Everyday AI Podcast – An AI and ChatGPT Podcast - EP 557: OpenAI and Meta's war on AI talent, will Gemini CLI kill Claude Code? 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 in 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. There's an AI talent more going on.

Starting point is 00:00:48 AI is doing Salesforce's job. And one new AI from Google could kind of wipe out a big part of Claude's business. All right. There's a lot of drama, new developments and pretty breaking news happening in the AI world. And if you don't have hours to spend each and every day keeping up and trying to figure out what matters and what doesn't, well, don't. Instead, spend your Mondays with us at Everyday AI. What's going on, y'all? My name's Jordan Wilson and welcome to Everyday AI.

Starting point is 00:01:27 This is your daily live stream podcast and free daily newsletter helping everyday people like you and me, not just learn what's happening in the world of AI, but how we can leverage that information to grow our companies and. our careers. If that's what you're trying to do, Mondays are a great way to start. Yeah, you don't have to have a case of the Mondays when you can just keep up and get ahead with everything that's happening in the world of AI News. That's our weekly AI News that matters show on Monday. So, thank you for tuning in to this unedited, unscripted, live stream and podcast, but you got to go really take to the next level with our free daily newsletter. So if you haven't already, please go to your everyday AI.com. So there, not only can you sign up for that free daily newsletter,

Starting point is 00:02:12 where we recap each day's podcast and give you everything else you need to know happening in the world of AI, but also on our website, there's more than 550 back episodes that you can go listen to, watch, read at any time all for free. It is a generative AI university. So without any further ado, let's get into it. Live stream audience, good to see. it. Joe getting sweaty all up in Fort Lauderdale. Thanks for joining us. Brian,

Starting point is 00:02:43 join it from Minnesota. Good to see it as always. Marie. We got the trade partners happening here on the YouTube machine. Nathan joined us from Australia. Good day, Nathan. Let's get after it. Dr. Harvey Castro and Michael, shall we? So first piece of AI news. The AI labs are winning copyright suits. We never would have thought that, right? So federal judges have allowed AI firms to train on copyrighted books. So federal judges in California ruled in the past week for two different AI companies. And they said that AI developers, Anthropic and meta, can legally train their large language models on copyrighted books, marking a significant early victory for the AI industry. But this is in the end of easy.

Starting point is 00:03:38 of these two cases, FYI, I would expect both of these to head to higher courts. But these initial rulings from federal judges are some of the first in the U.S. to address how copyright laws applies to AI systems, setting some early precedence, but leaving the larger legal fight unresolved. So, Judge William Alsup determined that Anthropics use of millions of copyrighted books to train its chatbot Claude was legal for the books that the company paid for, but said that Anthropic must still face claims regarding a pirated library of over seven million books. I just see now Anthropic hiring hundreds of interns to go to rummage sales and find as many of those seven million books as possible to go scan them because that's reportedly all that Anthropic did is

Starting point is 00:04:34 They bought the books. They kind of took the covers off and scanned them. And then they kind of were able to prove at least somewhat in court that, hey, we don't spit out verbatim copies of these books. So in this initial ruling for Anthropic, they're good to go, at least on some of the books. In a separate case, Judge Vince Chabara sided with META in a similar lawsuit that meta was facing on copyrighted materials, finding that R. arguments from a group of 12 authors, including Sarah Silverman, were insufficient to rule on a

Starting point is 00:05:10 copyright infringement related to Meta's Lama AI motto. So, yeah, essentially what the matter ruling said was the judge more or less said, yeah, we can't really rule against meta because the defendants did a kind of a poor job of going after meta for specifically what they were asking for. So intellectual property experts cited by the Associated Press say that these cases are likely to be appealed and could eventually reach the U.S. Supreme Court, meaning the ultimate outcome for AI training on copyrighted works remains uncertain in the long run, but a pretty big win in the short term for AI companies. So right now, U.S. copyright law, it's a little messy and I know we have listeners from all

Starting point is 00:06:01 over the world. But here's the thing. The biggest AI companies are in the U.S. So I do assume that whatever kind of precedents from a legal perspective are set here, even though those may not be applied unilaterally throughout the globe, obviously, that's going to be the benchmark that other courts and other countries start with, right? Whatever rulings, if any, were made in the U.S. because that is where the biggest AI players are. So U.S. copyright laws grant creators exclusive rights over reproductions, distributions, and some derivative works raising questions about whether AI generated outputs might infringe on those rights. So legal experts warn that while these rulings give AI company some breathing room,

