GPT 4.5 Comparison, Where's the Moat?, RIP Skype, Sulek Earns Pro Card, Cursor for Figma

You're watching TVPN live from the temple of technology, the fortress of finance. That's right, the capital of capital. Today is Friday, February 28th. And we gotta say happy birthday to our vice president, Ben. He's back in the studio.

Ben.

They tried. The enemies tried to take him down.

They could not.

They couldn't do it. He's too powerful. But we got a great show for you today. We're covering GPT 4.5. We're going to answer the question, what do you get when you spend 10 times as much money pre training and LLM? There's been a bunch of conversation on the timeline about this. We're going to break it all down. Also, Skype is shutting down. I'm sure we'll have to do a little moment of silence for that. Tell some stories about our early Skype usage days, give you a little background on the company and how they wound up there. And then the timeline has just been in turmoil. So we're going to be spending a ton of time on the timeline, to say the least. Yeah, it's been spicy.

Say the least this.

This morning. There's been drama and there's been tons of fun posts, as always, so. So stay tuned. We got a great show for you today. Jordy, how you feeling?

I am feeling good. I'm a little frustrated because Ben didn't tell us that it was his birthday until about five minutes ago. And you know how people are. You know, they try to be nonchalant. You know, they'll wait till it's. It's. It's almost. You know, we got an hour till it's noon.

Yeah.

And he just drops. Oh, yeah. It's actually my birthday.

Yeah, yeah, yeah, yeah.

Cool, cool. You're so nonchalant. No, tell me. Yep, tell. He's. Ben's gonna mute me. But no, this goes out to all the nonchalant birthday boys out there. Tell us. Tell us it's your birthday. Yeah, A week, a month before, a week before, a day before, in the morning of.

Yeah. And I want to know so I can set up Singing Telegram Mariachis. We were at breakfast with Ben earlier today. I should have had the whole staff come out and sing him Happy birthday.

Yeah, I would have done that. Pancakes with candles.

Exactly. Pancakes with candles. But he robbed us of that opportunity anyway.

And I will never forgive you for that, Ben.

Anyway, Jordy, would you like to hear a joke?

I love your jokes.

You love my jokes, but these aren't my jokes.

You didn't write this joke.

I didn't write this joke. This is what this is, is a joke written by a model that cost $5 trillion.

Actually, we should get into the actual cost. But for those that are maybe listening for the first time, this is my. John has this thing. This is John's eval. Yeah, your Coogan's eval. And we should. We should actually. You got a coin?

We got to coin this.

So Coogan's eval is where you take the new model and you say, tell me a funny joke. Make sure it's actually a funny joke, like a stand up set and not just a knock knock joke or something simple.

Literally.

My favorite comedians are Shane Gillis, Dave Chappelle. It's. And so why don't you read out the joke and buckle up?

I mean, you're gonna be in tears. We should get some tissues because you're gonna be laughing so hard. So. Hey, you ever notice how technology makes us think we're way smarter than we actually are? Like, I realized this when my wifi went out the other day. I immediately panic because without Google, I'm basically just an ape holding an iPhone. Seriously, if you took me back to the middle ages and asked me how to build a simple pulley system, I'd. You got YouTube here, everybody. AGI achieved. AGI achieved.

Hang it up, AGI.

Everyone's out of a job. It's incredible. Amazing. Of course, that's not the goal of these models. Their job is to work in the mines.

Are you going to.

Oh, you want me to keep going? It's so bad. Okay, so. And it's not just the Internet. Remember when we thought Alexa and Siri were going to make our lives easier? Now it's just me arguing with a speaker speaker about whether I said play Kanye or order pad Thai. My credit card's maxed out, but, hey, my Spotify playlist is fire.

I like how it just throws in this completely random non sequitur.

Yeah, I don't get. Why is it maxed out? Because he's paying for, like, AI tools.

Ordering pad Thai.

I don't know. Speaking of tech, I think the weirdest part about AI isn't that it's smarter than us. It's that it's becoming more human. We're programmed. We programmed AI to mimic humans. And now AI lies, makes mistakes, and blames other people. We invented digital middle management. Congratulations, humanity. We've automated incompetence. Wow. A real knee slapper, Right?

Perfect.

I mean, I did that. Was that stolen from Shane Gillis or Dave Chappelle?

I Think what we need to do is feed your lines back into the model and ask it to judge your delivery. Because it really might come down to the delivery.

Maybe I'm setting it up.

You're setting it up to fail. You don't want to believe that the model can be as funny as you are, but anyways, we're not. We're not. We don't have AGI yet.

We're not out of a job yet. So stay tuned. But I mean that realistically, that's not why, like, they're not even trying to make these models that funny. It is eval. I think it's a good eval. I think that they should be able to do that at some point. And it, and, and the humor eval is a way of asking, you know, the model, like, can you do something that you haven't been trained on? Right.

Yeah.

The move 37 of writing is a joke, a novel joke.

In my mind, the real benchmark is a true knee slapper.

I think so. I think so. But anyway, we got a much more serious analysis from Andre Karpathy. He says GPT 4.5 and an interactive comparison. Today marks the release of GPT 4.5 by OpenAI. I've been looking forward to this for two years, ever since GPT 4 was released, because this release offers a qualitative measurement of the slope of improvement you get out of scaling, pre training. Compute. This is something we've talked about. Yeah, I'm super scale pilled. Scaling is real. But the question is, if you scale the model up 10x, do you get a 10x improvement? Do you get a 1x 2x improvement?

I don't know.

If you get 10%, do you get 100x like sometimes, you know. And he talks about the evolution here. So each 0.5 in the version is roughly 10x pre training, compute. Now recall that GPT1 barely generates coherent text. GPT2 was a confused toy. GPT 2.5 was skipped straight into GPT3, which was even more interesting. Then GPT 3.5 crossed the threshold where it was enough to actually ship as a product and sparked OpenAI's ChatGPT moment. And if you remember, when GPT 3.5 was getting really good, there was this da Vinci model. Is GPT 3.5002 dash DaVinci? A great name, of course, very memorable. But that was the moment when people in the developer community were toying with it and being like, okay, this thing is actually like passing the Turing test. Like, it sounds like human and it is remarkable. It's remarkable innovation. And so GPT4 in turn also felt better. But I'll say that it definitely felt more subtle. I remember being part of a hackathon trying to find concrete prompts where GPT4 outperformed 3.5. They definitely existed, but clear and but clear and concrete slam dunk examples were difficult to find. It's that everything was just a little bit better, but in a diffuse way. The word choice was a bit more creative under understanding of nuance in the prompt was improved. Analogies made a bit more sense, the model was a little bit funnier. World knowledge and understanding was improved at the edges of rare domains. Hallucinations were a bit less frequent. The vibes were just a bit better. It felt like the water that rises all boats where everything gets slightly improved by 20%. So it is with that expectation that I went into testing GPT 4.5 which I had access to for a few days and which saw 10x more pre training compute than GPT4 and I feel like once again I'm in the same hackathon two years ago. Everything is a little bit better and it's awesome, but also not exactly in ways that are trivial to point to. Still, it is incredibly interesting and exciting as another qualitative measurement of a certain slope of capability that comes for free just from pre training a bigger model. Keep in mind that GPT 4.5 was only trained with pre training, supervised fine tuning and RLHF. So this is not yet a reasoning model. Therefore this model release does not push forward model capability in cases where reasoning is critical like math and code. In these cases, training with RL reinforcement learning and gaining thinking is incredibly important and works better even if it is on top of an older base model GPT4ish capability or so the state of the art here remains the full O1. Presumably OpenAI will now be looking to further train with reinforcement learning on top of GPT 4.5 model to allow it to think and push model capability in these domains. However, we do we do actually expect to see an improvement in tasks that are not reasoning heavy and I would say those tasks are more EQ as opposed to iq. So it's more it's more just better to chat with and actually have a conversation with related and bottlenecked by world knowledge, creativity, analogy making, general understanding, humor et cetera.

So definitely bottlenecked by humor as I've seen.

So these are the tasks that I was most interested during my vibe checks. So below I thought it'd Be fun to highlight five funny slash amusing prompts these capabilities and organize them into an interactive LM arena light right here on X using a combination of images and polls in a thread. Sadly, X does not allow you to include both an image and a poll in a single post. So I have to alternate posts that give the image and then so basically he put he left it up for eight hours. This got a million views, lots of people voted and he broke down the results and interestingly he says okay, so I didn't super expect the results of the GPT4 versus GPT 4.5 poll from earlier today today, but GPT 4.5 outperformed GPT 4 on only one of the five questions. Wild people prefer GPT 4 in 4 out of 5 questions. Awkward and it's always odd. I mean the splits are not huge. 56% of people preferred GPT 4.5 in one question. The margins were 10 20%. It wasn't like it wasn't like 100%. 90% of people thought it was one way or another.

Still to 6040 for every one person that prefers, you know, GPT 4 or 5 in question 3 two people prefer GPT 4. That is pretty significant.

Yeah. And he says to be honest, I found this a bit surprising as I personally found GPT 4.5 responses to Be better in all cases. But maybe I'm just a high taste tester. I think that's probably true because you built this Andre, and you're like one of the greatest programmers of all time and so you probably have high IQ. Yeah. But the thing to look for is that GPT4 more often says Trump has.

A refined taste in another kind of model.

The thing to look for is that GPT4 more often says stuff that on the face of it looks fine and type checks as making sense. But if you really think about it longer or more carefully, you will often more you will more often catch it saying things that are a bit of an odd thing to say or a little too formulaic, a too basic, a little too cringey or a little too tropey. Slightly reassuring. A number of people noted similar surprise in the replies. I noticed. The few I noticed as an example for the roast, 4.5 is punchier for the story 4.5 narrative jumped in, had dialogue and hinted at a unique storyline. B was a little bit more schematic for the poem. 4.5 is obviously way better. The rhyme scheme and meter of B are so unsophisticated. A has to be 4.5 the voters have poor taste. I like that. So yeah, either high taste testers are noticing the new and unique structure but the low taste ones are overwhelming the poll. All you plebs don't.

Normies.

Yeah you normies. Lovies don't. Can't. If you can't tell, honestly, if you can't tell the difference, if you can't tell the difference between 4.5 and 4, like I don't even know if you should be allowed to be a member of society.

It's interesting. You have to imagine OpenAI is trying to continuously dominate the narrative, right? It's sort of dominant. They're sort of ahead on revenue but very, you know, well beyond.

But it's like a knockout drag out fight in terms of attention and what's.

It's really a war for attention. So they had Grok last week. Was it over the. It was Friday.

It started last week, but then the app rolled out, the voice mode rolled out. The reasoning.

The new Claude Model 3.7 and OpenAI, like there, there's some value in being the last model to drop right. In just terms of, you know, sucking the air out of the room.

And I believe that this, that this model will be the top of the stack in benchmarks and, and OpenAI has always done this cadence where they come out with the best and then a couple months later there's a whole flood of everyone else catching up and all of a sudden they're beating them by like a few percentage and then OpenAI comes back and is back on top. And so if you average out like, like OpenAI is not always on the top of the rankings on every single benchmark but they probably have the most number of days on top because they're on top like you know, for about half the year and then towards the back half of the year everyone else is kind of chipping away at their leads and different benchmarks and stuff. But, but it's fun. So he closes out and says so we'll wait for the larger.

Also this feeling, it's hard to know if it's accurate, but there's a general feeling that I have, and I'm sure others do that that, that the other foundation model companies are launching stuff as fast as they possibly can and OpenAI is maybe sort of timing things more and that, who knows?

You know, I kind of disagree with that. I mean, I think that, I mean the narrative is like anthropic has some insane voice model that they haven't released for safety reasons. And I think if anything all The GROK push and stuff. That's really pushing the cadence at OpenAI.

I'm not saying they're not feeling the pressure. I'm just saying they seem to be. They seem to be keep. You know, you can imagine if they wanted to launch GPT5.

Yeah.

You know, could do they feel the urgent. They felt clearly an urgency to come out with four or five.

Sure.

Maybe they don't. And they're feeling the pressure to get five out, but there's not. They're sort of like trying to stay just perpetually slightly ahead.

Yep.

And focusing on quality. Whereas GROK was just. We have massive amount of compute. We're just going to brute force spend launch as fast as possible iterate. GROK was clearly not actually ready for production.

Yeah.

It didn't have the reinforcement learning that made it so you would just get these obscene answers that never should have been live. Right. It clearly was more of even. And it was almost like a beta version of the product in many ways. It was responding to stuff that was more of an alpha level product where that stuff should have been caught.

It's so funny to me that the NonProfit board at OpenAI was upset with Sam for launching ChatGPT, the very first version. And they were like, he's moving too fast. He's too dangerous. Can you imagine the nonprofit board watching Elon launch Grok 3? And they're just like, no, no reinforcement learning. No reinforcement learning. It's like giving you gas. Like how to make a bomb stuff. It's generating intellectual property that's complet infringing. Like it has sexy mode. And they're just like, what? Like actually like Sam would be much better.

Yeah. On second.

Bring Sam back. Yeah, yeah. Actually we should have bagged Sam and kept Elon out of the.

Turns out our biggest nonprofit backer was the most degenerate. Yeah, yeah, yeah.

That's hilarious.

Safety pill.

Yeah.

Is just accelerate, accelerate, accelerate. Totally, totally the entire time.

Yeah.

