It’s been a year since my last Ask Me Anything session. A lot has happened since then across AI, marketplaces, macro, and the broader tech ecosystem.

Here are the questions we covered:

• 4:22 Why does AI feel so widely feared or disliked right now?
• 8:48 Why is AI driving massive progress while politics and public systems lag behind?
• 13:34 What are the real opportunities in commercializing AI today?
• 14:10 Will startups still need human co-founders in an AI-first world?
• 17:51 How important is a technical co-founder in the age of AI?
• 20:00 Will intelligence (IQ) become irrelevant as AI improves?
• 20:18 What skills should young professionals focus on in an AI-driven world?
• 22:48 How should education evolve in the age of AI (and how should kids be taught)?
• 26:40 What drives investor decisions at the earliest stages of a startup?
• 28:23 How can pre-seed founders raise capital, especially outside the US?
• 30:11 How might graph neural networks impact marketplaces?
• 31:32 What does it take to win in marketplaces in regions like Latin America?
• 33:10 What creates real defensibility in AI companies vs hype-driven growth?
• 35:38 Are we in an AI bubble—and what does it mean for investors?
• 37:30 What is the right funding path for startups that require large upfront capital?
• 38:54 What proof do investors need before funding an early-stage startup?
• 39:40 How has your marketplace investment thesis evolved in the age of AI?
• 42:02 Where can founders find strong fractional developers?
• 43:15 What defines AGI—and how should we think about it today?
• 45:08 How important is the initial “wedge” in building a marketplace?
• 46:29 How do you evaluate whether AI adoption actually creates value?
• 48:00 What founder traits matter most today?
• 49:45 If you were starting today, what would you build and why?
• 52:32 Who are you beyond your professional identity?
• 55:11 Do you still have insecurities—and how do you think about them?
• 57:10 What would you do if you weren’t an entrepreneur?
• 59:45 What signals show a marketplace is reaching liquidity and product-market fit?
• 1:01:19 What is the core defensibility of a marketplace from day one?
• 1:02:24 What is Quince and why has it been so successful?
• 1:04:22 Which “boring” industries will produce the next big companies?
• 1:06:36 What is the biggest blind spot among VCs today?
• 1:08:15 Which AI sectors are overcrowded right now?
• 1:09:35 Can marketplaces succeed around complex, multi-service life events?
• 1:11:41 How have fundraising expectations changed in the AI era?
• 1:16:07 Should graduates join startups or big companies in 2026?
• 1:16:21 Is being a generalist still a viable career path?
• 1:18:35 Do investors prefer warm intros or cold outreach?
• 1:19:55 Should startups use services as a wedge before becoming SaaS?
• 1:22:02 What games or tools are best for child development and learning?
• 1:24:39 Will AI cause massive job losses or unemployment?
• 1:30:57 Can venture-scale companies be built in Latin America today?
• 1:32:26 How should AI be used in operational decision-making?
• 1:36:00 What separates massive marketplaces from niche ones?
• 1:37:24 How much decision-making should be delegated to AI vs humans?
• 1:39:10 What matters most in early-stage B2C startups: traction or insight?

If you prefer, you can listen to the episode in the embedded podcast player.

In addition to the above YouTube video and embedded podcast player, you can also listen to the podcast on iTunes and Spotify.

Transcript

Hi everyone. I hope you’re having a wonderful week. It’s been, frankly, over a year since we’ve done a ask Me Anything session, and so much has happened in AI, in macro and geopolitics, et cetera. So I figured it was the time had come to answer all of your questions and every round possible.

So with that, any further ado let’s get the show on the road. Welcome to episode 53. Ask Me Anything.

Great. So I’ve received a lot of questions that you guys submitted ahead of time which I’ve decided to go through, I guess one by one. Obviously, feel free to send questions along the way as the show is going on.

Along first fundamental question that someone asked me was why does it seem that everyone hates AI right now? Like, why is AI so hated? And I thought long and hard about that, and whenever new technology appears, there’s always like backlash. So let me give you really interesting examples. So back in the day when, writing in a way was invented.

Socrates was complaining that writing would make people lazy. They would no longer use their memory, et cetera. And now the funny thing and the irony of this is the only reason we know that is because Plato actually wrote down Socrates sayings. And so had writing not invented in terms of preserving knowledge, allowing to build upon the knowledge of other people, we wouldn’t have that.

The knowledge and expertise that we have today. And this has been true through a habit history. So when the printing press was invented, same thing, worried that somehow when the Bible was written, you’d lose the connection through the, to the church. When newspapers were invented, the main criticism was, oh my God you’re no longer going to get your news from the pulpit, and that’s going to be a big issue.

And of course, none of us have gotten our news from the pulpit. It’s not an issue. Whatsoever. When the bicycle was invented, people were saying it’s going to lead to a crisis of morality because women are going to be able to take their bikes and have affairs versus being stuck in one specific location.

And of course all of that was hogwash, right? It didn’t really change anything. It just made it just made our lives better. And so these concept of like crisis of like morality and technology kept happening. The TV television people thought it was going to create these zombie people in front of TV that are not using your brain in any way, shape, or form.

And the same is true of the intranet with Wikipedia, oh, or the student’s going to be learning, memorizing, et cetera, versus having access to information. And so right now people are worried the same thing about AI. It’s going to take all the jobs, which has been a concern people with that forever.

I’m going to address that in a different question. Maybe it’ll uprise and take over from us, as in all the so be movies, et cetera. So number one, general concern with new technologies where people are not comfortable with it and come up with all these really crazy scary scenarios. Number two, I’d say AI arrived at a moment of the zeitgeist where VCs and tech founders are no longer respected and they’re more frowned upon, criticized, et cetera.

So it’s not like in the late or early 2010s. But right now they feel that they have become the villains, right? Like the last Superman movie is like some tech billionaire as the villain. And the cultural zeitgeist is no longer pro technology. If anything, it’s anti-technology. And of course, things like social media, both positives and negatives.

And yes, they can be used for promoting, fostering democracy but can also lead to mental health crisis in young women, et cetera. So the, because the world doesn’t have the same positive view at a moment where they’re afraid of technology, I can see why people are uncomfortable. And then last but not least as always, it’s been, it’s very easy to imagine the jobs that are going to be lost because of AI.

It’s always much harder to imagine the jobs that can be created. And so people can foresee a world where maybe the jobs they have or no longer required, and there’s going to be fundamental change. People are risk averse. The, our amygdala it has like this fear response we’re hypersensitive to fear because 10,000 years ago, from an evolutionary prior G perspective, if you were in the savanna and there was a ruffling of the leaves the people that were really afraid and that it might be a tiger that could eat them survived.

And so the risk averse people are the ones who survived. In general, we’re afraid of change. So I understand why there’s this fundamental fear of AI. So that was a question from Tom.

Question from Emmanuel. Question number two. We live in this moment where AI and we’re seeing an extraordinary productivity revolution in science with new discoveries because of AI often doing a lot of the research or finding mathematical proofs. We’re seeing an explosion in creativity in startups because of AI where it’s easier to build startups than it’s ever been. And we’re seeing it in finance as well. And yet when we look at our political systems and our political processes, things seem more broken slower than ever before.

The quality of the people there seems to be declining, if anything. Why is that? And, it’s probably one of the big philosophical paradoxes of the 21st century than the one hand with the best tools, the best people working on changing the world in fundamental ways. And the other way, the other side you have political systems that are supposed to be for the public good.

They don’t seem to be doing a particularly good job. And there’s a number of, frankly, fundamental reasons for that. So first of all the markets are not great at necessarily allocating and dealing with public services. Which is why the public sector has been created. The issue is the reason one works better than the other, or it is as follows.

When you are building a startup, for instance, it’s a meritocracy. And if you do something good, you’re rewarded for it. And if not, you run outta money and it’s a very quick feedback. Very quickly if what you’re doing is working or not and the rewards keep accruing to the winners.

And your objective is very clear. Find product market fit. Create a business model sustainable, scale your business. And very quickly, if things work or not. And it weeds out the losing ideas and the losing people.

The political processes are very different. The feedback loops are very slow. It’s very hard to tell if you’re a good policymaker or a bad policymaker, or if you’re a good politician or a bad politician. And so 10 years in, you may still not know the answer to that. And because the systems move reasonably slowly by design, by the way. It takes sometimes decades for bad decisions to have culminated to the point where they lead to bad outcomes.

And because it’s much slower. And also the objectives are different, right? Like in the venture startup e ecosystem is like you invest in the startup, it works or it doesn’t work. You find product work you scale. And the other one, the, your main objectives to be reelected. And the political cycles are too short.

