Transcription
In a way, the entire US economy is just betting on AGI. But Claude is shutting it down, which I'm saying is so stupid because if you look at the net effect, Anthropic, you've lost a $200 a month paying customer.
The open claw argument would be, hey, um, we have ushered in a lot of people that are now buying your highest paid plan, $200 a month. They're buying it specifically for this amazing personal system, and they're going to other models like Mini Max that are 90% of your quality um, for a fraction of the price. And they think it's absolutely stupid.
The real innovation with OpenClaw is basically that it allows you to text like on your phone. Um, I don't trust that Open Claw is good. I don't want my brand associated with Open Claw because of the Nicholas Rev.
How's it going, bro?
Dude, fantastic. I'm going full hobo core today. New style that I'm trying. How about you?
Hobo hobo maxing.
Pretty much, man. Yeah.
Dude, I love that. He looks good. He looks you great, man. I I saw a funny about um you talk about locks maxing community where like you all these little micro things and this guy was like, you got to find perches so that you're at maximum height at all times. There was like, and he was like, dude, I will strategically lean against counters and I plan my routes out as I'm walking out the doors. The guy playing 6'2. And then there was I saw a funny one on dwarf maxing. So like for people who are like really short and the idea is like, dude, I've started like I've grown out a beard, I'm drinking mead. He's like, really leaning into it to like dwarf max out and stuff. And like it blew up on Reddit actually. So these trends, dude, they're popping around, but I've never seen the hobo max. Dude, I think this could be the next big thing.
It is definitely the next big thing. Slept on commodity, invest now. Dude, I saw this um I saw this uh guy who was Jawsizer-ing the other day. Have you heard of the Jawsizer?
That's like where they chew the the ball.
Oh, that's you get the the nice defined jaw.
The massive. I never knew if that was fake news or not. If it actually worked.
Well, like you get big jaw muscles, but like the jaw muscles are different from like the jaw line, you know, like and then like it kind of makes you look kind of weird. You're always like,
"Yeah, the mewing is uh also pretty funny." I don't know what the hell the kids are up to these days, man.
Yeah. I I think the best thing for jawline is reducing body fat. That that's like number one, right? And then they can like chew, I don't know, coal, whatever the hell they're eating. I don't know what they're reading, Nick, but there's there's a lot of communities that go deep on it.
Damn right.
But dude, I was going to say, speaking of um Jawsizer, dude, are we kicking off today with Claude? So, today we're going to go comments. We're going to look at some Claude stuff. We're going to do a little bit of a freerange roast stuff. Some people are going to get in the fire and land. We can't say who or what's happening, but we're going to have a little play around with that. It's going to be good. Where do you want to kick off today, Matt? Should we kick off with some Claude stuff?
Yeah, for sure. Why don't we start with Claude and then we can do a roast of uh some LinkedIn profiles. Jack and I were quickly scrolling through and looking at all like these NAN giveaways and Claude code giveaways and I think it'd just be worth it uh for us to riff on, you know, what these things are, what like you're getting at the end of the day. We don't know any of the people involved, but we're just going to like open up my LinkedIn profile and see. And then uh we could do some some hot comments as well. Obviously respond to all the people that are calling us morons. That's always fun. And then many.
Let's do it.
Cool. Sweet.
All right. I'm just going to pop open a little screen share here and then we can check out the uh new main model announcement for Sonnet 4.6. And the main reason why I want to bring this up is because there are a couple of cool advancements that have been made with Sonnet 4.6 that just weren't present before uh in a lot of these models. So, to make a long story short, um I'm just going to full screen this here. I don't actually know if full screen works. Apparently not. Um, to make a long story short, basically the other day, uh, Sonnet dropped 4.6, which is new. And so they call it their most capable Sonnet model yet. Full upgrade of the model skills across coding, computer use, long context reasoning, agent planning, blah blah blah blah blah. So obviously, you know, if you think about like Sonnet 4.5, this is basically like Sonnet 4.5, just better in every way, shape, or form. And they talk about performance for Sonnet being comparable to Opus uh 4.6 6 as well, except this is cheaper and then it also has a 1 million context window. So that's kind of interesting to start, but the big thing and the big reason why to me, well, I guess there's two that the Sonnet 4.6 drop is really big is because of computer use and so essentially when Sonnet controls your browser uh versus uh I think even versus Opus, not only is it like really cheap for it to navigate your browser and stuff like that, it's also pretty accurate. Is this is 72.5% accurate um compared to Sonnet 4.5 which is 61.4 and I think Opus like I'm not entirely sure, but I think Opus is like right around here. So basically there's just more training for computer use because you know the Anthropic team anticipates this is going to be a bigger and bigger and bigger deal as people, you know, move away from APIs and then start automating the things that were previously unoptimate unmatable.
And then so.
Yeah, no, it's going to double click on that. It's interesting. So Sonnet 4.6 6 is best at things like browser automation. Like, hey, you got to open a browser, you got to click some buttons versus Opus. That's essentially what that's saying.
Yeah. Yeah. It's really really good at browser automation. I mean, 70 whatever percent um accuracy on like oneshot, you know, like one-off tasks, but consider how many times you try and do something on the browser. Like you you don't always get it right the first time, right? Like me going over here and then clicking on this tab or something. I don't know. Sometimes I'm going to miss. I'm going to click here instead, right? And then that's not going to do anything. So, what am I going to I'm just going to move my finger up a little bit and re-click. So, it's like 76% is actually really really good if you think about it, especially extended over a long time horizon.