But yeah, to Elon's credit, he doesn't. He, he. He, you know, makes the claim that. That Sam is. Is disingenuous and, and yeah, snaky or whatever. He's not saying that Sam is reckless exactly.

It's actually a very different critique. Yeah. And I, and I think, I think it's good. Like I think that this shows that, you know, just scaling up. You know, like there was definitely a, a drum beat from the AI doomer community that, that GPT 4.5 GPT 5. You scale up the pre training by. By 10x they were so scale pilled that they said when you scale up by 10x, you're going to get a 10x more dangerous model. You're going to get 10x more risk. And that does not feel like what's going on right now at all. It feels like, yeah, these things are still like friendly and nice and they make mistakes and they're just simulating humans. And humans are default, like kind of good. But also sometimes midwits and sometimes they, they, you know, come up with good stuff and you know, it's all about how you implement them. And if somebody uses one of these models and creates a wrapper that does something bad, then yeah, you're going to get a bad outcome. But the model itself is not going to, you know, break out and have a mind of its own and destroy everything. So Andre finishes out by saying one really bad mistake that bugs me is in the GPT 4.4 versus 4.5 conversation. The one generated by 4.5. 4.5 asks. Still buffering your responses like it's dial up Internet. This is really bad because it clearly borrows tropes from early days computing. Huh. Interesting. Anyway, let's go to some of the other takes. Michael Magnano is is this comment is interesting? I think this is about it.

No, the from main reaction there. I don't know if we have the answer, but it says is it possible the 4.0update last month was already a distillation of 4.5? I think the results would be less concerning if so.

Huh? Oh yeah, because maybe, maybe we got some of that already baked in. That's hard to tell.

Such a small, you know, difference.

Yeah, yeah, yeah. It's hard to tell because like if you're just trying to.

And this goes back to people right now. People right now are almost comforted in the fact that you're getting chapters of an LLM, right? Sort of like version 1, 2, etc. And they all have different naming seams but over time you can imagine they're just perpetually updating. But that's uncomfortable to be. You know, humans get smarter or dumber over time, right? Depending on their environment and their viewpoint of the world and knowledge and all that stuff. But it's not necessarily perpetually updating throughout even a day where you wake up and your employee in the morning is brilliant and then they're slightly dumber later in the day nerfed.

I mean the vibes around the cadence going from ChatGPT which launched on 3.5 to GPT 4, everyone saw ChatGPT in 3.5 and we were like, we crossed the Turing test. AI is here. This is going to be massively disruptive. Like there's going to be no more marketing jobs because it can just write copy. Like, like there's going to be mass unemployment. Like people even, even beyond, like the doomers who are saying this is going to kill us. A lot of people were like, this is going to fundamentally change the world very rapidly because it's happening at Internet scale. It's not, it's not an atoms company, it's a bits company. And so this can be rolled out into every single company and layoffs can happen now. Right. And so people were people worried. And then three months later it was like, oh we did, we did. We went from three to four, which feels like a 25%, 33% increase. It just felt like, oh, okay, if we're continuing at this pace, like the acceleration is real. And, and. But then over the last two years we've just been incrementing on fours. 441-4540.

People were starting to call that out. But Sam, if you're out there in the markets trying to raise 40 on 350, 350 billion or whatever the price is with the moss around, you want there to be this sense of hey, there's this sort of impending super intelligence. And you know, every, if you're needing to raise that kind of capital, you need superintelligence to perpetually be right around the corner. Even, even Elon was posting, I think earlier this week. We are on the, it's the eve of super intelligence. You know, whatever. Everybody sort of. If the models were to stagnate completely, there would be a massive pullback in new investment. Because you see the difference between 4 and 4, 5. That's not going to get the incremental, whatever $20 billion of investment.

It's such an interesting dance because if you go too hard and you say super intelligence is right around the corner, well then the regulators are going to come out. People are going to be protesting in the streets. Yeah, the doomers are going to have a bigger argument saying look, you can just look at what Sam is saying. He's going to say it's the most powerful God in a box. Like we have to control this, we have to regulate it. But if you don't say, hey, super intelligence is coming, well then why should we invest a trillion dollars in the capex like we. In order to make these types of investments, it's got to be the real deal. And so finding that Balance is really, really tricky in my opinion.

I think Satya has, has the right balance, which is that AI in its current state and for the near future is this sort of almost like a co pilot thought partner. It doesn't yet have total agency. Right. It's not actually replacing an employee, but it's being a work sort of partner for you. And that still justifies a lot of capex, but I don't think it justifies an incremental $30 billion investment in OpenAI at the moment in time.

Interesting. I think people massively underestimate how much work there is to do.

One point of view, I forget who was talking about this earlier this week, is that there's so much value to be created and captured from the current technology side. It would take a. Even if we had, even if all model development paused and we had a 10 year shot and we just went to implement, to implement the technology, we could still, there would still be numerous.

Numerous, you know, and that's what happened with the dot com boom. Like the Internet rolled out. Dial up was pretty terrible, pretty limited. But even in like I don't know, 2000, what 2004, 2005, people were starting to get high speed Internet, DSL, cable, stuff that you where you could play video games online, you could stream videos. And then there was a question about like okay, well what if your Internet gets 10 times faster? Well like you're not going to watch 10 movies at the same time, like and yes, you could stream 8k video instead of 1080p but most people just are fine with 1080p, still entertaining and valuable. And so, and so like we accelerated the speed and number of connections to the Internet and then it took us 20 years to actually reap that benefit it even though the tech kept getting better and the Internet has gotten faster and now you can download an 8K video very quickly if you have the right line. But most people are just like yeah, I just need, you know, a couple gig download and I'm fine. Yeah, it is interesting. Well, let's move on to some reactions. We got Michael Magnano. He says technology is not a moat, never has been. There are so many examples of incumbents copying startups technologies and leveraging their distribution to crush competition. And AI coding models are quickly making technology even less defensible. With coding models like Claude, assistants like Cursor and no code tools like Lovable and Bolt, it's becoming stupidly easy to copy existing products and apps with virtually no effort. Turns out the stuff that actually matters is what always mattered. Strong products, real network effects, ruthless distribution, and brands people actually care about. The winners will double down on those things, while everyone else wonders why their perfect tech keeps getting cloned into irrelevance.

So the most relevant sort of point related to all of this is that we've talked about this on the show before, this idea of easy come, easy go. And so the issue with Cursor is if you can go to $100 million building a product that was sort of open source, it was like VS code for code. They forked it, get to $100 million in revenue, lovable and bolt come out. You know, lovable goes from 0 to 17 million in three months with a small team. But the same technology that enables them to do that also enables other people to come and compete away their market. Right. So Windsurf AI, which is a Cursor competitor, they have gotten to a $40 million run rate in a very short amount of time. And so historically, if you look at these businesses that. That have this sort of explosive growth, sometimes there's sort of specific market dynamics that allowed that to happen or allowed a specific team to do that, or relationships that enabled it. But oftentimes, if a pure play software company explodes in growth and there's no network effects, which Michael references here, or sort of durable distribution advantage. Right. That value creation can erode very quickly. And so what I think we're all interested to see how this plays out is all these companies that have this explosive revenue growth, one, it's great because when people say, oh, there's this big bubble, you can point and say, well, real value is corrected, but. But real customers are handing over dollars to pay for these products. And, you know, Cursor would have gone public, you know, a year ago if it was 2021. Right?

Yeah.

But what I think we're all waiting to see play out is does this revenue ultimately get competed away by other teams? That there's teams that will use Cursor to compete with cursor. Right. Which is. So we'll see. I think we're generally bullish on these companies in the short to medium term, but unclear how they sort of maintain and continue to create value when people just say, you know, there will be a version of Cursor that allows somebody to vibe code a free version of Cursor. Right. How far? Like, I'm actually surprised we haven't seen somebody say, I use Cursor to rebuild cursor.

And that would definitely go viral. We've seen some people do additional forks and there's been some pushback.

I think there was that yc. Yeah.

And so, yeah, there has been pushback. I wonder, are you a bull or Bear on DocuSign? DocuSign is like the quintessential. You should be able to rebuild that with cursor. Right. It's just forms and filings. But. But maybe those ten, twenty, hundred thousand employees they have are just the best sales guys. They're embedded in every organization. They're calling you on your birthday, they're celebrating you.

They're not like, Ben, they tell you their birthday that morning. They wake up and say, hey, it's your favorite sales rep. Yeah.

You clone DocuSign for, you know, $5 in cursor credits or whatever. You have Devin on the case for the weekend and you got yourself a DocuSign clone. How are you distributing it? You need 10,000 sales guys.

Maybe, Maybe. You know, I just, I don't think we know. I want, you know, the bigger threat to DocuSign is, again, so Google Workspace, I think, just rolled out a Docs document signing product. We already pay for Google Workspace. It's one click away. It's not going to be as good, but for sure, Google's like calendar booking link throttled the growth of Calendly. Yeah, like that. And so I think the bigger threat to some of these sort of commodity products, the thing is, I mean, document signing has the viral sort of effect.

Where you sign up and then you sign up.

But that also enables distinguished Baked in the product. Yeah, distribution is baked in. But I mean, we've said this maybe five times on the show before. How has DocuSign not said, hey, our product's actually pretty simple. We should be able to maintain it with a team of 100 engineers. We actually don't need a thousand. Yeah, let's sort of give back. Let's free these people to go work on other more interesting products.

Set them free.

Set them free.

Instead of slaving away in the. In the docusign minds. Yeah.

And just let them go work on more interesting things and then, you know, give that. Give that cash back to shareholders invested in new products. It just seems like a very uninspiring company. But. But we'll see.

Yeah. Anyway, Scott Woo, CEO of Cognition, chimed in with his take on GPT 4.5. I thought this is an interesting data point because he's obviously using these models very deeply. He says GPT 4.5 has been awesome to work with. Of course, they build this into Devin. You can run Devin on a number of different LLMs. And he says our agentic coding benchmarks. On our agentic coding benchmark, it already shows massive improvements over 0, 1 and 4.0. Excited to see the model's continued trajectory on code. One interesting data point. Although GPT 4.5 and Claude 3.7 Sonnet score similarly on our overall benchmark, we find that GPT 4.5 spikes more heavily on tasks involving architecture and cross system interactions, whereas Claude 3.7 Sonnet spikes more on raw coding and code editing. As AI takes on increasingly complex tasks, we believe that multimodal agents, multimodal agents that incorporate each model's unique strengths will perform best. Just like in your brain, you have different sections of your brain to do different tasks on a team. You have that developer who's really good at architecture and you have that other guy who's just a beast at writing code all night long.

I might be off here with Devin and Cognition's position where as the underlying models get dramatically better, their product gets better and more valuable. That's a great position to be in, given the complexity of building out the sort of team like product, like individual contributor s product that's integrated into a team. Will Cognition be better positioned than a cursor? Where. How long until you can get Devin to build a version of Cursor and just say, hey, fork. Yeah, you know, we might be 10 years away from that, but it's, it's not even that.

I mean, like the mental model I have for, for Cursor versus Devin is like completely separate things. One is like this.

No, totally. I'm just saying.

And with Devin, I almost see it as like, you could see them like a McKinsey almost, where it's like, there's a large company and they're like, hey, we have this app and it runs on iOS and we need to get it to run on Android. And so you just need to take that code and transfer it. And it's just going to be a lot of work and it's got to be done perfectly. And so we just. You. We're going to give you $10 million to do it because it would cost 50 million to do it with, you know, some human outsourcing team and, and cognition just goes and cranks and like gets it done. And then eventually you're doing that all over your organization. And it's much more for like these, these like corporate projects than like, you know.

Yeah. Or backlogs, etc.

Yeah, yeah, exactly. I, I've seen a lot of, of case studies like that, but it's very, it's very different. And then also, yeah, it's, it's very, it's very like it's a B2B product. They are in open access, you can just go install it. But it is expensive and it's a very different sales motion, I think.

Yeah.

Anyway, let's go to Chubby. Hilarious name with the most serious post talking about OpenAI. Mark Chen, who I actually got dinner with a couple months ago, great guy Mark Chen, says, we found a new paradigm through reasoning which we're also scaling. That's right, because GPT4.5 was clearly proven that pre training has definitely reached a wall. GPT5 must be a banger. Last chance, says Chubby. And so a lot of people throwing a little bit of a little question on, you know, hey, everyone is going to be shifting to, oh no, no, no, we don't focus on just bigger, bigger models, we focus on reasoning because like that's the next paradigm. But I think that's fine. I don't care where you're optimizing your AI system, I don't care if it's. All the gains are in pre training, all the gains are in post training, all the gains are in RLHF. When RLHF came out, people were like, oh, so like now like to scale and improve a model you just need to have a bunch of people like writing answers like this is an AI and it's like, I don't care, I just want, just give me good results, makes me laugh, make me laugh.

John just wants to laugh.

But also, also do the research reports and make them great, you know, and, and, and, and you know, throw scale, all of these things. And I think that there's probably going to be these cycles of memes online where people are like, oh, it, like scale doesn't matter at all. It's all about reasoning. Oh, RLHF doesn't matter. It's all about post training or pre training or, you know, fine tuning or a custom style.

I do think it's fair for people like Chubby to feel like they have the right to try to hold OpenAI and these other model companies who have made these sort of massive audacious claims and they've put out a lot of communications on, on this topic to try to say, yeah, last, last chance. You know, GPT5's gotta be a banger. They're basically holding. But it's also funny to be in.

