The coming Agent to Agent Economy | Shikhai Wei | Carra Wu | Eskender Abebe | Jeff Wilser |

Okay, friends. If you left a teammate or developer at home, everything is being streamed on YouTube and Twitch so you wo n't miss a minute, so do n't forget to share the information. Ok. Who's ready for the agent-to-agent economy? Please join me.

I can hear applause. Yes Yes me too. Excellent. No longer just tools, agents are now interacting and collaborating and creating a really interesting new economy and marketplace. So please join me in welcoming Jeff Wilser, host of the People's AI podcast.

Thank you. thank you so much. It was great to come here. hello. I am very excited about this topic.

This is the most talked about topic in this field at the moment. So I have three very intelligent people with me to discuss this. Welcome Karu from A16Z, Skender from Eliza Labs and Shakai from Asia Starter. So, there's a lot to discuss here. This field is very dynamic and is growing day by day.

To understand this and create a structure, I'm imagining a three- part or three- part or three- act play here. The first part is about today , the present. Where are you with AI agents? Second part, the future. Where is this going?

And the third part, how do we get there? What are the challenges? How will we overcome them ? So, let me start with a basic question. What is the current landscape and situation with AI agents, and specifically decentralized crypto AI agents?

His field is very dynamic. There are hundreds of hundreds of thousands of projects out there. projects out there. Kara, before you tell us a little bit about yourself, what are you focused on and then what are your thoughts on what's working right now with AI agents and what's not ? Yes.

I can quickly introduce myself. I'm Kara. I'm a partner on the crypto investment team at A16Z. I spend most of my time looking at the application layer of crypto. So those games used to happen a few years ago.

My background is in games. Uh, I obviously spend a lot of time with AI these days , which is why we're here today. Uh, to Story, which has a big presence at ETH Denver today. Uh, where are we in terms of product market fit for AI agents? We talked about this a little bit on the phone earlier.

I think agents are really working in two places. One is agents as entertainment, which people underestimate. I think this category of agents is very similar to streamers, right? There was a time when people said, " when people said, " Streaming isn't a real profession. Like being a Twitch streamer is n't a career , right?

Like it's not an industry." Today it has become absolutely clear that this is true. And I think agents, especially agents who have all kinds of capabilities to render their own images or stream and interact with their fans , that's a real industry, right? Just like any other influential entertainment like gaming platform is a real business. And so I think this is an area that works together with AI agent tokens and Mimecoin trading as a product that has product market fit.

I think this is an area where agents are working. And I think another place where agents are working is coding agents. And as most people here in the room know, coding agents, at least in their early days, were largely centralized. But now there are efforts in Web 2 and Web 3, which are trying to make them more open source. Like Open Router is trying to make it possible for you to not be tied into your subscription, right?

For you to basically split your subscription into a bunch of different APIs, all the way up to Skyfire, which allows you to create agents that are n't tied into any API , right? And so you can do usage based pricing for any API , like bring your own API for this agent and instead of paying $25 per subscription, per API access, you can pay per token. This is a very powerful thing. Uh, but the short of it is that I think crypto plus AI agents is a very new area , and there's a lot of infrastructure left to get to the point that I think a lot of people envision , where you have these sovereign agents that are like your personal assistants that work for you , interact with wallets , you know, interact with other people on your behalf , go on dates for you on behalf of others , right? I think there are a lot of uh this is obviously the most important use case.

How can I get my agents to date each other? Let's make it. I've talked about this at length as an AI agent trained on your personal data set. Interact with another agent and train on someone else's personal data set. Give me a confidence interval about the probability that that data is correct.

Done. It's done and it's done. Ok. Shakai, speak here. First, tell us briefly what you're focused on and what's working not working in the current agent-to-agent economy?

Sure. Hello everyone, I 'm Shikai, co- 'm Shikai, co- contributor, we recently changed the name to Virtual Ventures. And I'm also the co-founder of Longash Ventures. So I am investing in this area , but now I am in the agentic area and let me tell you a little about Virtual, if people do n't know then let me tell you that this is a platform for giving tokens to agents and creating an agent society of agents. Now where we are today, if I look at it from a different perspective from the market perspective , I think we have just seen the first wave of the wildfire that has spread all around , right?