The reality is the things that happen in the world take time to move. Like in the last 50 years, a billion and a half people came out of poverty in China and India. But that took 40 or 50 years. Nothing happened in two years. And right now in the US you’re electing Congress every two years.

You’re electing presidents or prime ministers in the west every four to five years. And these timelines very little really happens. So it’s very hard to tell if someone’s being effective or ineffective. And so as a result, that world moves extremely slowly and I expect will continue to move very slowly.

And by the way, as I think through the impact of AI in society, I suspect as with most of these things, people are overestimating the impact in the short term and underestimating the impact of the long term. And the reason they’re overestimating the impact in the short term is if you are intact right now, you’re like, this is changing everything.

 All the jobs are good in the sphere. The world will be funnily different two years from now than it is today. But this is not the way the world works, right? Culturally. We move slowly politically, we move slowly. And if you think of where most of GDP is today, it’s in public services. It’s in large enterprise and these move extraordinarily so slowly, yeah.

When do I think that the DMV will use AI to make the process of getting their driver’s license faster? I think it’s going to take forever, right? So I think we’re going to see GDP productivity, funnily inflected by AI. You need it to see into the public services, which are 40 to 60% of GDP in most western countries and in a large enterprise.

And these are very slow adopters. So it’s going to take a while but ultimately we’ll change society in ways we cannot begin a fathom today.

So LinkedIn user, we are developing first nationalization network with the AI agents at Jacobian Labs. So it’s this fall during the FJ Labs thesis. Do you have any thoughts on the commercialization AI prospect?

Not enough information to answer the question. Maybe, I guess is the answer. Just send us that we’ll review and let you know, but obviously yeah, commercializing AI in some way, shape, or form makes a lot of sense.

We’re developing, so Alessandro, a co-founder matching platform, calls the founder’s Junction who believe that with AI is reshaping the job market and the internal landscape internals will always need a human co-founder. Investors. Do you agree with this view?

First of all, founder dating is a huge deal, right? Like finding the right co-founder matters a lot in terms of building a company. And so do I think that with AI, you’re going to be in a position to help people find better co-founders. Absolutely! Right, like there’s not been a very clear process. People take their friends, but the friends may not be the most suited for whatever skill sets they need. People look in the networks of random. And so do I think there’s a need for foundry dating and finding people that work together. Whatever you’re looking for, by the way. A CO may need a COO or a CTO may need someone to help them define the business model and fundraise.

So, I think there’s definitely a need for it. Now do I think that most officers in the near term, cause you have humans running them. Absolutely. I think your co-founder will be a human rather than OpenClaw. Absolutely! Now, do I also think you’re going to be using OpenClaw as your super smart assistant to do research and help? Absolutely!

Maybe not OpenClaw in the near term. In the near term, it’ll be an Open Claude type agent embedded or provided by the core AI LLMs like Claude or OpenAI that will offer an Open Claude equivalent without any of the security concerns and risks that you’re seeing today.

To answer your question, yes, I think the founders will continue to play an important role in building companies. Most of the founders will be human founders, even though you will be using AI. And I do think it makes a lot of sense to actually use AI to find better founders and to improve the co-founder dating process.

And by the way, I really would do a co-founder dating process, meaning you should totally do like projects with ’em define tasks and see if we work well on them together. You should totally hang out meet their friends, meet their girlfriends. You should go to dinner. Like really make sure that this is someone you can see yourself working with on a regular basis for a very long time.

Okay, moving on to the next okay. Remember a question that was related to founders that was actually interesting. Let me go through the list of questions that were pre-submitted.

 In the age of AI how important is a technical co-founder and should we focus on finding a technical co-founder versus someone with relevant vertical industry occurrence? Now the answer of course to this question is, it depends. As is the answer probably to most questions. If you’re building an AI startup with a foundational LLM model, then yes, you absolutely need a CTO, who’s absolutely fantastic.

If you’re building a company that is using applied AI to then probably not that hard to build in a way, makes way more sense to find someone who’s going to have legitimacy and help you sell into the general contractors and the subcontractors. The answer is, it depends. But if you’re an open AI or a foundational model, for sure, you need extraordinary tech talents.

If you’re building applied AI companies, yes you need good talent, but in a way the CTO is less key than might’ve been in the past. In fact, if I think of the marketplace is we build and invest in the things we care the most about. Are unit economics, can you make them work? Do you find product market?

So what is your customer acquisition channel? So in a way, understanding how you scale your customer acquisition ma matters a lot more. And make sure the unit economics work than getting the tech because the tech is more commoditized. And there’s more and more things you can do either easily with tech today, I mean with vibe, coding and cursor or lovable if you’re doing something very simple, et cetera. But in general, there are categories where yes your tactile matters a lot.

Okay, let’s go to the next batch of questions. See if there’s any this is a question from Julia. I recently had a conversation with someone super early at OpenAI who basically said he is trying to build a new startup now because IQ will become irrelevant in two years. This is a thought provoking statement. Do you think it bears some element of truth? And if that’s true, what do you think the, are the most vital characteristics, skills for entrepreneurs and ambitious professionals to focus on?

It’s interesting. I can go both ways on this one. I can make the argument that the very best, smartest people are going to use AI so much more effectively, they’ll be even more valuable. So, the 10 x developer is going to be a hundred x developer, in which case intelligence is not commoditized and actually continues to be a key success factor. But I can also make the case that because now intelligence and you have the tools that are so intelligent, you could be an average developer or an average person and get outcome.

And that is or products output that is extremely valuable. And as such, catches up and intelligence becomes commoditized. I suspect the former feels more true to me or is more true, and it feels more true to me than the than the latter. Right now I’m seeing the very best coders being more value than ever before.

The very best employees using tools in a way to be much more productive. Now, will that change at some point? And intelligence will be commoditized. Maybe doesn’t feel to be that way to me today. That said, the average intelligence seems to be going up dramatically as everyone’s improving productivity. Everyone using these tools extremely effectively.

And so what would I do if I was thinking, if I was in college today and I wanted to make sure I’m ready for the workforce? Play with all the tools, like play with Runway, play with Sora, play with Midjourney, play with Claude, play with cursor, play with Lovable. Install your OpenClaw.

Figure out what you can do to create scalable, repeatable systems. See what they’re good at. Test the limits. And there’s so much to play with today. So I’d be basically throwing all the spaghetti in the world, pursuing your creativity and figuring out, what resonates and what works for you.

Let me see what next question that was sent ahead of time. Question from Lisa. What type of school or education have you chosen for your son and how did you come about making that decision? So this is interesting because I’ve been through a few iterations here and actually a few changes over the years in my thinking.

And the first school I took my son to is a school in New York called The Ecole and the philosophy of that school, it’s a French American school, it’s amazing. And the thinking and the theory is you have the rigor of the French system with the public speaking and team building of the American system.

And he’s been there for two years. Man. He likes it. But when I reflect upon in the age of AI, is this the correct way to teach our kids where you have a teacher of variable quality spewing facts at kids of variable quality, typically to the lows common denominator where you’re repeating the same and teaching the same thing daily for three to four days.

It’s pretty slow process. And the answer to me is it doesn’t feel intuitively correct. If I took Socrates from 300 BC and I brought ’em to the world today, or he would not recognize the world. We go to space. We have these crazy, magical devices with the sum total humanities knowledge in our pockets.

We fly from one end of the world to the other in ours. And yet, the way we educate our kids hasn’t fundamentally changed in 2,500 years. And so the idea that you should be using AI to teach the kid exactly at the right level makes a lot of sense to me. So there’s this school out of originally called Alpha School where they use AI tools to basically get your kid to the very maximum of his potential.

So they’ve realized that you want to teach them to the point where they get 85% of the answers correctly because 99% is too easy. If it’s 50%, it’s too hard. And so in every discipline you want to get them at about 85% and you want to see how far you can get them. And on two hours of curriculum per day, they basically can cover the normal curriculum, and then they use the rest of the free time to lean into the kids’ natural inclination to have them do whatever works for ’em.

Now my son is four. And he’s years ahead of math, right? Like for fun, he’s doing multiplications divisions, he understands basic algebra he loves playing with numbers and he’s understands negative numbers, et cetera. And at the same time, he’s not pretty good socially. And so a school that is more bespoke for him where they’re can challenge him mathematically and frankly linguistically as well, where he is very verbose and eloquent while helping him develop his social skills, which are lacking, I think makes a lot more sense.

So starting next fall, I am bringing my son to office school in New York, which was created I think this year. So it’s first class right now. It’s a small school. And it’s going to be an experiment. It’s going to be an alpha test. And if I like it, if he likes it we’re going to take Amelie there probably as well.

Now, you know what’s interesting is one of their objectives is for the kids to love school. And the most kids don’t love school. It’s too easy, it’s too hard, it’s not interesting, et cetera. And I brought my son, who’s a bit shy to a shadow day where he went to check out the school and I was worried because he doesn’t do well in new environments, new people.