A nod and be like, hey, $300 billion company, last chance, last chance no, no. Who are you to decide whether or happens.

But yeah, there's this balance between the, the. Our. Our. It's. It's more. So. Yeah. Chubby is probably a customer of OpenAI.

Yep.

They're allowed to have strong opinions.

Totally.

They're marketed something to.

Yeah.

But at the same time, Open AI, it's squarely within their right to try to massage the narrative and say, well, actually this is the thing that's more important. Yeah. Yeah. So I think both, both sides have. Have clear.

Well, let's go to another hater. Do you know Gary Marcus?

I do not. Okay.

Gary Marcus is, is a fascinating figure in AI. He wrote a piece called Deep Learning is Hitting a Wall. He, he, he was at. He sold an AI company to Uber, I believe, and his whole thing is, what is it? Symbolic symbol manipulation. So he believes that like, you cannot just throw a big pile of data at a basic algorithm, like the GPT3 algorithm, and then just do predict next word. Predicts next word. Predict next word. And then get God or get the genius human. He maintains that like, humans represent information in their brains in different ways than that. It's not just predict the next word every two seconds. Right.

And so special and unique snowflakes.

He's, he's somewhat right because there are clearly like the reasoning and being able to search the web and use Python and all these different tools. We're building out the machine brain in multiple different sectors. But he's also been like massively wrong for a long time because he was saying like, deep learning will never scale. And it's like, I'm talking to.

He's the Michael Burry of models.

Exactly, exactly.

And so you only got it. You know, if he's right right now, he's going to be pretty.

A lot of people have been like, this guy is a joke. Like, why is he, why is he testifying in front of Congress at the same time? You know, he, I think he does have a PhD. He is like, he's like a serious researcher. He's just taken a very contrarian stance. But so has Noam Chomsky. And Noam Chomsky was a linguist at mit and, and had.

Yeah.

And had a lot of, A lot of theories on the way language is constructed and, and saw and argued that you could not rep. Like you could not represent a human through like a sloppy algorithm. There needed to be some sort of more fundamental truth. And so symbol manipulation basically says, like, you know, we need to have, you know, a concept baked into the model of like, what A table is how it interacts with a chair. All these different things. And so it's been, it's been controversial because there haven't been as many gains in the symbol manipulation world. But Gary's been, you know, happily taking a victory lap I guess over GPT 4.5. Not just purely proving that pre training scale is all you need, but let's find out. So Chubby again says judging by the mood, GPT 4.5 is the first big failure for OpenAI. Too expensive, too little improvement. Improvement and often inferior to GPT4O even in comparison. And creative answers in community tests like what Andre Karpathy posted. And it's interesting because too expensive, don't care at all. They're going to distill this model and get it a thousand times cheaper in two seconds. Too little improvement. That is interesting. But again you have to run the, the, the capex to profit calculation and the GPT4 training run I think was like under a hundred million dollars or something which is still like a lot of money at the time, but $100 million and they're making billions. And so you, if you train $100 million run and then generate billions, you generate billions a year forever potentially if you can be sticky.

Not necessarily super high margin, not yet.

But once you bake it down onto silicon onto a chip.

I'm just saying, I'm just.

Yeah, yeah, like true, like they're not profitable right now. But just the idea of any of these AI companies throwing a billion at or 10 billion, when you think about how much usage there's going to be, it's not as crazy as you think. And so the question is, especially when.

You consider there's models that no one uses that have raised $100 million and have spent most of it with nothing to show for it. Yeah, OpenAI looks like they're allocating fairly efficiently.

And you can say the same thing about like Amazon.com they built out huge data centers 10 times they 10x their capex like 10 times. Right. Did, did the experience of Amazon.com get that much better? No, but they printed money the whole time because they have a great product on top of it. And so like you don't necessarily need a 10x improvement in the core product. If the business is working and it's generating cash, you can underwrite those capex expenses all day. And, and that's what I think a lot of people, that's why a lot of people are jumping into this because Elon is saying look, I'm going to Spend billions of dollars on Grok and Xai. But then I'm going to vend that into Tesla. I was thinking about this, like having a deeply integrated Grok and Xai integration in a Tesla is going to be amazing. There's so many times when I'm, when I'm driving into the office and I'm talking to ChatGPT about, hey, here are some topics that I want to talk about. Can you do a deep research on this one? Just using the voice mode and if that was integrated into a Tesla. So it's basically self driving itself and you're just having a conversation with Grok. Okay, today I gotta, I gotta talk to Jordy about the show. I gotta pay this bill. I gotta to schedule a meeting. And it's got a plan. Yeah, yeah, just cataloging all of this. Oh, okay, you did legs yesterday, hit chest. Here's what you do.

I'm constantly in the car. I'm constantly in the car. You know, I spend enough time on the road, you know, these days because of all the traffic that if I get in the car for 90 minutes and I'm driving, I'm going to have 20 thoughts that I want to be able to execute on in that moment, but I end up having to like wait until I'm at a stoplight. Write it down. Yeah, keep going. And then I've built this huge backlog of tasks where if you can have this sort of omnipresent AI that, that says, hey Grok, can you, you know, call, you know, the cable company? I want to increase my. Yeah, I want to get fiber. I have like, you know, I want to upgrade my plan exactly. Like that's an item you're checking off the to do list.

Exactly. And I'm sure your car, all your cars have some sort of voice recognition. But it's awful, right? It's like, like call John. And it's like, oh, did you want me to, you know, call you a tow truck or whatever? It like always gets it wrong. No one ever uses it. Imagine if it's, if it's, you know, truly frontier, best in class. And so the question is not, does spending a billion dollars on Capex to train a Frontier model get you something that's a thousand times better than the competition? It's just, does it get you a valuable asset that then you can vend in and sell all over the place? And so can Elon sell Grok to X Premium members and make the money back there? Can he sell it into Tesla? Can he sell it into other Things like, like can you hook it up to Neuralink? Like, can he make money from the money he spent? That's, that's the only thing that matters at the end of the day is capitalism. Right. Anyway, so, you know, people are pouring cold water.

Tough timing too. So, yeah, they're pouring cold water on it, calling it the first big failure of OpenAI. Meanwhile, poor Sam had his first kid like last week. It's rough, this guy.

And so Gary Marcus says, a big surprise only to you and the many other OpenAI fans who refuse to listen to me. When I patiently and endlessly explained that this would be exactly what would happen for the rest of us, this was all but inevitable. And so he's taking a victory lap. But it's, but it's odd because it's like, okay, like go do the thing then, Gary. Like go, go do your approach. Like do the symbolic, the symbol manipulation and, or beat them.

Start a podcast.

He might, he might. But yeah, I mean, it's easy to be a pessimist and say that approach isn't working. Like the, like, like the, the starship isn't going to work to get us to Mars. I don't care. Build something better that will get us to Mars. Then if you think the slingshot we were talking about to Delian about this, like if he wants to get there with the space plane, he'll do that.

Like, I don't know. I don't know. I mean that's an obvious critique of Gary, but at the same time, if you are highly researched on a specific topic, you are an expert on that topic in many ways. Gary, you could argue he's an exited AI founder. He's been talking about this for years.

Yep.

He's perfectly entitled to have an opinion on the world changing technology that everybody else. And it's fine if he take, I think he can take a victory lap. We're not, we're not, he's not trying to say that I can do it.

He has been for years. For years. He was saying, he was saying my approach is better than deep learning.

Okay, so he has, that's why.

Because, because he's, he hasn't been advancing the ball on his thing. It would be one thing if it was like they did this and then symbol manipulation. His approach outperforms and it's like, oh my God, he was right. Like we should have, we were going, going down the wrong path. Gary was right. Instead he's just saying like, my, their thing sucks, but my thing really sucks.

Yeah. Yeah, okay.

Which is Kind of like my read on it. But I don't know.

I mean, still, you know, he can't take a vic. He can say, hey, I'm right. And OpenAI won't respond to this, but if they. They could or would or wanted to, they would say, cool, like, let's see your benchmarks.

Exactly. Exactly.

He wouldn't. Sounds like he wouldn't have a good answer.

Yeah, well, now he would say I'm out of the game. I'm not. I'm. I'm working on policy and stuff. I'm not. I'm not really building. I'm not trying to compete in that. But it is, it is a little frustrating because, you know, like, the Deep SEQ team over there, they're making, they're saying, hey, we have a slightly different approach and we're getting. We're getting better performance with less investment. I mean, you know, the actual investments. Debatable. But there were clearly some optimizations there where they, where they got close to the frontier with obviously less than what it cost to train 4.5. And so I'm much more. I'm much more sympathetic to like a deep seek stand saying, like, hey, the, the best investment in, in this technology in the next wave here is on the optimization side, not on the scale side side. I'm receptive to that.

Yeah.

But if you're just saying, like, this isn't working, but also, like, my thing.

Is, okay, anyway, I was right, but I also failed.

Anyway, let's move on. Let's close out with another attempt at humor. Dan Shipper says GPT 4.5 is actually so effing funny. Be me. Be Dan Shipper. Wake up at 5am Journal for 45 minutes about the dream I had last night. Dream was in second person. Probably inspired by a obscure neuroscience paper I read last Tuesday. Brew coffee using beans handpicked by monks on a remote Guatemalan mountain. Sit down to write today's newsletter. Accidentally outline four new business ideas. Resist. Resist the temptation to launch them all by noon. Meditate. Feel mind literally expanding into previously undiscovered cognitive dimensions. Tweet an insight from my meditation session immediately gained 200 new subscribers. Open laptop top to write an article on mental models. Realize the mental, the ultimate mental model, was within me all along. It's only 9am why aren't you laughing?

I mean, what's wrong?

I mean, do you not understand the humor? Did it go over your head, Jordy?

Yeah, it flew over my head.

Because if you're. If you don't. If you can't tell the difference between five and four. You're. You're not really like high iq.

Yeah. I mean, Claude actually got fed this and laughed. Yeah, it actually played a laugh track. That was its only response. So the models are making each other laugh? Yeah, they're not making us laugh just yet.

I do think there's something cool about this. I think personalization.

What if the model is actually sophisticated enough that it knows we'll laugh harder if it just delivers a terrible. Like a terrible joke and just botches it and we go, haha, you dumb model.

It's. Yeah, it's being sneaky behind the scenes, hiding its power level.

Yeah, exactly.

Secretly, it could be the best. It could put every comedian out of business.

Yeah, but it's like playing basketball with a kid.

Exactly. Oh, no, I dropped the ball. I guess you scored. Yeah, that's what it's doing.

Yeah.

Anyway, I'm glad he thought it was funny. I. What is interesting here is that this was personalized. So he fed it in his. His timeline or some more context around him, and so it clearly made jokes that were highly personalized to him. And I think that that is cool. And I think there is something there. And we've talked about this with the GROK stuff. It creates a more unique interaction, unique product. Especially when I think the most relevant, again, won't be humor, it'll be research where it will know. Hey, Jordi, you're an expert in angel investing and ads and entrepreneurship, so I don't need to explain to you what a safe is, but you're not an expert in mining, you know, rare minerals. So I'm gonna break down. Yeah, yeah. I'm gonna explain to you in granular detail. Okay. Lithium ion. Where does that come from? And I'm gonna explain it to you like I'm five. And then I'm gonna go deeper and I'm gonna work with you to understand where your knowledge is and get you up to speed really quickly. I think that's gonna be big a. Amazing. Anyway. Well, one of the companies that I love that's implementing AI better than anybody else is, of course, Ramp.

Better than any.

Go to ramp.com Time is money saved, easy to use, corporate cards, bill payments, accounting, and a whole lot more all in one place. And it is. I mean, obviously I'm joking around, but it is true. RAMP is a perfect place to. To implement AI. They. When you take a picture of the receipt that gets scanned, of course an LLM processes that. And it's gotten so much better. And it's the exact type of organization that you want to be with. Where they're, I mean the cognition guys work there like they have, they have an insane team that can go and look at whatever the frontier model is, swap it in in two seconds. They move very fast, they move very quickly. And so the product just keeps compounding in a way that's, you know, great.

From an invest, you know, know from a company that you would potentially join. There are companies out there that are just sort of boilerplate SaaS that maybe have a lot of revenue today like DocuSign, but presumably over time will, will be more like cable where some people keep paying but you know, they get sort of replaced. These sort of businesses that are more, you know, this is for informational purposes only obviously. But the businesses that are more in these sort of regulated spaces that require payment Rails and banking licenses and stuff like that are super well positioned to implement AI for customers. But not in the same way. You can't just use the version of cursor in 10 years and generate ramp because banks are going to say I'm sorry we work with a smaller number of partners.

It's just great. There's obviously a lot of larger companies, more established companies that have been trying to implement AI McKinsey's printing money. All of the, all of the consultants are going around to every business. Here's how you can use AI. RAMP actually has an obvious narrative here which is, you know, receipt classification and iterating over your financials to get you more accurate data more quickly. Anyway, go check it out.

A couple comments before we jump into the next section. Vishal in the chat says guys react to Zielinski versus Trump. I said said too political for us. Vishal look nasty though and he says makes sense. Jordy, we don't need another all in. Which is a great point. We're trying to hold the line on politics. People you know will ask you to cover politics because they get a lot of attention. But we are a technology and business show. Yeah, we are firmly.

