Everyday AI Podcast – An AI and ChatGPT Podcast - Ep 749: OpenAI’s C Suite shakeup, another huge Claude leak, Gemini’s BIG small model and more AI News That Matters

Episode Date: April 6, 2026

Claude Code’s code leaked. Are they done for? 😱OpenAI’s C suite won’t really look the same this week as it did last week. 🪞Microsoft may have a new AI challenger in …. Slack? 🥊And the biggest AI news this week might have been a small model from Google. But seriously….. do NOT sleep on Gemma 4. You’ve been warned. 🥱Don’t spend hours a day trying to keep up with AI or wonder about what AI updates matter. That’s what we do for you on Mondays. OpenAI’s C Suite shakeup, another huge Claude leak, Gemini’s BIG small model and more AI News That Matters — An Everyday AI Chat with Jordan WilsonNewsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion on LinkedIn: Thoughts on this? Join the convo on LinkedIn and connect with other AI leaders.Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:OpenAI C-suite Leadership Shakeup ExplainedAnthropic Claude Code Source Leak AnalysisOpenClaw Subscription API Shutdown ImpactSlackbot Launches 30 New AI FeaturesGoogle Gemma 4 Open Source Model ReleaseGemma 4 Apache License and PerformanceOpenAI $122 Billion Funding Round DetailsAnthropic Claude Code Reverse Engineering IssueTimestamps:00:00 OpenAI leadership shakeup details04:09 Leadership changes at OpenAI08:21 Changes to OpenClaw API usage11:19 Slack's major AI updates15:26 Slack's new AI and CRM features20:29 Benefits of smaller, local models23:53 Google's strategy with Gemini26:43 OpenAI's massive funding round30:13 The Anthropic source code leak32:49 Copyright concerns for AI code36:49 AI news roundup and updates38:48 Learning AI and free resourcesKeywords: OpenAI, C-suite shakeup, executive health leave, leadership transitions, Brad Lightcap, special projects, $100 billion fundraising, OpenAI IPO, investor uneasiness, Fiji Simo, Greg Brockman, Jason Kwan, Sarah Friar, Denise Dresser, Kate Roche, AGI deployment, Chargebee T, Anthropic, Claude code leak, OpenClaw, Opus 4-6, agentic tools, compute strain, API usage, Pro and Max plans, OpenAI ChatGPT integration, Google GemSend Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

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. This week, we had some gaudy funding announcements, staffing drama, new models,

huge AI shakeups, and more. And by huge, I do mean huge. I mean, we got another huge leak from Anthropic, not the one from last week, but a new one as its clawed code source code has essentially been leaked and duplicated on the internet, which could mean some absolutely crazy things in the near future. Open AI closed a 100 plus billion dollar fundraising round, yet at the same time, like half of its C-suite will look different this week than last week. Slack, yes, that Slack, could be primed to challenge the likes of Microsoft in terms of collaborative AI capability soon.

And the biggest story of the week in AI might have been one that you overlooked, because it was a big story about a smaller model from Google. And y'all know I have been sure been about small language models and local AI since 2023. So I might actually go bonkers on this one because I think it's a really big deal. All right. If you can't spend like 10 hours a day following AI like I do and you get lost and feel like you're getting left behind, don't worry.

I waste all those hours and I tell you what matters and what doesn't on our Monday's show when we bring you the AI news that matters. So let's get straight into it. And what's going on? If you're new here, my name's Jordan Wilson. This is Everyday AI. This is your daily live stream podcast and free daily newsletter helping everyday business leaders like you and me make sense of the nonstop AI craziness. And I help you grow your company and career using AI. So if that's what you're trying to do, it starts here, the unedited, unscripted live stream podcast. But make sure to go to our website, Your EverydayAI.com. Go sign up. up for the free daily newsletter.

And we're going to be recapping all of these stories and everything else. You need you need to know to be the smartest person in AI, your company. All right. Like I said, please don't spend like five hours a day, try to see what's happening in AI and how that's going to impact your company or your business. On Mondays, we do the AI news that matters. If you are new here, Mondays, we do this show on Wednesdays. We go deep, usually doing a demo hands-on.

with one new tool, one new model. On Fridays, we have a new show called Feature Fridays, where we do kind of the smaller features that are going to make a big difference in your workflow. Then Tuesdays and Thursdays, we kind of rotate different shows. All right. Enough of that. Let's get into our first big story. And that's, well, Open AI, at least in the C-suite, is going to look a lot different this week than last week.

