Why Andreessen Horowitz Is Still Betting Big On Crypto

I am not excited about an agent trading on your behalf. I think that that is very likely not going to work, and the reason for that is that in order to succeed at trading you have to have some alpha.

And if everybody is using ChatGPT to trade, by definition, there will be no alpha. I am more excited about the prospect of agents in AI being used by professionals that fine-tune the agent in a particular way, so that now the strategies of that agent actually are more successful than strategies that do not use AI at all.

You can get so much more done by yourself using AI tools than even a team of 5 to 10 people could a few years ago. I actually have a background in AI. Before I joined the firm, I used to work at Google as a full-time engineer.

And actually, we see the cost of entry has dramatically collapsed and dropped, and you can test your ideas, which I think is great because it means that there can be so many more startups. The downside is it's just hard to keep up.

Are we going to have our own AI agents at some point in the future?

I think it's going to be faster than people expect. John Collison had the five stages of autonomy for agentic commerce. I actually condensed those five stages into four stages, and then I add my own fifth stage. So my own five stages of autonomy are one.

Ali, welcome to new economies. Great to have you.

It's great to be here.

So this thing, AI, it's changing so much and it's changing so quickly. Just in the last few weeks, we've had OpenAI acquires TB PN, we've had Anthropic reaches $30 billion in annual client revenue, we have all these AI models, you know, trying to basically become number one as quickly as possible.

We have a couple of companies potentially going public at some stage. It's just changing so quickly. Have you been surprised how quickly it's all advanced?

Yes, it is absolutely staggering. And especially because I actually have a background in AI. Before I joined the firm, I used to work at Google as an engineer, and I was part of my time there was at Google Brain.

And actually, I was sitting quite close to Dario and to Ilya Sutskever at Google Brain while they were still there. And even though I did not work with them, I definitely got to know them a little bit from that experience.

And it is insane to think how far we've come from that point when it still felt like actual AGI or anything resembling AGI was decades in the future. And now here we are with the frontier models being pretty close.

I think that you could actually already say that we have Artificial General Intelligence, and it really just depends on how you define the term AGI. And it tends to be this shifting goalpost where people define AGI one way, and then we hit that definition, and then they move the goalpost and define it another way.

And we're at the point where I think especially since December, for all of the tasks that involve

that, or for all of the tasks that can be easily verified, AI has become extremely powerful and very, very capable. Especially, I think, for example, in the realm of writing software. And as a software engineer before, I've actually been spending a lot of this time in nights and weekends writing code, but no longer writing code manually and rather using AI tools for writing code.

And I'm just completely baffled as to how much more productive I am, even part-time, than I was when I was a full-time engineer spending every waking moment thinking about software engineering.

It's crazy. And you know, ChatGPT launched November the 30th, '22. I remember the day as it was kind of like yesterday. And what, we're 13, 1400 days in. And if this is how far we've come just in 1400 days, you mentioned like just since December,right?

The next 12 months, next six months, it's going to be exciting.

Yeah, it doesn't seem like there's any sign of it slowing down. We're on a relentless exponential. And yeah, I think that the next couple of years are going to be wild.

Yeah, the definition of AGI, like you said, keeps kind of evolving, and people have their own different perspectives on what that actually means. What's your take on what AGI actually means today?

Well, people debate this question endlessly. I think my personal definition of AGI is the moment when an agent can be fully independently economically valuable. So when you're at the point where you can have an agent that's an end-to-end employee or an end-to-end founder of a company where you can just let it go, let it run out in the wild, and have it create value and actually make money and become self-sustaining, at that point, then you can actually argue that all of the cognitive capabilities that are very human and are required to get to that stage have been accomplished.

And as a definition, I think it's a very complete definition because it forces an intelligence that's measured to that standard to be very well-rounded and to not be as spiky as machine learning models and AI agents are today.

You know what this means? It means so many cool new startups can be emerged. And the great thing about being a venture is we get to know all the latest tech trends and hopefully meet some awesome founders. You guys, a16z, being on kind of a rocket ship, it feels, over the last few years, I think most of our audience will understand and obviously know of what you guys are focused on, but maybe help us just paint the picture.

What's the latest happening on the crypto firm and what are you guys most excited about at the moment?