I mean my, my answer on Zelensky vs Trump. I see Trump as the founder of Truth Social more than the president. I mean that's something he does politics but we're interested in him as a consumer software found.

I think they're arguing about the minds.

And so once we, once we hear Zelensky's business acumen then we can comment but in terms of the politics stuff, not really interesting to us.

So Calder says watching John get better at transitions is like watching Tiger woods practices swing. So thank you. Better every day.

Every day.

Eventually you won't know if it's an ad.

Yeah, yeah. It's all blended together.

It's going to be perfectly.

Calder, we'll have to have you on. On the show. And when we have you on the show, we'll have you call in not via Skype, but via Zoom or FaceTime. Because Skype is out of business. They're shutting it down. Breaking news here on the show.

We're actually breaking this news. It broke earlier elsewhere, but we're breaking.

It on the show. She'll monot says end of an era and says exclusive. Microsoft is finally shutting down Skype in May. It's the end of an era. Skype could have been WhatsApp+Zoom in one, but it was a terrible Ballmer esque Ballmer era Microsoft acquisition for $8.5 billion. Let's go. Historical size.

Historic. Yeah. So it was, it was maybe terrible for Microsoft, but it was fantastic for the Skype shareholders.

Yes.

And so we should celebrate that.

Yes. And one. So Skype fasteners, I'm sure they took.

That illiquid, you know, they got the returns from the illiquid liquid Skype stock, rolled it into other illiquid assets that have since, you know, multiplied.

Hey, you're getting ahead. We got the whole breakdown right here.

Great.

2003 company founded in Estonia by the founders of Kaza, which was. Which was a competitor to Napster, if you're familiar. The P2P File Sharing Service used to pirate music. I was using Kaza back then. That was fun. Download some corn tracks or Limb Bizkit albums. It was fantastic. This is like the best era of the Internet. The Kazaa money meant they didn't need.

To raise truck in 2003.

Really?

Yeah.

Wait, what are you.

I'm actually.

Wait, what do you mean you got run over?

This is some lore. I got run over by an off duty ambulance driver who was driving a pickup truck.

Is this why you can't bench three plates yet?

Yes, that's exactly why I was actually. I went, you know, got taken in a real ambulance to the hospital and they said, you know, your son's fine, but he's never going to be able to bench three plates. And that stuck with me. It put a chip on my shoulder.

Well, when you do it, you know, we're going to be.

But I'm going to prove him wrong. I'm not going to listen to that doctor back in 2003. I'm going to, you know, do the impossible.

Yeah. Once you do it, we're going to Saber open a bottle of Dom Perignon and Saber open a bottle of protein shake. Chug that.

Perfect.

So you didn't need to raise me 2003, 2005. They have 54 million users, 7 million in revenue and 60 million in 05. They're printing just.

Estonia's got to be a great place to start an illegal peer to peer file sharing platform.

Very fast Internet there. Do you know this story? So Estonia's eastern bloc used to be part of the USSR.

Is that where a lot of the VPNs are?

I don't know if there's that many, but there are some, there are some fantastic entrepreneurs there. So they break away from Russia and they invest very, very heavily in the Internet. They get very, very high Internet speeds. And so they have tons of kids that get online and learn all sorts of. Very early in my YouTube career I had a research assistant who was an Estonian guy, he was obsessed with Estonia and was always like we should cover this Estonian story. And I'm like, I don't know if I should do another Estonia thing, but I like the way, I like where your head's at, kid. And but I did do one on Bolt and the founder of that company, which is an Uber competitor and the kid built this business into billions and billions of dollars kind of Uber for the rest of the world and has been fantastic, fantastically successful. And, and one of his, I think one of his angel investors was this guy Jan who was a founder of Skype who got a bunch of money out of Skype and became a big investor. And that guy also invested in FTX very early on. He's a fascinating character. So he's been, he's been all over Silicon Valley. And so in 2005 they get acquired by eBay for 2.6 billion. EBay claimed it would help buyer seller communication but that never made sense to shib deal and ebay wrote it down by 1.7 billion in two in 2007. By 2009 they have 400 million users, 100 million DAUs and $700 million in revenue. And so 65 of Stripe gets acquired by Silver Lake, A16Z, CPP, IB and Index for 1.9 billion valuing the company at 2.75 billion. A16Z wrote a 50 million check, 50 million dollar check out of their, their 300 million dollar first fund. People thought they were crazy for investing in a late stage private equity deal. That is crazy. And I remember that being on their, on their. I Remember that being on their portfolio page. And I didn't realize that they did it post acquisition, which is very, very rare for, for VC. So in 2011.

Yeah, but it's sort of some interesting lore given that they've turned into more of this sort of multi stage asset manager that, that resembles a private equity.

Started back in 20 2009. They've been in the game for 16 years in these odd deals. So if you're out there saying, oh, they're completely pivoting their model, maybe not. Maybe they've cut their teeth.

Yeah, they didn't return the fund even though it was a good outcome, but certainly got some of the way there.

Yeah, I mean they invested at one point at 2.75 and they sold it at 8.5. So 3X, they probably got 150 mil out on that $50 million. Check, check. But that's a, that's a $300 million fund. So they returned half the fund on that deal, which is pretty good. And it was just two years. So by 2011 there are 600 million users, 170 million DA MAUs, 1 billion in revenue. Balmer buys it for 8.5.

They probably recycled the investment and just reinvested that the 3X and potentially got.

Exactly. So a 16Z looks really smart with early DPI. And so they're, this is their first fund and they're saying, hey, you gave us 300, we already got 150 back. We got a bunch of other bangers in here. Let's do the second fund, the third fund, the Growth fund, the Dynamism fund. All these different things. At Ballmer era, Microsoft, innovation slowed. Skype never really made the move from desktop to mobile. Skype had poor UI technical debt, crazy, and never got much mobile adoption. Meanwhile, Zoom was founded in 2011 and WhatsApp took off. WhatsApp has $2.5 billion and is probably worth at least 10, 100 billion. Yeah, that sounds all right. Zoom has 300 million, is worth 25 billion. Still blows my mind that Zoom trades lower than it did pre Pandemic. When it had fewer users, it went from 10 million to 300 million.

Another interesting 10 million users pre pandemic.

Yeah, it was small, but then it was, but then Google was way, way behind and there were a bunch of problems that actually Google internally was using Hangouts because they're like, wait, we got a doggy food. Our product. But once they realized how bad it was, everyone was like, this is awful. Like it's, it's, it's Breaking all the time. We can't get 50 people in there crashing. So what's the choice? We either fix this problem or we go to Zoom. And there's this story about they put all their best engineers on it. Even, like, the AI guys came in for a couple days and we're like, okay, let's figure out what's going on. Let's optimize this. Let's go deep seek mode on this.

It is wild. I have zero. You know, I have so many memories using Skype as a kid, playing Starcraft, other video games. Having Skype would be something that I would get home from school. You know, I'd finish soccer practice or something like that. I would turn it on. It would just be on almost until. Basically until dinner or sometimes even later, you know, doing, you know, homework, hanging out. Yeah, this was.

I used a lot of these irc.

Were you using Skype when you were gaming? No. In your early part of your career?

A little bit, actually.

I was trying to make a joke about you being old.

I mean, yeah, Skype was the thing. Yeah. You escape with someone definitely in college, Skype was like the default. It was like the best video chatting platform. For sure. Yeah. But the other interesting tidbit is Tony.

Bates, like, Game Boys over Skype.

It was actually Pong. That was the one. Yeah. And. And a lot of telegraphs, you know, like, that was the main thing if you wanted to raise.

Get on Skype and pray, practice your.

Exactly, exactly. I'm a beast at this. Morse code. Morse code, yeah, Morse code. And carrier pigeons, obviously.

Yeah, yeah, from time to time.

A lot. Lot of scrolls, passing around on horseback. This is kind of the vibe back when, you know, draft dodging, trying to not go to Vietnam. This is like my cut, my classic stuff. Exactly. Tony Bates, the CEO of Skype at the acquisition by Microsoft, was considered a rising star at Microsoft and was actually one of the top folks considered to be in the running for CEO of Microsoft, but was but left once Nadella got the gig. Anyway, fun to look down memory lane. I use Skype a lot in my life. I use Skype regularly until 2018 because it was the industry norm for one of my companies. In 2018, they started requiring Microsoft login and they lost many of the users they had left. And then he did add a correction. Skype 1.0 did raise VC dollars. They raised 18.8 million from DFJ in index. Mangrove and Bessemer a year before they got acquired for 2.6 billion. Mangrove was the first investor and got 100x on the deal in two years. And so you had a story about like a pen pal on Skype. What was your experience with Skype?

No, I would just use it while playing Starcraft.

There you go.

And I just remember a bunch of my friends after school get on Skype because I lived in a, I grew up in wine country and everybody was super spread out. So if you, your parents weren't gonna drive you around, you just kind of had to go home after school and get online. So.

And so if you wanted to get a bunch of your Skype friends together out in wine country for a long weekend at a nice house, what would you do?

Well, I'd go to wander.com, of course.

Fantastic. And you'd find your happy place.

Find your happy place.

Find your happy place. We've had another Find your happy place. Find your happy place.

Thank you to Tyler for the recommendation. He said next time you'll wander jingles for sure. Make sure to sing it for sure.

Flip that book of wander with inspiring views, hotel grade amenities, dreamy beds and top tier cleaning and 247 concierge service. It's a vacation home, but better folks. What's not to like? And we got a post from Kyle Tibbets over at Wander. He says international expansion has kicked off at Wander with pins popping up on the map in new places like Quebec, Cabo and Costa Rica. Stay tuned for more. I never leave the United States, but if you do, good luck to you. Book a wander while you're abroad.

You are really America pilled.

And I'm so America.

Try not to leave at all.

I'm not leaving. You're gonna have to drag me out.

I'm not leaving.

I'm not leaving. But fortunately there are hundreds of wanderers in the United States and I'm gonna visit all of them.

Yep.

I'm gonna, I'm gonna drive to all 50 states.

I'm gonna checking it twice.

I'm gonna go to stay in every single wander in America and then I'm sure I'll see some videos of people.

In the issue is they added a hundred in the last month. So if you were trying to get go to every wander and they keep this place.

Yeah. Three.

Never. You have to be doing three a day. You wouldn't even actually get to spend the night in every single one. But you just pop in, you know, remote, unlock it, jump in and anyways. But the pace is ridiculous. And we got some more news.

We got some more news.

Some really breaking news.

This shook the tech industry. This this shook the tech industry, I would say, more than any of the AI news. This was the biggest story this week, maybe this month. I think we're going to look back on 2025 and say this was the year of Sam Sulek getting his IFBB Pro card. And so Cody Maine says, extremely niche tweet for my fellow Jim Bros. But Sam Sulek earning an IFBB Pro card in his second ever show is one of the coolest things I've seen in a long time. He's 23 years old. He's got no coach. He documented every step of his journey on YouTube. And he's a certified freak. Look at that, man. That is the pinnacle of excellence. And we got another post from Zach. He says, sam Solick Sulek is the ultimate example of obsession. He just became an IFBB Pro. He was looking absolutely diced, grainy. He was peeled. It was fantastic. Winning two, he winning first two bodybuilding shows. He posted daily YouTube videos for two years documenting every single workout. He gained 4 million subscribers. This is what it takes takes. Apply this mentality to your goals. And I really think we are the Sam Sulek of tech and business podcasting in the sense that we always post North Star every single day. And it's silly. And you can look at this and say, oh, he's just some bodybuilder, he's just some vlog or whatever. But there is a lesson in here about the obsession mentality and just doing things daily and improving daily over time. You see it with, why is Jake Paul big? You know, you might not like his content, but he posted. It's every day, bro. He literally posted a blog every single. A vlog every single day.

Here's something about a coach, what a coach often provides as much as advice. Advice is free on the Internet. You can go get advice anywhere. It's infinitely available. A coach provides accountability. And so Sam, by just sharing every single workout, has the accountability of hundreds of thousands or millions of people, right, that are going to say, you didn't push yourself as hard as you could there. Right?

Right.

And they're going to call him out on it. And it's. And it's part of the, part of the program. But, but yeah, I think it's, I.

Think it's people with his 3D renders. He posted a new 3D render every single day for something like 10 years.

I think Joe Rogan, famous thought leader, had a, had a quote that was something to the effect of, how would you live your life if you, if your life was you know a movie, right? And you had, you knew that you had an audience, right? You probably wouldn't doom scroll and you would go to bed early because you know your main character, you're a hero, right? You'd get snuggle up in your eight sleep and you'd crash. You'd get that good sleep score. But yeah, there's this idea of like, how would you work out in the gym if you actually had an audience, right? And a lot of people just go through the motions in their life. They don't take it very seriously. But take your life seriously. Sam takes his Life seriously. He's 23, IFBB Pro. He's completely natural to future.

Future Terminator. I'm really pushing for that. That's our, that's our 2025, 2026 goal. Please reboot Terminator cast Sam Sulek. The lines will be around the block at every theater in America.

Truly, it actually would be a blockbuster.

It would be an absolute blockbuster.

And it's a perfect movie where you don't need. It's. It's about the ridiculousness of it. You don't need the best actor.

