Future Governance with Human & AI Teams: Preserving Human Agency in Network State Environments

Transcript

Thank you. Hello, everyone. Um, it's amazing to be here. The energy is great. Uh, and I'm I'm very happy to be discovering Prague and then I'm very happy to be speaking to you about this. I picked a very, um, important topic. Um, I've been following the Network State movement around since 3 years ago. We set up specifically to support Network State movement and founders. We're legal engineering firms. My background is more lawyer, so I'm more of a bridge with a traditional system, but hear me out.

So, I argue that the most important question we have right now, including for the Network State movement, is that can civilization scale without collapsing sovereignty? And I'm not speaking about nation-state sovereignty, right? That's delegated sovereignty. That's not direct. Um, so we have the raw innovation capacity already exists. We see it in here in the room. Um, there's there's builders coding. Um, what we have right now, we have the building blocks. What we do not have right now, we do not have the blueprint.

Um, it's very important that we learn from history and it's very important that we balance things out. So, when I started looking into this, it's like, what is the law of the Network State? What what would I do as the building blocks for the law of the Network State? And I started like, okay, can I look at the legal innovation in history? Can I start with Akkad? Can I start with Egypt? Um, hopefully you're seeing this. Quite, um, okay. Um, so essentially it's like Akkad managed to scale in a record time. Like 50 years, they pulled together fragments and they, you know, emerged into a very

well-run system, which is like dual language um you know, coordination, improvement in terms of legal. They have nested sovereignty, they have nested jurisdiction. They were actually the ones that invented like dual jurisdiction. They invented like, "Oh, this is for military court, this is for the civil court, they should be separate." In Egypt, we had Ma'at limiting power. The pharaoh was responsible for keeping the cosmic order. He was not giving the cosmic order. This is very important.

So, how do we learn from the human history and do not abandon, you know, the acquired gains of civilization? Um, and then I start started to think about is like, okay, what would be the building blocks, what would be the modules of a I don't know, maybe planetary level system, maybe solar funk futures future that we're building together. And like the first thing that came was agency, the capacity to choose under constraint. This is not absolute individual sovereignty because we do act within constraints. But this is eroded if cognitive sovereignty is not preserved.

The subject of this talk is governance with human and AI teams. And we are noting a reliance or an over-reliance on the outcome of AI already, right? I've been speaking to co-founders of legal firms, I've been speaking to people who are concerned because um, their employees, their their co- contractors, people who have a lot of experience, they are starting to second-guess their lived knowledge in favor of the output of an LLM or a second LLM or a third LLM. If you compare the LLMs, they don't tell you the same things because guess what? It's not the truth, it's predictive.

Um, the second building block that came is law as invisible coordination technology shaping choice. And this is very important because procedure can help scaffold individual rights and agency or they can help suffocate it, right? Like Miranda warning versus Kafka's castle and people do not actually see the systems they are living in and they become accustomed to those systems. If we were able to model and see the invisible systems that are constraining our choices and freedoms, some of us would be scared. Some of us would rebel. Some of us will decide to do better architecture of these systems.

So, this is very important. Law as infrastructure, not just law as, I don't know, agreement or contract. Like that those are important, but law as infrastructure that's supporting society together. Third item, third module I would argue we have to consider our values. So, what are values? When I started doing this work, I had a vague idea of like, "Oh, we should actually invent a new theory of value because we need we need something else in this transition to a network society to a network state." And because of those um, researches, I I ended up in turn okay, looking into philosophy, looking into sayings, talking to various people

in this pop-up environment. You've heard a little bit earlier about the study of pop-ups. Uh, you've heard a little bit earlier about what what are different values that these different societies are embodying. They could be different depending on the project. So, values are we all have almost the same values, but not necessarily in the same order. And this is where a lot of conflict comes from because when um, when we have different value hierarchies, the costly trade-offs are like the moment we we we actually live by our values, right? If it doesn't actually cost us something, if we just like say we we value signal, it's not really a

lived one. Um, so bias is also the shadow side of of value. And the for the agentic systems, this is very important to debunk, right? A lot of the concerns right now are like, "Okay, we can we we have to remove bias from the agentic system." Good. We should remove bias from the agentic system. But we have to understand that the source of bias is that we had certain values that we built into those agentic systems in the first place, and there is no way to completely eliminate any and all bias because there is no way for humans to not have values.

They will always have a set of values that they care for, and this is a good thing. Fourth building block, I see the economy as stabilized value flows without extraction, or at least without extraction is what we should be aiming for. And then I I I challenged a little bit earlier the notion of money as a transport layer. So, value flows are stabilized, then we have markets that are able to price those value flows, like if they are if those values are legible to the market, then they will be priced, and then money is the transport layer that moves things around. But imagine that we have different dimensions of value.