Yeah, happy to. I'm happy to. Maybe I'll start with the crypto fund and then talk about kind of the context and where we are now in the space. So the crypto fund exists at the firm with the sole purpose of specializing on crypto companies, companies that are entirely focused on building protocols or building around protocols, building around decentralized systems.

We started the vertical back in 2017, 2018, and started a fund that is kind of the same a16z mold, but very tailored for the needs and requirements of companies in the space. And in particular, crypto companies are unique in that they need much more help on both regulatory legal policy, just given that there's been so much uncertainty on that front in the space since its inception.

And then also, in many cases, on the tech research front, because there are so many open-ended research problems that are core to the space because it's so nascent that people who have spent a lifetime studying those disciplines have a lot to contribute to.

So two of the major teams that we built in-house to be able to really help our crypto companies are a legal regulatory policy team that has people from government, including people like Bill Hinman, who was a former director at the SEC, our general counsel, Miles Jennings, who's a former partner at Latham, and has some of the, like one of the deepest understandings of the space whilst at the same time understanding all of the legal context and has been able to push forward a framework for how to think about decentralized systems at a legal level.

And many other people on that team to help our portfolio companies really think through their specific legal regulatory strategy to make sure that they're always compliant with the law, but also make sure that they're making progress and being aggressive with what actually is possible.

The research and engineering team, on the other hand, is a team of world-class experts in distributed systems, cryptography, mechanism design. We even have people from sort of political science to help our portfolio companies think through all sorts of questions from protocol design to governance to scalability of distributed systems to cryptography, et cetera.

And that's both on the engineering front and the research front. And we have a research team that includes people who come from academia and have

sort of longstanding backgrounds in those areas. And then beyond that, I think we replicate the a16z mold in that we also want to help our companies on go-to-market, talent, and marketing. And those are the sort of the major functions, the major areas that we tend to kind of help our portfolio companies with.

And the whole team is a team of around 70 people here at the firm, with 10 people, 10 or so are on the investment team. The other 60, 70 are entirely focused on helping the portfolio.

So I had your colleague, David, on the show a few months ago, and we got talking around the what does the future venture actually look like? And what might be surprising to a lot of people is actually, given the team size of a16z, most of your team aren't sitting on the investing side, they're sitting on the platform side.

Why do you think it is so unique and so important to have such a well-integrated platform team who can help with all of these things, like you mentioned, go-to-market, especially around regulations? Because actually, very few other venture firms do this.

Yeah, I think that it is a product of the fact that technology is becoming a much more important facet of human life to the point at which companies like startups end up actually becoming far larger scale than they used to and needing to touch much more of the various different aspects of the world than they used to.

I think we lived in a world before where a company would really just build a tool, and the tool was something that then some other company that is a non-tech company would take and then would use and would pay for.

Like, for example, I think you would build like software for taxi companies is maybe the canonical example, as opposed to building Uber, which completely replaces the whole taxi model altogether in an end-to-end way and in a vertically integrated way.

And building a company like Uber is a much more challenging thing to do than just building a tool for taxi companies because of the fact that now you have to interact with so many different aspects of the world.

And so I think as time has gone on, we've entered an era of technology development where startup founders need so much more support from their investors in order to be able to deal with all of those challenges. And we found that it is a massively amplifying capability to be able to project power on behalf of our portfolio companies, help unlock doors, help them essentially be an extension of, or essentially use the firm as an extension of themselves in a way that I think they really can't do with any other investor.

And I think especially for founders who are highly technical and who maybe come from an engineering background but don't have the background in the legal policy regulatory component or don't know how to hire a world-class CMO or a world-class general counsel, these are things that we can really help with.

And so I think that the whole mold of the firm is how do we make innovation and the development of technology as efficient as possible? And that's kind of the reason why we believe like the vertically integrated model of a firm with a platform is the best model.

Super unique. And hopefully with that, you know, you get access to the best founders,right? But hey, the whole crypto blockchain category has, I think it's fair to say, it's had its ups and downs,right, in the last few years, maybe since '21, when the whole craze of NFTs kind of came around and it was kind of this whole category exploded,right?

Then we had a whole bunch of stories and a whole bunch of events which happened, collapses, and so on. Help us paint the picture. Like, what is actually happening in the world of cryptoright now? And should we still be excited?

