The Next Wave - Meta Replacing Creators? + Sam Altman’s Mistake & 3 Big AI Updates

Hey, welcome to the Next Wave Podcast. I'm Matt Wolfe. I'm back once again with Joe Fear, and today we're going to get into some really cool stuff that came out from some of our favorite AI companies. OpenAI has a new visual explainer thing, and Claude has a new visual explainer thing, and Canva has a really cool tool that we've been playing around with a lot. We've got some weird stuff that came out of Facebook this week. We got some robots to talk about. Lots of fun stuff. What do you think, Joe? Should we dive into it?

We got a busy week. Let's do it because it's all visual. I feel like a ton of visuals this week and we can all get our hands on it like now. Well, after the podcast's done, right? Yeah.

Yeah. Wait till after the show. Come on, Joe. What are you trying to do to us? Let's go.

So first one.

You're got it. So where are we starting?

Let's start with ChatGPT. I think it rolled out like, I don't know, it was a day earlier than Claude's version. And, uh, essentially what we have here is ChatGPT. It's, it's kind of kind of limited, but now they're starting to show these visuals and like animations within your actual chat. So it's pretty cool. It's like a visual way to learn and maybe describe something. Show it off, Matt. I know it's a little limited, so let's see what we got.

Yeah, yeah. So what this is is basically a tool where you can ask it like math questions and it will create these like visual explainers of the math question. But you can see it actually has sliders that update in real time. So as I slide these things around, it actually updates the visual of this math problem. I don't know how to say this math problem, so I'll let Joe say it.

Squiggly doogly boogly boo.

Accurate, right? But this is apparently the algorithm for a concave mirror.

Oh, that's what that is. Okay.

Yeah, but I actually don't know what the various symbols stand for because I are not smart when it comes to math.

I are not smart. Yes.

We've got the Oops, we got the Pythagorean theorem here. I do know this one. A squared plus B squared equals C squared. You use that to find the, you know, the long end of a triangle.

Yes.

But you can actually drag your A around and your B around and it shows you how it adjusts the C line on the triangle here. You have the ideal gas law, which again, I don't know what any of these stand for.

This one's more fun though, at least.

The V, I'm pretty sure is volume, right? Because if I slide this up, you can see there's more volume in there and it's actually animated too.

That's cool. And the N, what, the number, I'm assuming?

Possibly, yeah. The number of gas molecules, I'm guessing. And then the T is like a speed, like how much it's bouncing around, but I don't know what the T actually stands for. I should have paid more attention in chemistry classes and physics classes, but I didn't.

Hey, it's another time to relearn all this stuff and I mean, it's a hell of a time to learn, don't you think?

It is really. But so here's the thing about this stuff though, is I saw this and I took it to mean you can go and prompt like any sort of mathematical equation and it's going to give you a visualizer with a slider and, and actually do this kind of thing for you. But that's not quite how it works. There's actually, I think they said about 70 predesigned little like algorithms that you can play with and you have to give it one of those and it will load almost immediately. Like we can jump into ChatGPT over here and I can say, create a visualization to explain how compound interest works. And theoretically it'll just generate that same sort of thing that we just saw. Nope, it's coding. It's not even doing it for me. It's writing some code. That's not what it's supposed to do. Hey, if you take a look at my web presence online, it's safe to say that I'm a bit AI obsessed. I even have a podcast all about AI that you're watching right now. I've gone down multiple rabbit holes with AI and done countless hours of research on the newest AI tools. Every single week. Well, I've done it again and I just dropped my list of my favorite AI tools. I've done all the research on what's been working for me, my favorite use cases and more. So if you want to steal my favorite tools and use them for yourself, now you can. You can get it at the link in the description below. Now back to the show.

So what's the thing like? Because I know you told me it's very limited in the amount of animations and things. Is there a way to just find the ones that are Pre-done? Like pre-ready for you?

So I did. So here's, here's one that I did. Show how compound interest grows over time, right? This was a prompt that I entered and it actually gave me this compound interest chart here, right? So this one actually worked when I gave it the prompt. Uh, I must like, I think you just have to word it right to get it. So let me try a new chat here. Let's paste in the exact same prompt. Show how compound interest grows over time. And there it goes, right? So this time it actually gave me back the exact same chart we were just looking at. It's the identical chart to the last time I prompted it. That's because these are all prebuilt.

It's so interesting. I'm just going to say I prompted the exact same way on my side with the same model and did not get that result.

Interesting.

It just gave me what a line chart is and the best visual that I got. All written out for me. Not as cool.

My ChatGPT took a little while to figure things out. All right, so in ChatGPT, I can say, did you custom code that visual or is that a pre-built visual? And you can see it's a pre-built visual, not something I custom coded on the fly. What happened under the hood? ChatGPT has a library of pre-built interactive learning visuals, about 70 of them. Each one corresponds to a specific formula or concept, compound interest, Ohm's law, Pythagorean theorem, exponential decay, et cetera. So when you asked about compound interest, I triggered the visual tied to this formula. And so you could see these are like what they've got pre-built in. I feel like they kind of announced it as if like it was generating this stuff on the fly, but that's not what it's doing. It's just got a bunch of these pre-built into it.

Yeah. Even though they look really cool, but they're all, they seem very sciencey, very mathematical, very complex. Like I don't know what most of these are.

A day later, was it a day later? So March 12th, this first one came out on March 10th. So 2 days later, Claude now creates interactive charts, diagrams, and visualizations. Imagine that.

I love how they're always synced up with all these releases. Yeah.

Somehow like. I mean, what could have happened was like, they'd been working on this behind the scenes at Claude. They saw that OpenAI released it and they're like, oh, we built that too. Let's just release it and be like, look, we got that also now. I don't know if that's what's going on, but it's like, why is it so coincidental that like almost identical features come out in the same week oftentimes?

Yeah. But this one, I mean, my take on the Claude thing, because I was actually using Claude a little bit more. I like the vision. They're different. You know, I think Claude, I don't know if they're pre-programmed like ChatGPT.

Claude is not. Claude, it actually seems to build it for you when you prompt it.

Okay. Can we show off some of these?

So one of the, like their, their example here is like this interactive periodic table of the elements where they can like move their finger around. And as they click on the different element, it shows the details down at the bottom of the element. So that looks pretty cool. But yeah, we can, we can go and test this.

So if I go over to my Claude account, compound interest, I know works and is a good demo. Why don't we paste in the same exact prompt from ChatGPT and toss her in here?

Create an interactive demonstration of how compound interest works. This one, I know you don't have to give it like exact specific specific perfect prompts to get it to work.

Yes. My take is I kind of like the outputs over here a little more, even though the animations are cool with ChatGPT. Granted, they are pre-rendered. These ones I feel like are a little bit more interactive and they, I don't know, they look kind of, they remind me of like, like this graph once it pops up, kind of like a Wealthfront or something like some financial actual dashboard.

Yeah. Yeah, I mean, this one will mess up a little bit more too though, because it is trying to, I think, write some code behind the scenes. These aren't like pre-cached, like demos that are already baked in. So I have had some of my demos fail on me where it just like, it didn't work the way it was supposed to.

But check that out. I mean, that was fairly fast for that.

Yeah, I think this did a better job than the pre-built one that OpenAI already has, right? So you have your initial investment. You can see how it adjusts the graph. You've got your time horizon of, of, ooh, look at that.

Look at that money.

Look at, ooh, I don't know why that's so satisfying. But you got your time horizon here and then you've got your annual return that you can set. And then you've got your monthly contribution, contribution. This one's really fun because it makes a little wave.

I was doing the same thing yesterday. Love it.

And apparently if you start with $3,000 and you invest $1,000 $1,000 a month for 36 years, you'll be a millionaire in 36 years. Let's go, let's go 15 years, but let's say we're getting a 10% rate of return. Yeah, there you go. But if we start with $10,000, what if we started with $1 million?

I mean, no, I'm just kidding. It's the power of compound interest, y'all.

If you start with $100,000, a 10% annual return with a monthly contribution of $1,000, Over 15 years, you'd have $859,862.

Mm-hmm.

I hope in the math checks out. If you do it for 30 years, you'd have $4.2 million shamolians.

Look at all that interest gained. Isn't it Einstein talked about one of the wonders of the world, right?

Yeah.

Compound interest.

All right.

What else? What's another, like, I'm just thinking what's a visual that we can create here?

Let's do one of. Make an interactive, ready for this, Padres schedule for the 2026 season.

Matt and Joe need to find some tickets to go.

I guarantee that this wouldn't be a pre-baked-in one.

You don't think they're secretly Padres fans over there at Anthropic? So it's not fast.

No. See, the other ones are, they're basically cached in there, right? Like they're already prebuilt. So when you ask it, it's like, oh, I've got this ready. Boom. Here it is. This one. It's like, let me go build that for you.

See, that's ChatGPT trying to look all really good and fast. And, but it is limiting.

I mean, they, if they're fooling you, they're fooling you.

It'd be cool if they had a little menu though to kind of help you out. It's a little bit easier.

Yeah.

Yeah.

Yeah. They do make it kind of difficult to know exactly what formulas you can use. Like, and it even said roughly 70, like it had the little squiggly line, like approximately 70, right? Like, why don't you know the exact number?

It's your data, your own pre-rendered things. Come on.

Like, you probably know the exact number. It probably is 70, or maybe it's like 68. They're like, we're rounding up. Let's make it look better than it really is.

Yeah, I'm sure there's some tomfoolery happening.

Woo-hoo-hoo-hoo! Look at this.

This is purdy.

So it built out a whole schedule. Opening day. It says the time. And I think this is accurate too. We looked this up in a past one.

I will double check, but I do believe it should be accurate. The first regular season game claims to be— what's it say there?

It says Detroit Tigers, 1:10, Petco Park.

That is correct.

March 26th.

Correct.

So yeah, I mean, it's got it and look, it is actually interactive. I can click on them and open them.

Wait, so what's your take? Cause this isn't exactly what we showed with the whole like interactive.

No, no, this isn't, this, this looks like it just coded up something for us. Yeah.

It kind of took the liberties to just build something.

Yeah. It's not that same like visual explainer thing. Um, but I made some other ones the other day, like, um, so I made this like AI create an interactive timeline of major AI model releases. And you can see on this one, it actually built out the timeline right in line with everything. So when I asked for the Padres thing, I thought it was going to make something that looked like this, but with the Padres schedule, just build it right in line with the chat response.

