Sam Lessin on AI investing

Sam Lessin is a GP at Slow Ventures, an early-stage VC firm based out of San Francisco, Boston, and New York. In the last decade, they have invested in the earliest rounds of over 500 companies including Solana, Robinhood, Venmo, Airtable, Slack, Front, Allbirds, Postmates, and more. He's the cofounder of Fin. He was VP of Product Management at Meta. He has a degree from Harvard.

In this episode, we cover a range of topics including:
- AI technology is great, but AI investing is not a good idea. Why?
- Why is AI just an extending innovation as opposed to a platform shift?
- Why can't $100B platform wins be manufactured?
- Do AI startups have moats?
- AI moment vs cloud computing moment
- AI regulation
- Seed equity "killing fields"
- Meaningful AI use cases

Sam's deck: https://docsend.com/view/e6nd457kgbg8zich

Sam's favorite book: The Lessons of History (Authors: Will and Ariel Durant)

--------
Where to find Prateek Joshi:

Newsletter: https://prateekjoshi.substack.com 
Website: https://prateekj.com 
LinkedIn: https://www.linkedin.com/in/prateek-joshi-91047b19 
Twitter: https://twitter.com/prateekvjoshi 

Transcript

Prateek Joshi (00:01.093)
Sam, thank you so much for joining me today.

sam lessin (00:03.458)
Thanks for having me. Happy to be here.

Prateek Joshi (00:05.641)
Yeah, you had some very epic takes on AI. So I thought it would be fun to just dive in. On this podcast, we mostly talk about why AI might work, but I think it's very important to know and have a strong case for what might and could go wrong. All right. So you mentioned that AI is a great and impactful technology, but AI investing is fueled by magical thinking and that it's

it heavily favors incumbent. So why is that the case?

sam lessin (00:38.742)
Look, I think the thing to understand is, again, just as you said, it's very important for everyone who doesn't know my shtick to understand I'm not anti-AI. I actually think AI is going to make a lot of people an incredible amount of money. I think if you're a big platform, you'll do fabulously. I also think if you're very, very small, you will be able to get more done. You will be able to act as though you're bigger. Just as the cloud and the move to cloud, it was great for tiny companies. I think AI will be great for tiny companies who are just consumers of the APIs.

I think the problem for a Silicon Valley style tech investing though is, what you're looking for is companies that aren't going to be really successful, profitable, small businesses. You're looking for companies that can be the next 10, 20, $30 billion outcome, where a new startup can outcompete big companies. I don't think that's realistic in the case of AI. I think there's a few reasons for that. One is,

If you think about it, like there's this hardware thing everyone understands, right? Which is that AI is, it's a CapEx problem, right? It's not just a intelligence or can you build a better interface question, right? Like it's really just a question of like, do you, do you have access to the machines you need to do something really epic or different? And the reality is it's super constrained. I mean, the reality is open AI, right? Which is as close to a, this is still a startup, but like, it's basically a big company, right? From a capitalization perspective.

they have trouble getting access to machines versus the Googles and the Facebooks and the Microsofts. The world is part of the Microsoft deal is they literally couldn't get the hardware they needed, right, like without Microsoft. So that's one obvious thing. I think another thing is data, right? Which is, look, you know, when you think about differentiation in the AI world, you know, having access to data is absolutely critical. If you have the same access everyone else has, you're not gonna build something that is that differentiated. You know, everyone.

you know, copyright aside can crawl the same public internet and we'll kind of get within 20% the same-ish answers. The question is what can you do that's meaningfully different? Can you build a information pump or access that no one else has? There might be some pockets where that's true, where for whatever reason, some small company has some special deal, but if you think where almost all the data is in the world and all the share of voice, it really is with big companies, not small companies, right? And so I think that's the second reason. The third is distribution, right? Which is, you know, people have...

sam lessin (02:53.994)
this is now many months ago, but for the last year, I've been pitching, we're going to be the Adobe of AI. And I'm like, Adobe is the Adobe of AI, right? They have the distribution, they have all the pieces. For them, the idea that they can just add in generative fill, that's great. It's awesome. Like no one's going to build a new company on generative fill. Generative fill is another tool you add in, right, in a lot of ways. And so I think people misunderstand the lessons of history, right? Which is...

