Hello and welcome everybody to The Attention Mechanism. My name is Justin Robert Young, joined as always by the inimitable Andrew Main. You know, it's very funny because we are doing this on a Tuesday afternoon as opposed to a Monday afternoon. And in that intervening 24 hours, We had a buyout offer for the nonprofit wing of OpenAI by Elon Musk. which is its own paragraph of an offer. We had a new... web agent that is a competitor to OpenAI's operator.
hit the scene in the last 10 hours. So I don't know how much either of us have really gotten to play with it, but still something to talk about. And that all happened in the intervening time when we would otherwise be having this conversation. The buffet is out there. Where are we going first, man? So I want to share just a quick anecdote. It happened yesterday. So I was meeting some friends at an AI company yesterday.
and was in a city at SF. And this is sort of understand why people in SF maybe have a distorted view of how the rest of the world works. I text with a buddy who lives in the city, my friend Lucas, who's another AI developer. And I'm like, hey, man, I'm in the city. You want to hang out? So that's great.
And then I says, and again, all of San Francisco, I'm in the building there. And I said, where are you at? He goes, oh, he names, he says, you know, Mission Street. I'm in a building there. I'm like. I'm at the building there. And then I go, what floor are you on? He goes, 19th floor. I say to the person across me, what floor are we on? He goes, 19th floor. And it's a co-working space.
I go, step out into the hallway, steps out of the hallway, and there's Lucas, and there's some people saying, I wanted you guys to meet Lucas. Here's Lucas, here you can meet. And SF is like that, where... We tend to think that everybody knows what everybody else knows. You think that everybody knows that this is the new thing. And so things come kind of fast.
And you kind of get this sort of distorted view too of kind of how the world works, you know, because it's just like, oh yeah, well, the guy that runs that company is five blocks away. You know, the people who built that billion dollar startup are having lunch over there. And it's kind of like being in Hollywood and being in the entertainment industry where it's no big deal, but you also don't realize that most of the people are normal people. Most people are not tech people or whatever.
And so when these things come at you, it's the kind of thing where, you know, we've said here, because when Operator came out, I think two weeks ago, three weeks ago, we were like, yes, this is going to be the year that you're going to see a lot of these other tools coming out.
Yeah. And you're getting, and we said, you know, we've been talking about this for like two years that was going to go from chat interface to browser. You're going to, your chat was going to control, you know, it's going to start controlling the browser. And so we're seeing that and that's going to be a lot of acceleration there.
So browser use, I think, is a great one to start with because, you know, we saw OpenAI come out with Operator. And then now you have another tool that's out there. Operator, limited to, I think, just pro users, which mean people are paying $200 a month.
A lot of people are like, oh, that's frustrating. It's like, well... it's a research preview you know it's i think it's useful for some things but it's in development you know these other things too we're gonna see I think the big trend of the week, though, is just to see how fast other things come out, which we're...
they're not slowing down. And that's what I will, I will give a shout out to convergence.ai. That is the proxy service that is getting a lot of attention today that launched today, Tuesday, February 11th. Lord knows what we'll launch between now and when we release another episode, but... I wasn't even looking at that. I was looking at browser use, which is another one. Oh, really? Yeah. We were looking at two different new agentic browser searches.
so proxy i played around with a little bit this morning and I will say it is very fast, and it did a pretty good job of, you know, I gave it a task of, you know, for PX3, the politics show that I do. I'm like, find me the top news stories of the day. Make sure that you are taking from a wide ideological viewpoint. And Convergence's solution to that was to go to CNN, the BBC, and Fox News and take the top headlines. Now, it didn't do it perfect.
It did not have the links correct to them. It told me that it was giving me the links. The links weren't real. And to be totally honest, I don't know if the headlines were even real, but it did have the sense to say, okay, what is ideological diversity, at least in terms of this agentic perspective. It thought that those three were different places on the spectrum. Makes sense.
I tried to do an operator. Operator had a harder time trying to, it went to a site, allsides.com to basically, that's how it looked to solve the ideological viewpoint question it wound up. choking there, but it shows you how exciting this field is right now that this stuff is moving that fast.