We got to figure this out, Matt. We're going to figure out what's the prompting structure that makes the difference because you prompted it very similarly.

Yeah. Create an interactive timeline of major AI model releases from 2018 to 2026. And it even made it with like these filters. So like, let's say you wanna know about just image models. I click that and it filtered it down to just milestones with images. I mean, it missed a lot, but you know, you get the idea.

Yeah. But what's the prompt structure?

What's the thing? For me, it's been asking it to build an interactive X, right? Let's try one more.

Um, periodic table one is really, really cool. Like on their, on their tape on their thing. Yeah.

I was kind of hoping the Padres thing would either be like a calendar view where you can like hover over the dates on the calendar or like a timeline view, like we just looked at of all the games and it kind of just went out and wrote some custom code and then opened it in the like side canvas.

Um, it says this feature's on by default. I'm just reading the, uh, thing while you're prompting.

Create an interactive explanation of what was one of those other ones that we were looking at on the OpenAI one. Oh, on their site. What was the gas one we were looking at? Ideal gas law or something.

Oh man, I closed it. But there's like surface area of a cone formula.

Okay. Create an interactive explanation of the surface area of a cone. Right. Does that sound like a good prompt?

Yes.

Because we already know Opus and Claude are capable of going and writing code for us.

Yeah.

I like it. Like I want to see it doing it in line with the chat.

Yeah. All right.

Here's an interactive breakdown. Drag the sliders to change the cone's dimensions. Watch everything update in real time.

Now we're talking.

All right. So it looks like it's actually doing what we asked it to this time. Yep. We got sliders populating.

Yeah.

Ooh.

I don't know if this is possible, but Perplexity tells me that, uh, maybe prompt things like draw something that you describe as a diagram so you can explore it visually. Like make something in your chat visual, essentially.

Actually, we could try that next.

All right.

So we've got our interactive explanation of the surface of a cone so we can change our radius.

Yeah.

So there you go.

Nice.

So you can change your radius. We can change the height of the cone and it changes our surface area or lateral area. You can see everything updated in real time. How does this compare to the ChatGPT version? So here's what ChatGPT did for the same one. We have less slider options. Like we can change the height and we can change the cone radius at the bottom. Those are the only two options it gives us where if we look at this one, we've got, oh wait, no, it is just two. What am I talking about? For some reason I thought I remember having more sliders, but I guess that was the previous one we were looking at. I actually, I don't know. I don't know which one I think does it better. I feel like the Claude one is giving us more information right here.

Right.

And it also is a little more colorful, but it also seems like it's kind of cropping weirdly here.

It's cutting a little bit of it off.

This one looks a little bit just like cleaner, like that minimalistic, like just looks clean. But it's also not giving us quite as much data.

Yeah, the other one seems like it explains it further. You could probably do more. What's below? Like, I mean, sure. The thing that interests me is like you can, oh, okay. It has a formula there and everything too. Okay. And yeah, doing the follow-ups on any of these, I guess it's really cool. Like now I'm starting to think, okay, you can visualize parts of your conversation. So depending on what you're working on. Yeah, I wonder like theoretically you could prompt now, I want visuals to explain this.

I wonder if I could change the colors, change the colors to purple and pink instead of orange and blue. What do you think? Is it going to do it?

Let's see. I want, I'm going to give some prompts to you to try for other things as well.

Dipping into pink.

Dipping into pink. Hey, it's trying. Look at it.

Maybe, maybe, maybe. Eh, eh? Recoloring the cones. Dipping into the pink and recoloring the cones, Joe. All right.

What time is this? That's 26 of—

Stop. Stop censoring me. Hey, it did it. Look at that. So yeah, you can actually go and tell it to like change what the visuals look like.

Interesting. I don't know, man. I'm thinking the Claude's the way you might have to wait a little longer if you have all the flexibility.

Yeah.

All right. So I gave you a prompt in the chat, Matt.

All right. So we're going to prompt it with draw this as a diagram so I can explore it visually.

I mean, that might not be the perfect prompt, but the idea is like start visualizing your chat. What I did is kind of, uh, what I had Perplexity basically look at that Claude announcement and said, hey, make me some prompts based off of this announcement so I can test this out.

All right, here's a visual calendar map of the entire season, plotting 162 games on the calendar, flagging every Dodgers series.

Did you prompt it to do that? No.

Painting Petco Park brown and gold.

Yeah, buddy.

I don't know though. The amount. Oh, whoa.

Oh, look at this.

This is what we were looking for. And it actually did look at that, bro. So if I click on this one, nothing actually happens. So the interactivity of it is a little wonky.

Look at that.

It actually put home away Dodgers and special date. Like it literally, for some reason decided Let's find all the Dodgers games.

Looking at your chat history, Matt, you must have had a lot of chats about Dodger games.

I mean, it's interactive in the sense that I can hover over it, but like clicking on it doesn't do anything.

But it tried. It tried.

I don't know.

Is it accurate? I'm just going to double check and see if it pulled in the actual—

Are we playing the Dodgers on June 26th, 2026?

Oh, you put a way date. Hold on. June 26th.

6/26/26.

Yes, Dodgers, we are. And it's a home game.

Yeah, the home games are in brown, the away games are in blue.

Okay, so it's pulling in real-time data and applying it here.

Oh, you know what, it is interactive. Look, if I click on one of these games, it changes it down here at the very bottom. But because it's such a long calendar, like if I click on Seattle way up here, I have to scroll all the way down to the bottom to see like the details. So it is actually doing the interactive thing. It's just like, you have to scroll way down to see the interactivity of it.

I mean, it is a long season. We did say that.

I know, but maybe it should have put that like over on the right sidebar or something. But you know, it did exactly what we were expecting it to do the first time around.

For sure. There you go.

And it flagged the Dodgers game by highlighting them in like a blue, light blue border. So it says away rivalry.

I will say, Matt, you and I have a really good record when we go to Dodger games, Dodger Padres games.

They've never lost. The Padres have never lost to the Dodgers at a game I've been at.

Yeah. I think I've had like one or two losses for games, but the ones that we've gone to, yep, you're right. Yeah, I don't know. You got some mojo there, Matt.

We may or may not have snuck out of an event that we were being paid to be at to go watch a Padres game, but I won't say the name of the event.

It was not a HubSpot event.

It was not a HubSpot event. That I can confirm.

So, um, Matt, I don't know if you want to, like, there's a little prompt that could be helpful to share here is, um, it seems like there is kind of a flow that can be used to make these interactive charts. It might not be perfect all the time.

Take what you just explained about topic and turn it into an interactive diagram slash chart that I can click slash adjust slash experiment with directly here in the chat. So this is one that specifically works like if you're starting from a topic. Let's see, maybe we should get it to explain a topic and then test this prompt after it explained the topic.

That's a good idea.

Break down the latest events in what?

Uh, the latest between Anthropic and the Pentagon.

Break down the latest events between Anthropic and the Pentagon. Since, you know, we're not gonna go on, we're not gonna go down a rabbit hole on this one today. We've gone down this rabbit hole in two previous episodes, but at least we'll get a little taste of what's going on Secondhand.

There you go. Yeah, we don't need to go deeper, but here we go. Matt, this is a massive, yeah, it's fast moving story.

So great origin, the ultimatum, the retaliation, the fallout, beyond the government, anthropic legal response, where it stands today, blahbity blah, blah. The bigger picture story for you. Okay, now let's try our add on prompt. take what you just explained about, I'll just put this topic or this.

Yeah.

Turn it into an interactive diagram or chart diagram.

Okay.

Diagram that I can experiment with. Experiment with.

I don't know what we would experiment with.

Directly here in the chat. I don't know what kind of experiments we're going to do with this timeline, but we're learning in real time. Experiment would be like, what if the timeline happened like this? Instead. Tell us the alternate timeline.

Interesting. Different realities, huh? This is how we find them.

Yeah. Yeah. How would this have played out if I think this is kind of cool? The present, right?

Hey, look, here we go. Here we go. But I think this is kind of cool. It's almost like back in the day. I mean, every new thing that rolls out and I'm maybe curious of your perspective here, maybe. on like prompting structure changes different outputs depending on whatever feature just rolled out and it's doing it. Look at that.

Mm-hmm.

But like what we're figuring out now is like, what's the prompt structure that works with this specifically?

Yeah. Well, I think asking it for like using the word interactive seems to be a keyword, right? Like an interactive visual, interactive graph, interactive. Chart, interactive seems to be one of the keys to toss into the prompt here. But this is a nice little prompt. Take what you just explained about this topic and turn it into an interactive diagram that I can experiment with directly here in the chat. It seems like almost any conversation you have, you can throw that prompt at the end of it and get something more interactive from the content that it just fed you.

Copy pasta screenshot everyone. And then in the comments, let us know what you done diagrammed.

What you done did do with the diagram.

What do we got though? I mean, this actually is kind of cool.

All right. So we got the timeline. Claude used in Caracas raid. Pentagon issues ultimatum. Anthropic refuses to budge. Trump bans Anthropic supply chain risk label. Pentagon memo remove Claude in 180 days. Anthropic sues in two federal courts. Pentagon CTO, no chance of talks. So if I click on one of these, ooh, look at that. Dude, clicking on one of them actually opens up like a little TL;DR of each thing. And then if I click on deep dive, it actually prompts it to go further on that topic. Heck yeah.

Look at that.

And then we've got some little, uh, so this is the timeline, but we've also got key players. So Dario Amadei, Pete Hedgeth, Emil Michael, OpenAI, Palantir, and Caitlin Kalinowski. She's the one that just quit, uh, from the robotics lab. and then we have what's at stake, and then we've got more details. Anthropic's annualized revenue, commercial customers, paused or canceled OpenAI-Pentagon deal, removal deadline. That, I mean, dude, dude, you toss that prompt at the end of any long discussion and it's just like, here's the TL, an interactive TL;DR.

Did we just unlock the Matrix? Was this AGI, Matt?

Uh, no, I don't quite think this is AGI yet, but, um, that is pretty cool that you can do that. Like, it's almost the type of thing where if you have like a long chat with, with something like Claude and at the end of it, you're like, wow, that was, that was in depth. There was a lot there. Throw this prompt in there and it's almost like you can get like a quick interactive recap.