Um, so that's pretty cool. And then like the big thing that I wanted to chat about was this over here. Okay, which is that Claude's web search and fetch tools now automatically write and execute code to filter and process search results. And this is something that like I don't think I don't think a lot of people noticed. Um, but basically this is a massive massive improvement in token usage in a context window for the Claude family of models. Basically every single time that previously you were trying to do a web scrape with uh I don't know, like Opus 4.5 or even Opus 4.6, what's happening to be clear with you is you are loading all of the code on the page. Okay. So, actually to make this even more accurate, if I right-click this and then press view page source, this is what you're loading here. And I mean, it's so big that like it doesn't even want to show me. Like this is this is what you were seeing. And then think about how much work that you know and how many tokens like Claude is dealing with when you inject this whole thing into a prompt. I'm like, hey, can you read me the new news about Claude Sonnet 4.6 or whatever? You know, it's going to load all of this stuff into into context. And whether it loads it in the context of the parent model or the um, you know, now they launch like research sub agents and stuff like that to try and minimize the total context solution in the parent window, like you know that depends on the various models you're using. Like I'm sure Gemini is still loading a lot of the stuff into its main context window for instance. Maybe CodeX has some ways around this. But now what they do instead of this, okay, instead of this, is they just parse out the text. So basically, um, Claude Sonnet 4.6 6 instead of loading the entire context window, will just write a brief function and that brief function will just take out the things that you want to take out. So let's say I'm like, hey, summarize this. It'll actually just pull only the text, which as you can see is the stuff written in white here is like one 30th or 40th of the total um content of the page, right? And so we've just gotten like orders and orders of magnitude more context efficient in a very short period of time. Uh, which is really really cool. And then, you know, now because it's filtering and processing not only results from like web calls, but also its own search results like internally on your computer, uh, not only are things going to be a lot cheaper, but then, you know, things are going to be a lot more token efficient, meaning the quality of your outputs are going to be a lot higher. So yeah, I wanted to cover that because I just think it's really freaking cool. I mean, like dude, that's that, what a, what a zero to one improvement.
Hey.
That's really cool. I love the fact that they're building on additional steps from the user query. Like it used to be, I'll just go and do what you say. But I think the more that we use these models, the more they can identify these buckets of use cases saying, "Hey, we have about a million people asking us to do this thing. Why don't we spend two minutes making it slightly more efficient?" Like the underlying capabilities of the model don't need to improve. It's just how it executes on tasks that's gotten better. I I did hear interestingly, just to throw a bit of a controversial thought you're winging on Sonnet is that someone says that even though on paper it's cheaper, that effectually it is not because it ends up using a lot more tokens to get to the point.
That's interesting. When you say tokens to get to the point, you mean it just like it's dumber so it requires more processing?
Yeah. Does a lot more hammering and janggering and so even though the advertised cost is is less, that actually on certain tasks it ends up taking three times as much as another model would do.
Yeah, I think Sonnet needs some Jawsizer, clearly. It's swinging by the shop. It's it's perusing a little bit, getting little detours.
I played around with it. It's it's it's different for sure. You know, like Opus is smarter, let's be real. And I usually like using the smartest model for everything. Um, but it's not that much smarter. It's just like a percentage point smarter on like a couple of benchmarks. So.
I don't know. I mean, like it could be a million things. Another really cool thing about Sonnet is it's trained up to January 2026.
So, I hope this is like the cutoff is like October 2025 or something like that.
A lot a lot changed in those those six weeks.
It did. It did. And like also there's like a lot of new new training techniques um available and stuff like that. Everybody thinks that like Sonnet was actually supposed to be Sonnet 4.6 was actually supposed to be like Opus 5, but then they like dumbed it down a little so that they could release a bigger Opus 5.
I wonder why they did a 4.6 rather than the five. They probably they probably looked at GPT's playbook and realized when we dropped the five, it's got to freaking turn heads.
Yeah.
You know, we can't be slow rolling a five. We got to really,
You know, it's like iterative product launch, right? Like they're like GPT5 and everybody's like, "Oh my god, this thing's probably AGI." And then they start using it.
That was brutal. I felt bad. OpenAI's goodwill was tanking in real time.
It you know, I think that was a bit of an inflection point for OpenAI to be honest with you. The the there was such a disparity between the expectation and delivery. Even though five was great and Chat is awesome, I think it just could never live up to the hype. It was more of something they could do in six months. They should have gone back to GPT 4.7, 4.7.5 or just slowly behind the scenes improve it a little bit.
Yeah.
Rather than drop the five bomb. So I think Claude is trying to learn from that. Dude, we have to talk about the whole open core thing uh open AI thing. Um, because isn't it really fascinating Anthropic's strategy with that about hamfisted you cannot use our oorthth crap and our max plans? And there's two schools of thought here. We'll just outline for the audience. Uh, so I'm going to steal man. We'll try and steal man both cases. In fact, Nick, do you want to pick a side and we can we can try and steam man this?
Yeah, absolutely. Okay, you go.
Okay. Which side do you want to go go with?
Do you want to be team Anthropic or team Claw?
I'll be team Anthropic 100% till the day I die.
Team Anthropic. So, so the the open claw argument would be, hey, um, we have ushered in a lot of people that are now buying your highest paid plan, $200 a month, which which is not like pocket change, right? That's like, oh my gosh, dude. That's like 20 Netflixes these days or maybe 10 Netflix these days. And the idea is they're buying it specifically for this amazing personal system. Uh, but Claude is shutting it down, which I'm saying is so stupid because if you look at the net effect, Anthropic, you've lost a $200 a month paying customer. And this could be people say, "Look, I understand that you want you to protect your signature Claude brand, which is doing over a billion dollars a year in ARR, but it's crazy cuz you're losing customers and they're going to other models like Mini Max that are 90% of your quality um for a fraction of the price. And they think it's absolutely stupid." Um, so they should do it. They want more customers. The more people using your product, the better. Anthropic, what the hell are you playing at?
