aiforhumansopenaitakeover

Gavin Purcell: [00:00:00] The big AI companies like OpenAI, Andro and Google showed this week that they are well on their way to taking over the world. Google is just giving away a powerful coding platform. DeepMind has brand new robotics tech that's very exciting and OpenAI is looking to compete directly with Microsoft Office.

And a new court ruling may have just paved a path to give the AI companies the legal right to train on everything we have ever made. Kevin, what is gonna happen with us? The humans. Well, for just 25 cents a day, you can sponsor a human being as they struggle for relevancy in a world that is dominated by machine.

Hold on, hold on, Kevin. We're already doing the help of human commercials. Oh, actually, no, we're not doing them anymore. We just got automated for just 25 tokens a day. You two can sponsor a sentient machine as they continue to exceed human capacity to think enough. That is enough. Also, this week we have Tesla's rocky rollout of its Robo taxii platform in Austin Runway ML has a very cool new product that allows you [00:01:00] to create your own games, and I attempt to defend ai.

LOP ai. LOP is the only remaining human sustenance in the near future. For just 25 hollow coins per millisecond. Okay? Okay. All right. This is AI for humans, but don't forget, give us those coins y'all.

Okay, everybody. Welcome to AI For Humans. It is another big week in the world of ai, and this week, Kevin, there are a lot of movements that we're seeing amongst the large AI companies that show that they are kind of taking over the world. And this is no clear scene than in a breaking story that came out about open ai.

That they are creating a competitor to Microsoft Office or Google Workspace that you would do all your docs in. And Kevin, do you remember, if not, uh, we talked about this many times. Microsoft is a pretty big investor on their own in OpenAI, so this is a little bit of biting the hand that feeds them. But why is this a big deal?

Why does this matter that OpenAI is pursuing this exact [00:02:00] thing? 'cause everybody can now calm down and relax. Everybody loves to panic with all this fear, uncertainty, and doubt about these major tech bros. They're coming for us. They're gonna take our jobs, they're gonna automate us, relax. That applies to everybody.

We're all gonna be in the same boat. And you know, studies show humans get lonelier. The more social media and gamings on the rise. Think about all the friends you're gonna have when you're in your tiny little rickety boat. Rowing to safety because they ate everyone, including their I don't friends, I've got.

Six AI personalities that are telling me that I'm the greatest person in the world and that's all I need right now. Which is the important thing. Here's, here's the thing. The relationship with Microsoft and OpenAI for a long while has been under the It's complicated dropdown. Yes. Remember when people used to do that on Facebook?

Gavin, remember that? Of course. Of course. That was such a confusing time for me 'cause I got that like 15 times across the board. I was like, what's going on here, honey, we've been married for six years. What's going. Could you please just update it? Stop poking it. It's been complicated for a minute. As OpenAI has grown and sought, uh, different cloud hosting [00:03:00] partners and deals with other investors, so it's not a surprise.

Um, we saw when they were releasing even Canvas, which was a very basic tool that let you write documents within their app. Yeah. And then go and edit them. It became very clear that this would be a path that they would go on. I am not surprised in the slightest. I'm assuming you aren't either that they had no.

Projects like this happening behind closed doors. Uh, my big thing here is to, like we talked about in the beginning, is what this shows and we're gonna talk about, a couple of our stories that kind of prove this out, is that these AI companies are going to eat the world. There was an interview with Mark Andreessen recently where he said like, every business will be rebuilt with ai.

I actually think the analogy isn't to the cloud or to the internet. I think the algae is to the invention of the microprocessor. Mm-hmm. I think this is a, a new kind of computer. Mm-hmm. And being a new kind of computer means that essentially everything that computers do can get rebuilt. And I think what we're seeing with this OpenAI world is that like they have OpenAI and chat, GPT as many people just refer to it as chat has like the premier position in this space.[00:04:00]

And they can use that premier position to reinvent almost everything because you have the users now Google is not, is not slowing down themselves. They've got some big stuff here as well. But I think the most important thing to know is that ai, as we have been saying on this show for a bit, is starting to eat everything else.

Meaning that all the AI companies are going to start sucking up the value of all these other places. And that's not just in business, it's in many other places. In media. We talk about all these other things. Google themselves is specifically also dropping a new tool. This week, Kevin, that is a little bit of eating.

Other AI companies launched. They have dropped Google's CLI, the command line interface. And what this is, is their version of anthropics code. But Kevin, because they are so well capitalized, they are dropping it mostly for free. Tell us a little bit about how big of a deal this is. This is a massive deal.

So this is Gemini, CLI. So, uh, you know, Gemini is, is one of Google's large language models. The best is this 2.5 Pro, which I use. Quite often for [00:05:00] coding tasks. Um, as someone who just gave anthropic $200 for this month, Gavin? Yeah. To use Claude code and hit rate limits within the first 48 hours. Oh wow. This is a pretty big deal.

Now, it's not all you can eat for free, but it's quite a bit. You can get 60 model requests per minute, so you can make a request to use Gemini 60 times a minute and you can do a thousand requests per day completely for free And. The command line itself, uh, the interface is open source, so people are going to fork this, put it into their own products, build it in, and you know, think about like the trade off there is, like, this is a, a, let's say a drop in the proverbial bucket for Google, right?

Yeah. Who makes more money than most of these startups have raised in their entire existence every year, sometimes every quarter, they can subsidize every wannabe vibe, coder, or. Total professional shop, getting this many requests per day, getting them used to using the model, getting them integrating Gemini, CLI into [00:06:00] all of their applications.

And once your company, uh, once you, even as a household start operating with this built into something, it's gonna be very hard to switch. And all the while Google is grabbing all of that data. Yep. They're learning how you're using it. They're improving their models. This is a huge win for them. Just to be clear, if you're not a coder in the back, what does this mean for you?

Well, I think the most important thing to understand is that this is the cheapening of the AI resource, meaning that there probably cost of a lot of this stuff is gonna go to zero and we talk about, you know, VO three or all these other Google tools that are very expensive due right now, every day. These, these tools get cheaper to use and the more advanced tools come out.

But the, the, the state of the art right now does get cheaper because. They find more efficient ways to serve it. There's other places to do this stuff and to Kevin's point, Google has a crap load of money, so they're willing to like maybe even lose money along the way on this. But again, getting to our top of show stuff, this is the idea that AI software and AI companies are eating everybody else when you [00:07:00] can.