Prateek Joshi (03:01.149)
Hehehehe

sam lessin (03:19.67)
the move from no software to software, and then from desktop software to the internet. Those were fundamentally disruptive things that changed the business ecosystem so dramatically that it created opportunities for different companies to win. What I think you see now is that actually AI slots beautifully on top of cloud, right? It literally is just another cool tool that fits perfectly into Amazon Web Services. It fits perfectly into kind of a lot of the platforms that already exist. And so again, I think it's very exciting, but I think what you'll find is that the researchers

at the highest end and even the product people at the middle end who want to do the most epic work, they're gonna find that working at big platforms is far more profitable and will come with far better answers than in most cases starting new companies. So investors desperately want AI to be a thing. They desperately want it. It's perfect. If you're a Silicon Valley investor, AI is the greatest story you've ever heard of, right? It requires a lot of capital, which is good because you have the money, right? It is...

hugely disruptive, it's going to change everything. And like, so you can plow a huge amount of money into companies that you can own a lot of that have this story that they're going to change the entire world. Nothing could be better, but you have to be careful that investors who like to be on Twitter a lot, you know, will tell you the story they want to hear, not the story that's actually real.

Prateek Joshi (04:35.449)
Yeah, amazing. And you know, you published that epic deck, like it's so many slides filled with a lot of information, a lot of takes. And you mentioned that AI is not a platform shift, but rather a classic extending innovation. So where is the line that separates extending innovation versus a platform shift? Like, how do you recognize one versus the other?

sam lessin (05:02.734)
I mean, I think the answer is, if you think about this, and again, it's tough if you haven't worked at a big company, but if you work at, if you have a sense of what it would be like to work at a big company, I worked at Facebook when it wasn't anywhere near as big as it was, but it was big enough. There's a question if you're on the executive team and someone shows you a new magical piece of technology. Do you say, oh shit, or do you say, fuck yeah, right? And if you say, oh shit, right, this is a, then like that is potentially a platform shift. If you say, fuck yeah,

that's an extending innovation, right? And I think what you'll find is AI is a fuck yeah for those types of companies. They're like, absolutely, this is amazing for us, not an old shit. Google might have an old shit sort of with an interface thing and how their business model works. So like there might be something there that you can argue about, Sundar sitting there and there's this new thing, what does he think? But these companies have been working on AI for a decade. They're clearly bought in, they clearly see the value, they clearly know, I mean, they already have most of the researchers, right? So.

I think what people miss and take is that distinction. Even mobile, so mobile is an interesting one in this history because people like to talk about mobile as though this was this big platform shift. They're like, oh, first there was software, then there was the internet, then there was mobile, now there's AI. Every 10 years, we get a new disruptive innovation that changed everything. Mobile didn't change anything. Mobile actually was just more internet. What's interesting, because I actually was around for this, was there was this moment during the mobile shift.

where even Facebook was like, oh shit, right? Because if you thought about it, like it was a web-based company, traffic was all of a sudden moving to mobile, there were no mobile teams. So there was this moment, brief moment where a company like Facebook said, oh my God, is this a new platform? And is this disruptive? And it turns out that it wasn't, right? It turns out that actually, like you had to change some resources, but it actually translated beautifully and more internet benefited Facebook, took it from a hundred billion dollar company to a trillion dollar company, right? And so I think that's kind of the thing you'll see.

Will there be exceptions? Of course there will be exceptions. There are always exceptions, right? Like you can't, but I think the thing to keep in mind, and this is especially, I'm a professional investor, is look, you can't, you have to be structurally right, versus just randomly getting right because something changed, it wasn't exactly as history plays out. So I will for sure accept the fact that if you make a billion, every bet in the world on this,

sam lessin (07:25.69)
Somewhere, something will end up making you money, right? Somewhere, something will end up breaking the rule, I'm saying, but I think if you're trying to be scientific about this and think about where to put your effort in particular, like I actually care much more about builders than I do investors. You say, is this the place to go try to compete with the big platform? The answer is no, right? The answer is definitely not. Someone might randomly make some money, but this is not the type of platform shift that is exciting if you wanna go start a company.

Prateek Joshi (07:52.401)
Right. And you made the point that a hundred billion dollar platform wins, that never manufactured and can never be in terms of, in terms of where they can happen, for example. But what are the characteristics of one such thing? If some, if we were to see this, how would we know that, oh, this kind of maybe looks like something that could lead to that?

sam lessin (08:14.218)
Well, I mean, look, this is where I'll talk my own book. You know, I believe it or not, and you can laugh about it, but less than you think, I think crypto fits this perfectly. Right. Like, and I think, you know, like it's going to have its ups and down. It's obviously been a crazy year or so and a half. Um, but I think it's perfect. Right. In terms of this, why is hyper disruptive, right? It is a different, it's a software intellectual breakthrough with Bitcoin.