No. And that's the thing. If you do the full rebound reboot. The first Terminator, Arnold still was working on his English and his acting, I'm pretty sure. And so he's the bad guy and he doesn't really say that much. He's actually like the monster. So most of what he's doing is just marching and shooting and stuff. He's not really delivering that many lines. Then by the second one, they like reboot him and he becomes the good Terminator plays into all the AI stuff. I would totally watch it. The Terminator franchise, really, really cool. But it's gone downhill with all these like kind of shoddy spin offs and we're in like series seven. Go get James Cameron back. Can you believe it's a James Cameron movie? That's crazy. Get Sam Sulek, get James Cameron, get it in the same room and make a great Terminator reboot. I would absolutely watch.

Crazy to think that if James Cameron didn't. Didn't survive his reading and driving era. Remember he was driving trucks while reading. Very, very sketchy. But we would never have Terminator and we would never maybe Titanic, Avatar. We never have Sam Suk cast as Terminator.

Yeah. Well, if you're thinking about going for your IFBB Pro card, you know that sleep is the number one way to increase your testosterone naturally. And yeah, maybe you're competing in the naturals. And if you Want to get great sleep? We recommend getting an aids.

I would. I would love if someone in our community actually went to be. To be a bodybuilder. We would put the full force of the brotherhood exactly behind them. And we'd even pay off judges and.

Do whatever it takes.

Whatever it took.

And so eight Sleep. Of course, it's nights that fuel your best days. Turn any bed into the ultimate sleeping experience. Go to 8sleep.comtbpn to check out.

What's your sleep score?

Ooh, I don't know. It might have been a rough night.

I actually had a rough night.

Eight Sleep. The app.

My rough night is your best.

You got me beat a couple days. Oh, what, what you got? I got a 91.

Yeah, you beat me.

Yes. Let's go.

I got an 86. Okay.

I got.

I just didn't get. I just put. Didn't put the time into it.

Quality wasn't great. Routine wasn't quite on point. But we're getting better.

I had to pick someone up early this morning, earlier than usual.

It's a journey.

Talking about soon. Soon.

Well, let's go to some launches.

By the way, we gotta mention it, we have a link in the. In the YouTube and we have a code. Just the code is TBPN or there we go. Sleep.com TVPN. Get a few hundred dollars off your eight sleep. Enter the the challenge and try to, you know, beat us on our sleep scores.

Yep. Fantastic. Well, let's go to some launches. We got Splash Robotics coming out of Y Combinator. They are building unmanned patrol boats to defend our nation's critical infrastructure and secure our borders. They're able to autonomously detect and intercept threats faster than any human operator. I love stuff like this. You know, all the memes are like, I guess we're doing boats now. I guess we're doing military technology. And I saw someone reacting to this like, did the DoD just take over YC or something? But this is the perfect hack project for a YC team to go build. I'm so happy with these projects is this, this is a surrounding competitor, probably at some scale. But you know, with these YC companies, like, who knows, like, they're so early that they could wind up finding some little niche here and there. It's very unclear that it's like, yeah, this is a real shot across the bow to the billion dollar company. That is like working with the DoD. This could be something that goes into Coast Guard or, or responds to an entirely different program area. It could wind up being something that's more commercial focused. It could be something that, that gets acquired, hired or merged into another company. It's just a bunch of people building some cool stuff and so I'm here for it.

One of my favorite YouTube channels is this guy. Hi Sutton. He's a defense analyst. You know this guy, he just talks about naval warfare and he makes these like one shot monologue videos about interesting tech.

Okay.

So he'll basically start the video and being like, yeah, I didn't really. People have said, oh, get a better microphone, get a camera, all this stuff. Not doing it, just making this video. And these videos are super engaging. Great. He just nerds out about. You can go watch this video. Essential guide to 2024 naval warfare in the Black Sea. And so it'll talk about this guy and it'll talk about all the ways that they're using, you know, autonomous technologies and so Splash Robotics. I wouldn't be surprised if they're using his videos to inform their product development. Very cool. They talk about how they'll, they'll actually take jet skis and turn them into sort of these remote controlled bombs and you know, eventually would potentially be autonomous. So cool. Cool to nerd out on.

Awesome. Well, congrats to the team and let us know how it goes. Let's move on to Brendan Uribe, the former co founder of Oculus VR with Palmer Luckey. He has a new company called Sesame. He says we're exploring a future where the computer isn't just a tool. It's a partner with a truly natural voice and personality. No big claims, just early work. We're excited to share and a bunch of people were shouting this out. Toby Lutke, the founder and CEO of Shopify, said, man, Sesame's voice model is absolutely insane. You have got to try this demo and post the link. Still puts up 1k likes. Let's see it built different. Love to see someone include a link.

I want to try to play it. I don't know if we're.

Nikunj also says this is the most her moment I've felt in the last few years. It's truly uncanny. Recommend trying this now, but in the comments Gauravar Vora says, what am I missing? This felt trash. And Nikun just says what like so a little bit of different of different experiences. You just gotta go try it for yourself. I don't think we're gonna pull it up, but it seems like a cool, cool chat focus, voice focused AI tool. So excited for them to launch that and we recommend that you go make your own decision about whether or not you like it or not. But speaking of AI, we got a post from Kushi. She says, where is cursor for Figma? And you got an answer.

Figma is cursor for Figma.

That's true.

You've been talking with the team over there?

Yep.

They got some big, big plans in the works.

Yeah. It's really shocking that a founder mode CEO who's a teal fellow and has been in the game.

We just got dollar breakup fee.

Yeah, Billion dollar breakup fee and has been studying this stuff.

Billion dollars of non dilutive financing.

He's actually heard of AI, you know, and he's been on the same timeline that we've been on and so he is well aware of it and building plenty of tools to make Figma easier to use and more diffuse, both on the prompt to design side and then also deployment into actual app. Obviously you can imagine what where Figma goes on either side of the design program. They've built great Figma. I think they've built this great core design tool that's very, very rich, very deep. It makes a ton of sense they would extend in both directions.

I think Figma should ultimately do this sort of quarterly release model. I think they have more release than someone like flexport. Right. They need to be dominating sort of.

And it's highly visual.

Yeah, it's highly visual. Very cool. Having them switch and really start to push, you know, having Dylan push the team to say we are going to be shipping, you know, these sort of incredible advancements in the product every single quarter would be cool to see.

I was thinking about this, like have you ever seen those like pictures where people use Ms. Paint and then they paint like the Mona Lisa and it's like the most incredible thing. Or they do art in Microsoft Excel. I really want to see an account pop up that's like Figma Abuse basically where someone built like the Sistine Chapel in Figma and it's like pixel perfect.

But like with the individual pixels.

Yeah, exactly, exactly. So just having someone there obsessively abuse Figma to do something that it can do, but it was certainly not designed to do. I think that is highly viral potential.

And I think it'd be a great marketing in the age of AI generated images. I want the artisan designer to be making an image pixel by pixel.

I want to see the time lapse.

With the face cam hand colored, hand painted.

Exactly.

Go and do it. We'll cover it here live.

That's great. Well, let's Move on to Weber, who says, introducing Flora, your intelligent canvas. Every creative AI tool thoughtfully connected. We've built Flora, Flora, Fauna, AI side by side with creative professionals, from art students to designers at top agencies like Pentagram, to give them speed, control and collaboration in one seamless system. This feels like Figma adjacent, potentially. Is that right?

Yeah.

We built a library of creative workflows from cinematic storyboards to typography variations designed by creatives with taste.

Feels more oriented around generating things like copy versus design. But I'm sure they're thinking about it.

Yeah, like less prosumer, less professional and maybe more like. What's that with that other design tool, Visual Electric. No, the one that's the Australian girl runs. You know who I'm talking about.

How do we not.

How am I blanking on this? It's worth like billions, Right?

I know.

I'm talking about Can. Canva. Canva.

I was going to say if you search design tool, it will be the.

Yeah, but Canva was always like a little bit more consumer, like make a meme, make a postcard, make an invite, and then it's gone kind of upmarket. Figma has been more of like what the best designers in the world use and now they're going a little bit more down market to something where an individual programmer could use it. And this feels a little bit more like pop in, make something. And they're starting with, you know, something that's a little bit more organic, UGC driven than, hey, we're plugging this right into enterprises. On day one, they've said they built a library of creative workflows. From cinematic storyboarding to typography variations. These workflows could save weeks of iteration to get access. Repost the first post above and comment Flora. That's hilarious. Little, little algorithm hacking on your. On your launch video. 1000 repost posts and 3000 likes. Half a million views.

Very smart. You don't see that as being. You don't see that on. When people release podcasts on Instagram now, they'll say comment for the link. Yeah, and I'll DM me the link.

Right.

And it's sort of a way to just hack the algorithm to think that there's a ton of interest in a post, but clearly there was a bunch of real interest here and go watch the demo. It's very cool how they, you know, link these models together and this feels something. Something closer to cognition or Devon for me. And that if they can own the workflow as the underlying models get better and they're connecting them in different ways to generate final outputs. They'll be well positioned and their product will improve as the underlying models get better.

Well, it's cool to see people exploring new tools and yeah, love to play with that and see if we can generate some content for the show with it or something. Thing. Let's play around with it. Well anyway, let's move on to the timeline which has been in turmoil.

Vishal in the chat just asked for this, so his timing was perfect.

Fantastic. Packy McCormick says Andrew versus Jason Calacanis. Who you got? And a lot of people are sounding off in the comments.

Reggie in the comments. Yeah, drone strike his ass.

Now we're talking. Well, I've got the popcorn. And so this all started with there's been news in the. Just in the world that like there might be some Epstein files coming out, some Epstein leaks or something. Palmer posts a picture of the black book with Jason Calacanis phone numbers listed there. I guess this is a Rolodex from the. The late Mr. Epstein and. But this is actually old news. This is not the years ago.

This is not to be confused with the client list.

Exactly. Yes, yes, yes.

And Jason doesn't, you know, it doesn't necessarily make. Certainly doesn't make Jason look good. Yeah.

But there are lots of people in that. In, in the black book because they met him for any sort of reason.

In this case, they had at least four of his phone numbers. So it sounds like it looks like.

Three or three in an address. Right. I think. But yeah. So obviously they met and. And Matt Griffin, grim quote tweets Palmer and says this is the face you make when your weekend just got a little bit more complicated. We see you, Jason. So the Android guys are really coming for jcal. And Jason responded, he says, and now Matt Grimm, Palmer Lucky's partner, is using a Photoshop photo of me being twice as fat as I ever was and trying to slander me by insinuating I was involved with Epstein. Matt knows full well that I wasn't, that I've said I wasn't and that there is no news here. Like thousands of.

So the issue here, just double clicking on the Photoshop thing is how do we know this is Photoshopped? It doesn't even. If you were going to go to the lengths of photoshopping a picture, you would, you would maybe take it a bit further, I would say because he looks. Yeah. Like those J.D.

Vance ones where he's like full red or like the Gigachad one. Yeah, yeah.

Take it a Little bit further. This is just very believable. The other thing is, you know, Jason has been vocal about being on Ozempic and these things and he used to have an, some, some extra weight on him. So the issue with Jason is he does tend to say things that aren't true from time to time. And so I do believe there's some factually correct things in here related to this not being the client list, things like that. But at the same time like thousands.

Of others, I might, I kind of.

Think it might be a real pick. I think it might be a real pick.

And ted in the 1990s I am in his black book which has been out for over a decade. That's true. And has hundreds of other ted. Ted Tech finance, New Yorkers in it. What is Ted?

Yeah, so, so it is a real picture. So a few, a few of the guys in the chat, Rory Hescal and Brian. Matt actually has a screenshot of the source he pulled it from.

Okay. Okay, here's a real photo. Jason, I mean we respect, we respect someone who's looking great now. He's looking younger than ever. He's getting diced. He's on Ozempic and I see an IFBB pro card potentially in the future. Know this one not looking great but he's looking better now. Keep going. Apologize to Palmer, clean up the relationship and then hit the gym. Yeah, go subscribe to Sam Sulek.

The information.

The Arnold Classic.

He basically open sourced.

I'm not asking him to go open.

People have been classic. People have been. People have been investing in creators. I'd like to see Jason Calacanis's launch.

Yeah.

Invest in, you know, give, you know, invest in Sam Solick. I would kill to be in that.

I would, I would actually that's a great.

We should try to.

I would kill for that.

Try to put that together.

Yeah, we should.

John would, would kill somebody to get Allocation.

Oh man. I think he's going to be extremely successful. I'm super bullish on Sam Sulek and so yeah, little bit of, little bit of spicy timeline today. You know we love the Anduril folks. We love jcal. There is one problem that we need to address which is the fact that Jason Calacanis still does not run ads on the all in podcast. Podcast and that you have to imagine.

With the, the all in and turmoil now. You have to imagine he, he's seeing a window to potentially get some ads on there. Right people?

Yeah.

Sachs is out of the picture.

Do you think is the reason why he doesn't run ads. Do you think that's what Epstein told him?

Say, hey, my final wish to you.

Don't run ads.

I know we weren't that close. I know you never came to the island.

Yeah.

But just please, whatever you do, don't run out. Don't run ads. And Jason maybe just wants to respect it. Seems like they were maybe they certainly were friend wish. They certainly were friendly. We don't know if they were friends. Yeah, yeah, yeah, they were, I don't know, acquaintances.

I, I, I think, I think J. Cal has the chance to turn it all around, apologize to Palmer, clean up the relationship, and then run some ads for Andre.

It really is a stain on. It's rough, Jason, and it's rough.

