Meta's New Model Unveiled Amidst Character.AI Bankruptcy

0:00 podcast. Today, we have a big announcement from Meta, who has just released their very first frontier model. This is something competing directly with OpenAI and Anthropic for the first time ever. This is, for me, really exciting because we have a new player. Meta's always been there with AI, but they've really lagged behind. They've focused on the open source stuff. They had a place in the market, but it wasn't anything that I was actively using. This has changed. I'm actually actively using and recommending Meta for certain circumstances. And with this new model, I'm actually excited with some of the stuff that's... that's coming out here. So we'll be covering that on the show today. Also, there is a Ukrainian drone

0:33 strike that has knocked out three different Yandex data centers, which were training Yandex GPT. It's crippled a lot of the Russia's digital economy, but we'll get into that. And mostly from a perspective of kind of how some of these data centers are a choke point for what's happening with AI. Also, we have Character AI that has just filed for bankruptcy. There is a wild story, if you have been following along what's happened with Character AI from the people leaving Google to go started and then going back to Google and everything there. And as far as all the acquihires that have happened over the last number of years with AI, this is an interesting story that applies to a bunch of different startups. Also, Epic

1:10 Games is going what they're calling, quote, AI first. There is an unannounced Disney shooter. The launch has been delayed and they're saying they're going to go all in with using AI to finish it up. At the same time, we have Amex that has just introduced what they're calling a new framework to prevent AI agents from making unauthorized.

1:28 purchases. This is something that I think is very needed. And when you have these tools like Meta's Muse that can go out and actually make purchases for you, this is something that we're going to want to pay a lot of attention to. We also have a whole bunch of AI labs that are buying up dead startups data to train models. This is something that's incredibly valuable. I personally have been approached by a number of companies to do this. And I am doing a lot with data licensing.

1:50 So I'm going to get into where I think the market is for all of that. Before we do, I actually have a school community I wanted to tell you about where I help people grow and scale and start businesses using AI. So if you have a business that you'd like to grow with AI, or if you'd like to start a side hustle, my goal for this AI hustle school community is for you to be making if it's a new project, $10 ,000 a month from a business that is run and automated by AI. We have a bunch of different channels and kind of strategies that we would recommend. We have building apps. We have creating a social media like content creation type business. And then we have a number of other things that we've have done. So inside

2:28 the community, this is stuff that me and my business partner, Jamie, are actively working on and building. So we show you the workflows we're using, and we also show you the revenue numbers from our businesses. And then we show you the strategies we'd recommend to do this. I think there's over 100 videos that we've recorded. We have a bunch of different like courses or subcourses inside of there, just depending on what your strengths and weaknesses are. We have an assessment. So you basically go in. Take the assessment. It'll tell you what you're most likely to be successful with. And by the way, if you already have a business and you're like, I already know what I'm doing, we have a bunch of courses on growing, scaling

2:58 those businesses with paid ads and free ads or free marketing using AI. So there's content for everyone in there. It's only $19 a month. We're trying to make this affordable for people, not make this like a $300 tool for growing. But we really believe you get a lot of value out of it. And I'm super excited to see what you create.

3:16 If you join it, we have about 1300 members right now and everyone's kind of sharing what they're working on. So it's an awesome community and I would love to see you in there. I'll leave a link in the description. All right, let's talk about what Zuckerberg is saying. He just said, Introducing Muse Wonder, our new family of frontier models with state -of -the -art performance for personal and professional agentic work. If you remember, they had Muse Spark before, which is kind of their low -level version. They have Muse Wonderful Max. Haha, wonderful is the name of it. It's two separate words, obviously wonderful.

3:45 I like it. They have Muse Wonder, and then they're actually comparing this to Claude Fable 5 .1 and GPT -6 Astra. and Claude Opus 5 .5. And it's actually pretty comparable on a bunch of different benchmarks. On the personal super intelligence index, it is at 59. It's beating out everybody. On the Harvey Lab AA, it's at 13. It is beating out everyone. The next closest is, and this is their wonderful max, by the way. So anyways, they kind of have like their top tier, and then they have like a slightly lower tier. Their top tier is smoking people on a bunch of different benchmarks. Obviously, this can be cherry pick for whatever benchmarks. They're the best at one that I do think is a pretty good benchmark

4:22 though is the deep SWE 1 .1 and they got 79 % on that so the reason I bring that up is because basically what they have said they focused a lot on with this particular launch is Creating a tool that could deeply understand information about you. And I guess that's probably a Facebook specialty, right? They've got a file on every person and they're generating as much data on you as they can. I'm sure my wife hates that about meta, but it makes it incredibly useful for an AI tool where it understands everything you've said and knows what's important and relevant.