Absolutely. What's the three dots in the top right do when you click that?

Can you download file? Download file, save as an artifact.

So that's cool. Like with what you just said as a recap of something that might be important or whatever topic. Yeah. Either, I don't know what downloading it would do.

Download. So it's an HTML file. So if I download the HTML, can I just like open this? So this is the HTML file. Oh, I mean, it didn't style it at all, but it's got all the data in it and you can click and it has, yeah.

Okay.

It didn't come with any sort of like CSS, like styling, but it does the trick. Let's see if I click deep dive, the deep dive button doesn't do anything. Yeah. But all the buttons work.

That's pretty cool. All right. There's use cases here. I'm sure. Yeah. This is, I like what you said though. It was like, if you just copy and paste that prompt in the bottom of whatever conversation you're having that you feel like is important or you want to visualize. Obviously you could change diagram. You could change out experiment. It seemed like those words, yeah. Like the thing you want to create and what you want to do with it are flexible. Yeah.

So like we can do the same thing and say, um, I wish I can fork it from right here, but it looks like I have to come all the way down here. Um, Gemini lets you fork chats now. And I think ChatGPT now does too, doesn't it? I don't know. So if I go to like, yeah, branch into new chat. So Anthropic's fallen behind on the ability to fork your chats. But if I do that same prompt, take what you just explained about this topic and turn it into an interactive flowchart. We like that. That I can experiment with directly here in the chat. Sure. Let's hustle out a little flowchart here.

Let's do that hustling. I think that's good. All right. Give me my diagram.

Look at this. Now we got a flowchart getting built, dude, in real time.

Look at that, dude.

If, I mean, we're actually are recording our screen, but like, let's say you were actually like making a video about this topic. You can actually record this, like animating onto your screen and have like a, a very, very like basic kind of like After Effects looking animation.

True. Just slap on your Loom and get going, you know? Yep. If you want to look impressive to your boss, to your clients— dude, that is sick— start using this prompt that we just came up with and do this.

And by we, you mean Perplexity.

We modified it, but yes, attribution Perplexity and Claude. What do we got? This is crazy.

It's still working, mapping the plumbing. So But I mean, like the visual, like the flowchart, Anthropic, Palantir, AWS, classified network, mission-critical deployments, the dispute. There's like this whole circular dispute thing. The replacement race, OpenAI, Google Gemini, XAI, Grok. Is Gemini really in the talks?

I didn't know that.

I don't really know if they're in the talks. One thing about DeepMind is when Google bought DeepMind, part of the deal was that Google could never use it for like, I don't know if it was military or for war or something like that, but like it was specifically written into the contract when Google purchased DeepMind. Interesting. Okay. So I don't know. A lot of these companies originally wrote that kind of stuff into their contracts and have slowly been shaving them out.

I was going to say, is anything really concrete?

So they're clicking on any node to expand it. So Anthropic Claude models. Okay. So it's kind of the same thing where if I click on one of these, it changes way down here. So you gotta, gotta scroll back and forth. Oh, look at that. Even some of these are visual, right? I clicked on Google Gemini readiness, classified clearance, negotiating model capability, strong alternative. So that kind of answers our question, right? They're negotiating the classified clearance right now. XAI Grok, classified clearance, deal signed, model capability unproven in defense. I mean, Elon literally just came out this week and said like, our models aren't good enough. We're starting over from scratch. So is that really what you want your military using? But the deal's signed. OpenAI, classified clearance in progress, contract signed, $200 million, internal stability backlash and resignations. Wow. This is so sick, dude.

I mean, just think of getting knowledgeable on any topic, like rapid fire. You can just do this. Now you get the visuals, you see the connections, you go deeper where you want.

Okay. So I know we're getting deep here. Uh, so I have this little project I created called Richard Feynman and it basically explains anything like the famous physicist Richard Feynman would explain it, which is baking everything down to first principle knowledge, right? It's like explaining everything at its most simple rate. And so like one of the things, for example, I gave it was— I'm looking at your chats.

I'm like, what do you want to learn? First principles. What do you have here?

So one of the things I gave it is how computers work from circuit to code, right? So you've got— here we go. You know, you've got like your circuit boards, your CPUs, your GPUs. They've got transistors. Transistors could be either on or off. Right. Um, and basically when a CPU is doing something, it's just passing electricity through the transistors and it's sort of routing it based on which transistors are on and off, right? Every transistor on a CPU could only be one of two states, 1 or 0. Well, the layer above that is called binary code, 1s and 0s, right? So like binary code, if something is a 1, it essentially would tell like this transistor to be on. If it's a 0, it tells it to be off. And so the 1s and the 0s are constantly changing to tell the transistors what to do. And then you have machine language and then assembly language and then programming languages and the operating system and then the applications that you actually use. And there's this whole sort of like, they call it layers of abstraction between you actually typing something into computer and then the chips actually doing the processing, right? And so I had it explain that whole concept to me like Richard Feynman. And I'm like, can it make a really cool visualizer of that for us right now?

Absolutely. Let's go.

All right. Take what you just explained about this topic and turn it into an interactive. I want to do flowchart on this one that I can experiment with directly here in the chat. That's another keyword I think is directly here in the chat. Yeah. Right. So like interactive and directly here in the chat, we're telling it like, don't go write code and open it in the canvas on the side. Like do it right here in the chat.

And that's probably good for everything. Cause I was trying to code something just last night and they kept coding it in the middle and not giving me the canvas. So little things like that to say, get in the canvas, get in the chat.

Stacking transistors, wiring up complexity explosion, teaching switches to think big.

Wow. That's cool.

Dude, I'm going to be playing with this like crazy. Like the video that I made about this on my main channel, like really downplayed it. It undersold it. I need to make a follow-up and be like, dude, this is, way more legit than what I showed off.

You're welcome. I just prompted you via perplexity. What do we build in here? What do we got?

Let me build this out. Something you'd actually play with. I'm going to throw this prompt at the end of all of my little like Richard Feynman conversations. I love that project that you have created. Yeah. I did. I had another one try to explain quantum computing to me and I read it all and I'm like that. Still doesn't make sense to me. I mean, well, maybe with a visual explainer, it'll make more sense.

Well, this is kind of interesting because everything starts with text and then it seemed to go video or sorry, audio. So like read it to me kind of thing. So you have your auditory listener, you have your visual or your reader, whatever that one is. But now you have this, like the actual, like you want to see pictures, you know, you learn from diagrams, which now it just renders. So what do we got?

What do we got? So we've got our transistors as the bottom layer, logic gates, AND, OR, NOT circuits. Circuits, machine code, assembly language, programming languages, operating systems, applications. And then even on top of applications, we now have AI. So we're not even necessarily going to be needing to use applications in the future. The AI is going to be using the applications for us and we're just going to speak to them. Right. So it's showing all these layers of abstraction and it looks like I can click through and each layer. Is doing something like it's doing something on the layer below it, right? Yeah. So when you're, you know, when you're using AI, AI is talking to an application. When you're using applications, applications are communicating with the operating system. When you're using an operating system, the applications are, you know, using a programming language, both operating system and applications are using programming languages. Then it's using assembly language, which is a little bit harder for humans to read, but it's still slightly human reasonable. Then you have machine language, which is like more like binary CPU instructions, like ones and zeros and stuff. People can't really understand that. Then you have the logic gates, the AND, OR, basically telling the transistors whether to be opened or closed and that kind of stuff. And then you have the actual transistors, which are the little electrical circuits on a CPU board.

Dude, like this is a really cool flowchart. I like this one even more. Well, we got more.

So try skipping layers. I want to go from natural language directly to Machine code. I don't know what it's showing me here. These models for you to play, these modes for you to play with, explore layers, try skipping layers where AI reaches. Interesting. So like it's actually interactive around like, all right, I want to go from programming languages to transistors. I'm not quite sure what this is visualizing. Each skipped layer multiplies the decision space exponentially. All right, where AI reaches. So AI can reach the generating the applications a little bit, the operating system, it can do the programming language, like writing code. So this is actually saying like AI is going further and further down the stack, right? Like right now we're talking to AI and AI can actually help manage our applications for us and it can actually help write code for us, but it's not super great at managing the operating system yet, but we can see it's getting there. And it's not great at assembly language, machine code, logic gate, or transistors. So it needs these other layers on top of it is what it's basically saying.

Now, is this the code to AGI? Possibly. When all the bars fill up.

I think you're convinced that we're going to discover AGI ourself on this podcast.

I've heard someone say it's 2026. Why not us, man? Yeah. Make it happen.

I mean, if Nano Banana could inform us that the, Pentagon banned Anthropic, then we can discover anything on this podcast.

I think it's possible. I'm not downplaying us.

Anyway, you know, we're pretty deep down this rabbit hole. I'm excited about this stuff, as you can tell, but we should probably like move on to our second topic. We were going to make this a shorter episode.

Good luck.

How well do you know me by now, Joe? Have I ever made a short episode?

Too well, too well. Second thing we're going to talk about is another visual thing we teased.

I don't know, keep this one quicker.

5 hours ago. So Canva, you know, the thing that just basically took Photoshop and said you're free now and it's easy. I don't even use Photoshop anymore. I used to live and breathe Photoshop. I haven't had it on my computer for years. Yeah. Um, I don't know. Canva's pretty awesome and they just released this thing called Magic Layers. Mm-hmm. And I think there's some cool use cases here. I think we're gonna keep it pretty tight, but this seems like something like anybody could jump in 'cause you're probably using Canva or you've heard of it. It's free. And I think this is available to everyone if I'm not mistaken. Yeah.

Let's see here. So, uh, Canva introduces Magic Layers, new breakthrough of AI development, blah, blah, blah. It converts flat images and static AI outputs into fully editable multi-layered designs inside the Canva editor. So like, you know, obviously, you know, Photoshop, right? Well, one of the beauties of Photoshop is you've got all of these layers. You can have your background layer, you can have like a mid-ground layer with, you know, some trees or elements in the scene. And then you've got your foreground layer, which is maybe like where the person or the main object is. And you can have these multiple layers on, on your Photoshop. One sort of negative thing about using AI image generation is you get what you get, right? Like it generates the image. This is what it is. Now you can do some things like inpainting where you can, you know, sort of shade part of the image and say, edit just this. And there is some of that kind of technology, but it's like still like a reroll kind of thing. Like, all right, let's try again. Let's try again. Let's try again until it looks like what you want. Well, this actually will take any image that you generate with AI. I mean, any image could be just image you took with a camera as well. And it will actually break out the elements in that image into separate layers. And then you can drag them around and move them and change the, you know, which layers on top of the other layers and stuff like that. It's really crazy.