That's because Open AI is undergoing some significant leadership shakeups, as several top executives have stepped back for health reasons, marking a pivotal moment for the company's direction and public messaging. So Fiji, Simo, OpenAI's head of AGI deployment, who was formerly their CEO of Application. She got a newer title, but she's taking at least several weeks of medical leave to address a relapse in a medical condition that she had going on,

a chronic medical condition that she has publicly discussed. And CMO's departure triggers a broader reshuffling of responsibilities among OpenAI's tech executives, sorry, top executives impacting their product, business, and operations oversight. So the big one here is their chief operating officer, Brad Lightcap, is moving to a new role, focused on special projects, including a potential $10 billion joint venture with private equity firms to distribute open AI technology and would report directly to, CEO, Sam Altman. Then President Greg Bruckman will temporarily lead the product organization while chief

strategy officer Jason Kwan, chief financial officer Sarah Friar, and chief revenue officer Denise Dressor will split oversight of business and operations. And then you also had chief marketing officer Kate Ruch stepping down to focus on her cancer recovery. So yeah, essentially, I mean, that's not every single C-suite position at Open AI, but that's literally like half of them are either have a new face in the role or the role is just temporarily being shifted around. And this is pretty important because these changes are coming as Open AI is increasing its focus on its core business with some high-stakes dealmaking, which could impact how the company partners with other firms and also how they

expand internationally. But the biggest thing, obviously, is, well, is this going to impact their IPO that is going to be coming this year? So we'll see investors may see this as some short-term uneasiness, but maybe some long-term stability. We've talked about in the last couple of weeks on the show as we were recapping some of Open AI's recent moves. And if you really care about it, we did do a dedicated show specifically on this. It was episode 742. It was called OpenAI's Most Chaotic Week yet Sora's dead. New models are coming and what it means for you.

Pretty good. Listen, I would say if you are a chat chbt user, if your company is a chat GPT user, I think it's actually an important episode to go listen to. So make sure you go listen to that, 742. For me, right, I don't know too much. But if I'm one of opening eyes biggest investors now, right, obviously it's a tight race with anthropic with Google. Some Google news I think is actually going to shake up the industry that we're going to talk about here in a couple of minutes.

So at least having all these things happening at once, right, obviously two huge losses for Open AI, you know, dealing with some health issues with some of their top executives, but then also reshuffling a handful of other people, you know, so I don't know. It seems like we might see some short-term uneasiness, but potentially maybe some longer-term stability once they get this sorted out. All right. Speaking of Sorted Out, Anthropic has finally sorted out their official stance on OpenClaw. And this one is extremely important for a couple of reasons. Well, one, if you missed this a couple of weeks ago, InVity CEO, Jensen Wong, yeah, the CEO, technically the biggest and most important company.

company of the world said that OpenClaw was the most important piece of open source software ever. All right. And for the most part, the majority of people were using Anthropic to power OpenClawe, right? It kind of started almost as this unofficial, you know, partnership with this open source piece of software. At the time, it was called ClaudeBot, changed his name a couple times anyways. But for the most part, what it started, it was recommended that you use Anthropics models.

of people, they kind of found a little shortcut, right? Wasn't really a shortcut, but, you know, using their auth or logging in and using their Anthropic subscription, which at the time, right, was kind of a gray area. And then a couple of weeks ago, Anthropic came out with a statement that didn't exactly shoot it down because they didn't want to piss developers off, right? Because people love especially using Opus 4-6 with OpenClaw, and they were using their clawed subscription versus using the API, right? So a couple of weeks ago, Anthropic kind of came out with this wishy-washy statement saying like,

oh, well, you shouldn't be using it. Well, they didn't even say that directly. But over the weekend, big news, they literally said no more, we're shutting it down. So Anthropic announced that over the weekend, subscribers to Claude Pro and Max plans will no longer be able to use their subscriptions to connect Claude AI models to third-party agentic tools, such as OpenClawn. They did obviously mention OpenClawn by name. So Anthropic said that this change is due to heavy strain such usage puts on their compute and engineering resources, as well as the need to serve a broad user base more reliably.