And so I left him a bit insecure and uncertain. I came back to see him and he was like, I love the school. Like I want to stay. Why am I going back to a normal school? So I’m excited to see how it plays out.

Question from Luis on the stream. From your experience investing in hundreds of marketplaces in today’s early stage environment, what ultimately drives investors’ decision? The most interest, intrinsic strength of the products and the market opportunity are factories like early traction, narrative, et cetera. More introductions to the ecosystem. In other words, you believe there’s still real space for investors to cover and back exceptional marketplaces, marketplace ideas, purely in their fundamentals.

Before the signal is validated by the crowd if you’re backing a very early stage founder. The signals are very early, right? Like that often there is no crowd. The big funds, the Sequoias of the world have raised so much money that they’re writing big checks once things are proven and there’s an emergent winner.

So absolutely there’s role for pre-seed investors and seed investors to back the right founders and the right ideas early as they’re at the early moments of product market fit and figuring out the distribution channels and union economics and retention and cohorts before the crowd validates it.

The crowd being, I guess both a combination of users that scale the business and of VCs with big brand names. That is side invest. So absolutely, there’s still a big role to be played today because many people are not investing all that early, given the level of capital delivery.

So today, if you’re in VC, probably you should be in the seed or, multi, like a hundred million dollar fund or multi-billion dollar funds. So you can keep doubling down in, in like the emergent winners.

Question from Ideal. This is an entirely different kind of question because you’re basically invest in online marketplaces. Can you provide leads for pre-seed investors for non-US based pre-revenue startups for game changing projects? Like earthquake protection system?

So presuming that those are venture backable businesses, meaning that they can scale two hundreds of millions or billions of dollars of revenues, because there are many ideas that are not venture backable.

And so let’s think through how you would get funded if you’re a pre-seed founder. And the answer actually is there are very few. There are not that many pre-seed VCs to begin with. There are few, and they’re usually highly focused these days, mostly on AI. So non-US pre-seed, honestly, what I would probably, and pre-revenue, I would probably start with the old adage of fool’s friends and family.

The good news of the world we live in today is it’s cheaper than it’s ever been to, to build startups and to start scaling them and start getting revenues. And so with several hundred thousand dollars in funding, which most people should find a way to be able to get right, like our friends went to great schools, maybe working as doctors, bankers, lawyers, right?

If you have 20 friends who give you 10 k that’s 200 k, you should be able to go very far. And so this way you can get some level of traction that should allow you to then go and raise a proper seed grant for a couple million dollars. Given that there’s not that many pre-seed founders or pre-seed funds.

Question from Mahesh. Graph neural networks are becoming more and more pertinent to discovering new applications, new pathways. Do you have any relevant thoughts on how this is relevant in marketplaces?

 So first of all, I care about at the end of the day I like marketplaces because they’re where it takes most. They’re scalable, they’re capital efficient, but I’m not like wedded to marketplaces, right? Like what I care more is can we bring tech to the world to make things cheaper, better, faster?

Now can I think of use cases for graph neural networks and marketplaces? Absolutely! There’s many marketplaces that don’t work without a human because the matching, the supply and demand is broken and too complex and there are too many variables that are not clear. And so I can absolutely imagine a world where, in a category where you have all these inputs, all these variables, all these, like having an agent in the middle that does the matching and the introduction, et cetera, probably makes a lot of sense. So I can imagine it becoming relevant in this category. But regardless, I think they’re reasonably relevant.

Nachogorriti on Twitch: greetings from Buenos Aires catching up with your content, just watched episode 52. I love the point Zillow being more exposed than Airbnb and DoorDash because low frequency and low management layer. Correct! We’re actually building on that exact thesis with remix and now native real estate search engine for Latam, eight month, one 50 K monthly visits for very good after eight month to realtors and the B2B pipeline, how do you see the Latam opportunity. What does it take to own the category here?

So in Latam there’s not an MLS and a way you can you can create your own inventory and create value reasonably in a less competed space. There have been a few companies that have done pretty well in real estate in Latam where I’d like to think of VivaReal in Brazil.

Do I think there’s a big option, need to go after the market, next generation? If you want real estate portal using AI, absolutely. Not sure if it’s Latam in general versus a specific country, right? Like usually in these categories, you need liquidity, you need density, you need the listings.

That’s that perhaps. To the extent you’re a search engine and you have sources for listings, it’s easier to fix in the past. TBD but do I think there’s a big opportunity in, in, in going after real estate with next generation tools? Yeah, absolutely!

Okay. Continuing the questions. For another one from Lisa, one of the clearest signs that an AI company is real defensibility rather than temporary velocity. It’s an interesting question. Because what we’re seeing right now in the current AI bubble is a lot of companies launching with the exact same product essentially.

So you have a Stanford team and the MIT team and the Princeton team and the Harvard team, and they all raised 20, 30, 50, a hundred million offering variations of the same product and often doesn’t feel particularly defensible, right? The one week one is ahead, another week, another one is ahead because there’s so much pressure to win, they’re all offering their products at a negative at a negative gross margin. And you are seeing businesses scale massively, ElevenLabs, or Lovable or a Cursor. That in a way we were all incorrect in not investing in, because we’re like, what’s the defensibility while they’ve been scaling the issue is they get scaled because there’s so much capital that’s willing to fund the growth at negative margin.

So TBD, how this ends up playing out. I worry that a lot of these are going to die. And frankly, a lot of these might be taken over by, Claude and ChatGPT I’m sure they’re going straight after cursor and after lovable. And yet these seem to be doing well for now. So these feel less defensible.

Now, the things that feel more defensible to answer the question are. If somehow you’re built on proprietary data sets that no one else has access to, if you’re solving specific vertical problems that no one else is going after. And so the versus the foundational models, those seem in a way riskier.

Like I suspect, right now, ChatGPT is an 86% market share, but it ebbs and flows. Gemini is going after it. Claude is going after it. The, there are weeks where Claude is better or Gemini is better. Then weeks where ChatGPT is better. That is a game of kings. I’m a bit skeptical. In fact, someone else asked me a question.

 Lemme go to that question that Tatiana asked. What is the massive seed round that was just announced. So we have La Koons company, AMI just raised a billion in seed at three oh 5 billion. What does it mean about the future of AI and how should investors think about technology versus valuation risk at this stage?

And to be clearly, we are in an AI bubble. People were willing to fund because the prize for winning is so high. People were willing to throw essentially infinite money at any price in order to win. But do I think this ends up in tiers because most companies are going to fail and many investors who invest at very high prices, they’re not going to see returning their capital. Absolutely!

That said, in the meantime, is going to lay the foundation for the extraordinary 25 years of productivity improvements and economic growth, we’re going to see the same way that the railroad bubble laid the foundations for all the railroad tracks around the us that led to a massive productivity boost in the economy for the decades to come.

The same way that the bubble in the late nineties laid all the fiber that led to the internet revolution in 2000’s – 2010’s just took a while for it to happen. So we’re in AI bubble. I hope it keeps inflating to be honest, because even if we, even though we’ve been disciplined, I worry that when a burst, the companies that are currently having a hard time raising because they’re not AI, are going to have an even harder time raising.

And frankly, in the meantime, with all this capital going, think of all this capital going into R&D, a lot of it is like money losing, but it’s going to be great for society even though a lot of these companies are going to die. So we’re an AI bubble, but it’s okay.

Friends and family, it is not a path I can take getting the MVP at the stage required VC level kind of investment.

That doesn’t sound like a venture. So there are different types of businesses in the world, right? The ones that need 10, 20, 30, 50 million to build a large to turn the lights on are not, frankly, particularly VC backable. They’re the ones that are VC backable are the ones where a couple hundred k you can get the prototype and get revenues, and then you get your million dollar pre-seed round and you get more revenues and more proof, and then you get 3 million.

The ones that need 20, 30, 50 million to get off the ground, they either belong in like large enterprise that are in the category or people that have succeeded before that have extra capital, but they’re not appropriate for normal founders because that’s not the way the VC treadmill works, or the VC treadmill is your full friends and family for a couple hundred K.

Then your million dollar pre-seed, then your $3 million seed, then your 7 million A, then your 15 million your 15,000,025 million B. Now in AI, you have different numbers than these, but that remains the type of numbers for non-AI companies that you’re seeing. Let’s see what other questions came on.

Alessandro: We’re close competing the MVP for our co-founder matching platform as these 500 founders on the wait list? I understand you invest in early stage startup, you require proof of revenue. Proof of revenue is not necessary Alessandro, but definitely proof of product market fit, that it works, that people like it, that there’s retention and you need to know what your business model is going to be.