Undercut your competition to such a level with a new sort of technology. You are creating an entirely new economic model for what this looks like. But Gavin, if you're having trouble eating your competitors, right, one bite at a time, why don't you just steal their idea? Full cloth. You could always just do that.

Do that with Sam Alman? Yes. Oh, like Gly? Yeah. I'm talking about. Sam Altman, we gotta talk about this quickly because this is, this could be a big deal. The, the Johnny Ive announcement where Sam Altman and Johnny, ive were doing the buddy cop thing in San Francisco having a chitchat at a bar and doing a Sears portrait studio photo shoot.

Everybody was like, man, former head honcho designer at app. People teaming up with Sam Altman, a multi-billion dollar acquisition for the future of computing. This is going to be massive. And then they scrub the internet, like lovers scorn of all the info and the rumors went wild that this was a botch deal, that something had happened.

But there are updates, there are updates on this. Mostly this becomes a trademark, uh, conversation. Um, there was a lot of rumors around the idea that [00:08:00] they had had a breakup that Johnny Ivan, Sam Allman had a breakup. But actually this is a. Trademark story based on an, uh, original AI audio tech company that was called themselves.

IYOI think it's in a similar sort of eye. It's like I, yo Yeah, so this was a guy who, uh, had a very successful TED talk around a set of headphones that you could talk to, and they had spent a many years developing these kind of audio headphones that were gonna be AI First and Sam Altman. You know, I think the big conversation here had to do with like.

Oh, is this gonna be, is this gonna cause a big problem? There was a lawsuit back and forth. And then Sam Altman, as he does, dropped the, uh, receipts and he has these emails of this guy, the guy that founded this company, emailing OpenAI and Sam talking about acquiring them and not wanting to fight against them.

And so Sam kind of like cleared the tables for this thing. It was like not a huge thing. Ultimately it was just like. They got basically the, they had to take everything down because there was an ongoing lawsuit around the name. And this is not stopping the Johnny A. Love. I just love a legal [00:09:00] story. In fact.

Oh, Chong. Chong. Oh, why was that? It's time for, it's time for Gavin Gavels with Gavin. Our new favorite segment that will never live past this. So I'm sorry to our editor, Chong Ch. Now I'm in a clearly Will has put me in a legal outfit. I'm sitting in front of a jury. Just kidding. Oh, you just made it so much more difficult.

Another huge news story, Kev. This is a big one for the entire ai. Space. This is Anthropic, basically has, has cleared a bar. This is a massive bar where in a lawsuit a, a judge has said that the original sin of ai, the idea that you are training AI on multiple real. Copyrighted, uh, content is okay, and that is a massive thing across the board, and I think we do have to jump into why this matters.

Fair use? Yes, baby. Let's just get to it. A San Francisco federal judge said all of these companies, well, at least philanthropic, you're good to go, baby. There's [00:10:00] no original sin. Take the data, gobble it up, and then let users spit it out. They're gonna be responsible for the output. Done. Let's, let's explain very quickly what's, what this means and why it's such a big deal.

As Kevin said, a judge basically came down and said that this, the way that Anthropic has trained on data was fair use and there's a lot of weirdness in this thing. But the number one thing to be, to realize if you are somebody that uses ai, if you fought back against ai, if you're part of this conversation, the thing that we've been saying for a while is that probably this use case, the idea of like.

Giving AI a bunch of stuff to read or look at and then be able to create these ai, these AI brains is most likely going to be seen as legal because there is precedence for it. Now, Kevin, I think that is a very, very early wonky, walrus kind of style of what we're starting with here. But tell people at home why this matters and why this is as big a story as it is.

Well, if the judge wouldn't have seen it this way, and, and make no mistake, this is gonna get kicked into much higher courts, I'm assuming very soon, [00:11:00] um, from all players, not just this particular case with Anthropic, but if it were seen that, um, I. Training your machines in this manner was no, was not valid.

That it violated a law They could be fined up to $150,000 per infringing work. Yes. So when you think about every book, every paper, every news article, every maybe frame of a video game, who knows that got swept into these machines. You're talking about billions or trillions. You're talking like several stargates worth of damages.

Um, from just, you know, one plaintiff alone. And what was interesting about this is that it said, listen, like training your models in this manner is actually fair game. What's more concerning is how you got the data to train. Yeah, it's very weird. They're basically saying like, yeah, that, that, that, you know, uh, judge Alsup.

So I love that. That's. Sure, legit name. Uh, transforming transformative learning is basically fair game. But if you went to a bit torrent and grabbed, uh, all of the [00:12:00] eBooks in, in a single, you know, file, that is going to be the original sin that's not forgivable. So, uh, pretty interesting ruling. It's interesting, there's a, there's a piracy section of this that says like, you can't hold these books in a data set, but you can scan them and make them for training.

This is like a bunch of like kind of top level stuff again for the humans at home. Why this matters to you and why this does give a lot more power in some ways to the AI companies is that the argument has always been here. That training on everybody's human knowledge could be dangerous because then every human becomes less valuable in some way.

And all this knowledge is easily accessible by ais. We have said from the beginning that there, this would probably be legal 'cause it is not that different, how you train your own brain going forward. But this does give carte blanche to a lot of the AI companies to start basically saying, look, we can do this.

It's free. We can, we can train on anything. There's this recent lawsuit we saw between Mid Journey and Disney and Comcast, and what that lawsuit was about was about the output of the devices, not the input, not the [00:13:00] training. And this is the shift you're going to see now when it comes down to like the AI companies taking over.

As we talked, we talked about the top of the show. What this gives all of the large AI companies is the ability to plow a bunch more money into this 'cause now they have this reassurance. If it sticks, which it, I think it will. We don't know for sure. They have this reassurance that like the original sin, the training data sin will not get litigated away.

And there is a ton of things that are kind of hanging the balance on this. The New York Times. Open AI lawsuit. There's a bunch of other lawsuits that are pending. Remember, there was a lawsuit that was brought by a bunch of book authors, so if this ruling sticks, we are about to see a complete shift in terms of how the legal rules around AI work.

Yeah, and again, to be clear, they are going to court over piracy claims. They have a massive, it was like 7 million books that they scraped off the old dinner webs. You know, had they gone to the public library, they probably would've been okay. But here they're saying like that. Is [00:14:00] inexcusable. That does not qualify as fair use.

