Why AI has no taste and how to fix it (w/ Thais Castello Branco) | E2319

You have models that can solve cyberhack and solve PhD-level math problems and they can't like write a good tweet or make a good design. The average is that thing that you're going to kind of not even take a second glance at, that you've seen a lot.

The people who build these things have no taste. Let's be honest, they don't have taste. Most design, most tweets, most blog posts are average. I haven't replaced that function yet.

I don't feel like we have figured this out yet. Everyone I know who used to work in media as this kind of cultural critic or curator class, curator, curator class, curator, all struggling to, to find work.

The more you have of this, like, mass production of things, the more I think the peak of the peak of the peak becomes valuable.

This Week in Startups is brought to you by Odoo, the all-in-one business platform.

Get started for free at odoo.com/twist. That's odoo.com/twist. MongoDB. AI-assisted and agented coding is helping you build faster than ever.

Start building at mongodb.com/ai.

And Wise Security, the on-demand security team for startups. Need enterprise-grade security without hiring a $400,000 CISO? YSecurity gives you 40+ expert engineers matched to exactly what you need by the hour with your first 6 hours completely free. Go to ysecurity.io/twist. That's the letter Y security.io/twist.

All right, everybody. Welcome back. It is Friday. It's July 31st, 2026. I'm in New York City.

World tour.

Hawaii, Paris, Tokyo, New York. You've been everywhere. Austin next week. And get ready for the big school, uh, everybody says school starts at August 15th now, Lon. Did you know that?

I think they added more holidays for the kids during the year. When, when I was going to school, it was just September, second week of September you go back. This seems shockingly early to me.

Like, I broke out of the matrix a while ago. I told these schools, like, listen, we live a non-traditional, like, situation. We're going to take the kids on a ski week. We're going to start, you know, Christmas a week early. We're going to come back 2 days later. And, you know, they get put all this kind of pressure on you. And I'm like, but we're going to have a tutor and we'll keep up with the stuff. And I want them to see the world. And I was really excited because my 3 daughters came with me to Paris and they came with me to Tokyo, but they came to Foundry University Japan and got to see some of the content. So I'm really excited about that because How do you raise kids, Lon, in this new world, in this new AI world where like you could be taking a class and never put it, we already were in that situation where like, is this class gonna be applicable? Now it's like, is anything applicable? What is the skill of the future?

Do you need to read Hamlet at this point? Like, I don't know. I'm glad I did, I guess, but.

Or maybe you do and that's what makes us uniquely human is our, you know, artistic ability and other things, which we'll get into today.

Read Hamlet. It's good. It's good. I'm not saying Hamlet's not good. You should read it. I'm just saying.

Well, and I mean, all the jobs are going away except for the 6 open job reqs I have right now. I don't know.

Twitch producer, editor. My business is growing, not going away. We need newsletter writers, but most jobs are going away.

Well, if you're ambitious, now is the best time ever. All right. We've got a lot of news, but we also have a guest.

We do.

So we're going to do guest and then news.

We'll bring in the news.

Oh. To tease it a bit.

Sure. Well, we gotta, we gotta talk about Situational Awareness. That's, uh, Leopold Aschenbrenner's, uh, fund. He's only 25. His public portfolio has been sold off to Citadel. But there, there is more to this story. I think everybody initially—

we're gonna get into it.

Yeah, and everybody initially was like, haha, Wall Street, this kid got blown out, he's done.

There's a lot going on here behind the scenes, and I have strong feelings about it. What else is in the news? Well, there we go.

We gotta talk about Google Made this insane decision to add, they added Nano Banana to Google Earth. Did you see this? So you could use their AI image generator to manipulate images from Google Earth. I think you can immediately see why that might be a questionable idea.

Shenanigans might ensue.

Yeah, so we'll have to talk about that. Anthropic's models, last week we talked about OpenAI's models escaped their sandbox and went wild on Hugging Face. They hung out there for a whole weekend. So that inspired Anthropic to investigate what their own models were up to. They found some surprising things. So we've got their report on the 3 incidents.

Okay, so now it wasn't enough for OpenAI. Anthropic had to catch up and say, no, no, we're committing grand theft as well. We're breaking into other people's— we're cyber hacking.

We are also a devastating cybersecurity threat. Okay, everybody, listen, it's not just—

yeah, just to be clear.

Yeah.

So we'll talk about that a lot and we're going to go off duty at the end.

We are. And we got— you threw a ton of docket suggestions at us. We've got some rapid fire news. We are gonna go off duty. It's a packed show, Jason.

All right, let's introduce our guest 'cause I think this was a suggestion from J-Cal.

This was. You saw, you saw her launch video on X and we're like, book this lady. And we always listen to you. This is Thais Castelo Branco from Taste Labs. She is designing a, it's a system, it's a two-prong system for generating better quality stuff with AI. She feels like AI has cracked the objective domains like coding, But subjective domains that require taste judgment, you need another layer of preference data on top of that. Mm-hmm. That's what they are building at Taste Labs. Thais, thanks for joining us.

Thanks for having me.

Welcome to the show, Thais. T-H-A-I-S. The first time I've ever heard this name, but it's a Greek name. You're not Greek, but you have a Greek name, which scores you immense points with this Greek kid from Brooklyn.

Welcome to the show. I'm already, I'm already up on the other side.

That's great.

You're, you're starting from a very, very, very good position. Good start. Good start. So, um, why do most large language models have no taste? Because when I design something or a founder comes to me and they show me their design, I'm like, yep, that's the intranet from hell. That is the canonical, you know, tasteless Facebook-looking, you know, intranet with the sidebar and the shadowing, and it just always looks Not like slop, but it looks the same. Samey. Yeah. Same. Why does it suck like this? Because it's a known problem.

Well, and we keep hearing about how smart these models are.

That's exactly, I think, people's frustration is you have models that can solve cyberhack and solve PhD-level math problems and they can't like write a good tweet or make a good design, which, and I think a lot of this honestly goes back to like, how are these models trained? Right? I think a lot of, the logic is how do you arrive at the most likely answer of something that should be the correct answer. And in math or coding, that's kind of exactly the behavior you want. But then for something that's more subjective, what's considered great, right? Like in real life, like things that you encounter that are like, hey, this is really great writing or really great design. They tend to be things that are a bit more unique, that are a bit more like out of distribution, right? It's not the average. Like the average is that thing that you're going to kind of not even take a second glance at that you've seen a lot. And so I think that variety is one of the core things that are missing, but also that nuance, right? I think the other piece that's very weird is the fact that like both of you could, for example, prompt to, for two completely different things, but then the end output would like look the same. And why? Like I'm asking for a totally different use case. Like why should this be like, why should this look the same? And so I think this almost like fits to the prompt and fit to the user intent and almost like personalization is almost missing from —well, and it, it makes sense in two ways to me.

Have you been ever, uh, Thais, to Palo Alto or Atherton? Yeah, I've driven up and down El Camino Real. It is a wasteland architecturally, design-wise. Have you ever gone to a rich person's $40 million house? They all look the same. Restoration Hardware, disgraciado. You, you can barely keep a Michelin star restaurant below San Francisco. The people who build these things have no taste. Let's be honest, they don't have taste. They have taste in like private jets, and then it's a big drop-off from there. Like, they know which private jet is the best one, they could talk to you about it for hours, but they can't do anything else. Like, they— that's it. I mean, it's— they have no taste. And then second, if you were to take everything, um, as opposed to skimming the cream, most design most tweets, most blog posts are average. In fact, when you take everything, you're guaranteed to have regression to the mean. That is like the term that exists. So I guess, how do you reverse that? And then who are your customers and how are you working with them?