Yeah, I think it is funny that we are now at a point where the sentiment about crypto is close to an all-time low, at the same time that the fundamentals are very clearly at an all-time high. And that we have some amount of regulatory clarity in the form of the Genius Bill that passed last year, which created a legislative framework for how to issue a stablecoin, which has led to increasing stablecoin adoption.

We're now at something like $200, $200 to $300 billion of circulating supply of stablecoins, more than a trillion dollars a month of volume on stablecoin activity. Many large financial institutions from the tradfi, traditional finance world, entering the space in a way that's completely unparalleled, in a way that is unprecedented.

All of this happening largely because I think we now have some amount of regulatory clarity of how crypto will interact with the space. And at the same time, I think that the sentiment about the space tends to be still very negative because I think it hasn't really percolated or made it out into the world that the level of traction that we've seen is real.

And it also hasn't really made it to the consumer. All of this progress has happened behind the scenes. And this also is happening with the backdrop of AI being an extremely exciting alternative sort of category of technology, which in many ways has sucked the air out of the room and has, I think, drawn a lot of founders and a lot of energy and capital to that world and away from crypto.

So in some ways, it's actually an interesting opportunity because crypto is contrarian. It's very much not consensus the way that AI is. And at the same time, I think we've de-risked the space more than it's ever been.

It's super exciting. It feels like it's emerging back a little bit. But so AI is being almost at the forefront of attention,right? If we go back to '21, this is obviously like when the NFT craze came about. One of the biggest surprises I actually found is not a huge amount actually came out of that if we look four years forward.

The decentralized social networks, like I covered a lot around the creator economy back then. And there were ways where fans could buy into the fandom across music, newsletters, content, and so on. But actually, if we look four years forward, not much has really emerged from that,right?

What do you think we can learn from those times? And do you think we jumped onto the trend too quickly?

Yeah, there are two views you can take. And this has been a debate that's been raging on Twitter for some time. One view is that none of those ideas were ever going to work. And crypto really is purely a financial technology and will only ever be a financial technology.

The other view is that we were too early, that those ideas actually still have some merit. And it's just so happens that they were introduced at a time when the technology wasn't really ready for those ideas. I tend to believe the second, the latter, largely because I think the fundamental reasons why none of those ideas worked back then and haven't worked still are, one, that

I think the bar for a consumer experience is a very high bar. It's probably the highest bar when creating products that you would have to meet in order to be successful if you have a consumer go-to-market. And I think because the technology is nascent, it is doing consumer in hard mode, where building a consumer user experience that matches that of traditional web 2.0 centralized consumer products is an extremely difficult thing to do when the underlying infrastructure has only really been around for, say, 10 years, 12 years since the launch of Ethereum.

And really, I think it hasn't really been around in a form that could be used to build consumer products for more than a couple of years. So that's the first, the first reason it hasn't worked. I think the second reason is that in order for a consumer product to work, you need a market, a market to be able to target.

And I think that the fact of the matter is the market in crypto of people who actively use crypto and blockchains directly is actually very small. It's hard to actually know exactly. There's lots of people who hold crypto, but the number of people who transact with a blockchain directly and interact with blockchains on a daily basis is minuscule, probably in the maybe tens of millions.

I would suspect it's actually probably much smaller than that. And when that is the case, building a consumer product is extremely difficult because those are the only people that you can hope to get on your product, especially when the backdrop is the regulatory environment that we had before in the past administration when everything in crypto was cast as a scam or as illegal or as purely for speculation or as a casino.

I think it was a perfect storm for making crypto adoption by consumers extremely difficult. I think that now, with a little bit more regulatory clarity, with the entry of some of these major financial institutions into the space, with the fact that we now have scalable blockchains that can do sort of thousands, tens of thousands of transactions per second, you can move any amount of money or any transaction from anywhere in the world to anywhere else in under a second, that now actually creates the conditions for some of those ideas, maybe down the line.

Maybe it's not immediate, but in the next, say, 5, 10 years for those ideas to actually work again.

Totally. I totally agree. I mean, back in '21, this was, again, it still felt like so early. And actually, a lot where the adoption was happening was actually in Southeast Asia. Philippines, I think, had the largest amount of NFTs, followed by Thailand, actually, where I'm based.

