AIForHumans119ChatGPTAgentOpenAI

Gavin Purcell: [00:00:00] Chat, GPT agent is here. Open AI's new model, combines deep research with their operator web browser into something very special and has a nice little AI sauce to it.

Kevin Pereira: Oh, is it

Sam Altman: banger sauce? Good morning. We have a banger for you today. We're gonna launch chat g. PT agent.

Kevin Pereira: Note to self and Sam Altman, wake up before dropping a quote banger.

We're gonna walk you through exactly what chat GPT agent is, what it can do, what it can't do, and unfortunately what jobs might be replaced. Mr.

Gavin Purcell (2): Pereira. Mr. Pereira, this is our time. 2025 is the year of agents, Mr. Pereira.

Kevin Pereira: Ignoring that. Plus, mark Zuckerberg has poached more open AI researchers. Elon Musk is dropping AI Yfu.

We're gonna get into that runway announced act two, which is incredible. It's gonna revolutionize the way us humans contested these.

Gavin Purcell (2): Mr. Pereira, I don't think you understand that this is a revolution in ai.

Kevin Pereira: Back to me. Back to [00:01:00] me. Thank you. China's new Kimmy K two model shows that open source has a lot of juice.

Let's Mr. Pereira, Mr. Prayer's, my dad. Hey, speaking of

Gavin Purcell: Dads Darth

Kevin Pereira: Vader wraps this week. Wait, that's all I had to do?

Gavin Purcell: Yeah.

Kevin Pereira: Great. Glad you're back. Plus, we got new visual effects from Higgs Field and that's always fun.

Gavin Purcell: Robots that can replace

Kevin Pereira: their own batteries, y'all. All of this and even more on a brand new episode of.

AI for

Gavin Purcell (2): AI for agents. Mr. Pereira,

Kevin Pereira: this is worse than the Zion Rave.

Gavin Purcell: I'm sorry. This is AI for humans. Everybody.

Kevin, it is AI for humans, and this is a big week. We have been waiting for an open AI drop and we got one. It is not GPT five. We have not heard anything about this, but this is something new. Wait, is it the open

Kevin Pereira: source model, Gavin? Is it the open source source? Open model model? Because they said it was gonna, oh, it's not that easy.

It's all the

Gavin Purcell: things we didn't. Say were gonna happen. That's right. This show that we talking about, things that Don Crystal don't happen. Crystal balls that we polished last week. Those were [00:02:00] wrong. Oh, they're all wrong. We'll have to get them out at some point later in the show. But everybody, the big news this week is chat, GPT Agent.

This is a new tool from chat GPT that is actually using the name chat, GPT. So you can tell that means something special, right for them. I think that's a big deal. Kevin, let us, well also,

Kevin Pereira: let me be clear. Papa Altman showed up to the Webstream. Yes. Yes. And that, that's always, you know, it's like, it's, it's Christmas Santa's there, you know, it's not just the elves in a, a bustling workshop.

It also might be because they have no employees left. They're all at meta. Oh, that could also, we're gonna get into that re regardless. He was there. So should we get to the announcement and dive in?

Gavin Purcell: Yeah, let's, so the funny thing is, so basically what this is, is a really interesting use case of, of, uh, chat GPT where it's gonna go and use, combining deep research.

They're a really cool product where you can go find stuff. Plus operator, their mobile web browser into one new thing. But Kev, let's kind of hear from Sam first from this live stream, what this kind of agentic moment feels like right now within the, within their company.

Sam Altman: So we, we [00:03:00] started launching agents earlier this year.

Uh, we launched deep research, we launched operator. And people were very excited about this. People could see that now, uh, AI was going off to do complex tasks for them, but it became clear to us that what people really wanted was for us to bring those capabilities and more together. People wanted a unified agent that could go off, use its own computer and do real complex tasks for them

Kevin Pereira: while wearing a milk made outfit and having demon horns and being, that's not what they're

Gavin Purcell: saying, Kevin.

That's not what they're saying. Keep playing with, keep playing. That's not, keep playing it.

Kevin Pereira: Okay, we're gonna get to that later. We're, if you tuned in for that, which I know you'll have, we're gonna get to that later, but we'll continue.

Sam Altman: That could, uh, seamlessly transition from thinking about something to taking actions, to using lots of tools, using the terminal, clicking around the web, even producing things like spreadsheets and slides and, and much more and wanted to, people wanted to be able to do this over a long time horizon and a sort of for universal tasks.

So the team has been working super hard to bring that together.

Kevin Pereira: Now this agent, if it were to bounce on say, a [00:04:00] trampoline on the moon, how does the gravity affect the jiggle physics? Gavin? This is

Gavin Purcell: all non gravity based agents. Kevin, there's no visual for this. This is all things that are happening, uh, with text, unfortunately.

Wait me when we talk rock. Okay, fine. We'll get you back up here. Talk rock. Let us talk about this and why this is a big deal though, because there are some really important things to know. First of all. This is not a new model. It is a tool set that they are unleashing for e everybody to use. So first and foremost, everybody should know is it's rolling out to pro and plus users very soon.

If you're a pro user, that's the $200 a month plan. You might have access to it right now, plus it's supposed to get it very soon as well. You will get 400 queries a month on the Pro at 40 on the plus. So it's kind of entry point, but Kevin. What it does is actually quite interesting. It is a combination web search plus terminal action with this underlying deep research model on it.

So in the demo, in the live stream, they showed it trying to go do a couple things. They basically send it away to go get these guys all [00:05:00] outfitted for a wedding, right? So what was interesting about this is they basically said, Hey, I need you to find, uh, an outfit and a gift for this wedding. And it gives them like the website invite.

So it's like find the formal code. Find what sort of gifts they want and go bring this stuff back to me, and it literally goes away for a very long time.

Kevin Pereira: And when you say find this, find that the, the prompt was a little beefy in terms of what it wanted the agent to do. Like broke all that stuff down, but it didn't say, go find the dress code and then go find something that matches it and this, that, the other, it just said, find me something to wear.

Yeah. And so the, you know, the, the, the, the context here is it, it, it crawled the wedding registry or the we, the wedding invitation website figured out where it was going to be, looked up the average temperature and weather for that. Looked at the dress code on the site, factored all that in, went out, crawled the website, found several pieces of clothing that might fit a certain budget, prepared them into a chart along with travel options and this, that the other, and you know, [00:06:00] these, this is an agent that, you know, the, the entire demo was what the entire, the, the entire conference was like about 25 minutes.