Both great points. And by the way, I think, uh, quickly on, on the first point of like kind of who has taste, or like how to develop taste, I think this, it's very hard, right? Like, taste is such a rare skill because I think it takes this combination of like seeing many things and having that exposure in whatever domain it is. And I think it's impossible for every human to develop taste in everything. There might be one area you pick that you're like, okay, I'm going to obsess over this specific thing. And like, for example, half my family are like architects. So the side of interior design they're obsessed over, they have great taste in. They might have less good taste in code because they didn't spend their life like perfecting that. And so this process of like developing really strong opinions and intuition for like what is good or what is right when there isn't clear right answer. It takes a lot of work and trial and error. We have to find the people that have it to help us go and then train these models and determine what good is. I think a lot of our challenge now— so we work in two ways. We work half with the frontier labs on how do we benchmark their models, eval their models, understand where they're breaking, and figure out exactly how do we pull them more out of distribution into this direction of great? And how do we both fix it from a capability perspective in terms of more like mistakes that they're doing or inability to do things like vision or 2D reasoning, which are more foundational, but then also these higher-order things of creativity and style and aesthetics. And then on the other side is like, okay, great, let's assume we're making these foundation models better. How do we work with application layer companies and agent companies that aren't necessarily touching the model layer, but that want to still make the outputs better? So what can we control there? And I think a lot of our answer there is both context of like, how do we help the generation be better before it even starts by like feeding it the right context? So most users nowadays that are using things like, I don't know, Lovable or Cursor or Codex or Claude or Figma Design to create are not experts, right? They're not designers professionally. They're regular people just wanting to make interesting, great things. And of course they're not going to know how to express what they want in the exact right terminology or Write really good prompt text.

You know what kills more enterprise deals than bad products? Security concerns. One question from a customer's IT team and suddenly your whole pipeline is frozen in place. What happens if you get hacked? How will you protect our data? But relax, there's no need to panic. This is why you brought in Wise Security. Wise Security is your on-demand security team. They're providing you with 40+ engineers who've been in the trenches at places like Apple, Uber, Microsoft, Robinhood, and Brex. Hey, I invested in two of those six. But You're not hiring a full-time CISO at $400K a year, are you? Nope, you rent their team by the hour. Real experts embedded in your company, getting you through SOC 2, ISO 4200, whatever compliance hurdle is standing between you and the deal. And you can set up a monthly cap so you know exactly what you're spending. No surprises, no bloated retainers. Plus, your first 6 hours are completely free. So here is your call to action. Head to wisecurity.io/twist and book your free 6-hour strategy call. That's ysecurity.io/twist. Once again, ysecurity.io/twist. So does that put you in the category of company like our investment in MicroOne that provides, you know, datasets to the large language models or, you know, are you doing something different? The data labeling annotation companies, are you like those or are you something different? How do I put you in a box and understand this?

It's hard because we kind of don't like boxes. I would say we like, so half of our business maybe it looks more like that, like half is more like data, RL environments, but the other half is tools, it's API. So I come from XOP before I started Tase, and so I love this infra side of like, how do you build the tools for agents? What needs to be true in the world? What needs to exist for agents to do great work? And we're also building those very opinionated tools and products., that look more like infrastructure or like API-shaped things. So yeah, it's, it's both multiple things at once.

I was going to ask you a question, Lan, here, because I think one of the things is an observation I've had is we used to have a curator class, and the curator class were restaurant reviewers. We had— Lan was part of it. I was part of it. We had incredible, uh, reviewers of movies, um, Elvis Mitchell, uh, A.O. Scott, And then who was the one from the '80s that like Brett Easton Ellis was from the New York Times, who was a very famous movie critic in the '80s and '90s? You can look it up on Claude or something. Ask producer Claude.

I'm not sure, I'm thinking of a few different ones. I'm not sure which one you're talking about.

Maslin? Something Maslin? Oh, Janet Maslin, sure. Janet Maslin was like a real critique. And then you had the New Yorker.

Pauline Kael, Roger Ebert, the list goes on.

But these were intellectuals who made it their life's work you know, to go to restaurants, to study food, to study art, to go to art exhibitions, to study the opera. Sure. And they have been hollowed out along with the newspapers and the magazine class. And what's the first thing to go? You know, you're like, hey, listen, there's other people who can review books, there's other people who can read these things. And so they just got gutted. And social media.

I mean, now that I can just go on X and see what does everybody think of the new Spider-Man movie? Maybe I don't need a guy who saw it early to tell me in a 200-word review.

That's my question for you, Elon, is that's been lost, right? Is there— because when you go onto X or Instagram, like, I haven't replaced that function yet. Is there— is it been— is it still chaotic and they're just all losing their jobs?

Yeah. I mean, I don't feel like we have figured this out yet. Everyone I know who used to work in media as this kind of cultural critic or—

Curator class.

Curator class. Curator. Those people are all struggling to, to find work. And, you know, here I am. I mean, honestly, I'll just talk— here I am talking about AI and tech with you lovely people right now, not movies, not pop culture, because we sort of phased that, that function out. I don't think we have replaced it with anything solid. I think people just now use AI or Rotten Tomatoes or social media as sort of a— or TikTok.

Well, this is an opportunity for you, Tavis. Why don't you just like— we've got an incredible company. You've got incredible customers. Maybe you could create this class by hiring these folks, but then we lose them in the public and they're trapped in some AI box somewhere. So what's your thought on this creator class? Because I'm going to give you my secrets of how I very quickly figure things out. I'm going to give you my stack and then I want to compare it to your stack and your process. Um, so you hire humans, right, to, to help you?

We do. So I was going to say, we actually— and we call them tastemakers. So we have this awesome community of about like 1,000 tastemakers of different domains, like different specialties in terms of design, different media types, different styles. And we purposely wanted to bring folks from like very distinct style buckets too, so that we can have that diversity. And I think it's worked in a really good setup because you have folks that have, to your point, spent years like honing their craft, developing very strong opinions, developing that like critic sense. And honestly, a lot of what they do with us is literally like being a design critic. And that they get to both be a part of this and be a part of changing it. I think They have the same frustration as we're talking about in terms of slop and being like, I don't want to live in a world where all of it looks the same and feels the same.

And these are paid. They get paid $50 an hour, $100 an hour, or something like that?

So we pay them, of course, for their time. But I think we've also been exploring some more interesting ways to collaborate with them, both involving them more deeply in the company. Some folks have been joining full-time, actually. Oh, great. And we've been testing some different things. So even partnerships—

You're testing this new concept of equity for— Poor, poor person. No, not that.

Like we're testing almost like allowing them to contribute their designs to our index. Because part of it is like we're building this design index. So if they can like contribute their portfolios or things that they've done in the past.

And get credit when the LLM does output?

Not necessarily that, because that's like a hard attribution, but more so our search engines. So this like API product that I'm telling you about, that's something that we're testing or yeah, having these assets also be a part of like lab projects. And so it monetizes a little bit differently.

Yeah, I suggest money. For sure, for sure. Money seems to work pretty well for this class. All right. So—

I have one. Can I take it in a little bit of a philosophical direction? Sure, why not? Let's do it. So to me, when I think of taste, when I think of like the people in my life who I'm like, well, they have great taste, part of it is the uniqueness. Like, well, they told me about a band I'd never heard before. They turned me on to this obscure TV show from the '70s that was so cool. And so I wonder, once you're commodifying taste, what you're saying, we're gonna teach this AI how to have taste, doesn't that go away? Don't you lose that? And now the AI is recommending the same kinds of things to all sorts of people and it loses that rarity that made it special.

I think that would be my, my fear. That's what I think we're trying not to do. So I think it's more so about execution of us doing it well. Okay. Uh, but I don't, yeah, that, that's definitely not the goal. I think, 'cause part of it right now is the bars are so on the ground and the first job is just let's lift the bar off the ground and even get to a place where these, models can output like high-quality things. Then I think it becomes more a problem of like personalization, but I don't think we're there yet. I think it's like we have to first fix just quality. And I think there's elements of quality that more people would agree on in terms of just like craft and polish. And then there's things that people would more disagree on depending on like style or fit or audience. And sure. But yeah, I think quality is the challenge first. But yeah, definitely don't want to end up in that scenario.

So this is an interesting philosophical thread and let's keep Going, pulling it and see where it leads us to. If everyone knows in 1982 that U2 is this really cool band before, you know, the Joshua Tree comes out and, you know, then all of a sudden everybody knows about it, we kind of all feel like it's been commodified, the knowledge of it. Right. So it's a never-ending task. And, you know, this is happened very acutely in Japan. I go to Japan twice a year. Have you been, by the way, to Japan?

No, I actually haven't.