So the company did explain that its subscription plans were not designed to handle the high volume for tools like OpenClaw and these automated usage patterns seen with other third party agent tools. Right. So yeah, some people were using Anthropics API, right? You see people easily saying that they're going through hundreds or thousands of dollars in Anthropic API credits, but there was kind of a very well-known workaround that most people were using just logging in with their Claude account. And then if they, you know, have a $20 plan, well, that $20 plan is probably not going to get you much. You know, if OpenClaw gets out of bed on a $20 anthraic plan, you're already hit your rate. limit, but I think people, you know, on the $100 or $200 max plans, we're getting pretty decent usage out of this to power their OpenClaw setups, but not anymore. If anything, this is fantastic news for a couple of people. Number one, Open AI. Obviously, Open AI kind of ACWA hired the OpenClaw project. Not really, but they hired the loan developer, the lone full-time employee of OpenClaw.

So, you know, there's obviously that kind of correlates. And, you know, the Open AI team is obviously reminding everyone, yes, you can use your ChadGBT subscription inside of OpenClawn. So I think it's going to mean a lot of new business for Open AI. But also a lot for Google. Yeah, we're going to talk about their new impressive Gemma 4 model, not technically customers, because it's open. It's open source and you can run it locally, but also for the Chinese models that are super cheap, right, via OpenRour. router, something like that, they're going to see a boom as well. So it'll be interesting to see how usage changes on something like open router that tracks kind of all third parties,

you know, API usage at least through open router. So in theory, you might say, okay, well, Anthropics might go up. This might make them a ton of money. Or people might, right, I saw a lot of testing this out. They're like, okay, I didn't do anything. I did this for one day, didn't use it a ton, right? Because it happened over the weekend. And they're like, I spent $60. This is not sustainable, right? Barely used it. I can't be spending $60 on this every single day because with some of the, you know,

some of the Kwen models, some of the ZAI models, right, that would cover more than a month of heavy usage with their best models because they're much cheaper. All right. So it should be interesting to see how this story continues to unfold because I don't think it's done. 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 AI studio. Powered by Adobe's Creative Agent, Firefly AI Assistant lets you start with your vision, 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 Prophemy. 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. Adobe Firefly AI assistant now in public beta. See it today at Firefly.adobie.com. All right. Our next piece of AI news, Slack is getting into the fold.

And could the company that was, you know, once most known for just being a messaging company, are they going full AI workspace? Well, it looks like they are. And maybe they're trying to challenge at least a little sliver of Microsoft's business. So Slack has rolled out more than 30 new AI cap capabilities for its Slack bot, marking the biggest update since Salesforce really acquired the company back in 2021. So yeah, this is by far the biggest update to Slack that has ever happened. So the overhaul turns Slack bot from a basic assistant into a full featured enterprise agent, able to take meeting notes across any video platform, run on users desktop, execute tasks through third party tools, and even function as a light CRM for small businesses.

So yeah, if you've never used Slack or Slackbot, right, maybe your organization is Microsoft Teams, you're like, okay, what the heck is this? So Slackbot was essentially an AI powered assistant that Slack has had for a long time, But for the most part, it was just, you know, to help you summarize long threads, help you get caught up, things like that. But now they're literally turning this into not just an agent that can kind of live across your entire Slack ecosystem, but it can do a lot of things that couldn't be done before, right?

Such as the meeting summarization, now bringing some of these even Salesforce CRM, right, a lighter, lighter weight version of it into Slack and having Slackbot be able to work with it. Also, Slackbot now has AI skills, which are reusable instructions, right? So if you know skills, well, you know skills, right, but they're reusable instruction sets that let teams automate and share common workflows. Then you also have a new deep research mode that allows Slackbot to conduct multi-step investigations that take several minutes moving beyond the instant response model that is typical

of most chatbots. Also, you have the model context. protocol means that Slackbot can now interact with over 2,500 Slack Marketplace apps and more than 6,000 Salesforce app exchange tools. And plus, it can create and edit Google Docs and slides. And then the other thing that I kind of mentioned earlier, but yes, that means SlackBot can now join any of your meetings, whether it's Zoom, Google Meet, or others by accessing local audio, capturing discussions, summarizing decisions, and the logging action items directly in Salesforce's CRM.

So essentially, if you are not a Microsoft co-pilot organization and you use Slack, which is a big chunk, or if you are even a Microsoft organization, but you also use Slack instead of teams and maybe you're not getting what you want out of co-pilot, right? I think co-pilot, FYI, I think I've said this at least once or twice. I was bearish on Kilpilot for like 2025, but 2026, I am bullish. They've had a great couple of weeks. I've been chatting with the team a little bit behind the scenes.