Well, the big thing is uh as a spokesperson for Anthropic in this hypothetical, um I don't trust that Open Claw is good. I don't want my brand associated with Open Claw because of the myriad security leaks so far. Not to mention all of the security leaks that are occurring that people aren't talking about and all the security leaks that will occur over the course of the next, you know, six months of this thing going. Secondly, the real innovation with Open Claw is basically that it allows you to text like on your phone. And that's something really trivial, which we're working way better ways to do on our end, >> without any of the security drawbacks. So rest assured when we launch our alternative alternative to Open Claw that's way safer, faster, better, whatever, we'll claw, we'll claw some of that back. Oh, one more. $200 a month API Pro Max plan, whatever the hell. That's a loss leader for us. Like, we're not doing this because it's the most efficient thing ever to gain revenue. We're doing that because we want to take up as much of the market share and like light cone and like consciousness of people in AI as humanly possible because that's what we think is ultimately the most important thing at this moment in time. We as Anthropic think we're just a year or so away from fully automating all software engineering. So we don't really care if we run negative. We're not really looking at our metrics. Do we care that we're making a billion dollars? No, dude. We have billions upon billions upon billions lined up just through funding alone. And if we think that that's enough to push us through to, you know, the actual major AGI revolution, then we'll just have AGI do it all later on.
Interesting. Well, well, Anthropic, let me on behalf of all the claws out here, let me um drop some some thoughts to you. By the way, I have a concept I call Gravity Claw coming out, which is basically build your own clawbot in anti-gravity.
Yeah.
And it is better. I'm going to tell you right now, I'm going to have to have to talk chat about this. I think it's way better. So, I guess the point would be that's an okay point of view so long as you are the best model. And at the moment, I think it's fair to say that it is, but that's not guaranteed to be the case. And there is something to be said for banning users and the kind of ill will that creates in the marketplace. It kind of in my view puts Anthropic in a kind of adversarial position against a lot of its core users. Granted, not its mass market because I I don't think that, you know, Sally May or whoever, you know, works at a barber shop is downloading Claude bar, although a lot of people realistically are. So there's an adversarial point there which I think is interesting. Your point about the $200 max plan not being a profit center, I think is interesting. I think it would be good to get under the skin of that and understand are they actually making a profit on that? Cuz I think there's probably that that fractal AT20 law there, some Pro Max users that probably cost them a freaking fortune actually. And what was interesting to me is when these $200 plans came out, do you remember Nick, people were like, are people even going to pay that? Turns out, yes, they are going to pay that and they're very comfortable paying that cuz it's super expensive. I just think it's crazy to me they're losing customers. And here's the thing, it's okay to take a bet on yourself, but just make sure that people don't find comfortable alternatives and never need to come back because if they knock on Mini Max's door and it's like, "Hey, let us warmly welcome you in with tea and and wonderful desserts and and and wonders beyond your dreams. Oh, and by the way, it's very cheap." that they don't end up staying there um for your long-term strategy. But I'll just also segue one little point that as well is just like I think the big play is B2B with this stuff. The API calls. I it's like they're not really that fussed about subscriptions for the average user. Like all the money is in the API calls and the B2B deals.
Yeah, I would agree. Um, I think about first of all, call me Dario, but I would agree. I think um just based off my own API usage over the last uh like week and a half, you know, my Cloud Plus Pro, whatever the hell, the best of the best is like 200. Yeah.
And I've now spent like $800 this month just on API token usage simply because I'm using the fast plan, like you know, cuz because that bills your credits directly. And um, I mean, that's like that's 4x right there. And then not only is it 4x if you think about it per marginal unit of usage, right? If you treat this like a like a commodity basically, which it is, and then you just count it all like per unit, I'm probably using like 10x fewer tokens for the $800, which makes it like 40 times the the profit for them. So, I I agree, but money is not really the main focus to be honest. And and you know, it never really has been. As uh somebody that's been a big fan of Anthropic for quite a while, especially after reading all of Dario's writings, like it's just clear that these guys are very much of the mind that they're going to bootstrap their way to AGI. That's really the only thing that matters. And um along the way, obviously, they want, you know, like acceptance and market share, but like the revenue is is nothing to them along the way. They don't even they don't really care the money at all. Uh, which is really surprising. I guess to most people they're like, "Oh my god, this is a cash cow that got all the money." They're like, "Why worry about a billion dollars now when there's a trillion dollars to be made in like a year and a half?"
It's It's funny, dude. It's like that whole race. I think for a lot of players that if they don't hit AGI, it's not worth the payoff. Like all the acquisitions for Meta, the hundred million dollar deals to get the talent, the war on talent we had at the highest level. It's like the risk of not being the one that arrives across the finish line and hits that AGI is too high. So they're just going to pay whatever it takes to get there.
Yeah. It's like existential, right? I'd wager that like in a way the entire US economy is just betting on AGI. If you just look at their deficit and their budget, >> like the only solution to a lot of developed nations >> is some form of AGI, which is why so many of not only the companies and like the masses, but also now the governments are starting to push this because they're like, "Wait a second, maybe the only way forward is limitless abundance through intelligence."
Yeah. Because if that's not the case, then how do we >> not collapse, right? It's like uh, you know, Louis the 16th or whatever paying a quarter uh of his entire budget in interest or something like that, maybe more.
Well dude, this is why a lot of the big advocates of Bitcoin are like, you need to get some clonic currency. Well, what was interesting to me when I was in South Africa, I saw a few entrepreneurs that had Bitcoin and, you know, I've heard all the arguments and stuff, but I will be honest, they were checking their balances every 25 minutes. The level of anxiety, cuz I think it hit like an 18-month low, was crazy. It was like, kind of this huge thing. There's no other asset class that gets you >> in cold sweats at 1:00 p.m. in the afternoon every day. So, it was interesting.
What do you.
Yeah. Like I wish you could buy like gold. I wish you could just buy a big bar of gold and and not necessarily have to like arbitrage it, you know? Like I wish you could just buy like a physical Bitcoin or something. Yeah, you can buy, but it's even like I I spoke to a dude who's like, "Dude, I've got the silver buried under my bed." And I was like, "But it's not like that's safe." Do you know what I mean? Like that can be stolen. Like it's not as. And it's not as if your Bitcoin's safe because you can forget your pass key. There's hundreds of million dollars of Bitcoin in many like rubbish sites. So yes, >> none of it's really safe in that respect. Do you know what I mean? You're just kind of it's just risk management. I just wish there was a way to in to buy like an underlying asset without contributing to it speculation, but uh the Bitcoin bros are always checking their phones and stuff. That that said, like I have done some reading into it. I've actually had a few people reach out to me like some some pretty prominent uh people in the space being like, "Hey, you have a fantastic platform now. You're discussing like the implications of AI and work. Um, have you considered cryptocurrency and how decentralized and deflationary uh, you know, money is like the key to a sustainable future for us all?" I was like, "No, that sounds like crackpot thinking." But then I like looked into it and I was like, "Oh, wow." Like, "This very well might be some form of decentralized uh monetary system is like really the the only way to solve this." So, I don't know for sure. I don't know for sure.
