort every single episode we've ever done by category.

Starting point is 00:02:16 Learn from the world's leading experts all for free. So make sure you go check that out. All right. Let's get into it. Let's go over the AI news for this week, July 21st. So first, meta, even though that they've been a leader in open source AI, I might start going closed source, according to a new report. So, Meta is weighing some major shifts toward closed AI models amid some leadership change.

Starting point is 00:02:51 So Meta's new superintelligence lab is considering abandoning its leading open source AI model, behemoth, which has not come out yet, to focus instead on developing a closed model, according to reporting from the New York Times. So the internal discussions were led by new, leader there at MSL, Alexander Wang, the former Scale AI founder and CEO, and Meta's newly appointed chief AI officer. So he recently joined Meta after Meta invested $14.3 billion in his old company, Scale AI.

Starting point is 00:03:27 So for years, like I said, Meta has championed open sourcing its AI models, arguing that public development speeds up innovation and broadens access for developers. But the conversations are apparently going in a different direction. A switch to closed models would mark a pretty dramatic shift in Meta's AI strategy, potentially limiting outside access and collaboration on its most advanced AI technology. This move could also reshape the AI landscape as Meta is one of the few major tech companies to have consistently released its biggest AI models for public use. This comes right on the heels of Meta CEO,

Starting point is 00:04:05 Mark Zuckerberg announcing that meta would spend hundreds of billions of dollars in compute a move to lead in AI and attract top researchers to the company. Also related here and kind of ironic, Open AI is prepping any day or any week now to release an Openweight's version of its model. We've heard reports that it's going to be kind of like an 03 mini level. So yeah, it's pretty interesting turn of events here. Again, this is just reporting, but it's from the New York Times. So you have to take it pretty seriously. I wouldn't be shocked, though. I've always kind of quietly opined and not as much on the show, which by the way, I've been wanting to. You know, I always put out things in our newsletter like, hey, what do you want to, you know, what do you want to hear about? And a lot of people don't want to

Starting point is 00:04:58 hear about meta. So I've actually had some thoughts like I never really thought meta for the long run would be able to competitively play in the open source only. Right. Right now, they're essentially fueling their AI side, right? Because they're like, oh, well, how can meta invest, you know, hundreds of billions of dollars into compute and, you know, pay these researchers $200 million sometimes apiece, you know, if they have a free and open source model. Well, there's some, some enterprise agreements for larger customers.

Starting point is 00:05:34 But obviously they have their social network, you know, conglomerate with Facebook, Instagram, WhatsApp, et cetera. So they are essentially taking a lot of that money and just investing it on the AI side. But at a certain point, it's going to run out. Right. So you can't just be open source forever. I would assume that if this does come to fruition, that meta would still keep some closed, some closed models and some open models.

Starting point is 00:06:06 I would assume they're quote-unquote transformer models. Their non-reasoning models would probably remain open or open-esque. And then they're probably reasoning models and their largest and most powerful would be closed. And I actually think that's not a bad game plan for a lot of companies to follow, if I'm being honest. It's kind of what on a much smaller scale, what Google's doing. They're doing it with their many, many models, their Gemma models, but their Gemma models are open source and their Gemma models are really, really good. So we'll see what happens with this report, but it's pretty interesting nonetheless. And yes, finally for one week, we didn't talk about meta poaching a bunch of people, even though they did punch a bunch of more researchers.

Starting point is 00:06:55 It's not going to be one of our top stories. All right. Moving on. So U.S. President Donald Trump is. expected to unveil policy guidelines for AI that call for easing regulations and expanding energy sources for data centers. So he is supposed to have a keynote at an event this week on Wednesday, I believe, and it's part of the AI action plan that he'll be talking about.

Starting point is 00:07:23 And that's going to be released this week. And it will include executive orders and a nationwide promotional campaign led by the White House Office of Science and Technology Policy. So the plan urges Congress to consider federal legislation that would preempt state-level oversight of AI aiming for a more unified national approach that obviously didn't work earlier because that was kind of bundled in as pork to a spending bill. But the U.S. Senate voted that down 99 to 1. That was kind of Trump's provision that he wanted. essentially no one else really wanted it.