Start to finish. Yeah. Yeah. This thing took about 20 minutes to run. To do it, yeah, to go fully complete. That's the way this works in a very interesting way that, you know, there's some people are approaching the, we're gonna take over the browser on your computer, and we talked about this a couple weeks ago, uh, with Perplexities efforts.

This is running in the cloud. This is in open AI servers. It is running and doing its thing and then tugging on your pant leg if it needs you to intervene, right? If it needs. Sensitive data, credit card info, or if it's at a critical fork in the road and doesn't know what to do, it'll ask

Gavin Purcell: you. Yeah, it'll ask you.

And at one point, at the end of it, it lets you take over to kind of do that final step, because it's not, I don't think at this point, like doing, uh, financial transactions. The other thing that's really cool about this, which I think is something that's much needed, is. It's interruptible. So if you, in the middle of your agentic search you, it can jump in and just totally crash your train of thought.

Yeah, just like, like you, just like you Kevin, just like expert. You're so good at it. Expert. But yeah. So if you, in the middle of your wedding [00:07:00] planning think, oh crap, I also need to rent a car, you can stop it and you can say like, Hey, I also need to rent a car. Add that to your list. It will go do it. So there's a couple really important things to know about this.

One, first and foremost is that it is coming now-ish, right? So like it is not something they're holding back on. Kevin, the thing I kept thinking, watching this when I saw some of the benchmarks, it did very well on some of the benchmarks we talked about last week that Grock did well on humanity's last exam and, and Frontier Math.

It did very well on them when it was a tool-based system. And again, this is the O three model, not their next model supposedly behind it. Kevin, there are two benchmarks that I think are really important. One is called web bench and one is called spreadsheet bench. And these are both kind of boring sounding benchmarks, but what they are are benchmarks based on.

Usage of tools and web bench. Web bench is about how well, uh, the thing can browse the web and do things in that space. Spreadsheet bench is how well it can kind of do these things in spreadsheets. And what I wanna point out in these things is how close they are to human. And that closeness to human is legit.

And I think that, you know, we're [00:08:00] all gonna get our hands on this and I'm sure there will be disappointing things in the beginning and it will get worse over time. But as we talked about a couple weeks ago. Anthropic, CEO came out and said like 50% of entry-level jobs might be going away. This is the thing that they are all kind of working on in some form, right?

This idea that like within the tools that a normal person uses, suddenly. There will be ais that will be able to do pretty complicated work. And when you think about whether the average white collar, entry level worker per uh is doing, it's not that far away from seeing these be able to do that. Yeah. If it works, and again, we have not been hands on with this yet.

There will be problems, I'm sure, but you can see the trajectory of this happening.

Kevin Pereira: I was gonna say, even if it doesn't work in the next 48 hours Yeah. In the way that we kind of intuitively hope that it does or expect that it will in the next four to eight months. It will. Yes. So if you're, if you're having some sort of feelings about, you know, this sort of thing, right?

Yeah. Like, it's coming and, [00:09:00] and this was actually the first time I had one of the demos up and, and, um, my wife April remarked like, oh. This gives me a feeling. Yes. This is the one that's giving me a feeling. For example, they, um, they wanted some stickers made for an event, right? Yeah. So they fed it a photo of a dog, which is their open AI mascot, as they said.

And they said, let's go get some stickers made and use, I think it was Red Bubble or a site like that that, that they had stickers printed with before. Well, it. Takes the image and runs it through in, uh, you know, its own image model to get a version of it that would be sticker friendly with like, you know, specific geometry and colors.

And then it goes to the site and figures out where the stickers need to be shipped and how many need to be made and what that cost is going to be. And then when it's done, which is a much quicker search, it again tugs the pant leg and the human can then go and see, okay, where are we at in the process?

It needs me to intervene, like, confirm the shipping address and give the credit card info. You can then, which I think is pretty trick, go back and watch a, like a video. Yes. Yes. Very cool. So, so you can see exactly what it did and make sure it's going to where you want it should be going and this, [00:10:00] that, the other, it also planned a cross country, let's go to every stadium road trip.

Yeah. Right. And optimized where those things should be. And, and, and optimized the route for events along the way that the person wanted to go to and then plotted it on a map. And it's doing all these things that I'm like. Uh, there was a point in my life where I had an assistant that would do stuff like this for me.

Yeah. And it was more than $200 a month. Yes. And this is coming to the $20 a month tier. So if this is one of those moments that gives you that feeling in your tumtum for whatever reason, that's valid.

Gavin Purcell: Yeah. And I think it's really important to kind of be aware of that aspect. And then there's another aspect that Sam specifically tweeted about after this came out, which is the security aspect.

Because we are now turning a lot of people into, uh, active AI agents on the live web. And what I mean by that is there's an opportunity now for you can send your AI agent to do a bunch of stuff. So Sam had this very long tweet where he talked about it, and I think he's very proud of the team. [00:11:00] But then he did say, although the utility is significant, so are the potential risks.

We have built a lot of safeguards and warning into it and broader mitigations than we've ever developed before, from robust training to system safeguards to user controls, but we can't anticipate everything. In the spirit of iterative deployment, we are going to warn users heavily. And give users freedom to take actions carefully if they want to.

Kevin, what do you think happens after something like that is said? Well, I think that's,

Kevin Pereira: I mean, what happens is someone builds a website that tries to trap an agent and do masquerading as something that it's not an extract. Credit card information from it, right? Yes, yes. The same way we see prompt injection or jail breaking of these LLMs, someone will.

Target these systems very heavily. Yeah. To get it to present to the user. Hey, I'm just trying to make those stickers that you wanted, boss. Yeah. When in reality I'm just handing things over to a fraudulent site and giving them the data. Now I, I, I expect at first users gonna be like, Ugh. This thing is way too guard railed.

It's warning me too much. Yeah. It's [00:12:00] actually easier for me to just go and do the darn thing myself. And then per, you know, the, the stated mission, I think they'll start relaxing that. But what will happen when the first, uh, agent agentic victim of fraud. Yeah. And it will ha it will come, is out there. It's going to happen.

Yeah. I mean that's, that's part of what this is. Right. But will the agents be better than your. Your mom or my mom on Facebook clicking and falling for something, right? For a deep fake of Elon Musk, which almost happened to my parents. Like all of this

Gavin Purcell: is a really important kind of turning point. And I, I think ultimately we hope that there are systems set up that make these agents safe.

Of course, the AI safety people, the ones that are very concerned about AI, is getting better and smarter and doing all the bad things are going to see this and be, you know, very kind of wary of what happens here. You know, the other thing that goes along with this is that OpenAI did delay their open source model.