Starting point is 00:06:50 the broader issue of compensating rights holders and limiting AI outputs that resemble original works is far, from settled. And I would agree with that. And this isn't going anywhere anytime soon. And like we've talked about previously on the show, the biggest domino to fall is obviously going to be the New York Times versus open AI case that has been open now since late 2023. It was December 2020. And I do assume that whatever happens there will have a cascading effect on the rest of the industry. And Another big one that could fall before that is the Disney and Universal lawsuit against AI Image Generating Company mid-jurney that we covered on our Hot Take Tuesday about two weeks ago. So if you care about copyright law or if you're just intrigued, those are some big cases

Starting point is 00:07:45 to keep an eye on outside of these two that were temporarily at least ruled in the favor of meta and anthropic. So yeah, what do you, what do you guys think? What do you guys think? I like this comment here from Michael on YouTube. He said, OMG, AI companies are about to ravage the free little libraries. Yeah, yeah. It's going to be interesting.

Starting point is 00:08:14 Even on this very small piece where the judge essentially said, hey, Anthropic bought these books and scan them. And the companies weren't able to show. that you could directly reproduce their works in these large language models. So, yeah, I'm wondering if there's going to be a resurgence of book buying, specifically just to train large language models. Clearly, it's already happening, and I would guess with this recent ruling, yeah, that might set a wave of similar aspects in motion. All right.

Starting point is 00:08:54 more lawsuits. This one with OpenAI. So OpenAI has been hit with a lawsuit from I.O. A startup led by Jason Rugolo over the name of its upcoming hardware startup with Joddy Ives called I.O. For what they are saying is a hardware device, not a lot of details known, but more or less, this is a trademark battle. And a lot of people. thought that this partnership between OpenAI and famed Apple designer, Johnny Ive, was kind of halted because it was essentially scrubbed from a lot of places online, including Open AI's website. So a lot of people were like, oh, this means the project has gone under.

Starting point is 00:09:42 It looks like it's just a temporary trademark dispute. So more on this, the dispute became public after OpenAI CEO Sam Altman posted private email exchanges with Ruegelow on Twitter showing that the latter had pitched Open AI for a 10 million dollar investment and later sought a partnership. Altman decided to decline to invest, citing that Open AI was quote unquote working on something competitive and later referenced Johnny Ive, the former Apple designer, as leading Open AI's new hardware division. The lawsuit, though, claims that Open AI knew about IOs. I think that's how I'm going to say it out loud here, right? I.O, which is the Open AI, Johnny Ives collaboration, IO versus I.O. I.YO.

Starting point is 00:10:40 All right. So Open AI, the lawsuit claims that OpenAI knew about IOs technology and branding through meetings with Altman's investment firm and Johnny Ives design company. as early as 2022, according to the lawsuit. And like I said, OpenAI did recently scrub all I.O branding and mentions from its website after a court granted a temporary restraining order in IOs favor on June 22nd following the trademark lawsuit that was filed on June night. So OpenAI's hardware team testified that its device is not. an in ear or wearable device, which is what iOS product is, right? It was essentially kind of like, I won't say clunky, but it's a large, think of it as like an

Starting point is 00:11:36 AirPod, but like much larger. It almost fills your ear. And then there's some AI capabilities in there. So essentially, we got some confirmation on what this new project from OpenAI is not. So it is not something that is going to go in your ear. And it's apparently not going to be called I.O. They may rename this or they may fight these trademark claims. So yeah, pretty interesting here on how Sam Allman decided to deal with this,

Starting point is 00:12:08 essentially got ahead of it after people assumed, right? So essentially it just went quiet, right? All of this Open AI and Johnny I's partnership information that Open AI had posted about on social media and on their website went down. And then, you know, the internet rumors started swirling around like, oh, my gosh, this thing is already, you know, this multi-billion dollar acquisition is already in the tanks. And then essentially, Sam Altman came out, shared the receipts, shared the screenshots, right where Rugalow was kind of hounding open AI to invest in his IEO technology. They said no. And then now we see I.O.

Starting point is 00:12:50 A new hardware division come out. So, yeah, I don't think this lawsuit is going to really lead to anything, right? It's not going to take down this new hardware company that is working on a product that we said or that we reported earlier might not come out until late 2026. And a lot of people are saying it is a kind of a third hardware device. It's not supposed to compete with a laptop or a phone. So a lot of people are saying it's kind of a hardware piece that you might slip into a pocket or something like that. And it's going to potentially pick up, you know, information from your surroundings and then pair that with your open AI data. That's kind of the best reporting that we've seen so far.