Starting point is 00:08:01 So essentially all the US senators said, no, we're going to allow states to still legislate AI as they see fit. But pretty spicy take here coming in from President Trump here talking about, yeah, he's still going to push this national approach, which, you know, I had plenty of takes on. This is a news portion, you know, but it's interesting that even after that was pretty much unanimously, almost unanimously shot down, that he's still trying to push this more national approach on some instances. It makes sense. In reality, if you live here in the U.S., you can probably understand how it wouldn't make a lot of sense, right? A state like California

Starting point is 00:08:46 has much different needs than a state like Ohio or West Virginia or Mississippi. So I don't really necessarily agree with the national approach, although when you look at it at a global scale, it would help. But in terms of the people living here, you know, keeping an equitable access to the most powerful technology of our time, I don't know if federal level would work necessarily. So the proposal would focus on executive branch actions and messaging rather than sweeping long-term strategy. Industry leaders like OpenAI, Meta, and Google, have pushed for fewer regulations compared to the previous administration. And this new plan aligns with their interest by promising reduced regulatory barriers and support for innovation.

Starting point is 00:09:36 All right. Let's talk AI startups that don't have a product, but have already raised billions of dollars. Yeah, as weird as it sounds, there's quite a few of those new AI startups that don't really have a product or service. Yeah, and have raised billions of dollars. But the latest is Thinking Machines Lab. An AI startup launched by former OpenAI researchers and leaders, including Mira Murati. And they've raised a record-breaking $2 billion seed round. This is their seed round.

Starting point is 00:10:14 $2 billion. So right now that funding values the company at $12 billion, making it one of the highest-valued early-stage AI companies ever. So the founders, led by Maradi left Open AI to build new AI technology, signaling a significant shift in the competitive landscape of AI research. The company's emergence from stealth mode and massive funding reflect intense investor interest in next generation AI platforms and tools. The news comes amid a broader wave of talent movement with Open AI,

Starting point is 00:10:48 meta, and Google all competing fiercely to attract top AI researchers through lucrative offers and strategic hires. So this new fundraising round from Thinking Machines Labs and high valuations suggests that investors believe thinking machine labs could be a major player in the AI industry, potentially rivaling already established giants. Speaking of Giants, Google adding some pretty cool features and functionality to their traditional search. So Google has launched its Gemini 2.5 Pro AI model inside their Google searches AI mode,

Starting point is 00:11:31 along with a new deep search capability and an AI powered feature for calling local businesses. That's going to be wild. And if you haven't used AI mode yet in search, you know, kind of leadership at Google has pretty much said this could be the default search in the future. it's really good. I'll probably at some point do just a live stream podcast on the new AI mode and the new improved AI mode, which I don't have yet, even though I'm subscribed to the altar plan. It hasn't rolled out, but it's a slow rollout. But let's talk about what's new in this push.

Starting point is 00:12:10 Well, first, the updates will at first only be available to US subscribers. And if you have the Google AI Pro, which is that $20 a month plan, or, their AI Ultra, which I think that one, that one's a little price here. I think that's 250. So pretty big step here from Google bringing their most powerful Gemini 2.5 Pro model. So by bringing it to AI mode, it's designed to handle advanced reasoning, math, and coding questions, providing users with more comprehensive answers and links for further learning. Also, now there's deep search powered by Gemini 2.5 Pro, and it can issue,

Starting point is 00:12:51 hundreds of searches synthesize information from multiple sources and generate fully cited reports in minutes, making it especially useful for in-depth research or for major life decisions that you would normally be doing a ton of Google rabbit-hulling for. Here's the cool thing, though. There's a new, agentic local calling feature, right? It's only for certain industries, but you know on Google Maps and, you know, it'll show their direction, their hours, all these things. But therefore, for those paid users of Google, if you're logged into your account when you're doing the search, there's a new agentic local calling feature that allows users to instead hand off that phone call to Google's AI to call local businesses to check prices or service availability.