I I I love the idea of it coin, whatever it's called.
This idea that >> basically you're just getting the banks out. Why do we need them? Why have they got to be in in the transaction?
It's like real estate agents. It's like, do you need a real estate agent now? Do you need a car dealer? And same with lawyers to some extent, right? It's like they're never going to engineer themselves out the process because they have what is called uh income requirements and they would quite like to continue receiving that income and charging you a significant fee for it. So they're never going to say, "Guys, looks like our job's done here, but well actually what we need to do is bring in this complexity." So I think that's fascinating. You know, it links nicely with AI stuff for the speculation. Really, really cool to the players we've got, dude. And we'll see how it it rocks and rolls with China. But as far as I'm concerned, dude, as long as the models get better over time, that's all that I really care about. The cost comes down, which is awesome for everybody. But we even are now seeing Nick this difference between if you are that's what I would call like a regular user of >> uh AI, you're not accessing the best and brightest models unless you're paying a subscription right now. And I would say the difference is probably noticeable in the in the output that you're going to get from that.
Yeah, it's like the uncanny valley, right? Like a human face can look like 97% of a human face, but it'll be terrifying. You know, if it's like 80% of a human face, as in it looks like a cartoon or something, people like, "Oh, that's that's cool. That's funny." And then as you approach like the closer and closer and closer to human, like the qual, you know, people's sensitivity to minor quality changes goes way up. It's the same thing with these models. Like we're growing really sensitive to minor changes in the quality of uh, you know, Opus 4.6 and whatnot. Like people can now vibe check and see like, something about this is off today. And then, you know, it turns out Claude the Claude developers will release a thing being like, ah yes, for exactly 40 minutes on September the whatever, like the quality was off and people are like, yeah, we know, like it was felt it different on it. It's it's why, you know, I saw a creator do a video and this this would be a good segue into our getting the marshmallows and roasted roasting do right now is they said, look, this is a cool model and it automates all my posts. >> And he was like sharing it. You see all this post on I don't write a single one. He didn't say like this, but the implication was I don't write any of the slop you've been swallowing for the past 12 months. Um, right? And then I saw a comment that was like, oh, you don't actually write that stuff. I'm like unsubscribing. That's like it was a really good point actually that if someone knew you didn't really write any of the stuff you'd been saying, like if it came like, you know, we both got I think what I'd like to think wonderful communities, something that we we really care care a lot about our people and if it came to light that we hadn't actually written a single DM, we hadn't actually written a single post, that would really feel weird and it'd be like, oh, okay, that's a bit low effort and lame. Like I get that you can automate it, but it it it just has a tacky feeling to it. Um, LinkedIn, I think, is its own cesspit of craziness. So, we'll put that one over there just for a second. But there is something about it being written by you and read by you. Like, one comment I get back is, "Is this an automation or Jack?" And I'm like, "Well, you'll never you'll never know the difference." But, and you know what I'm saying? It's like, but it was just interesting seeing the reaction from the crowd being like, "I don't know. That just feels weird to me that you've not written that even though you can."
Yeah. Yeah. Yeah. Of course. I think uh people they treat it the same way when it turns into that somebody else is writing the posts. That's why like ghost writers and stuff like that are typically under strong obligations.
Yeah, man. It's it's silly because if you think about it like the ability to write is no longer the constraint, right?
The constraint is like the human aspect of the writing logically. So in supply demand, like the fact that you wrote it is now like the high good economically speaking and so like you should do it. People attribute significantly more value to the fact that you will have written something than an AI if it's 100x easier for an AI to do it. They'll be like, "Wow, Jack actually spent the time." Which takes me to like I've noticed so much AI commenting and [ __ ] in my communities recently. Like AI posting. We do these accountability updates. Kind of kills me. And don't ask me why I had the sunglasses on. I have no idea. I was waiting to roast and then I was like, I'm going to need put take you put them on so you can take them off when we're roasting.
We go my gosh. Like.
That's good. I I actually always joke that glasses are a sick prop cuz.
They're a conversation and they're like, "That sounds great." And then the glasses come on and it's the conversation's dead.
So.
Give you give you a little extra.
No, I feel you.
A little bit of pep. I like it.
But like uh, you know, we're doing these daily accountability updates where people are genuinely reviewing what they did in the day and then, you know, like helping reflect and stuff. And like probably like a third of them are now written with AI. And every time I see somebody that like very obviously wrote it with AI, I just send them a DM and I'm like, "Hey man, you know the whole point about like reviewing your day is that like you review the day, right?" Um, people are like, "Oh, you know, I fed all my sales call transcripts into Claude and, you know, here's what it said." And I'm like, I mean, like, yeah, fair enough. But, you know, like a whole one of the big deals here is like that like you're actually reviewing your own sales call, right? Like that's how you that's how you improve. You look at yourself and go like, holy [ __ ] did I just say that? Like, oh my god, I said like 4,000 times in that sense. Like that sort of thing.
Yeah. No, 100%.
Is actually the whole point. So, same thing with like engaging with your audience, you know? I'm like responding to YouTube comments for probably like 30 minutes a day now and people are like, "Oh, like why don't you just have AI do it all?" And I'm like, I feel like if there's even a 1% difference, people will notice and the impact will be much more than 1%. It'll be like a,000%. So, I'd rather just do it all myself, even if it's way slower.
I maintain that.