Starting point is 00:13:35 But yeah, some slight road bumps on the on the path so far. Yeah, live stream audience, what do you think? Is this a legitimate complaint? I don't know. It's not the same name. It's not I-O. It's I-Y-O. And yeah, there's some receipts going back a couple of years showing that Jason Rugo Lo was trying to meet with OpenAI and Sam Altman and, you know, getting some conversations going.

Starting point is 00:14:09 But, you know, ultimately, it seemed like Sam Altman said, yeah, we're not going. going to do that. We're going to just do something better. So I don't know how, you know, how far back Open AIs I. I.O. naming went or if that's something that they came up with recently, I guess we'll find that out through court discovery if this does go that route. But I would assume that this is going to either get squash pretty quickly or open AIs just going to announce a new or a different name. All right.

Starting point is 00:14:40 Next piece of AI News, 11 Labs has launched 11 AI. a voice assistant that executes tasks in real time. So 11 labs, which a lot of people know as the, just kind of probably the leader in AI text to speech. They've introduced their voice-activated AI assistant that can perform actionable tasks across connected digital tools, marking a pretty big step beyond just their text-to-speech platform. The Alpha version of 11.

Starting point is 00:15:15 A.I. is now live at 11.a.I. That's the website, allowing users to issue spoken commands, such as planning their day, updating project management tools, or researching prospects using integrated services. So unlike most existing voice assistants that are limited to simple conversation, 11A.I is designed to carry out sequential and productive actions by connecting with third-party apps and also internal tools. The system currently supports integrations with platforms like perplexity, linear, Slack, hacker news, and Google calendars with connections planned to roll out weekly. And here's the pretty exciting thing.

Starting point is 00:16:02 It's built on the Model Context Protocol, or MCP, a standardized API framework for AI that enables seamless integration with services such as Salesforce, HubSpot, Gmail, and Zapier. It also supports custom MCP servers for enterprise workflows. I'm excited about this one, personally, if I'm being honest, I don't use 11 labs a ton. It's one of those pieces of software that I've been paying for since it came out. And I don't know why, you know, they have a pretty affordable entry level plan. And I'll just go in there every once in a while just to kind of play around with the latest

Starting point is 00:16:43 capabilities and see what one of the leaders in the space is doing. But this might change it. I might now actually be using this 11 AI a lot if they continue to invest in it. So right now, the assistant uses a permissions model, allowing users to specify what actions it can take within connected applications, which could address any privacy and security concerns. 11 Labs is positioning 11A.I as a direct competitor to new actionable agents. such as the new one from perplexity, their iOS app, which I think is really good. Amazon's Alexa Plus, Alexa, that's finally getting smarter.

Starting point is 00:17:23 And we saw that up to a million people have been invited to try out Alexa Plus. It's been a very, very slow rollout, right? But Alexa, through their partnership with Anthropic Claw, is apparently going to be a little less dumb in the near future. So we'll see how this stacks up against the others. if you follow the show at all, you've seen that Apple's smarter Siri has gotten pushed back like, you know, 20 times. It gets pushed more than a kid on a swing. You know, it just keeps getting punted down and along the road. So who knows if we'll ever actually see a smarter AI Siri that you can talk to conversationally and you can have these follow-up questions. But hey, I like what perplexity.

Starting point is 00:18:09 And now 11 labs are doing by competing in the space. where there is a huge gap and also 11 labs taking a pretty unique approach by leveraging the very popular MCP standard that was developed by Anthropic and has been adopted by just about everyone. Adobe just introduced an entirely new way to create, bringing the power and precision of its creative suite into one conversational experience. Meet Firefly AI assistant, now live in the Adobe Firefly app, the all in one, creative. A.I. Studio. Powered by Adobe's creative agent, Firefly AI assistant lets you start with your vision,

Starting point is 00:18:56 just describe what you want, and shape the outcome as it takes form with the assistant. The assistant orchestrates multi-step workflows, drawing on 60 plus pro-grade tools across Adobe Creative Cloud apps, including Photoshop, Illustrator, Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director.