Starting point is 00:13:37 Then it aggregates the responses to present a range of options for the user. Yeah, I can't wait to have access to that. Like what it first came out, the services were a little limited. You know, there's things like, you know, hair care, automotive, right? So it was only certain, you know, service-based industries at first. We'll see if they expand that. But I'm going to find a reason to care about one of those service-based industries, even if there's not a lot.

Starting point is 00:14:06 Or maybe I won't be as sad when, I don't know, you know, a pipe in my house breaks because I'll be able to use this, you know, and be able to have the Google AI agent call, like, you know, I don't know, 15 plumbers and, you know, get quotes in a availability right away. It's like, let's be honest, it's things that the humans don't really like doing. And it's probably also something that the, uh, the companies don't like doing as well. So this move signals Google's investment in agenic AI capabilities where the systems can take actions on behalf of the users, potentially reshaping how people interact with search and local businesses. So while these features are currently limited to paying. subscribers, Google is expected to expand access in the future, which could have significant implications

Starting point is 00:14:56 for businesses relying on local search visibility. All right. Speaking of surge, OpenAI will begin taking a cut from merchants with their new updated chat GPT shopping, looking to move beyond its current reliance on just premium subscriptions for revenue. So instead, they will get a cut of purchases made that start, inside of chat chvety. So this new e-commerce feature is still under development,

Starting point is 00:15:26 but early versions are being shown to brands with financial terms under discussion. The company's partnership with Spotify, as an example, was announced in April, and it allows for integrated checkout technology inside of chat chvety, and that enables those transactions to be completed without even leaving the chat chfifty platform. So currently, chat chaptaintee T displays products, product recommendations with links to retailers, but the planned update will keep users in app for the entire of their entirety of their purchases.

Starting point is 00:15:59 So merchants fulfilling orders through chat GPD will pay a commission to open AI. And it's anywhere, it's been floated anywhere from around like 2%, something like that, you know, found inside, you know, the normal chat GVT or via their chat GPD deep research tool. What will be interesting too is if the new chat GBT agent will also have. a 2% cut. You know, that could be interesting, right? Especially if you kind of hand the keys over to chat ChpT agent, yeah, could get, could get spicy there.

Starting point is 00:16:33 So the rise of AIO, which is AI optimization. So we've all heard about SEO search engine optimization, but the rise of AIO is prompting brands and agencies to rethink how they get products noticed by AI models, much like It's been doing for SEO search engines for decades. OpenAI has previously stated that it has no active plans to pursue advertising, but it is now openly exploring new revenue streams, including advertising and affiliate fees according to recent comments from company executives. All right.

Starting point is 00:17:09 Let's take a very quick break for a word from our partners. This podcast is supported by Google. Hi folks, Paige Bailey here from the Google DeepMind Devril team. For our developers out there, we know there's a constant tradeoff between model intelligence, speed, and cost. Gemini 2.5 Flash aims right at that challenge. It's got the speed you expect from Flash, but with upgraded reasoning power. And crucially, we've added controls, like setting thinking budgets, so you can decide how much reasoning to apply, optimizing for latency and costs. So try out Gemini2.5 Flash at AIS Studio.gov.com and let us know what you build.

Starting point is 00:17:44 All right. So let's all learn together. Well, that's what essentially all the AI labs are doing and all kind of at once. So just in the past week or so, there's been a wave of student focused learning tools that are emerging from leading AI labs. So just in the past week, we've seen kind of leaks and reports as well that Anthropic Open AI and Google, are all developing and beta testing features to turn their chat bots, respectively Claude, ChatGBTGBT and Gemini into more interactive, organized study assistance for students. So Anthropics is called study projects for Claude. OpenAIs is called Study Together for ChatGPT. And Google's is called guided learning for Gemini. And all three aim to help users build structured learning paths,