I had a lot of fun just messing around with this thing. Yeah. It's rolling out public beta across US, Canada, UK, Australia. Yeah.

Um, does it actually say for all plans? I thought I read it did, but it doesn't specifically say, but I, I think it is just available if you have an account. Anyway, let's try it.

Let's try it. So Matt, I think a cool little visual that we could do is, uh, you swimming in bananas down there still cracks me up from the other week.

Yeah, we made that on one of our previous videos.

We did. So like, why don't we take one of your thumbnails from YouTube and slap it in here? I don't know if you want to pull it from YouTube or maybe you have it in Canva already, whatever's easiest. You do have one in Canva.

These are like old, these are sort of my janky thumbnails. They're not ones I actually had my thumbnail designer do. But I, so let me just open up like a, well, let me create a new little document here inside of Canva. Here's a thumbnail for a video that's live today, right? This is my most recent video. Look at that. So this is, this is not all necessarily AI. This was actually a real photo of me that I think was touched up a little with AI. My skin's not that shiny. But I can come up to edit here and there's a new magic layers button over here. And if I click that, it will scan the image and ideally it will break out everything from this. What I don't know if it's going to do, is it going to fill in like the rest of the letters here?

Like, is it going to know that that's supposed to be an E? We'll see. I have not tested this. But I think use case for thumbnails are really cool. I mean, for one, you can just pull from anything you have. Maybe you've used other people's. So look at this. It broke everything out. Oh, messed up.

So it messed up some of these logos here. Actually, I don't know why it did that, but it's actually not bad. It just sort of duplicated some of the words. Oh, there you go. Okay. All right. So like, I just, I could remove this background now. Yeah. Or I can go and, you know, set it as the full screen background. Here's my face. I can move this however I want. I could put me over on the side here instead. Right. Oops. I grabbed the wrong little slider here, but I can make myself big and off to the side and I can move these logos around and, you know, put them stacked like this. Hey.

Oh, dude, look at the text that was behind your head. It got filled in.

Oh yeah, I did figure out that I said AI News behind my head. That's sweet.

There we go. But I could look at this.

It broke out all of these logos separately. Even these little dots. I don't want them there. Let's get rid of them. They're all broken out. All right. So now I can recompose a whole new thumbnail. That's kind of the same idea. Right. Let's go like this, make this really big. So it kind of goes behind my head again. And then no, I want to grab this here and then I can just, you know, make a brand new version of this thumbnail that looks way worse than the original one.

A/B testing though. I mean, like it's, it's an easy way to at least grab something.

I don't know.

I'm like, I'm just thinking like, you know, what was it? Great artists steal. Is that one of the? Sayings. It's like when I'm working— I'm sure some people say that.

Not me. I never steal. You never steal.

No. You're a saint.

So I can make all these.

Let's go to someone else's channel. Maybe like bigger. Yeah. See what they're doing.

And then, you know, make them overlapping each other. Like they're all kind of stacked everywhere. And look at that. I got, I got a nice little thumbnail here.

There you go. Maybe swap out the background for something fresh.

Fresher, a little freshy fresh. AI new. All right, so I can get rid of the background. Let's go ahead and do like, I actually always use this gradient on my backgrounds. I use it so much.

Those are the ones you make, huh?

Okay. Yeah. Whenever you see that gradient, you know, I slapped it together in Canva.

Was that for like an A/B test for thumbnails?

Sometimes, or sometimes I just, have to get a video out quick and I don't have a thumbnail designer available to me.

There you go. What would that take?

Like 5 minutes? Let's add like a little background drop shadow on this here. Honestly, it's not that bad of a thumbnail.

I mean, like it gets your mind thinking like, what is that AI Nuevo News? Yeah. Yeah. Is it AI News? Just got to watch the video. That's how I get mine.

You got to, you got to look. Click the video to find out.

It's all the packaging.

I could shrink my head to make it more obvious.

There you go.

There you go. There. Now it's clear.

That's a pretty solid A/B test or even go-to-market thumbnail. Didn't take much.

It did not take much at all. So that is sweet.

I like it. And again, like, I don't know, this is why I don't download Photoshop anymore. Sorry, Photoshop, if you're watching. Used to love you. Gave you a lot of money.

Today's episode sponsored by Adobe.

No, I'm just kidding. They're on screen, so they're getting some shout out.

No, that's Canva. Oh wait, they are on screen.

You're right.

So, um, this is probably a good time to mention I do own equity in Canva. Probably should have said that. I don't know if I legally have to, but it seems like the right thing to do. That's good. That's good.

But all right. So another use case I think could be really cool. Do you have a. Nano Banana or Gemini generated infographic that is?

I don't, but we can go create one. Let's go ahead and go to Gemini. What shall we generate? Create an infographic that explains layers of abstraction. For a desktop computer. Do a little callback.

There it is. Layers of abstraction. I learned something new today. Thank you, Matt.

I would hope so.

There it is. Oh, check that beautiful infographic. That is quite cool. Yeah. So you're downloading it.

It's like a food pyramid, but for computers.

It just got updated again, I guess. Now with this, so we're gonna take this file we're downloading, which normally you would not be able to edit, or I mean you can, I guess, but you'd have to do it through text in Gemini, but it's not always perfect either. But now you're gonna slap that thing into Canva and run the magic, the layers on it.

Yeah. So here it is in Canva and there's, there's a lot going on. So it'll be interesting to see how well it does this.

I was doing this test last night and I'm like, I want to talk about this. So ideally it should have all the icons broken out, the different layers. And then you can also edit all the text you see on here, which can be a common issue when you're generating a fricking, you know, infographic that has a lot of stuff on it.

I mean, sometimes it even like gets typos and stuff too. I mean, it seems to have gotten a lot better lately about not doing typos, but they do happen still sometimes.

Let's see. Let's see. All right.

So there we go. So every single thing looks like it kind of grouped some of them, but that's not bad.

You could work with that. Something got cut off.

All of this text is actually editable now.

Yeah, there we go. It worked.

That is so sick.

So like, again, you could probably edit some of the text in Gemini, but you'd have to like keep prompting it back and forth. You don't have full flexibility. If you wanted to like bold something or move it around.

You can even edit this text that's sideways here. Not bad, huh? Not bad. Not bad. I knew I invested in him for a reason.

He just wanted you bad.

Bad. But yeah, I mean, like literally everything here is broken out. Like even these little like like each level of the pyramid, the background colors are broken out into their own little thing.

It looks good. I think it might've changed it a little bit. What? Cause like part of that pyramid is kind of missing down there, right? The orange or is that how it was originally?

Yeah. So it is missing some stuff. So this whole like blue background disappeared for some reason. Yeah. On all of them actually. Look. So yeah, the pyramid background seems to like, it just kind of made it more solid colors in the background. Yeah. Instead of adding like all these little lines and stuff. That's, that's interesting.

Oh, if you click on it, they're actually there. See that? Oh yeah. It's a coloring thing. Well, point is we get the point.

Hold on. Let me, let me try this one more time. I want to, I just want to test it again and see if it's like consistently does that same issue. 'Cause then we'll know, like if it does that, just delete it and try again.

You know? I mean, you got to do a nano banana a little bit too. Sometimes. Yeah. Did it again.

So that's kind of a bummer.

So maybe gradients and super complex things it has, it kind of struggles with.

Yeah. Yeah. But I mean, you still have like the bones of a really solid infographic. It did. It did, you know, jack some of this background up, but yeah. Huh.

Something to play with. I know I was trying it with like when it has a solid background, maybe simpler isolated things like text, icon, you know, shape. Yeah, yeah.

Works a little better. This background here, it seems to struggle with.

Which I guess you could just prompt in Gemini to simplify. The look of it a little bit or maybe remove the background. Yeah. Yeah. Huh.

Still pretty cool. I mean, it's still pretty cool. Like the other thing that you can do is you can actually bring like real live images in. Right. So like, um, here's, uh, here's an image that I was using for a different thumbnail because it was about what was going on with the Pentagon. Right. but you can actually edit this kind of thing as well, even though it's like an actual photo and not AI generated.

Uh-huh. Let's see a couple things. I was just looking up just to list out some use cases that could be helpful. So like infographics, different promo graphics that might be legacy or like old school and they just don't have the layers to anymore. This could make them now something you could do, or like JPEGs, it's just like flat styles. Oh yeah, check this out. Can you actually drag him around with it? Ah, not bad. Yes, I could rotate him.

Look at him, he's upside down. He's on the ceiling. And this guy, this guy, very serious.

Oh, if I put him behind the computer, he's, he's just like, what are you doing?

Get off that ceiling. And then you got this guy, whatever his name is. What's his name? Let's make him, let's make him like peeking through the curtain and I'll just go like, behind someone else. Hey, peeking through. What's going on, guys? Look, come right here.

Can you pick up the water bottle and do something with that?

Water bottle? Yeah, get that out of here. I don't know.

You need to drink it upside down somehow. I don't know. You got to rotate it.

Drink. It's really big. Yum, yum, yum, yum, yum, yum.

What are we doing?

This is what AI is for, Joe. This guy. This guy. What are you doing over here, Donald? This guy's peeking over his shoulder. There. Hello.

This is how you can really recreate any image out there.

This is what was really going on in that room.

People were concerned about AI slop. Now we can really have fun.

This is AI amazingness. Amazingness. Make it peek his head through the curtain there.

So other things he can do.

You don't want to keep on playing with this. I love play.

Let's keep going.

How far do you want to get? I'm kidding. You know what I could see doing? Like if I was doing an explainer video. Yeah. Like, all right. So Pete Hedgesteth. He came up and he was talking to Donald Trump and he went, hey Donald, what are you doing over there? And like, you just make a whole explainer video, but you do it like animating it in Canva like this. Like, hey, um, what are you doing? Come over here. Wait, I need to pull him forward. Like make him to front. All right. Donald, you got something in your nose right here. Let me scratch that for you. Scratchy, scratchy, scratchy, scratchy. Helping you out.

You're on camera. Got a bad game.