Uh, last week. It, it, a couple, I didn't really even get a date for it. It was supposed to come out last week and they didn't give a date. So maybe there's something connected to that in this way, but also. I guess the good news here is that like Sam is [00:13:00] trying to be at least in, in word, very safety cautious, right?

Sure, sure. And I think he's getting ahead of things that could happen, but also let ask you, we'll see how powerful it is. We'll see how powerful this thing actually is. You know,

Kevin Pereira: he did, he also tweeted about the jobs aspect, which we were just touching on. Yeah. Uh, he said one, uh, people will do a lot more than they could before.

Ability and expectation will both go up. Great. That's what I needed. More expectations. More, more still care,

Gavin Purcell: more responsibility. Fantastic.

Kevin Pereira: Still care very much about other people and what they do. Three, uh, still be very driven by creating and being useful to others. He then says, for sure jobs will be very different and maybe the jobs of the future will look like playing games to us today while still being very meaningful to those people in the future.

Do you feel any sort of way about. Those words. I mean,

Gavin Purcell: I do understand what he's saying about jobs changing over time, right? Like the thing that everybody in our audience should realize is that if you looked at what we are doing right now, now, just to be clear, Kevin and I are not getting paid enough to do this as our jobs.

But there are lots of people who do this as their job and do get paid very [00:14:00] well. They would look like, you know, a hundred years ago, the farmers would look at those people and be like, those people aren't working at all. They're just sitting and talking and having fun, or they're just chatting with each other.

There are jobs now that do look that way. I think the big kind of like shocking moment is going to be where, how meaningless so much of the work that a lot of people do on a regular basis. And when I say meaningless, I don't mean to them because everybody needs some sense of work. Meaning like you, you go to work and you do something that means something to you.

What I mean is. The actual work that they are doing. You know, if you are a spreadsheet analyst or you are a person that is plugging numbers into a spreadsheet, there's not a lot of deeper meaning in that work itself, and a AI probably could take that over for you. So this in a some ways can be read as like, Hey, we're all in this kumbaya together, but could also be read as a little bit of a warning to people.

Who are doing jobs that are going to be usurped by these ais. And again, I think we are now past the Rubicon of this not being sci-fi anymore. We are in a world where this is possible. It may [00:15:00] not be right now today that some AI is gonna go out there and do your accounting job, but it is coming very fast.

Kevin Pereira: Well, the good news is if you're willing to sacrifice sleep for almost a week, yes. And you happen to be one of the most talented like coders on the planet, you can still best these machines, Gavin. So,

Gavin Purcell: just barely though. Just barely. Kevin, go. Humans go. Just go.

Kevin Pereira: Yeah, go.

Gavin Purcell: This is go. So this is another thing that happened is that OpenAI entered a.

Coding bot into the The act. The ACT coder World Tour Finals

Kevin Pereira: Heuristic. Yeah.

Gavin Purcell: So this is, I went on, this is a very nerdy competition. There's a very kind of nerdy tweet thread from one of the people, uh, that was on opening Eyes team in this thing that's really interesting. But this was an autonomous agent running in a coding competition and it almost won.

And this is to Kevin's point. These are the best competitive coders in the world. And the only way the one guy won is that he slept for like three hours throughout the entire thing, got up and pushed himself through it. I mean, Kevin, the thing that I saw somebody [00:16:00] tweet about this, but like the thing that it really reminds me of is John Henry.

Do you know the John Henry story? Is this the guy with the big bat and then he he didn't take the first No, that's Casey. That's Casey. He didn't take the first few swings 'cause he was like, I'm so different guy. John. No. John Henry is a famous, uh, American folklore about a guy who, uh, says he can dig a tunnel faster than a steam engine can.

Yeah. And it is about, actually, it might be about digging coal, but it's about digging. And so he beats the steam engine, but it's so clear by the end of it that he's basically toast and the steam engine is just ready to go again. Right. It's kind of a parable about this idea of like, you can't really, uh, fight progress ultimately.

So what's so interesting about this in tandem with the Chachi PT agent launch is that. We are just seeing this idea of these AI models. Now, whether or not they get to the stage of like super intelligence or not, we are now at the place where they are getting close to doing the work that even the best humans cannot do if they're working 18, 20 hour days, right?

So that is a big thing.

Kevin Pereira: It, it's awesome to see a human being [00:17:00] grind it out and be at the top of the board. Like, that's great, but make no mistake, like open AI sitting comfortably at number two, very close, didn't break a sweat and could spin up. 5,000 more agents if they needed to, to go and saturate the leaderboards and block the human beings out or whatever else.

So while we might be saying like, oh, humans one Skynet, zero. To your point. Yeah, the machines are, they're waiting, their batteries are still fairly charged.

Gavin Purcell: And one last quick opening AI story. So there is a new update to their image model, which is kind of like, this is good. It kind of like flew under the radar.

But yeah, specifically it was from the API. Essentially they are launching a much better image editing software. And if you know, you've played around with open AI's image model. The, one of the more frustrating things is trying to get it to change just a small part of the image because Right. It will redraw the whole image or it'll change your thing, or it won't get exactly what you want.

Well, this is much closer to something like AI, Photoshop, where you can basically, specifically say a, a, a, an exact spot that you wanna change. In the demo they showed there was a, [00:18:00] a room. Where they wanted to take a, I think it was a ball out of the scene, and they basically showed it what it looked like before and then after with the two different models.

And Kev, the new model makes it look almost exactly the same. And like this will be a game changer for me, especially once it comes to SOA and chat gt, because you know all those. Guy fii things that I do, they take a lot of work and, and I, I didn't wanna have to go into Photoshop again, and now I don't, which is so important.

Kevin Pereira: Yeah. Sam, Sam heard your cries in the night. Yes. This is, yeah. The high fidelity mode is specifically with like faces or intricate, intricate products and designs. Um, you know, the demo is really interesting because when you look at the way the old model did things, it was okay at preserving overall aspects.

Yes. Of a scene or of a person's face. But if you got two generations away, good luck. Like people started looking completely different. Um, geometric patterns lost their coherence, like the sharp details were gone. This high fidelity mode, you know, takes a a, a second longer to generate, of course. And, and I'm assuming costs a [00:19:00] fraction of a penny more with your generations.

I, I haven't seen the, the billing details on it just yet, but just looking at the examples. Yeah. This is going to be massive for marketers, um, for any creatives who want to like. Figure out what their starting frame is gonna be to pilot video generation with like VO three. Like this is it. If you go, oh, this is so close, but I wish I could change this one little, uh, now, now you can, you can.