Right. And then you can talk about kind of DeFi. You can talk about kind of Solana is like tack ons on top of that fundamental set of ideas, but it's a fundamental breakthrough in thinking and approach, which is so disruptive to how we think about finance and money in a lot of ways, that it is incredibly scary. Right now that means there are regulatory concerns. You know, the reality is, is interestingly, because it's taken a while, cause it's a decentralized versus centralized movement in a lot of ways, it's taking a while to get going.

a lot of the smartest banks have actually gotten wise to it. So on one hand, they're saying crypto is bad, we don't like crypto. On the other hand, they're absolutely adopting a bunch of crypto primitives and ideas in the idea of stable coins and everything. So it is always more complicated than black and white. But you say like, if you said like, we have two kind of trains running, one is the AI train and the other is the crypto train in terms of what truly is a disruptive platform, right? It's crypto not AI because crypto is like the type of thing where if you work at a big bank, you're going, oh shit, not hell yeah. Same thing with,

I'll call it one third one just for fun, which is the other big theory of the recent history has been, oh, metaverse is going to be this big disruptive thing. That was kind of during the pandemic work from home translates to virtual world. Again, I think people are totally misunderstood that right. Because again, you know who metaverse is great for big platforms. It's too expensive. It's too complicated. Right. You know, the people can build headsets are meta and Apple, right? Like this is not a good place to be a startup.

And so I think these are the types of things to keep in mind.

Prateek Joshi (10:09.565)
And when you talked about mobile, obviously mobile was, you said it's just more internet, but it fundamentally changed the way people consume stuff. The reason it was such a big shift is that people had these little things in their hands and anybody who figured out how to supply stuff to this thing, people consumed it and you made, obviously you made a lot of money. So in that situation, isn't...

Isn't AI changing the way people consume information or is it just like a nice tool to have on top of existing? Is it just more internet but with a different modality?

sam lessin (10:47.618)
You know, I think this is a great debate to have, right? And I think there are probably niches where it will change. Like I do think the good example, if you want to argue with me, is clearly Stack Overflow is fucked, right? And clearly, ChatGPT is a way better way to consume like answers to coding questions, right? So there are places where it's gonna make sense, right? That are very specific and now, now is that a new $100 billion platform? I don't think so, right? You know, and the reality is, that's a great example of like,

sometimes overfitting, which is there couldn't possibly be a more perfect problem for AI, right? If you think about it, like it's perfect, right? For the AI, for large language models. Why? Tons of open source code, right? Because code has to run, there's like an objective function, like does it work or does it not work, right? In a lot of ways, and machines in a lot of ways can run that code and test it. It is like a perfectly clean data set or relatively speaking, it's super clean in terms of the approach.

So it's tons of free data, it's got that kind of property of being validatable in and of itself in the digital world. And it's just obviously engineers love it and they use it all the time. So they're like, oh my God, this works so well for this. And they kind of over pattern actually, it's gonna work for everything else. And I think what you're gonna find is that it's a unique, it's a very unique place where I agree it does change behavior. How valuable is it? Again, it's open data.

It's fine, you know, like other people will do the same thing. Like everyone can have their coding assistant. There's no secret sauce to it. Um, but yeah, that's kind of how I would.

Prateek Joshi (12:20.685)
Right, right. Yeah, yeah. You made another interesting point in that deck where you talk about how AI is too much of a repeat of the cloud compute moment where customers were locked into their contract and infrastructure and there's no way now companies are gonna fall for that again. Now, is that, it's a question one.

Are people like, since we have the data, will companies be way more diligent not getting locked into open AI? And also what are the mechanisms to make sure that you don't fall for it again?

sam lessin (12:59.678)
Yeah, I mean, look, I think especially what happened last week, I don't know if you were following with OpenAI and kind of the disaster over there in terms of trust, right? But like, just people aren't that stupid, right? And if you think about it, it's like, you know, I think that what you'll find is that with a lot of this AI stuff, if version one in the cloud was like, okay, who's my cloud provider? I mean, hilariously, when I started my first company called drop.io, literally 15 years ago.

One of the things we focused on, we were a very large AWS user initially. We kind of did a lot of like kind of drop box of storage, but then we immediately moved to this model of multi-housing or multi-homing cloud stuff. And we had this idea that you could kind of like switch between crowds and like play off them, right? You know what I mean? As an individual or as a group. And you know, at the time that was a little ahead of reality, even though it was a cool concept, now everyone does that, right? The big companies shop contracts, the small companies, you just say, well, where's it's all the same commodity. It's cheapest.

The only lock-in in cloud computing really is that Amazon web service has done a very clever thing, which is no one of their services is actually that differentiated or valuable, but they just have so many that you buy into the overall ecosystem, you get stuck in it. Right. And so I think when you think about like how to not do that so that you don't get extorted, right. It really just comes down to thinking about when you architect an application, you're like, what are you using AI for? Like where are, like, are you pushing data to it?