Starting point is 00:18:44 visualize complex concepts and generate detailed study guides. So these new tools will focus on guiding students through coursework and adapting tutoring based on personal goals and materials, not just answering questions on demand. So early versions keep the existing chat interfaces but introduce new workflows and responses tailored for ongoing learning and collaboration. Pretty interesting here that we have these like simultaneous rollouts. kind of shows that there's kind of a mounting pressure for AI companies to capture the growing student demographic, kind of like how Google gave away like a year plus of its paid Gemini

Starting point is 00:19:28 product to students. And we talked about on the show last week, kind of the new initiative, you know, supported by OpenAI, Microsoft and Anthropic to train more than 400,000 teachers in the U.S. different large language models. So a really big push here. Live stream audience, what do you think? My thought is like, I hope that this isn't just for people with an EDU account. I would love to have this feature because I, like one thing I use large language models for a lot is for learning. So I have, you know, GPTs and jams and projects that take longer form content from a large language model and essentially do this, but not as good as a job.

Starting point is 00:20:12 You know, from what I've seen, there's some nice, you know, user interface user experience to it. So I hope this is something that's rolled out to all users, not just students, but, you know, I don't know. I guess if they get it, no one else good on that. Hopefully they use it. You know, you dumb kids start using it. All right.

Starting point is 00:20:30 Our next piece of AI news. Anthropic has announced a new financial analyst solution aimed at helping financial services. financial services firms streamline and verify their data analysis. So this new solution unifies diverse financial data sources, including market feeds and internal databases from platforms like Databricks and Snowflakes into a single interface. So data verification is built in with direct hyperlinks to original sources, allowing users to cross-check information instantly and hopefully reduce errors.

Starting point is 00:21:10 Anthropic claims the tool, enables complex financial analysis, such as Monte Carlo simulations and risk modeling in minutes rather than hours. So the solution runs on a domain-specific version of Anthropics' latest Claude models, including Claude Opus 4, Claude for Enterprise, and Claude Code. Key partnerships with data providers like DeLupa, Facts that Morning Star, Palantir, Pitchbook, and S&P Global ensure access to comprehensive financial financial. financial data sets. Implementation support comes right now from major consulting firms like Deloitte, KPMG, and PWC. The product is available now on the AWS marketplace with Google Cloud

Starting point is 00:21:55 Marketplace availability coming soon. So you might be wondering like, all right, Jordan, this seems like a pretty niche story to include. And FYI, y'all, I spend all week, like I always have a running list because obviously we cover the AI news every single day. So I usually have like 30 stories for consideration and we only really do nine or 10. So you might be confused. Okay, this is a pretty niche story. Why are you bringing this up? Because I think this is actually part of a bigger trend.

Starting point is 00:22:26 And I've been talking about this for a long time, domain specific models. So this is one of the first times outside of health that we've seen Anthropics, specifically, you know, come up with a domain specific models. And I think that's going to be a pretty big trend in 2026, right? We've seen it a lot with health. You know, Google has a lot of health models. But I think this is, and then you've had, you know, all the big AI labs obviously work, you know, on the financial side before. But that's usually one-off partnerships, right, in helping big financial firms essentially spin off their own version of those models.

Starting point is 00:23:07 But those are essentially fine tunes of a main model. So the difference here, a domain-specific model, that's the future, right? Domain-specific models, small language models, model routers, I think are going to be way more common than these, you know, multiple trillion parameter giant, you know, general use case models that we all rely on right now. You know, I've always said, and, you know, I mentioned this in a lot of, like, keynotes that I do, you know, going back to like two years ago, I've been talking about this, that I really think it's, you know, probably from 2028 on, it's only going to be domain specific models, right?

Starting point is 00:23:47 That's how you increase the quality of the output. That's how you decrease hallucinations. And that's ultimately what business professionals want, right? So it's almost like, you know, we're going to have hundreds of fine-tuned models on specific niches, right? So even financial analysts, right, there's probably going to be like 20 clawed. you know, domain specific models around that. There's going to be 20 from Google, 20 from Open AI, right? I don't know how they're going to update them.