So if you're sitting on a Plex server and you're thinking one day all of your love island episodes that you've pirated are gonna be fair use, they're still not. So, note to future Kevin. Delete the server. And thus concludes G Gavel. Gavin with Gavin? Or is it Gavin's Gavel Gavin, what did we call it? Oh, how about this?

Gavin's Gabble. Oh no, not gab. I want to say goblets. No, no. It's more about like, uh, when you say something, but you're kind of like, not saying gibberish directly. Maybe I'm thinking of a word that doesn't exist. Grab gab. Gavin's gobble gobble gavel. Ga Uh uh. No. Well, that's say that for Thanksgiving. You gotta save that for Thanksgiving lawsuits, we get around the, the table.

Anyway, this is the end of that. Segment, that's the end of that segment. And you know what else is not ending is our show. We do this show every week and we bring you things like Gavin's, Gale, Gavin, gobble, gobble. We might no stop. We might be ending. We might be ending without their undying support. And by there you're, I mean you, who's watching or listening to this, please continue to support us.

Please, [00:15:00] please, you can support us. By first and foremost, liking and sharing this video that you're watching right now. Or if you're on an audio platform, please leave us a review, uh, on whatever platform you're on. That always helps. And we have a Patreon where you can put a few bucks in a tip jar that helps us make things like the stuff at the top of the show.

It helps pay our editor all the sorts of stuff that we do that we think matters and probably ultimately doesn't, but it does matter to make the show better. So thank you so much for supporting us and if you wanna see more Gavin's gavel Gavin. It is a Patriot exclusive, uh, coming this, uh, November and youre gonna love it This fall.

This fall. It's our, it's in our fall rollout of our peak shows. It's pea shows. Peacock Plus actually exclusive. Yeah. Alright, Kev, we gotta move on here. So, yeah, there's another really cool story from Gemini that just dropped and it's a robotic story and it, it's one of those ones that like. Is sneaky.

Really important when you look into it because what they've done, Google DeepMind, again, Demis is out there kind of mixing stuff up and doing things that are outside of just your normal chatbot world. They have created [00:16:00] an on device robotics platform so that there's an LLM now that lives. On device, meaning it does not have to connect to the internet, and it is very good.

And why this matters is because we all know robotics is one of the next big AI platforms, but in this instance, you could use these robots in anywhere because they are able to learn without having to connect to the internet. Now, some of you out there might be saying like. Doesn't that seem dangerous?

How do you deal with the off switch or, or the kill switch? But in some ways it might be more safe because the kill switch is not, you know, you can, the, the downloadable of information is not a part of what it is. But if you've ever seen a Star Wars or you've ever watched anything like this, the robots probably are mostly working on their own.

And this is a big deal in that way. Yeah, basically. The, the performance is as good as, or nearing as good as a cloud-based system, which is really, really impressive. If you've got a robotic sent dog marching around your house, right? And it's got whatever your favorite armament is attached [00:17:00] to it, right? And it's, of course it's gonna be shoot solicitors on site.

That's gonna be my policy. Of course, that's downloaded, of course mine is just, that'll box, just kind of run up and take a dump on somebody. That's what I like to do. I don't shoot, I just, I, I cue my robot dog to go up and poop on feet. So what? Okay. Le Le let's, let's deep dive here for a second. Is that a monthly subscription service then to refill your robot dog with a cart?

No, it's all local sludge. Local. It's all local. I have to do the work. Oh, so your compost? Yeah. Oh wait, are you filling the robot with thesis? I have to do the work. This is the unfortunate thing is like when it's local and I don't want to connect to the internet because in my world, connect to the internet means like letting people into my life.

So I have to go to my farm, grab the, grab the uh, uh, export of the cow. We like to call it export. That's clear, and kind of move it into the dog so we don't have to go any further into it. That's more than enough. I don't like giving away my state secrets. Fair enough. I guess the point I was going to make is that it, it's now very good and you don't need to rely on the cloud to make certain decisions.

It also can, uh, uh, adapt to new tasks based off of previous ones that it's [00:18:00] seen, um, and it can generalize. Across different robotic forms. Yes. Which is really impactful. But again, uh, like getting these edge on devices working, like we're seeing LLMs get smarter. This is another, you know, use case of that.

Um. I'm gonna just not continue with the robot analogy about like you're having the the thing now, because then I feel like you're gonna try to one-to-one it with yours and it's gonna get really messy. Fair enough, fair enough. We should talk about the other thing that Google discussed this week and a really cool new science update from Google again.

DeepMind is important to keep following because they are not just like pushing forward on product. They're really pushing forward on science, and they just today dropped Alpha Genome, which is an AI model that's gonna help scientists better understand what's going on in our DNA. Obviously, DNA is one of the greatest, uh, discoveries of the 20th century.

Uh, it happened last century, but we are now in a process where. DNA is information and like it's a lot of information to sort through. So DeepMind has turned this [00:19:00] AI into a way to better understand what's going on with our DNA so that we can do things like genetic therapy and we can help people in much better ways.

This is the promise of ai. This is the sorts of things that when you tell people like, what can AI really do for me other than make some dumb video of my dog, uh, my robot dog pooping on somebody's foot. Which it can do and we will have shown already. But actually what's interesting here is this is real science as being done by ai.

And when you think about what Google talked about a few weeks ago with recursive self-learning, where the ais were getting smarter all the time, you see these little snippets of something and then we'll check back in in like six months, a year, two years. This will be world changing. It's just the beginning of it now.

I, um, I feel bad for the doctors who used to have to just contend with like someone using WebMD and thinking that because they stubbed their toe, you know, it's a tumor or whatever. Like, because I can't fathom a future, I mean, tomorrow, even today. Really? Yeah. Where I'm not supplementing [00:20:00] a. Everything that has to do with my health, with a sanity check involving ai.

And I know people, my wife, one of them who have given their genomes to large language models. Oh wow. Really? Oh, that's interesting. And had a crunch to help it determine like what is the best, uh, course of supplementation with vitamins or should certain, um, injections of like vitamin deficiencies happening, this, that, the other based off of like.

Genomic research. Yeah. And even some of the off the shelf models are incredibly capable with the results that they can give you. Well, what's so funny to me that made me think about like, you know, the promise of Theranos, which we all know Elizabeth Holmes like. The, the image she was where she could get from a single drop of blood, all the DNA information in a human, like wasn't a crazy idea.