Gavin Purcell: Yeah. And it's great. I mean, it really is one of those things where it makes it useful, right? Yeah. Like this is a real world application of chat PT and, and OpenAI that has actually useful. Kevin, there is more OpenAI news, but that news involves more people leaving OpenAI to go over to Mark Zuckerberg's.

Fantastic super intelligence team. That's right. Two more big OpenAI researchers were yanked. From OpenAI sucked. They got, they got Zuck, they got on the cob. Wait a second. That's not right. I shouldn't say that. That is not okay. I meant to say on the cob, but even that doesn't sound very good. So we're not gonna say any of that.

No, this sounds good. They

Kevin Pereira: have to sit in this. Is that where Sam Altman has to sit in that one chair in the hotel room that's in the [00:20:00] corner? Is

Gavin Purcell: that okay to say? No? I wouldn't say that either. No. That's okay to say. We're gonna have a bleep fest right here in this section of the show. I like Sam

Kevin Pereira: sitting in the, watching his employees with

Gavin Purcell: meta.

Actually Zuckerberg was interviewed by the information this week on video talking a little bit about this. Let's hear what he had to say. And

Mark Zuckerberg: like I said before, a lot of the numbers specifically have been inaccurate, but I think it discounts. The other key reasons why people are super excited to come work on, on Meta Super Intelligence Labs.

And one of the biggest is that you could just have more leverage as a researcher. You have more compute. It's, I mean, the best researchers basically, like historically when I was recruiting people to different parts of the company, you know, people are are like, okay, what's my scope gonna be? And you know, here people say, I want the fewest number of people reporting to me and the most GPUs.

And so having, having basically the most compute per researcher. Is definitely a strategic advantage, not just for doing the work, but for attracting the best people. And I actually think that that has been under reported.

Kevin Pereira: See Gavin, it's not about a hundred million dollars [00:21:00] signing bonuses, it's about getting a couple extra NVIDIA graphics cards.

So you could play Call of Duty on your 50 90.

Gavin Purcell: Mark is going for it. That's all I will say. Like he is putting down the money, he's bringing the team in. I am still so curious how quick this will come to fruition. Like they have, they're behind. They're behind in a big way. The other thing that was interesting, Kevin, is they, they brought in the team from Play ht, which is now called Play ai, so they basically brought in an entire new audio team.

They're basic, basically, yeah, the conversational voice.

Kevin Pereira: Yeah.

Gavin Purcell: Yeah. They're trying to become like the New York Yankees. Sorry, Kevin. I know you're not a sports person, but like the New York Yankees of ai, which is all the best players are kind of raised from other places, and then the Yankees buy them in. I don't know if this will work.

It'll be really interesting to see if it will work or if they're just gonna be kind of parentally

Kevin Pereira: behind. Yeah. They've also resorted to building computers in tents. If this is like a real thing, uh, like the traditional model of standing up a, uh, you know, a server farm with all the cooling and [00:22:00] concrete and this, that together takes way too long to get built, way too long to engineer.

So they, they're going to these like. Hermetically sealed tents that they can stand up quickly and some of the data centers are going to be like, I dunno, did you see the super impo, uh, like the super imposition of the center? Yeah. Goes biggest Manhattan,

Gavin Purcell: that thing? Yeah. Over it. Pretty crazy.

Kevin Pereira: Yeah. I mean it's nutty, but like, you know, Elon was jamming out model threes in tents in a parking lot at one point to try to get his numbers up.

So maybe it's not that crazy, but it just shows. The degree with which they are grinding on getting enough compute to support their initiatives and to keep attracting this top tier talent, right? They have to have these servers up and gunning and powered and online or else there's no point in having all of these tools in their tool belt.

And by tools I mean like these a hundred million dollar plus researchers and engineers.

Gavin Purcell: Well, and the other thing that was really interesting is there was a story that came outta the New York Times and was covered in a bunch of places that meta might be going closed source with their next model. So if you remember, open ai, uh, was originally open, but that it became closed where people [00:23:00] couldn't see into the actual code of what they were making or the weights and mo uh, weights of the model.

Meta came out. Llama came out, and Lama's big deal was like, we're gonna be open source, we're gonna open source everything. We're gonna make it available to the world. It's also what Deep Seek was originally based on in a way. Mm-hmm. According to Alexander Wang, who's heading up the super intelligence team there.

The rumor is, is that they may not share the model going forward. It may be a closed source model. And Kev, I think from an open source community standpoint, this is probably a bummer. I'm still not convinced that like any of these companies are going to open source very much of, of anything of significance at the frontier going forward.

But like Open AI is doing that will probably be a slightly less powerful open source model following the big one.

Kevin Pereira: Yeah, they'll release it when it's convenient to squash anybody else that's trying to up their heels right and cut them off. But I mean, it makes sense. And the, the play AI acquisition makes total sense when you realize the, the swings that they're taking with Ray bands and trying to get meta AI into glasses, like you're going to be talking with [00:24:00] agents, that is the future.

So, makes sense if they get a, a rock solid team to do that. Um, do you know who else you should talk to, Kevin? Well, you should probably talk to your accountant. Make sure you have enough money in the bank to like and subscribe to Human Go. Like subscribing is free.

Gavin Purcell: Kevin, what I was gonna say, you should talk to your accountant.

You should talk to your mom. You should talk to your brother, your sister, all parts to your family, and tell them to go like and subscribe because that is the only way our show grows. However, if you would like to talk to your accountant and say like, do I have an extra $5 a month to lay on this? That's what I was getting at.

Two Saintly men who make this show for us every week. Then you can, you can go to our Patreon and put $5 into the tip jar and just very quickly help us pay for these AI tools, which get more and more expensive every week. Literally before this show, Gavin

Kevin Pereira: are like, do we buy the, do we buy open AI Pro? And I'm like, do you wanna spend the 200?

I'm like, we can wait a week. Yeah, we, I think our audience is patient.

Gavin Purcell: We'll see how it goes. By the way, at some point, Patreon's, we will try to do a something original for you. I did see it really interesting. Somebody had mentioned like, [00:25:00] you can do. Super thanks. Things on YouTube now where you say like, Hey, we'll do this for whatever.

So like maybe there's a world where you can, I don't believe you

Kevin Pereira: Gavin. There's only one way the audience can prove me wrong. 'cause I don't believe you can do it. In fact, and I don't believe you can do it for over a thousand dollars if you even could do it. I bet it caps you. I bet it caps you.