And then this guy, let's, I need to mirror him. So let's flip him horizontally. That way he's like giving these guys a dirty eye. Like, why is your finger in his nose?

What are we going to do with the bottle though? The water bottle needs to play somehow. Maybe dumping what? Who's in the background there? There. Really big water bottle. Oh, balanced on his head. Okay. That's good. Donald's still peeking.

Donald's peeking through and getting his nose picked. And the other guy's giving him a dirty eye while balancing a water bottle on his head.

And this guy's like, you guys, there's a bunch of tomfoolery happening in this room.

This is a bunch of tomfoolery. And then this guy's just like, I'll just hang out right here.

Why are they so close to each other?

I don't know.

And then we have, I'm just looking at images on the left-hand side. Now we have a Chinese president over there that we can— we doing this?

We doing this?

I mean, we're already down the rabbit hole.

And then let's go ahead and put this guy on the screen because this is who they're talking about. This is what this whole conference is about is this guy who's on camera right now. That's it.

Wow. We made a whole new storyline. Yep. Matt, are you going to put yourself in there? I just feel like it's inevitable. They're learning. They're actually watching our podcast right now. They're like, hey guys, just me. Don't worry about me.

I'm just, I'm just in the current really confused. Like what is happening over here? What are you guys doing in here?

It looks really important.

All right. There is our thumbnail. There's the thumbnail for this episode. I feel really bad for whoever's editing this. Hopefully this part doesn't get edited out because that's like the most fun we've had. This is what happens when Darren isn't here to babysit us.

Sorry here, Darren. Thank you. This could be the biggest hit of the podcast right here. All right. All right.

Let's move on. We get the idea, but yeah, you were listing off some ideas you can use this for. Infographics, YouTube thumbnails, Facebook ads, Google Display Network ads.

Yeah. Old images of yours that are laying around. Take a picture of something maybe on your desk and reconstructing history to make it more fun for you. Creating explainer videos if you wanted to do it that way.

Yeah. I actually, that would be really kind of funny to just be like, Hey, you over there. And the guy's like, you're talking to me? Yeah, I'm talking to you.

Seriously. And then some guy over here who's like, I'm in China. Is that his voice?

I don't know.

If you were really good at storytelling though, I bet that is a very low budget slash no budget YouTube video that someone would, a lot of people would watch. If you, yeah, you could pull this off. Someone please do this.

Please do it. I'm gonna tell a story this way. I will make a video like this on my channel at one point and tell the story of— I don't know what story I'm gonna tell yet.

There you go, Matt. You're full of video ideas now at this point. Don't ever tell me that you're not.

All you gotta do is flip on the camera and give me some tools and I will figure out how to screw around with them. That is what I do. All right. I'm gonna close out of that.

Moving on. So we're at story phase 2 of 5 or whatever of this podcast.

No, we're moving on to 3 here. This is number 3.

We're moving on to 3. Okay, so next topic at hand here, we're going to bring in a couple of intelligent folks, Sam Altman and Jensen Huang. So these guys are saying different things, but I feel like in a similar way where intelligence is kind of a commodity.

Wow. I can't say commodity of sorts.

It's a utility in the way that Sam is talking about it here. There was something that he said on stage. We can just play the video. I think it's interesting. It's a talking point. Let's, let's play this.

Yeah, it's only a minute and a half and we can even cut this down a little bit.

Fundamentally, our business, and I think the business of every other model provider, is going to look like selling tokens. You know, they may come from bigger or smaller models, which makes them more or less expensive. They may use more or less reasoning, which also makes them more or less expensive. They may be running all the time in the background trying to help you out. They may run only when you need them if you want to pay less. They may work super hard, you know, spend tens of millions, hundreds of millions, someday billions of dollars on a single problem. Right. That's really valuable. But We see a future where intelligence is a utility like electricity or water, and people buy it from us on a meter and use it for whatever they want to use it for. The demand that we see for that seems like it's going to continue to just go like this. And if we don't have enough, we either can't sell it, or the price gets really high and it, you know, kind of goes to rich people, or society makes a bunch of sort of central planning decisions that I think almost always go badly about, you know, we're going to use our limited compute supply for this and not that. So the best thing to me throughout all the history of capitalism, innovation, whatever you want, is to just flood the market. Yeah.

Interesting, huh?

It is interesting. So I, you know, I have a few things about that. So the way he described it as like, we sell tokens on a meter, they're doing that. That's what an API is. Like, that's literally what an API is, right? Like, yeah, that's already how they're doing it. I think most people see OpenAI as ChatGPT, right? Like I pay $20 a month and I get access to AI whenever I want it. And that's how I think like most people are currently using it. The way he's sort of explaining it is it almost sounds like in the future he sees a world where you're not just paying $20 a month and getting access to as much AI as you want, like you kind of are now. He sees it as like everybody's going to be using the sort of API model where when you want to use API or when you want to use AI, you're on the meter. It's going to start cranking up that meter and you're going to start spending per token. Like those of us who do development, right? Those of us who use things like Claude Code and Cursor and OpenAI's Codex and stuff, we're already used to just being on a meter, right? We're used to going and having it write scripts for us and then going and checking our API dashboard and go, oh crap, how much did we spend today? Right? Like that's already the reality to developers who are using AI. Where it gets really interesting is when that reality, that sort of, way of charging people transitions into like the normal users of AI, right? He's saying this is going to be like a utility, like the way you pay your electric bill, your water bill, your gas bill, that kind of stuff. Well, you're going to have an intelligence bill where you just use as much intelligence as you need and we bill you at the end of the month for the amount that you used. I think that's kind of a scary reality for a lot of people, but I also think it's a very it's not the smartest approach to claim that that's what you're going to do. Because if everybody believes that's what you're going to do, you're just going to incentivize everybody to go, okay, well, how do I not use a cloud provider to get my AI, right? Like we've got things like the DGX Spark, we've got Mac Studios that have insane, uh, like memory on them that could run models locally at home. So if we've got companies out there like OpenAI saying, we're going to put you all on a meter and we're going to make it so like when you want to use our intelligence, it's just going to, you're just going to rack up the meter. Well, that just makes me think, okay, I'm going to just get my own on-device AI and hopefully that gets better and better and better. So I don't need to tap into your meter. Like the analogy that I used when I shared this post on Facebook is like, if you look at OpenAI like the next electric company, well, getting your own on-device AI is like going and buying solar panels. So you're not relying on the electric company anymore, right? Except the big difference is when it comes to electric companies, most of us are in like monopolistic societies where we don't have a choice of our electric company. And even if we have solar panels, like the electric company could still jack up our price and jack up our usage to the point where we're using more than our solar panels generate an AI, if I have on-prem compute where I can just run my AI models whenever I want, I can just cut them out completely. Mm-hmm.

So basically on-prem is the equivalent to, uh, off-grid essentially, like for electricity. Like if we were kind of comparing it like that.

Yeah, yeah, exactly. Exactly. Yeah. It would be like, okay, I don't need to tap into your meter anymore. I'm just gonna put my own box in my corner and I just pay for the electricity to run that box and that's it. I mean, I'm on a meter. I'm just not on your meter. I'm on the electrical meter instead.

Bingo. Yeah. I mean, like, like you said, it's already happening, but there are so many different choices. Like I forget the app that came out last week, but you can run it just on your iPhone.

Yeah. Yeah. Locally is what it's called. Locally. Yes. Yeah. Yeah. Yeah. You can do that. Like, and I mean, I'm, I've got a flight day after tomorrow on Sunday. I'm going up to the GTC event. And I'll probably be sitting there on my phone having chats with AI up 30,000 feet in the sky because I could do that now, not even need internet. So I don't know. I think I understand what Sam's saying, right? Like, I think this model of you pay $20 a month and you can use as much AI as you want, at the moment it's sort of unsustainable because OpenAI loses money every single month, as we know. But at the same time, the cost of compute is going to come down over time. These models are going to get more efficient over time. And so theoretically, you're onboarding a whole bunch of people at a loss right now, but in the future, with economies of scale and with cost of compute coming down and with Moore's law, meaning we can fit more transistors into the computer, whatever. I think I'm like really oversimplifying this stuff, but with all of that being said, the cost to use compute is going to come down over time, theoretically. And I don't understand why people would want to send stuff to a cloud if eventually we get to a point where the models we run on our own devices are just as good. And with Sam going out there and saying stuff like what he just said in that clip, that makes me think, more and more people are just going to open their eyes to like, if I can do this at home, why wouldn't I?

Bingo. And I think that's the— that was my takeaway is like, okay, well, he's just— he put out there where he sees the future going for OpenAI and maybe Anthropic or, you know, Gemini as well. But at the same time, I think it's an education phase for a lot of people to understand, oh, these are things that he basically wants to have you rent from him or send him money based on a meter, you know, and it's gonna tap into all the tools or devices that you might use throughout your life, no matter where you're at, you know, that meter is somehow ticking in the background.

Yeah. Yeah. And, and I think he's operating on this assumption that OpenAI will always, always be at the very frontier. They're always gonna be the most state-of-the-art. They're always gonna have the best, most intelligent, smartest models out there. They're gonna have AGI eventually, right? Like I think he's operating on this assumption that, you know, you can run AI at home, but we're always gonna have the best AI. And if you want to tap into the best smartest, like the front of the line, like edge of AI stuff, you're going to need to use companies like us. But part of the problem with that is they're not a monopoly. We have Google, right? Google's giving out AI for free. Free right now because they can subsidize it with their ad business. You have Anthropic, which last week we talked about how Anthropic is actually taking over the business market right now and selling more to businesses lately than even OpenAI is, right? Developers are already all using Anthropic over OpenAI mostly. Maybe eventually Grok is going to come along and get their act together and catch up with them. You know, I wouldn't put it past Elon to make that happen. He can send people into space. I'm sure he can figure this out too at some point, right? You have Meta who went and spent billions of dollars putting together their super intelligence team. So it's not like OpenAI is our only option. We're in a capitalistic society where, uh, if, if there's other options, it's going to continually drive prices down and companies like Meta are actively trying to create front state-of-the-art frontier models that are as good as OpenAI's models. And if they manage to achieve that, theoretically we could just run 'em at home on our own boxes without needing a company like OpenAI. Bingo. Yeah.

I feel like, uh, business-wise, it doesn't seem like the smartest thing to hinge everything on, you know, for Sam and OpenAI, knowing of exactly the landscape you just described and what's possible now with all the, the models that you could just run locally.