So I know that they're cooking up some big updates here over the next few quarters. However, this is a pretty big play from Slack, right? Literally starting to venture outside of just messaging, right? bring of being able to just the two things that I think are really worth noting it slack bot now being able to instantly you know transcribe any of your meetings that's huge right and then being able to log action items directly into a lighter weight version of sales force CRM right because sales force obviously owns slack so I mean that's big as well so once those features do get rolled out they are on tiered rollouts right some

business users get this before kind of normal paid users. But this is a pretty big play here, right? I wouldn't want to be Zoom right now because Zoom, you know, had an attempt to do this as well. You know, last year they had some pretty big announcements, right, going into a more kind of AI collaborative platform. And that's what we have happening here with Slack. But I think kind of the low key thing here that might actually make this super, sticky is just the, you know, the lightweight Salesforce CRM option and automatically transcribing

your calls across multiple platforms straight into that. Because I've always said, right, I've always said for five years. I'm like, man, the thing that Slack is missing is some type of CRM, right? Because otherwise, like, right, I think, because I used to work with startups a lot. And I'd be like, startups are trying to use Slack as a CRM before the features were there. Right. And then you just spend so much time looking for things, but now that the combination of these new agenic AI capabilities within Slackbot, but also the the call transcription and the CRM, this is, again, I'm not going to say watch out Microsoft, but they're at least coming for a slice of that pie. All right. Now to Microsoft's actual big competitor. That's Google and some big news because

Google has launched Gemma 4. All right. So Gemma 4's release marks a major shift in AI accessibility and capability with Google's new open weight models now available under the more permissive Apache 2.0 license. All right. So here's what you need to know. So Google has released Gemma 4, a updated family of AI models spanning four different sizes from a 2 billion parameter.

edge model that you can literally run unlike a Raspberry Pi and obviously even like older smartphones all the way up to a 31 billion parameter dense model, which that model currently ranks third on the Arita AI Open Model Leaderboard. So the Apache 2.0 license is a major change, removing previous restrictions and allowing broader commercial and enterprise use. So all four Gemma4 models are built on the same research as Gemini 3, Google's proprietary frontier model, and are designed for both on-device and workstation class deployments. So the lineup includes the effective 2B or the E2B and then the effective 4B or the E4B. Those are the edge models for phones and compact hardware.

Then you have the 26B mixture of experts. So that's something that you can run on base. basically any newer laptop, right? Because it's a mixture of experts. So it doesn't take or doesn't require quite as much, you know, RAM. And then you have the 31 B dense models. And that's more optimized for offline use on developer hardware and consumer GPUs. So yeah, the 31 B model that is the one that right now is third.

All right. Yes. All right. If you don't know the math here, let me break the math down for you. All right. So I'm going to actually go ahead and bring up Google's blog posts because I don't know if I'm going to have time to do a show on this. I really want to, but let me just give you the quick rundown of why I think this is actually super important. All right. So I'm just going to bring up for our live stream audience here.

This is just a screenshot from Google's announcement blog posts. And here's why this matters. All right. And I'm going to try to simplify this. So if you don't know a lot about, you know, edge models or offline. open source, don't worry, right? So proprietary models, that's what most people use, right? You go to chatchibt.com, you go to clod.aI, whatever.

Your Microsoft copilot, it uses a cloud model, right? And so you pay for the inference, you pay the cloud, right? Because you're running this, and these are huge models, right? The chat chit, the Claude, the Gemini, et cetera. But then you do have kind of offline models that you can download. Those are open source models. for the most part, anything that's decent is like obnoxiously large. Right.

So I have a little chart here from Google that just shows the kind of model performance versus size. So essentially the Gemma 4, 31B thinking. So you will you will have to have like a Mac studio, right? Like I have a Mac studio. I'm going to be running Gemma 4. You can run the Gemma 4, 26B. That's a mixture of experts, so you're not getting the full power. But even that one, if you have a newer MacBook, right, you should be able to run that one fairly fine.

But this is about one. I got to zoom out here to make sure I'm doing doing my math here. Yeah, I'm doing my math correctly. It scores higher on the arena than models that are about 20 times its size. All right. So essentially now, especially if. If you care about open source, and I do think open source is important, even though, right, frontier models with all their protection are safe to use.