Where are you getting data from when you make a call? Who are you making it to? It's not that hard to switch from a, I make one call to I can bid them off each other, or I can set up my infrastructure to switch very easily between them. Right. You know, again, there are slight trade-offs to this, but you know, it's already, it's not that, I mean, I've built several, unlike most venture capitals, I actually like to build stuff. And so I've now built several projects, you know, some of which I actually might commercialize, right. Cause they're kind of cool.

I'm using, you know, a bunch of these APIs that are open and like, we all know it's just not that difficult if you just treat them as services to be like, I'm just going to like be able to use multiple services and switch between them. At which point, you know, the idea of being a cloud customer, right? Like a cloud service is very difficult. It's very similar to Uber and Lyft, right? It's like as a consumer, you know, part of the problem, part of the reason Uber is in a much bigger company.

sam lessin (15:11.286)
Part of the reason that Lyft is now almost bankrupt, right? And like this has played out, and like so much venture capital was played into it, is it was far too easy for consumers to sit there on their phone and just say, I'll just use both, like I don't care. You know what I mean? Whatever is cheaper or faster this moment. And I think that's what you're gonna see on the other side of cloud, a lot of this stuff.

Prateek Joshi (15:30.441)
Yeah, I think you made a very interesting point here where customers are now, there are so many models available and they'll just pick the one that suits them the most. Usually it's the cheapest one. And so does that mean based on today's reality, is there room for a model routing product or a company where basically what they're saying is that, hey, we'll index all the available models and based on the request, we'll route it to the cheapest one or is it more of a again, a feature in a big way?

sam lessin (15:59.694)
Look, someone will build that. Many people will build that. You know, Cloudflare will build that. You know what I mean? Like lots of people will build things that do that. There'll be open source ones. There'll be closed source ones. Like a router is always a good idea. The question is, is it a good business? And like, again, the answer in my mind is like, probably not. Right. Like, you know, like, you know, because it's too easy to build and it's unclear how are you going to differentiate yourself from every other possible router. So will there be a piece of software that exists?

it's an open source project you include and you say, AI router, right? And I just, sure, why not, right? Like there'll be many of those. I just, again, and if you're an individual, here's what I would say, let's say you're an individual engineer or you have a team of three and you say, we're gonna build the best AI router in the world and we're gonna evangelize it and build an open source project around it and we're gonna charge a hundred dollars a year for it. You could make a fortune as an individual, right? But.

That's not a venture capital opportunity, right, is the way I would think about it, right? So again, like I wanna be clear, like I do think lots of people and individuals will become extremely wealthy building very high quality, very specific things that appeal in specific places. I just don't think it's a place to put a bunch of venture capital.

Prateek Joshi (17:13.469)
Yeah, and I think you make a very good point. It's more likely to end up as an Apache project, getting standardized open source, and people can just take the tool and run with it versus it becoming a standalone business. Coming back to AI startups and the mode, and right now, as you said, most of it's so much open source tools. Everyone uses the same APIs, and everyone just offers stuff on top of their big cloud platforms.

ease of access. Does that mean more people will build more stuff, but you have to figure out how you differentiate because everyone has access to the same things. So how can a startup attack this problem? If you want to build a meaningful business, is it going to be like classic SaaS business where sure you can use tools, but you have to build like an actual real business?

sam lessin (18:11.09)
Look, the thing I'm really bullish about, especially if you're an individual engineer or know a specific domain or you have some sort of real personal differentiation, again, I want to be clear. I think you can make a huge amount of money and a very good business being tight and building something of value and distributing it. It's never been easier, right? You can as a small team, as an individual do amazing stuff today that you couldn't do 10 years ago, right? And so I think that the key for me is don't over capitalize.

Don't convince yourself you're building a trillion dollar company or a hundred billion dollar company instead be like, I'm going to make $50 million a year for me and my team with a super profitable, specific, very high quality thing. Like I think that's extremely doable. And so I think, I don't know, that's kind of how I think about this stuff. Again, like I think there'll be a lot of entrepreneurs.

You know, the reality is, if you make 50 million, and making up numbers, let's pretend you build a great tool and like it's awesome and super purpose built and like everyone loves it in a very small niche, you're able to make $50 million a year. Well, guess what? Over 20 years, what would you make $2 billion? Right? Like, I'm sorry, no, over 20 years, make a billion dollars. That's a great, you're doing fabulously. Like that's very valuable. Like I'm not saying that's not a good thing. I'm just saying it's a bad venture capital opportunity.

Prateek Joshi (19:26.481)
Right. Also, among all the ideas you've talked about, many AI startup ideas, they're not venture backable, but you've given a, instead of like, yes, no, maybe, you've given a maybe to the marketplace idea, right? So why are you less skeptical? Or basically, why do you think this might work as opposed to all the other startup idea that's gonna just burn money?

sam lessin (19:41.815)
Yeah.

sam lessin (19:52.929)
Um...