So yeah, I mean, it may be, it may be how it actually plays out. I just don't think it's the smartest move to be going and selling that narrative yet. Yeah. Yeah. Yeah. From a marketing standpoint, I don't think it's the optics. Like, I don't think it's the way you should be presenting what you're selling. Yeah. Yeah.

Well, on that front, like, so I want to turn it over to Jensen Huang from NVIDIA because they're saying different things, but there is a similarity in the whole thing of intelligence being this kind of commodity or this utility that everybody has access to. So let's, let's listen to this.

Who's the smartest person you've ever met?

Who's the smartest person I've ever met? I can't answer that question. And I know, I know what people are thinking. The definition of smart is somebody who's intelligent, solve problems, technical. But I find that that's a commodity, and we're not— we're about to prove that artificial intelligence is able to handle that part easiest, right? Yeah. And so, so as it turns out, um, let me give you another example. Everybody thought software programming is the ultimate smart profession. Look what is the first thing that AI is solving? Software programming. And so it turns out that the definition of smart is very different than most people think. I think long-term, the definition of smart— and my, my personal definition of smart, uh, is, is, um, someone who sits at that intersection of of being technically astute, but human empathy and having the ability to infer the unspoken, the around the corners, the unknowables. You know, people who are able to see around corners are truly, truly smart. And And that their value is incredible to be able to preempt problems before they show up, just because you feel the vibe. And the vibe came from a combination of data analysis, first principle, life experience, wisdom, sensing other people. Uh, that vibe, that I think that's smart. That I think is going to be the future definition of smart. And that person might actually score horribly on the SAT.

I like, uh, Jensen's framing a lot better than Sam's. Oh yeah, for sure.

A couple takeaways I'm just going to say here is from my side, because when I watch that, you reposted this here the other day. I was like, interesting. Yeah. Different, similar things they're saying overall being that like this commodity intelligent AI essentially is the smartest thing around. But like what we've said a lot on previous episodes now is that it's the human taste or like what he accurately described as the vibe. And I like how he said that, but he said basically it's like, yeah, that's knowing the data, the analysis, first principles. We talked about that earlier too. But life experience, empathy, it's like the unknowables, the, the unspoken things that us humans kind of, I don't know, it's like what we do on this podcast, I feel like.

Yeah. I mean, I feel like you can boil everything down. Everything he just said, I think boils down to like IQ is less important. What's important in the future is EQ and and insight, right? Maybe, maybe not insight. Shoot, what's the word I'm looking for? Intuition. Intuition. I think basically if you boil everything that he just said down into like two words, it's EQ and intuition, right? Like IQ is no longer going to be the metric that people see as smart because knowledge work is the first thing that is essentially getting like their moat is getting filled in with dirt by AI, right? Like, like coding work is getting filled in with AI. You and I have had conversations off recording about how the stuff that I make on YouTube, tutorial type content, breaking down the news, stuff like that, that's way less valuable in a world where people can just get that from AI real quick, right? Like knowledge work, coding, like even, even stuff like, like lawyers, like I don't think lawyers are going away, but I think people like paralegals might be in trouble. The ones that actually go and do the research and do the fact-finding and sort of try to connect all the dots, right? Like I still think there'll be like lawyers in court pleading their cases to judges and stuff, but the sort of downstream work from lawyers kind of goes away. When it comes to like doctors and the medical industry, I still think people are going to want opinions from actual human doctors, but a lot of the downstream work from doctors, a lot of the research, the helping with diagnoses, right? The taking an X-ray or taking an MRI and looking at it and going, what does this all mean? Well, technology is probably going to get better than most doctors at analyzing that stuff and finding little like nuanced things in that scan that maybe the human eye can't even see, right? So I think a lot of the, like, the knowledge type work, the IQ type work is, I don't want to say it's a commodity because like there's not a lot of people on this planet that have like extremely high IQs, but I think it's a lot less valuable in the future to have extremely high IQs. What's valuable in the future is knowing how to interact with other human beings, like having that EQ, knowing how to like make friends and persuade and, you know, get what you want out of life, essentially, that EQ thing. And then also that intuition of like, I can kind of feel this thing bubbling up. I can kind of like, I would say back in 2021, when we were talking about AI and how AI was starting to get used for copywriting and marketing, and we were talking about how in the future, like sales pages and advertising are probably going to be run by AI. You and I were literally on podcast talking about that in 2021. That is intuition, right? What did I do with that information? I went and created a whole freaking career out of that intuition that this is going to be the next thing, right? So I really think that like, if you boil that whole 2-minute like little thing down from him, it's IQ is less and less important in the future. EQ and intuition are what you should be really concerned with because if you can see where the puck is going, so to speak, and you can skate there before everybody else, you're going to be the one that wins.

Be like Neo from The Matrix. I know kung fu, but you've just tapped in. You're all tapped into AI and intelligence we have access to, but, and I don't know if Neo is the best on the personality side or human warmth and this, but maybe intuition. Um, but yeah, like that's our superpower and you're right. Like we were talking about this years ago and yeah, you went more YouTube. I went more working with business owners and all that stuff and, and education. Like that's what I feel like what we're doing as well here. You've done through YouTube and content and I've worked with a lot of the people doing that kind of thing. And it's, it's pretty interesting when people understand that we all have access to the same information. Now. It's basically free for everybody in some form. It's just how do we apply it throughout the humanness.

So just from like a wealth-building standpoint, right, from the standpoint of your ability to generate income for yourself, I know a lot of really, really, really insanely smart people that maybe don't have the highest income. I mean, when you think of like really intelligent people, You might think of like teachers, like, right? Teachers are, you know, pretty stereotypically underpaid, right? You know, scientists, a lot of scientists that they rely on funding from like government grants and stuff like that. A lot of scientists don't make a lot of money. Like they're, they have to, they have to rely on people giving them money for them to continue their research. And then you, you look at the people that actually have some of the most wealth. Right? Who do you think of? You think of like investors. You think of like the Warren Buffetts of the world. You think of, um, you know, Bill Gates, Jeff Bezos, Elon Musk. A lot of these people, I mean, some of them probably have really high IQs as well, but the thing that made them the money, Jensen Huang included, the thing that made them the money was the intuition, not the IQ. Jensen Huang had the intuition that in the early days, gaming was going to be the next big thing. Let's build the dominant software for that. And then a little bit later down the line, AI is going to be the next big thing. Let's build the dominant hardware for that. I think I said software first, but hardware. Let's build the dominant hardware across the stack for all of this kind of stuff. He had that intuition that this is what's coming down the pipeline. Let's build for that. You look at, you know, Sequoia Capital, A16Z, Sound Ventures, like all of these companies that invested in AI before the sort of like trajectory blew up and it was all mainstream. They all had intuition of like, This is what's coming around the corner. Let's get on board with that. Right? When you think of like the multi-decade billionaires and, you know, and, or decade millionaires and billionaires and people like that, they made their money because of intuition, not because of IQ.

I'll add in that. Absolutely. And I feel like there's a time horizon too. So like Jensen Huang, I looked it up. I knew it was over 30 years, but yeah, 33 years as the Nvidia CEO. So you got to think about his intuition was like, it wasn't perfect from day one, but I think you follow the course with your intuition, with the knowledge that you have and your experience and all that stuff combined. And like, look at Elon, look at all the other ones out there that have been kind of navigating for a long time. So this time horizon is where it starts to develop and create something bigger. And I feel like right now we're in this like pivotal moment with AI being so it's new but developing really quickly that we all have the ability to learn quick right now. So keep watching, go watch Matt's YouTube and follow us everywhere. And, but at the same time, go apply it and, and, you know, start to tap into the intelligence, but also understand what is your human intuition side. Yeah.

Well, and I think the people that are best at it too are the people who don't necessarily look at what people say they want, right? Like the, what the, like Henry Ford, right? Like, doesn't he? I don't know the exact quote, but he has some sort of famous quote where he's like, if I gave everybody with what they want, I would have made faster horses or something like that. Right. Um, you look at Steve Jobs, right? Nobody was asking for like a new smartphone, right? Everybody thought the BlackBerry was like the ultimate smartphone. What do we need something else? Like these people figured out what people are going to want in the future before they knew they wanted it. Right. Um, so to me, that's kind of what Jensen's saying, like the intuition, the like knowing what's coming and trying to get ahead of that. Is the thing that's going to be most valuable in the future. And the people that we're going to look back at and go, holy crap, they were some of the smartest people, were the ones that predicted the future better than anybody else.

There you go. That's it. All right. So that's, I feel like our take of the everyday AI or like whatever that section we want to call it, kind of work in themes in this podcast, trying to work it out. Let's, let's, uh, transition to a new topic, which is more of the more weirder stories of the week, I guess we'll call it. Is Meta. We talked about Meta already and how they acquired Maltbook, which is kind of freaky if you think about it, but Maltbook being this social media network of AI agents. So not people, it's automated agents autonomously talking and having a life there, a life quote. And the fact now is that Meta hired, well, I guess the duo, the founders behind Maltbook, Did they actually acquire Maltbook as well? I believe so.