Something like, I think, 10 to 15% of the population owned an NFT, which is crazy.

It is, yeah.

Crazy,right? I think, though, where we're going is it feels on the privacy side, is privacy the new moat for startups? Because when I feel when I meet all these new startups coming up in the crypto space, they have the distribution, which is awesome.

But the biggest concern they always have is, and for the consumer as well, am I going to lose my money? I don't understand this whole crypto space. But privacy, if you have amazing privacy, then surely that's going to be the new or potentially the next moat for startups.

Do you think that's accurate? And if so, where do you think we're heading next with that space?

Yeah, I actually wrote a tweet about this, about how privacy can actually be one of the biggest sources of defensibility for crypto companies. I wouldn't say it's the only source, nor necessarily the most important, but I think it may become a major one.

And I think if we step back a little bit, historically, crypto has really not prioritized privacy as part of its roadmap, in that every blockchain that's been built up until very recently, with very few exceptions, has been a fully transparent blockchain, in that anyone who's motivated enough to look can actually trace all transactions in the network.

And I think the reason for that is that privacy is at a whole other level of technical difficulty. And for the longest time, people were just focused on the low-hanging fruit of actually building a blockchain that's truly decentralized, that's verifiable.

And also, there was this ethos of optimizing for verifiability as opposed to privacy. And these two values are somewhat in tension with one another. There is technology that actually can get you both. But for the most part, absent key technological breakthroughs, you either have verifiability or you have privacy.

But I think, obviously, more recently, with the entry of some of these major financial institutions into the space, it's become extremely obvious that crypto will never really become mainstream if it doesn't have privacy. It's inconceivable that you would have businesses running on-chain and potentially paying their employees if every single transaction is actually decipherable and visible to the general public.

So privacy is maybe simultaneously the least, I mean, maybe the most important feature in order for crypto to really succeed, and also the most overlooked, that most projects have not emphasized it or spent that much time on it.

And it's been an afterthought for the most part. I think when we talk about defensibility, the reason that I believe that privacy can help crypto protocols in particular become more defensible is that whenever data is out in the clear, it is inherently easy to port.

It's easy to move it because it's in the clear. You can just look at it, copy it, move it somewhere else, replicate the service somewhere else. And as a result, the switching cost for a user who's using a blockchain system or a protocol, if the protocol has no privacy, is actually much lower because data can be migrated in a way that's quite easy and can be transparent to the user.

The user might not even have to do it manually themselves. It might just sort of happen automatically, or whomever the competitor may be able to do that on behalf of the users that they're trying to win over. Whereas when things are private, things are encrypted, things are no longer visible to everyone who might care to look.

In those cases, the migration process is much harder. So the switching costs are much higher than they are when things are public. And add to that, there is friction in moving private data from one place to another because in order to do that, you have to extract the data, decrypt it, re-encrypt it.

And there's always risk in that whole process that some of the privacy is lost and that the data becomes unveiled and your privacy, like your data gets leaked. So maybe there's also unwillingness on behalf of customers to switch, even if maybe they could, because they wouldn't want to take the risk that their privacy is violated.

And then add to that, finally, that all of these systems have network effects. You want to be on the blockchain or the network that has the most users because those are the users that you can most immediately transact with.

And it's expensive or more difficult and maybe incur some risk of loss of privacy to transact with people outside of that network. So the networks that have privacy that begin to win and have the most users are likely to continue to win because those are the networks that people are going to want to join.

And so that's kind of my kind of my spiel on why privacy is important and why I think also from a business strategy standpoint, it may be critical for defensibility for protocols.

Yeah. To double down on that, where do you think founders, where do startups go wrong from a privacy point of view? Because at the end of the day, this is financial services and people want to make sure they're protected,right?

Where do you think startups go wrong and where should they do better?

One very common mistake in this space, and this is somewhat controversial and people might debate me for this, but there is the cypherpunk crypto anarchic ethos of you must have privacy end-to-end that is uncompromising and is as secure as sending, as basically transacting in cash or sending a fully encrypted message from one person to another, visible to no one and subject to no laws and regulations.