4:52 This is basically how they serve you ads. And you can imagine that the data you put in here, I can't be for certain, but you would imagine that this data is also the greatest way to collect data. feed it into their ads engine to understand more about you. So what's interesting is they have their Metamuse tool, which is kind of like ChatGPT work or the Claude cowork. And I have it set up to every morning. I have it running through all of my emails, sorting my emails, deleting the spam, deleting ones that aren't supposed to be certain places, forwarding emails to team members. And then it goes and writes a draft reply to every single one. And I go review and, and I, you know, instructed on how to do

5:28 this. I actually, if I'm being fully honest. love this every day. It was very hard for me to keep track and keep on top of all my emails. I have a lot of projects going on, a lot of partners and things that people are working on. So staying on top of everything was pretty tricky. This has finally, for the first time ever, made it feel doable. So I'm really excited that they now have this frontier model that is going to be running all of that. I will say when they show their like top of benchmarks for their like wonderful model, that is only if you are paying subscriber to Muse code, their like coding platform, which is, you know, competing with check or cloud code or, you know, any of these other subscriptions,

6:02 but you still get Muse, you still get their like app that goes and checks your emails and all that. That's all free. It's just, if you're going to be doing coding and some more heavy work where you burn through a lot of tokens and you need that top tier intelligence that you'll get that. But for most everybody, I would say this is an awesome way to get started for completely free.

6:20 Okay, the Ukraine has dropped a bunch of bombs and knocked out a bunch of different data centers in Russia. This is disrupting payments, maps, cloud services all over Russia. Of course, the... russia and ukraine are escalating things in a lot of different departments there's all of the political drama with the ukraine i think targeting oil refineries in russia and then trump had to deal with russia for oil refineries and he's like okay there's a there's like an energy ceasefire russia and ukraine have agreed not to shoot at each other's energy whatever so there's like stuff geopolitically going on there. But I guess data centers were not in the ceasefire. So they took out some data centers. One of them is in

7:02 Cherveskensky. That's 29 petaflops. Those taken offline. Another one is in Lyapunov. 20 petaflops were taken down. And then one in the Lucian, which is 20 petaflops. Basically, these are a bunch of supercomputers and they're running on NVIDIA A100 GPUs. So like these are very big data centers, very expensive facilities. And with all of that, Yandex, which is kind of like Russia's Google, because I'm assuming they didn't let Google fairly compete in the market, although I don't know that. deep history of Yandex, but it's the big Google in Russia and their search engine has about 70 % of the Russian market.

7:36 So it's the main thing and it was hitting them. But just like Google, we have Google maps, we have Google Gmail, we have YouTube. Yandex did the same thing where they created all of the software services. I must be honest, China has the same thing, right? With Alibaba and stuff, they got a couple of the big winners and they own a ton of different things. So hitting these data centers was bad for a lot of reasons, but also Yandex was in the middle of training Yandex GPT. They're trying to like, like their look, we got off. most of the American technology for most things, but they were still using our AI models because we're the frontier AI models, just like China. They're trying to train their

8:09 latest Russian AI models. And I think this is interesting because whether it's China or the United States, at the end of the day, when people talk about AI safety and AI alignment and how AI is going to take over the world. I mean, all it takes is for a data center to be down and like the whole service can't run anymore. And like there is all sorts of literal ways to unplug the AI. So I don't know when people are overhyped on the dangers of AI and how it's going to take over the world. If you take out the data center, basically Yandex can't be training their stuff anymore. And I'm not advocating for terrorism, but if there really was a world ending killer robot AI, there's a... unplug port there's an

8:50 unplug for all of them so in my opinion this is why the ai fears are grossly overhyped people are in control of these things people are creating these things and most importantly people are funding these things these data centers don't run without absorbent amounts of electricity and resources and money and so all this has to be paid for okay let's talk about what's going on with character ai this is not a great not a great story for them this is five years into when they were starting Which if you guys remember, this was like the hottest, one of the hottest startups. Everyone was using it. But I think it was a lot of like high school students just having like really long chats with them. There