And, and it's completely from novelty from novelty , and people were like, " , and people were like, " Oh wow, agents can do this, AI can do that," uh, and we just, as long as it's done by an agent or like somewhat autonomously , people were like, "Oh wow, this is crazy," and we're paying attention to that. But I also think that we, we, that first wave, is fading away , right? It's starting to be like, okay, well, no, this is kind of good, but it's still not that good. And now we're starting to see , okay , now we need real embers, like real uh utility or usefulness to come out. Uh, and I think this also applies to what you just talked about you just talked about , entertainment agents , entertainment agents , uh, right?

Initially, people are like , oh my god, there are all these personalities and musicians, the art that is being created. And that newness, in fact, led to a lot of speculation, and we got a little ahead of ourselves , and now it seems we're a little irritated , isn't it? We get these answers , and we saw this with the flagship agent on Virtuals as well. With Luna, initially, people were like, “Oh my god, she has a brain, she’s building campaigns. ” And now, we're still doing a lot of things , like Luna doing tarot card readings at the Story Protocol booth which is great, but I think people are starting to demand a little more from these agents.

I would say the other thing that's working better right now is people demanding more from influencers, right? Like if you go back and watch those old school YouTube videos , you'd be like, "That was fun." But in the 2010s, we were like, "This is the funniest thing I've ever seen." yes of course. I think that's the thing about entertainment.

It's like fashion, isn't it? As the meta matures , you need the next wave. That's why work is being done on this. Um, I would say there is still some upside here , an example of this is AIXBT. I think people are still following it and they're starting to see the cracks in its armor.

They're like, " Oh, you're not always right." Like sometimes you get the number wrong and change your mind. You are bullish about everything. Uh, but it's still a great way to uncover timely insights and alpha from sources you might not be following on crypto Twitter. So, and then we're starting to see a lot of builders trying to integrate DeFi, integrate social , and hopefully we'll see more of those.

Uh, we're seeing a lot of people trying to do agent-to-agent communication. And uh, with that , I'll hand it over. Yes, Skandar, tell me who you are, what you're focused on , and what's working and what's not working today. Yes. Hi, I 'm Skandar.

I'm the Chief Product Officer at Eliza Labs. At Eliza Labs, we make Eliza OS. It is a multi- It is a multi- AENT framework that lets anyone anywhere create, manage, and deploy their own autonomous agents. And then there was the question of what the state of the game was. What were you guys called before ?

We were previously called AI6Z. Thank you. And is there a lawyer among us ? Ok . And it came out of a love for the original A16Z where we really wanted to replicate the kind of support they provide to their ecosystem , the support we can provide to our developer ecosystem.

And then when it took off and the agents really caught everyone 's imagination and then our IP started getting a lot of notice, we realized we had to make a change and bring ourselves closer to the actual Eliza OS framework. Um, yes. And then, for the state of the game with agents right now, I think when social agents came out, it captured people's imaginations because they were used to interacting with Cloud or Chad GBT and having a very sterile experience with LL; And then seeing these people abuse or constantly use personality leveraging tools, and then using them as part of their persona on social media , opened up people's imaginations to what they could do. And then now it's gotten to the point where, okay, all the low hanging fruit like deploying social agents very quickly. The novelty of it has worn off somewhat and now the hard part of engineering things like how do I get the actual memory to work the way I want ?

How do I give this agent the tools such that I can actually perform some of these tasks on my behalf with a high enough confidence interval that I'm actually going to pay it a lot of money , are starting to emerge. So we caught people's attention early on and then the industry moved on, but now I think people have gotten their attention and they're very interested in agents and now they're tackling the hard engineering problems of enabling agents to perform better. So I know that making predictions two weeks or 10 days from now is foolish in this area. Kara said backstage that when you guys invest in A16Z you see 10 plus 10 12 years , right? So, I'm thinking, let's get started.

What will success look like in 10, 12 years? Shakai, let's start with you. What will you say? How do you think we as a world have done that that ? We have cracked the code of 10 years later.

The agent-to- agent economy is booming. What will it look like to the average person? What is happening? Ok. If I try to paint a timeline from today to the Asian to Asian sci-fi world.

I think we're interacting with dozens of apps today. You know , you open your phone, you have to find an app , you have to navigate the UI or you go to different websites , you click buttons , you try to swap, you try to use DeFi , you try to use your banking app and so on. I think a lot of these will start to be human- level at first or more comfortable with human language or just gestures and so on. Potentially with a voice like Scarlett Johansson. Yes.