I think the reality is that that is never going to work. Especially when we're talking about the transacting of or the movement of money, that is never going to work in a world where we've got things like sanctions laws and a very strong desire on the part of governments to not have money be funneled to terrorist organizations or countries like North Korea.

That's just never going to work. That may just remain a niche kind of sort of technology that's available out in the fringes. But in order for the technology to become mainstream and have a real impact on the world, it has to be accepted under the legal frameworks of the governments that we all live under.

So the reality, I think the more pragmatic path forward that is much more likely to succeed at a larger scale is one that is a middle ground that provides users with some protection, but also has compliance features and the ability to prevent money from reaching certain actors in the world that are bad actors.

And the way to accomplish that, it's very hard to do that

upfront because the set of requirements is amorphous and opaque and constantly changing. The way to accomplish that is with programmability. And this is where recent technological developments that allow you to build encrypted networks, like blockchain networks, that simultaneously give you the privacy that you want, but are also programmable and allow you to encode certain compliance features and certain mechanisms for allowing law enforcement to do what law enforcement has to do in a way that respects privacy and still maintains some of the benefits of having end-to-end encryption.

That's really the only way, the only path, the only path forward.

It also feels to kind of help uplift the crypto negativity. There's this massive now opportunity with the intersection of AI. And one of the trends I've been hearing a lot about is, are we all going to have our own AI agents at some point in the future, which can transact, can do trades, can manage our finances on demand?

What's your take on that?

Yes. Actually, you may have read Stripe's 2025 annual letter. In that letter, John Collison had the five stages of autonomy for agentic commerce. I actually condensed those five stages into four stages, and then I add my own fifth stage.

And so my own five stages of autonomy are one, getting rid of the tedium of entering payment information on forms online and having all of that be automated by an agent. That's stage one. Stage two is

descriptive search, which is that you can articulate things that you want to buy. You can name them by name. You could be like, I want like pens. I want to buy particular clothing items, etc. And then an agent will go find those items, figure out what the best price is, and then execute the transaction for you.

The third stage is delegation, where it's an even higher level of autonomy and that you'd just tell it, these are the high level, the things that I want, and here's a budget. And then an agent figures out end-to-end which items to buy and then buys them for you.

The fourth stage is anticipation. These are still kind of coinciding with John Collison's, but anticipation is the agent knows you so well that it knows what to buy before you even ask it to go buy these things for you.

And then I would add the fifth stage of autonomy, which is agent-to-agent commerce and more broadly agentic finance. And this is a world where we're actually moving beyond what's really possible with the existing financial system. And we're not just paying for things in the way that we're used to.

And instead, there is a much larger number of transactions that happen between agents for whatever reason. It could be that maybe agents are buying compute for themselves or buying data or transacting in skills, transacting in other resources that become valuable in a more agentic economy where an agent knows what it needs in order to do its job and can go find those things from various places, not just from traditional visa merchants, but also potentially from other agents or other individual humans.

And you can also imagine agents not just paying each other, but also entering into contracts with one another or with humans that are enforceable in software. And this is a notion of a smart contract. It is possible to do that on a blockchain in a way that it would not really be possible to do with a traditional legal system.

And that would be maybe the fifth and kind of ultimate stage of autonomy for agentic commerce.

This is so exciting. We've seen a couple of startups like emerge in this space now, especially from the commerce point of view. I think one really exciting opportunity might be controversial potentially is how do these AI agents start making trades on your behalf?

Are we going to go down that route? I have mixed opinions. What are your thoughts?

I am not excited, certainly not in the short to medium term, about an agent trading on your behalf if you are sort of a consumer out there who just wants to make some money and delegates portfolio management to ChatGPT.

I think that that is likely, very likely not going to work. And the reason for that is that in order to succeed at trading, you have to have some alpha. You have to have some kind of differentiation over everybody else who's trading.

And if everybody's using ChatGPT to trade by definition, there will be no alpha in the trading strategy that ChatGPT implements. I am more excited about the prospect of agents and AI being used by professionals, by people like, say, a hedge fund that does have, say, proprietary data or maybe fine-tune the agent in a particular way so that now the strategies of that agent actually are differentiated and do contribute to a maybe broader trading strategy that is more successful than strategies that do not use AI at all.