Starting point is 00:19:37 Adobe Firefly AI assistant now in public beta. See it today at firefly.adopi.com. All right. some breaking news here. The U.S. Senate has struck a deal to pause state AI regulation in exchange for federal funding. That is if this big beautiful bill, as it's actually being called here in the U.S. passes. So this is contingent, and this is just hours old. So I haven't even seen a lot of mainstream media.

Starting point is 00:20:19 report on this yet, but here's what's happening and here's why this is important. So U.S. lawmakers have reached a deal to temporarily block states from regulating artificial intelligence for five years if they want access to $500 million in federal AI infrastructure and deployment funds, according to recent reports. So this is from the Hill. So the compromise was negotiated between senators Marsha Blackburn and Senate Commerce Chair Ted Cruz. And it reduced the original proposed moratorium from 10 years to five. So essentially, this is part of this big, larger bill that happens a lot in the U.S. And if you're wondering, like, what is this? This is essentially a budget bill. And it's been codenamed the big, beautiful bill. I don't understand.

Starting point is 00:21:15 But it has not passed yet. And this kind of of AI piece. Yeah, there's like dozens of things that are just grouped in this bill. You know, a lot of pork, what we call it, you know, here in the U.S., everyone's, you know, trying to get their certain provisions in or out. So this spending bill is, you know, quite ridiculous if you look at it. And it's always ridiculous, right? And this isn't a political thing, right? But just how the U.S. in general and how the government spends money is out of control. anyways, there was this 10-year ban that said, but there wasn't an option. So before that was included in this bill that was apparently going to cause a lot of U.S.

Starting point is 00:21:57 senators to vote against it, specifically Republicans who are in the majority. So they needed to get more Republicans on board in this 10-year AI state ban was apparently and reportedly a pretty big holding point. from keeping the larger bill from moving forward. So now there's kind of this compromise that is a five year ban that states essentially can't come up with new laws regulating AI and instead have to kind of follow the federal government's lead. But also now there is at least choice.

Starting point is 00:22:33 So it's more or less, hey, if you want a big piece, or if you want your state's piece of this federal funding, billions of dollars in federal funding for AI, broadband and some other things, you have to kind of stop regulating AI at the state level. So it seems like at least right now with this latest compromise that it's gone down, it's a little less strict and instead of 10 years, it's five. Also, the updated provision now exempts state laws focused on unfair or deceptive practices, children's online safety, child sexual abuse material, and publicity rights.

Starting point is 00:23:13 So Blackburn is a vocal advocate for online child safety, highlighted the need to protect progress made by states in regulating big tech and AI. So the deal comes as Congress continues to struggle with passing comprehensive legislation to oversee the digital space and protect consumers from potential AI harms. So the new measure survived review by the Senate parliamentarian Elizabeth McDonough, who ruled it did not violate the bird rule and could be included in the reconciliation bill. So yeah, there's kind of a lot in the weeds here, but essentially how this 10-year ban that was included in this bill, right, you have to vote for the whole bill.

Starting point is 00:24:01 So, you know, this 10-year ban on states regulating AI was apparently causing a rift in the Republican party that has the majority. So all they have to do is just get all of their members. to be on board with the rest of this bill to get it sent through. But now this AI compromise could be one of the things that now pushes them to get this through. So some lawmakers, including senators Ron Johnson, Josh Howley and rep, Madjury Taylor Green, have previously opposed the AI moratorium. And it remains unclear if their concerns are fully addressed in this new version.

Starting point is 00:24:42 So now the Senate is expected to vote on the broader package as early as today, which includes this new updated AI provision as they try to meet President's, as they tried to meet President Trump's July 4th deadline. And that's an arbitrary deadline. You know, it's not like if this thing doesn't pass, it's not going to happen. But this one's pretty interesting, y'all. And I was kind of shocked to see this 10-year ban included in the original Bill, and I said, you know, at least online and some online discussions, I'm like, there's a zero percent

Starting point is 00:25:18 chance this thing gets passed as is. Number one, it violated the bird rule. There's some questions on the constitutionality of this clause, but now it is more of like an opt-in. It's more of like, hey, if you want hundreds of millions of dollars, right, the federal government has set aside hundreds of billions of dollars in funding in infrastructure projects. And they're essentially saying if you want a piece of this big $100 billion AI pie, your state essentially has to agree not to come up with any state laws regulating AI's development. And one of the big reasons, and if you don't pay attention a lot to politics, you know, I was a political reporter in a former life.