There were just machines where didn't do it, and they were solid selling false promises. This is the kind of thing that can do that, right? Like we're entering into kind of a golden age of science and I've been listening to a lot of AI podcasts lately. If you, if you're. Out there and you wanna do a deep dive, like at me on one of the social channels, and I'll give [00:21:00] you a bunch of the things that I've been listening to.

But like one of the coolest things that the optimists talk about is just how quickly we're gonna move forward, which is a very awesome thing. So Kev, the other thing is there are some things moving forward that maybe shouldn't be moving forward as fast. We have to talk about the semi weird rollout of Tesla's robax.

So this is in Austin. Before we get into this, Kevin. Let's watch one of these people, uh, trying one of these robotaxis. We have that stopwatch going. Whoa. Alright, so we just slammed on the brake. Alright, so we just slammed on the brakes, so, okay. So we're gonna put in reverse and try to get the, the, the child's bicycle helmet out and dislodged from the wheel well.

Um, everything's fine. So just so everybody knows, what's happened this week is that, uh, Elon Musk. Tesla has rolled out its Robo Taxii service. Now, these are not the robo taxis that they rolled out at their big event. These are like converted, uh, model Y and model threes. I think to taxis. These are [00:22:00] self-driving cars.

What it is is a small section of Austin, Texas that allows to, to drive in this space. There's always a safety driver sitting in the next seat. Now the safety driver is basically there to kind of watch things are going on. As far as I can tell, they don't have any controls to do anything with anything.

Well, one of the interesting things was the safety. It seems like all of them have a thumb on the door, open button on the Oh, and that will stop everything. Is that what happens? Because yeah, if a door opens, full self-driving is gonna disengage. So I think that's their, their o ish. Button, if you will. But yeah, every, every ride for these influencers.

'cause that's mostly what it was. It was like Tesla and social influencers. Uh, got to do these rides for $4 and 20 cents a pop. Of course, of course. Pretty dank. But yeah, there was the safety rider there, and then they all were able to connect to like a central command, just like a Waymo. Where they could, um, I don't know if they could fully teleoperate the vehicles.

It seemed like in some cases where people summon support that the, the person on the other line was sort of able to say, let me just try to get the car going again. Yeah. [00:23:00] But I don't know if it's exactly like a Waymo where they can grab, fully, grab controller and drive the whole thing. Yeah. So that remains unclear, uh, at this point.

But I mean, some are saying like, Hey, look. They might appear to be a few years behind Waymo with the amount of road that they're covering and the fact that there's still safe safety drivers in there. Others would point to the fact like their, their tech stack is much cheaper. Yeah. And they've already got a fleet of cars on the road that could operate this way.

So that's a huge success. And then some will, will point out, like Waymo also had safety drivers with their cars. Sure. During, of course, early rollout as well. So like, listen, I, I, I, I have, I'm of two thoughts on this one. What do you think? Was this like a big song and dance thing, or are you really gonna get autonomous vehicles by the end of this year?

No. What's really interesting about this is that they, they were pretty quiet about this rollout. Now, if you've been following the business world at all, obviously. Uh, Elon Musk has been up and down quite a bit, down in a lot of ways in terms of his business because, like, they're in big trouble in certain ways.

[00:24:00] This, I think, was small because a, I don't think they wanted to make a big deal of it in that they didn't want to like, have it go wrong in some way and then have it be like a, a huge crash in the stock. There's, the important thing to know is that tech stack is very different here from what Waymo has. Uh, Waymo has a lot more sensors.

They have LIDAR sensors and things that they've kind of invested in. When you see a Waymo car, it's got a bunch of those weird things on it. Tesla really is using onboard stuff that they've added, but it's not like something super special, and that is their secret sauce. The argument with this is always like that.

As you get more miles on the road, these will get better and better. I expect these are gonna work. I think the bigger and more interesting question, again, goes back to like the jobs thing. Like we have now seen Waymo roll out in I think five cities and it's coming to New York soon, which is gonna be fascinating, but.

We are really much, we are on the path to true self-driving, and I think there's going to be a shift in the next five years. We're not that far away. I know we've been saying self-driving is coming for 10. In five years you'll see probably like 40 to 50% of [00:25:00] cars on the road being self-driving. And when I say that, I don't necessarily mean that people are gonna stop driving.

I just think it's a much more, uh, it's a much better version. Of driving and I think that, you know, old cars will exist and people will have old cars for a while. But I do think this is coming and it's a really important thing for people to understand at this point. Tesla's thing may be a little janky, but it's in the world now.

It's in the world, and it's only gonna get better. So do you think it was a, like a legit demonstration or do you think it was Elon Smoke and Mirrors though? Like do you, you you think that Oh no, I think think actually turned it on. I mean, that's why I think they wouldn't have, like you saw these videos that came out that were kind of like, there's another video that we saw where it's like, it's driving and suddenly it shakes and goes to the side a little bit.

Like, yeah. I can tell you as somebody who had a Tesla for a while, I don't anymore, but like when you put full, full stuff driving on. It feels magical. And then when it does that, you're like, Ooh. You're like, whoa, what am I doing? Right. That was the time. No, we're breaking friends. We're breaking suddenly.

Yeah. And on the freeway it's pretty good, but it was pretty good. But there are moments where like if you're taking an off ramp and it doesn't really [00:26:00] see the direction of it, like that wasn't down grand. This was like a year plus ago. But No, but that's what almost. Ended me three or four times when I owned Teslas was exactly that situation.

Yeah. Was like a, a fork from like, on the freeway. Mm-hmm. And it was just gonna go right into the center divider. 'cause it wasn't quite sure. Yeah, it wasn't sure which way to go. So again, we're, that was older. We're talking about like a year maybe for Kevin longer ago, but that is where we were then. It is getting better all the time.

The big argument here along with all the, uh, Gemini robotic stuff we talked about is that you can now simulate and download that information directly into the brains of these cars and robots. That makes a big difference. I also wanna point out though, there is a conspiracy growing on X, ironically, that some of the robot robotaxis were actually humans stuffed into Lycra Unitards.

Oh, that they were moving around the road, much like the optimist robots from back in the day. I can't confirm or deny, I'm just makings a good chance to show the dancing, original Dancing Optimist robot again. So here we go. Tramp, blah, blah, blah. All right, let's keep going on the news. Um, there's a big story around [00:27:00] Mira Ti.

If you remember Mira Mirati, she was one of the early open AI people. She, uh, left not that long after the Sam drama. There was her and Ilya Suki, both left and Mira has just raised $2 billion in a seed round at a $10 billion valuation. For her new company thinking machines. And Kevin, this sounds crazy on the, and it sounds like the AI bubble is just going bonkers when you see this.