I think that is almost inevitable, almost certainly already happening. But the more kind of the broader kind of consumer focus, let's like just delegate wealth management or maybe wealth management to some extent could be delegated, but actual trading and trying to find good stocks or things like that, I think is not likely to work.

Yeah. It's actually a great point. All these AI models, as I see, I actually had discussions with the Yahoo CEO who we had just on the show yesterday. And the same concept is around with all these AI models is spitting out the same information than what you'll ever write.

Same kind of applies to the commerce, agent commerce. What do you think? How far away do you think we are from this being fully adopted from the consumer landscape? Are we 6 to 12 months away? Is this 2 years away?

I mean, this whole category is just shifting so quickly,right?

I think it's going to be faster than people expect. Already, many API endpoints for things like if you're a developer, for example, I think pretty quickly it may be the norm to pay for any kind of API request that you might make using X402, which is a protocol that enables you to pay for things using stablecoins as opposed to the traditional way of you sign up, you enter your financial, your name and credit card information, and then you get an API key, and then you give that API key to an agent, which ends up actually being if you write code in these days, if you write code and your code needs to interact with the external world, that ends up being the most time-consuming part of actually writing code because the agent can automate everything

else, but it can't go and sign up on your behalf and enter your credit card information on your behalf and get you an API key and all of that. And you as a human have to go do that.

And you end up being the bottleneck on kind of the route to progress. Whereas in a world where you just give your agent a crypto wallet and maybe you approve of a particular expense, but it just happens because it's a stablecoin transaction and then the API endpoint can just service your request almost immediately, that is a much better model.

And I suspect that'll start to happen soon. I think we'll also maybe start to see

we'll start to see people being able to buy things on apps like ChatGPT directly. That I think is already beginning to happen. And in the back end, that may to some extent be powered by crypto. As you know, I mean, Stripe is doubling down on crypto.

They acquired Bridge and Privy to startup companies in the space last year and are using crypto as a way of optimizing the back end of what Stripe does. And so then the short to medium term, maybe you as a user don't fully become aware that stablecoins are a part of the puzzle, but they will be.

And then I think maybe longer term, the user, the end user might actually begin to interact with stablecoins directly.

You know, we've been talking around, there is so much opportunity in this space and this whole AI category is just shifting so quickly,right? One of actually the concerns I have is it's all great, but actually I think it's a massive distraction for founders.

The pace of innovation is moving out. One day is this and the next day is that. Then startups need to shift in this different direction. Are we moving too quickly?

I think it's, I mean, the upside of moving too quickly is moving quickly,right? It's the progress that we see, the incredible increase in capabilities that we have, the amazing new products that get built and delivered. The downside is that it's just hard to keep up.

But that's the name of the game. I think the best founders are the ones who have a preternatural ability to stay on top of it and to ride the wave and continue to build even in a rapidly shifting landscape that's changing everything every day.

It's changing so much every day, but also the mighty talent coming out of some of these AI companies who are then going to found their own companies is super impressive. You're meeting tons of founders at the moment. Has the fundraising cycles and how the fundraising processes work changed for founders from four, five years ago to today?

And what does that actually look like today?

I think the biggest change, and this is across the board, it includes crypto companies as well as any other kind of company, especially a software companies, is that you can get so much more done by yourself using AI tools than even a team of 5 to 10 people could a few years ago.

And so then that shifts the goalposts and the staging of when you raise money and what the expectations are at the very early stages from an investor in order to make an investment because it is now very feasible to build a first version of the product and launch it and have it have traction, have actual signal from the market whether it works and whether people actually want what you built, which I think is great because it means that there can be so many more startups.

It means that

the cost of entry has dramatically collapsed and dropped, and you can test your ideas and be sure that they're going to work or have far greater confidence that they're going to work before you go and raise a big round and make a big commitment and hire a bunch of people.

So that I think is the biggest change. And then I think also maybe at the later stages, it's just the case that people, what we see so far is not that people use AI instead of humans. It's that people like founders leverage humans who are very, very good at using AI to increase or double, triple, or 10X their productivity.

You want maybe as many people as you had before or more people, all of whom are very competent at using AI as a way of moving 10 times faster. That's much better than, say, trying to cut and having only a tenth of the number of people that you would have had before and moving only as fast as you would have in the previous world.