Rough. And, and we know that Palmer does have the ability to turn the other cheek. You know, he was going after Zuck and Meta and Facebook for a long time and they cleaned that up. They cleaned up that relationship. Something happened, things changed and they clearly, you know, had a conversation. Palmer went and looked at the Orion headset, said, hey, this is great, you know.

Yeah.

And so I would, I would love to see that turn around because I hate when the big dogs in tech fight.

Fighting.

I don't like it.

Hate to see it.

Yeah, I hate to see it. I just want one. One Solidified podcasters. Technology.

Podcasters and founders should never fight.

Never. Never. Anyway, well, if you love Jason Calacanis, don't work at Anduril. I guess Devin Luton says crazy Andrew Camp.

That's an amazing extension. I want to see an ad quick billboard that says, if you love Jason Calacan, you'll hate working.

You'll hate working at Anduril. Yeah. If you're listening to this week in startups, probably don't work at Anduril because, yeah, you'll be, it'll be rough for you. But they have a crazy Anduril campaign here at Boston South Station. They put up a bunch of different posters. They're literally obsessed with America. Don't work at Anduril. They have the graffiti, some great design going on, and people are loving the advertisements. And if you're looking to do some out of home, we recommend adquick.com baby be out of home advertising made easy and measurable. Say goodbye to the headaches of out of home advertising.

That's an ad quick line that we.

Could add quick only Ad quick combines.

Technology, enables you to launch hyper targeted out of home campaigns.

We, we're still working on, we're still workshopping that, but it combines Technology out of home expertise and data to enable efficient, seamless ad buying across the globe. Go to ad quick.com to check it out.

They work with a bunch of of fantastic companies too. If you work with Ad Quick, you're working with Shopify, Carvana, Door Dash, Poshmark, Butcherbox, Ramp, Wall Street Journal, Warby Parker, Discount Tire, Cool, the H and R Block, Lyft, Instacart, Compass, Compass, Basecamp, Equinox, Pluto, TV, Duck, Duck Go, Nike, NFL, Amazon, Drizzly, GoPuff. They got all the big dogs up there. Go be a big dog. Run some out of home at Ads.

With Ad Quick and you can follow Chris on X. He said hey Elon Musk, look at this. And it's a post from someone flagging. The fact that Google is buying Apple search ads on the grok term is just too beautiful a circle of life for me. Google is buying Apple search ads on rock term.

Hey, you wanted this extremely politically incorrect LLM. How about about the most, the most politically correct, politically correct one?

Do you want the opposite?

Yeah. And so yeah, I don't know how I didn't know Google had a Gemini.

We talked about this. They spun it out because it was so hard to find because I was complaining, they listened to me. I was complaining about how I paid all this money and I couldn't get to Gemini. Although I'm sure when I download the app, who knows what model I'll be on. And the model probably won't even tell me what model I'm on and what the context window will be. But that's not how people use use these things. And so they say unlock the power of Google AI. On your phone, you search for Grok, you get Google Gemini. Don't make the mistake, you know, be sharp, don't get fooled. If you're looking for Grok sexy mode, you're not going to find it in Google Gemini. Let's go to Brent Liang who we've done some work with. We love this guy. He says the Lean Startup thinking has been has probably done more harm than good. Lots of low conviction, three month pivots and folding based on an abysmally small sample size encourages neurotic and constantly skeptical. Almost the opposite of being an entrepreneur. What do you think, Jordy?

I mean the thing that. Yeah, I think the lean like the whole thing. The best thing you can do if you're early in your startup journey is just read all the books, read 0 to 1. Read the Lean startup Founders podcast. Founders podcast. I think founders podcast is advice or Wisdom that's more timeless. Right. Where, where Senra is not saying you need to fully follow the McDonald's founder and just do what they said. He's just saying this is his approach to work. Learn from it, take it, apply it. I read the lean Startup back in the day. I think it's this fine line between sometimes if you're willing to just persevere and look stupid and be okay with the fact that you don't have traction and just keep pushing forward, trying to get, sometimes you get paid off, you know, you get paid for it. If you look at, you know, crypto founders, the ones that have been successful have not followed this approach for the most part. They've sort of just said, you know, they stayed, stayed lean, but more so they're willing to look stupid for two years so that when the market rips they look like a genius. Right. And so it's definitely. This is a vintage post. I think we're now seeing. If you're going and building in hard tech, for example, you can't afford to follow purely the lean startup approach. It's a model that I think works really well for PMF or die style products where it just says intense focus, stay really lean. But yeah, it's not going to work. If you're trying to build Anduril, you need to have this sort of bold, ambitious plan and be very patient and execute it over many, many years and you know, realize that, you know, you could do years of work before being paid for it. And that's part of the game.

Yeah, I think this is the lean startup thinking has kind of mutated a little bit. Like the original idea of the Lean startup was just that in the dot com era, in order to build a website you had to raise $10 million and build a data center. Yeah, like I remember as a kid, like my dad worked at a.

You hand built some of those data.

Centers and I literally did. I would rack servers as like a summer day job. And, and, and this was just for like a company with a website.

You got it when, when, when we're, you know, 60 and some. And on posters in the comments, you know, you chirping at you, you said.

I was servers myself.

I was racking servers when I was.

You're wearing diapers. Yeah, but it's true, like, like the default, the default path to building even just a website, like we want to build a car, a car auction website or, or we want to build any startup. It was raised $10 million because you are going to need to buy Servers hire a lot of different people because there are not these abstraction layers of aws. There's no cursor. There's not even Python. Like, the code needs to be written in Java or C. Well, what happens when you write in Java and C? You need five times as many people because everything takes longer to write. And so the lean startup was just getting the word out that it is possible to now build a company.

Yeah.

In a lower cost, lower capex intensive way. And that was true.

Yeah. He basically just coined what was happening.

Exactly.

In some circles. Said, you know, this is a way. It's founder mode. Right. Where founder mode didn't have a name.

Exactly.

And then. But people kind of were aware that certain companies just operated, you know, and CEOs just operated differently.

Yeah. And. And so people. He. He. People were. He was just saying, like, hey, we don't need these $10 million seed rounds to build websites anymore. You can build on top of these services and stay lean. And then once you find traction, talk to your customers, get out of the building. These were all these ideas of, like, don't go and waste all this money and. And do a $10 million experiment. Do a $1 million experiment. But I agree with Brent that pivoting every three months, folding based on an abysmally low sample size, that is really wrong. You should. You should make sure that you really tested things, actually understand the product. And ideally, you have a ton of conviction about the. The direction that the company's going before you even start. So you're like, yeah, the first three months have been rough, but I still know that there's an insight here. I still know that there's some fundamental value of what I'm creating. Anyway, let's move on to Rohit. He says, excellent portrait of Tyler Cowan in the Economist. Lots of known beats, but this is the most underrated podcast heart. And so the Economist did a whole profile on Tyler Cowan. I highly recommend you go check it out. But I love this quote. He says, Cowan's life. Cowan's emotional life remains a mystery. He told me he did not experience regret. I don't know what the function of it is, he said. Is it to signal thoughtfulness? To stop you making further mistakes? It's like revenge. I don't understand it. Cowan also said he doesn't understand envy or anger. He didn't know. Know what? He didn't know what he should be envious of. He didn't get lonely by himself or in company. He actually said, why bother? When he told me he had never been depressed. I asked him to clarify what he meant. He'd never been clinically depressed. Depressed for a month, For a week and afternoon. I looked up from my notebook and an enormous smile, one I had not seen before, spread across the whole of Cowan's face. Like for a whole afternoon? He asked hugely grinning. Just mogged.

So on the on the topic of our earlier subject, GPT4.5, Tyler Cowan was quoted today saying, I'm more positive on4.5 than almost anyone else I have read. I view it as a model that attempts to improve on the dimension of aesthetics only as we know from Kant's third Critique, that is about the hardest achievement possible. And so somebody did the b me prompt. So so be me. Tyler Cowan, wake up at 5:23am Precisely. Eat exactly seven blueberries, three Brazil nuts, and a quarter cup of Icelandic skier sky Sky. I don't know Productivity jpeg Write two blog posts, one on obscure Romanian cinema, the other on the economics of parking meters. Check email reply exclusively in lowercase Very interesting. Thanks for this. Send perhaps send Attend breakfast meeting with unnamed intelligence officials. Discuss Singaporean food stalls instead of geopolitics.

Yep.

Lunchtime arrives. Ask waiter what is your least popular dish? Waiter visibly confused.

Popular dish.

Recommend some fermented herring special. Wonderful. I'll have that extra spicy. Genuinely enjoy it. Back to Office Podcast Guest waiting. It's a chess grandmaster who's also an amateur botanist and former dj.

That's great.

So I actually think this is like pretty.

That's pretty good. And. And that's just like so many different threads pulling all together. There must be a lot of different data in the training model if you're pulling all that together. Because some of that stuff, you really have to follow all of his blog posts, all of his tweets, all of his podcast transcripts to understand or talk to him in person to really get that flavor. So yeah, good job. Maybe potentially underrated. Maybe the early haters are wrong. We'll see, we'll see as it rolls out. Either way, it's just nice to have a more aesthetic model. I think that's good. Even if it's not, there's actually maybe even a better outcome if it's just a nicer, friendlier, more aesthetic experience. It doesn't necessarily need to be 190 IQ solve everything.

Generally, what Carpathy was kind of alluding to just said differently, which was if you're a high taste tester, you're going to notice the difference here. If you're a lobby word cell, you're Going to maybe miss.

I need to. We. We need to try the be me prompt. And maybe that should be the new prompt. See how it goes.

Be me golden retriever start day. Jordy Hayes run around the house, bark a little bit turbo S PNG Put on sunglasses indoors.

It's great. Well, we got a post from Katherine Boyle. J.D. vance is going to be headlining the American Dynamism Summit next month.

This is Coachella for get. For tech bros that like guns.

Yes. And Coachella for tech bros that don't like guns. Had Gavin or Gavin Newsom and Dr. Chelsea Clinton at the upfront summit. And so this is the opposite of the all in summit says Chris Frolick.

And wait, Chelsea Clinton has a venture fund? She has a venture fund.

What? I didn't know that.

How have we not. We've said before we would never raise around.

Yes, yes, yes. Also, when did she get her doctorate?

Metrodora Ventures.

Interesting. Anyway, yeah, upfront summit happened this week. It was very popular. I think Delian stopped by. Some other people texted me that they were in town.

Yeah. So she has a values conscious venture capital firm investing in health and learning.

Huh.

Maybe she get back the PMF or die founders.

That'd be great.

They also don't want to raise. But you know ventures about value add more than capital. I bet you do.

We. What were the values that she cares about?

Health and learning.

Oh, she should back Sam Sulek. He's all about health.

That's so true. That's so true.

He's the pinnacle of health. Look at the picture.

Uptrend try to get Sam suck but he just was busy.

They got Chamillionaire a long time ago. Did you know that?

Yeah. They've had a Chameleonair as like a venture lineup.

It was very fun. I think Andreessen had Nas and then upfront had Chamillionaire. And so you'd hop on a call and Chamillionaire would just be hanging out with you talking sassy sass.

Talking sass.

Great. Anyway, speaking of the PMF or die guys, let's go to Blake. He says standards I have for new employees slash partners. Work 12 plus hours a day, make massive sacrifices, desire to pay their dues, won't break under pressure. These are 100% within one's control.

So the, the. The key here.

Yep.

Is it's perfectly fine for a leader to say this if they live all those things. Blake is you can go on PMF or die xyz, which is not our website, and see exactly how much he's working, tracked every single second of the day. He's making massive sacrifices. He's not leaving the cage for 90, you know, potentially 90 days. Pain is dues potentially ever.

He hasn't broken under pressure.

He hasn't broken under pressure. I mean there was a lot happened. There was a lot of pressure week one. The thing, the thing we didn't fully calculate with the cage is that the, there would be hundreds of people basically, you know, demanding that, demanding attention in the chat. So we'll get to talk about that. We're actually going to be doing a press conference on the PMF or dive stream at 1pm Pacific in just 30 minutes.

So we gotta, we gotta check on him because if PMF or die.xyz shows Blake in the seat for 11 hours, 59 minutes, it's not going to look good for him. Yeah, he's got it. He's got to be down, he's got got work. Cell number 17's got to be operating at full capacity.

Totally.

Let's go to Elon Musk. He's quote tweeting someone who is doing another AI game dev demo. He says, wow, this is cool. Danny says, I believe that Grok can be a serious tool to build quality games. I've been pushing the limits of Grok 3 by having it code this retro doom inspired gothic shooter from scratch. It's been less than 24 hours to of work and I think I got it to a pretty good place and the demo is pretty cool. But yeah, there's another post in here that I wanted to highlight that was, that was fun about how if you want to go viral, the best way is to code a game right now and screen share the video. Let me see if I can find that one yourself.

Building the game. You build the game and then play it.

And then you build it.

Play it, build it, play it.

Yeah. Christoph says if you want to go viral right now, start making a video game with cursor and post the progress. I'm sick of seeing it but that's what gets the impressions right now. And so yeah, maybe we should build a TBPN game go viral. Hey, we built this in 10 seconds.

I think we just need a bait. We need to be focus one post a day from each of us. Baiting an Elon repost.

That's good.

Wow. Elon is awesome and Grok is so awesome. I love Grok. Grok. I, I stopped using all. I stopped using ChatGPT and everything else and I just use Grok. It is the best.