You need to know how much you’re going to charge to whom. At least the theoretical unit economics could look like. It can’t just be we launch, we’ll figure it out later. That is not the way we invest. There are a lot of people that, that do that. It’s just not us. That’s not the approach we have.

Boris: Great initiative. I’m curious about if your investor thesis around marketplace has evolved since 2022. Have you become more risk-averse with pre-seed or marketplace investments, shifted more validating AI opportunities inside.

So Boris, that was episode 52. It was my podcast last week, which is investing in marketplaces in the age of AI. We continue to be very bullish on marketplaces. And we and all the marketplaces use AI. They use AI to translate the listings and to translate the conversations buyers and sellers so they can be global. So you pan European startups for the first time.

You’re using AI to have a one click listing where you take a photo and boom, title, description, price category, all prefilled for you, improving productivity. You use AI to do better matching with supply and demand. So we’re still investing in marketplaces. And they’re all using AI more effectively. And we’re more seed investors than pre-seed.

Meaning we like things to be live and have unit economics. Now the categories were more B2B these days and consumer facing, but there’s fun stuff happening. Even consumer facing, we’re like a live commerce company called Palmstreet, which is like rare plant marketplace. We’re doing we’re investors in a firetruck company or fire engine company, like 30 KAOV called Garage.

So there’s a lot of interesting things that happening with layers of services added on it. So we’re bread and butter because I don’t want to avoid competing in the game of Kings with infinite capital and negative gross margin in the AI bubble. So we’re indirectly exposed to it because it, while A. we have amazing investments in things like figure AI which is doing very well, and B. all of our companies use AI.

But there are vertical applications of AI versus being foundational AI models themselves. And I actually think that’s where a lot of the interesting opportunities lie today in terms of like reasonable, you can build big businesses with very little capital and you don’t need the same super shallow pool of AI engineers.

Yoni: Any tips on where to find reliable fractional full stack developers (AWS + Angular) to help improve an existing SaaS MVP?

 Depends how good you need them to be. But there are a lot of places like Toptal which allows you to find amazing, oh no, we, you said fractional. I would go to Fiverr or go to Upwork. The issue is you’re gonna need to do selection. So one of the ways I would do selection on Upwork or Fiverr, by the way, is you create a spec. You get 20, 30, or 40 people apply. You look at the best five. You give them the first 10% of the job, you hire five of them, and then you see the one that delivers the best and you like working with the best. So you’re overpaying for the first 10%, five x, and then you find the one you like and boom, that’s the one.

So in a way, you don’t necessarily even need an interview. You can just validate based on the work they do. And that’s the way I’ve hired a lot of people on Fiverr and on Upwork over the years. Okay. LinkedIn User: somehow, no name showing. Long time, no see. Wanna fund our AGI Endeavor? Recent breakthrough, paying for a demo.

What is AGI exactly, right? General intelligence. Currently, our GPT can pass the Turing tests. So is that AGI, is that not AGI? I suspect that the way we’re gonna define intelligence is going to change. The way it works from my perspective is AI is super human in certain capabilities, right?

Like in terms of ebbing away in math problems, et cetera. And it’s way beyond human intelligence. It’s significantly better, significantly faster, significantly more patient, by the way, human minds work, which is. With limited data. We create concepts, which is the opposite of the way these LLMs work, which are, there is infinite data.

They get patterns. Maybe just profoundly different, may just be two different ways of creating thought patterns and processes. So it’s not completely obvious to me not completely obvious to me that there’s a human, we’re going to replicate human thinking. I think we’re going to have profoundly different ways for the AIs to think and that’s okay.

And so yeah, it’ll, it, it’ll be interesting. But would I suspect that whatever your AGI endeavor is, is going to cost infinite money. So if it’s capital efficient and happy to look at it. If you need hundreds of millions. We, sadly I wish I had more capital. Are not the guys.

George, in your experience, how important is using the right initial wedge when building a marketplace? What makes the wedge strong enough to expand into a larger ecosystem? So when you launch a marketplace, you have no barrier entry. Just to be clear, like anyone at the beginning can build the same thing. Your wedge if you want, your bare your, what is going to differentiate you over time is liquidity, right?

In these marketplaces, ever more buyers, presumably more sellers or more sellers, brings there more buyers. Once as a buyer, I go there and I find whatever I’m looking for, and as a seller, and it could be of anything, a product service, there’s someone to buy what I’m selling, that’s when you have your wedge.

So it takes time to get, to build. Day zero, you have zero barrier entry, but within 2, 3, 4 years. Your barrier to entry is actually the liquidity that you have. So find, create early liquidity between your buyers and sellers. And as you get your early liquidity, that creates your barrier to entry over time as it becomes bigger and bigger. And these things, as I mentioned, have a tendency to be winner, or it takes most because everyone more buyers brings and more sellers or more sellers brings ever more buyers.

Let’s continue on the questions and the pre-sent questions. What metrics matter most when you’re evaluating whether AI adoption is actually to marketplaces? Okay, so yeah, so whether it’s sticky or not, we look at retention. We look at retention when it comes to whether or not an AI company is being successful, right?

So a lot of the AI companies have massive churn. And so that’s one of the things that makes me worry that they’re not very sicky. Maybe they have product market fit, but they definitely don’t have barrier entry. It used to be that I used Runway for making videos, and now I’m using Sora.

So I’m on ChatGPT. It used to be that I was using Midjourney for almost all of the photos and images I was creating for my blog, which itself had replaced stock photography. And now I’m using ChatGPT more and more. So I would look at cohorts, I would look at retention and not just one month retention, but six month retention, 12 month retention.

Like the better products usually have a U shape. You use them, maybe you use them less, but at some point you come back to them. And cohorts, retention curves matter dramatically.

Boris: check out Djini. It’s Ukrainian HR marketplace mostly of your software developers. Yep. Good idea to recommend that for people that are looking for software developers.

Okay. Continued questions. What founder trade do you value more today than you did eight decade ago? Honestly, the traits that I value haven’t changed very much. I love people that are extremely eloquent and visionary and therefore that can hire a better team, sell better to VCs, talk to press, get better deals, et cetera, but also know how to execute that, their attention to detail. They focus on unit economics, et cetera.

Now, the one trait that, that sadly is not a requirement for success, it’s like being a kind person. You have a lot of assholes. And the problem is because some people Steve Jobs or Travis got away with being assholes. It encourages emboldened or allows people to just not be kind.

And, but life is too short for dealing with assholes. And I’m in a position where I don’t need to, and so I want to work with like kind people. Now that said, or a lot of founders are arrogant. Absolutely. Is that bad? Nah, you need a some level of like delusional self-confidence to build a startup, right?

Like the five-year survival rate of a startup is like 7%. And so you need to believe the odds do not apply to you. So arrogance, narcissism I can probably deal with. Being an asshole, definitely not. But has it changed? Not really. I was ready already had that belief system before. Okay.

Question from Jeff. If you were graduating from Princeton and maybe just leaving McKinsey or consulting in 2026 what do you think you would be building right now and why? Now, clearly I’d be building something in AI. This is where the world is and it’s changing and it’s interesting now, it depends.

So if I was 23, depends on the skillset. I’d say there are multiple viable path. You can join a rocket ship and latch on. Go work for open AI philanthropic you can build an AI. Now building an AI, the thing is the bigger the game of Kings is like, am I going to own humanoid robots?

And you have figure and optimist. Am I going to own the underlying LLM? So you already have like big winners there. And then you have some of the verticals. I suspect I would go for applying AI in categories that are so like old and broken and antiquated where everything’s done by pen and paper and relationships in a category that is of interest to me because obviously you as a founder don’t work at a vacuum. You have your own set of interests, you have your own set of skills, and so you want to solve a problem that’s big enough that’s monetizable, but that you actually care about. And whatever your background is, I would focus on that. And so maybe your parents are paying from the construction industry, so maybe go and optimize that.

Maybe you work in the food industry and there’s, and you have so many profound problems in terms of like turnover of employees, sourcing of different materials, et cetera. So I can think of applying AI to automate processes and bring efficiency to many categories that I haven’t been addressed before.

And I’d probably be working in that now. Which one specifically? I don’t know because I haven’t been thinking about it because I’ve been too busy between, my funds, Midas, the kids, et cetera. But definitely an interesting thought experiment and something I actually tend to allocate time to on a go forward basis in terms of thinking through, okay, if I wasn’t doing FJ Labs today and building Midas, what should I be building?

And the answer obviously is something in AI, but what it is for me today is interesting. I don’t know the answer to that, but I definitely, it’s a question that it warrants asking, and I will ask myself in the coming weeks, month and year what it could look like.