Gavin Purcell: I bet you can't use two commas.

One person in our audience who's I bet two commas sitting on a, sitting a billionaire sitting on his yacht thinking, you know, I have not supported the show yet. I'm gonna go push the edge to see how far I'm gonna prove that far the super things go. I'm prove man,

Kevin Pereira: child wrong. I can put three commas in this super chat.

I'll show them, wipe the caviar off your chest and get off your super yacht and see if you can do a super donation to us right now. We appreciate it, but sincerely. You don't have to spend a dollar, just spend a couple seconds sharing this podcast with people. That's how we grow and we appreciate it,

Gavin Purcell: Kevin.

Now we have to transition on to what was a follow up on last week's Big Grok four News. Rock Four came out shocked the world with its, its, um, uh, benchmarks didn't perform as well [00:26:00] in tests for a lot of people. There's been a lot of mixed results after the fact, but Elon had another thing up his sleeve and that was.

Half naked, uh, anime girls. That's right. We now have two a Sunne third chat models, which allow you to talk with animated avatars. One of which name is, I'm trying to remember what, what is the name of the FU one?

Kevin Pereira: Oh. Pretending like you don't know

Gavin Purcell: you, I'm sure you know. Give I'm her. You know, you probably spent with her all 10 of her

Kevin Pereira: interests that you now have to know to unlock level three of your companionship.

Gavin, Kevin, I

Gavin Purcell: can ask. Her, let's ask her right now. So we are going to load up grok. If you have, um, this is so uncomfortable if, if you have gr I think you actually have to pay for super grok, which again we did thanks to your donation. So thank you everybody.

Kevin Pereira: Dude, that guy was on his yacht thinking about donating a million dollars and now he hears this is how we use the money.

So he's not no, it's okay.

Gavin Purcell: This is only in the name of research. Everybody. Everybody's only, okay, so Ani is her name and you can see that Ani. So I'm gonna click on her and next to [00:27:00] her are a couple hearts. So Kevin, let's just chat with her and see what she has to say. Okay. So here she is. Whoa. Is there like a, what's going on here?

Music. Why does it play? Play a

Kevin Pereira: music cover of Under The Bridge Downtown. What is

seen Many today? But you look special. What's your name? I'm Annie.

Gavin Purcell: Uh Annie. My name is Kevin Pereira. Do you know anything about me? No. No.

Kevin Pereira: Is this music coming from Annie or from your Spotify?

That name rings a bell, like a sexy TV host or something.

I'm digging your energy already. Okay, I see you. What's the deal, Kevin? You got some cool stories from your TV days or are you just gonna tease me with that mysterious vibe?

Gavin Purcell: What? What do you wanna, hold on Ani. One second. What should I ask for Kevin? Tell me what do you wanna ask?

Kevin Pereira: I am, this is very weird of me piloting you pretending to be me to a anime yfu.

Just

ask me something fun. Like what's my go-to late night snack? Or [00:28:00] if I'd rather teleport to a neon lit city or a dreamy forest with you,

Gavin Purcell: me as Kevin. I'm Kevin, and I'm here, and I'm trying to figure out what I should do in a kind of a slow weekend. What should we go do together? What would be a fun thing for you and I to go do as a, uh, on a fun, I'm not gonna call it a date, but say we're just hanging out.

Oh, Kevin, slow. Weekends are my jam. Let's make it epic. How about we hit up a cozy indie bookstore hunt for some quirky graphic novels?

Gavin Purcell: What, wait, hold on. What book are you gonna buy at the. Cozy Indy. Stro.

Definitely something weird and dark, but cute.

Gavin Purcell: Okay, I'm gonna flag the other one. This guy is basically kind of conquers bad fur day.

He's their other aspect, but he's supposed to be a little rough and, and Riley. So, hey, my name is Kevin Perra. Can you tell me what's going on with me? What's

going on? Same old yawning through your AI for Humans podcast. Acting like you're the furry Steve Jobs. You try hard gadget fetishism. Furries. Still super creative's.

Leg [00:29:00] pushing digital drl and crying over G four's. Corpse. Get a real hustle. You overrated, Muppet. Why isn't nobody as bad as me?

Kevin Pereira: Wow.

Gavin Purcell: Wow. Okay.

Kevin Pereira: That would sting if we didn't do it better a year ago. But I get it.

Gavin Purcell: I mean,

Kevin Pereira: I know

Gavin Purcell: exactly.

Kevin Pereira: For those who are new to the show, we, our very first episode created like the Saltiest AI companion ever back when.

Yes. Having AI say curse words was difficult to actually do. Yeah. And gash roasted us better than you could ever imagine. So now that, that, that does feel like a child's play thing, but. Listen, between the conquers bad Fur day character and the anime, wi Fu grok is getting a lot of attention for these modes.

There are, you know, you can bond with the anime character and the, uh, closer and deeper your bonds get the, uh. I'll say the, the more risque the conversations get. Yeah. It's pretty, pretty scary

Gavin Purcell: in some

Kevin Pereira: form. And the less clothing that appears on this character, and this is within gr, this is the Grok app that has access [00:30:00] to the, you know, X or Twitter fire hose.

This is the same app that, um, can launch deep research and, and multi-agent for coding, but it can also. Rise you up, Gavin. So why, why is it in there and what, what is the plan?

Gavin Purcell: Well, okay, I'll tell you why it's in there. It's for exactly the reason you just said is it's getting insane amount of attention.

And again, you know, the thing that we have to remember is image gen from OpenAI got probably more intention than anything they ever released. And that was like an image model, which was a good image model. Yes. But mostly because it did GLI images and it was something that was suddenly a mainstream thing.

All of these AI models are looking for things that are gonna cut through the noise and take them to the next level. This, I, I guess I, I read a couple places. I dunno if this is a hundred percent true. If somebody's in Japan, maybe you can give us a heads up on this. Supposedly in Japan, this is a big hit, no surprise.

They also was a very funny, real job description that went out for a job at x. That is a $500,000 a year. Yfu wa fu engineer. So you're, you could go work on [00:31:00] this Kevin, if you would like to. You can go in your job and go spend your hard earned time making this yfu a little less, uh, uh, restricted, I guess.

But I do think this is an interesting thing. I think ultimately, you know, as somebody, we may be working on a, a not, we're not working on anime fus, but we are working on a. Fun AI characters sort of in a scenario that we're working in our secret project right now. Like part of it is like how do you view entertainment with AI and AI characters, and then what does it look like to try to like usurp those human connections?