Starting point is 00:24:16 I'm sure the AI will figure that out, but pretty big news. And kind of the reason I'm including it, because I do think it is indicative of a larger trend that's starting to happen in 2025. But I think in 2026 is when it's really going to make a splash. And I think the combination of domain routing or what I call mixture of models, Some people call it mixture of experts. My explanation that I went over in the 2025 AI roadmap predictions series is a little different in the mixture of models. But yeah, that's where I see this going.

Starting point is 00:24:50 You know, you essentially have a routing model that will then rely on in the future hundreds or maybe thousands of domain specific models. And those are essentially or will, I think, essentially be small language models. All right. Let's get to some drama. So the U.S. government's decision to let NVIDIA resume sales of its H-20 AI chips to China has sparked concern among some U.S. lawmakers, including rep John Moolinare, who chairs the House Select Committee on the Chinese Communist Party. All right. Let me break this down. So essentially, the U.S. government restricted NVIDIA's ability. to export certain chips from China,

Starting point is 00:25:40 which was a pretty big hit to Nvidia's bottom line. Then recently, NVIDIA CEO Jensen Wong met with President Trump, and they left out of that meeting with assurances that licenses for NVIDIA's H20 chip would be granted. So essentially, they reversed it. So essentially, it was like,

Starting point is 00:26:00 hey, you can't send these super powerful AI chips to China because we're in an AI arms race with them. And then they reversed it. And now you have Republicans, which is important, saying like, yeah, we, we don't like this. And why that's important is it's pretty rare for anyone to, not anyone, but on a key issue like this, to object to the president in your own party, which is why it's pretty important here that it's a Republican, you know, making these objections. So Mulanar warned in a letter to the Commerce Secretary that the H20 chip described as both cost-effective and powerful could significantly boost China's AI development outpacing their local Chinese technology. So the Commerce Secretary said the policy shift was part of a broader rare earth materials deal with China and emphasized that Beijing would only receive Nvidia's fourth best chip, not a.

Starting point is 00:27:02 its top performing models. Mulanar, hopefully I got that right, Moulinard argued that the comparisons should be made to China's current domestic chip capabilities, not just to NVIDIA's product lineup, and suggested tighter export controls based on even small technical improvements. So the congressman raised concerns

Starting point is 00:27:22 that selling large numbers of age 20 chips to China could enable China to develop advanced open access AI models, citing deepseeks, R1 model as an example of how China could leverage U.S. technology to gain global market share. Moulinard stressed that U.S. exports policy should ensure American companies, not Chinese firms, set the global standard for AI infrastructure and warned of the risk if Chinese models trained on U.S. technology become widely adopted. This one gets tricky.

Starting point is 00:27:57 And trust me, I'm no geopolitical expert, but I did stay at a holiday and express. ones. Okay, I'm kidding. I actually have a decent understanding of this topic here. So, so here's what it boils down to. Right now, AI is an arms race, all right? Not in the way of a military arms race, but kind of, right, especially because AI is being used in the military now, right? So there's always been this kind of struggle, you could say, between the U.S. and China on who is going to become the first kind of global or the most prominent global superpower when it comes to AI. And here's why it's kind of important, right?

Starting point is 00:28:45 And we actually talked about this a couple episodes ago, which if you missed this episode, I highly recommend going back and listening to episode 564. That was with Dr. Ben Gertzl, who is one of the most famous. people in AI ever. He coined the term artificial general intelligence. So here's why this kind of AI arms race struggle between the U.S. and China really matters a lot. And we talked about it very briefly there in episode 564. Essentially, once we kind of quote unquote achieve AGI or artificial general intelligence, that's when any AI system like maybe Chad GBT's agent can go and do essentially economically meaningful work better than a human.