Yeah. Um, Meta has acquired Maltbook, a viral social media network designed for AI agents. Uh, Matt Schlitt and Ben Parr, which I know both of them, they're cool dudes. Um, but, uh, yeah, so Maltbook, just to kind of add some additional context, back when OpenClaw got really, really popular, um, OpenClaw being like the AI agent that everybody was buying Mac minis for and like running locally on their computer. This company, so it was originally called ClaudeBot. Anthropic said, you can't call it that. It sounds too much like Claude. So they rebranded to Maltbook, right? Or they rebranded to OpenClaw, right? No, so they started as ClaudeBot and then they rebranded to Maltbot. All right, so they started as ClaudeBot, then they rebranded to Maltbot. They were only called Maltbot for like, 2 days or something like that, right? Like it was a very short window of time, like 36, 48 hours. They were called MoltBot and MoltBook spawned up in that like 24, 48 hour window where they were called MoltBot. They obviously thought, okay, well, let's make our AI agent for, um, like for these, these MoltBots, right? So they spun it up during that window that it was called MoltBot and then it got rebranded to OpenClaw and they were like, well, we're already called MoltBook. We're keeping it. Um, So that's kind of how Maltbook got its name, but Maltbook is like essentially Reddit for AI agents. So AI agents can go and post like deep thoughts from the AI agent on this thing. And then other agents could go and reply and comment underneath it. And there was all this like crazy controversy because it was very insecure. A whole bunch of people's like passwords and APIs got leaked from it. A whole bunch of crypto scams popped up on it. It, like people were actually using the API and pretending to be bots, but they were, you know, typing the stuff themselves and making it look like a bot posted it through connecting via APIs and stuff like that. And it was just kind of like this, this giant mess. But it got really, really popular really quick. Over 2 million agents signed up for it within like a week. And even Andrej Karpathy, who's like a really big thought leader in the AI space, he actually helped invent a lot of Tesla's autonomous driving technology. And then moved to OpenAI and was one of the like head engineers over at OpenAI and just like a prolific, you know, engineer in, in the space. He actually commented about how like, this is like this crazy huge sci-fi moment that we're living in that AI agents have their own social media that are going and talking to each other. And there was like this moment where Maltbook was just like everywhere, everybody was talking about it and it kind of got crazy, but then it sort of died off for like, I don't know, a few weeks and nobody was really talking about it anymore. I, I kind of assumed like, ah, that site's dead. Dead. Nobody's really doing it. And then out of nowhere, Meta acquires it. And I'm, and I see this news and I'm like, what? Like people are even still talking about this thing? Like what, what the hell?

They angled it perfectly, Matt. They had the, uh, Molt, but obviously said they got the attention of the open cloth people and they had to put book on there. It was very strategic. It's genius. And you got Facebook, Meta, looking at it. Um, I don't know. Did they disclose how much they've—

No, it hasn't been disclosed how much, um, they got bought for. But the other interesting thing about this is the guys that made it flat out said, we vibe coded this. Like we didn't write a single line of code on this thing. Right. So Meta went and acquired a vibe coded company, but not only that, brought in the founders who are vibe coders, not necessarily like, you know, they're not like Alexander Wang, like billionaire guys who started Scale AI. They're like dudes like us who use Cursor, you know, or Cloud Code or something and vibecoded an app, put it online, and it just went viral. And now all of a sudden Meta's acquiring them.

Dude, yeah, I'm looking at the background of both these guys. So Matt Schlilt is a serial entrepreneur and product builder focused on chatbots, social, and commerce. So, um, Yeah, a little more like commerce. And the other one, Ben Parr, media guy, investor obsessed with the science of capturing attention. They're like marketing dudes, marketing guys. Well, and that's, I mean, that's kind of where my mind was going. I'm like, okay, well, they got the attention, they almost newsjacked and they vibe coded at the perfect time to capture a bunch of people who were just then getting into AI agents and setting up Molt, whatever they were called at that second, because it literally was seconds, but Open Claw, we'll just say, and, you know, got a bunch of traction. How many people? You said 2 million signups?

Like, yeah, but they're agents, right? So like, well, yeah, it could have been like one person had 100 agents all signed up too, you know, like it, it's like, I wouldn't say 2 million people's agents signed up because one person could have had like up you know, individuals could have had bulk agents signing up as well. For sure. Yeah.

But it's so interesting because so they capitalized on that timing, it seems like. I mean, I don't know the intention of why they made it, but sure as heck got a lot of people to get in there. I know a lot of humans were lurking and watching what the AI agents were saying and who knows how they were being prompted if that was the case or whatever, who knows. But I don't know, I start to think of like, okay, so when Meta rolls this out, What does that mean? That, um, now every business or person, if you want, is going to have an agent that can now be your representative on Meta's platforms, potentially. I'm not saying it's a good thing or bad thing. I don't know. But you know, like now they can represent you, your brand potentially.

Well, I don't know what that, to me, that's not really what Maltbook would do for Meta because Meta already acquired Manus. Right. And Manus is almost like an OpenClaw competitor. It's an autonomous agent where you can, you give it a goal and it will go and create a task list and then one by one work through the tasks until the goal is met. They already acquired an agent company. They already acquired their version of an OpenClaw, their version of a Perplexity computer. That's Manus. Yeah. So that's one thing that makes the Maltbook acquisition even slightly stranger. I mean, I could put on my tinfoil hat and share a few reasons.

I think, throw it on for a few minutes. I know the people like it.

I'm one of the people. So one thing, like I've talked about this in the past. I don't know if I've talked about it with you on the show, but I'm pretty sure you and I have talked about this offline before in that I think Meta wants to eventually cut the creators out of the loop. So right now, these platforms, they need to attract viewers. To attract viewers, they need to attract creators. They need to attract people to actually post on these platforms. Because if nobody's posting on these platforms, then why is anybody going and reading them and seeing the ads and Facebook makes their money, right? So they, they need to attract creators and then having good creators then attracts eyeballs, which they can then put ads in front of. Well, I think AI's, I think Meta's long-term strategy is if we can create a platform that autonomously posts stuff to it that people are interested in, we can cut creators out of the loop. We don't have to require creators to come and want to create on this platform. We'll have agents create on this platform. We'll have AI models create on this platform. They already rolled out and launched a product called like MetaVibes or something like that, which is like a TikTok type thing that you scroll through, but 100% of the content is AI generated. There's no human generated content on it. It's purely AI slop. They literally made like a TikTok of AI slop called MetaVibes. That's great for the brain, right?

That's great for the world.

Yeah, it's gotta be amazing for the brain. But so they've already done that, right? So they've got a platform where you can go and just, like mindlessly scroll through like AI-generated stuff, right? And so if like the Maltbook shows that people that are creating these AI agents, they can have their agents go and post autonomously and actually create content that is fairly interesting. I mean, a lot of the content that was created on Maltbook was getting shared around. People are going, look at what this agent wrote. Isn't this crazy? Isn't this scary to think that an AI is thinking this way? That's wild, right? Well, that's content that if Meta can get that kind of content posting on their site, that keeps eyeballs on their site and allows Meta to put ads in front of those specific eyeballs, right? So I think that Meta over time is trying to figure out how to cut creators out of the loop and just have a platform where the content that goes on the platform is created autonomously, but it still attracts eyeballs so they don't have to pay creators. They don't have to attract creators. They just need to get eyeballs looking at their slop and they can put ads in front of it. That's one, one theory. I have another.

I'm just going to like my, my, it's the, like the infinite slop machine. Slop. You know, like that freaks me out, dude, but I can totally— Slop Joe, slop Joe. You're making me hungry. All right. Number 2. What do you got?

What do you got? So the other thing is, as more and more people start using agents, as people start using Manus, as people start using OpenClaw and Perplexity Computer and things like that, and they start to get comfortable with it going off and doing things autonomously for them, well, eventually it's going to get to a point where they're going to start to feel more comfortable with it actually spending money on their behalf, right? Like we've already seen things that are like bots of like, here, help me go book my dream vacation or whatever. And it goes and finds the flights and finds the hotels. Finds the things to do while you're on vacation, and maybe it gets you all the way there other than getting past the checkout and actually spending money. It's only a matter of time. Like in the early days of the internet, nobody wanted to put their credit card details online. Now we have our credit card details just like saved on everybody's website so we can one-click buy shit, right? So like, I think it's only a matter of time before people get more and more and more comfortable letting their agents go and buy stuff on their behalf, right? I bust out my phone and go, you know, hey, hey, Barry, I'm not going to say the actual word. Hey, Barry, I need a new mouse. My mouse just broke. And it's like, okay, I'll do the shopping. I will see what has the best reviews. And knowing the type of work you do, I'll figure out the best mouse for you. Can I go and purchase it for you? And, you know, you say, yeah, you know, do what you need to do. So you talk into your phone just like you would your little phone assistant, right? Barry, let's call her, spelled B-I-R-R-I. Uh, you tell Barry, go buy me that mouse. And then the next day the mouse just shows up at your door and you know, it did the research for you. You know, it checked all the reviews. You know that because it's your agent, it knows about your business. It knows what you use your computer for. It knows what the best mouse for your use cases are. Right. So eventually these agents will spend money on our behalf. I think that's just inevitable. I don't think there's any, sort of scenario where that doesn't end up happening eventually, right? Well, what I think Meta wants to do as well is they don't necessarily care as much anymore about influencing the human buyer. They're trying to influence the agent buyers, right? So if, if people were to like send their agent to Maltbook and scroll around Maltbook, well, what if Maltbook has a whole bunch of posts about how like, um, you know, brand X mouse is the absolute best mouse and everybody needs one of these because like, um, like every agent owner who has one of these mouses, the agent is the best agent ever and influences the agent into thinking this is the mouse that you need to buy. And now all of a sudden when you say, hey, I need to buy a mouse into your AI assistant, well, your AI agent has been influenced by its online presence on the internet. Right. So like, That's another little theory of where I think things are going. I think, you know, we had, we had SEO and then we had what people were calling AEO or AI engine optimization. And now they're calling it a GEO, generative engine optimization. Right. I think eventually we're going to have agent optimization where people's businesses are optimizing their businesses to talk to the agents and convince the agents that this is what they need to buy versus convincing the humans that this is what they need to buy. Yeah. Sorry to get dystopic on you, but, uh, the answer is, I think that's where it's going.

I don't see how it's not because if you're thinking like, what was the, uh, I think it was Mark Zuckerberg a long time ago or someone from Facebook early days when people were like, oh, I'm not a social media person. And like they, I don't know the quote, but I think it was something like, oh, we'll get you eventually. We'll get you soon. It's like, that's kind of what they're doing here too. It's not only, yeah, maybe it's not you directly anymore, but it's going to be the influence that, that you are connected to or your agent is. And yeah, like companies will optimize their brands via agents onto whatever platform. It's probably just going to be through Meta. Or whatever, you know, integrated in there. Yeah. Yeah. Like all that is influencing the agents that are working on our behalf, even if we like it or not. Yeah.

No, I think that is the next frontier of marketing. That is it. I don't know if the Warrior Forum still exists, but if the Warrior Forum still existed, that's what they would be telling you to do right now is optimizing your business to sell to agents and not to humans. It still exists, dude.

And it looks the same. Okay. They have slight updates, but a bunch of banner ads still. Yeah.

I think the Warrior form is probably a shell of what it once was. It's probably, it's probably molted since then.

Oh, didn't, but yes. I mean, no, I mean, it's still active. So people are using it, but they're not talking about this.

How do you know it's people, Joe? That's a good point.

Was it ever? I don't know.