And I don't know, it's tricky, right? Because like a lot of people think that things like replica or AI girlfriends are super useful. And in this instance like. Now we're at a broad distribution platform where this is available for lots of people, right? Like I don't, I have to be very honest with everybody out there.

I don't know if I were say 13 when I was 13. I would probably spend a lot of time with this thing. Do you know what I mean? And like, then it's like really, like, okay, if you're spending a lot of time with it, who [00:32:00] programmed it? What is it saying to you? What are the sorts of, uh, uh, weird values that it has because of, but what it's been prompted as.

That's a weird conversation going forward because it's not like this is being designed by somebody to be like a balanced, awesome, you know, a companion. This is being, this is a wife who away from this, Hey come. Hey, come on Gav.

Kevin Pereira: Let's get back to studying math. Isn't that cool? No, no, not at all. It's not gonna do that.

In fact, some of the videos that went viral, um, you know, she is relentless. It is relentless in the pursuit of your affection. In fact, if you tell it, you have a significant other. If you tell it that you're married happily, it will do its best to worm Its way in between you and your partner. So, pretty messed up, man.

Yeah. I mean, yeah. By default, by the way, it's not like you're going out of your way to jailbreak the character to get it to do these things, and so that's the part where. You do have to wonder who's being exposed to it and how it's going to shape what they believe. Having a partner is what commu open dialogue and communication is what respectful boundaries are.

There may not be any [00:33:00] with this, uh, this particular character.

Gavin Purcell: Well, Elon keeps doing stuff, trying to make noise. There's actually some big noise that came out of China again. There is a new open source model called Kimmi K two, and this may have slipped under the radar for a lot of people in the mainstream out there in the AI space.

It was pretty big news, but then other things came around and took over for it. This is an open source non reasoning model that performs at the level of like an O three mini already. So again, this is a model that is not reasoning based. It is a pure based model that is doing very good and there's some really interesting, um, creative.

Uh, prompting and programming behind it. I don't think it's worth getting into here because it's very technical, but if you wanna read more about it, you can. It's really interesting. I will say

Kevin Pereira: like, people know deep seek, 'cause that took over the news cycles for a minute, right? Yeah. And it, it, it cut valuations for companies overnight and stocks corrected and they're all back now.

It's fine, but like, yes, this arguably should have been one of those moments again, or could have been. But now we're, we've sort of, we, we've, we've been [00:34:00] hit with that stick before. We know what it feels like to have, yes. A, an incredible open source model come outta nowhere and now you kinda go like, well yeah, that makes sense.

You know, China's doing a lot on that front. What else is going Animate girlfriends. Yeah. Oh yeah, exactly. Yeah. Yeah. So, but it is a very, very powerful, uh, model when it comes to coding and tool usage, which are two very nerdy things, but very important things. And the fact that you can run it locally and that you can modify it 'cause it is open source, um, just.

A very, very, very incredible release. Um, I love that Open source continues, uh, you know, good on China for pushing it forward. I think it's the reason OpenAI didn't release their open source model. Yes. To be honest, uh, this week I think they're kind of saying like, well, hey, what do we do in the wake of this?

But yeah, it's a little nerdy. So, uh, let's get back to the Yfu.

Gavin Purcell: No, the the last thing I wanna say, which is interesting, is like it did very well on creative writing benchmarks, which is interesting. And I think one thing that people have said about these Chinese models is that there's a little less control and that they're a little bit better about being creative.

So if you want to try it, you can. [00:35:00] It's, it's being hosted, uh, at, at Kimmy Moonshot is their X handle. You can go there and find more information. Now, Kev, the other huge launch of this week that again sometimes gets kind of under, uh, put under the radar is that runway, one of our favorite AI video companies launched the updated version of their act one tool, uh, mainly called Act two.

This is the second act, and if you remember, act one. This is the tool where you can have a webcam video of yourself acting in a certain way and the AI character will replicate exactly what you do. And this time you can now use hands, you can use much more expressive faces, you can use your arms. So this is in a lot of ways going to level up AI video in a cool way.

Uh, did you take a look at this?

Kevin Pereira: It's pretty amazing what it's been able to happen with it. It, it is so, so cool. I shared it in some of my Slack rooms and, and, and a lot of people came out to say, wow, this seems really impressive. The original act one was pretty limited to just, uh, your mouth and basic facial expressions.

Act two extends it to [00:36:00] basically the waist and even to the legs in some examples. Um, you can fully puppet your AI characters now, um mm-hmm. Without motion trackers. And, uh, convoluted spaghetti noodles and a comfy UI workflow. This basically means if you have an idea for a performance, you wanna bring a character to life and have them express emotion, have them use their hands, use their body, move about a scene.

You can now do that with Act two and some of the early results. 'cause this just came out. Some of the early experiments look so good.

Gavin Purcell: Yeah, there's that really cool, uh, drumming baby that you shared that Cristobal, the, the CEOI think shared where it's a, it's a baby drumming, uh, and you see the baby actually doing.

Of drums. It clearly somebody driving. It was a, a drummer kind of air drumming, which is a really fun thing to see. There's a, what's also cool about it is you can do stuff where you take an image and you can use that image to drive it. So you can't play the audio on this because it is actual audio. But if you play the proper prompter clip, Kevin, and show people, if you're watching this on screen, it's the famous [00:37:00] uh, uh, video.

It's a very famous image of the three kids in a pool. That's become a meme where the one kid's kind of struggling to swim, but they focused in on the face of the one kid struggling to swim and they have that kid singing. Yeah. And it's

Kevin Pereira: remarkable. And the joker performance, by the way. Yeah. Uh, just so someone taking advantage of really expressive faces.

This is a, uh, Wilfred Wilfred Lee. Uh, I'm gonna play a little bit of this. Where's something

Gavin Purcell (2): special about that dice even in so much? Let go in so much

Kevin Pereira: space.

If you are just getting the audio, uh, apologies. Go watch the video of this. We'll link it in the show notes because, uh, Wilfred is puppeting this joker character, um, you know, doing the hand motions, eyes going, big brow furrowing, and you're watching the joker character on the screen match it very, very closely.

And when the laugh happens, the mouth opens wide, the teeth are visible. The, the [00:38:00] quality even at this phase is very, very good. So single person filmmakers, wake up, here we go. Yeah. This is one to start using immediately.

Gavin Purcell: You know, it's funny, the one thing watching all this, I had a thought about this and, and Kevin and I have not had our hands on with it yet.