Starting point is 00:26:03 So I understand this, right? And this happens all the time. you know, the federal, state level everywhere, where sometimes, you know, you're voting on a bill and there's, you know, dozens or hundreds of provisions, right? So a lot of times they're, they're trying to balance, you know, one of these things that are causing people to not vote for this whole bill. So the Republican majority can pass their spending bill more or less. And this AI, you know, not allowing any states for 10 years to come up with any laws surrounding AI was absolutely bonkers. So I still think five years is kind of wild. I would have expected maybe a

Starting point is 00:26:44 two or three year potential ban kind of aligned with current president's Trump term. But what they're doing right now is essentially making it an opt-in. Like, yeah, if you do come up with any laws against AI, you're just not getting this big piece of the pie. Joe here saying, I don't have any confidence in the U.S. government to figure out how to regulate AI. States are at least a bit more agile, but I doubt they could figure it out either. You know, Joe brings up a pretty good point. If I'm being honest, I kind of agree with Joe. I've covered, you know, both state, local, and a little bit of federal politics in my time previously as a reporter.

Starting point is 00:27:31 and even at the federal level, I mean, you have Congress people who literally don't even know what the internet is or how it works. So I don't have any level of confidence in the federal government to legislate or regulate AI development. But I think that's the whole point. I think right now this current administration doesn't really want any rules or regulations as it comes to AI's development. Because essentially, it is an AI arms race. the U.S. against China. So the federal government doesn't want any states, you know, holding their federal progress back.

Starting point is 00:28:09 But, you know, all you really have to look to right now is number one, is this going to be the final compromise that makes its way in this big, beautiful bill that will be voted on any hour now. And what is California going to do? Because California right now is the only state that actually matters because that is where all of the big AI labs, are located, right? Google, Open AI, meta, I guess the one big exception to that rule is Microsoft, which is headquartered in Washington. So, but for the most part, you know, anthraffic as well,

Starting point is 00:28:44 in California. Almost every single big AI company is in California. So, you know, we'll have to see if this is actually what happens and what California decides to do, you know, are, do they care about that federal funding or might they? just passed some AI regulation since the majority of the world's AI development actually comes out of the state of California. So keep an eye on what California does. Speaking of California, one of the big companies there, Anthropic, just released a pretty big update to their Claude platform. So Anthropic has rolled out some new features for Claude. It's consumer-facing AI app and website, making it easier for non-technical users.

Starting point is 00:29:30 to build software simply by chatting with Claw. So the new update centers on artifacts. One of my favorite AI features in Anthropic was pretty far ahead to everyone else by rolling this type of integration out into their platform. But the new update to artifacts, right? And if you don't know artifacts, it's now if you've heard of ChatGPT's Canvas mode or Google's canvas mode, it's essentially that. Think of it as a side-by-side window and you can dump a bunch of

Starting point is 00:30:00 data, you know, on the left side and then Claude Artifacts will build something on the right side. And if you've been following our AI at work on Wednesday series that we started about a month ago, we do a lot of work inside Claude Artifacts and Chad Chb-T Canvas and Google Gemini Canvas. So this new update to artifacts essentially is you can build and share artifacts, which isn't new. But the new thing is, is you can embed their Claude. AI technology in any artifacts that you create. So this is pretty cool, right? So essentially now you can build an app without any code, right?

Starting point is 00:30:44 But even within that web app, right, you can embed AI functionality in there without really being a developer or even knowing how it all works. Right. So you could, as an example, upload all your data, create a little web app. And then in that web app that you could then share with people, you can embed AI technology. So the users could go in and essentially use a version of Claude that lives inside of your artifact. All right. So pretty exciting development there from Anthropic.

Starting point is 00:31:15 But if I'm being honest, Google's had this, right? And we demoed this, I think, three weeks ago on our AI at work Wednesday. And it's actually so crazy because it's like hard to find inside Google Gemini canvas mode and they didn't really make a big deal out of it when they announced it at their i.O. conference in May it's just this little Google Gemini button and it's just there and it works. And the cool thing that I like a little bit more about the Google Gemini version of their embedded canvas or sorry, embedding Gemini in your canvas iterations is it's free to use. So right now the downside with Anthropics is you have to

Starting point is 00:31:58 have a clawed account. So if you share this openly and publicly, right, it doesn't go out of your kind of limits. It goes in everyone else's. So front end users, if they want to use that embedded kind of clawed functionality within this new artifacts, they have to be logged in if they want to use that. And it goes essentially from their usage, which we know, according to Claude, you know, if you've used it, if you hear me talk about this all the time, yeah, Claude limits are extremely low. So, like I said, that means that right now, subscribers on the highest tier plan, the $200 a month Claude Max plan will have way more capacity to actually use this and

Starting point is 00:32:44 leverage it on the back end. But yeah, right now, if you're a normal paid, you know, $20 a month, Claude subscriber, I don't think this is going to add a ton of utility, especially when Google has it embedded in their canvas mode and it's free, right? Even on the back end, you don't have to have, you don't even have to be logged in to use it. So, you know, I'd like this advancement from Claude, but right now, Google kind of already beat them to the punch.