Look, it's really, I don't think it will work to be clear. Right? I don't wanna like be overly categorized as like, oh, AI marketplaces will work. I do think that market, if you can find places that historically buyers and sellers couldn't connect, but for some reason with AI, right? They now can. I just believe very abstractly and intellectually that like that could like finding new places where buyers couldn't connect before and they can now with AI is not the-

craziest thing I've ever heard. The reason I'm probably less skeptical about it is actually one of the things I've built, and it is with some proprietary data and things like that, is effectively a private marketplace for buyers and sellers of a certain specific thing that does, in my mind, kind of work. Now, I don't think it's a multi-billion dollar company, but I do think there are those niches that exist, is connecting people is always a valuable thing to be thinking about and doing. It's hard.

You know, I grew up at Facebook. It's hard to shake the knowledge that those types of connections are scarce and valuable. And I think a lot of the biggest companies in the world, including TikTok, kind of connects people with content, things like that. But forming new connections is kind of the way to make a fortune. And so, again, I don't wanna think, I think people are like, oh, well, what's gonna be the AI of eBay? I'm like, it's fucking eBay. Like, it's not that complicated. eBay will be, like, so.

It's not like, I don't think there's gonna be, I don't think there's like some huge category, but I think there'll be, there are missed opportunities in connecting people, right, to make a lot of money.

Prateek Joshi (21:32.289)
Right. Just shifting gears a little bit, you've talked about killing fields, which I think it's a good name, it's kind of, it stays in your memory. In the killing fields, you're talking about how they are the two hallmarks of disaster and early stage capital investments. Can you just talk about what they are and why they cause so much trouble here in the early stage?

sam lessin (21:59.674)
Well, look, I just think the killing fields, at least in my head, that idea comes from World War I, where you basically had these literally killing fields where you'd have these soldiers that would run at each other in war, out of the trenches at each other, and just get mowed down by machine guns. It's one of the most upsetting things you could possibly imagine. And the reason is that happened is because warfare had changed. Technology of warfare had changed, but...

people like the tactics hadn't evolved yet, right? And so you had this thing where people were just like literally, you know, pointlessly just slaughtering humans, right, like on both sides. And there was no way they were gonna get across the killing field. And I think that's kind of what you're seeing with, you know, with capital. People can bring a lot of capital, a lot of effort, a lot of passion, a lot of human capital to bear, to try to do great projects. But if fundamentally the configuration of the war is set up where like that stuff is just gonna get obliterated.

then it's a huge waste of everything. And so that's kind of what I see as the killing fields of AI. Again, I haven't thought about this before, but I might have to put your thing out. I might have to make kind of a screenshot essay on this. It really is, I think one metaphor in your mind is the history of Uber and Lyft. That was a killing field for money. It was just plowing money into a thing. It was a war of attrition. And I just don't think that, the only way to win a war of attrition is to have the most money.

right? And like the reality is you almost certainly don't. And the funny part about it is, let's pretend that you even do let's bring you our Uber, right? You do have the most money, you can win. But it's a pretty shitty victory for everyone, right? Everyone gets super deluded. Sure, the investors make some return, but it's not great, right? Because again, like it is literally just, you know, it's, there's a there's actually a great old, very old Star Trek on this.

where like two countries go to war and the way they go to war is basically by just incinerating their own citizens and who can tolerate the pain of incinerating the most of their own citizens? And it's like brutal, but those are the types of games that as a builder and as an investor, I want to avoid is what I would say.

Prateek Joshi (24:10.309)
Right. And since you're at this intersect, kind of interesting intersection where, and you've done some amazing crypto investments, you've been deep into that field. And now, obviously, with AI, you've gone deep there as well. And given where we are, the world is very hungry for compute in general. Like people want to run many, many models, algorithms, inference. So do you think there's room for decentralized compute?