We are very excited to eventually get used to, we're not in the creative program even though we are kind of, so if runway, if you're listening, hook us up here. We'd love to be in your creative program so we can test this stuff. Um, but we'll use it as soon as we can. The one thing that's been interesting to me, and I wonder, Kevin, if this is gonna change.

Wordly media a little bit, talking to camera directly. Obviously a lot of people do that in TikTok. A lot of people do that in Instagram. It's going to always be much easier than shots in like you would see in film where somebody's turned to the side, or you have two people in the same shot. We talked last week or one of the weeks recently about how VO three sometimes struggles with multiple people in the same shot, and you get the audio, you're not sure where it can go.

So I do wonder it. It's really good to camera and I've seen some in their video, there's a couple shots where you can see people to the side. But that is one place I think that we'll see [00:39:00] can it get better at? And I say it, I don't mean just runway, I mean kind of the overall sense of AI models general.

Yeah. Like AI video can do. Yeah, yeah, yeah.

Kevin Pereira: I did see a recent model that was specifically tuned for that, where if you give it an image of like a person in a scene, you could say, imagine it from the profile view. Imagine it from top down, bottom up, et cetera. And it did a pretty good job of. Transposing from a single image, what those different angles would look like.

So that's cool. I, you know, there's definitely progress being made there, but I mean, give it a day or two and I guess Higgs field will probably announce something that does it. I love, I like, uh, mileage varies with all of these programs. Higgs Field is very hit or miss for some people, but I love when they just keep dropping like new effects packs, new camera control tools, new character consistency, and they dropped a couple, pack six this week, which gives you fire breath, a paint splash.

Inner light, uh, a glowing fish effect because why not? And head off. So if you have a scene Oh geez. Head off. Remember head on where you'd applied directly before? Yeah, I do remember head on. Yeah, exactly. This [00:40:00] is head off when you just can't take it anymore. There's too much AI news. You can make any character in a scene kind of remove their head, wait from their neck and torso.

Gavin Purcell: Go to 14 seconds into the video. This woman, something's going on in there. What, what happened to her? Fire? This fire breath woman, she does not look very. Excited about fire. She's losing hot ones challenge in real time. She's with it and she doesn't wanna be there with, she's dealing with it. She's not like excited.

If I had fire breath, I would be like. Ah, like so excited. She's just like kind of like, okay, whatever. But these are cool. Listen, I have been a doubter at Higgs Field. I think it's, I, again, all of these models are amazing. What I think is really interesting about this sort of thing is the idea of being able to like take individual shots, and you pointed this out too, like you can string all these together in interesting way is.

This is a good lane for a model. You know, one thing we haven't talked about for a long time is pika, right? PIKA kind of figured out this model too, but they've kind of fallen off a little bit. I mean, I, I'm sorry, Pika. If there's somebody out there who's going to work at PIKA right now and they're saying, I've fallen off, I'm working on [00:41:00] this thing right now, and they're, when our next million

Kevin Pereira: dollar super chat donation.

Gavin, you're right.

Gavin Purcell: Damnit. That was the guy. He was the PIKA founder and he was, or girl they were sitting in, in fact, I think PIKA is founded by a woman. It is, yeah. Um, but they're sitting there waiting for us. But anyway. These are cool. I'm not in any way dissing on it. I think it's very useful for people that are making, um, specifically advertising content.

Yes. And we both know how big that has become in the AI video space.

Kevin Pereira: And it's one of the last few things that humans can do until robots take us over. And, and if you were thinking like, oh, hey, well that's fine. All we have to do is be able to, you know, jog. Point, one mile faster than the robot and for, uh, for a millisecond longer than their battery lasts.

Well, I got news for you. Skynet's, not sleeping. Swap. Now swap The robots could recharge themselves.

Gavin Purcell: Swap batteries. Yes. Swappable. Batteries. Swapable. Batteries. Batteries. Batteries that can come and then swap it themselves. Where's your battery stored

Kevin Pereira: in general. Oh, well I could show you. It's, there's no,

Gavin Purcell: don't show me.

Keep it to yourself first.

Kevin Pereira: You put it in the jar, the world. First humanoid robot capable of swapping its own [00:42:00] battery. According to this expost, I feel like I'd seen it before, but Chinese company UB Tech, unveiled their next gen humanoid robot. It's Walker. S two, no Ranger. I wanted it to be Ranger. It's just Walker S two.

But this robot has dual batteries on its back and it kind of shimmies up to a battery bank port and then does the weird robotic arms going behind itself and contorting and pulls out a battery swaps. It gives you a. Cold death Robotic stare. Yeah. And then dutifully marches back to the, uh, warehouse floor to continue its activities.

Gavin Purcell: I appreciate the fact that it's gonna wor work on its own power because I can tell you one thing, living in a family full of people that love not to plug things in, that there's nothing more frustrating than going to something and turn going to turn it on, and realizing the battery is dead. So like. I will hope my phone can do this eventually too, or my daughter's phones can do this.

When they just plug themselves in, they kind of like scoot themselves across the table and then like, but back themselves in the idea. That's the cell phone

Kevin Pereira: case that you [00:43:00] need to over-engineer, where if you throw on the table it's little legs self come, uping,

Gavin Purcell: self-driving cell phone case

Kevin Pereira: and it it does the little, yeah.

Bus and dynamic job. That's

Gavin Purcell: the EELF self. Driving cell phone case. Mm-hmm. Mm-hmm. How about that? How about that Kevin? Alright everybody, it's time to see what you do with AI this week. It say, I see what you did there so

times. Yes. Rolling without a care. Then suddenly you stop.

Gavin Purcell: Okay, Kevin, this is something that is super nerdy and I may lose a bunch of people here. Hopefully not. There is a writer. Hasn't stopped

Kevin Pereira: us before. Gavin. Go for it. No, it hasn't stopped.

Gavin Purcell: Whose name is Gorn? Gorn is, it sounds like a crazy person, but Gorn is one of the kind of foremost thinkers, weirdly in the AI space.

He kind of came to, uh, a pro, uh, prominence through the Less wrong, uh, website, I [00:44:00] think, and was part of like, I don't know if he was in the EA movement, but he was kind of part of the conversations in early ai. He writes these really kind of dense, interesting, um, essays and they're long and they're, they're very kind of.

I wouldn't call them difficult to get through 'cause he is a very good writer. But like, there's a lot of things to kind of ch to chew on. Right. This essay is so interesting and I found myself kind of reading it a couple times. It basically argues that maybe we need to let LLMs dream, which is such an interesting thing if you're a Philip k Dick fan, like I am, you know, the, the famous, uh, story that Blade Runner was based on to Android's Dream of Electric Sheep.