If you're on a paid plane, you don't have to worry about. But there's obviously some huge benefits to having local models. But you are essentially getting a model that is about the same power. I had to look this up, even though I knew this in my head. If you go back one year, right, the best models were GPT 40. It was, I believe, Gemini 25, and it was Claude Sonnet 3-7. All right. A year ago, this Gemma 4 is better than all of those on both the scientific benchmarks

and on the arena scores. And arena scores are user preference, right? So it's all a blind taste test. So y'all, literally, you can go download a model right now. Run it on your machine, no internet, nothing. And it is, it's not going to be blazing fast like you would expect, right? But on my Mac studio, I did pay a little bit more for it. But it'll run fairly quickly.

This thing is better than what we had a year ago than what you were paying for. And now it's free. You can download it on a local machine. So pretty cool. A couple other details. here. So the larger models, all right, so yeah, they do outperform competitors up to 20 times their size on benchmark tests. And the cool thing, all Gemma4 models are multimodal, so they're able to process video images and text natively, while the Edge versions also support native audio

input for speech recognition too. Yeah, we'll share in our newsletter. There's actually, Google has a dedicated app that you can download that lets you run the locally on your phone as well. So yeah, if you're, I don't know, somewhere where there's literally zero cell service, you can still chat with a, you know, obviously the edge versions aren't frontier level, but, you know, they're probably at least like two years old frontier level,

if that makes sense. So the edge models feature a 128K token context window and are up to four times faster than previous Gemma versions, and they use up to 60% less. battery. Also, according to Google, the Gemma models have been downloaded more than 400 million times and already there's more than 100,000 community created variants since their initial release. And I think that number is going to let go bonkers because, y'all, I cannot emphasize enough. This is huge. This is huge. And this, if I'm being honest, like Google just woke up

and chose violets, right? Right before the Easter holiday here in the U.S. They're like, yeah, we're going to put out a model. We're going to open source it. Not only that, but changing the permissions to Apache. Because before, it was a little more restrictive and you couldn't do as many things, right? I would assume that we're going to see within a few weeks, we're going to see tens of thousands, if not hundreds of thousands already variants of Gemma 4.

I do think this is probably going to be the most talked about. open model ever. And this is it if nothing else, this puts a ton of pressure on the frontier labs because if I'm being honest, Google doesn't necessarily need to win with Gemini, right? Because they could put out something like this with Gemma 4. And oh, now all of a sudden this might as an example, right, this might take away, you know, a 10% of the pie from all the other big labs that are getting paid. So Google doesn't even need to necessarily worry about making more money with Gemini. Obviously, they will.

The product continues to improve. But they can literally just create models that are good enough for so many customers. Right. So in theory, Google might be losing money every time they put out a model like this that is a step change for open source. But it might not matter because, right, if it's, that good, if it's that good, it's only going to make their Gemini ecosystem stronger because you're going to get more and more people using it. It is based on the same research as Gemini

3. So for companies that have maybe been on the fence about AI or if they need a local kind of open source model in their stack and they get it and it works, right, then they're more likely going to be using Google's other products. All right. Two more quick stories. I told you, I might go a little long on that one. All right. So next, Open AI has closed a record-breaking funding round as they officially closed. It's $122 billion round with a post-money valuation of $852 billion. So the company's latest round was co-led by SoftBank with also major contributions from

Anderson-Harrowitz, D.E. Shaw Ventures, and participation from Microsoft, Amazon, and Nvidia. So the fundraising surge comes as OpenAI's ChedGPT, now boasts officially over 900 million weekly active users, including more than 50 million paid subscribers. Also

in a release, so OpenAI said that they're now generating more than $2 billion in revenue every month after reporting $13 billion in revenue for 2025, but the company is still not yet profitable and continues to burn a lot of cash. So yeah,

Last reports we saw said that OpenAI did not expect to be profitable until about 2030. So the latest capital raise included $3 billion from individual investors and, for the first time, allowed participation through bank channels, widening access beyond major firms. So Amazon committed up to $50 billion, Nvidia invested $30 billion and SoftBank added another $30 billion with additional funds coming from a broader pool of investors. So Microsoft already a key partner with over $13 billion. previously invested also joined around, though its latest commitment was not specified in the announcement. Yeah, so we've talked about this a little bit on the show, but this was essentially the end,

right? This was the end of the round. We had already heard that it was going to be above 100 billion, but now it is done. It is close. 122 billion record breaking, right? And this means that Open AI is likely to IPO or very likely to IPO beyond the $852 billion, right, because that's their post money evaluation right now. I've said that it's very likely that they're going to IPO past a trillion dollars unless something majorly wrong happens in the next couple of weeks, which you never know with all the AI or open AI drama. It's very possible.