Is that fun? Is that even possible? And two, can that even work as a business?

sam lessin (24:44.594)
I think probably not is the answer. I've actually looked at this many times in my career. You go back to like SETI at home, right? Like remember that, and you look at the history. There's been so many attempts at kind of being decentralized. I actually invested in a company that originally was doing this with crypto six years ago. They were like, look, it was more for Bitcoin, for mining than anything else. But like there's people talk about edge compute and how to get access to that. There's all these types of things. And everyone's like, oh, decentralization, decentralization. Here's my observation about that.

is decentralization is always gonna be worse than centralization for anything that's about efficiency. It's not worse for security, it's not worse for trust. There are a lot of things that can be better for, but you will never outcompete a centralized system on efficiency.

right with a decentralized system, at least not for a long time. So maybe there's some spot instance. If you said, look, the world desperately needs compute. We've run out of centralized compute. Let's go find shittier, lower quality versions of it, right? Containerized random instances around the world and like kind of make a few bucks selling them. I guess that might work for a little while in a super constrained environment in theory. But I just think in practice, it's much more likely that centralized solutions that can be better managed.

right, and like have lower cost of capital and that will just win in anything that's about efficiency. Now, again, that doesn't mean it's always the right answer. Like there are times you pay a cost, decentralization is a cost you pay to increase security, to increase trust, right, when you can't, you don't want, when decentralization is scary or bad, but I don't think that decentralized compute is like a long-term good idea, right, if the problem is efficiency.

Prateek Joshi (26:29.169)
Yeah, that's actually a very fantastic take on it because there are like really smart people who are trying to figure out, hey, can this possibly work? And I think it's good that the great people are attempting to do this, but I think you're right. If, when it comes to compute, people wanna pay the highest, the combo of highest quality and lowest amount of money to for like a given unit of compute. And it's always gonna be harder with the.

sam lessin (26:51.986)
So here's what I say, you know where there will be a decentralized compute market? The black market. So there will be, here's a very plausible future, which is most of the world says, look, centralized is going to be more efficient, better, it's where we go. There's going to be a bunch of people who want to run bad models, do bad things, right? And the biggest companies will say no, right? They'll be like, we're not going to like, Amazon will kick you off, right? Like, you know, Google is not going to allow it. Open AI, they won't allow it.

And so will it be like a dark market for decentralized compute, which is more expensive and shittier, but like anyone can use probably, right? I'm not sure that's the business I want to be in, but I think that is a plausible reason why there will be that thing. It just won't, you know, it won't be what you want. Right.

Prateek Joshi (27:38.753)
Right, right. Yeah, that's, yes, I think there's always, I think even throughout history, anytime the governance instruction governments they don't allow something, there's always like a parallel black market going on where the quality is low, price is high, but it exists just cause people wanna transact. So I think that's a great point. Let's talk about the broader like regulation movement that's been going on. People are like, you know, government agencies,

Everyone's getting together to figure out, hey, how do we regulate this thing? And the exec order came out and the recent another fiasco with the OpenAI board drama. But NetNet, what do you think about this kind of movement to regulate AI? Is there merit to it? If so, what are they getting right and what are they getting wrong?

sam lessin (28:31.278)
Yeah, look, it's a little bit of a complicated topic to talk about entirely. What I would say at a high level is this, is regulating AI a good idea? Yes. Right. Like the reality is it's like a super powerful tool, right? That can, you know, people talk about AGI being super dangerous or scary. Again, I think that there is merit to that. I don't think it's actually the problem of the day, right? I think the problem with the day is that it's kind of like having a dirty nuke, right? People worry about countries getting nuclear weapons. They should be.

But the reality is, is like, you don't need to build a full nuclear weapon. You take a bunch of plutonium and wrap it with some other explosives. You destroy a city, right? And so to me, we're kind of in that phase where AI is super dangerous. Um, you know, it's especially mixed with social media. The only way to manipulate people's beliefs, et cetera. It's kind of, again, it's, it is social media on absolute steroids. It is bot farms on steroids. If it used to cost you a ton of money to set up a Russian bot farm and try to have an influence campaign.

over some social issue, now it's much cheaper, and we should expect it to happen much more, right? And so I think any attempt to say, how do we regulate this, or how do we as a country manage this, totally valid, right? And I do think, now the question of what actually is being done, and is it effective, that's really hard, right? I think the unfortunate reality is that in a globalized world, which we live in, right, it's really hard to manage this type of stuff, and to go a step further, I'd say,

A lot of what people put out there is a little bit misdirection. I do worry that there's regulatory capture in Washington, D.C. Where the big companies say, well, this is how you should regulate AI. I say, well, of course you believe that because that's extremely preferential to you. Right? Like, it's not really the right answer. The right answers, unfortunately, for regulating AI are difficult. So it's stuff like, I'll give you an example. National identity, digital identity. It's like, it is insane that as a country, we can't identify in the United States or in any country.