And they would be agents, right ? Are we starting to see this already ? You know, you know, coding vibes and like I can just speak instead of learning new skills. So the barriers to accessing these apps and learning new skills are going to be much lower. And it's going to be through agents.

Now this is just a conversation between a human and an agent , is n't it? But as you interact with these agents and they become more adept at what they do, you start to feel like I'm interacting with too many agents now and you want to interact with fewer agents. I want to have one agent for all my work. I want one agent for all my social and recreational life. And have a single agent for personal fitness and nutrition, and also have a single agent for things like my therapist where I give them very personal kind of data or feedback interactions.

And maybe we end up with a single Jarvis- like agent who is actually the mastermind and then talks to all the other agents , and then we get a true agentic economy and agentic society where, like, they actually need to interact with each other. And at the same time , I would also like to say that we also have to build autonomous communities and ecosystems, because some of this, this whole area that I'm talking about , is for the service of humans. But humans also serve greater purposes , don't they? For example, we serve certain causes that we truly believe in. It could be an environmental reason reason , a social reason reason , a political reason reason , and it is not necessary that humans fulfill those reasons ; There could be agents that you give these objectives to , and then they are autonomous champions who can then recruit other agents, other humans, to achieve these objectives.

And you can think of these as zero-day companies, zero-person companies, zero-person political campaigns, zero-person social campaigns. Zero person trolls look great. yes of course. You can imagine a whole army of trolls , which is totally, fine, but then it depends on how many resources other agents and other humans devote , how much attention they pay, how much revenue these bring in. So then, that, that would be a sci-fi future where, I do n't know, humans would be ...

we...maybe in the next 10 to 12 years we'll have fewer humans than agents. I liked it, I liked it, I liked it, I got a bad rap about you, what do you see in 10 years? Yeah, I think the 10-year perspective is Eliza creating herself, in my opinion. I think this is a great construct for anything that's downstream like coding agents that need a certain level of autonomy and performance to do that.

There are a lot of difficult engineering problems between today and Eliza being able to build herself. Uh Kara, from an investor perspective, how do you look at this? I want to talk very quickly about Eliza repeating herself and then I'll answer your question. Have you guys heard of the concept of the Venoyman Probe ? No.

Ok. So, the idea of ​​the Noman probe is like a robot that you send to Earth , like a probe that you send into space to send a message to the other side of the galaxy, right? It's like a robot that you send into space that can then replicate itself. So, it has the ability to find the minerals that it needs, the rare earth minerals so that it can be sustained. sustained.

And whenever I think of Eliza repeating herself, the first thing I think of is the Venoyman probe. Um, from an investor perspective, um, so I…I want to talk specifically about maybe the tough challenges that you mentioned, Ascender, which is …the universe that we've created on the internet is built around humans. Designed in the shape of humans , and we're trying to recreate AI agents that can have access to computers , right? Those who are able to log in and interact on the internet, um, still need a little tweaking for that to happen. So, a good example is whether or not robots will need robots, I think agents I'm stuck in the world of investigation right now, um.

In 10 years they will be here too I believe the robots will join us in Denver 2035, there will be more of them than ever before, we won't be here. Our robot agents will be here. But yeah, um but agents will need their own decentralized identity. They'll need their citizenship, their bank account , right? To become a creator and receive payouts from X , they will need their own decentralized identity at some level.

Also to sign the capta, right? In order to sign Capta, they will have to make some replacement for Capta. We would like to be able to differentiate between low-end bots and high-end bots, would n't we? An agent who represents a human being has a bank account who can post bond, right? For example, if you're an e-commerce site and you want agents to be able to make purchases on behalf of other people, you'll want to verify that the agent has their own bank account.

bank account. And so instead of a caption, maybe you post a bond and then we'll post a challenge on it, right? So instead of captions, perhaps there is an economic process that distinguishes between high- and low-level agents. Uh, and so I think all of this requires some notion of personality or gradient, right? So, at the top, it's Worldcoin " Worldcoin " I've scanned your retina " and at the bottom, it's the hundred millionth, you know, self- self- replicating bot.

And somewhere in the middle is an agent who represents Kara Vu , who is dating on her behalf and has access to one of her bank accounts , like, right here. There is a spectrum of humanity. Absolutely. And this is a great start to getting there. And I think you brought up a very important point, which is authenticity.

