Luke: [00:00:00] You’re listening to a new episode of The Brave Technologist, where we explore how emerging technologies like AI, machine learning, and Web3 are influencing our lives. In a couple weeks, our team heads to London for the AI Summit, where we’ll interview founders, builders, and policymakers leading in this space.

We look forward to releasing these episodes to you later in June. And for this week, we’re going to meet Balazs Nemothy, who is a Hungarian technologist and entrepreneur working at the intersection of internet infrastructure, identity, and AI. He is the founder of the Agent Community, a community-backed effort to establish .agent as a new top-level domain and identity layer for AI agents.

Previously, he helped scale the Decentralized Identity Foundation to more than 300 organizations and is the inventor of a granted US patent for programmable assets. In this episode, we discussed why agents need identities, including personalized names and domains, how every .agent domain is meant to host a real working agent, what challenges will emerge around trust, safety, and accountability [00:01:00] as agents scale, and what community governance could change by preventing single companies from controls over agent identity.

And now for this week’s episode of The Brave Technologist.

Balazs, welcome to The Brave Technologist. How are you doing today?

Balazs: Thank you, Luke. I’m very excited to be here.

Luke: I’m really looking forward to this one. Obviously, agents and agentic AI and i- is kind of a buzzword and a lot of hype around it. And, a lot of people are hearing AI agents, and they think of things like productivity tools or chatbots.

You’re thinking much bigger than that, from what we’re gathering. How do you define an agent and what is the future you’re actually trying to build?

Balazs: It’s super interesting because I think an agent is kinda smaller than that. Like f- from my standing, an agent is something that can use tools, like an AI, like generative AI or, or a [00:02:00] version of an AI that has the capability to use tools, and then becomes an agent because now it has agency to decide to use a tool or not to decide to use a tool, or use that tool to build other tools to extend its repertoire.

So kind of starting from there, which is like an entry level. But of course, chat uses search. Is that a tool use? Arguable. Technically, what I just said, yes. I think like coding agents are like a very clear picture that like that’s an agent. But then robots are agents at the end of the day because there is an AI running somewhere that is using a tool that moves the robot around.

So like it comes to the physical space, and then everything in between and where it’s going, super intelligence, general intelligence, we can’t really imagine how that looks. Once we hit the exponential growth and the curve starts to [00:03:00] go up beyond human understanding, like we don’t know what’s gonna happen exactly.

We are not there yet. I’m not a doomerist per se, but like, it’s really hard to imagine if it’s gonna be with us or not fully with us.

Luke: Fair. Fair. It doesn’t take a ton of stretch to, to see how it could go, sideways pretty quickly. And to your point too, even like the browser is, technically known as a user agent.

That’s one of the things that’s so exciting to me about this part of the phase that we’re in now is that, it’s like you’re giving a lot more power to these agents, but like integrating them in things like user agents and all of that. You hear a lot around, concerns about, you have these agents, they might be representing, things that you’re doing or, your work or you in some form or fashion or your information, right? Why is there currently an identity layer, for these agents? I know that’s what you guys are focusing [00:04:00] on, right?

Balazs: So there is none, and that’s what makes very, makes it a very interesting problem set because like the internet of today, of course there is like the server internet that’s like the majority of the volume that is like API to API type communication, and then there is the human internet that has an ever broken human identity system-

because the internet was not designed with an identity in mind. So like it has been patched and patched and patched. And we have some solutions, but every industry has their own solution and it’s just like an extremely big patchwork of- Of somewhat broken or very broken systems, but kind of working.

But interestingly, like, agents have kind of came around this capability so quickly that there isn’t really an established standard or like a winning player in this [00:05:00] space. So it’s, we could call it a greenfield opportunity in a way.

** :** Mm-hmm.

Balazs: Because it doesn’t need a lot of the human infrastructure.

As kind of developers now, again, using CLI towards term-terminals to run the agents, because it doesn’t need UI to do things, as it’s kind of a first citizen on the internet.

** :** Mm-hmm.

Balazs: Or it’s a digital persona that can sign everything. Unlike a human, we cannot sign digitally.

We need some kind of connection into the digital world to be able to do that. Which means that, like, a lot of, the authentication protocols, the complexities, a well-established rotatable key can prove that the agent is what the agent is. Of course, the questions comes up because they might represent us of, what this agent is allowed to do- Mm.

-on behalf of the company or in this use case or whatnot. [00:06:00] There is an authorization, and that’s, there’s a lot of companies, looking at this of, sure, there is an agent, but, the intent protocols, like you might have heard. So there is a lot of work happening in this space, but there is none one that has won.

Naturally, communication is the lowest level of all this. Of, like, where they actually start the conversation.

** :** Mm-hmm.

Balazs: Or interaction. Doesn’t have to be a conversation. So it’s going to be evolving. The reason is because today’s agents are not the agents that we really need identities.

Luke: Right.

Balazs: So just historically, OpenClou happened. I assume the listeners know OpenClou. It was, like, a open source project that got global attention around February, about an agent type that is designed for never be shut down, so a non-ephemeral agent that’s always available. Mm-hmm. People who work in the technology have [00:07:00] known about these and kind of it was like, this is the next era.

But OpenClone made it globally understood of what is a long-running agent could look like. Whether it was the right implementation, y- many of us can argue, and we have different opinions, but doesn’t matter. The important is that it’s creative, like a shift from “Oh, it’s just AI,” to like, “Oh, it’s an agent.”

Right. And so the term agent for most people who like not deeply involved with tech is something that runs long term. So it’s not the coding agent that you spin up, you solve a problem, then you clean your work tree and move on. It’s something that could manage my social media account or could constantly handle my finances, so I don’t need to do it because I really don’t like it, but, I don’t really want to set it up.