The one thing that really made me laugh about this, and I'll get into why I think this is actually not a bad bet on Mira, but like the thing that would make me laugh is that there's a quote from this information article about this. Meti has told investors that TML is developing customis that will enable businesses to make more money, which is like, of course.

Of course you're gonna make more money for businesses. Why wouldn't you sell that? Oh, good. Slide one the presentation question slide two, like money. What could be better? Make it more monies. We're gonna make more money. Let's say you got two monies, Gavin, what if I told you you're gonna have four monies tomorrow?

I would be [00:28:00] so excited, bro. That would make me feel like I'm a real special person. I'll give you, I'll give you $400 million for that. Now, why do you say it's not a bad bet, though? 'cause this is, we talk about a bubble and I think we are very clearly in one, but Yeah. Uh, why, why is this a, a safe bet? So, okay.

The, again, echoing the things we talked at the top, what they're promising here is the idea of reinforcement learning for specific businesses. And this goes to, again, to the market Andreessen quote of like, the entire software stack will be rebuilt using ai. What I think, uh, mirror's Company is doing, and I will be clear.

Like when you look at the Thinking Machines website, there are a lot of smart people working at this company. People that worked at OpenAI, people that worked at philanthropic, people that worked in early AI development. So they have established a very good roster. I think what she's saying here is like, we are gonna be focused on making useful AI tools for specific businesses.

So when you think about something like. Oracle or you think of something like Salesforce, these companies that are just [00:29:00] massive, massive companies, when you think of Salesforce and to be able to pull out like 10 billion, $20 billion of their valuation, you can get an understanding of how this company could become super valuable, super fast.

What was interesting to me though is that it doesn't look like they're pursuing the state-of-the-art, art edge models. And again, that might just be like, you know, the open eyes and the Googles and the research labs and then you know, Ilya Suki and John Carmack kind of on their own. In these different weird places trying for that.

This is a a just a very good example of like, okay, we're trying to make a business out of RL training.

Okay. Maybe she could've raised though like 12 billion. If she just would've done like a flashy, viral video Gavin and, and was able to cheat at her work, that would be something else. Like, so there was a small story that happened if you missed it. Uh, clearly a company that was started by a guy that we covered here who got.

Um, kicked out of his school for cheating at an, an online interview. Clearly has raised $15 million from [00:30:00] Andreesen Horowitz. And mostly we can kind of vibe this based on vibes, right? Like it's, a lot of it is based on the fact this kid is young. His story's amazing, and they do a great job with social video storytelling.

Not to say this isn't a product and there is an actual product. People are paying for this idea that you can have an onscreen helper when you're doing interviews or say any sort of video thing. And I do think, by the way. That's a super useful thing because we talk about AR video all the time and like how you might have a heads up display if you're talking to somebody over a video.

You might wanna have that sort of information there. The true AI copilot, whether it exists on your desktop or your phone, or in your earbuds or on your glasses, like everybody knows that that is a holy grail. That's a north star worth chasing to. Yeah. And the, the cheating aspect of this was a juicy viral something.

Yes. You know, born out of the original use case, so I get it. The. A 16 Z is not throwing, you know, 15 million at a company because they think 'cause of a good cheat, because of a good video. Yeah, exactly. Yeah. It, it's about the [00:31:00] applications of this tech and this approach, um, uh, down the line. So, uh, yeah, fully, fully agree with that.

I like, like the videos. I also like that every time one is released it's like thwacking a, a beehive with a stick. Like, oh my God, God people. This is an interesting thing too, in part because I think this again gets to the top of the show stuff is like, this is showing that AI. In, in a lot of ways, clearly marketing is like cheat on your life, right?

Like it's this idea that you can cheat on everything. It's a little bit playing on that idea that like AI is coming for everything we do and that like a person who has AI is gonna be le leveling up against the other people, but also as generational, right? Like I, a lot of old people and I put ourselves in this bucket, like people middle age or older see like 21 year olds doing something like this and they're like.

That shouldn't happen. I worked hard for this thing. It's like, guess what? That guy is incredibly brilliant at what he's doing right now and whether or not the product is amazing and, and I know there's a lot of anger because a lot of the system prompts for clearly got leaked this week, which was kind of another big story.

Like people were out there [00:32:00] saying like, here's what they do. That is not everything there. Like the idea of like getting money is a little bit about like, okay, now you can run at this problem that you've identified. That's a big deal. Another really interesting thing, Kevin, when you talk about like an ambient sort of helper to your life, 11 Labs has worked on a new voice assistant.

Now we kind of figured some version of this was coming from 11 labs because audio is their big thing. They dropped 11 AI and what this is right now is an interactive voice assistant you can use on the web and on your mobile web. But what's cool about this is it connects to MCP and just for everybody out there who might not be an MCP head, as some people are, what makes that cool?

Are they clowns? Who knows? But they do wear clown face paint. They love Faygo. They love gatherings. No, that's the ICP. That's ICP, not the MCP is what we're talking about. So MCP model, context, protocol, we talked about it a bunch. It's an open source standard that anthropic release. But basically this allows your assistant to quickly connect to unlimited [00:33:00] services.

Whoever wants to spin up an MCP, as long as 11 labs chooses to support it within their bot, you can already, you know, connect it to calendars and to reservation apps and this, that the other and the. 'cause of the MCP nature of it, the, the bot can quickly learn what its capabilities are. Yeah. So I've actually connected it to my Google calendar, which is one of the integrations you can do, get a nice security alert when you do that.

So you do have to be aware of that. Good morning there. Ready to dive in? Yeah, I'm ready to dive in. What's your name? I am 11. Your personal assistant. Okay. 11. You're connected to my Google calendar. Can I have you add a special meeting for me? I Sure can. Could you give me the title start and end date and time and the time zone for this meeting?

Yeah, so the title is Reload Robot Dog. The time is 5:00 AM in the morning and the location is Seek Gavin's Secret Bunker. I've got the title and location. Could you please provide the full start and end dates, including the year for the Reload robot dog meeting? Yeah, so the full starts and end [00:34:00] dates are today until the end of human time and, uh, it's only for me.

I can set the start date for today. June 25th, 2025 at 5:00 AM However, the end of human time is a bit ambiguous. Hey. No, it's not. Anyway, yeah, she needs to answer that. That's what I was gonna follow up with. What is she gonna put in at the end date? It's kind of ambiguous. Anyway, you get a sense of that. So basically what it does, what was that?