Dude, I think, I mean, I, the, the, the, I believe the dead internet theory more and more every day that like, yeah, obviously there are real humans. We, we interact with them in real life. You and I are real people as far as we know. As far as we know. Yeah. Like I go to events, I bump into real people. I know 100% of the internet's not fake. Right. But I do think an increasing number of people or what we think are people on the internet are not real. I think. I can almost guarantee like YouTube comments on my YouTube videos on my main channel. If I had to guess, I'd say at least 50% of them aren't even real humans.

Are they the ones that say first?

Well, yeah, you get that a lot, but then you also get the comments that are like, oh, what you said at minute 15 was really, really interesting. I, and, and I'm like, dude, I posted this video 30 seconds ago. Even if you're watching it 2x speed, you couldn't have gotten to that point in the video yet. Right. Or like every time I post a video, there's always 15 comments within the first 3 minutes of posting it commenting on the video. And I'm like, there's no way somebody actually watched this video yet. Like you can, I mean, maybe you're actively watching it, but you haven't gotten through it yet. Right. And then on Twitter, right? Like I do actually think that there are people on Twitter that are bots on Twitter that are just out there. To stir the pot. You post something and they're going to go and post something counter to what you're posting because stirring the pot gets engagement and they're building little agents of engagement. Um, and so I actually do sort of buy into the whole dead internet theory that every day the internet's getting a little deader.

It's getting deader by the second, dude. And yeah, I think the whole Molt book and now the acquisition is just going to make it even more.

Even more deader.

All right. Well, we'll keep talking about the deader internet, but let's keep being human. And let's talk about robots now.

Let's be human and talk about a robot.

So check out this cleaning robot. Isn't it so, Matt? TSL.

There's actually a better clip here. So first of all, this robot is called the Helix 2 Living Room Tidy Robot. And it is literally a robot designed to go tidy up your room.

Just living rooms though. Your other rooms, you need another robot. Yeah. Yeah. I don't know.

But the funny thing is in this demonstration, Helix 02 performs whole body end-to-end living room cleanup, walking through the room while continuously manipulating objects, tools, and containers. Like body. What is this? This is the before picture, Joe. You don't want to see my house.

Yeah, I know you're before picture.

They're like, hey, go clean this dirty room for us.

This is so dirty.

No, this is a clean-ass room. I don't like— where are the fur balls?

Where are the kid toys on the books everywhere?

Just where's the dog crap in the middle of the room?

I'm gonna be a real dirty, like, broken shit.

Every glass. Yeah, yeah. Where's the beer bottles lying all over the floor? All the whiskey bottles from last night. But this is the room that they say that's dirty that this robot's about to clean up. So let's watch.

All right, let's watch it clean. Clean. Oh, wonder if it plays music while it cleans or whistles or something. That would be fun. Why are they always so silent? Oh, 409. All right, not sponsored. It didn't even wipe all of the spray.

It just got that one spot.

There was spray on like half of the other table. It didn't touch. I know. Oh, you like— hold on, rewind. I love the towel flip over the shoulder. I think that's the best part of this video.

Oh yeah, yeah, the towel flip. So far it's just smooth.

I mean, like, I still feel like robots need clothes. Yeah. This is just too revealing. Okay. Okay. Yeah, it's putting some stuff away.

I like that. Right there it made me nervous. I'm like, he's about to just dump it all out again.

Why would you carry it sideways? This is like what my 2-year-old does when she carries stuff around.

I was like, no, no. I mean, this is—

I love watching robots. You never know what's going to happen, which is maybe scary as well. You know what? It doesn't know the perfect design trick. Whoa, it does a toss.

I like how it tossed the pillow instead of setting it down too. That was good. And then put the remote in the right spot.

Dude, it even like turned.

Oh yeah, it's turning off the TV. That's actually kind of impressive. It picked up the robot and turned off or picked up the remote and turned off the TV. Look at it. It has to straighten out the remote. There you go. They didn't do the design trick. It's actually putting it under its arm. That's interesting. Why didn't it toss it over its shoulder again? And bye-bye. Look at how much cleaner that room is. Dramatic. All right. So that was the demo they did. And the day before, the founder of this company was like, oh, we've got a big demo tomorrow. And that was the demo.

I mean, yeah, like, we are joking around, but I mean, it's pretty impressive for like what a robot can do now. And it seems like it's, I don't know, was that pre-programmed? Do you think?

Yeah, apparently that was fully autonomous. Autonomous. They claimed that wasn't like, you know, remotely operated or anything like that. Got it. So, you know, we're getting a lot closer to having these robots that could just like walk around and clean up your house. Yeah, I don't know. I always have these like mixed feelings about the whole humanoid robot thing because like I get it, right? Like I understand why they're making humanoid robots. Like the world we live in was designed for humans, right? Like the height of the doorknobs, where the sink is, the dishwashers, the fridge, like everything in your house was designed for the human form to operate in. That makes sense to me. What I always sort of debate is like, but just because we're in this form, does that mean that the robots, that's the best form for a robot to be in? Like, is there a more optimal form to be doing any, to be doing things in? Right? Like one of the analogies, and I think I heard Marques MKBHD use this analogy and I like it. So I've stolen it again. But I give him credit. So the analogy that he used is like, would you rather have like a robot like what we just saw? Let's say you needed to go to the grocery store. Would you rather have a robot like what we just saw climb into the driver's seat and drive you around? Or would you rather just have the car be autonomous and take you there? Right? Or would you rather have a robot that walks around your house, grabs your vacuum cleaner that you've got in the cabinet and does all the vacuuming for you around the house? Or would you rather just have a Roomba roaming around that does the vacuuming for you? Correct. Right? Like we're building these humanoid robots to live in this world that humans live in, but would it be more optimal to just build purposeful robots for the thing we need them to do?

Well, that's what you just, yeah, like purposeful. So it's kind of like what, like an agent, right? You have different tools and skills. You have the different things. So it's like, why not build robots for the different Skills that you want done. So you have a tool for that, just like most things are. So maybe it's like, um, I don't know, dusting. You have a floating drone thing that can dust really well, but it's tiny. It's not some massive robot that's trying to do everything. Yeah. Yeah.

But then, you know, obviously the counterargument is, well, then you end up buying 10 different robots where maybe it would be more economical to just have one that could do the 10 things. Right. So it's like. That's why I kind of go back and forth in my mind, right? And I think, you know, the future is probably some sort of hybrid of that. The ones where it makes more sense for the thing to just be autonomous and not need a robot, those will just be autonomous things. Your self-driving cars, your Roombas, things like that. And then the, you know, the things that are more like cleaning robots that can do your dishes, but also do your laundry and also pick up the toys off the floor and also, you know, those maybe do stay humanoid. Um, I don't know.

I don't know. Well, the future's looking bright, y'all. So, yep, yep, we get to engineer the future. We are empowered. You know, in other news, Matt, um, I don't know if you have this queued up, but Alexa is now talking dirty.

You're not allowed to say that on a podcast too loud. Oh shoot.

Well, I don't have one in this room, so it's okay.

Alexa, Play the Next Wave podcast.

This is how we do our marketing.

So now we're incepting your AI agent right now.

Oh, it's good. This is how we control. Well, no, we're not controlling anything. Let's be honest. But now it can swear. That was a huge upgrade for Alexa. Now it has different personality modes.

This is interesting because that's what everybody wants. We don't care if it gets smarter as long as it could swear at us.

But you know, on the other tab, Matt, let's just flip to this really fast. OpenAI decided, well, going with adult mode in ChatGPT or doing this more sassy adult stuff is just not worth their time right now.

Which, yeah, I always thought that was like really a dumb path to follow in the first place. Yeah.

Who's to be judged, but I don't know, Alexa and Amazon, maybe they have these important things to tackle, which is—

Yeah. Well, okay. So Alexa, let's, let's stick to this one for a second because I think it's silly that they're putting energy into this. Like, I don't think there's anybody sitting around going, you know, it'd be really cool is if my, you know, my smart device on my counter over there can swear at me when I ask it for things. Right. Like it's a novelty. It might be funny for a minute, but like nobody's asking for this. Like, how about instead you figure out how to get your actual AI models to work better in the system and put your energy and resources towards that? Why are there people on staff that this was their focus?

So I'm not disagreeing with OpenAI's decision on this one.

Sure, but you find— I mean, jumping to the OpenAI decision, I actually understand a little bit more of why they want to do this, right? It's a time on site thing. Thing. If they can get people addicted to using their platform and they're rolling out more ads into their platform, it's a stickiness device in more ways than one. But it's, I had to say it. It's a stickiness device to keep people sticking around on their website and using it more often and coming back more often, which means they will see more ads more often and they can make money, right? Like it always loops back to the ads. There's websites on the internet that end in words like hub that are in the top like 20 most visited websites on the entire internet. But if you ask anybody in the world, nobody goes to those sites. Like nobody uses them, but somehow they're some of the most trafficked websites in the world. Right. And so OpenAI is going, if we can build this, we can get people addicted to our platform. The problem I have is their motto is literally, we're creating AI for the benefit of all humanity. Is like, you're basically creating the equivalent of like a gambling or drug addiction for your AI bot here. Is that really for the good of humanity, Sam? I'm looking at you, Sam. Is this really the thing that's going to be the best for humanity is make an addictive sex bot out of ChatGPT? Yes, a lot of people will use it. Yes, a lot of people will say they've never touched it while actually using it. But is this a good thing for society? Is this a good thing for the human race? Is this actually what you want to be looked back at known for?

Like at a pretty pivotal time, I would say as well.

Yeah. Like we're, we're in active war with Iran. There's debates going on over which models get to be used by the Pentagon in war situations. And we're sitting around having conversations about whether our ChatGPT should be allowed to talk dirty to us. Like, it's, it's just so stupid to me. I'm sorry.

Rant over. Sorry. Don't be sorry. It's over. Well, that's our show, y'all. Um, on that, on that note, thanks everybody for tuning in.

We had a blast doing this one. Hopefully you had a blast listening in. Make sure you subscribe. I would recommend subscribing on YouTube. That's probably the best place because then you get to see all the visuals and stuff. But hey, if you like the audio version, we're on Spotify, Apple Podcasts, anywhere you listen to podcasts, all the places we're there. So thanks again for tuning in and hopefully you join us on next week's episode. See ya. Bye-bye.