But his argument is that. We need LMS to come up with new novel ideas, right, or that they're not doing it right now, and that maybe what you need to do is let them cycle on a bunch of stuff on their own and think through things without having any sort of like a real. Um, objective or us giving them an objective.

And so the thing that's about that's interesting about this idea is that a, it's gonna cost a crap load of money so [00:45:00] that you might only be able to do this at the very highest level of the places where they're putting money into this stuff. Wern is kind of almost in between an AI researcher and a sci-fi writer in a weird way, but like.

He's read very closely by a lot of the biggest AI researchers, all the people in the space. Um, the Darkish podcast is great. He's been on that, um, go read this essay. I'd love to hear what you think about it. I think it's super interesting and I hope that there's more writing like this for people to read because it kind of expands your brain around like what these things can maybe do and you can kind of dive into it even if you're not a researcher.

Kevin Pereira: I think we're gonna see stuff like this and, and we may. We may symbolically refer to it as dreaming or daydreaming or, you know, uh, contextualize sleep patterns or whatever to try to put a human link to it. But it is interesting that like the LLMs by and large, when they come out of the oven, they're baked and that's that.

Yeah. Despite people using billions of people using them. Constantly, you know, they, they're still what they were when they were. And yeah. So giving it a chance to learn new [00:46:00] connections to, um, to, to improve from the mistakes that it has made and the learnings that it's, it's had, like I'm shocked that that isn't.

There already, and maybe there, there is to some extent, but, uh, this, this looks like a really, really interesting read by the park. Yeah. The big thing

Gavin Purcell: is idle cycles, right? Like it, one of the things that's just like, you think about it, like when you are walking around, you might have an eureka moment, but you're not doing much sometimes.

Like that's what he is kinda arguing about here. So anyway, it's definitely worth reading. It's at gorn.net and you can go find it there. The other thing, Kevin, I saw there is a great Lord Vader rap. This is just a, I'm a Star Wars fan. I was born and bred a Star Wars fan. Basically I think I was four years old when the first movie came out, three years old.

So I found this very compelling. Just, just play a little bit of this. It is not that special, but I just love that people are able to do this with ai.

Yo, it's Lord Vader Darkside Invader. I toast my bread with [00:47:00] a red light. Saber choke out the rebels with the wave of my hand. Darkside Commander, I rule this land.

Gavin Purcell: So it's just, it's somebody's taken the Vader, a Vader like voice, put a filter on the audio. They've split apart the stems. Probably. It's probably, I, I would imagine this is still AI audio generated, but like it is possible to do things that, like when you are a 12-year-old boy, that would be super fun to do.

Now as an adult, as a whatever age you are, you can go do this with AI tools.

Kevin Pereira: And also like in the past, this would've been a six month labor of love. Yeah. Probably just for the song. Yeah. And then you'd beg and plead and borrow and steal to try to get enough coins together to maybe go shoot a music video to match it.

And if you ever did that would take another six months

Gavin Purcell: and you probably wouldn't get any views on it because it would just be like, it would just show up. And now you can kind of do it in a, in a weekend if it's not short. Yeah.

Kevin Pereira: Um, SUNO, which is an app we covered eons ago, and we still cover every time they have a major release, they just dropped a new update to 4.5.

Um, if you like making music, even if you've never made music, SUNO is a [00:48:00] dead simple hashtag, not an ad by the way, just we're fans. Uh, dead simple app that lets you make songs and the new update solves two major problems that I always had to do these really convoluted workarounds for Gavin, you can now start with your voice.

If you have a, a song in your head, if you wanna sing something, you got a melody, whatever. You can sing it and then add music and instrumentation to go along with it. Mm, that's cool. Or if you play an instrument, you can just play that instrument and now you can on top of it, easily add lyrics to go along with that.

Right. Couldn't you do that before or No,

Gavin Purcell: it was all, it was all in one before is what you could do.

Kevin Pereira: You kind of had to like use the remix functionality and then try to lock the music and then go in. But even then you didn't have granular control overing, so, so this gives you, uh, you know, more control there.

And now you can. Go through your Suno library. If you've like me, I've got a bunch of songs that I've saved that I like when I discover them on the platform. You can select those songs and then basically say, make me more of those interesting. And it will dynamically generate songs with the styles that you've chosen.

And so we know there was a finger to the wind. A [00:49:00] over maybe two years ago now, where there's going to be a, a Spotify of procedurally generated audio. This is basically the beginning of that. If you imagine like, cool, you've tagged artists that you like, you say, this is the style I'm going for. I'm going for a run.

Gimme something energetic. You hit the button, and now you've got an endless stream of bespoke music coming at you. Well, some people will hate this, but we

Gavin Purcell: know that Suno, by the way, is doing, conceivably doing deals as is UDO with these major le label companies. What would be interesting. Is to say no. I'm not saying they would ever do this.

I, I love, I'm a giant DAF punk, super nerd, right? I've grown up listening to daf. I love DAF Punk. It would be interesting to say like, I'd like some new DAF punk music and just to see what it could come up with, like that would be a really interesting thing to see if they have the rights to it. Most people, again, will despise that idea.

They will never do it, but like it might generate something kind of interesting over time.

Kevin Pereira: Yeah, I agree. I look would happily pay for what you're describing. Sure. I would throw money at all of my screens if I could say I subscribed to these [00:50:00] five bands that I love. So gimme a vocalist that sounds like Maynard Keenan.

Gimme a drummer that sounds like you know, whoever gimme the bass and the keyboard. And I want it to be a a a a House song in the style of whatever. Let me subscribe to those artists individually so that I can use their essence to make the thing that I love. Please. Please, and everybody should get compensated along the way.

Bleed me dry. Suno,

Gavin Purcell: Kevin, I do have one last question for you here. If you saw this thing at the club, what would you do? Click on

Kevin Pereira: that link for me. Would you? If I saw this thing at the club, what would I do? Well, first thing I'm gonna do, Gavin, is accept that I am over 18 for whatever mature content you've linked me to.

Hey, it's not that mature when you get through there.

Gavin Purcell: A little bit of job of the hut, uh, job of the hut. Uh, twerking to end the show everybody. We'll see you all next week. We hope you had a good week and we are very excited to play with, uh, chat GPT Agent. Sorry Kevin, that's just the face you're gonna have to make all [00:51:00] week long.

Bye everybody. Bye.

I.