Starting point is 00:29:35 And you can make arguments on the definition. Are we there? Are we not? It doesn't matter. But what you need to get there is you need Nvidia chips. No other chips, at least right now, will do. No other chips will compare. So that's why there's been some of these kind of tight exports on Nvidia's chips

Starting point is 00:29:52 because the previous administration said, hey, we want to keep that global superpower here. But Nvidia obviously wants to expand their line of business, which also has benefits, right? When Nvidia is able to sell chips to other countries with fewer restrictions, their revenue goes up. Let's be honest, their stock price goes up. And so many people have invested into Nvidia. So, you know, there's kind of pros and the cons. But the big sticky one here, ultimately, it's going to be who has the most.

Starting point is 00:30:26 most powerful AI chips to develop AGI first, artificial general intelligence, because whoever does kind of achieve that first, the path to go from AGI to ASI, according to people who know, like Dr. Ben Gertzl, is actually or could actually be very short, right? And so whoever does figure this out first is going to have an insurmountable advantage in terms of being a global superpower, right? You can almost like wipe out decades of, you know, failing to compete in other areas where you're talking about economic growth, military health, education, whatever. If you can achieve AGI plus ASI before everyone else in doing so with these chips, you have an unfair advantage that I don't know if any other nation would ever really be able

Starting point is 00:31:17 to dethrone, right? Like whoever hits that first is going to have an extreme advantage when it comes to kind of like global superpowers. All right. Last but not least, saving the juicy stuff for last. Open AI has unveiled ChatGPT agent, a new tool designed to handle complex, multi-step tasks for users, signaling a major push into the fast-growing field of AI agents. So ChatGPT is agent is powered by a new, unnamed reasoning model built specifically for handling tasks that require multiple tools such as browsing, analyzing visual data, and using a virtual

Starting point is 00:31:59 terminal. So the tool can perform actions like checking a user's calendar, planning and purchasing ingredients for meals, and generating business research reports for slide decks. So OpenAI essentially combined its operator and deep research teams and their modes technically to develop chat GPT agent blending those capabilities with hands. and on automation. So essentially, you know, you had these chat GPT called them, or Open AI called them agents, right?

Starting point is 00:32:29 Operator, which can kind of see and visually interact with webpages. And then deep research, which more like goes through web pages in bulk, but really only looks at the text side. So the agent, chat chad chad chvety agent combines the pros and the cons of both of those, right? It's speed, it's accuracy. it's, you know, better visual understanding. But then on top of that, it has the ability to create PowerPoints. It has the ability to create Microsoft Excel spreadsheets.

Starting point is 00:33:00 It has a terminal, right? It has an actual virtual computer that I can use, a terminal where it can pull, you know, third-party public APIs. So it's actually extremely powerful and technically extremely dangerous. So right now, most tasks are taking between 50s. 15 to 30 minutes and there are safeguards in place for high risk actions. For example, the tool always asks the user for permission before making bookings or sending emails and it does restrict financial transactions for now.

Starting point is 00:33:34 There's also a watch mode, which is enabled for sensitive actions, pausing the agent if the user navigates away from certain tabs, particularly for financial websites. So it has already rolled out to Chad ChpT Pro users, those on the $200 a month plan. So I've been using it since it came out. And it should be rolling out. Open AI said it will roll out today for other paid users. So if you are on Plus and Teams, opening I did say it would start rolling out Monday.

Starting point is 00:34:06 So we'll see if it actually is rolling out today or if it starts today and, you know, maybe takes a couple of days or a couple of weeks. We'll see. Regardless, very, very impressive so far, right? I do assume I'm going to write on the on the chat chbt Pro plan. I think you get 400 agent queries a month. I'm literally compiling a list right now of all the different agent queries that I'm going to run. It's going to be good.