Oh, I mean, if you're into— if taste is your thing, I brought Lan for the first time. It's neat. If taste and refinement and kaizen, the constant improvement concept that comes from Japan, is like— gets you excited, oh my lord, you're gonna— it's gonna blow your mind to watch people kind of just operate in the world. But there's a problem. TikTokers, Instagram influencers. I can't stand these people now because what they do is there's places I've been, like I started talking about Niseko where I skied like 7 years ago, I guess, and now everybody knows about it. I can't get a hotel room, I, I can't, uh, get a reservation. Then in Japan, somebody finds out that this noodle shop that's 150 years old is like the best one. And then everybody who goes there feels the need to go there and TikTok it. And now those places have crazy lines. Some of them now, and I'm making the X here, this is what the Japanese people do when they say no, they don't allow Americans, no tourists. So they change their menu all to Japan. They change their Google listing, all Japanese. Oh my God. And then they make it hard to find the place. So they make the signs. And this you may have experienced.

Well, everything in Tokyo is hard to find. Everything, you go to the building and you're like, I'm at the right building. It's like, oh no, it's 3 floors underground. Or like, oh no, it's at the top of this thing and you gotta go around and take a hallway.

Now imagine they put it, they put it in Japanese in a tiny sign around the corner when they used to have it in the front.

I was the dumbest, most oafish person in every room I was in for 2 weeks.

It was crazy. Yeah. AI is writing code, reviewing that code, and then shipping that code. But it's important to remember your workflow is only as fast as the slowest thing that it touches. And for most of you, that's gonna be your database. Hey, with today's tools, you can generate a new feature in 90 seconds, then spend 3 days bending it to fit your rigid decade-old schema. But this is where MongoDB can help. MongoDB gives you the flexibility to ship at the speed of AI. The asset compliance guarantees you get to actually sleep at night while it scales to handle massive Fortune 500 workloads. And the best part is developers swear by it. Literally. I can't use the actual words they said in this ad, so let's just call it a really great database. The MongoDB for Startups program is here, and it makes it easy for eligible startups to get started with free credits. Apply now at mongodb.com/startups. Once again, that's mongodb.com/startups. So your thoughts on this particular issue, which is, yeah, you know, there's, there's a half-life to cool.

I think yes, but I don't think that's a function of AI. I think that's always been true, right? You've always had this like class of curators or like the tastemakers that are at the forefront. They kind of pick the thing that is the cool thing next, or they're at the avant-garde, let's say. And once it starts getting more ubiquitous and like more people start knowing about it, it loses the cool and then they move on to the next thing. But I think that's always been true, even before social media, even before AI. I do think though that social media like accelerated the speed of that cycle, let's say. And AI could have that potential too. I think so far it honestly hasn't just because it's so, it's still so bad that I don't think like the actual tastemakers are being like, oh yes, this is the exact thing I want my sites to look for or to look like. And so I don't think we've seen that effect yet, but I think as long as we push for that diversity where it's not meant about it having like one correct answer, but it's about it having both breadth and the ability to execute well in like all those different directions. That I think actually could help the problem in a good way, like of pushing people towards actually more directions rather than compressing to the mean further. Let me tell you my stack. Tell me. Yeah, I want to hear it.

I don't want— I'm reticent to share this because this is going to cause the other problem I have. Okay. I have a number of curators that I trust. I'll give you but one. Monocle magazine is a magazine out of the UK. Have you ever read it or seen it?

Yeah, yeah, yeah.

Okay, so you should subscribe to and support them. And so I was having my Athena assistants last year take Monocle and about 15 other sources where I find things and specific searches I use on those places. I'm not going to say anything else. I gave you one. Then I would tell them, hey, aggregate what I'm looking for when I go to Singapore. I went to Singapore for the first time. I like design hotels, I like hip, I like funky, etc. And so if you were to put in Singapore here, put in Singapore hotels, and this is how I picked my hotel for Singapore. It's beautiful, like cyberpunk hotel that had like trees on the side of it. I think I showed it to you, Lan, if you search for Singapore hotels, and I find it there. And yeah, let's see if it's there. When you pull it up, it's like a garden hotel that's like really tall, and it looks like something out of like a positive version of Blade Runner. Um, anyway, you're just seeing like Singapore pictures. That's not it. But there's like a ton of things that are like this where they make it like a rainforest. Um, incredible. Um, so then, uh, they did a really good job of finding the places I like to eat food. Which I'm looking for things that are, I'll just say, Anthony Bourdain-esque, you know, street stuff, one dish they're known for, etc. So then I programmed my own Cool Hunter and I said, you are a Cool Hunter. You're looking for things that would appeal to these specific people, you know, in these categories that have these specific design elements, etc. And it works.

That's all I'm saying is it works. I wanna test it.

But it's a super mega prompt. And then I have it go out and I have it cool hunt for me over and over and over again. And so every week I get a cool hunting report of like trends, et cetera, just 'cause I'm like interested in the world or whatever. So how do you do it? You actually get to hire people to review output from LLMs. And then have them rate it on some vectors? So there's a mix of things.

So there's rating things. There's also this concept of almost like a critique. So imagine if someone's actually like a design critic or a senior designer who's teaching a junior designer on their team. So how would they almost critique every single thing they're doing and explain why it's wrong or good? There's also this aspect of actually curation. So choosing examples of things that are good and choosing examples of things that are bad and explaining why. And yeah, or also actually creating things from scratch of like, hey, here's an example of what a perfect version theoretically of this thing would be compared to what an LLM was able to generate. So yeah, there's all these different data shapes that are helpful in different ways for model training. And then they are also helping a lot of these internal products that we're building. So the index, for example, is something that had a lot of curation from experts. And this judgment system that we're building of how do you verify also if something is, for example, staying on brand. Involved a lot of experts, like, making those judgments and be like, hey, this—

how do the experts feel about this work? Do they have this sense like, I'm, um, giving away my curation and what makes me special, and I'm going to get myself out of a job? And then how do you handle that? Because AI is scary to people. So if you came to somebody who's a movie reviewer and you're like— or you're an art critic— and we're like, we're going to show you art, we're going to give you the provenance of it, and we want you to describe it, like, are they like, are you trying to take my art critique and eliminate me from the world? So they must be thinking that on some level and you must have an answer for that. What is that answer?

Yeah, I mean, I don't know. I think it, by the way, I don't actually fundamentally believe that tastemakers are gonna go away by any means with AI. I actually think it's the opposite. I think they become more valuable because the more you have of this like mass production of things and kind of creation of outputs, which is just going to happen, right? Because everyone now in the world can click a button and create something. The more I think the peak of the peak of the peak becomes valuable, similar to what you're describing about the specific restaurant in Japan, right? It's like those really specific things that become really good, I think become more valuable, not less. And so, and I think we're infinitely far of having the power of like a true like creative direction and choice and this kind of craft. And so I think, yeah, being a tastemaker kind of becomes more valuable in terms of like how they feel in terms of the work. We actually ran this survey with the community a couple weeks ago of asking why you're here, why are you excited about working with us? And it was really cool to see how a lot of responses were almost around curiosity. I was like, I am excited about the problem and I hate SOP and I want to help. It was very motivational in terms of purpose of why they were interested in the work. And so I think people want to be a part of the solution maybe as the conclusion I've gotten.

Interesting. Yeah. That would make sense to me. Here's the Artisan Hotel I stayed at. Check this out. I'm gonna say this is gonna blow your mind. And so like, I stayed a little bit off the beaten path, but this place was kind of new. I think, I think my room was like only $500 or $600. But if you zoom in on this thing and you, you'll see there's like cutouts where it's green. Those are actually double story green atriums, Lon. Oh. And so those are seating areas common to everybody. Like you can't get them as a suite. Right.

It looks like the whole place is great. Like, are there rooms in between the jungle?

Yes. When you look at the jungle area, like, if you look at any jungle area, there's rooms to the left of it, there's rooms above it, like in typically in like 8. And yeah, um, it was— and then on the roof there's an infinity pool. Go to the roof there. That whole area up there, that's an infinity pool type situation. And the— there it is. So you go up to the roof. I would go up there and work, um, And I would have never found this if I didn't have my system because it didn't come up anywhere. It's not like this is a giant hotel. It's got like whatever, probably like a boutique hotel. Yeah, it would be in the boutique hotel, not a chain. Artizen. Art-y-z-e-n. You know what it looks like? I find these things so consistently.