But presumably with all the momentum that Open AI has, I don't see that changing, right? And also, if you are listening to this show every day, you have to realize we leave it in the AI bubble, right? Because you might say, oh, no, like Anthropics winning 2026. And I agree, Anthropic is winning 2026. But still, outside of our little AI bubble, no one, no one's heard of Anthropic, right? Let me just, let me be very honest, right? People know Google Gemini, because now Google is advertising literally everywhere. And, you know, AI, people just think, oh, chat, you.

So for the most part, outside of, you know, your general, you know, people involved in tech, you know, people who follow AI, right? For the most part, people just know Open AI. They just know chat GPT. Oh, that's AI. Right. So pretty big news here as Open AI prepares for its IPO. All right. And our last big news story of the week. Yeah, Anthropic, another leak. This one, maybe not as good, right? We covered their last leak, last week, last week, last leak. Let's try that one five times fast. Last leak last week, right? That's where we found out about their new mythos model

and a couple of other agetic capabilities, but this one not as good. And I'm going to give you like a random hot take after we get through the actual news. So a major leak of Anthropics Claude Code has exposed the full architecture, unreleased features, and internal performance data of the AI coding tool.

So the leak handed competitors a very detailed roadmap of unreleased features and deep technical insights, potentially reshaping the competitive landscape. So according to reports, here's how it happened. While there was a debugging file that was accidentally bundled into a routine update and pushed to the public registry leading to the exposure. So the leaked file pointed to Anthropics' own cloud storage, whoops, and contained nearly 2,000 files,

and 500,000 lines of source code, a lot of that being for Claude Code. So within hours, the code base was mirror and analyzed across GitHub, quickly gaining thousands of stars as developers dissected the material. So Anthropic confirmed that no sensitive customer data or credentials were exposed, attributing the incident to a packaging error rather than a security breach. So the source code included dozens of feature flags for capabilities not yet released, such as session review for improvement, persistent, assistant mode, and remote control from phones or browsers. The leak provides a clear picture of Anthropics plan for longer autonomous tasks,

deeper memory, and multi-agent collaboration features central to enterprise strategy as it prepares also to go public like OpenAI. Here's the interesting part. outside developers have already reverse engineered parts of quad code, which prompted taked down requests from Anthropic. Many of those were originally granted, and then a lot of them were reversed. And here's the kind of spicy or interesting thing, y'all. And I'm just going to go ahead and share a random, since I'm already sharing my screen in another tab here.

I'm going to go ahead and share a random tweet that I just posted. Yeah, as if I don't spend enough time talking about AI. So here's what I said because I think maybe instead of just me riffing it, I'll just read what I wrote. So I said, okay, this is crazy, but hear me out. Anthropic has publicly claimed that Claude pretty much writes 100% of its own code. The Supreme Court, the courts ruled last month that AI generated work cannot be copyrighted. Claude code got leaked. Someone ported it to Python, called it Claw code, got 100,000 GitHub stars, and then clones like this were framed as clean room rewrites or new derivative works.

So by admitting that Clawed code writes nearly 100% of its code, I'm asking here, did Anthropic just kind of shoot itself in the foot? Because I wonder what will happen when or if Wall Street figures out what this means before their IPO. And well, what this means? Well, the Chinese Claude code clones are going to go crazy this month. All right. So long story short. Many senior people from Anthropic have said that Claude writes 100% of their coat, right? And it's been multiple people at Anthropic have said this.

So why does this matter? Well, if you accidentally leak all of your code or if it gets leaked, right? and people take it and they rewrite it using AI, right? That could be considered derivative. Yet, if Anthropics says that they've, you know, 100% of the code was written with AI, and the Supreme Court said that if something is 100% written with AI, if it's not protected by copyright, essentially, Claude Code, right? This is a unicorn in and of itself.

This is one product, but probably Anthropics' most popular product. now in theory, I mean, we'll see it's going to be a lot of work for Anthropics legal team, which we know their Anthropics legal team is always sending out a lot of letters. We'll just say that. Right. But

this is going to be interesting because China, we know, does not usually pay too much attention to copyright. And there is literally versions of this in the wild of the actual code of Claude Code. Right?