Yeah, I would say that, and that would be a task that I would have thrown at deep research or O3 was search. Not necessarily operator, because I find that operator just really is optimized, that model they used for it. I forget what they call it. But anyhow, it's optimized around more tool use than so much that. But again, eventually in the perfect world, we'll just... ask for a thing and the AI will figure out what to do. We'll figure out the smart way to do it, yeah.
Yeah, I was curious to see, because I know that one of the things with operators is that it chokes on tech, so I was curious to see whether or not Convergence did better with that. Proxy seems more like a perplexity or a research tool and not really a browser operating tool. Maybe it's not proxy then. They do have a new thing. Let me make sure that I'm getting this name. All right, because this looks kind of like a perplexity kind of thing.
There's browser use, which we mentioned, which they show their score. They say that they beat operator on the... their tool is better on web agent accuracy. Of course, as we find out, there's a lot of these different kinds of metrics for that. Let's see. Alright, here we go. Let's find what the hell I'm talking about. It is. I was having a conversation today. Proxy.convergence.ai, that is what I'm talking about. But that was a browser.
It opened up a window for me, but it didn't look like it controlled or I could do anything functions in there. Yeah, I mean, it did for me. So it has the templates. So here, let me share my video. Here we go. Like if I put in like a form URL, it'll fill it out. I don't know. I don't know. We can try. So we'll do one of their... Let's just say... This is the search I did earlier. Top political news from a variety of ideological sources. So it has opened up Proxy's view. It is now starting...
It has gone to Google. It has Googled Fox News. Yeah. It is now clicking on the Fox News headlines. And then if it did what it did last time... And it's on now the Fox News homepage. And there we go. Top political news from Fox News. It has accurately copied the headlines. Okay, so I thought that when it said link, it was going to link me to that story. It is just linking me to foxnews.com. so it is not technically the wrong links. Now, it says it completed the task.
which I asked it from a wide variety of ideological sources. It only got it from Fox News this time. Last time it got it from three different sites. But that is what proxy.convergence.com is. is. But that would be a... Yeah, no, it's like trying to log in. Like I asked if I could log in and create a Google form.
So it is trying to do a login thing. So I think it, I mean, it seems like it's agentic like that. You know, the examples I gave were all like perplexity style ones, but it does look like it can do other stuff. I don't know if it's actually going to.
you know be able to do it i'm not saying it can't it just it looks like right now it's optimized more for like data retrieval not necessarily a browser operation but i could be mistaken on that but anyhow and then that's the point is that you know these things are moving very fast and you know, these tools overlap in their capabilities, you know, and what they're able to do. And so I could say like, yeah, it's funny because You said one thing, I'm thinking of something else and whatnot.
It is absolutely crazy. And so I think that, and it's frustrating for people because you're thinking like, how do I stay up to date on that? You know, I was talking to somebody the other day who runs a very big company. and was showing them stuff. And he's like, man, I just don't have the time for this anymore. And you're like, man, some of these decisions could... Be greatly benefited by understanding what the state of the art is, yeah.
So before our call, I was showing justinmake.com, which is in a category of Zapier and these other tools. If this, then that. And I like make.com's interface because I thought it pretty well. And they did a great integration of, you know, GPT-4 or just the OpenAI API.
And it's a little bit of a technical, even though it's not, it's no code, you know, but it's a thing where you can have a thing. If I get an email in my inbox, have chat CPT, take all my notes and summarize them and then do this. Or if I upload a thing allows you to chain a bunch of things together. And the challenge is they've gotten rid of the coding and things that I'd built before that I'd spent way too long building in code. I could just automate away with this.
You still have to know each individual step exists. For example, I wanted to build a thing that would convert. our podcasts automatically transcribe them. And if I did it the default way and I use, let's say, OpenEyes API, that only lets you upload 25 megabytes, which wouldn't cut it for an hour-long podcast.
But I knew DeepGram's API lets me do things up to like a gigabyte or two gigs. So I'm like, how long is their API instead? Not a thing that people necessarily know. Not a thing that you understand. And then you might think, well, this tool can't do what I need. I need to go use something else. kind of where, you know, we were talking to, I was talking to our friend Brett today about the idea of automation engineers. Forget AI engineer, prompt engineer is just an automation.