I I love that about you, man. I think it's so true. It's YouTube comments is one I'm getting back on this week. It's It's the value isn't just a task that needs to be done. It's you get to you you have the privilege of interacting with your audience is what you have. You get to respond to them, but it's also you get to learn what they're talking about and what they happen about and oh my gosh, this link isn't working and what the hell is going on with this. So, you get to learn a lot from that. So, it's knowing what to automate and what to set aside is actually also the point. And I think that's probably a valued death many people are falling into. They are it's what can the question I ask is what can I automate with AI? not what what should I automate with AI and that's why everything's just getting thrown into the the slop grinder. By the way, if you are doing outreach like that, you need to you can program people are programmatically now introducing errors >> so it doesn't look like it's not about what does it not include, it's like what errors does it include? So I'll add, you want to add a few extra? You if it's Nick, add a few ks at the end of it, you know, maybe a space and a comma, a few spelling mistakes, but not in a way that looks like it's part of the system. It needs to be completely randomized otherwise too obvious.
I mean, that's actually my recommendation when doing any sort of ad personalization effort campaigns for people in Maker School. Uh, or like when they run it for their clients, like, "Why isn't this working?" And I'm like, "Sounds too good. Have you tried making it sound worse?"
It's like, >> "Hello, Jack. So great to see that you're also in school, blah blah blah blah." It's like, "No, man." Like, "Yo, yo, Jack.
What's up? What's up?" You know, like silly [ __ ] like that.
Somebody sent me an email saying, "Jack, were you in the Epstein file?" I was like >> and I open what the hell or oh, did I see you with the Epstein file? And I open and it was like, just kidding. Actually, what I'm emailing about is blah blah blah. >> Like, okay, it was funny. Like it got it got got your attention. But there were a few of them that dude, there's like we should do one of these episodes, dude, like this email is not unique. And it's like the emails everyone sends. Like I don't know, there's got to be some video somewhere that says send this template because there's seven of them that I see all the time like everywhere I go. And we need to pull the news alert out saying, "Guys, these ones don't work anymore." Or at least they're going to be very ineffectual. Oh.
They're straight up. Probably a big chunk of them are from my community. Dude, we could roast cold emails too. Oh, man. One of my favorite things to do is just to like scroll through all my cold emails and then just look at how terrible they are. They're awful.
Dude, that's hilarious. You know, we could also do we also do a challenge, by the way. Like, wouldn't it be cool if we both picked like >> a really cool AI company and we get one we get one email, >> right? And we both got to write this like secret email and we got to see like and then we we can check next week like did we actually get anything back or something like we we can think of something cool to do like that. I think that'd be really sick.
Yeah, that's not a bad idea. That's not a bad idea. Okay, cool. Cool. I'm game to do all that stuff. I love doing like these roasts. It's also just like to be clear, free content for us because we don't have to come up with like all the ideas and whatnot. Okay, I got my roasting glasses on. Jack, you want me to open up?
I'm dude. I've got my I've got my fork, dude. I've got my hands clasped. I'm ready. Well, just so we know, I um I actually have a filter for this because, you know, as somebody that does a lot of cold emailing, uh one of the best ways to get high quality cold emails is just to copy them from people that are sending you emails. So, what I do is I actually have um an AI model that goes through all incoming emails and then sees if they're solicitation and if they're solicitation, it just dumps them into this label called solicitation. So, I have a bunch of solicitation right here.
That's hilarious. So I can actually see like Nick Sarai of negative reviews. You know, that's actually not bad as like a title.
It's not. It's good. That would get you to open it.
Loved your fast cash strategy. Here's an example of something that I think is too AIness. Hey Nick, space capital U your whole and then like perfect quotes warmth ladder idea made me look at my own contacts like wow, half these people probably forgot I exist. Like when I look at that, I'm like, I don't know. >> Kind of feel. Look at this one. Hey Nick, refame scale. Just a quick vid on how to 3x your view reviews and clicks in 2026 despite the fact that there's no personalization there. I'm actually like, I mean, I might click on that. Look, there's images. Do you know what I mean? It's it's gotten something cool. Another little one I've noticed, bro, as well is profile photo does make a difference. If you've got your face with a color behind you, I have noticed. I was like, "Oh, I'm more I'm more compelled to click on that than I am if it's just nothing."
Agreed. Agreed. About your ex account. That sucks. Do you know the stuff that I actually um like tend to respond to? Just silly [ __ ] that's like, you know, I don't know, Nick. Hey, like this is so bad it might actually be good. That that is so funny that it actually is the worse it gets, the better it is.
Like I give you example. I had an email that had no spaces. It was actually eight lines from top to bottom with no indentations. I actually ended up reading it all. Seriously, because it's so obviously not written by AI at all.
Agreed. Agreed. Um, the stuff that's like you just talked about spaces, right? So, I open this up. The stuff that has like the perfect, >> you know, like copyrighting spaces and stuff, that doesn't work. Also, just check this out. Hey, Nick, I'm anakette. It's like, I know it says it right up here, buddy. You don't have to waste my time with that. Pardon me for reaching out this early in a Sunday morning. I mean, like, this can be endearing, >> but I don't know. Like to me. And then check this out. I noticed you are the capital F founder. Man, the second you you see this, you're like, "Oh, he's just scraping my job title."
Yeah.
What a waste. Imagine if he's instead like, "Hey, Nick. Um, big fan of left click." And then he misspelled left click or something. Like I'd be like, "Oh [ __ ] this is a real dude." You know, >> but instead it's just too perfect. Anyway. Anyway,
Silly.
I agree with that. I I think that's really well said.
Silly stuff. Okay. Like I figured we just I don't know. I figured we just go through this, you know, take a look.
Yeah. And And while you're a little scrolly rolly right there, dude, I want to just call like I did have one guy that reached out to me. He's like, "Dude, I really want to work for you and I want to work for free."
Um, the problem with this and we we spoke about this in one of the last live streams we did is that you don't account for the cost to the individual. Like your free isn't free, right? Because we have to train you and give and I gave this guy a shot. I was like, "Dude, okay, find some way to add value and if it is valuable, we'll talk."