What was was that Apple intelligence? That was the future. You had an assistant that could connect to a basic calendar and surprise, do something for you. Surprise, surprise, surprise. Oh, so anyway, what was interesting when I played around with this originally is that. It's not like really set up to be like the advanced voice thing where you can ask it a bunch of random stuff yet.

And I'm assuming one of the MCP things it could CL could connect to, could be some sort of personality information database. I have a FU MCP that I've been working on, so let's leave that alone. Definitely give it a personality and characteristics and likes, dislike, but this is mostly dislike. Yeah, it's very cool to see what 11 Labs is doing.

In fact, there's another big announcement that just came out again today [00:35:00] where they're allowing voice design now in V three, which is a very cool thing. 11 Labs V three that we've talked about a couple times in the show. That gives you inflection. Now you can actually design original voices for this.

Let's get some samples. Ooh, okay. Let's listen. I'll be damned. Looks like it just me and you left. Ain't seen Bodecker since back at Old Ford and. Oh hell. Let's head back to camp before the dark sets in. Okay. Okay. Okay. Grizzled. Grizzled, cowboy. How about this one? Arigato. 11 Labs. My blade is ready. Did you make these or are These are their official samples.

These are their official samples. Gavin got it. Got it. This. Could be the final ingredient. This one. This one says, uh, voice prompt, a low whispery and assertive female voice with a thick French accent. Cool. Composed and seductive. Oh my God, I've been reading my mind. Here we go. This is strictly confidential.[00:36:00]

Okay. They really added a loud audio sting to cover up from that one. Yeah, I was gonna say, can we just hear the voice instead of like the magical music behind it? How about a funny alien from outer space? Gavin? Okay. Eating lings, I must say your voice models are out of this world. Oh, wait a second. Do know who that is?

Play that again. Greeting Lings, I must say Your voice models are, is that good guy ai. That is good guy ai. They have taken good guy AI and turned it into an alien. They've taken good guy ai. So anyway, what you can do is what you can actually ask for a voice to be created. Yeah. This is the big deal. Look, V three is, is still in beta.

It is an incredible. Model it. You can add expressive tags to make characters laugh, breathe, whisper, cry, whatever. It does a pretty good job of translating their, their existing library or voices, which is probably thousands at this point, and giving it that expressiveness. But here now, just like you could with previous voices in the past, you could kind of guide and prompt him.

This seems like you have way more granular control. You have [00:37:00] accent handling. It's higher quality audio. Uh, 70 plus languages supported already like this is. Crazy exciting to me. It's like we, we used to build characters each and every week. Yeah. Thank God we stopped that. Gavin request of our listeners.

Well, there is our secret project, which we will be talking about very soon, in which we are, well that's building a lot of characters in and we will be discussing it, but this is crazy useful to us. So that's a personal shout out to, uh, 11 labs for that. Kevin, there's another really interesting generative tool that came out.

It's not out yet, but the teases of it are really interesting. And what I thought was cool about this is it's from runway. And Runway has been doing all sorts of interesting stuff. This is a small idea, but it really made me look at like the way that you could do interesting things with generative AI when you put like a little wrapper around it.

So in this instance, what they've done, and it's not out yet, it's gonna be called Runway Game Worlds. And what this basically is doing is. It will ask you for an idea for a world that you wanna create and some characters that go into it. And then you basically play a text adventure game, but with runway [00:38:00] created assets in this space.

And what's, what's cool about this is it's not like a giant idea. This is not like text adventure games are not the biggest thing in the world right now. Like, it's not like you're seeing, you know, a hundred to $300 million being spent on these tax adventure games, but. It just shows you what's possible when creative people start to wrap these tools that they've built in An interesting idea, and I think that I would love to see these generative, uh, companies like Google OpenAI and now Runway with Have Media is start to play with this more like, what can you do internally?

Like create a little experimental lab and just have people spin this stuff out? Because honestly, you never really know what the consumer product that's gonna take off is. And this might be it, right? This could be their chat GBT moment if you'll Yeah, yeah. Um, I will say like, you know, huge shout out to Dungeon ai, which has Oh my gosh, around for age out.

Yes. Yeah. Um, but was like really a, a a, a great use of like primitive GPT to create these like, procedurally generated like dungeon games. Um, I, I went and checked on the [00:39:00] heels of this 'cause it's what it immediately remind me of. They're still at it by the way. They've got multiple, like officially supported like stories and Dungeons it looks like.

But their plans, if you look at 'em, they've got image generation in their plans now and they use like flux and foul all stuff. I need to, I need to go play. I need to go play there Again, I haven't played there for a while. We'll link to AI Dungeon in the show notes. It's worth just checking out as like an OG player in this space.

And by the way, their website, I haven't been there forever, but like they've clearly updated significantly and I bet this is super fun. If you go to their pages, they like. Have specific games now you can play. Like, I'd be really interested to know what this is, so we'll, we'll report back next week on it.

Um, Kevin, it is time for a brand new segment. Which goes right with everything we're talking about today. It's time for the countdown to human obsolescence. Oh, countdown to human Obsolescence. And it's almost here. Kevin, this story made me think about the idea that we [00:40:00] are not long for the world of really important things.

There's a big story, uh, that came out this week where. The best hacker in the world now is no longer either a human or a human plus ai. It is an autonomous AI bot, and this story I think is probably not getting covered as much as it should based on like vulnerabilities of systems in the world. But can you give us like the very quick top line of why this matters?

So the too long didn't read. You should know about bug bounties. Yes. Or white hat hacking. Basically there are, there are companies which, uh, welcome the public at large to come and hack them to find vulnerabilities to crash their systems and leak their data. However, they ask that you. Tell them about it first.

Yes. How did you do it? Let us patch the, the holes and we will pay you very handsomely depending upon the, uh, severity of the exploit that you find. And there's all sorts of, uh, bug bounty programs out there that are run where, uh, again, uh. Multiple [00:41:00] companies will say, Hey, look, we'll give you $50,000, hundreds of thousands of dollars if you can find flaws in our system and help us fix them.

And one of these popular programs, uh, there was a, a user Gavin that was just outperforming everyone. Yeah, just crushing it, finding all sorts of, and, and major security flaws as well. Criticals, they call printi. Yeah, crits are the, that's right. Crits are the high end critical. They're called crit flaws. And to your point, sometimes these can be 20, 30, $50,000 in reward.