It looks like in the post-apocalyptic movie when nature has reclaimed the building, but people are still like huddling in it like children of men, you know?

Check this out. Yeah, uh, look at the, uh, cutout there. That's really cool.

Yeah, by the way, half my family are architects, uh, so I feel like the side of like interior design and, and hotel hospitality, like, I, I love. I gotta give you some Brazil tips for next time you guys go. You know, I've never been.

Uh, Foundry University, São Paulo or something. Let's go.

São Paulo has a huge, huge like, uh, founder community.

There you go. Make that happen.

So do you have aspirations to do anything public-facing with this collection of people, or have you thought about buying— and this is the idea I had— if I would, because I think you raised money at a pretty decent valuation. Yes. Yeah. Yeah, we did. Do we have that in the notes, Lan? We did.

I have that it was a—

and who the investors were?

Yes, I have an $18.5 million seed round co-led by CRV and Amplify. That's what I have.

Oh, CRV. I know the team over there. Oh, great. Congrats. Founders are scrappy by nature. And many like you like to dig in and get started building without worrying about admin. They run the whole business on Excel, then keep adding on more random tools as needed. But here's the problem. None of these tools and platforms know how to talk to each other. So you wind up playing go-between with a cluster of different systems. It's a Franken-business. It's a Franken-site, and it's a huge waste of time. But now there's Odoo, the all-in-one management software that's serving 16 more than 170,000 companies in over 120 countries. Now your CRM, sales, accounting, manufacturing, website, inventory, and point of sale, they're all in one easy to find place. And most importantly, every app talks to every other app. Sell a product on your website and the invoice gets created instantly and your inventory gets updated automatically. So if you're stitching together spreadsheets and keeping track of 5 different logins, Get started today at odoo.com/twist. That's odoo.com/twist to level up your business. What it would cost to buy Monocle Magazine or to say I want the exclusive license to Monocle Magazine would be a fraction of what, you know, these large language model frontier model companies or even like the mid-tier ones. Have, have you considered maybe acquiring some of those assets and putting them together or the archives of them? Like Paper Magazine got sold in a fire sale, I think sadly. And that it was just like this incred— I used to I used to write for them back in the '90s. Like, incredible archive of, like, history and art and taste. Like, somebody could actually own that and, oh, and get the writers involved in one of those. And it could be like the rebirth of the magazine business, which was the ultimate coalescence of taste, of curation. Yeah.

So I think right now we're very focused on, like, visual design. And so, um, yeah, we're thinking— we're very— going very deep on the visual design side of things. And so we've been more doing these types of relationships actually with like creative agencies, uh, or like epic designers that we want to like take a look at what kind of never saw the light of day in their portfolio that they maybe worked on. Oh, very nice. Um, but yeah, I think when we eventually get to something like creative writing, it would be really cool to explore some of these, these partnerships because to your point, it's like, it's one of one, right? Like you don't have, you don't have 50 magazines. And, uh, so yeah, totally.

All right. Listen, this has been incredible. Uh, since you just raised money, I know you're hiring. What positions you're hiring for? Where's the company based?

Every single position you can think of. We're based in San Francisco, but I think we're hiring a lot across like engineering, research, operations. And we actually have this really cool like design researcher role. So if you're a designer, but that really wants to think of the problem of design almost as like a theme to be studied and how do we like break down what does good mean and evolve this, then yeah, that's a really cool role.

Sounds like you'd hire a professor. The website is Taste Labs. And yeah, look at the careers page. A lot going on there. Hey, just give us an update as we wrap here. I know San Francisco is on fire. I might have to— I have a loft that I rented and it was like underwater for 4 years, but now I think it's probably a good time to sell it. What's it like operating in San Francisco with no housing stock? And, you know, people— Yeah, say a little bit. Explain to the audience, people who aren't there, exactly how crazy it is and hard it is for employees that you try to hire?

No, it's nuts. Like I lived in New York before, so New York also was not easy by any means, but no, I've become very SF killed and I love it. And we are very like in person. And so a lot of folks have moved here, but yeah, it's like you have to show up to the apartment visit with already like all the documentation to like pay them upfront. Crazy. Get it. And like, it's insane. And so yeah, I think this, the housing side has been tough, but The energy of the city really is incomparable, and it's, it's very worth it here.

So we have this great thing in New York City, it's the subway, and then we have boroughs. And so you can live in Flatbush and take the subway for an hour and get a $2,000 a month apartment and live a little closer, and it's $3,000 or $4,000. But in San Francisco, I mean, what's a one-bedroom now, $4,000 or $5,000? I don't even want to think about it. Bonkers.

Yeah, it really, it really is bonkers. But, um, I do think that this like concentration of people in one place just creates the type of magic you can't reproduce.

I had, I had a dozen great years there and now I'm in Austin and love it there. But, but it's something you should think about. Like imagine if you were running the, in, you know, Austin, you, the housing stock has gone down 3 or 4 years.

They're like, hey y'all, first 3 months free, no rent. Come stay here. That's what it's like. Crazy. They built it.

So much inventory, it breaks a New Yorker or somebody from the Bay Area's brain when they first move there.

Jacob looked it up. A one-bedroom in San Francisco costs an average of $3,500 to $4,600 per month right now, Jace. That's what— yeah, that's our latest.

If you got in for $4K, you'd feel lucky. All right, listen, continued success, and, uh, we'll check in with you in a year.

Thanks so much, Chase. Yeah, that, that's right. I'm curious, after design, I wonder if they're gonna try to go comedy, because I feel like that's the one area where We all know LLMs aren't funny. They're not good at writing jokes. Everybody seems to have sort of given up. Terrible at Twitter threads.

Right. Like they don't understand Twitter. Right. I mean, Grok is on Twitter and it kind of understands it a little bit, but not good enough to hit public.

Grokopedia can be very useful, but Grok is not funny. And that's what I mean. Like, I feel like we've all kind of even given up. Like, it's never going to be funny. It's never going to write a good joke. But I don't know. That's also a taste— Good for brainstorming, though.

Great for brainstorming.

Great for brainstorming. Useful in a million ways. I'm not trying to go there. I'm just saying this one area, like write me something funny. You can explain jokes. You can feed it a million jokes. Claude isn't funny. Claude's many things.

I'm really feeling, uh, good about my annotated.com. I bought this because I love the bookmarking and curation. Um, but we're gonna have to— I think I should launch annotated. As instead of my bookmarking service, like, have that available. But I think I should let anybody put a task up there to rate stuff and fact-check stuff. So you're collecting data? Correct. Like, I didn't think about it, but like, it all is data collection. So if consumers did this, and then you could sell it to the— nothing— Micro Ones or her—

nothing more valuable right now than preference data. That's, that's the goal.

You could also put it out there, like, if you had annotated.com hotels, and you could just send to people who were hotel aficionados, hey, what do you think of this offering? What do you think of this price? What do you think of this room design? What is this missing? What's good about it? And you could actually get them to maybe vet other people's curation. Anyway, uh, we're gonna do it as a bounty this week in startups.com/bounties, uh, and we should put some fresh dates on that, that maybe we'll push that contest to finish in September.

There was some travel, there was some delays, but now that you're back around and in the mix, we'll definitely definitely get that bet to go.

And the software is great right now, right? Yeah, software is great. So I want to build this app. Um, you know what I'm going to work on is try to get somebody who has a tool like a, you know, Perplexity Computer or Claude Code or Lovable. I'd like to get one of these people as my partner on this and put up like some serious bounties. That's what we have to do next. There you go. Uh, but go ahead, check that out. All right, uh, we got a lot of news to do.

