Alignment Is Intelligence: AI Digital Twins & the Future of DAOs

Transcript

From coordination to alignment

Welcome everyone. I'll try to be quick here. I know this is the last session of the day. Again, my name is James Young. Uh, and today I changed the title of my slide from coordination to alignment. So, alignment is intelligence. and we are at uh an Ethereum based event and just want to give acknowledgement out that Ethereum has solved this hard trust problem and I'm going to get into it a little bit here and this is not a small achievement. It's one of the most important coordination breakthroughs uh that I think we have experienced in our time. Uh I'm going to talk about communities and Dows and for those uh who who who in the audience knows what a

DAO is. Okay. And who's has experience and who's been in a DAO before? All right. So what I'm about to say is probably not surprising. Um Dows have taught us a lot of things that may feel uncomfortable. And uh this is what I'm going to be talking about. Uh what I've learned over the last few years uh I'll get into my bio a little bit more in detail in the next slide. Uh is that coordination is necessary but it's not sufficient. And what I'm going to be addressing is a potential missing layer. Um, and why I

Ethereum as a coordination breakthrough

AI makes it uh impossible to ignore uh because I think we're coming into an era where communities are not just decentralized. They will use AI to scale culture. I'm going to go to the next slide. Um, it's not going to the next slide. While I try and figure this out, I just want to uh give a shout out to Anelie. uh she is actually my co-founder and my wife. Today is Valentine's Day, so I felt that it was a good thing to to say. Um so uh I'm just privileged to have been able to be uh with her for over 30 years now.

Why DAOs feel uncomfortable

And so I just want to say happy Valentine's Day to Oh, there we go. Okay. So, uh why me? uh as mentioned um briefly before is I helped co-write the Mollik Dow white paper. Uh I was fascinated uh with Ethereum when I saw the Dow. I'm not a trader. Uh I don't have an economics background. Uh I come from a game design background. Uh, and just the idea of coordination and incentive alignment was hugely fascinating to me. When I saw the DAO, I felt like it was a video game that kind of peaked out into the real

James’ background (MolochDAO, Collab.Land)

world. And, uh, I've been fascinated ever since, even more so now than ever. Uh, I've been working on with Unelie on Collaband now coming up on six years. So, we've seen over 60,000 tokenized communities flourish and some of them a lot of them stagger or decay and I'm going to address that and talk about that today. And then also I've been working on a a side project uh called Beacon for about a year knowing that uh coordination for AI agents uh was going to be important and while in the middle of working on Beacon I've been keeping up with other

EIPs like 8004 uh and I'll talk about that in a minute. So, we're going to talk about Dows and um why Dows keep failing. Uh here's a list of reasons. Um and everyone has seen this. Uh a Dow launches, it gains momentum, and it slowly decays, some faster than others. And I don't think it's necessarily a governance problem. I don't think governance is impossible. I think it can scale. Um the issue is is that uh communities and how they grow are very

Why DAOs decay over time

different than onchain market metrics. Uh markets coordinate assets or money. Uh but communities really coordinate meaning and meaning is a lot harder. Part of the reason why I think we have not done a great job with Dows, who's heard of good hearts law? So when you when a measure becomes a target, so a KPI, a goal, when when that measure of whether you're reaching that KPI or goal becomes a target, it ceases to be a good measure. So, we're talking about voting interactions, we're talking about tokens, token price, we're talking

Markets coordinate assets, communities coordinate meaning

about TVL, and we're talking about just onchain engagement. Um, and it is kind of painful because where do we go from here? Uh I think in the early versions and up to this point in time uh Dows have optimized what they could measure. Uh but the moment you optimize it, it ceases to be a good measure. And I think that a lot of Dows have goodhearted themselves. Um not because incentives don't work. It's because incentives were pointed at the wrong metrics.

Goodhart’s Law & broken DAO metrics

And this is where I'm going to talk about this issue of trust. And Ethereum is also marketed as trustwear. And I think Ethereum has solved hard trust, cryptographic verification, immutable identity, economic security. But uh there is a second trust problem that it's very difficult I think for blockchains to to solve. um it's about behavioral alignment when you get a group of people together and this is again the key insight is that I think coordination is necessary so you have to have a shared asset you have to have

this meaning where people can build up a treasury but it's not sufficient it's the first step the insight here is that tokenizing a community is really just a bootstrap event And we need something that can keep alignment over time. And this is why you see if you're in a DAO, huge fights, whale attacks, huge proposals, spending hours and hours reading proposals or side chatting with others. Um, when proposals are about to pass.