Okay. Question from Margo. If we remove premier identity startups, investments, performance perhaps even financial success. Who are you truly? Is this person enough?

And it’s interesting. So in the US people often define themselves by the job they have. And of course, the job they have is only a tiny percentage of who they truly are, right?

Like your personality, your needs, your desires, your dreams, your aspirations. Now I try to be my true, authentic self at all times. And so I think it comes across in the way I speak. But you’re still seeing through my blog, through the podcast, the professional version of me. And so to answer the question is the look.

Look, I think that the meaning of life is to be yourself, true, authentic self, whatever that is. And we’re all built differently with different predispositions desires, needs, et cetera. And honestly at this point, like actually I’m fully fulfilled by being who I am. Like. I love all the things I love, like from being a father and a parent, from playing with the kids, to playing with my friends, to playing video games, to reading books, to writing my blog, which is actually these days, not mostly not about business, to interacting with my friends, to actually being yeah, the patriarch of the family in the positive sense of the term. To, yeah, playing tennis, playing paddle, et cetera. Yeah. The life I have is extraordinary. I literally think I’m living the best life that’s ever been lived. Definitely the best life I can live. And I am fully fulfilled. And so if I was not, for whatever reason, could not be working today’s world, I’d be very fulfilled and happy regardless.

The external identity driven by work is nice, but actually and I think it’s a source of purpose because I think at least one of my purposes, help harness the deflationary power of tech to solve the world’s problems, to make things better, cheaper, faster for the masses and to try to address a combination of inequality, of opportunity, climate change, and the global mental and physical wellbeing crisis.

But even if I didn’t have that, I find extraordinary source of purpose through playing my kids, raising my kids, plaguing my friends, et cetera.

Another question from Margot, you give the impression to have an infinite confidence that’s super rational and to be very poised. Do you have any insecurities.

So I’ll start actually by answering the question. In the past growing up, I had many insecurities. So because I was really good at being very smart and at getting good grades, I defined myself by that. But I was very insecure socially, right? Like by virtue of being younger than my peers by the fact that I never had a girlfriend or friends, et cetera.

Like I, I had my first girlfriend at 27. Was it a source of insecurity to not have a girlfriend when I was 26 or never had a girlfriend? The answer is yes, right? Today much more comfortable and with who I am and so don’t have specific insecurities, so I’d say, I guess the answer is no, no real fears.

But are there things that really graded me that I don’t love in life? Absolutely I a bore aging, like the, I used to be the youngest at everything I did and now often I’m the oldest. Do I like that? Absolutely not. And so I rage, rage against the dying of the lights. And that’s why I work really hard to stay fit be sharp and yeah, say keep my youthful energy, hopefully forever.

But definitely for as long as possible. Not sure it’s an insecurity per se, but definitely something that annoys me and I’m working very hard to fight against father time because yeah, the, there’s so much to do and we live in such extraordinary times and we’re so privileged to be it that, to have the energy the health to be able to live it to the fullest.

Same thing. Like I want to be able to play with my kids in a very meaningful way. And last question for Margot. If you could not have become a founder and an entrepreneur, what do you think, what job would you think you would’ve liked to explore? That one’s hard because I really a bore traditional structures where like the nine to five job, the having a boss, like I consider myself unemployable.

So if tech wasn’t a think, I suspect I’d still be entrepreneurial if possible in another form of industry category. Now, if entrepreneurship itself is not possible, if’s way harder because then I’d have to find a job that like matches more my way of thinking and I’m not quite sure why that could be.

Interesting experiment for another life that I hope I never have to do because I love what I do and I love the flexibility and the freedom and the creativity. In a way entrepreneurship is my form of creative expression. Taking something from zero to one and creating something out of nothing, and I am not sure what else would be so fulfilling.

So no idea. I guess this is the honest answer. Could I have been in private equity or consulting or banking? Absolutely. But would I love it day to day, minute by minute? And I think the answer is no. So there are many things I could be very good at. I could be a professor. I’d be a fantastic economics or mathematics professor, but again, would I love it?

And the repetition over the years of the same course material, I don’t know. It’s too slow not scalable enough. I don’t think it would feed my soul. But yeah, actually professor is probably a reasonably good one. But not sure it’d be as fulfilling, but for sure I wouldn’t find it as fulfilling in a way I scratch my professor itch by doing this podcast, by answering the questions of the audience and the users by thinking through things I want to share in a way playing with unicorns has always been about what are all the things I wish I knew when I was 23 and starting as a first time founder that I now know that I can share with you.

And I find that it’d be more interesting, more scalable than having classes. And I used to teach at classes at Columbia Business School or Center for Business School, et cetera. And yes, you’re teaching your amazing people, but they’re small class, not super scalable. And the content didn’t change that much.

Now it’s whatever crosses my mind, create the material, poof, put the podcast, and it is as and when there are ideas and that are relevant.

George: in early stage marketplaces, what are the clearest signs that are platforms about a break out and the cold, sharp problem, rather than remain stuck in low liquidity?

 If the sell through rate of the items on your site, if you’re selling products about 25% or more, you start to have liquidity. If you’re a services marketplace and you start accounting for 25% or more of the revenues of your supply, you start having liquidity. And the way to make sure you get there is don’t overflow.

I guess it depends on the marketplace, but the biggest mistake marketplace founders can make is having too much supply. If you have too much supply, they’re not going to be engaged, they’re not going to reply. The buyers are going to be overwhelmed with choice. It’s much better. You have the very best supply for whatever category, zip code, et cetera.

Find them demand, get them liquidity. Then scale a bit more and scale a bit more. On this slide, I think it’s a bit more demand and keep matching. A sign that you have product market fit is when your customer acquisition costs are declining, and that’s when the users start coming back, bringing their friends and your unit economics keep improving.

But early signs of liquidity is typically, yeah, 2020 5% sell through rate is a usually a good sign. That in a used good marketplace, at least you have liquidity. Okay, going back to the questions from that were pre-submitted.

Lewis Gonzales: if you’re starting Global Marketplace from scratch today, what would you prioritize most as your core defensibility from day one?

Liquidity, brand, community, technology especially with AI becoming increasingly accessible. I’ve answered this before, but basically day zero, you have no moat, no barrier to entry. Your barrier entry over time becomes liquidity. Once you actually start getting more buyers, bring more sellers, more sellers, bring more buyers.

So focus on union economics. Whatever your scalable, repeatable strategy is, for scaling supply and demand keep doing it. Keep matching it, keep getting liquidity. So liquidity in marketplaces trumps everything. And in fact, imagine that somehow the top of funnel, like the agents would be the ones transacting on behalf of users, they would transact where there is liquidity.

So your ultimate defensibility is in liquidity. So liquidity. Liquidity. And when in that more liquidity.

I see you invested in Quince. Can you tell us more about them and what is their ambition in the future? So Quince is one of the fund returners for FJ Labs. They’re doing extraordinary well.

They’re in affordable luxury marketplace and direct to consumer brands. The marketplace because they’re in an asset live model. The founder said it’s extraordinary. We invested in them since the beginning and I guess the pitch for them, the elevator pitch is it’s the quality of a Macy’s, the pricing of a Costco and the logistics of Shein or Temu.

And they’ve grown extraordinarily from whatever, like a hundred million to 300 million a billion in sales to I think over ran 2 billion last year. It’s still growing like crazy, and they just raised at a 10 billion valuation from iconic. So where do they go from there? So first of all, it’s extremely rare that a company at this scale, like a billion in 24 revenues is still growing a hundred percent year on year.

That like never happens. And they’re still in the very beginning of their journey when you think of the categories they’re in, when you think of the geographies there, they just launched Canada this year. I think they’re going to start launching in Europe. So they’re at the beginning of international expansion.

They’re at the beginning of a category expansion. I can foresee a world where they’re in the tens of billions of revenues in five to 10 years. And this is a company, don’t you can keep winning. The company is already in a dominant position and it can keep winning. There’s so I’m hoping that it keeps winning, that it keeps scaling, it keeps doing extremely well on a go forward basis.

Quince is already a fund returner and I’m hoping will continue to be a fund returner in the future and ever more and one of the biggest winners ever for FJ Labs.

Gael: what markets today look boring or unsexy, but will produce the next generation billion dollar companies? So everyone right now is focused on the big war and the foundational models, right?

And yes, this is the multi-trillion dollar opportunity and the ChatGPT versus Claude versus Grok, whatever. And this is where all the attention, all the money is going in, right? So when we looked and in my last podcast, 75% of the venture dollar is went to AI. And 95% of the YC companies were AI foundational model type companies.

We’re fighting over who the game of Kings. What is completely unsexy right now actually are things like marketplaces. They, we have amazing companies in the, in, in the portfolio that they’re growing from 10 million in GMV per year to 30 to a hundred or whatever. And because people saw growth of zero to a billion or billions very rapidly in the AI space, they’re not excited by this anymore.