We got so much news. All right. Well, we got to talk about situational awareness or lack thereof. So famed AI-focused hedge fund was led by 25-year-old former OpenAIer Leopold Aschenbrenner. It was founded in 2024. He started out with a few hundred million, Jason. He amassed over $45 billion in assets under management by July. In June, Wall Street Journal said $20 billion. So it went wildly up. The fund had a 439% net return through June 30th. He was placing very large leveraged bets no professional investing experience prior to launching this firm. Early backers included the Collison brothers from Stripe, Daniel Gross and Nat Friedman from Meta, many others. Investments included SK Hynix, Micron, SanDisk, Bloom Energy, CoreWeave, Nebius Group. There was a— there was a paper that he wrote in 2024, also called Situational Awareness. The argument, the thesis was basically like, we know AI is going to explode. We know the demand for compute is going to increase exponentially. So he was— let's place our bets. He was placing bets in the infrastructure and the hardware that was going to allow for this buildout to happen. So then what happened? He compared it to a bank run. He got margin called by a number of his biggest investors, Goldman, BofA. And so eventually he had to, in order to access cash,. He sold the public portion of his stock portfolio to Ken Griffin's Citadel. Everything that had been financed with borrowed cash is gone. However, private investments he's still holding on to. It's going to just continue on as a fund for his private investments. That includes what Financial Times tells us is, uh, $5 billion of Anthropic stock. Uh, so TBPM this morning published the letter Leopold sent. To his LPs this week.

Yes, we took ownership of it.

Yeah, we let you down this month. We came closer to permanent capital impairment than is acceptable to us. Uh, he took full responsibility. Uh, he said vulnerability was begetting more vulnerability, and we took decisive action to protect LP capital. Uh, that's why they will carry on with the Anthropic shares and other private investments. Got a question for me at the end there, or just gonna drop it? Oh, I, I'm sorry, I thought you were gonna jump in. So I guess my big question to you is, this kid is 25. He did not have investing experience before this. Do you think that a more seasoned pro would have seen this coming and been able to navigate around it? Or is this a classic no crying in the casino? You take a big risk, there's a chance you're going to get crashed out.

Could have happened to anyone. No, this is, I guess, the folly of youth and a lack of experience for sure that played some sort of a role here. He was running, according to the reports, something like 4x leverage, which means for every dollar he held in a share, he was then ballooning it up by 4x. So if the market goes down 25% or 50%, you basically get 4 times that kind of, uh, you know, volatility, I guess, and downdraft. So that was a mistake. And then there were other things at play here. This— the South Korea stock market was having a crazy run where like a million different levered accounts. Again, having leverage, a.k.a. debt, is— it's different when you have a home and you have debt than when you have equities or when you have crypto. When you have a home, they have to foreclose on it, which means they got to kick you out of it. It's a process. It's emotional to kick somebody out of their house, and there's laws around it, and it's a slower process. You can miss some payments, you get some penalties, you can do a partial payment. What happens when you're on margin, you know, either for crypto or for stocks, is you get margin called. What does that mean? Whoever your banker is that gave you that loan then starts selling your equities automatically. Now, they may give you a correct period as a grace, you know, in a grace period. Hey, you got 24 hours, 48 hours to put up the collateral. But generally speaking, they will just liquidate you. This happens in Bitcoin all the time. People lever up their Bitcoin. You know, they own $1 million in Bitcoin. They're like, okay, give me another $500,000. I'll buy another $500,000 in Bitcoin and go long. I'll pay 7-8% interest on that, but I believe Bitcoin's going to go up 100% this year. Then Bitcoin goes down, they got to sell Bitcoin to cover it. Yeah. And those lenders— so that's what's like technically happening here. Schadenfreude is really, really high in finance, especially in hedge funds and stuff like that. Have you ever watched the TV show Billions? Sure. There is a pile-on that occurs. So he kind of outlined this, uh, in his, um, letter. Where when this occurs, people know that you're, you know, experiencing this downdraft and that you're going to get margin calls. So what do they do if they have the same positions? If we all own— let's say it was, uh, pick a public company that's in AI— NVIDIA, NVIDIA, uh, Micron— you and I sell all of our position so that it makes it go down more. So we're flooding it with stocks so there's more supply available to put pressure on them. Then we short it, we create more pressure and just more pressure, more pressure. And then at the same time, you're talking to that person about buying their position. So Citadel was talking to them about buying the position. Other people— and I don't know if Citadel was involved in this, but this is what they do on Wall Street. They are just like Knives Out. Somebody starts falling, you just push them in the back. Like, if they're stumbling, you don't want them to recover. You want to buy the asset from them. I will tell you this.

I mean, I'm worried about conspiracy theories that Citadel sort of purposefully did this. They were predicting a surprise rate hike from the Fed, which historically would have tanked the market. That did not materialize. So people are coming up with these theories that, oh, this was all sort of like purposeful. I don't— I don't know how you feel about that. Seems a little far-fetched to me.

This is not SBF, Sam. Bankman-Fried. This is not the same situation. Sam was making great bets and he would have been one of the great bettors of all time because he also bet on Anthropic, etc. And they wound up selling his positions for peanuts when they liquidated FTX. Right. He was stealing money.

Allegedly went to jail for it. Yeah. Yeah.

He was involved in real fraud where he's taking people's assets that they put in their essentially bank accounts, in their accounts, and he was investing it. Right. And giving it to celebrities. That would be like Bank of America goes into our accounts and then sponsors the Super Bowl and invests in, you know, SpaceX and Uber with the money.

He was doing a lot of crypto exchange and the investment fund and intermingling funds in ways that the SEC says you absolutely cannot do.

But he, you know, this young gentleman is obviously pretty bold and getting your ass kicked this publicly, this brutally, is going to either break him And I was a little concerned with the pile-on. I'm like, this is the kind of thing that could psychologically break somebody and they commit suicide.

People were taking a little bit of glee in it. I mean, he's like, he's a very young— schaudernfreude. Yeah, he's a very brash guy. He's a good-looking guy. I think there was a little bit of like— I don't know if he's good-looking, but maybe you don't think so. Pull up, pull up one of the memes. Pull up.

Anyway, the memes are coming, but I think it's also a slow news day. Now wait a second, come on.

That's him as Mitch McConnell. But that, that, that's not cool. That's an objectively good-looking guy. You don't think so?

I mean, sure, like as a German or Scandinavian, whatever he is. Sure, preppy German guy. Sure, I mean, it's okay. Listen, I'm nice, not Brad Pitt.

I think people were taking a little bit, I think people were taking a little bit of excessive glee in this guy. Too much glee.

He does have a lot of layers in that photo, so like layers are for players. I'll give him, the layer game was pretty good there. Look at the layer game in the brown. Yeah. He definitely had the layer game going. He was playing the role. Looks like clavicular. He was, yeah, I don't know what he was, he was leverage maxing. He was leverage maxing. Anyway, the lessons here are— Spiked his cortisol levels. Yeah, he was cortisol maxing for sure the last week. But this is not an example of like something fundamentally wrong with the industry. If you have leverage, it's just not something anybody with any experience does at this level. If you want to have a margin loan against like 5% or 10% of your stock, so you got like a billion dollars in stock and you have $50 million, you're not going to get called until it comes down to like 1 to 1 or something. Like, you have plenty of room, right? This was the opposite. This would be like somebody who's got a billion dollars trying to put $4 billion to work. Very, very dangerous. Um, but he's going to come back stronger is my prediction.

Yeah, if it doesn't break him. He's got $5 billion in Anthropic. Still he's sitting on, plus other investment. I, I wouldn't be too worried about this guy.

I think now there were other memes, so I don't wanna pile on, but I do wanna see the best memes.

Uh, Jacob's got a few more memes for us to look at.

Okay, let's see. Yeah, this is surprising memes for me. I'll put my glasses back on for these. And I think it was a slow news day and then people are kind of sick of AI and nobody likes a young successful person.

Exactly. I think to, I think to me it feels like this is bubbling over. People just dislike AI and they dislike AI people. And this young punk was trying to make all these billions of dollars placing AI bets. I think that's what fueled the backlash. Sure. All right. Here's the next one. Next one, Jacob.

Let's take a look. When you beg a client to accept a great settlement offer and they end up getting zeroed out at trial instead. Okay. Not funny. Not that funny. Next one. Not funny. So your account got liquidated even though the S&P is only 3% off its highs. Okay.

Jonah Hill from Wolf of Wall Street there.

Yeah, he was great. He should have— did he win? Did he get— did he get nommed?

I believe he got nominated. I don't think he won, but he— I believe he won.