Oh, just understanding, especially with a knowledge of AI, how can we automate these things that deliver the kind of results that are going to be what they are? When you think about elements of our technological stack that are enhanced by AI some more than others, automation is
a huge tool. And we're already seeing it in places where repetition is key and speed is key. They're not necessarily the most enriching parts of our internet world it's a lot of like a lean generation and a content slop and stuff like that but We haven't seen, I think, fully unleashed automation's power on you know, good things. Good content. Speed of good content. Like, can I, as a one possible two-man shop, uh, become
a far greater news alerting service than I would be otherwise? Can I spin out my content into varying different places with less time being put into it? These are things that automation can provide. So I did a quick demo for Justin before to show him some of the workflows I've been putting together.
It still takes time to put these things together. Eventually, AI systems can only take you so far. Even if I have the smartest human in the world, they're going to still say, can I use this account? Can I use that account? Can I do this? I was playing around this. I said, man, if, and I'm sure there are people out there to do this,
I would have hired somebody. I would make a list of things that I want to automate. I would love to just have hired somebody to say, go build this for me. And when we talk about new opportunities is. You know, I would pay a couple thousand dollars if somebody set some of the stuff up for me between my books and my other stuff and what I'd pay for my business stuff would be in the tens of thousands of dollars to do a really good pipeline.
And then every month pay somebody a fee to just make sure the stuff is up and running. I just think there's so much opportunity out there if you're just entrepreneurially and you want to learn how to use these tools. yeah yeah you know i mean it's funny the point that we're at now is that You can see the tools that will build the future. You can touch all of the tools. It's them working together and then eventually once somebody creates a minimum viable product.
that form being perfected and built upon and built upon and built upon, that remains to be seen. But one of the things that we've talked about a lot... is something that gets a lot of conversation in our modern world. In fact, it's again popped up in the political realm is whether or not social media is healthy. Healthy relationships that you have, especially on the younger age spectrum.
on social media and how much of it is a dopamine factory that can be problematic. There is a world now with automation and especially with the agentic stuff where you could say, You could automate a job of a social media intern. And by that, I don't mean somebody who's creating posts, but literally managing your social media and bringing you... a reflexive-to-your-demands summary of it on whatever
basis you would like, right? And so now you can still have some kind of really, like, I don't ever log into Facebook, but would I like to know if somebody that I knew from college commented on a post that automatically cross posts from Instagram?
Yeah, I would like to do it. Probably not enough to log into Facebook, but if that information could reach me, I would be interested in it. And that's all the tools are there right now. It's just a matter of... that being put together and becoming a product, you know? Yeah, I think...
I've been working on kind of a follow-up piece to my article I wrote about the 0% unemployment economy, the idea that how AI could accelerate employment. And I've avoided trying to get too much into like, what are the jobs, the future and stuff, because that's just...
incredibly, a little bit more speculative than I feel comfortable doing because it's just, you know, it's also, it's also the fastest way to be like either an absolute genius or a total, total dollar, right? If you wait five minutes. I mean, imagine trying to explain Mr. Beast or PewDiePie in like 20 years ago. Yeah. That, that, that, the, you know, that.
The budgets of, you know, we started working in TV around the same time YouTube launched and trying to explain to somebody the budgets on these YouTube shows are going to exceed television, broadcast television shows. In fact, I think I remember... We were scouting a bar that we used as a shooting location for the first pilot we did. And that was the first time that I saw on the television the news that Google had bought YouTube. Well, that was the big mystery. How's YouTube going to make money?
Yeah, yeah. I mean, and also, you know, we still don't know, but we do know that it is the cornerstone of cult. Yeah. Yeah. And I think that that's that's it's so hard. So the idea of, you know, an influencer and stuff and whatnot and kind of the example I started to think about was.
Things that seem like minor gigs are minor things now, maybe bigger, or things that you think are a thing now will transform. You would think that there was sort of a one-to-one... transition from being a local radio DJ to a podcast. But they were the last, if ever, people to make that transition because I'm on real radio. I'm in the real world, not this fake thing. And other people in the fake thing are like, cool, I'm going to go build this up.
You know, local DJs and radio hosts were like, oh, shoot, I got to get into this thing. Cool. You got to compete with 10,000 other people that were believing in it when nobody else was there. And I think that's one of the things that's hard to predict is to say that like. you would go like, oh, well, who...