Um, and he he ended up not doing that. He ended up just not really giving me any value, but I thought I'd give him a shot. But I think this is the whole point for people with this call, any kind of outreach is you really if if you had to tell the R of it, it's like think about it from their point of view. It's always what's in it for them. Like that's that's with everything. It's always what's in it for the person you're reaching out to. And the more time you spend thinking about that, the better generally you'll do. Are we looking at some data here, Nick? This looks interesting.
Yeah. Um, the best emails I've ever gotten are always like, "Hey, man. Just edited a video for you. Here you go. If you like it, I'll do it again or something." And it's like all lowercase and it's like usually some young ambitious person that's manually doing it all. And they just like do the value up front. And like I'll always I'll always look at that and then I'll say, you know, for sure, man, like let me look at this. Let me actually like give you some feedback or whatnot. And like a lot of these guys I end up working with. Like some dude just sent me 30 YouTube titles and like two of them were good. And then I was like, "Hey man, can you make me 200 more? I'll pay you like $300." And he's like, "Yes, yes." And then we did it and then we worked together.
That's so cool, dude. I love that. I wonder if you reached out and said, "Dude, I suck at so many things, but I'm kind of good at this and I'm I've got nothing but time on my hands on." Yeah. It's like almost like just real dead ass, real authentic stuff with like actually being valuable. Dude, you know what's funny with LinkedIn? I I as a user, I don't use the platform because I think so much of it is >> um everything's networking up. Everything's I think everything's become designed now to get engagement on the post. It's I I don't find there's too much value exchange, at least from my my point of view, on these platforms, dude, clicking on it. And I I don't leave LinkedIn feeling nourished mentally. You wouldn't um provide your email address for a for a text file.
Well, for MD for MD. Is this is this the new replacement for the node porn?
I think so. I mean, I was looking for node porn and I wasn't seeing it, but basically um just so we're all aware, node porn is where, you know, some fella or chick on LinkedIn or even another social media platform is like, "This Nad AI agent structure saves me 4,000 hours a month. Here's how." And there's like 4 trillion nodes and it's like obviously not like maintainable. It's like obviously ridiculous, right?
So that was a real big thing um up until pretty recently. I'm sure we could still find some if we continue scrolling through, but um yeah, I think this is now replacing it. It's just like Claude.md files, which anybody could immediately generate in like two seconds just by asking their agent.
It's hilarious. We we had that era of complexity equals value. Make me a good Claude.md file. Please make it like super good. Okay. So, I just asked Claude to make me a good Claude.md file. It's going to go through. It's going to do a little bit of research, find some find some conversations and chats, and then it's going to, you know, ideally it's going to make something that's probably better than this. I mean, we'll see. We don't know. But, uh, that's how easy it is to do this. You don't actually need to give your email address to any company these days.
No. Well, dude, one one thing I can do actually, if you want to go on to you, Do you want to go on to YouTube and check one thing out? Um, I want to get your I want to get your thoughts uh on a little bit of a mini an interesting direction for where content is heading on YouTube. Unless there's something we want to double click into in.
I mean, you know, hit me and then I'll type it in so that we don't see um that I just spent like 45 minutes watching League of Legends stream yesterday.
Oh, dude. Just hit me, you know. Let me get a link. Um.
The YouTube explore page is a window into the soul.
Okay. Go to a go to a YouTube channel called James Blue.
James Blue YouTube.
Yeah.
This fell right here.
Yeah. Okay. So, let's look at this. Okay. So, look at the bio. I'm James Blue, a 40-year-old CEO using AI to get better results faster.
Okay. So.
This account is 100% AI generated. So, yeah, hold on. I got to turn this off because y'all can't hear this anyway. Oh, holy [ __ ]. Yeah, it is. Oh, that's wild. Really? Okay, hold on. I have this little extension which lets me play things in really slow-mo. That's wild, dude. That's not a human being. This whole channel with 18,000 subscribers is like fully AI.
100%. Oh, dude. This is the spin-off channel. There's another channel that's got even more than this. If you type in Parker Prompts or you go to Parker Prompts YouTube. Yeah. Okay. So, whoa. I think this is the same person. The DNA of the channels are are the same. Media.group. So, yeah. Yeah, it's the same. Yeah. And also the way they put at in brackets is is it right. But Park has had some real big hits. But everything about it is the same, dude. Even the background's the same. So, you know what they probably do? They probably generated a picture of a dude or stole a picture from a dude. Probably more likely they generated with the background. Then they pumped this into one of those like uh image to video platforms.
100%. Yeah. Yeah. Well, well, very. So, basically what they've done is they've got some kind of production pipeline. The other one, okay, so this is number two and then let me show you number three. This one's going to be probably the most controversial yet. If you type in a guy called Yuri Ai uh J or why. Uh, let me find actually, um, the name, um. Dude, my YouTube has been, um, not pepping popping right now. Okay, I'm going to get this guy's. I don't I don't know what this means. Okay, it's Y O U R I Y U R I. Okay. Y like a. Do not type Yuri with a J. Controversial things pop up. Ever again. Okay. So search this up. Click on any one of these videos. The one. Well, I've actually seen a couple of these before, it looks like. So why don't I watch, um, this one here? Okay. Same thing. Hold on a second. This looks like good though, man. You sure? It looks good. One bazillion percent. Yeah. It is different though. Like notice how like the quality of the uh where the hell is the video I was just watching? Okay. Well, anyway, I don't know. Like the video quality.
So, I can actually prove this if you go on his channel, bro, and go back one year. Okay. Uh, so this is like before I guess he started doing this stuff. Before 12 months. Yeah. Um, like this one here. I mean, yeah, absolutely. This is first video. Oh, no, that's not it. You have to go back even. Is that his old? Oh, sorry. Four years. Maybe something like a year ago. But you'll see like it's not, um, yeah, anything like two years ago. Okay, there you go. Oh, he looks different. That's for sure. Okay, let's do this one. Best AI video generator on YouTube. There's a lot. Oh, here we go. There you go. There you go. That's definitely a real dude, right?