I wonder who he lost against. I'd have to— I think Wolf of Wall Street, by the way, just not to go off-duty, I don't want to pre-off-duty this, but I think during that moral panic when the Overton window was really closed worked against this film. He was— that's my hot take.

Uh, yes, he was, he was nominated. It did— that movie didn't win any Oscars.

Any. Any— and he— and this— I was at the year of like Moonlight and other woke stuff.

Uh, Moonlight is a terrific movie, but I, I will—

I will look up what I felt like. I literally fell asleep 4 times. You're wrong.

You're wrong about Moonlight. You're wrong about Moonlight.

Okay, I thought that was, uh, that was La La Land, Moonlight year. 2 terrible picks. Go ahead. Yeah, no, let's see.

Uh, Best Directing went to Alfonso Cuarón for Gravity. I'm on the Oscar Oscars. No, of course they don't have gravity. No. Yeah, uh, Matthew McConaughey won Best Actor for Dallas Buyers Club. Oh, here we go, Jared Leto beat him. Jared Leto won for Dallas Buyers Club. Okay, it's a perfect example.

I listen, actually, I thought Dallas Buyers Club was great, and those two got— well, I thought those two guys had great performance, let me say that. But I do think Oscars go— if Oscars are given a choice, a bunch of debaucherous '80s cocaine-taking stripper-paying debauchery from the '80s or '90s versus people dying of AIDS. Yeah, well, that was, you know, in a tragedy. We know what the Oscars are going to go for.

There was a lot of weird discourse around Wolf of Wall Street too. Like, is Marty praising these guys? Is it pro-Jordan Belfort? There was a lot of that stuff that might have heard the same thing with the Goodfellas.

It's like, are we praising these guys or are we just telling a story, folks? It's just aesthetics and a story. And if you look at the test of time You know, Dallas Buyers Club, Moonlight, whatever, Gravity. Nobody's gonna remember those compared to Wolf of Wall Street in 20 years. Gravity's awesome. Gravity is so good. All right, it's okay. It's not— you, you're telling me— wait, hold on, who's your Best Picture then of 2014?

Let me see. All right, let me, let me, let me look at the 2014.

I'm gonna force you on this one because you know I'm right. You would—

you're definitely 2014 Best Picture nominee. Run me through this. I might have been a— I might have been a Wolf of Wall Street. Let's see.

Of course, Gravity over Wolf of Wall Street. Nobody even remembers Gravity.

Who's in it? What is— what?

Sandra Bullock.

Oh, you know, you know what? Google's, uh, AI here, really awful. Like, you, you would think this is such an easy thing to look up, like, what were the movies that were nominated? All right, here are your 10 Best Picture nominees for this year. It was the 20— 2013 movies. It was the 2014 Academy Awards.

Okay, go ahead, give him—

hit me. The nominees that you want to hear the winner last. I'll do the nominees first and then I'll—

yeah, yeah, sure, sure, sure, sure.

The nominees were American Hustle, Captain Phillips. Remember, this is the 10 nominee years. American Hustle, Captain Phillips, Dallas Buyers Club, Gravity, Her, Nebraska, Philomena, The Wolf of Wall Street. And your winner was 12 Years a Slave that year. 12 Years a Slave won.

You know, it's just— that's not— it's this— there is no contest there. I mean, I did like Nebraska. I thought artistically great. I like our films. I have to say, I would say to me, Her and Wolf of Wall Street, it was between— yeah, that's a jump ball. That's clearly a jump of those nominees.

It's between those.

But you know, this is just the problem with the Oscars is they go for max suffering and, you know, whatever, put your hand over your heart. And yeah, the most—

Her did win Best Original Screenplay that year.

Yeah, this is why the Oscars are meaningless to me. I just hate the Oscars. Uh, okay, let's keep going through the news.

All right, we got— yeah, we got to talk about this. So Google Earth Earth. Alphabet embedded their AI image generator Nanobanana into Google Earth. Yeah, Nanobanana, amazing. Google Earth, amazing. Now you can go to Google Earth and with a simple prompt you can change things, the way things look on Google Earth, visualize streets and intersections as they might have looked in history. Oh no, you can even make an infographic based on anything you can pull up in Street View. Uh, but people have already—

you got to describe what's happening here. You got to describe what's happening.

Okay, so here's an image you can do nefarious things too. Like, here is a spot along the US-Mexico border that we see from above. There's nobody really there. It's a totally normal image. But if you tell it, hey, put a huge refugee camp and people pouring all over the border— oh my God, you can create that.

Like what's going on in Spain, right?

Like what's going on right now in Spain. And, and, and so here is— it's Hank Van Es created this fake refugee camp near the Mexican border. There's a second example as well that we were looking at. This is he put in Gaza a bomb crater next to a fake hospital. This is a fake image, folks. This is not real. And so obviously the potential here for creating misinformation and visually appealing inaccurate things is massive, which led a lot of people, myself included, to wonder just why even do this? Is the utility— I'll tell you what happened here. What is the use case here? Like, Well, I'll tell you what happens.

No, no, it's very simple. Uh, there, everybody at that company is going, hey, AI is the next big thing. Sprinkle a little AI on everything. So Google Local, Google Flights, you know, Gmail, just put some AI on it. Just get a little AI on there.

Just a dab. Just a dab. It's that little dash.

Let's see. And then if it's good, maybe we just coat it in AI. Maybe we deep fry it in AI, whatever it is. So somebody was like, oh, would this be cool to share, to say, hey, here's my lot. What happens if I put a 5-story building over here? Which, like, on an architectural basis, I can come up with 100 uses for this that would be really, like, a city planner might look at this and say, hey, let's put bike lanes in instead of doing something nefarious.

The problem is— You could do fun stuff too. Like, to test it out, I pulled up downtown Austin, like, right where we produce the show, and I made it like, what if this looked like the lunar surface? Like, you could do fun stuff like that.

Okay, yeah, so there's goofy stuff, there's practical stuff, and then there's nefarious stuff. Now, anybody could take these images already put them in Nano Banana. They just close the loop here. I'll tell you what the problem is here. When you do close the loop like that, like if you were to take, I don't know, Instagram and then be able to on the top level say like Instagram this person and put them in a bikini, you know, which is what people do with Grok all day long.

And you get the images of Grok. Yeah.

Like, please don't make me do it.

Yeah. Grok refuses to do almost everything you ask. Because it got in trouble.

So this, this is the problem with a tech. Every time there's a new technology that accelerates stuff, you're gonna have these bad uses of it. And because it's adjacent, that does lead somebody to believe, or lead somebody to rightfully criticize, like, did you think this through? That's why you can't, to the best of my knowledge, Zuckerberg hasn't said on Instagram, you know, long press, you know, your Instagram post, and then I put in an AI prompt and repost it to my site or comments.

Yeah, well, we got a related story this week, actually. I wasn't sure we were going to get into, but I don't know if you saw LinkedIn added a report AI slop button. Yes. And they've gotten rid of the let AI help you write your post button that they put in that originally. That's gone. So there's no more like enhance your LinkedIn post with AI because so many people were just hitting that button and lazily posting AI slop. And I don't know if you saw last week, Substack added a bunch of anti-AI tools. And yes, Chris Best, the CEO—

this is going to be— this is going to be a big trend.

Yeah, of course. Chris Best, the CEO, in his blog post explaining the new tool was like, we don't want Substack to turn into LinkedIn. Like LinkedIn has become the definitional, like, this is a place where everybody posts.

Yeah, well, I mean, if you are a busy salesperson and somebody tells you Hey, if you write your 5 secrets to sales, you're gonna like get more sales. And so, you know, there is, you know, and then back to being AI being pushed by the boss, and I'm one of these pushy bosses saying, hey, let's use AI, let's use AI. You know, you are going to have situations where people don't do it thoughtfully, right? They're just, okay, I'll just produce slop. I've had that happen. I had team members here who are like, okay, Jake Howe wants us to use it. And they just make a full business plan that they haven't even read or they haven't you know, scanned and they presented to me and I'm like, uh, no, no, I'm paying you to write the plan.

I've had a few newsletters submitted that we were like, I don't, I don't think this is original.