Even though these companies are capital efficient, they need a lot less capital. They have amazing unit economics. They have amazing gross margin. And there are a lot of industries that where you could use AI to make them more efficient from public services, to construction, to retail, et cetera.

Where I think there’s massive opportunities. There are many categories where combination of opaque, fragmented data or needing a lot of people for intermediation, you can imagine a world where these agents could actually improve the economics, make the category larger, et cetera.

So I would say boring old industries that have not yet been touched by technology where for the first time you could use agents in order to scale and make the category more interesting and efficient, of which there are essentially infinite, right? Most of the economy has not yet been touched by AI to only the super early adopters and tech thing.

That’s case. What is the biggest blind spot you currently see among venture capitalists? Definitely everyone’s piling in all AI all the time. Doesn’t matter, the valuation, doesn’t matter the gross margin structure. We need to be in, because the win is going to be huge and it’s very bubbly.

It feels like 2021 all over again. It feels like 2006 real estate where it only goes up. It never goes down. It feels like 98, 99, 2000 tech bubble. At the same time someone is going to win and the rewards would be huge. But would I be coming in right now at these insane valuations and anthropic and OpenAI?

I guess the answer is no. Could they still grow a lot from where they are? And is this the biggest opportunity of them all, possibly. But if you were early, that’s great. If coming in now, it wouldn’t make me feel really comfortable. And so I would be, the more what we are, like, the more boring applied AI investors that I the way describe our strategy is the smart way to invest in AI.

We invest in companies that use AI super effectively to have higher margin, to have lower cost micro acquisition costs, to have higher conversion rates. To me, that’s the correct way to play this. And it’s yeah, definitely not what VC other VCs are doing.

Let’s see about the questions that were submitted by email. In the meantime, while you can still keep posting questions here let’s look at here.

Muresh: which categories/ subcategories within the AI space of potential, which ones are overcrowded based on the pitches in the discussions you have with other super smart investors and VCs? I feel that the foundational model game is super crowded, right?

The xAI and Mistral and also true in the verticals like Runway versus Sora and Midjourney, et cetera. So that feels extremely crowded to what I suspect will be a winner. It takes most category, maybe it’ll be two, you may be a drop wins B2B and ChatGPT wins consumer and Gemini keeps its, some market share.

But do I see 20 winners in this space? No it feels 1990s search engine wars, AltaVista versus the Lycos versus Yahoo, et cetera. And then all of a sudden Google comes along. So I would not be funding more foundational models. And I would be focusing, as I said, on the applying AI to categories that that people have not been using it for right now, but definitely harder to raise in these categories. And because it’s not deemed pure core AI.

George: have you seen marketplaces succeed when the value isn’t a single transaction, but rather coordinating several services around a larger life event? Yes. We are investors in a marketplace around weddings. That is doing pretty well. They have a massive market share of weddings in Europe.

Of course the name is going to come back to me at some point soon. And of course, and the way they monetize is helping you find your caterer and your venue and the photographer and the person providing the cake, et cetera, et cetera. So they are coordinating around many services for around one large life event.

So wedding is definitely an example of that. Can I think it can happen in other, at other large life events. Perhaps we’d have to define what those life events are, right? Like death obviously is a big traction for people like liquidating estates and estate cells, et cetera. And, like graduating college eh, the pro, the thing is.

When you graduate college, you need, maybe you need a car, maybe you need a job, maybe you need a housing. But all these are well done by sites that do that full-time. So would I create one site for all these things? I’m not so sure. Versus the verticals that are already best in class for each of these categories.

Same thing on like moving cities. So there’s a bunch of companies that help you move cities and they’re doing okay. No one’s great. Because again, if I’m moving New City and I need to find apartment, Zillow is great. You don’t need to go to site specifically for moving. So I think wedding, make weddings make a lot of sense, would be definitely a fair amount of sense. What are the other large life events worth thinking about? Okay, continuing on the proposed questions.

Godfrey: question number one, how has your FJ Labs fundraising matrix changed, especially in recent month given AI’s rapid impact on B2C and B2B market, B2C marketplace in terms of traction, ran size, valuation?

So are our valuations going up dramatically at the mean and frankly even median? Yes. Because of AI, you’re seeing bigger seed rounds. You, we just saw a billion dollar seed, round billion raised, 3.5 billion pre. So clearly the valuations that people are commending, especially in AI, are a lot higher.

But we are avoiding that AI hype and so we are still focusing. Like in 21 when everyone was saying, oh, your matrix that out of date, it doesn’t make sense anymore, et cetera. And of course I was correct, it came back, I was correct, meaning it came back with a vengeance. And the number is reset.

So if you remove from the equation all the AI hype companies the matrix is still viable, right? So we still want you to be at like 500 K to 750 K a month in GMV with a 15% take rate when you’re raising your series A and you’re raising 10 at 30 pre or seven at 23 pre or something like that. We still want 2.5 to 5 billion in GMV per month.

That’s expecting, by the way, 10, 15% take rate curated A 2, 3, 4% take rate B2B. We expect much higher GMV when you’re raising your series B of whatever, 50 million or 53. So the matrix is still correct but doesn’t apply in AI where people are paying insane prices at seed, pre-seed, A, B, whatever.

But if you were building a company, I would recommend you to stay close to it because if you raise too much money at too high a price, it will kill you. It’s one of the biggest reasons companies fail. They don’t grow into the valuations and they fail to raise the next round. If you’re a VC, I would recommend it to you to stick close to the matrix because if you overpay, you’re going to have bad returns, and the VC asset class is already not doing very well have good returns.

Question number two, since AI makes building software much easier. How much do early stage VCs value technical co-founder? Oh, yeah, I answered that before. Which I, as I said was the answer is it depends, and it depends on the category you’re in. If you need a technical co-founder because what you’re doing is extremely hard then you should have one. If it’s if you’re building an open next generation OpenAI, have a technical co-founder.

Okay. Rosa Bluda, what are you missing in life if you’re missing anything? Honestly, I am, I’m really think I’m living the best life that’s ever been loved. I don’t think I’m missing anything. Life I’m healthy life that my family’s doing great.

I’m doing well, like life is, it’s extraordinary privileged and I’m full of gratitude for the life that I have. Don’t think I’m missing anything. Maybe I don’t know what I don’t know. And there are things I’m missing I don’t even realize I’m missing. But yeah.

Next question. Does Palantir have a rival? There’s a French Palantir called Arlequin AI. Not spelled in a funny way as most tech companies are, but there’s a more interesting one called Fundamentals which is, because Palantir, it’s hard to tell how much of a tech company is it versus a services company, right? Like their implementation is 6 month to 18 month.

Most of the revenues they have is coming from the implementation services versus recurring SaaS fees. And Fundamental, and you have it in, they use AI obviously, and they do integration in two to three days, and most of the revenues are through subscription. So to me, that’s the most interesting Palantir competitor up and coming.

Do you have a preferred artist? I’m talking about a painter. Not really. Writers, more so. Painters. Yeah, no, probably, I guess not Look, do I appreciate art and what artists are trying to do? Absolutely. But not sure I would, I have an answer to that question.

Okay. Continuing on the Quest pre-submitted questions. I recently graduated from master’s program Matteo, and interested in AI startups. If you’re graduating in 2026 today and want to build, would you start your career at a large company or an early stage startup? And for someone with a generalist profile, is that still a viable path today? And what skills would you prioritize broadly on technical and non-technical?

So in general, I think you learn faster and better in, in startups than you do in large enterprise. The, when I graduated from college, I went to McKinsey. It was like business school, except they pay me, but it would’ve been just as viable to join a startup partly seed A or B, maybe a B sage, but not too big.

Otherwise, you’re going to have a role that’s very pigeonholed and you’re not going to be able to learn as much as you would otherwise. So you want one that has enough product market fit and funding that’s going to continue to do well but not so establish that the role is like very cookie cutter and that you can do and prove yourself out and follow your passion and learn as much as you can.

So I would join an early stage startup probably in AI, probably in the Bay Area and move there right now if I was graduating college. And to figure out like what is the best path. Versus joining a large company. Now, again, OpenAI maybe is okay now if you’re an engineer. Now if you’re a generalist, which is your case probably then the smaller companies make more sense.

And do I think there’s a path for generalists? Absolutely. I think there’s more path today in a way, for generalists than ever before because as a generalist, you can actually use the tools of AI to get tech out the door very quickly. You can learn to vibe code pretty quickly, right? Like with Cursor, you can things are much easier for you as a generalist that is smart using the AI tools than they were ever before.