Who's a citizen of a citizen who isn't? We support free speech for our citizens. That's great. Like it's one thing to say like, oh, is this a Russian bot farm or is this an American in America, right? It's another thing to be like, oh shit, is this an AI that's pretending to be this or has this agenda? Where is this coming from? Like when you look back at the history of the internet, I think I totally understand why it was born without an identity. It makes total sense. It made it grow much faster, dah, dah. But like the fact that it isn't baked in at like TCPIP level of like who is speaking.

sam lessin (30:55.302)
is a huge fucking problem, right? And like that is, when you think about regulating AI, identity is the answer, right? It's stuff like that. And the problem is it's extremely hard to pull off, right? Both because it's, people see a problem like AI, they wanna regulate AI. They're like, we need to stop this thing. You're like, well, that's actually a very unsophisticated take and not gonna work, right? And so, I don't know, I do think the government should be talking about regulating AI, and I think we need to do what we can. I do think it's dangerous in a lot of ways, and like that's the government's role.

But I think that a lot of the really right answers are very hard for people to understand.

Prateek Joshi (31:30.889)
Yeah, that's actually a very, very interesting take. A lot of the debate is happening around how, hey, instead of regulating the AI infrastructure, why don't you regulate the applications? And then there's already laws around that, but I think identity is a very, very critical part of this because AI, one of the biggest, most popular use cases or like bad use of AI is like misinformation. Hey, can you, how can you disrupt an election? And it's very effective, right, with that tool.

sam lessin (31:55.963)
Yeah, incredible. I mean, you have to imagine, I think this is the sad part, is like you just, you look at what's going on with Hamas, right, and a lot of the promotion of people like promoting Hamas and that, and like, you're just like, I have absolutely no idea, I have no idea if this is real or not, right, but it does clearly have impact, right, and so that's the thing is like when you have a thing, you're like, I have no, people like, I don't know if this picture is real. Fine, that's an issue too, right, but before we even get to there, it's like,

are these millions of voices talking about this stuff, are they real or not, right? And like, I think that we just don't have any way to do that well. So.

Prateek Joshi (32:33.089)
Yeah, I think that's a very interesting point. All right, with that, we're at the rapid fire round. I'll ask a series of questions and would love to get your answers. All right, question number one, what's your favorite book?

sam lessin (32:46.794)
Ooh, so I have a lot of favorite books. I just mentioned on Twitter though, and the one I would recommend is Lessons of History by Will and Ariel Durant. I think it's like incredible. It's only a hundred pages. It's super short and easy to read. Actually, my six-year-old just asked me for a podcast, and I said, you have to listen to this instead. It's totally over his head. He's never gonna understand it, but I do think it's the thing to read right now as we kind of descend into a lot of chaos, is like, let's just pop up and talk about Big Picture where we're at.

Prateek Joshi (33:14.593)
Yeah, I saw the tweet, loved the recommendation. I'm a huge history nerd, so I'm gonna, no, not yet, that's why I was really intrigued. Yeah, I will. All right, next question. What has been an important but overlooked trend in the last 12 months?

sam lessin (33:19.138)
Have you read the book? He's good.

sam lessin (33:31.65)
God in the last 12 months. Um, I'll talk my own book. I'd say in the last 12 months I think the resilience and growth of Solana is overlooked, you know in a lot of cases It's you know, people thought after the collapse of FTX like oh This is an FTX thing of dead and did take a big dump But like the reality is like it's now trading higher than ever before because it's super strong You know, they're there's really interesting software being built. There's a huge amount of volume going through it

I've always been, I mean, I was a seed investor in Solana, it's been a great financial return for us, I'm a huge believer in it. But I do think that the death of crypto is greatly overstated and then within crypto, the death of Solana was extremely overstated in the last year and it's been cool to see people really start to come back. So, but I think the world hasn't realized that yet.

Prateek Joshi (34:19.629)
I think it's a great position to be in. When you're contrarian and right, I think it's a fantastic position to be in. And I know Solana has a huge, huge win for you and the firm. So that's a kudos on that. All right, next question. What separates great products from the good ones?

sam lessin (34:25.171)
Really.

sam lessin (34:30.606)
Thank you.

sam lessin (34:38.626)
These are hard questions. I mean.

sam lessin (34:45.046)
Look, I don't have a great answer to that. But what I would say is I do actually think it's a little bit like investing. If you think about starting a company or starting a new product, you know, you're invest, you are an investor, you're starting, you're investing your time. And I think the answer is that to me, the great products are the ones that are contrarian and right. And what I mean by contrarian is you have to do something meaningfully different than everyone else. If you're doing something that's like the same as everyone else is like, you're never going to like, you have to people, if you pitch your startup idea or your company or a product idea, people like, we love it. It's probably wrong.

You need some people like, I fucking love it, and then a lot of people will say, I hate it. And then I think the second thing is you actually need to execute. It's not just about the idea. So that's the being right part. It's like you always have to be very contrary and then build something amazing. So look, probably the thing that I would put, I was involved in from the very, very get-go that most pattern measures is Venmo. So I was literally, when Cortina and Ickrum were starting it, I was around the table when the ideas were coming together. I helped with the IP.