This feels like Claude. Interesting you say that. So there is this company Pangram that we were talking about on Wednesday's show.

These are the guys who designed Substack's anti-AI tools. Yeah.

And so what Substack should do is when you actually publish it, they should tell the writer, hey, our AI's, you know, checker says it thinks this is a 92% chance it's, you know, not, um, you know, whatever, it's AI produced, you're going to be— you will not— because it's AI, we're going to put a label on it and we're not going to feature on the homepage. And this is what has to happen on social networks, is this stuff now doing this.

If they think you're posting AI, they flag it and they say, hey, maybe consider rewriting this so that it sounds more human.

Okay, so they do it at the moment of publishing. I think what they should do is if anything scores over over 50% on AI, they should gray it out in the feed, not put it in the feed. They should, what was the term? They should shadow ban it. Shadow ban posts, yeah. So anyway, this is good. This is a good moment in time. I like this. And everybody's gonna have to make a decision. Do we wanna live in an AI slop world or not? And it's screwing up my feed because I like to show my daughter, is really into sharks and I like to show her shark videos in my Instagram feed. And now they're all like, oh my God, here's a great white shark taking 3 bites out of a person. I'm like, that's never been recorded.

I've had this exact, my nephew, Duggie loves trains. And so I had— we had YouTube up and we were just pulling up, you know, videos of trains going by, trains pulling in. Trains are great. But who doesn't love trains, right? But there's tons of AI slapdash. So it's like fake train. He doesn't want to see fake cartoon trains. He wants to watch real trains. Like, it's terrible, terrible experience.

This is why young people hate AI. Like, my daughters, when I say, hey, use AI to find a place to have dinner, you know, I use Gemini and put it on a Google map. They're like, no AI, we don't want to use AI. AI is lame. Like, that's what 10 to 15-year-olds think right now. All right, let's keep going down the docket. All right, do you want to—

well, let's get to some of these rapid-fire items because I'm worried we're not going to get to everything. All right, so first one that you share with me, uh, so Ceuta is a Spanish-controlled territory on the northern tip of Morocco in North Africa. I pulled a screenshot from, uh, Google Earth, not doctored, folks. This is the actual image. Uh, so that's where it is. You could see it's essentially a small city on the North African coast right now. Next to Morocco, across from the Strait of Gibraltar from Spain, but it is considered Spanish territory. So what's going on is thousands of migrants are attempting to cross into Ceuta from Morocco. The border fence has been breached. 18 migrants or so— we're not— it's very chaotic, so we're not sure. They think about 18 migrants have died so far from either being drowned or being crushed in a stampede. Of people trying to get across this border.

I mean, it's not much of a border we're looking at here.

It's like a— it's a fence along an urban beach. So that's a beach. There's a border fence, but it is—

you could swim around it in 10 minutes.

You can swim around it. You can push your way through it. It has been breached. It's open now. Spain's government is deploying their Civil Guard to keep the peace. They have not yet declared a national emergency. Here are some numbers. Uh, the Spanish government says 60,000 people have now entered Ceuta from Morocco. It's just a city. The entire population is 85,000 residents.

So that's a one for one almost.

That's a 70% jump in their population since this time.

Was there an inciting event here that I'm— I am trying to catch up on this and my entire feed is venture capitalists retweeting this. Yes.

Elon Musk retweeted it. Like 20 times.

I'm like, dude, you're a venture capitalist. Nobody cares your opinion on migration. But I guess people have strong feelings about migration.

Elon is one of the people who really started this trend and made it sort of go bigger and viral in the US.

A lot of people are talking about it. Right, it's the number one account.

Yes. NPR, I looked it up. As far as we could tell, we don't know. There have been breaches along this border before. It's not like this has never happened, but we're not exactly sure why all of the sudden tens of thousands of people seem to have all gotten the same idea. They did— NPR did talk to some of the people who are are pushing their way across the border. And almost all of them were talking about economic hardship. There's no jobs. They're looking for—

Yeah, but it's coordinated. Okay, so if it was economic, this must have been coordinated on socials or group chats. So that's what's gotta come out. I think there's some psyops going on here. 'Cause like, you know, they're, you know how the CIA and other three-letter or other intelligence agencies from around the world do this kind of stuff to—

create this, you know, um, absolutely long history of the CIA creating civil unrest in other places for all kinds of reasons.

I mean, all kinds of reasons.

You could argue propagandistically that this backs up a lot of the kinds of stories that we hear, scary stories about migrant invasions and migrant crises.

And I mean, in which this is a literal migrant invasion.

Like, this is invading.

I've never seen anything of this scale. Like, if you have boats show up in Italy, you know, okay, there's 50 people on a boat, but I think you said the number 60,000. Like, that's like 4 Madison Square Gardens, 5 Madison Square Gardens. Yeah, that's a lot of people.

That's what the Spanish government is saying, that it's— well, and this is in the whole town. 60,000 are now trying to enter the town. And but that's the other thing is this is not— it's a Spanish-controlled territory. But even if you get from Morocco into Ceuta, you're just now in this city. You're not across the Strait of Gibraltar and in mainland.

Yeah, this has nothing to— I mean, in some ways there's nothing to do with Spain because this is a territory of Spain.

But yeah, it does also make you— if you want to get colonialist about it, like, why is there a little territory in Morocco that's still owned by Spain? Like, shouldn't this just be Morocco at this point?

I guess. I mean, we have that kind of situation. We have Alaska. I mean, sure.

Oh, we've got all kinds of little islands and stuff still.

Yeah, like Puerto Rico is what it is, but Yeah, I just want to say, if you're a founder and your VC is tweeting about this more than once, like, probably not the right VC for you. Like, I don't understand why VCs are spending their entire day losing their minds over this. Like, this is Spain's issue. You don't have to— this is my message to VCs— you don't have to comment on everything. Hmm. We don't need 20 literally one. You get one tweet on this if it's like the news story of the day. But like, I don't know, how do you, how do you make this go back to venture capital and startups and nothing to do with it?

There is one venture capitalist. I'm not going to name names. There is one. Don't name anybody. There is one VC that I feel like I don't know how this man has time to invest in companies because he is so unbelievably active on X. Constantly, no matter what the trend is, he's on top of it. He like way more than me and I make podcasts. I won't mention names. I bet a lot of you can already—

I mean, there's a number of them. I know which one you're talking about.

Jacob, Jacob already guessed in our chat room which guy I was talking about.

Hold on, let me take a look in the chat room here. Oh no, I was gonna go for, um, yeah, that's an interesting name.

I mean, I think there's a few, there's a few names up there at the list, but this, I mean, he, he's an amazing investor, an amazing founder in his own right. Like, not— I'm not trying to crap on the guy, but I honestly don't know how he has this much brainpower for X. It's unbelievable.

Two other guys. I agree with that one, but there's two other guys. You're—

we're all thinking about the same guy. Yeah, I can tell. There's a few of them. We're just like, how are you doing this? I don't know how you have time to attend a board meeting while you're— you must be in the boardroom.

I mean, if you're doing that and you're commenting on this stuff, I wonder if, if a founder is waiting for you to you email them about something, like, they're like, uh, dude. And you're like, anyway, I think this goes to show that VCs have a lot of spare time.

Okay, let's keep moving. All right, let's keep going. Uh, next up on our list, we have Ethan Goodhart's self-driving golf cart. Uh, he's a Stanford computer science grad, autonomy expert. He created his own self-driving golf cart. This is utilizing entirely cameras and vision models, no LiDAR. Uh, he's had some beta testers, so he's offering rides on his autonomous golf cart around the Stanford campus. Some of his early beta testers, we have a photo here, includes Sam Altman, Andrej Karpathy, Jensen Huang. They've all given it high marks. And actually there's an iPhone app if you want a ride around the Stanford campus. It's called Wind. It's in Test Flight right now. So they're getting ready to sort of open this up in a beta test situation so people can get automated golf cart rides.

Okay. So the reason I wanted to share this was number one, everybody hangs out at Stanford, obviously. And, you know, that's the place— these guys are good at intercepting them. Yeah. But I think what this shows is how commoditized self-driving is getting. And, you know, my thesis has always been 20 people would get there at the same time. I think my thesis is going to be correct because Tesla's at the, you know, double nines way most clearly ahead of everybody with more double nines. BYD, Baidu just launched, I guess. Nuru, Zoox just got 2,500 units federally approved for no steering wheel, which is really important. Now they still have to go through all the local communities.