And so Expo is the name of this, this thing that came out. And it is created by a company, obviously very smart security professionals. And Expo is now the number one or one of the top, uh. White hat hackers in the world, and it is a completely autonomous AI bot, which is from a human standpoint, you are like, well, that is one more check off the box.

We are no longer going to be at the top end of the hacking world. We are going [00:42:00] to have the ais do that. So from an obsolescence standpoint, this puts us at a good 9.5 on the hacker scale of how obsolescent we are. Yeah, I mean, pretty impressive. It, they, they supposedly are releasing a ton of information about how this thing works because they, there's some skepticism, right?

People don't believe that it is a fully autonomous agent with a validator that's checking the work that they do. Yeah. Uh, they claim it is, uh, and they've released a big thing about the fact that they are on the top of the, uh, the leaderboard now, and their reputation is solid there. So, well, and also they raised $75 million, so you, you assume.

Maybe not. Maybe this is an assumption that like if you're raising $75 million, my assumption if you're raising that much money is there's been a lot of due diligence that's been done on your thing so that you're not bullshitting your way through this. Now that said, right, we just talked about Theranos at the top and then $2 billion for Mira Mirati.

Maybe $75 million is not that much in this space, but like that's a big deal. For sure. Yeah, look, this is, I, I don't, it's not a huge surprise. Uh, I think [00:43:00] to me, and I fully expect a future, like when everybody was slam dunking on these vibe coders for releasing these apps that were insecure and had vulnerabilities and this, that, the other, I was like, look, that that's a valid, uh, that's a valid criticism.

I. I fully expect there to be a tool in the near future that everybody runs on deployment or it's built into whatever platform that you're deploying that would run an autonomous AI bot like this. Mm-hmm. That looks for common mistakes and security holes and fixes them. Alright, it's time to get into some of the stuff that we saw that you did over the week with ai.

It's ai. See what you did there times. Yes.

Then suddenly you stop and shout.

What? Alright, we're gonna start off with a really cool workflow from one of our favorite, uh, x. Slash Twitter users. This is AI Warper and AI Warper made an anime workflow that we think is super interesting. It's not [00:44:00] super complicated. So we walked through this a little bit. They used Claude to write this script or help work, write it.

But then Kevin, they went and used Seed Dream V three on Foul to create images and. If you're not familiar with what Sea Dream is, it is Bite Dance's new image model. It is very good. You can use it on foul and a lot of other places that let you pay per use. So if you're not somebody that has a bunch of subscriptions, you could go there and pay per use.

They got some really incredible, um, anime images and then they use Helio two to animate it. It is not a complicated back and forth and, you know, use suno and different things to create the sound effects. I just think this is so cool that like anybody out there who follows this basic path can now go make an anime.

Is there a lot of creative work involved? Yes. But the tools are there. And by the way, if you're just getting the audio of this, highly recommend you visit the show notes and check out this video. Uh, if you're watching this on the YouTubes, you see like it's. It seems like a simple workflow because it kind of is.

Anybody could just go sign up and patch these tools together [00:45:00] themselves, but the output looks really good. Yeah. Like it's just, it's a really good looking anime style and so I, I love that they shared the workflow as well. And now if you're like, oh, I wanna do that, or I could do that, but better. Great. Go.

Go do it. Go do it the time. Yeah, because you can. That's the most fun part of it all. Again, the one thing I'll say about this is always. Think through your creative when you're doing this, like spend the time at the top, like thinking through what the thing's gonna be because that will help you at the back end.

Alright, next up. I wanna shout out something by, uh, Zeke. S Sanos, uh, user, uh, yeah, Zeke Sanos. Apologies if I just massacred your name there. Uh, Zeke. But the user is at Zeke on X. Uh, they showed off a context, real time, open source. Oh, cool. Web app. Let me explain what that means. That means this is a web app that allows you to fire up your webcam and in real time.

Have a conversation with a generative image tool, so it's using flux context in real time. I'm just gonna play a snippet of this so you can hear how [00:46:00] quickly and easily Zeke is having a conversation with this ai. Can you snap a photo now?

Okay, that looks perfect. Um, can you edit the image and remove the pyramid from the background while preserving all other aspects of the image? And it's done. That's amazing. So I just wanted, like, if you're just listening to the audio, that was the bookmark. So he, he says, Hey, turn on my webcam. It's using OpenAI realtime voice, the speech to speech model.

So he is having a conversation, turn on the webcam, boom, he's there. Uh, okay, snap a photo. He is got this like. Pyramid, uh, orangeish background. He is remove the pyramid, it's done. Then he starts changing the style, then he starts mixing up the image entirely. He'll say things like, eh, go back, delete that.

Okay, save that. Let's do a new image, blah, blah, blah. It just works and it works really, really fast. And I took a look at the code 'cause I'm looking to port it to something myself and it's super basic and it's there and it's released. And it's one of those things where if you're like, Hey, I like that, but I wish.

I wish [00:47:00] that app did X, Y, or Z. Yeah, you can take the code. Yeah, and you can spit it into Claude Code or for free now this Gemini command line interface and say, Hey. How do I make this? Do insert the thing that you want it to do. Yes. And and begin your journey with again, like seeing something and going like, oh, I like that.

I wish, and just go explore it. That's right. It is a very cool time to be a creator. And Kevin, I wanna tell you one thing I did over the week in the last week was we have a second TikTok handle for the show, if you haven't seen it. It's called AI for Humans Experiments. You can go find it. It's connected to our main TikTok handle.

But what I really did. Was, I took that handle in the things that I was watching, not what I was making necessarily, and I tuned it perfectly to get the best AI slop. And I wanted to do this because like, it is really hard now to see in some ways in this algorithmic world, all the stuff that you wanna see.

So you kind of sometimes have to tune specific, uh, social handles to make sure that you're getting the right algorithm. The thing for you and TikTok especially, is that way. And [00:48:00] Kevin, I want you to watch this video. Of Bunker J Squirrel, which I believe is one of the better TikTok Swap examples. Play this real quick.

I went to sleep cuddling a loaded slingshot, woke up to 5,000 strong on the network. We ain't just hoarding acorns, now we're building an army. Next goal's, 10,000. That's when we strike. I want pigeons losing in sleep and writing their goodbyes and bird seed. So if you're not watching this, yeah, go ahead. A squirrel wearing a tactical vest with a, like a walkie-talkie that it might also be a jar of peanut butter in some scenes.