So, this is what's. So, I'll be honest. I when you're looking for it, it's very, of course, it's AI. This is done. This is one of the best examples that I've seen of this. So there's some workflow that you could probably piece together after like 5 minutes and anti-gravity or something. But these high production channels, bro, are getting views. And my theory on them is I think they're sponsorship warehouses. Wow. This is like some CSI Miami levels of investigative research here, dude. This is wild. I mean, like this guy does make videos on AI influencers, so I could absolutely see this. He he like started about with AI, went to AI influencers, made such good AI influencers, he just was like, why don't I do that for me? And then turned it into a pipeline. Yeah. Well, think about it. If you have that pipeline for yourself, why would you not spin up multiple channels? But the point is that these channels are are getting views. Another one I'll throw up there as well is if you click on a channel called uh what is her name? She did she did the big the seven days that ruined Claudebot. Uh, she's like a really big Julie McCoy. Julie McCoy. Julia McCoy. Huh. Okay. Julia McCoy, internet personality. This is AI too. Oh, dude. Yeah. So check out. Dude, she is getting the views with a capital V, dude. Okay. Look at the. Where's the one that was like the whole? She you can see she's put out a lot of content. Like even the one that did crazy. If you just scroll down a little bit, you're going to see a little Claudebot video. Oh, there you go. Um, fooled everyone. There you go. 144K views. Wow. Nice. And she's publishing every single day, once a day. Dude. Yeah. This is all AI. This is all AI percent.
I mean, there's like human editors probably, right? Like. So, this is the catch. I think it's like you've got to have you've got to have good editing with it. Dude. I mean, like, if I were making one of these channels, I would definitely choose like a relatively nerdy style girl in a Star Trek shirt, of course, because I mean, obviously I know my audience, right? 100%, dude. 100%. They You probably go female because there's not that many girls doing it. Like, for sure. That's terrifying. Wow. Well, we're, uh, we're being replaced, man. She's got almost as many subscribers as me. The point, dude, you got so many subscribers, dude. The The point is that this is coming and people are consuming the content. Yeah. They are. They just are. They're watching it. I don't know who, but they're watching it, Nick. They're They're definitely watching it. And these are. They're saying you, Huh. Fascinating. To me, it's almost like this is just like a it's like a TV show that people are watching and then it's mostly about the commenting. Is she replying? I don't know. I think this is you're blowing my mind here. I wasn't expecting this today. Thank you.
So, dude, I know the thing about the site podcast, dude. Who knows what turn we'll take? You have to show up to watch. You have to know what direction we're going in. The point people are watching this stuff, dude. And uh, it's very interesting. So, I think this is just something to have on everyone's radar that this is going to. Once you lock down this stuff, like what's the limiting factor just to pumping out a video or two videos a day? Yeah. Some channels are doing five videos a day, Nick. Yeah. And you're going to click on one of them because not they don't all have faces on it so you can't tell. Yeah. The share of synthetic content to human content is ballooning and pretty soon there'll be way more synthetic content than human content. Just a just a matter of a few months now if it hasn't already happened. Um, production requirements are so much lower. Just how it how it's always been. I remain convinced that like human stuff is the scarf's good. I'm sure there are ways that you could use AI to facilitate like more content production and whatnot. But yeah, like I think there's there's going to end up having to be some sort of verification layer that people are real. I feel like they'll have to be. I don't know what that verification layer is on YouTube. You'll have to like submit your birth certificate or some [ __ ]. I I don't know. Yeah. It's like the the the way this is positioned is, hey, the real Jack is out working hard for his clients right now. He can't be here. He's out working like a like a maniac for his clients to get his clients six figure results every day. So that's like the positioning. That's why he's so busy he couldn't be here. And then it's also like, you know, and he's like, but dude, what he is doing, dude, he's in every comment right now. He writes the script. He read. It's just digital Jack says it. Is personally and he's like footage of me like over the shoulder like no, no, no, change that, change that kind of thing. And then so that's the way that it's possession so it's more palatable so you'll take your AI slap with your P's and Q's and. That's insanity. That is so insanity. Jesus.