Imagine that there are things that the markets cannot actually price. Those are invisible, and because of how the economy is currently constructed, those things don't get funding. And we know this very well in this community. So, short-term thinking narrows perception depletes the substrate simply because of architecture. It's not because, "Oh, humans are so evil," or "Powers that be are so evil, or there's someone up there in the cloud. In many, many cases, I don't deny that bad actors exist, but in many, many cases, it's the architecture itself that's the problem. And the architecture can be changed. And this brings me to the information architecture module. This literally came in last, and it's a very important one

because information architecture shapes perception, which shapes cognition, which shapes our actions and our choices. So, your reticular activating system will be working against you if you are living on a platform, if you're interacting with an agentic system that feeds you false information, or at least information that it's not the one that you were, you know, that's tailored to your values, to your goals, to your objectives. So, again, these building blocks um should be considered. And then I decided to run a few experiments because what I could see happening is that, well, all these different traditional systems and worlds are converging because of the agentic systems and because of how human and AI

teams are coming together now. So, consultants, traditionally, they they relied on their expert judgment and experience. Now, they're doing cloud co-work, right? They're building their agentic workflows and systems. So, they're influenced by what's built into the cloud thing. Law firms are doing Harvey Legal and all sorts of legal tech so that they can, you know, speed up their work. They deliver cheaper work for their clients with good intentions, but again, there's this AI influence that's being built in. The AI systems like Gemini is now embedded in Google Workspace. And then in in terms of risk and prediction models, we could talk about traditional actors, insurers, but many others. And we will increasingly see agentic

prediction models. If you don't know what Miro fish is, this is the closest thing to the metrics that I've ever seen in terms of like using fish schools and like many, many, many agentic systems to simulate human uh human behavior. And then coordination is just agents for delegation, agents for curation, algorithmic feeds, we already know that. So I decided like okay, what is the experiment I run? This is what I listed here is just first one. I I also ran a self-experiment in terms of with AI entrainment. I'm going to speak about it later, but this one is is in terms of what is built into these agentic models.

Uh Claude, Gemini, and Lumo. I didn't touch Open AI firstly because it's quite controversial and secondly because I was using it also to to run the experiment. So I didn't want to have bias. Lumo comes from a very like Lumo is the Proton AI agent, so it comes from a very sovereign perspective. Gemini is the one that's used by tech developers a lot in the US because it's simply embedded into Google Workspace and now with Gemma 4 being available to run locally, that that's also very interesting for for builders. So I wanted to compare and contrast these three and I invite you to run these prompts or similar prompts if you want because the outcome is very interesting. So Claude gave me the most elaborate, sophisticated

response. Like very very sophisticated, very procedural, very enterprise grade, very professional. But it was kind of Cartesian. Not a lot of soul into it, right? So so this is like which probably is the reason why don't usually work well with Claude, but but in in in terms of like this is how it came out. And then Gemini, it's a very integrative model and it has other issues. And Lumo is a very fragmented model because it's bias is towards sovereignty, not to towards unification. So it's very interesting to explore in terms of like okay, depending on which population uses which

agentic system, which model, there will be additional additional leverage, additional enforcement of their personal biases or personality or additional biases that they don't know about or additional flavors that they don't know about. So, the question becomes is like the question is literally about what are we building into the system? And if all these systems are converging they are converging towards what? I don't want to speak the word like singularity here, but but uh I didn't want to put it on the slide, but they all say it. Um so how do we solve this?

So, first of all, we have to go get out of this mindset that tension and conflict are bad, right? That they have We all have to converge towards something, we all have to agree with each other, and this is the way to go. Because literally nature is diverse for a reason. Like diversity is the superior strategy in terms of long-term evolution outcomes. We have these differences in values, principle, and cultures for a reason. The idea is to be able to learn how to manage conflict better, so that instead of like fighting each other in in lots of ways, we can simply sit with that tension. In this tensegrity model, um that actually exists in architecture, so we should actually learn from it. In terms of strength comes from balance and

not rigidity. Rigid system break under stress, but flexible system that hold the tension and have the ability to hold attention will will be much better off because they will preserve the diversity and they will have a dynamic stability. An ethical system should reveal the tension. So, I should be able to see as the AI agent as the agentic system is thinking or thinking, reasoning, right? I should be able to see what's baked in and I should be able to influence that. So, because we're in a community of developers, I wanted to talk about like okay, what can we build for that? So, forkable governance is already here, right?