9:26 was a lot of. There's a lot of drama about like not letting high school students use AI and like regulation. Anyways, they are now turning this down. They said that they had assets and liabilities between $50 million and $100 million. Their CEO, Colin Adams, made a statement and said, over the past year, Character AI is taking meaningful steps to sustain and improve its business in the face of significant illegal regulatory and operational headwinds. After a comprehensive review of the available alternatives, the board determined that This court -supervised sale process provides the company with the best opportunity to maximize the value for all stakeholders. The startup has agreed to sell some of their assets

10:02 to Disney for $20 million, as well as a non -exclusive perpetual license for its technology. And we already know that, I don't want to say they got gutted, but basically Google came in and basically did an acqui -hire of their former CEO. who was over at google before he started this company and they pulled him over and some of the tech but they left the company and so it seems like the company's just teetered out as basically if you had went to it in the past it was like a way for you to go and chat with it was like talk to einstein and there was a bot that's like respond einstein would respond it would talk to you like einstein so i think a lot of those things were novelties and while it was very popular

10:37 and had a spike in users at one moment it wasn't like durable a durable product you look at cloud code and it's people are using this to build software and websites are spending billions of dollars on it because it's helping replace what they would have spent on developers or developers are using it to replace the time they would have spent developing code and developers cost hundreds of thousands of dollars a year so you see a very big economic incentive with character ai just chatting with einstein wasn't really doing anything other than being novel and so it seems like that that got at the heart of the company and they weren't able to sustain the growth All right, Epic Games is going AI first. There's an unannounced

11:15 Disney shooter game and they have delayed the launch. The developers were told, quote, no more handwritten code and that all of the art would be generated with full AI workflows. This is all according to a report from Bloomberg. And this is... going to be a Marvel and Star Wars characters game, according to a report from Bloomberg, is a shooter game. The director, after he kind of, the game director, after he told all the staff they weren't allowed to use any more handwritten code and that they had to make all of the art with AI, a lot of people were upset about that.

11:47 He told them they're all specifically going to be using Claude. The project has basically slipped from their original March release window to now. So I think he's just like, look, this thing is super delayed. We were supposed to release this thing in March. It's now October. We've got to get this thing out. I'm sure he's got people breathing down his neck, his boss. And so he's just like. Just whatever you do, let's crank this thing out. We got to get it done as fast as possible. And I think it gets to an interesting into where obviously if he's telling people now they can't do handwritten code. I mean, to me, that sounds wild because I've been using obviously AI code in my startups for the last three years.

12:19 But there's a lot of people that do resist. And I talked to a lot of people that are working in organizations and I'm like, you know, developers. And I'm like, hey, like, are you using AI for everything? Like, oh, your company doesn't really know or care or whatever. To me, it's wild because I feel like once you try it, you get like. red pill then you can't go back but there's a lot of people that don't want to try it and they're just happy living how they've always done things and this isn't just for code this is in so many industries and all of a sudden what happens is the companies around you start competing or you fall behind on deadlines or your revenue's bad and it gets to a point where if you So basically

12:50 my recommendation is like, choose to use the progress or choose to use the efficiencies that AI can give you when it's your choice, because at some point it won't be your choice. You're not going to be able to compete with everyone that is using it. So before you get forced to do it, because your company's in a bad spot, like just choose to do it. And this is kind of where they're at. I mean, Mr. Release window by so much. Disney has invested $1 .5 billion into Epic Games. They did that in 2024 as part of the multi -year collaboration. Epic laid off over 1 ,000 employees earlier this year following a downturn in Fortnite engagement, which has been happening since 2025. So they have a bunch of their games that are getting less use than

13:26 before. and it looks like they're now going to move into ai and i'll be curious to see you know the quality if people are upset about quality or have concerns about quality i mean they've already spent so much time and money getting it to this point it was supposed to release in may and we're already in october i'm assuming they're just doing the finishing touches on the game at this point and then using ai is going to be 10 of the whatever 20 of the game is going to be made with ai so i don't think it's going to be distinguishable for how from however they were doing in the past but It is a precedent so that their next titles will probably be using this exact same AI mandate moving forward. So

13:59 I'll be curious to see how much faster they can get stuff out in the future. Okay, Amex has just introduced a framework to prevent AI agents from making unauthorized purchases. And there's an article over on TechCrunch and they say, personal AI agents are increasingly making transactions on their users' behalf. But given the non deterministic nature of large language models, AI bots are prone to doing what they're not supposed to, which in this case raises the possibility of AI misfiring and making unauthorized transactions.