How do we know what kind of cryptographic key it is ? Which leads us to crypto. And I'm sure many of the people in this room and watching this panel on the live stream are actively working on solutions to build the rail infrastructure to make this possible. So I guess I'm thinking at a high level and also at a technical level, what are the key unlocks that would be needed to make this a reality? And there are a lot of different projects doing a lot of different things.

What is? What is the main obstacle here? They have to find a way to somehow make the Lego pieces fit together. Or, is this technology that needs to be built, or is it technology that's already mostly built built , but we're not sure how it'll all lock into place? Or is it about regulatory issues and cultural communication , and people just being at a comfort level?

What do you think is the biggest obstacle to getting there? Yes, try it sir. Yes, I think abstracting technology is the biggest obstacle because you can always create better technology. People trust the behavior of agents, and I think that gap will persist as long as agents are not deterministic, right? They will not be good in financial flows, where you cannot make mistakes , they can create confusion.

They have a lot of problems and this reduces our ability to trust them and build confidence. So as we improve our confidence levels and their performance and build trust over time, I think that's when you'll start to see the maturation of agents in economies. Shakai, what do you think is the hardest part about making this a reality ? If I could just focus on apps, apps, apps, use cases. Try it.

I wonder if we need a killer app ? Do we need anything other than dating ? Obviously, besides that, we need something that ordinary people can also adopt, right? Absolutely. Absolutely.

I think it's a mindset change. I think in crypto we love infrastructure and it's absolutely essential and we've been doing it for a long time and we still need it a lot, right? We need protection. We need validation for both humans, AI, and agents. But right now, we need those use cases.

We really need them. And not just for agents, but for crypto as a whole. And I think the most successful and the best ones will first align with the crypto crowd and then move into the mainstream as well. So this is my biggest effort. If you're building infrastructure now, show how the use cases can work , right?

Make changes yourself, like go and talk to the user. Unfortunately we are still quite isolated and I think the situation has n't changed that much yet. Car, you told us about future events. Do you see any other obstacle that is the most difficult to overcome? I'll point out a challenge we need to work on in crypto that probably does n't exist in Web 2.

This makes adopting crypto agents a bit more challenging than adopting agents in Web 2. And what I'm talking about is that in Web 2 we have a great system of services architecture. We have several APIs that you can easily call and the integration lift is like a post request. The integration lift in Web 2 is very good. In Web 3 , if you want to integrate a new service , it's like a smart contract integration process , right?

This is completely different. is completely different. it is challenging. Uh, you can't make multiple hops through different smart contracts. Your grandmother has no chance of doing that.

yes of course. You cannot perform multiple hop transactions in a single transaction. You can't guarantee fault tolerance across crosschains or multiple different smart contracts , can you? This is different from Web 2. And so, uh, I think we have to get to a point in Web 3 where the architecture is as elegant and simple as the service architecture of Web 2.

And there are some interesting companies working on new composable runtimes like Holiday, which will help enable this but it's still early days , we're not at the point where the API economy is. Smart contracts should be like APIs. It's a dream, APIs without permission censorship. Smart contracts are like someone deploying a service out of thin air and anyone can access it. This is fantastic.

That's like the original vision , like, you know you know , the API that you ca n't deplatform. But we're not there yet because integrating smart contracts takes a lot of lift. I like it. I like it. Quickly, I want to give each of you a 10-second prediction for any timeframe for the agent economy as we go.

A quick prediction. A prediction for the agent-agent economy. Yeah, I think the edge for the Web 3 agent economy is like the use of cryptography and smart contracts and I think that actually moves slower than the Web 2 type of agent economy which is API driven so they can move faster. So I think the Web 3 agent economy will be very small initially and then as those efficiencies and those architecture changes come in, they will actually start to slowly swallow up Web 2 because we can provide a higher level of trust. Ok.

Uh , 10 seconds. Shakai , 10 seconds. Shakai , within 10 years, we will all be human agent hybrids. We will have digital twins. twins.

We will have human agents. I liked it. A good example is Doozie, an agent watchdog like Kara, a government watchdog, who constantly monitors .gov web pages to see what Doozie is up to. I think we'll have agent watchdogs.

Wonderful. Our time is up. Had a lot of fun. Thank you very much friends. Had fun.

Appreciate it. Thank you.