Starting point is 00:34:35 I'm going to have at least like one computer in my house. That's just going to be running the chat chbt agent kind of at all times. So let's now transition into what's new. what's next. So some of these are rumors. Some of these are things that just rolled out. So little tidbits that maybe weren't big enough to make the main news stories, but pretty noteworthy regardless. So Adobe rolled out custom AI sound effects in their Firefly app. There is a new version of OpenAI's O3 model being tested right now on the web dev arena. It's called O3 Alpha Responses 717. Yeah, great name. But it's like literally

Starting point is 00:35:18 pretty amazing when it comes to what it can code. So keep an eye out on that. Mistral had a big week. They open-sourced their speech tool boxed roll. They released their deep research mode, projects, multimodal reasoning, etc. Right after we saw reports that Apple is considering acquiring Mistral, so pretty interesting there. Manus, the agent, released data visualization features. You know what?

Starting point is 00:35:47 I've been kind of impressed with Manus from the far. I've used it. It was obviously first or, no, not first, but it was before Chachyb-T-A-Gent. So, you know what, if you think we should do a Manish show, just go ahead and drop Manus in the comments here on the live stream or, you know, reply Manis who are daily newsletter. Also, podcast people, we always, you know, you can always put a comment or send a text via the show notes and, you know, via Spotify. So yeah, just say Manus. I'll see if people actually want it. I don't know what people to do.

Starting point is 00:36:23 So you guys got to let me know. Claude is back on WinSurf. Yay. Okay. So, yeah, this is weird. So WinSurf was originally supposed to be reportedly going to get acquired by OpenAI. Instead, their top leadership got Aqua hired by Google, which left the rest of the company kind of up in the air.

Starting point is 00:36:41 So Cognition acquired WinSurf. And now Claude is back on WinSurf. because Claude, as one of the selectable models, was pulled previously when it was reported OpenAI was going to acquire them. Chad GPT's ImageGen now has styles that have rolled out to most users. So if you just click images and then there's a new style tab, so you can pre-stileize your images. There's a new Microsoft Copilot Vision update, and it can now, if you allow it to see your entire desktop, that's pretty cool. AWS, Amazon Web Services announced bedrock agent core platform for building agents. And like I said, as a reminder, chat chagip-t agents should be rolling out to plus users today.

Starting point is 00:37:31 All right. That's a wrap. Let me quickly recap the main stories. So first, according to New York Times reports, meta is considering shifting not completely to a closed source model. but at least starting to build closed source models. US President Donald Trump is set to announce a new and updated AI policy plan focused on deregulation and national standards. AI startup thinking machine labs raised a record $2 billion in their seed round.

Starting point is 00:38:05 Google has rolled out in Google search, Gemini 2.5 Pro, deep search, and AI powered local calling. Open AI is going to be charging commissions on chat GBT driven e-commerce sales. Some AI giants, and that just means everyone, is getting in on the student game as reportedly Anthropic. Open AI and Google are all going to be rolling out some kind of study together or learn together features for students in their respective chatbots. Anthropic launched a domain-specific AI financial analysis tool. We have a congressman questioning the Trump administration's move to allow Nvidia to sell AI chips to China. And then last but not least, Open AI launched their chat, GPT agent.

Starting point is 00:38:56 All right, that's a wrap, y'all, a lot as always. So we do this every single Monday. So maybe this is all you really care about. That's great. But, hey, we actually have the show Monday through Friday. So make sure you join us tomorrow. We're going to be going over for our hot take. Tuesday, agentic AI in the browser, the next frontier of artificial intelligence.

Starting point is 00:39:18 All right. And then make sure to join us for our AI working Wednesday as well. It will probably have an interview or two this Thursday and Friday. So, hey, you can just come Monday for the AI news that matters, or you can stay all week and become the smartest person in AI at your company or in your department. So if this is helpful, please share this with someone. Don't be, don't be a jerk. Don't keep this to yourself.

Starting point is 00:39:40 All right. It's better when we can all learn together and grow together. And then please go to your everyday AI.com. Sign up for the free day and the newsletter. Thank you for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks y'all. Meet Firefly AI assistant.

Starting point is 00:40:01 Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversation. interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. 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 Your Everyday

Starting point is 00:40:47 AI.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.

Everyday AI Podcast – An AI and ChatGPT Podcast - EP 571: Google’s AI makes phone calls for you, ChatGPT Agents and more AI News That Matters

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.