Starting point is 00:33:16 All right, let's keep going on to more big tech news. So Salesforce CEO Mark Benioif revealed that AI now handles 30 to 50% of the company's total workload, a major ship that highlights how deeply AI is being integrated into business operations. So, Betty Hoy emphasized that AI is now performing tasks at Salesforce, once done by humans, including software engineering and customer service, allowing current Salesforce employees to focus on what they call higher value work. The company also aims to have one billion AI agents by the end of the year as 65% of companies are experimenting with AI agents based on a April survey from management consulting firm, KPMG.

Starting point is 00:34:12 So Salesforce's AI product used by some big clients such as Disney has reached reportedly a 93% accuracy for tasks like customer service without human supervision. So Benny Oif noted that perfect accuracy is unrealistic, but Salesforce outpaces competitors, what they say is because of more extensive data in metadata data. So this is going to be pretty interesting to see, and this is the first time we've seen a huge company admit that their general work is being done by AI. Could this be exaggerated a little? bit, absolutely. Could this be a play by Salesforce's CEO to get eyes on Salesforce and their

Starting point is 00:35:08 agent force platform? Absolutely. I'm not buying it if I'm being honest. I don't know. I still haven't decided what we're going to do for our hot take Tuesday tomorrow. I put out a different poll on LinkedIn and a poll in our newsletter that you can sign up for at your everyday AI.com. And it was split. Some people, I think on LinkedIn, people wanted to hear about Salesforce and in our newsletter. People wanted to hear about something else. So maybe I'll save some of my spiciest takes on this. I'm just personally not buying it, right?

Starting point is 00:35:42 30 to 50% of your work? No. Because if I'm then on the board of sales, Salesforce, I'm going like WTF. why is it our revenue then up hardly at all why has our stock been getting straight up battered for the past year especially compared to other you know tech companies right so it seems like sales force is really trying to position itself as kind of an AI first an AI native and an agent native company yet their stock is not making that same rise as the AI companies and their revenue isn't soaring either. So I don't know. I'm not personally buying it. What do you all think? All right. Our next piece of AI news, Google is going free. Yeah. They just launched their free Gemini CLI tool to bring AI coding tools directly to the desktop. And this is something that could straight up kill Anthropics desktop tool.

Starting point is 00:36:56 Claude code. So Google has released Gemini CLI in the CLI stands for command line interface if you're not a dork like me. And this is a free essentially desktop coding tool that uses the command line interface like the terminal on Mac and it gives developers direct access to the Gemini 2.5 Pro models. And it's unlimited. it is free. Well, it's not unlimited, but it's essentially unlimited.

Starting point is 00:37:28 I think there's very few people that will be able to hit the number of tasks that, you know, that there's a limit on. So it's essentially unlimited. It is free. And it is open source. And when I saw this news, so, you know, I obviously saw it before it was announced, you know, and when it was announced, I'm like, yeah, this is, this is going to get, this is going to crush. This is going to crush not only Claude code. So Anthropic is going to have to figure out

Starting point is 00:38:00 other ways to monetize, but also this could impact some of the kind of desktop AI IDs out there a little bit as well, such as Claude or sorry, not just Claude Code, you know, potentially cursor, windsurf, Open AI's new codex, right? So Google making a huge play here by making this Gemini CLI, essentially a desktop coding agent free open source using the world's most powerful model right now. It is still Gemini 2.5 Pro. So it allows up to 60 model requests per minute and a thousand per day at no cost. So like I said, it is open source under the Apache 2.0 license and is available for installation on GitHub. So developers can use Gemini CLI. for coding tasks such as writing, debugging, generating content, conducting research, managing

Starting point is 00:39:00 tasks, software development, just about anything that you would want to do in the software development realm. The tool is integrated with Gemini Code Assist, which is Google's AI coding assistant, and is accessible as well in Visual Studio Code for users on free standard and enterprise plans. So, this one's, this one's interesting, right? And I'm obviously, you know, following a lot of discussion online. And I don't know. If I'm anthropic, I'm not super happy about this, right?