And if you think of the role, the CEO and the founding team in the go forward basis, the CEO is the generalist, so absolutely. Being a generalist is amazing. Now, as I said earlier, I would play with all the tools. I would, create an OpenClaw, play with Claude, play with it, with GPT, play with Cursor.

Become super familiar what you can do with them and see, and how far do the bleeding edge you can take it. And you would be shocked at how much you can improve your productivity, how much there is to learn, how much there is to do.

Let’s see. Alessandro, it seems to investors tend to fall into two camps. So those who prefer warm introductions and hate cold, average, and those who are open to cold average. Which camp do you fall into? So first of all investors prefer warm introductions, right? If there’s a founder I know, or a VC I know, or whatever that says, Hey, you need to talk to this founder who’s amazing.

Obviously I prefer that. But I’m open to cold outreach because not everyone went to Stanford and Harvard and Princeton and is in connected to the social connectivity of the networks that le allow you to meet the relevant founders in VCs. And so some of our best investments came from cold inbounds. They were in Brazil, but instead of being in Brazil in Sao Paulo Rio, they were in Belo Horizonte. That said that, some of the threshold is higher. It’s just there’s a lot more. We get two to 300 cold inbounds a week and the percentage that we invest in is a lot lower. So I we are open to cold inbound. If you can get a warm intro much better, but we’re open to it.

Andrew McCain. In the years since we last met in New York, your business selection criteria has changed my life. Ah, glad to hear that project. I’d love to get your feedback on follow up question here, regarding services, do you think there is value following the Palantir model of doing services heavy during scaling established customer relationships to durable moat, the product suite evolves and becomes more autonomous using AI to create true ARR. In other words, sources first approach, more of a go to market than is about AI ubiquity.

The answer of course is it depends. It depends on the category, depends on the customer profile and segment. I prefer non services approaches because the main feedback you’re going to get from VCs is like, are you a services company? How scalable is that? Versus are you a actual tech company?

That’s why I like fundamental more than I like Palantir. It really is a tech company. So to the extent that said, if you’re selling to governments often you need to be, it’s selling a service. The service layer, the installation, the relationship matters a lot.

So I guess the answer is, it depends. In general, I’d rather invest and have people build tech companies than services companies. And it’s some of the challenges that feedback that these companies face with fundraising, because the valuation of a service company is profoundly different from the valuation of a tech company.

But if it’s a go to market strategy, if it lock in the customer and then allows you to get these MRR or ARR contracts that are highly valuable to high margin. Then it’s okay. At the end of the day, what I care about is what is your go-to market strategy? What is your product market fit? What do the union economics look like?

What is your customer acquisition cost versus the net contribution margin for customer? And as long as those work, and if services is the way in, that’s fine, but it needs to be explicit that’s the way in and not the end goal.

Lisa, a bit of a different question, but I’m curious, what type of school education have you chosen for your son? No, I answered that question earlier. When I talked about office school.

Sonya, what are the kids development games and PC or Nintendo that you use? So you know what’s interesting is there’s so many educational tools out there. So first of all, my son, who’s four, is obsessed on YouTube with number blocks.

Like he’s doing multiplications for fun. It’s eight times eight is 64, like whatever, 27 times two it’s 54.28 times 2, 56. He does negative numbers. He does basic algebra, not because I am forcing him to learn math at four , when there are expectations for him to count to 25, it’s because it captures his interests.

And so he finds educational content on his own that he likes on YouTube, and I give him the iPad in the morning when he wakes up and the night before bed. And he basically follows number blocks and learns math. In fact, it, he’s inclined enough that he’s asked me to go to Russian math school in New York, so I’ve also signed him to Russian Math school.

But are there interesting games that you can play with your kids to foster their creativity learning. Absolutely. We just played together on the iPad a game called Lost in Play, which is an adventure game with puzzles where you need to use basically IQ type test or brain teasers to solve problems to get the story moving forward.

And there are a lot of these, again, that’s appropriate for a four or five, 6-year-old. As you get older, the reason I like building on accommodation of Minecraft and Roblox is the logic patterns of building there. And again, as a builder, not as a consumer is it teaches you coding in a way.

So it’s an interesting way through fun to teach kids coding. Are there more? Yeah, again, I don’t know the age of your kids Sonya but things like lost in play or amazing. And there are a bunch of kits you can order that are stem, where you can, your kids can build robots. There are a lot of things, but I would lean into their interests.

As I said, like I didn’t tell Fafa, okay, go learn math. He just decided he loved it and learned it. It’s part of the reason he’s so excited about going to AI school next year, which is Alpha.

Next question from Tom. Are you worried about job losses created by AI? So this is the perennial question. AI is going to take over all the jobs. There’s going to be 95% unemployment. It’s the end of the world, et cetera. And this is a fear that is universal and has been universal for hundreds of years, right? The. The Luddites were against the electronic loom in the early days, even though it made the life of people that were weaving is significantly better.

And this has been true throughout history and people have worried about all the job losses. But let’s say I take you back 26 years to 2000 and I told you, and we were now in March, 2000, look, in 2026, I’ve come back and the top four job categories of 2000 have disappeared. There are no more travel agents, there are no more bank tellers.

A trillion of local retail has disappeared because of online commerce. All of car manufacturing has been automated. And these are the top four job categories of us right now. Please now describe the economic conditions in 2026 and people would tell you, oh my God, mass unemployment, great depression, et cetera.

And yet today we have lower unemployment, higher employment, and two XGDP per capita than we did back then, despite all these job categories disappearing. Now of course I’m hearing right now, but this time it’s different. It’s happening faster than ever before. AI is replacing all these jobs, and so first of all, it’s not happening that much faster than ever before.

In 2011, 2012, when the first self-driving cars came to the fore, people were like, oh, the top job category in the US with 4.6 million jobs is truck driver. All these jobs are going to disappear. There’re going to be no more truck drivers. What are all these people going to do? They’re going to be, they’re going to be automated away.

And we are now, so this is like 2011, 2012, literally 15 years ago. We’re now 15 years later, not a single job of a truck driver as yet been automated by a self-driving truck. And we’re at the very beginning still of the self-driving AI revolution. Now, do I have any data in my mind that at some point in the future, 10, 20, 30 years, a hundred percent of vehicles in the road will be self-driving.

No doubt whatsoever. For sure, it makes sense. And they’ll also all be electric. But it’s going to take time. Like the first ones that get automated are the most expensive because the technology costs a lot of money. And culturally it takes time. A lot of people, first time they get a self-driving card they’re scared shitless that it’s going to, it’s going kill them, even though it’s seemingly safer than traditional cars.

So culture moves slower than technology. Technology moves very quickly, but governments will take a long time to adopt AI. Large enterprise will take a long time to adopt AI. These changes happen a lot slower than you think. So number one doesn’t move as fast as people think. Especially people in tech, because we’re at the forefront of tech. Number two, people don’t understand how many jobs are really going to be created or lost by AI because they, when the elasticity, they don’t understand where the elasticity of demand for a product or services.

So right now, one of the big thesis people have is, oh. Programmers are going to become obsolete. The AI will code itself. You will no longer need programmers. It’s a possible outcome, but it’s far from guaranteed that’s the most likely outcome.

In the 1980s, there was a job where people were called the spreadsheet, and the spreadsheets were done by humans. Highly paid, highly skilled humans were built a spreadsheet before something called Symphony, which now I guess would be equivalent with Excel, came to the fore. And Excel did lead to the destruction of all the jobs of spreadsheets. But you know what? It created the jobs of millions and millions of financial analysts who now had the tools to do financial modeling, financial analysis.

And so a couple thousand jobs disappeared. Millions of jobs were created. So when it comes to software engineering, for instance, you could make the case that has the cost of software development becomes very low, demand for it explodes. Companies that historically did not hire software developers like SMBs or governments or large enterprise at large scale, would start doing it.

And so I can actually make a case. I’m not guarantee that this is going to happen, that as it becomes so much cheaper to build software, demand for software will increase so much that actually employment increases. And that’s not including the fact that there’s so many new job categories that are going to be created.

Like in 2000 people could not imagine what the role of a social media manager will be or a Twitch gamer, caster or whatever. So many new jobs are being built and created that people have a hard time imagining. Am I worried about the job apocalypse? No. Are jobs going to change? Yes.

Will there be losers and who will need to be retrained and helped to adapt because the winner isn’t the losers as the job market evolves are often different. Absolutely. But am I worried about 95% unemployment rate and the great depression and we’re all out of a job and it happens overnight, absolutely not.

It goes against economics against everything that’s ever happened, against culture and the speed at which people are willing to adjust and adopt technology and the inertia built into our political systems, to our economic systems, et cetera. No, I do not think this time is different, but yes, I do think that as for usual, this technology will profoundly transform humanity in the way we lead our life.