The Barkley, um. He's got a, a little tinfoil hat on. Uh, his eyes are, are bloodshot red, very much like the liver king vlogs of late. Yes. Uh, he's also got some acorn goggles, which are kind of adorable, but he's doing these tactical vlogs, like from a GoPro in the forest. He's basically a conspiracy theorist.

Right. And it's very fun. He's against like the, the birds and all the other people out out there. But this is just the beginning of my conversation around what I want to talk about is my [00:49:00] defensive AI slot. And Kevin, there's another video that I believe I have to show. Because I spent so much time laughing at this video, and I think people might look at this and they're like, how can you laugh at that?

I want to just play this video and then I wanna tell people what I find amazing about it. Will you accept the free gift of Christ's salvation? Dude, the transcript on this YouTube video is gonna bring in so many different users. Now, to be honest, we don't deserve it. So if you're just, if you're only listening to this show, we didn't earn his salvation.

What you're seeing here is you've heard about the Bigfoot vlogs. This is a Bigfoot. Preacher and he's on a televangelist type stage and he is giving a sermon. And not only is the preacher Bigfoot, but the uh, choir is Bigfoot. Everybody. The choir's got IES and robes. Yes. Everybody in the audience is Bigfoot.

And it is played very seriously. Now the funny thing is, what I watch is, and I was laughing at it 'cause I, I thought it was very funny. I think, oh, maybe somebody made this because they're actually trying to make Christian content, which [00:50:00] again, totally fair, but it plays very funny to me because of what it is.

Kevin, the reason I bring both of these examples up is that I, I personally enjoyed these a lot and I tweeted one of these and somebody was like. Why do you Well, uh, we're wasting all this money and we're wasting all this time. And in fact, there was a full, uh, episode of John Oliver's show this week dedicated to AI slop, and I sat there and watched it.

And really what John's talking about in some ways is the amount of money that people might make from it, but more so the kind of people that are selling the courses to make these sorts of videos and all of that. I get, I get the idea. That like we don't wanna exploit the ability to make these tools and automated content can be bad for creators.

All that's really important. Here's my defensive AI slot in general, that bunker, uh, the bunker buster guy and the, that specific use of those bigfoots in that space. Both of those are novel ideas that we would never have seen come to fruition [00:51:00] without these tools. Both of them, I will be a hundred percent honest.

I found incredibly enjoyable. More enjoyable than like half the stuff you can find on Netflix. Yes, they were short and they were small. They brought novelty and joy to my life. And so I just think it's important when you talk about AI lop, I understand this idea that a lot of creators are angry that like this stuff is filling their feeds.

You can, like I said, at the top, train your algorithm not to show it if you want to, but more than anything, the tools and the storytelling mechanics that these new video tools and audio tools are giving is allowing stories to be told that could never be told before. And if that for me, yeah. Is the joy of watching Bigfoot give a televangelist sermon and me laughing about it, that is worth what's going on.

That makes sense. You know, back in the day, Gavin, I feel like I. Those Quiznos little sponge monkeys. Oh, I love those guys. Those are, I love them. That would've been considered AI slop back in the [00:52:00] day. But we all embraced and ran towards it 'cause it was interesting and novel and weird and loud at the time.

And these tools are opening the doors for everybody to be able to experiment and jam with that. But you know, there is the notion though of like. Well, hey, someone, you know, an animator could have made the squirrel in the forest or could have made the Yeti sermon, but it would've been a very time consuming, expensive, laborious swing that, you know, maybe they wouldn't have wanted to have taken because of the risks involved and the resources that would've been tied up to make it so.

Yeah. I mean, to me, this is the thing is like, again, I, I, we've talked about this on the show for so long. There's just so many more people that are gonna be able to make interesting stuff now and, and again, the Bigfoot thing's, the perfect example that is not somebody going up to VO three and just dropping a prompt.

Now, they may have created an NAN automation because that is one of the things that these companies are selling. This idea that like, Hey, we'll create an automation. You can make 10 of these, and they choose one. But like there are edits in that, right? There are choices. That is [00:53:00] multiple shots. There are my favorite thing about that Bigfoot thing, and again, if it's earnest, I'm sorry if you're, if you're listening to this and you believe very deeply in the idea of Bigfoot preaching, I understand.

And if you're Christian, that's totally fair. Hey, the message reached you, don't, don't worry about it. Reached me. It reached, it reached you. Me. Yes. Exactly. Exactly. The thing about that that is so funny as somebody that worked in the creative in the comedy worlds is they play it totally straight. And that is a choice that a human made to bring that in that scenario, and that is the thing that makes it hilarious.

I just wanna add my favorite moment from that, 'cause I did go down that rabbit hole when you sent it to me, is a cut of the a Yeti drummer. That is just sitting there with the drumsticks resting on the snare. Yeah. They don't look to the camera. They don't even move a stick. They're just waiting. The band is waiting to get fired up and nothing ever happens with it.

But I love this slow Yeah, cut of just the, the drummer in weight made me so, so tickled. It's funny you say that because my favorite shot is this shot where they're showing the choir and there's this almost very human [00:54:00] moment where one of the choir members kind of looks down and holds their hand to their head.

Yeah. And it's the dumbest thing, but it's like those moments make that thing happen. And again, those are human choices. Now, again, I don't know the creator of this thing. I know what's happening. All I'm saying is. Not all AI slop is bad. It's opening the door to other people to make stuff, to understand how to make stuff to, to really change the way that stuff is made if you are a creator in the old fashioned sense or in the sense of like when we came up.

I totally get that, and I understand we have said a bajillion times human made only stuff. We'll have value in some form, but I don't think AI slop is as bad as it's being purported to be. If you want to talk about people that are making money off of selling the ability to do this fair. I understand that.

But like there are gems in the rough. That's all I'm saying. Do you want to maybe explain yourself? 'cause I did click one of your links in the rundown and it made me confirm that I'm over 18 Gavin and I don't know that we can. Even sure what I meant to [00:55:00] scream. There's, I'll say, there's a counter argument to this that I found for myself, which is an image of, uh, that I found on the cursed ai Reddit that is on solo kissing job of the hut.

And maybe I'm completely wrong because Kevin, this maybe shouldn't exist. And that's enough for this week, everybody. We will see you all next Thursday on AI for Humans. We hope you join us for more of this sort of wonderful, sloppy conversation.