Starting point is 00:39:35 Because I think a lot of their revenue obviously is coming from people in software development, coders, engineers, et cetera. So when Google offers this free open source tool that is essentially unlimited, Yeah, I do assume they're going to get a big chunk of Claude code users, people that are relying on Claude code, which is a paid tool, and it's not cheap, as well as people using cursor. So it should be interesting to see what happens here as this continues to gain adoption. All right. Our last, well, technically story and a half here, there has been. an all-out

Starting point is 00:40:24 AI hiring assault over the last week in change with meta poaching top researchers and developers from open AI. So first, meta has hired at least eight

Starting point is 00:40:40 open AI researchers in the last week and a half, according to reports. So that is from the information and from the Wall Street Journal. So they've hired at least eight. Meta has hired at least eight researchers from OpenAI in the recent weeks. And this aggressive recruitment follows Meta's April launch of its Lama 4 AI models,

Starting point is 00:41:05 which reportedly fell short of CEO Mark Zuckerberg's expectations. So what happens, you know, is meta's Lama 4 didn't really make a big enough splash. And then ever since, they've been either acquiring companies, right, they spent 14, billion dollars for a 49% stake in scale AI and having its leadership team lead meta's new super intelligent team. So that was a multi-tiered play because not only do they now get a huge chunk of revenue from that 49% equity in scale AI and have its founder, or its CEO now running meta's super intelligence teams. But now a lot of their big tech competitors have obviously stopped using scale AI, which is, you know,

Starting point is 00:41:58 really big in the AI data labeling space and model evaluation. So not only is that a big investment in terms of meta's potential top line, but also it does hurt their competitors. And now, speaking of hurting their competitors, they're just going after anyone in everyone at OpenAI. So this hiring spree highlights the escalating competition for AI talent with OpenAI CEO Sam Altman last week, claiming that Meta has offered signing bonuses of up to $100 million in annual salaries at more than $100 million, though Meta's CTO Andrew Bosworth says that the real offers are not really that much and much more complex than that. The rivalry between meta and Open AI is intensifying as both companies race to build the most advanced AI systems and secure the top AI researchers in the field. Yeah.

Starting point is 00:42:59 And reportedly, these researchers that meta poached from Open AI are some of the researchers that helped develop their reasoning models and some of the brightest minds at the company. So pretty big. And then our last AI news story, which is related to that, well, on the backside, Open AI is reportedly kind of scrambling to kind of get all their duts in a row and strengthen their team. So Open AI is facing an intense talent war after Meta has recruited up to eight senior researchers from the company to join its superintelligence lab. So Zuckerberg, like we said, had. reportedly offered bonuses and salaries as high as $100 million, which is a figure confirmed by multiple sources. And like we said,

Starting point is 00:43:53 but Matt has said not really true. But here's the most recent new news on this ongoing escalation between the two companies. So Mark Chen, open AIs chief research officer, reportedly just addressed staff in a forceful memo on Slack, which was reportedly leaked to or obtained by wire. And it expressed a sense of violation and promising aggressive action to retain key researchers. So open AI leadership, including Chen and Altman, are working quote unquote, around the clock to talk to employees with competing offers,

Starting point is 00:44:36 recalibrating their own open AI or sorry, recalibrating Open AI's compensation and exploring new ways to reward and recognize top talent. So despite the push to keep researchers, Chen emphasized fairness and said that he would not retain employees at the expense of equity among staff. So the internal memo reportedly included encouragement from other open AI research leaders, urging staff to reach out if pressured by meta's exploding offers and highlighted the company's support system. Yeah. So even the, uh, the headline or the subhead here in Wired was open AI leadership responds to meta offers quote unquote. It's like someone has broken into our home.

Starting point is 00:45:25 Wow. So the competition for AI researchers is heating up across Silicon Valley with meta targeting talent from both open AI and Google, while Anthropic is seen as a less of a cultural fit for, uh, poaching leaders there. Also, Open AI employees are reportedly working up to 80 hours a week, prompting the company to kind of shut down for the week. But this is something that they've done reportedly almost every single year around the July 4th holiday here in the U.S. But that's why the timing here is particularly interesting.