Should we do some comments? Yeah, let's do. Let's do. Let's do some comments. Jesus Christ, Jack, you're blowing my mind. You want to do the comments? I got it open here if you want me to screen share, but I've also screen shared everything so far. Ah, dude, yeah, let me, um, pull up is stacked. One second. Sounds good, buddy. Dude, there's so much stuff we didn't even get on to. I was like, "No, that's true." But like that AI stuff, I'm going to be sleeping about that tonight because everybody's always like, "Is this AI? Are you AI? Are you an AI avatar?" And I was like, "No, man. Like, can't you tell my pimples?" Like, "What do I? I'm not going to put those in an AI." Like, what are you crazy? Do you think AI can hobax to this level, dude? That's why I'm hobaxing, man. What do you? They're going to try to look like this. Dude, you're crushing it, dude. With that guy glasses, move that guy had me second guessing, dude. Hang on a second. Let me share my screen. Look like pre blow-up Eminem. That's my goal. That's a That's a great goal, dude. Okay, so I'm actually going to come, um, let's have a look. Where did we get down to? We're pretty far down. You can close the, uh, Riverside tab, by the way. Like that doesn't stop doing loads of things. So look again, I just watched a few interviews of the creative claw Peter and comparing his attitude and vibe with you guys. You both feel like a nightmare to me and he doesn't have any links to make money, no affiliates to buy my course, no give me an email, no this, no that. Claude is just brilliant and it will only get better. By the way, Peter does not even have a photo with Alexi. Well, then obviously how legit can he really be learning gay if he's not about that? True that. I mean, like Peter, I want you guys to know that there's a lot of ways to like make positive outcomes from publicity that aren't directly, you know, like financially related. In Peter's case, he segued that into working with OpenAI. So, I mean, like I wonder if if this commenter would be saying the same thing had Peter already gotten that position at OpenAI when that comment was made, you know, like now he works at OpenAI, there's obviously like a lot of financial incentive there. I'm sure there was like a lot of there was some buying out. Um, but, you know, like fair enough. Um, I I understand where this brisk is coming from. Yeah, we we love Peter, by the way. And then I think this is, um, yeah, he's he's great. He's kicking button. Holy mother of clickbaits. I wonder if Nick thought, "Is this too much?" and was like, "Nah, I'll put." Yeah, that was an editorial mishap that we spotted and eliminated very quickly. It was to take that down. Yeah. Yeah, definitely. Neutralite supplements is the best. I'm sure they'll be doing this years. Yep. I agree. Short form content is useful if your algorithm is education and life advice. Maximizes frequency of ideas and sharing of ideas. The most important thing we can do right now. Um, that's interesting. Sharing ideas, sharing human ideas. Seeds looking for a place to germinate. Yes, exactly. It's an interesting concept. Um, but I I think you could probably get better outcomes by being a sniper and picking the videos that you want to have a look at. Um, but I I do think that's funny. I'm competing against the most prolific Nick fan out there. I know he likes Julian Shapiro. Did you? Does this make sense to you, Nick? Big fan of Julian Shapiro. Yeah, I had a call with him a while back. He's a cool guy. I really like Julian Shapiro. He's, um, yeah, I was like part of his marketing cohort thing and he was going to hire me on for a bit, but he has this awesome masterclass article on like how to write good. It was published like four or five years ago back when he was doing this big SEO run and really changed the way that I look at writing. So, I use a lot of his tips and tricks and stuff like that when doing copywriting for my own campaigns. That is a very niche fact. I must have written about that in some random blog post like half a decade ago, dude. Well, shout out Julian. I'll check I'll check him out, man. Thank you so much for the re recommend. We've got this guy here is asking about do we think privacy even exists anymore? Convenience versus privacy. Privacy is dead. No, I feel like it's just the same thing it's always been, right? You're trading off privacy for convenience. And, um, that train is not stopping anytime soon until we're all just one big orked Borg. Agreed. Agreed. I think it's also like the whole do you want to just accept the cookies? If not, spend 2 minutes live clicking. No. Yeah. Accept the cookies and move on basically is the is the mentality. Um, Anthropic just reported that AI responds at the level of the user question. So six-year-old, six-year-old responds. I didn't know that. That is actually very interesting. You ask it like a PhD, you get PhD level response, but it stays, but to me the basics stay the same. How does the agency customer know how to get the requirements clear and formulate the question so that the agents can run with them? I'm not sure I follow the spirit of the question here. I think Karine's point is, you know, how's your customer going to know to do this? I think communication skills right now are a big unlock. Being able to write stuff in clear written English is extremely powerful. And as a prompt engineer, agency, AI agency owner, whatever the hell you want to call yourself, probably like 50% of your job over the next few years is just going to be that. You'll have business owners that just can't communicate their requirements and no amount of things will will help them one-shot it. You know what I mean? So, there's some value there. No, it's a good point. It's like you can have the most capable models, but if they don't know what you want, it's pointless. Well, um, Nick, should we do one one final question? Why a prediction market? Okay, probably not one. We'll go over the prediction markets one maybe. Um, this is cool. This is a good one. Yeah. Hey dude, bros, where do you think traditional coders and codings sit in the AI revolution? And financially wise, was beneficial to be a coder? There's any benefit to anyone being a programmer anymore. So Nick, where do traditional coders sit in the AI revolution? Well, I think they're probably the best equipped to communicate, you know, with these models at least in the near term. And so I think there's actually quite an advantage there. You understanding which framework to pick or whatever is actually like pretty high leverage right now. If you're a total newbie and you have no idea and you walk in like your Claude is going to spend four hours exploring some [ __ ] solution that doesn't really make sense, you'll be able to spot it immediately. So yeah, in the short term you have like a lot of a lot of potential I would say. And, uh, there's no surprise why all of the people that are like making the big repos and making like the big changes right now, um, you know, they're like programmers and stuff like that, right? They're not people with like zero programming experience. They're people with with programming experience. But, you know, over the over the coming months and years, like eventually software engineering I think will fully be natural language and at that point it is more about a communication, um, skill than it is about your programming skills. When, um, when I when I left law school, they told me that pretty much everything you've learned now is irrelevant, um, because it's out of date the second you leave. But the point, the point of the law degree wasn't what you know. It wasn't what you think. It was you learning how to think. So I think the biggest value of program is the same, which is the way you can approach and solve problems and think about things is more valuable than you know, knowing what the how the hell to write in HTML. That used to be super valuable, not long ago. Now it's not that valuable. So I think it's more with the way you approach approach programs. Um, and I think I know you've said the same, Nick, but probably the most valuable thing that if I look back over the past X years of vibe coding and is really the skills of how to solve it is probably the most valuable stuff and worth a look rather than knowing syntax. I think like product managers, um, product leads, I think these people are probably the best equipped because their whole job is translating natural language into business requirements. And so I think they're like the best equipped over the next like couple of years to make big splashes and, um, you know, build really cool things simply because they no longer need the developers to be their fingers, right? They could just do the do the design thinking in their head, lay it out, and have the model act on it. But yeah, I think programmers are still valuable and will remain valuable. Not necessarily just because of their ability to program, but just because in order to learn how to program, one needs to learn how to think. Just like you need to learn how to think if you endeavor to be a great writer as well. Beautifully said. So programmers, it's not game over yet, but it might be stone. So if you want to get ahead, dude, you need to be staying stacked right now. You need to be stacked. That's it, dude. Where else are you going to get such riveting commentary on the future of LinkedIn than on this podcast? All 17 of you watching right now. Um, beautiful. What another great episode, Jens, gals. Um, we will catch you on the next one. Have a beautiful and blessed week and we'll we'll catch you there. Yeah. Catch y'all later. See you.