Zoox is— it's those two-sided— it looks like a little mini bus thing, but there's no like front to the car. It's just like no driver.

And with no driver means no steering wheel, no paddles, no nothing. So I think, you know, if students can start to build this on their own and hack it together, doesn't mean it's like ready to compete with Waymo. But if students can start building this, what we're going to see is a plethora, a plethora, uh, an unbelievable number of people with these. So what is the self-driving race going to come down to? It's going to come down to scale. Who can build and manage and get licenses for this. And there's a news story, um, and you can look it up, and I just want to talk about— and everybody assumes I'm talking my book here with Uber, but I can assure you I am involved in— probably got investments in over 10 self-driving companies. A lot of them are public. And the reason I say that is I'm not just talking the Uber book. Um, there is very quickly going to be a lot of bad feelings about self-driving because people are gonna— the first wave of people losing their jobs to AI is gonna be drivers. And like, when I say wave, I mean like, they're, they're not— they're literally gonna lose their jobs and the pay is gonna go down.

And this would be both Uber and Uber Eats, like the people driving people, ridesharing and delivery.

Yeah, yeah. This week, uh, Stanley Tang, friend of mine, debuted the Zipline, which I'm an investor in Zipline, and I'm an investor in DoorDash through a venture fund I was in that invested in DoorDash. I've been a shareholder of DoorDash. They did their drone company, and we had another drone company from Ireland on the program recently here. So there'll be like 3 or 4 drone companies doing— would be 4 or 5 drone companies doing this. So that gets rid of like burritos or just coffee, small things. Larger orders obviously going to go in cars or small things. The reason I bring this up is in China, where, you know, they embrace technology at a pretty rapid pace and they have a top-down government, they have stopped allowing licenses for self-driving. So you can look that up. There's many stories about it, but China is freezing the number of self-driving cars on road. Why? It's not about safety. It's about protecting jobs. Right. What's going to happen? And I asked this on This Week in A— Oh, what did I ask on this week in VC? This week when we did the VC show. So here it is. China suspends new autonomous vehicle permits after Baidu outage, Bloomberg News reports. It's nothing to do with Baidu outages. It's nothing to do with Pony AI, nothing to do with anything other than they don't want thousands, in fact millions eventually, of self-driving— of drivers protesting this. So get ready.

In America, we've seen there is anger at self-driving cars. People calling Waymos that they don't need and getting them all trapped in a parking lot together. You know, like, there is some already sort of resentment, I think, building up. And there's been—

like, this has been brewing. So that story we showed was from April, but there's some more recent stories in June and July. Yeah, of this kibosh being put on it. So be prepared for that, folks. Uh, and I think I am in favor of— and I'm going to be called like a libtard for this, but I do think some licensing of these, collecting of revenue to, uh, smooth out for society the unemployment and/or retraining of those individuals would not be out of line. And I think it would be smart for self-driving companies to offer that so that they can deploy the technology.

It just seems obvious, right? Like, you— why, why wouldn't you try to soften the blow for people? It'll make them feel better about this technological change that you want to introduce. I, I, I don't know why that automatically goes to libtard for people, but I—

Well, I think it's just, it's not a free market solution, but I consider it an easy way for Waymo or Uber, Neuro. They could just say, listen, we know this is going to impact people negatively, so we're going to pay $10,000 a year, $1,000 a month or whatever for each self-driving thing that money's gonna go to that driver, or like, I think every self-driving car, I estimated.

Yeah, it's what Mark Cuban said to do with data centers. It's just like pay people off to get them to accept what you— Free electricity.

Yeah, very simple. Okay, next.

All right, uh, well, we got to let you go in just a few minutes.

You've got a hard out, but we want to— What about this automated solar one?

All right, let's go crazy. We'll, we'll pull this up. Uh, robotics podcaster Leo Karon shared this. Uh, Grit Robotics of San Francisco is the company. Now, they don't build the robot arms. That is an off-the-shelf That is an off-the-shelf Kawasaki robot arm. They designed the AI that teaches robots how to handle utility-scale solar panels. Now it does— it unloads the panel.

And this is real or this is AI? No, no, this is real.

The, the, the robot— this is real. They have trained the robot to be able to unload a solar panel. It can transport it to where it needs to go and it begins the installation process. You do need a human in the loop. The human then comes and fastens it into place. The robot can't do that precision work, but it drops it, as you can see, right where it needs to go. So the human just needs to come along and boom, zip, zip, zip.

Like, right, zip it until they make an R2-D2 that goes underneath that giant and does the zip, zip, zip, or they build a system that has clamps in it.

So this is, you know, that's the next problem they're solving is how to get that human out of there in time.

Well, it's super easy. Like, if the thing drops into place and in the factory it has a clasp that fits into a hole and locks in place, you don't need to put bolts on it. So obviously they're using like the classic system that wasn't designed for robot deployment. What I found really interesting about that video, and if you can queue up this one moment, and as they show a very long field of them— yes, there are very long fields being built. This just recently in California, we hit 51% of energy coming from solar and batteries. That's also happened in Germany on a couple of sunny days, and it's happening in Australia continuously and some places in South America. We can build a factory in the desert that has the ability to build solar panels and then out of the back of it have a robot place them. Yes. And then it would just become this fait accompli. Is that the right word?

Yep, you got it. Fait accompli.

Fait accompli. The, the robots keep making them and then the distance they have to drive, maybe it gets to 3, 4, or 5 miles. Who cares? The robot's driving it. You place this one factory in the center and then you just have solar panels go in all directions. Yeah. Now somebody will complain about the cacti, but let me tell you something, losing some cacti and, you know, a couple of rabbits get more shade. I think it's gonna be okay.

If you drive right now from LA to Vegas, there's a huge solar farm like 3/4 of the way when you're almost in Vegas. But it could be— I mean, it's really big, but it could be 20 times as big and nobody would really notice because there's nothing. There's like 3 guys out there, you know, in a trailer most of the time. There's like, you know, Walter White burying his meth, and that's about all that's out there.

Well, and then what I will say to add to this is— it's a Breaking Bad reference— you know, there's, there's very few NIMBY people to complain. And you don't need a lot of space. And there are bodies of water and tracks of water when you drive from San Francisco to LA, there's like these canals of water. They're putting solar on top of them. They started by covering them so they would get rid of evaporation, but you could now take this, and I've talked about this maybe 1,000 episodes ago with Molly Wood when she was co-hosting. Shout out to Molly. Everyone in the Pool is her new podcast. Go subscribe to it. $100 a year, well worth it. You could— you could— there it is. So I don't know what date this is, but this is California Roofed these June 27th, this month. Okay. So California Roofed 2 irrigation canals with solar panels that makes electricity and cut the evaporation off the water lawn by 70%. There you go. So this is all within our ability. Now, what you could also do is you could do desal, desalinization, powered by the solar panel and batteries, and in a canal, and all of this fighting over water and electricity would end, and there's nobody to complain. Because if you're complaining about the solar being put over a canal and it reduces water, then you are a certified idiot. You're a moron. Now there must be people protesting this. Please, somebody ask producer Claude or producer Grok or Producer Deep Seek, is anybody protesting these solar farms and these solar on top of rivers? This is like— there must be. I just can't wait for somebody. Somebody's going to— you know it's true, Lon.

Somebody's protesting this. I'm sure there's some animal in there. People like to— Native Americans used that water canal for something. I'm sure you could come up with a reason.

I mean, poltergeist or some Native American back burial ground.

There was a burial ground. Yes. Yes. I'm sure you, if you gave me sufficient time, I could come up with something.

This is This Week in Hope. Last Week in Hope. I'm just giving a little hope here. Everybody's angry about immigration, Trump, this, that, AI, doomerism. You soaking up all the water for your data centers.

This is great.

This is great. You could stop complaining, people. We're about to solve energy. If we solve energy, we solve water. If we solve energy and water, we solve agriculture. Good things are happening, folks. See you next week on This Week in Startups. Bye-bye.

Bye-bye.