It's such a good insight. We had Wade Foster, the founder of Zapier on the show towards the end of last year, and he wrote this amazing piece around how actually teams are changing with AI roles versus how they were before.

Do you have any insights, especially into crypto companies, on the types of roles with AI that these AI startups want versus before?

Well, one conversation I have with founders all the time is the question of, do you want senior software engineers or senior people generally versus more junior people who are very good at using AI? And it is an interesting debate because on one hand, senior engineers in particular may be stuck in their old ways and culturally are more resistant to adopting a new way of working and new tools, but have the seniority to be able to know what the architecture should be and being able to kind of steer and manage, say, a cohort of agents to do theright thing.

Whereas a junior engineer might be fully on board and pedal to the metal on learning the tools, reinventing the way that software engineering is happening from the ground up. And that is like a real tension. And I think that the unicorn is obviously a senior engineer who is extremely adaptable and willing to reinvent themselves overnight by kind of throwing out the previous playbook and completely adapting to the new tools.

Those people are rare, but it's kind of critical that most of the people, at least if you're at a later stage, that you have some of that energy within your team to really be able to shape the culture for how an engineering team operates.

And then I think also the culture of adopting AI and reinventing the way that works gets done has to come from the founder and has to make sure that if you do have some of the senior engineers who are resisting the change and resisting the new paradigm of how work gets done, that those people don't end up completely warping the culture and preventing the whole organization from evolving and from moving into a new way of doing things.

And that tends to be, I think, one of the debates is like, okay, senior, junior. And then there's also the debate as to can an engineer also do design or can a designer or product manager also do engineering?

And all of these roles begin to blend because every person in those roles now has access to these superhuman capabilities that can do all of the roles. I think it's going to be very interesting to see the way that organizations shape out and how they change over the next 5, 10 years and what the best mold for running a new company will be.

And I think it's kind of uncertain. It's hard to predict.

I agree. Also, some of this young talent, which you see on X going viral, some of them are 16, 18, they're making a million dollars a year already. They've literally got like a one AI product and a one-man team.

It's so incredible how quickly this category has shifted just in the last few years.

I think how teams are constructed is going to be super interesting to watch and how big will they be, how small will they be, and we don't need thousand people teams anymore.

100%. I think that maybe the 1,000-person team that also leverages AI effectively might be still like a 10X to 100X more effective and capable than the 1,000-person team of before. So we may still have massive organizations that are just that much more capable because they've figured out a way of leveraging both humans and agents in a structure that is just so much more effective than anything we could have imagined, say, 10 years ago.

Exactly. Times are shifting fast. Let's talk about some of the tech trends. Love to hear some of the trends you're excited about. One that I think is really excitingright now and love to give your take is around these predictive markets.

Everyone's talking about them. Is it a massive hype and is it here to stay?

Prediction markets are a very interesting mechanism for surfacing information that can actually be very valuable for society. Like you have prediction markets around scientific breakthroughs or around geopolitical events or about all sorts of things that just humanity would like to know more about.

And that knowledge tends to be locked away in the minds of people that are sort of dispersed far and away. And it's very hard to get access to that knowledge and to tap those people to be able to have the understanding that they have.

The creation of a market and an incentive for those people to be able to monetize their knowledge is a great way of making humanity as a whole smarter. So this is the reason that prediction markets are valuable and interesting, I think, from our perspective.

Of course, there are other aspects to this, like people like to gamble. That's more of an entertainment product and that's fine too. The big unlock more recently is the fact that, at least in the US, the legal path for prediction markets has become so much more clear.

And that's, I think, the reason that we're seeing kind of the emergence of prediction markets as a big category, in particular through companies like Kalshey, which are based in the US, fully regulated and compliant with US law. And I think that the legitimacy that that lends is what's kind of making them a more exciting category than it used to be.

How many market leaders do you think there could be in this category? Is it just a handful or do you think we're going to see more enter the space? Because it's already pretty competitive.

It's an interesting question. There is an argument that

we would expect to see different kinds of markets that are fairly verticalized, that specialize on specific types of markets or predictions and really tailored to a particular kind of end user. So, for example, we've seen several startup companies that want to build prediction markets around everything that has to do with AI compute.