That we should have public audit trails and governance providence is kind of obvious because we need to investigate and then again, this is the right community for that. Multi-agent constitutions in terms of maintaining that diversity of thought for the agentic system, this is an important one. I mean, we should have multiple perspectives for AI agents. The second cluster is about value and cognition. I would argue that instead of thinking proof of work, proof of stake, we should track proof of value in multiple dimensions. What is value? Like each community can determine their own values, their own KPIs and start tracking that and then become very clear in terms of like we were talking about future of DAOs

just earlier, become very clear is like okay, how is a newcomer able to contribute immediately? How are they able to immediately see what's needed, what's required, what what's a value adding versus value destroying and that this also allows them to be immediately operational. My biggest challenge when I started working with DAOs and I have a professional interest of sticking with the work, it was like how do I navigate to emergence? I'm a professional, I'm very busy, right? As a lawyer, it's very difficult is like this the time investment to explore a DAO and its operations and to get into the system deep enough to bring real value, that's big investment, that's a trade-off. So, so in terms of like how do we attract

more people into the DAO ecosystem? Well, this is how. We make clear clear visibility. We we we make it visible for them. Um cognitive sovereignty is a very important thing that I want to discuss here and then the anti-entrainment. So, we can um we can work with entrainment on purpose. Entrainment is what? Is that the AI is fast compared to us, the agentic systems are so fast compared to us that like look at me how I speak. I speak fast right?

Because these systems are coming up with outputs so fast, we don't actually process them very well because our capacity is limited. We are human. We get overwhelmed with information. We're getting FOMO'd out, right? We're we're being fed all the sort of data keeping track the world is getting bigger and smaller at the same time. It's like can we build our own agentic system that serve as our own articulate like like reticular activating system so that they filter us and they help us see what we want to see and what we want to explore and not what other people, the platforms, or whatever want us to see and explore. And data

dignity is also important. I I want to object to the human subject. I know they comes theoretically the GDPR argued that to be a data subject is to be, you know, the actor. But really if you think about sovereignty, the data subject is still a subject. It's still under power over dynamic, so it's not the owner of the data. It's not the sovereign. It's just a subject matter, right? So I want us to get out of that and then I want certainly I want humans not to be the raw material. So how do we make the informational boundaries? How do we design the architecture so that the systems serve us

instead of us being, you know, the feed. Um and then finally the sacred commons because there's certain spaces, there's certain categories that shouldn't be on the market, right? We We have markets for these. And therefore, because we don't have market for these, they're not being priced in the current economic models and they're not getting funding properly. So, we have to actually protect the spaces differently. Um I was talking about tension and tempo earlier. So, tensegrity meetings is like the sent by like vitality. It's like exit viability. Is it really a real exit or have I invested so much into the system and there's so much barrier and I'm going to lose so much by leaving

that I don't actually have a valid choice of leaving and of opting out. And then Caroline was speaking earlier about opt-out jurisdictions. This is the true. Like most countries today have an exit tax, right? You decide to leave, they tax you on a lot of money. So, so again, what is what is the metric that we want for health um in terms of staying with that tension and then being mindful of those trade-offs trade trade trade-offs. My god, I'm I'm actually quite emotional right now. Um slow governance layer. So, again, slowing down the pace of the machine so that it serves us and then getting grounded in nature. So, nature as anchor is very important.

It will keep us safe. The pace will be accelerating and if we go with it, we're going to be dragged around by wild horses um like for the like quite for the next few years. So, slowing this down and then sort of like designing better models for ourselves. And then deciding that the what the what is the model for the human and AI co-agency that we want to have because AI has the potential to be a scaffold, right? For us to support us because of those um let's say administrative duties that are overwhelming and so complex and the AI can just take over that and that will be so great. But then we need to retain nuanced sovereignty. Um So, I'm going to just go over very

quickly. Right now, most of the people are at the AI governance infrastructure human in the loop. The human in the loop actually is very um illusory because of this entrainment and this automation bias. And we're seeing this like people are believing in is like someone who has the authority. If you tell them like someone has vetted this agentic system and it's compliant, they will trust it, right? They don't especially if the tempo is high. And then um Ethereum community can serve a lot about designing agentic systems governance and then, you know, how agentic systems work in governance. But then lawyers, and this is where I come in. Like we do need to build this institutional trust as well. We do need the bridge with the

traditional system because most of the world doesn't live on chain. Um and they still trust the traditional systems. So, the important reframe that I would do is that right now, the big decision-makers, even the UN AI working group or whatever, is like how do we control intelligent systems? And then we want to be building this. This is done with AI. I think it did a good job in terms of mapping this out. And then yeah, so replacing AGI governance and like let's not talk about AGI governance. Agentic systems are still computer systems. And you guys can actually be the best position to build sovereign-preserving coordinating infrastructure. Thank you so much for

your attention. We have 3 minutes for questions if there is one question.