If we asked somebody 20 years ago, who do you think are going to be the biggest stars of YouTube today? We'd probably go probably the top movie star stuff like this, you know, you know, you know, the Tom Cruise channel, Ben Affleck, all of that. And then they'll do it like. Go look at how many people actually watch a Ryan Reynolds video on his channel. A good amount, but not... And he's somebody who's smart about it, right? Let alone people who are lazy. I had a conversation with somebody.
who is one of these professional... college coaches that'll, you know, they'll have a kid who's graduating high school that they'll tell them what they want to do and they'll help guide their application process and blah, blah, blah. This person is a listener to the show. She reaches out and she's like, hey, I have a client that wants to go to Syracuse. Can you talk to me about Syracuse?
like sure yeah it's fine she's like what do you think the best reasons for somebody to go to a school she wants to be a broadcast journalism major what are the best reasons to do it and i'm like Well, the best reasons would be the networking. But you're not getting the same benefit that you used to 20 years ago when I graduated, let alone 30 years ago or 40 years ago when graduating from that school was... You punched yourself a ticket to the farm system of the world that you wanted.
It's hopefully will teach you discipline and hopefully will teach you that there are a lot of really ambitious people that will work really hard and you'll want to work really hard around them. But I was like, to be totally honest, you should tell this girl. The thing that you should do right now is just start working. Just be recording something. Record and post. something today.
Record and post something every day that you can. Learn the tools as fast as you can. And when you get your hands on better equipment, when you start to meet people with other skill sets, you'll know, oh, that's cool. You do that. Let's do this thing together. because No matter what, regardless of what your degree is, you're competing with. millions of people in a world that used to only have Hundreds. Maybe thousands. And that's what's changed, is that we have flattened out.
the expertise curve. We flattened out the authority curve. And AI is only going to further flatten that out, in my opinion. Although maybe it actually does the opposite, and it makes the people that really want to run fast, run faster. And maybe it makes the experience curve a little bit steeper. Well, I think that's the thing, is that I think that we've seen... You and I, I would say that we were both raised to be a bit entrepreneurial. You know, your mom's an incredible woman that, you know.
you know, watched her, you know, move from one opportunity to another, adapting skills, whatever like that. And I think she encouraged, you know, your brother and you'd think like that. My parents, both, you know, government workers, you know, my dad would be like, you know, he'd organize a neighborhood garage sale where everybody, every neighborhood, you know, the whole cul-de-sac would be doing that. And he had my brother and I out there selling hot dogs and stuff.
we'd go to Costco and you'd see that like, you know, like candy bars were like this much sense. Like, yeah, I do want to take this to school and sell them, you know, do that. So. That's not common. And I think that's the hard part is every time I would do a lecture, whether it be computer science students or art students. And, you know, my point would say, like, I can tell you who the most successful kid in class is going to be.
it's the one when I ask them what they did last weekend. And if they tell me if they're in a film school and they said, I went out and I made a movie with my friends that the teacher didn't ask for, I just did it, then they're going to be successful. If it's a writing class, if it's a kid that said, I spent all weekend writing, that's going to be that kid. If it's the kid that...
You know, me and my buddies got up late coding the dumbest thing you can imagine. You know, we're trying to control an RC car or whatever. They're going to make it. The one that's waiting for the next assignment and waiting for the teacher to tell them what to do. They're being trained for a role that is going to be harder to find opportunities.
Yeah. Yeah. There will be opportunities, but to find the good opportunities. And that's why I think everybody needs to think more entrepreneurial. And, and that is entrepreneur is like, what resources are out there? How do I make the best use of them? Yeah, because I think the good news is that there's going to be wider amounts of money out there, but the systems... Oh, the systems. The place where you wait in line and hopefully achieve something by attrition.
that's going to be hard because I think we're already seeing the crumbling of them. But then again, maybe there's new systems that arise. And that's always why writing an article like you're talking about is always so hard because even the smartest people... When you look at, you know, like, oh, like a futurist thought, you're like, oh, wow, look, like they totally predicted this thing. And then you read the entire article and it's like, yeah. And they had some wild guesses about other stuff.