Yes.

Repost. I also crushed your phone just starts melting.

I crushed my 911 into a, into a ball of metal and I bought a cyber truck.

The metal sidings on my cyber truck.

Yeah. With Crush911 melted down. 911 is the side side plating for sure.

We actually should do what of this. It would be worthwhile doing telling Jason Calacanis huge Tesla guy to say, hey, we're going to drive a vintage Ferrari. You drive a cybertruck and do one of those pulling battles and just let the cybertruck win. Post it from our account. Get the repost.

So we get the ratings.

We get the ratings.

We get the ratings as Jake Alex would say. But it is cool. I love that people are building cool games with cursor, with Grok, with all these different tools. Very, very fun. And we love to see people be able to instantiate like silly ideas like PMF or Die or PMF or Die XYZ in a very quick fashion which is only possible with the AI era. And so yeah, if you have some funny idea, something on your bucket list, go build it, post it. You might get quote tweeted by Elon Musk and get 3 million views on your, on your app, drive a bunch of followers and then start a podcast. Who knows? And anyway, let's move on to Lewis Hamilton.

Okay, I want to, I want to read this because this is, this sounds like something that Senra would say.

Yep.

And it, it's clearly.

I think he might be quoting David.

Sending A lot of the best athletes, you know, actors, etc have just been ripping Senra but clearly inspired by, by Timmy or potentially happening at the same time. Lewis Hamilton now over at Ferrari says don't ever compare me to anybody else. I'm the first and only black driver and that's, that's ever been in the sport. I'm built different. I've been through a lot. I've had my own journey. You can't compare me to another 40 year old Pastor present Formula One driver in history because they are nothing like me. I'm hungry, driven, don't have a wife and kids. I'm focused on one thing and that's winning. That's my number one priority. I love how locked in he is.

Is.

There's this quote, I think it's, it's Alonzo.

Yeah.

Where he's driving and he makes this like very, very like sketchy pass.

Yeah.

And he goes after they're, they're interviewing, he goes, oh, he's got a wife and kids at home. I knew he would break and it's like the craziest thing because you can tell he's you know they're both gunning for the corner y He's like yeah this guy's I don't want to racing again is like I don't want to die.

Wow, that is crazy. I hadn't thought thought about that.

And so when you're racing it's, it's you know the, you know I, I don't know what the actual I'm sure that BASE jumping is more dangerous than you know F1 but F1's one of the most dangerous you know high level.

Professional sports thankfully getting a lot less dangerous thanks to the halo and some.

Of this and 100% I I drive a lot slower than I did pre kids you know I I used to go up in the canyons and have a rip. I do that much less frequently now.

Yeah sense. Anyway let's move on to Sam lesson Sam says Elon's what did you get done this week at at AI Scale is in fact the AI future AI isn't going to be your professional assistant. It's going to be your professional manager. Getting spicy. A lot of people probably won't like this one over and over I Sam.

Sam doesn't get enough credit for being up there and the rage bait fame with with people like Sweaty Startup. Oh yeah, you know he's up there. Sam is more sophisticated, you know in.

His approach, higher brow.

He's not saying you know, I want my employees to work hard. He's saying you're going to be managed by AI and I think this is a cool take. You know we've talked before. You can imagine a world where a an individual who's never managed another person before is managing, you know, Devin and other agents to run the their business. But this is sort of flipping that on its head.

So he says over and over I see people pushing these rosy ideas that AI is going to be your assistant. Fine. Yes, you will get some cute AI that schedules for you in some form and sends sweet nothings in the morning about your day. It will prepare little reports for you that might make you feel all warm and fuzzy but that won't be the real leveraged impact. The real leveraged impact is that AI is going to be your manager. You are the order taker for it. Not not the other way around. Elon has got this right. The real value of asking formally what have you done this week and knowing who is pulling weight, culling those who don't and Optimizing, pushing those who are slash can. You couldn't do it effectively before AI at scale, but now you can. It might be nicely disguised as a coach. Cute. Hey, to be more productive to do XYZ. Or it looks like you could focus on X vs Y if you want to get that promotion. But what it will really be doing is snitching for your boss, your boss's boss and ultimately, ultimately the dreaded shareholder. Why? Well, that is where the productivity is going to come from. AI will clear the path for basic easy tasks, to make sure your time is leveraged fully and will make it easy for managers to deal with all the complexity. Cut to the chase of where and how value comes from and who is driving it. More transparency is generally good, but people who think this is some rosy machine works for man. Nah, it's the other way around. Interesting. Especially interesting in light of that controversial AI sweatshop company from YC this week. That, that it's like the. What did you get? I. No one really pieced it together, but that company could have just been called what did you get done this week? AI.

No, it's more of like what did you get done in the last hour? What do you get done in the last 10 minutes? And then comparing that over time previously. And knowledge workers have this thing where they sort of get assigned a project or they have a general area of response responsibility and then they just sort of have to. There's actually really delayed feedback. Everybody that's managed a team before, if you've started a company, you hire somebody, you sort of break down what they need to get done. You check in with them a week later and sometimes you're, you're saying, wow, this is amazing, incredible progress. You're crushing it. Sometimes you're sitting there and being like, well, what did you do the last two days? You're coming to me now confused. I wish you asked me, you know, two days ago.

Yep.

And even worse is like if, if something slips for a couple of weeks and it's like, oh, it didn't even get started.

Yep.

And I can sort of potentially be omnipresent and sort of be in real time guiding people towards, you know, the right outcomes. So I still think there's like a manager sitting above the AI coach manager that is sort of setting, you know, the goals. But maybe over time, you know that that fades and it's just we were all just working for, for the, who knows, hard working paperclips, the shareholders at.

The end of the day.

Yeah.

Well, let's go roll off bowtie.

That's actually insane. Just to kind of extrapolate there. You can imagine a world where shareholders in a public company say we're actually collectively as shareholders are going to vote to basically hire this AI consulting team that's going to micromanage, manage the management. It's, it doesn't seem unreasonable that, that that could happen and potentially allows shareholders to exert, you know, more, more day to day control over the management at a company.

Could be a lot of pushback. We'll see. I wonder what will happen with that post, if it will break containment and get some pushback. But for now I think it's a, it's a reasonable take. Who knows? We'll see. See, let's go to roll up Bota over at Sequoia. He says we remain destined to repeat the mistakes of the past. SPVs are making a comeback where the lead investor speaks for less than 10% of the capital, yet eagerly lines up the latest set of tourist chumps who think the story will end differently this time. It's only been three years. What do you think? Who is he subtweeting?

Yeah, crazy. I mean I'm sure he's just seen, seeing a number of rounds happening across his portfolio and saying wait, this person's doing, you know, a 20 million dollar investment and they only have $50 million under management. How does that work? And it's, oh, it's a, it's a manager who's. But yeah, again there's a bunch of really, you know, good players out there that use SPVs. They're great, you know, tool for investors and oftentimes they provide access to do great companies. But certainly there are many examples where you know, SPVs typically get sort of, if you're a founder putting together a round, you're, you're prioritizing people that have the capital ready to wire. Right. Or that's one capital call away to wire. SPVs are oftentimes a, not a last resort but you know, you're, maybe you make room for somebody to do an SPV because they're saying, oh I'm going to bring in this celebrity and this person and whatever. But again, I think, you know, I don't think he's subtweeting figure, but figure right now one of the risk of the rounds they're putting together is he's basically marketing. In my view, he's almost marketing to retail investors because if he needs to raise $2 billion at a $40 billion valuation, the people that will do Those rounds are tourist chumps who think they're getting in on the next Tesla. Yep. But are ultimately that when that company has been, you know, certainly a firm like Sequoia or Founders Fund is not doing figure at almost twice the price of Anduril.

Right. Yeah. And so there's clearly like a spectrum of kind of credibility in rounds. Not just what, you know, are you doing an SPV or going with a traditional VC firm, but even within the VC world, there are some, some funds that have very high GP commits where the partner who's doing the deal has a very material portion of their wealth in that fund. And so they are like, they basically solve the principal agent problem by putting their tying their net worth to that firm. And. And so, yeah, you might raise a $10 million Series A and $2 million of that might come from the partners of the GP or more in some cases. And so if it's a professional investor who's highly aligned, that can usually be the best possible signal or the best possible outcome.

But there's a good comment when we're good here from Hari. He says this is generally true. Responding to Roloff. This is generally true like 99% of the time. But when you stumble on a diamond in the rough, I'm still chasing the high for my second ever angel investment into an SPV with no, no institutional lead run by Naval into the Notion seed round in 2016. So, yeah, you know, I'm sure that was somewhere around 20 post and is now sitting at north of 10 billion. And he did. Did pretty well.

And I wonder if that SPV was the. Was the lead check or not because it's very different if it's like. Oh, there was.

Well, it said it had no institutional lead.

Oh really?

It basically. Oh, yeah, was the biggest. And his investor group was the biggest.

Yeah, yeah. I mean Naval's like pretty coated as SBB Runner. Anyway, let's move on. What comes after smartphones, AI glasses, you actually want to talk to, says Ani Midha. We at Andreessen Horowitz are excited to invest in Sesame's mission to bring the computer to life.

Interesting. This is a reference back to tech demo earlier. Earlier in the show where, you know, somebody was saying that this is the most her like moment that they've experienced. Other people were pushing back on it, but I just like the domain sesame.com.

It'S nice.

That's a good one. Yeah, that's a good one. Feels very, you know, it's very memorable.

Yeah. And then it still ties to like this idea of like open sesame. You're talking to a genie in a lamp. You have three wishes. Like there's like some fun lore there, but it's still a word you can spell. You can get the dot com.

It's open. This is a company that's competing directly with Meta Ray Bans, which is going to be tough because ASAP Rocky, famed rapper, I didn't know Style God, yeah, is working with Ray Ban. And so I saw Reggie was posting that. I'm sure ASAP Rocky will then be working on the Meta Ray Bans and.

You'Ll see and they have a crazy partnership with luxottica for to bring in house all the Ray Ban stuff and, and then obviously they're, they're rolling out llama and all these different, different tools. And that was one of the more attractive features of the Meta Ray Bans when I experienced them was taking the dog for a walk and being able to just say, hey, hey, Meta. What's, you know, like how old is this tree? Or something or whatever. It was like, it was, it was cool to be able to ask, you know, history questions. You're listening to a podcast and then you can just say, hey, Meta, like tell me more about this. And it becomes more of a conversation. Anyway, let's move on to Bridget Harris over at Foundation Fund. She says demand for yield will send the stable stablecoin supply way, way higher. Fintechs are realizing stablecoins are the best way to deliver yield to customers and it'll become status quo the same way Cashback did. Once customers adjust to getting yield on everything, it's hard to go back. Latin America is the first mover here. But popular US based fintech platforms are also starting to offer yield, which consumers are increasingly recognizing as the new normal. Right now, coinbase automatically offers 4.1% API on any USDC held on their platform. And there's a bunch of other companies that are getting into the game, into the stablecoin game. Robinhood's mentioned and they're pushing. Yeah, different API. First for subscribers, APY for subscribers. And so moving people into stablecoin.

It's so crazy that biology was formative and actually just making this category. And now the market cap is the stablecoin market cap is how many?

21 co was his company. He wanted to create a computer that could mine Bitcoin and also do a bunch of other stuff with stablecoins and micro payments using crypto. And now the technology is finally approaching reality. The micropayments haven't really rolled out yet.

But it's really painful that one of the areas of crypto that has the strongest product market fit, you couldn't really invest in the tokens themselves.

Yeah.

Because one one well tether if you.

Got into that deal, they're making a ton of money, they're printing money because they take a slice of the APY on top of it, the yield. But I mean it also has to work better in a high rate environment. Right. Because how can you get high yield on USDC in a zero interest rate environment? It's got to be lend it to.

FTX to try to make a 10x in a day question. You might lose it all.

Well, we got some personnel news. Joe Ghebia, the founder of Airbnb is moving on to Doge. He says he's excited to share. I'm bringing my designer brain and startup spirit into the government. My first project at Doge is improving the slow and paper based retirement process. Hilarious. Like that is a job job that no one in his position would ever.

Want, has ever done.

Has ever done.

I'm going to go be a product manager for the government.

It's insane. That beautiful.

I love it.

Thank you. Thank you for your service. Like I 100% want a smoother and less paper based retirement process. This is fantastic. Since leaving his operating role at Airbnb in 2022, he's been looking for the next design design challenge in the digital realm. And he can think of a few more important he and he says I can think a few more important ones than volunteering to improve the user experience within our government. If anyone else in good standing wants to help design beautiful, user friendly digital products, reach out. And so he's going over to the Office of Personnel Management trying to smooth things over with good design. And that makes a ton of sense. I love it.

So cool to see.

And you've seen just like the flaws with like the Medicare website took all this time. And like whenever you interact with any government website, you're just like this feels 20 years old. Like what's going on here? And so just having someone in there who can, you know, staff a team of designers, turn things around quickly, actually.

Implement changes most people haven't, you know, build a company like Airbnb and they go, they go into the opm, opm, yeah, Business. Other people's money.

Yeah.

Joe said, okay, I'm gonna go into the office, I'm gonna go into the opm, you know, sector and the Office of Personnel Management.

I love it. Yeah.

And you know, make Our country a better place to be.