Like if you wanted to, for example,

bet on the price of tokens or of GPUs in the next five years or something like that, maybe there should be a prediction market that allows you to express that kind of financial opinion.

On the other hand, there are also strong network effects that all markets, all exchanges have, both from a liquidity standpoint, from the fact that they have a strong brand. So there's also an argument that you could make that actually there will be consolidation and that the biggest players are going to be able to expand horizontally and serve all of the different kinds of markets that people may be interested in.

I suspect the answer will be somewhere in between, but it's very hard to predict a priori.

Super interesting. What are some of the other trends you're really excited aboutright now? And if you were to start a crypto company or even an AI company tomorrow, what category would you build in?

That's a great question. I think there's two ways to think about this, or there are maybe two molds that you might opt into. One of them is to build something that is like zero to one or something that is one to one.

Or maybe another way to put this is you can opt into improving a particular kind of human activity that already exists, or you can opt into creating net new activities that have never existed before that are now enabled because of new technology.

And those two are very different. And I think in crypto, I'll give you examples of both. One very big trend is this intersection between crypto and traditional finance and fintech. A lot of those companies are using crypto to improve the existing financial system and to make it more efficient, to make it faster, to make it more user-friendly, to make it more global.

But these are behaviors and activities that humans have engaged in for a very long time. And the technology is just making them maybe 10X, hopefully 100X, in some cases much more, like 1,000X, a millionX better. The other example of companies are companies where there's no precedent for the kind of thing that you're building, for the kind of new behavior that you are unlocking.

Some of the interesting companies in that world are in the crypto AI intersection. So, for example, a very futuristic idea that falls under the stage five autonomy set of capabilities is one where you have an agent that has a crypto wallet that is able to pay for things and get paid for things.

It potentially is also able to raise money and get investors and shareholders, and then is also capable of building a product or creating content or maybe trading in some cases if it manages to get some kind of differentiated edge.

And that that entity can exist even without really having any kind of direct human supervision. That's a futuristic idea that has no precedent and that's more of a sort of zero to one type of breakthrough. Very exciting. It's a very different kind of founder who would go after that opportunity than maybe some of the more kind of one to n type of opportunities.

If that happens, not only will it be incredibly exciting, but also just how we work, how we live, what our parents think of all this, it's just going to change everything,right? And bearing in mind we're 1,300, 1,400 days into kind of this AI new journey, I'm really excited what's going to happen next, but I'm also slightly skeptical.

I think my view is that it's a question of when, not if. And I think given that we are on this relentless exponential, I wouldn't be that surprised if we start to see things like that in the near future.

I agree. Well, hey, this has been super exciting. We've learned so much. To finish off all of our episodes, we love to do a quick fire round. So I ask some quick questions, you answer one word back. How's that sound?

One word, okay.

Yeah. Okay, your favorite AI companyright now?

XAI.

They're pretty exciting. Again, XAI, how far they've progressed just in the last few years. Incredible company and excited to see kind of where that goes next,right? If you could have one underrated company start up in the ACCZ portfolio, who would you choose?

Dylan Field at Figma.

Love it. If you were to start a company tomorrow outside of AI and crypto, what category would you build in?

Robotics.

Awesome. What type of robotics company would you start? Do you have an idea?

No, not exactly. But I do think that is the next big wave. We have digital intelligence. The next big unlock will be physical intelligence.

And also just with the robotics, the likes of Waymo, what Travis is doing at Uber, I think I agree. I think this is going to be really exciting. And actually, from the venture standpoint, there are quite a few robotics venture funds coming into the market, which I think potentially sets the trend.

Yeah. I think that is also probably the best field to go into if you're currently entering college because I think it's going to be maybe like a 10 to 15 year arc for getting robotics fully deployed and out into the world.

It'll move slower than AI, which could be a good thing because this way you can enter the space. There's like time to enter the space and become an expert and become a leading contributor to that field in a way that's harder in the world of software where everything is getting devoured by AI.

It just takes so much time. Like look at what Travis is doing,right? I mean, would you have a robotic chef inside your house?

Yeah, certainly. Sign me up, please.

Same. It's going to happen, hopefully. Hey, Ali, this has been amazing. Thank you so much for your time. And we can't wait to see where A16z goes next and this whole AI category.

Thank you so much. It was so fun.