So combine that, right, because it's really good. Claudecode is a great program. Myself, I like Codex a little bit more, but Claudecone is great. I use it every day, right? So now that's essentially out there. All of Anthropics hard work is out there and anyone may be able to just use it because someone that has ported it, a couple of people ported it to other programming languages. So between that is out there now and the fact that the Chinese open source models are

getting really, really good. And the proprietary clothes models are just, you know, maybe three months behind. It's not great timing for Anthropic for this to happen right before their IPO. We'll see if Wall Street figures this out, right? They're not digging in the details as much as some of us AI nerds. But regardless, not great news for Anthropic. All right. That's all for the big stories.

But let's quickly go over what's new and what's next. This is just kind of the bullet point overview of still some very important stories that didn't make our top list for the week. Some of these are rumors. Some of these are leaks. Let's get into them. First, Chad GPT launched in CarPlay, an updated version for both your, so if you update your iOS and your chat GPT app, you should be able to access that if you do have CarPlay in your car.

I can't wait to try that out. I don't know why I didn't this weekend. All right. Perplexity computer can now run workflows inside of Slack. All right. Oh, I know why I didn't try that. out yet. It's because I don't like the new iOS. I don't know. There's all these backgrounds and your text messages now. Anyone else out there? Just me. I'm an old man shaking my fist on the porch.

All right. Next, Google AI Studio launched Focus mode, enabling point and click UI app edits. I'm loving that one. Claude released Microsoft 365 connectors, enabling Outlook, OneDrive, and SharePoint access. Open AI just rolled out pay as you go. Codex seats for teams. So you only pay for what you use. Anthropic acquired coefficient bio for $400 million expanding life science AI with Genitech talent. Notebook LM's quizzes now show topic summaries. This is pretty cool. Also, study suggestions and a regenerate tool for focused practice. So yeah, you don't even have to reprompt anymore. It just does it for you. Microsoft is testing co-pilot advisors, which lets multiple AI personas debate topics and then it's summarized afterward.

The lead for Google's AI coding agent Jewels joined OpenAI's Codex team. Perplexity's deep research can now generate structured outputs like presentations, spreadsheets and websites. Perplexity computer released a feature to help users with their tax returns. Anthropic is testing Conway, an always-on-clod agent with extensions, webhooks, and browser integration. I can't wait for that one. I don't like the, you know, how Claude code uses or co-work uses the browser.

The new Quinn 36 plus preview reached number eight on Code Arena. A new image model was briefly tested on Arena and most were suspecting it's OpenAI's new GPT Image 2. So even though OpenAI killed Sora, it looks like they may be continuing to invest on the image side, AI hiring startup Merker reportedly suffered a data breach via its light LLM that may have exposed up to 4 terabytes of sensitive data from experts and clients like OpenAI, Anthropic, and meta. Speaking of meta, they launched a new Ray ban in Oakley AI glasses with prescriptions,

nutrition tracking, and WhatsApp summaries. A new Quinnipak poll finds that 55% of Americans believe AI will do more harm than good, citing job and education concerns. I agree with that one completely. Space X confidentially filed their IPO paperwork, reportedly targeting a June public listing. So I'm guessing that we'll see more traction on OpenAI and Anthropic around quarter three or four.

Google Workspace added Gemini to rewrite docs in users' writing style. Anyone else noticed this just pop up? I'm loving that one. Gmail launched AI inbox, but only for, for AI Ultra subscribers and beta in the US. So, all right, can't wait for that to just go to all paid users. Perplexity is now facing a lawsuit over alleged data sharing in private mode. Not good.

ZAI launched GLM 5V turbo, a native multimodal vision to code model, and OpenAI acquired TBPN, the team behind the Silicon Valley Tech Talk Show. All right, it's a lot of news this week. And if you missed any of it, don't worry. Spend your Mondays with us. Stop worrying every single week. Oh my gosh. What am I going to do?

How am I going to learn AI? What should my company be using? That's literally why I exist. This is all for free. I try to give you the unbiased, maybe with a little bit of opinion, a little bit of bad humor, every single day, Monday through Friday. So if this was helpful, please go to your everyday AI.com. Sign up for the free daily newsletter.

Thanks for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all. Meet Firefly AI Assistant. 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. 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 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.