And I mean, it is crazy. Like. Like, you think about his experience at Airbnb. He wasn't just the co founder who was there for a couple years. He was there all the way to 2022. So historic run, 15, 14 years, something like that. And then all. Also, Airbnb operated at a scale that is actually somewhat relevant to the government. You know, it's like millions and millions of people. They're dealing with payments, they're dealing with not. Maybe not Social Security numbers, but certainly like ID and. And KYC stuff. And so they've done a lot of different tools here. So, great post. Congratulations to you, Joe, and good luck on your mission to improve the retirement workflow.

Before we jump into the next post, we got a request from Brandon Jacoby. He's going to pull up a video right now that Zuck just post posted on his Instagram, but it's pulling it up. He doesn't have audio. I don't think we need. I don't think we need audio.

Look at this. Wow.

So Zuck posted this on his Instagram.

Oh, my God. This is incredible. People who called the top on Zuck?

Okay, but.

Wow.

But I. The live reaction. I have no idea. This is for. I don't think Zuck can move like this.

AI.

I think he's flexing a new.

I don't think so. I think he did it. I think it's real. Look at. Yeah, look at the jump. Yeah, that's accurate. The fingers count up. Everything looks good. They're having a blast. He's singing. This is amazing. Where is this? He's doubling down.

His wife's birthday or something.

Something like that. Wow. What an amazing, amazing post.

I love it.

You know. You know, in China, that would get you in prison. Did you see? Yeah.

Okay. Says your wife only turns 40 once.

There we go. What song is he singing? This is incredible.

Here, I'm gonna pull up the audio real quick.

I love it.

React.

It's so good. Yeah. Jack Ma did something similar.

Okay, so he's lip syncing.

Okay, that's great.

Still.

Yeah. Jack Ma did the rock star thing on stage and shortly after kind of disappeared, literally. Because it was like people were like, okay, he's becoming like a God.

Like, people are like, credit to Zuck. He practiced that to a degree that he.

Oh, yeah.

He looked.

Yeah. Like a real perfect blend of robot and bro. Clearly.

I mean, tested it. Totally, totally Mogging. All the people that said, oh, they were trying to do a poly market on if he was going to get divorced this year.

Oh yeah. Remember fake news?

They're back.

And yeah, I mean I think gamblers are in shambles. I think he did the right thing. You know, there's a lot of questions about like oh, should he back off the bro stuff? No, double down. Go crazier. Keep it going.

I mean this is sophisticated and wholesome stuff.

It is very wholesome.

40Th you're gonna put on this incredible show. She's gonna remember that forever. Yep.

The. The PR team behind this must be paid so well. It's like this clearly came.

Let him cook.

Let him cook.

The blue sequin suit.

That's gonna great. That's fun. Anyway, let's move on to some reports from Altimeter from Shield Monot. Incredible early vintages from Altimeter VC funds. Wonder how much of total returns are from Snowflake. They led the series C at $260 million post in 2015. That stake was worth 4.4 billion at the IPO. Let's do a little size gong for Altuner. Boom.

Thank you.

BC is a power law G game. And they crushed it on Snow Wild that Brad started with just $3 million in 2008. That's insane. Ran it up 61 net IRR, $6.5 billion realized on 2.8 billion in cat committed capital. They got a bunch of funds here, there a lot of rippers and it's great stuff.

Brad is just such a cool guy too. I ran into him at my local coffee shop. He spent. He's a server so he. He spends some time in Malibu and walked up to him said hi, we're in a handful of companies in the same companies but just such a cool down to earth dude who has been on an absolute tear. He had a rough go with the hedge fund.

Sure.

In 2022 I think it was when.

The market sold off 50%. Yeah.

When people were like seven long. Nvidia is over.

Yeah.

But that's the game.

And the interest rates went up and obviously if you're highly into tech, you're going to get burned.

And I feel, and I feel like back in the game as. As journalists like newcomer started leaking performance. This is a great time if you're an investor to leak your screenshots. Oh look at this. Leaked. I'm actually going oh, oh please don't share this. Please don't share that. I have 61% net IRR.

Yeah.

Billions of dollars. I would hate for that to get loose.

Anyways, let's move on to Nick who Posts a screenshot of Jeff Bezos. Liking his post. You know, you got to do some engagement bait for the billionaires that are hanging out on X all day. Nick apparently posted news. Amazon just dropped dropped their first quantum computing chip, codenamed Ocelot. We are so back. That could have been me posting that. But you know, Nick got to it first. He got the Bezos like. And I wanted to break this down a little bit. It's in the journal. Today Amazon unveils its first quantum computing chip. We heard this from Google a couple months ago. Google in December said it developed a new chip called Willow and that really at that moment the people that were watching quantum computing closely were like, okay, this is a break breakthrough where we're now starting to accelerate the development here. And then of course earlier this week we covered Satya Nadella on Dwarkesh Patel announcing the major Rana chip. And they claimed a quantum computing breakthrough by creating a new state of matter. Now Amazon's in the game. Their cloud computing business on Thursday unveiled its first ever quantum computing chip, which claims marks. Which it claims marks an important step in the development of useful, reliable quantum computers. That's what we want, folks. Folks. The chip, dubbed Ocelot, can lower the costs of reducing quantum computing errors by up to 90%, Amazon said. And there's. And the race is on. Everyone wants to build a practical quantum computer. The Amazon Ocelot chip is a prototype, not a full blown quantum system, said Oscar Painter, head of quantum hardware for aws.

To be honest, John, I'm a little bit frustrated. We were prepping our quantum quantum chip launch.

We got kind of edged out by this.

Satya goes on Dwarkesh. You know, even, even Google's working on one. They've got Willow. They're teasing that we kind of got beat to the punch. But I want to let people know we are still in the game. Ben has been working on it while we record. He's been working on quantum ship. He was never actually had his wisdom teeth removed.

He actually, he was grinding.

He was actually guan lab grinding.

Reducing quantum noise.

Yeah.

Creating new. Yep. Yeah, yeah. And this is going to change podcasting you. I mean you start compressing our audio feed with a quantum computer, it's going to be super high fidelity. The the best takes possible. Anyway, let's move on. Mike Taylor says I may have underestimated the power of my product appearing in AN Every post rip my calendar. 130 onboarding calls booked in 24 hours. Little 15 minute sleep slots. The guy is slammed his Product's ripping and congratulations to him. Have you ever had a calendar booked like this on a single day?

I have for party. I remember having this exact calendar set up. The my son was born on a Saturday and my calendar looked like this on Monday. It was kind of brutal. Yeah, I don't recommend that, but I'm sure Sam, all Sam Altman's look like that this week too. So sometimes you just got to do it, you know.

Yeah, you gotta, you gotta be in a certain frame of mind to really deliver and bring the same pitch, you know, 50 times back to back to back to back to back to back to Back to back. 15 minute segments. But you probably get really good at your pitch. If you do that and you get the reps in, you learn what works, what resonates, what doesn't. How to talk, how to let them talk.

I think I've done this, you know, not in a sort of micromanaging way, but portfolio companies where they'd say like, oh we not creating sales momentum. It's like cool screen share your calendar. I'm like, you're talking with, you're talking with like two customers a day.

Yeah, you should be talking.

You're never gonna actually be a top decile company from a growth standpoint unless you have some crazy product led growth motion unless you have this extreme volume because it really is just, you know, should, should be obvious. But it's just a numbers game. If, if he has a 10% hit rate but he's doing 20 calls a day, great. Adding he's going to be getting, you know, significant amount of traction from this.

Six minutes left. We're going through some quick bangers. I will read them. You give me your take. Let's start with Morgan Barrett. Big night for hard tech in California in America at Valar Atomics. So proud of Isaiah Taylor. They had a big announcement last night. Looked like something out of a sci fi movie. What was your take, Jordy?

Anyways, this was absolutely wild. I didn't know what to think. We got invited but couldn't make it. Couldn't make. Seriously. Would be easier for us to get to Austin and that side of town right now. But what a show. An absolute showman.

I was pissed because I feel like he really upstaged us on the production value. We're the ones that bring in the Hollywood designers and create elaborate sets. And his set was way above.

Yeah. People say, oh you guys have such a nice set. Yep.

And not anymore.

They say he's making us look.

Is the one that looks Cool.

So, yeah.

Anyway, congrats to the team over there. Let's move on to low Tam banger from Rob David. He says high yield, Harry City when it comes to accidentally sending people money. And it's just a dunk and, and there's a dunk reel and I don't know if you read this story, it's in the Journal today, Citigroup almost made a $81 trillion mistake. They accidentally moved to transfer 81 trillion to a customer. Yes, trillion. A bank employee mistakenly entered this absurdly high figure last year while trying to manually input a transfer to a client's escrow account.

Despite the fact they have 81 trillion, that seems a little bit.

Well, the annual GDP of the United States is around 25 trillion, so. No, Jordy, I don't think so.

Yeah, yeah.

Like.

Yeah. What, what is like Just to put this in perspective.

Yep.

They, they, I'm assuming they have what.

They have billions, I'm sure. But not trillions.

Hundreds of billions.

Yeah, maybe. And so there's a couple different examples of the fat finger transfer could happen is a lesson of technical go through.

Would it appear on the customer screen? Oh, you now have.

I have seen that happen. I've had friends who have said like, oh, some bank error happened and for a moment I had, you know, a couple billion dollars in my account because they messed it up and then they reverse it, of course.

So it's all fake.

Yeah, all the money's fake.

All the money's fake.

The money's in the computer. I talked to a hedge fund guy who actually said like there's money in the real world and there's money in the computer. And the computer money, like that's the stuff that goes up and down or something like that. It was a really funny thing. Anyway, if you're trying to make $81 trillion, you gotta start a hedge fund. You gotta go on public.com investing for those that take it seriously. They got multi asset investing industry leading yields and they're trusted by millions. And they're not going to send 81 trillion to anyone accidentally. They run a tight ship over there. Let's go to Nico. Over at default, they announced a $4 million new funding round from ABC. We covered it here on the show and they did this launch which I know you'll have something to say about. They sent out 100 branded champagne bottles in custom pelican cases. And Tom Osmond says solid sponsor for tbpn. Tbh. What do you think?

We love champagne on a long enough time horizon we will work with every champagne company.

Yes.

You know, as long as it's truly champagne.

Yep.

From the champagne region of France.

But this was a story. I actually heard this from Sam Blonde. He was at.

Yeah, didn't. Didn't they.

He's at Zenefits and then Brax, I think. And he said that one of the things that worked really well was sending people champagne. He was in sales.

Because you're not giving them a beverage, you're giving them a moment.

Yeah. Moment that they can take home. And then they open it with their wife and they think about it and they're having fun or something. Or their team. There's a lot of different ways we never.

We always offer champagne to Ben.

Yes, yes, yes. Anyway, we covered this earlier on the show and Mike Maples responded. We talked about his amazing BMW Z8. He says TB PNNers. Glad you like the Z8. I actually fell in love with it when BMW showed the Z07 concept back in the late 90s. At the time, I lived in Austin and the local dealer wouldn't even take my deposit because they were sold out instantly. So I called around and found a dealership in Shreveport, Louisiana, Indiana, willing to let me buy their allocation in 2001. I've had it ever since, and it's probably the one car I'd never sell. Cheers. We got to have him on the show. We got to ask him what other cars he owns. What else, what cars would he sell?

Yep.

You know, but fantastic car. And thanks for replying, Mike. We'd love to have you on the show, chat about cars. I've never driven the Z8, but I've heard really good things about it.

Everyone talks true classic.

Yeah. So thanks for engaging, Mike. We'd love to have you chat about cars. You're welcome to come in, call in anytime. Let's go to Riva. She says looking to hire a chief of staff.

By the way, we have basically a minute till we're starting to.

Okay, well, we got some promoted posts. We got. We got Riva. She's hiring a chief of staff. Trying the usual agencies, recruiters, but struggling to find someone vibey. So if you're looking for a chief of staff, go work for Reva. Who else we got? We got merch.

Let's save some of these for Monday.

Okay. Yeah, we'll save.

John's got an incredible podcasters high right now.

I don't want to stop. I want to stop. Well, we got. We got one more. We got one more we gotta do. We gotta do this. So it was Steve Jobs birthday on Monday and Bezel replied to us and said he's a minimalist watch king and in here he's wearing a beautiful Seiko. And I love that Bezel. We text them all the time. Can you clock this watch? And that's why we wanted to run an ad for Bezel today. Go to.

I know, I know. Brandon Jacoby is listening right now, and I think he's one of the biggest Steve Jobs fanboys that I know.

Lasers.

Brandon, I expect you to be in a Steve Jobs Seiko on Bezel. You should send us a screenshot and we'll feature it on the show.

Fantastic. Go to bezel shop. Over 22,500 luxury watches. They're fully authenticated in house by Bezel's team of experts.

That's a great place to end the show. Actually, the real place to end the show is is Sarah was. Was watching from home and she commented in the chat and said, you have 12 years. She's 28 right now. So she was referencing.

Oh, to learn. To learn the song.

Anyways, great, great day of timeline today. It's fun to rip through time. We're trying to do more timeline. There's a lot of news to cover, but we're trying to do. Yeah, we want to make sure the pod is snackable. So thank you for tuning in with us on a Friday. We are going to jump over to the PMF or Die stream and grip Patty and Blake on what they got done this week. They've certainly been busy and we are excited to grill them. So anyways, have a great weekend, brothers. We will see you on Monday.

See you on Monday. Thanks for watching.