Starting point is 00:46:03 And reportedly why this very forceful memo from Open AI went out is because Open AI is essentially going to have a week off, and meta is apparently not done poaching. So it's going to be interesting to revisit this in the coming weeks and see what happens and see if this is something that is eventually going to tilt the scales in favor to meta. So whether we see a Lama 4.5 or a Lama 5, I would expect their next version of their large language model to be extremely competitive,

Starting point is 00:46:37 whereas a lot of people, including myself, didn't even really expect a lot out of meta, meta's Lama 4, and I was not personally impressed, but I am impressed very much so by the recent moves from meta and what this ultimately means for the long term. All right, that's it for the AI news, but a really quick roundup of what's new and what's next. So some recent releases and some recent rumors or upcoming releases.

Starting point is 00:47:04 So Twitter's GROC 3.5 is no longer. Elon Musk confirmed that instead they're just going to skip 3.5 and said that GROC 4 will be out around July 4th. I wouldn't hold my breath. There has been about a couple hundred delays, you know, through everything that GROC has been, you know, talking about and, you know, releasing the weights and all that stuff. So I wouldn't hold your breath on that if you were a GROC fan. Google Gemini scheduled tasks is out to paid users. So check your account for that. There's no dedicated UI for it on the front end, FYI, but there is a setting for your scheduled task.

Starting point is 00:47:46 Google has also rolled out a Gemini model that runs on robots locally, which we covered in the newsletter last week. A lot of new Google rollouts. Also, their Ask Photos kind of feature that was previously delayed. for a long time is being slowly rolled out first to paid users where you can essentially talk with your history of Google Photos, which I'm personally looking forward to that one. Open AI is gradually rolling out limited connector features to standard chats. So what that means is before some of these connectors like Gmail, Google Calendar, Microsoft Teams, Outlook were only available to deep research chats, which a lot of people

Starting point is 00:48:31 We're like, okay, this is cool. But, you know, if I have to wait, you know, seven to 15 minutes, is it that useful? So now OpenAI has been rolling out some connectors to normal chats and not just deep research. ChatGBTGPT has also rolled out a Canva connector. If you're in, you know, big into the Canva design space and also a Slack connector will reportedly be rolling out next. And chat, GBT, pretty big one here actually, or Open AI rolled out. out their deep research to their API as well as webhooks. So that one's actually really big.

Starting point is 00:49:08 So I would now expect hundreds of new of these kind of deep research tools to be rolled out specifically for even specific verticals or type of work. So that's going to be an exciting one that Open AI now is opening that out, opening their deep research functionality out to thousands of tools. So yeah, keep an eye out for that. All right. That's it. That's a wrap, y'all.

Starting point is 00:49:35 Let me do the world's quickest recap of the AI news that matters for this week. So first, a federal judge has allowed at least temporarily meta and anthropic to train on copyrighted materials, although we're going to see some ongoing lawsuits about that. Speaking of lawsuits, OpenAI is facing a trademark case over their device name I.O versus Iyo. the other hardware startup, 11 Labs has launched 11 AI, a voice assistant that execute tasks in real time by using the MCP protocol. The Senate has struck a deal here in the U.S. to pause state AI regulation for five years in exchange for federal funding if that bill passes. Anthropic has unveiled a new update to its artifacts feature, allowing users to, to embed clawed functionality in its artifacts creations. Salesforce is apparently leveraging AI to do up to 30 to 50% of its company,

Starting point is 00:50:40 of its companies' day-to-day work. Google has launched a free and open source tool called Gemini, C-LI, or command line interface to bring AI coding tools directly to developers. Meta has hired up to eight open AI researchers amid an intense AI talent war. and Open AI on the back end is scrambling to retain top talent according to a recent report. If this was helpful, please let me know. Share this with your network. We spend so much time cutting through all the fluff, telling it to you just how it is.

Starting point is 00:51:14 So if this was helpful, please tell someone about it. If you're listening to on the podcast, thank you. Please make sure to follow the Everyday AI show and subscribe. Leave us a rating or review. If you're listening on Apple Podcast or Spotify, we'd really appreciate that. and I'd appreciate you tuning in tomorrow and every day for more everyday AI. Thanks y'all. Meet Firefly AI Assistant.

Starting point is 00:51:43 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 conversational interface. You direct the outcome while the assistant accelerates execution, Stay in 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.

Starting point is 00:52:19 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 EverydayAI.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 557: OpenAI and Meta's war on AI talent, will Gemini CLI kill Claude Code? AI News That Matters

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