Yeah, I and my my gut is telling me the thing to look for is You know? What are, an example I use is I, and then there I talk about if you're, you know, 5,000 years ago, a couple of Sumerian field hands watching the first ox pull a plow across a field thinking, oh man, we're doomed. This is the end of jobs. And I've kind of flippantly sort of said that's what we are now. We're looking at these tools. But then I have to think, well, the plow didn't look like a person. The ox didn't talk.
Now, robots are getting as agile as gymnasts and chat GPTs, conversational, etc. These things seem to have the form of people in many ways. And the problem is, it's like, well... You're more than how many bales of hay you can lift. And you're more than, you know, your actuary tables you can calculate or whatever. You're part of a network. You're part of a lot of other things, you know, the tenant or titles, occupation.
human you know and the fact that we we have to sort of step back and realize that and part of my point is like you know As the systems become more complex, there's going to be a lot of things you won't want them to be doing. But anyhow, I kind of think about like if I was trying to figure out, if they're trying to figure out what are the jobs of the future, well, the first thing to look around is say, okay.
Well, I don't know. What do we need? Like, oh, I've got my one tunic. You know, like, what if you have more tunics? How can I have more tunics? It takes me all day to make a tunic. Yeah. But I don't like the same tunic. What if I design tunics? Exactly. Everybody is, you know, you and because, well, how are we going to pay for this? Well, we just, you know, farmer just raised a lot more wheat. Maybe he's going to want that. So point is.
You look around you and you sort of say, like, what are some of the small things now? And those become big things later. Leisure was a small thing. Leisure became a big thing. Health care was a small thing. It became a big. Education was a small thing. It became a big thing. Travel was a small thing. Now it became a big thing. And that's the thing is to look at what are the small things we enjoy.
What are the small things that could have way more dimensions to it than we realize? And we expand it. And it's why you can't predict, you know. when the first pinball machine was made in the early 1800s that games would turn into things like Grand Theft Auto. Anybody who could draw a line between that thing and the other, that'd be amazing to me. And I think anyhow, my point is, I said, that's what we kind of look for that.
One of my favorite points that you've made, and I don't know if you've done it in text yet, but I know colloquially we've talked about it, is the idea that we are all out-of-work farmers. All of us. We are all sons and daughters of out-of-work farmers. It used to be the way that everybody made their money. Now it's how... You know, less than 10% of the world makes their money. 98% of all jobs are BS jobs we invented after we industrialized agriculture. 98% of all jobs. Name a job.
And if you're not like, you know, not working the cattle or a field. We just, that came up because we needed a thing. And I have friends like, well, I'm an accountant. What are you an accountant for? A movie studio. Well, guess what? Movie studio accounting is an entirely new category of job. You know, like, oh, wow. you know and again yeah it's this i'm a chef is that a new job like well uh not a lot of people can afford chefs yeah so no we are all techno wizards
compared to the farmer, compared to the farmer that got industrialized. And it turns out it's been better. So I have a talk coming up. I'm going to be at Reason Weekend in March. And I'll be there doing a panel, Brendan McCord and Nick Gillespie. And one of the things I want to talk about is...
How do we get to that future? And I think it's not, we can wait for things to do that. But my concern is People will be making a lot of dumb decisions between now and then and the dumb decisions could derail stuff we've seen this historically yeah that's why europe is now having an ai conference and panicking going oh my god we're going to lose on ai and losing ai could mean you know uh the difference between The West in Europe could look like Western and Eastern Germany.
And a lot of Europeans are like, no, we need to figure that out. But how do we avoid that scenario? Because as we have job loss, whatever. So that's one of the things I want to get into is like, what do we do? And I think encouraging entrepreneurship, encouraging people. to pick up new skills, encouraging tool ship, tool use. And I can just literally figuring out like, like how do you say this earlier? Like how do we create, you know, thousands of more companies? Yeah.
Well, we are certainly going to be on the road to that. We'll talk about that the next time that there's 50 new products that we can't even get straight exactly which one we're talking about on the next episode of the show. That's so funny. Oh, my God. Oh, yeah, the new browser thing. Yeah, okay, yeah, that's cool. Main, where can people find you? I'm andrewmain.com and at andrewmain.com. You can find me, Justin R. Young, everywhere on the internet. Until next time. Friends, we'll see you later.