Though it will take a lot longer, these people think. Yet again overestimating the short-term impact of AI and technology, underestimating the long-term impact.

Okay, Jorge: building decision intelligence infrastructure for the T-MEC/USMCA industrial corridor. Okay. I guess that means Mexico, US probably, and Mexico.

B2B2B model, targeting customer brokers, environment consultants and accounting firms, distribution channels. Do you see value in Latin American or industrial verticals in the market to fragmented build venture scale from there.

I’ll take a step back. Do I think you can build venture scalable businesses in Latin America? Absolutely. Think of Newbank in Brazil or Plata, which is a new bank where investors in Mexico, or Mercado Libre, et cetera. So first of all, the Latin American market is large growing, ever more sophisticated is starting to have its own VCs, from Kaszek to Monashees, et cetera.

So you can totally build successful venture backed startups in Latin America. Now specifically in your sector. I don’t know enough about the total addressable market size, unit economics, et cetera. But to the extent we’re talking like 10 billion plus dollar market where there’s probably enough margin structure, I suspect in the answer is yes. So yeah, reasonably positive.

Okay. LinkedIn user, I’m not sure why names are not always showing and sometimes they do. Hoi, Fabrice, I don’t know if you remember me from earlier episodes, I ran a marketplaces in the Netherlands. You provided advice during several earlier episodes. Sold the marketplace, used the money to now build insurance company heavily in integrating AI.

Great! Use AI, ready for customer service, fraud, pricing, claims processing. You have marketplace advice here. You’re more than welcome. And I think what you’re doing in terms of using AI to do improving in everything, customer service, fraud, pricing, claims, processing makes a lot of sense. We’re investors in a company in Europe called ACE Waves.

Ace Waves is a customer care company for marketplaces where they integrate and the AI replaces a big chunk of your customer care team on average allows you to lower your customer care costs by 50% while improving your NPS, improving customer satisfaction, et cetera. So definitely use AI for customer service for all those things. And every startup out there should be using the tools to their fullest extent possible.

Djordje: I’m probably butchering your name. Thank you for answering my question. I pitched to you in regard to our platform at Jacobian Labs bringing GNNs, I’m not sure what that is necessarily, commercialization in one, but you said your AI said pass. Is it possible to send it pitch deck or demo to you directly? Yeah, send me a LinkedIn InMail with the deck, et cetera. So working in my AI, by the way, the pitch Fabrice is just trying to give you feedback, et cetera. I’m going to try to make it more nuanced in terms of what it likes with this, unlike. What would it need to see that’s different for us to want to invest.

So don’t take the AI pass as an end all be all. And by the way, the team reviews all the pitches to Pitch Fabrice on my AI on the fabricegrinda.com. I haven’t done that yet, but it’s on the to-do list of the last batch of Pitch Fabrice. So yeah, send me an email we’ll review it.

Mention that you mentioned this conversation from this episode for reference. And yeah we’ll take a look. Now. Yeah. I don’t know how much traction you have. We typically, or post-launch, post revenue, post product market fit, but early. But post all those things. So not sure exactly where you’re at, but we will look at it.

Let me see if any other questions have come in the last few minutes. And if not, if you don’t have any final questions. We will bring this to a close. Let me go check. People have sent questions on WhatsApp.

Okay. I think we’re good. I think we’ve covered every question that’s been asked so far. So thank you for tuning in. As usual, I will post this, the transcript and the summary of this episode on my blog next Tuesday. And not sure what the next episode will be and when it will be just yet.

Perhaps what the questions people are asking earlier. In terms of what AI companies I should be building if I was building it today. Oh, actually wait a few more. Last bit of questions are popping in.

George, your experience, what separates marketplaces become truly massive platforms from those that remain niche or services businesses?

That the thing is, it’s hard to tell in the early days. Like Uber originally was a black car service, so it was very high end. It felt very niche. I was saying the other founder chose to do Stumble Upon instead of choosing to do Uber. He thought Uber was smaller. And it’s just when UberX came to the market, it became bigger.

Think of Airbnb. Airbnb originally was inflatable mattresses in people, living rooms felt like a very niche product, and of course became a much larger category. So follow market fed and see how big the category ends up being.

And sometimes you can create ginormous category. Just so happens that housing is an enormous category. And monetizing underutilized housing is a ginormous category. So added, been pitched like that would’ve been obvious, was big from the get go. Just wasn’t pitched that way in the beginning. So how do you know how big it is?

Often, even if something feels small, you can actually go to coin joint category, add other verticals, increase the temp, where the little bit is, that often the sky’s the limit. These things can end up being a lot bigger than you think.

LinkedIn user in the current phase of AI, how far would you allow decision making be done by AI and the level of human supervision?

 It depends what you’re doing, right? If you (A) use common sense of, when I ask AI to do research, which I do on a regular basis, I definitely cross-reference the results. Also ask AI to give you the counterfactual. So if it argues for something, say if you were arguing the opposite view, what would you think?

Also, ChatGPT is a huge sycophant. It tells you how amazing you are on a permanent basis. Ask for honest, realistic, no holds wide feedback very explicitly. Otherwise, you’re going to get a rosy, tainted answer as to as to what you’re doing. But in terms of like fundamental human decision fundamental important decisions, I would totally have human supervision right now for most tasks.

Now are there things that can be automated, like customer care for, oh, what’s the tracking number of my order or I didn’t arrive or whatever. Yes, absolutely, that you can have AI do it, but like mission critical things, use human supervision for now. Hallucinations, errors, biases, but it’s interesting. These are biases because it wants to please you, and so it ignores the downside. It tells you how amazing you are, et cetera. And so you need to be very careful on the type of questions you ask and how you diligence it. In fact, use multiple LLMs to test concepts and ideas to make sure that you’re getting a better a better perspective.

Quick questions. We’re in B2C we’re evaluating very early stage startups for matters more to you early traction, short, strong insight into a mass problem that incumbents have ignored. B2C is hard because you have inventory, there’s competition, et cetera. So I care about early traction and unit economics.

So for me it’s actually more than early traction is unit economics. But obviously does it need to be a problem that’s big enough that a warrant’s going after? Absolutely. But for sure, you, like in B2C, how do you market it? And how do you scale market scale marketing? The issue is customer acquisition costs are going up, and so it’s hard to make the margins often work. So making sure that you can find your economics working to me and scaling and repeatable is probably the most key.

Supervisor AI constantly asking, inform me information from all of our operational AIs. We allowed do light decision making. Yeah, that makes sense. And more than light impact supervisor human.

Yeah. That’s really the proper way to use agents and the way I would use my agents. So for instance if I have my OpenClaw, go and look in LinkedIn to see potential LPs for the funds and like by who could write 250 K to 500 K checks and different geos and think through when we could do a meeting.

So great. Do I let the OpenClaw then do I ask it to draft emails I could send? Yes. Do I automatically let it send the email without me reviewing it? Absolutely not. And is the and maybe it’ll do it for the long tail, but would I let it do it for, I don’t know if I’m pitching a hundred billion dollars pension fund that could write a $20 million check on the fund?

Absolutely not. Yes. Advise, draft, et cetera. And even then, I don’t love AI writing. I love my own writing, obviously I’m biased. When I wrote my huge, big thesis on the meaning of life this summer, which was a like 10,000 word piece on my perspective, on the meaning of life.

After I was done writing it, I uploaded it in AI in ChatGPT. I was like, okay, give me feedback. And it I essentially, except for the obvious mistakes of spelling mistakes, grammatical mistakes, et cetera, which I fixed. Using AI, I ignored all the advice. It’s oh your title is way too generic. The meaning of life. You need a punchy action thing.

The thing is way too long. You need to break it down in whatever, 27 pieces. Your examples are too opaque. And I basically is you know what? I love my own writing. I think the way you write is like too flowery and cumbersome and I hate the m dashes or whatever.

Yeah. Thank you for the advice, but no, thanks. I do my own writing. That said, I do getting feedback from AI. So for instance, yeah, I ask for ideas of things to write about, et cetera. I just like to do my own writing. And by the way, it did identify mistakes and repetitions, et cetera.

That did lead to fundamental improvements. But yeah, I think the way you’re using AI makes a lot of sense, is also the way I use AI, but look, I’m a AI super user. Like I talk to AI on a regular basis about everything. I test everything. I create everything from videos to images, to testing business models, to finding real estate. You name it, I use AI for it. Use it. It’ll make you more productive.

Okay, I think we’ve reached the end of the stream. Thank you all for joining in. This was interactive and fun. And I’ll see you in the next one, whatever the next one is, and whatever the topic may be, in few weeks, few months, we shall see.

Have a fantastic week!