Hard trust vs behavioral alignment

Um, and so this is where I think AI and crypto can really intersect. And one possible concept, and I'm just going to throw this out there today, is being able to have a digital twin. So this implementation is what we're calling like a digital twin agent. Uh, and uh, it's where members of a DAO all have a digital twin, but the agents are the only ones that are allowed in a DAO.

Tokenization as a bootstrap event

So no humans in the DAO. And the only goal for a member that is tied to that agent is to prove that the AI will make the same choices you would make over time. Uh we have the standards and the technologies and really the market time right now because I would argue that prediction markets are very popular at the moment. So there is this Overton window that's shifting and it's measurable, trackable and scalable. We have EIPs like 8004. We have account abstraction through 4337.

We have delegations. I highly recommend anyone in the audience that is looking toward agents to look at the MetaMask delegation toolkit which allows smart accounts to delegate uh some authority to their accounts. uh and there is no there's a minimization I believe and an exploration to figure out uh this type of shift because you're not then relying on these proxy metrics that we have had in previous versions of Dows but we're using repeated revealed preference over time

Introducing digital twin agents

and uh in a in AI parlament this is called coherence uh and so I'm going to describe how it works at least one example. Uh again, you can think about this as a user had playing a prediction market game with their digital twin agent. So what happens is an AI asks a user multiplechoice questions, but before delivering uh the questions to the user, the AI also guesses. You could use zk proof so that uh it's not uh leaked or revealed or it's held private. The AI will guess what the user would choose ahead of the user choosing. And after the user has

Coherence scoring & prediction market model

submitted the guess, the AI's guess is also revealed. If it guesses correctly, its coherence scores goes up. If it guesses incorrectly, its coherence score goes down. And I have an example the video of this. We I worked with a non-developer on this vibe coded this. You can see that there are different types of games you can play here. You're doing a solo practice. The AI is generating random scenarios for you to choose from. It's trying to guess what you would answer. So, it's coming up with a detailed personalized scenario and it's going to give you the multiple choice answers. You answer and you

submit. And here the user has surprised the AI. And you see the user's choice and you see the AI's predicted choice. You see the AI reasoning and then you see this 2%. This is what we're calling this coherence score. And so the idea is once you reach a threshold with your agent and you have a high enough coherent score, the agent can then begin doing things on your behalf. so how does this scale if I am just playing this prediction market game or

Delegation via AI agents

the AI is playing this prediction market game with me? How does this scale to other people and uh to Dows? Well, when you want when we're talking about scaling coherence and I mentioned that if you get a past a threshold in terms of a coherence score with with your agent, your agent can start automating tasks on your behalf. So, you can have this coherence between the human and the robot. uh but then once you reach a certain coherent score and other AI agents also reach a certain coherent score these agents can now delegate to one another.

Scaling coherence across DAOs

Now you have a semiautonomous DAO that's run that is running 24/7. Once you have a DAO that is coherent with itself, it can then organize with other Dows that are also coherent. And what you get is this network effect, this network effect of culture. And I think this is the missing piece that we are seeing in Dallas that I hope people and I I I want to get this idea out there so that others can experiment. I know that with AI, everyone's looking at like trading bots and other efficiencies

and this short-term gaining of financial assets. Um, but I think that we can use AI within the organization, create new types of communities and allow us to learn from previous versions of Dows and Dowo failings to really incorporate AI into the organization because strong if you can build a strong culture that becomes this magnet, this attractor pulling decisions forward through verifiable um actions. and patterns.

Culture as network effect

So, uh, it's time to put the AI the A back in Dows. Um, and really here what we're talking about is the human aspect of organization with AI. So that we can see the network effects of trust and belief, which I think is just called culture. And the more agents share a coherent co culture, the more valuable the culture becomes. Um and so yeah, I think that uh I want to implore everyone here to re-imag what AI can do uh not just to

“Put the A back in DAOs” & closing

you know increase your earnings but also help with organization and scaling organizations and communities. Thank you.