You know, my name's on the original patents that were filed. You know, I was their first investor. I was their first user basically. And like, that's a great example of a contrarian and right product that was great and is great. And it, because people thought it was fucking crazy, right. Um, but they were right.

Prateek Joshi (35:59.437)
And it reminds me of one of those products saying, the biggest or rather the worst things is not like negative feedback, it's indifference. Meaning people don't even care enough to like, they'll call you stupid. They'll just ignore you.

sam lessin (36:08.302)
Totally. Yeah. No, no, you can in general, I mean, you have to be careful with this. But it is if you can get people to both love you and hate you. That's great, right? Like you want to be mad at you, right? If you're not if someone's not mad at you or think you're crazy, you're not doing anything important.

Prateek Joshi (36:27.865)
Yeah, that's a very good point. You need both poles. If everyone's called, if everyone's unanimously calling you're terrible.

sam lessin (36:31.774)
Yeah, if literally, look, you probably want at least one person to think you're not crazy, right? Uh-huh.

Prateek Joshi (36:36.054)
Right, right. Yeah, 100%. All right, next question. As an investor, what have you changed your mind on recently?

sam lessin (36:45.178)
Let's have a question I ask people. I don't have a good answer to it myself. Um...

sam lessin (36:51.414)
What have I changed my mind on recently? Um, tick tock. So I would say that like, look, here's what I say. I, um, I think tick tock's a fascinating product. And like, there was a period I went through where I really, really enjoyed it and liked it. Not as kind of like an alternative to TV. You know, I understood the algorithms. I understood all the consternation about it. I was like, ah, this is like a fun toy and I enjoy it. You know, I'd say like, this is not in the last 12 months, but maybe

A year before that, I went through a phase where I really was really into it, and then I was like, I don't need it anymore, and I stopped. And now I'm actually in the camp, but really thinks it needs to be banned in the United States. I'm bummed by that, right? Like I wish that weren't the case, but I do think the reality is, is like, that's a thing where I think the national security concerns are just too great, and I don't think it's a toy, right? So that's kind of what I would say.

Prateek Joshi (37:41.829)
Next question, what's your biggest prediction for the next 12 months?

sam lessin (37:46.062)
Mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm

I think that all the AI investing is going to dry up and people are going to take a lot of losses on it We've already seen that starting to happen a little bit, but I do think that that's kind of my prediction is Okay, they're gonna be a handful of winners I'm not saying there won't be and like someone will make money There'll be any section but broadly speaking like I think AI is gonna go from being a very hot area to invest in to kind Of dead in the next 12 months as people wake up to the realities of what is useful for it not

Prateek Joshi (38:14.109)
All right, final question. What's your number one advice to founders who are starting out today?

sam lessin (38:17.634)
Thank you.

sam lessin (38:21.162)
Um, so the thing I've been pushing the most on with people is I think for the last, and this is a lot in the deck, which maybe you'll link to, and you know, I think it's been cool to watch it be received really well. It was like, I think that the question that most people ask when they go to raise money or start a company right now is what metrics do I need to raise my series A, right? The seat, you know, I'm going to do some seed rounds to raise my series A. I would say that as a, as a new founder, I would no longer be asking that question. Instead, I'd be asking the question of how much money do I need?

to get myself to a place where I have a good business that I can build on, right? And maybe I raise more money and maybe I don't, but I'd be totally uninterested in what are the metrics for a Series A and instead I'd be interested in what's gonna make this a great business that I wanna run and I can attract talent to and start growing, which is kind of old school, right? But I think that's the big thing people come to me now all the time and they say, I wanna seed invest in it and we're gonna do these things, we're gonna have these metrics and then we'll raise a Series A. I said, fine, what if you don't get a Series A? What if there is no Series A? Like what, how much money do you need to

actually build your business. And I think that's a much more interesting and important way to think about the world now than it was a year ago.

Prateek Joshi (39:25.745)
Amazing Sam. This has been a brilliant episode Obviously, I follow all your takes and it's always One thing it just it makes people think one way or the other good or like either They love you. They they obviously many people don't like you on Twitter But but one thing you cannot say is that they know that they ignore it that they will react and I feel like And obviously, you know the portfolio built over the years. I think it Just fantastic. So

sam lessin (39:41.378)
Fine.

Prateek Joshi (39:53.977)
Again, thank you so much for coming out of the show and sharing your insights.

sam lessin (39:56.154)
Thanks for having me and it's nice to be in touch. And yeah, good luck to you and everyone in kind of your community out there. It's a fun, crazy time, right? And I have to say, I enjoy not being ignored. You know, what's the, there's an old.

Prateek Joshi (40:04.849)
Right, right. And...