14:26 To get ahead of this problem, American Express is introducing a new framework designed to capture and authenticate a customer's purchase intent before an agent completes the transaction. The idea is that there should be a way to ensure that the customer's purchase instructions are carried throughout the transaction lifestyle.

14:41 American Express says from the purchase to everything that follows. So they're calling this the intent intelligence framework. And they said that they hope this is going to be kind of basically an industry standard. I like the concept here, right? It's like, what is the intent of the bot? The intent of the bot is if I tell my bot, hey, I need new headphones. Mine just broke. Can you go get me whatever the best rated headphones are under 100 bucks? I don't really care. I just need them just next week. as soon as possible or something like that. And it goes and buys them. Amex can see the original system prompt and it understands what the intent of the agent is. The intent of the agent is not to go buy

15:14 me $500 headphones. The intent of the agent is not to go buy me $10 headphones from the dollar store. I want something like in a specific range and I gave the specifications. And so it's going to basically before it allows the transaction to happen, it's going to allow it. It's going to make sure that it's following through with my original prop, which is cool. If there's a way to auto enable this or have this set up.

15:33 i think that a lot of people would find value in just having a financial institution before a purchase is made verify the backlog of thought and reasoning and the prompt that went into the bot that's about to make those purchases. And of course, like some people are like, I would never let AI make a purchase for me completely autopilot. And I get that, but some people will.

15:52 And I think it's good that they're making this technology available today. They, they're the VP of strategic partnerships and strategy. Hannah Bozden said these new transactions will have, will have a consumer, a merchant and an agent. And she said that they will be seamless, trusted and secure the way our card members expect them to be. So I'm curious to see how this works in practice once it's rolled out, but I'm excited about it. And a big shout out to American Express for putting in some time and money into figuring out this problem, which is. definitely plaguing the industry in a lot of ways. Okay. AI labs right now are buying up the data from old quote unquote dead startups to train AI models. This

16:32 is genius, right? Like if you have a company that's, they're like, Hey, we ran for 10 years and we were a stack overflow and we had a bunch of code that people would write questions and answers in code. Yes. Licensed stack overflows data set. Maybe people aren't going to their website anymore, but all of that data is incredibly useful for AI models. And if you have a company that maybe they were doing filters on photos, and there was an app to help with filters on photos, and now people are just using AI for it. Go and get all of the data on how those filters were made and train them into the AI model. There's so much information and knowledge that is quote unquote dark knowledge. It's just not publicly

17:08 available on the internet. These models can't go and scrape it, nor do I think we really should be scraping it at this point. I think people should be paying for the data that they're getting, especially when it's coming from private companies. And so with regards to all of that, I think that this is a phenomenal business model. It is called Simple Closure, the company that's doing this.

17:25 They have an asset hub. Apparently, they've brokered over 100 deals since April, and they expect to return more than $10 million to startups that have already closed by the end of the year. I've actually been approached by companies interested in doing this with some of my startups, some of them that aren't even closed. They're just not like, they're just autopilot startups that I have, apps or different things, and they want to license some of the data that's been created.

17:46 There is a lot of... there is a lot of stuff that is tricky, right? Like if you have user generated content, where like your users made content, you know, that has to be explicitly outlined in your terms of service. And, you know, your users all have to consent and opt in. So like, there's a lot of things that do make this more tricky. But if it But maybe it was part of your terms. Maybe you already had people opt in. And so you're good to go. People have long forgotten about the startup, but you still have the data available. So it's very interesting.

18:12 Google apparently outbid Merkur in the acquisition of Spirit Airlines and just buying all of their data. There was 100 million emails, 500 million Teams messages, and 176 ,000 employee records. And Google was... i think bid 10 million dollars for that 47 of simple closures shutdown deals now include code or data sales which is up from up from basically this being super taboo earlier on So this is pretty wild. Startup shutdowns are rising between 26 and 56 % annually per Carta and AngelList estimates. So this is something that could make a lot of money. Keelan Miskell, he got about $25 ,000 for his crypto startups code base and expects about $15 ,000 more from workspace data from Google, Notion, Figma, and

18:56 Slack. So they're just selling. I think in this case, right, the workspace data is really the good one because it's what your employees generated and that's just auto opt -in. So you have that. under your control all right guys thank you so much for tuning into the podcast today if you enjoyed the episode make sure to go check out the ai hustle school community if you want to grow and scale a business using ai and i'll catch you in the next episode