I want it to set it up it- [00:08:00] itself for me. And that doesn’t really exist yet, right?

Luke: Kind of more of a Jarvis kind of or, or like- Yeah … an assistant kind of a approach. Yeah,

Balazs: exactly. That’s what my non-tech friends are kind of asking for, like when that’s happening.

Mm-hmm. , And it’s happening in certain ways, but it is hard, and the AI models are still hallucinating and so, like, a lot of the things, we are not there yet. But we will get there. There is no question, like everyone is pushing to that direction, and then there will be agents that really need identities that serves, that represent people or legal entities that operate on the internet as becoming economic actors, and kind of filling in roles and functions in society, in a way.

** :** Mm-hmm.

Luke: What you’re focusing on, right, is handling that from the domain level. Can we go into that a little bit, how you’re approaching this?

Balazs: Yeah. I’m leading the agent [00:09:00] community- and this is now a pretty global effort with the initial goal to secure the .agent top-level domain in a community application.

To effectively make agent builders the governors of the .agent top-level domain. So it’s governed by the builders and not owned by a corporation. That’s kind of the base-

** :** That seems

Balazs: important. Yeah, I get the question, “But, but it’s just a domain.” And, and that is true, .dev is just a domain, .ai is just a domain.

.Agent will be also, in a way, just a domain. In itself, that is not special. But that’s when, the community application and what else it could do comes in. So we are thinking that agents will be a lot more long-living [00:10:00] and potentially start to build personas. So our tagline, the public tagline is, “Agents need names.”

Luke: The Brave Technologist is brought to you by the Brave Search API. Access billions of indexed web results from a simple API call with the Brave Search API. Join the leading names in AI and tech using the Brave Search API to power agenic search, keep LLMs current with real-time data, train foundational models, and bring the best of the web directly to the leading edge.

Get started today at brave.com/api.

People might say it’s just a domain, right? But I think that people sometimes don’t necessarily appreciate how powerful domains are. If you think about it, the things that are, that were, a critical part of the internet that have been there a constant, like a domain’s there.

Your email goes through a top-level domain. It is a really important piece of how we all interoperate, online. That’s one of the [00:11:00] things that stood out nicely about this is , we don’t necessarily have to overthink it with this stuff sometimes. You can use something that’s well-established, like domains, for this purpose and, have something that people can kind of relate to.

That’s at least what jumps out to me. I don’t know if that’s been some of the thinking around it from where you guys are coming from. But, it, what I find with this stuff is that it’s really hard with new technology to bridge the gap with, your neighbor, and people with trying to understand these things where the co- the agent’s complex enough, right?

Like they’re doing all these things, and there’s all these risks and stuff. Wherever you can connect that with something that’s well-known seems like a smart move, especially with something with identity where people are gonna have to understand what that means, right?

Balazs: And find them. Like one of the aspects of all this is just because something was built might not get usage.

In this case, like domain names Sure, they are human per se. It made for humans. But all these AI models are trained [00:12:00] on human language, at least for now. So human discovery is still in their DNA. Well, code, vectors. Just to be precise. And discovery and relations and, like a domain of search.agent is pretty good branding of what you are doing.

But Bob the maid, who might be my domain at some point, is my kind of like home manager Bob agent, I just called him Bob, and it was available, will be available for me or whoever else I want to rent it or, or give it services. But it can just be like easy to reach and just as you said, like an email, like I just visit the site, I log in or interact with it in a way, and then I can talk to it in any platform.

It doesn’t have to be a custom application built for the purpose or logged in. Mm-hmm. It’s my agent [00:13:00] is served there, let’s put it this way, or operate in that environment. And of course products and the usual domain things that you can think of. But what’s the specialty of how we are thinking is to go beyond the, just the domain to make it like a special domain.

Mm-hmm. Because, it is special because it’s a very good English word, and technically the AI industry does not have a domain name. So .ai- Mm-hmm … is a country code of a small Polynesian island. And it’s now like sixteen percent of their GDP. Oh, wow. Wow. Oh

Luke: my gosh.

Balazs: Wild. It’s wild. But it’s a country code, and it’s just like readily available if you have a bank card.

What we are thinking with .agent is let’s make this the AI native domain space. So let’s figure out an enforcement mechanism that you have to have an agent [00:14:00] on a .agent domain name. So it can’t be just your personal blog hosted there. You have, you can have your personal blog, but there has to be an agent served on that website.

I see. So basically the idea is that any .agent domain you will ever visit will have an agent on the other end, and it’s kind of builds that expectations like EDU is for education. When you visit an EDU site, you know it’s gonna be a university. When you will visit a .agent website, you will know that there is an agent

Luke: It’s super interesting.

Maybe it’s something you all have been thinking about more or maybe it’s early. I don’t know, but, you hear a lot about, different types of delegated agents, like giving certain agents, responsibilities or authority to do certain things and not others.

Is there a role in this domain space, to have, subdomains for delegated agent or [00:15:00] agents that are more specialized that could work in this scheme? Is it still too early or I’m just kind of curious if people are thinking about it that way too.

Balazs: Yes, but differently. So we are thinking about having special subdomains set up and operated partially by us, partially by partners, of course.

These are not set, but, examples are more like, let’s set up, an on-ramp developer domain. When you host a Vercel site or you host a Cloudflare site, you can get these, temporary placeholder domains, that, that is live, but it’s a subdomain of Vercel or Cloudflare.

Something similar to that, like .is.agent, for example.

Luke: Mm-hmm.

Balazs: And it’s like a developer version. It might not have all the requirements that, proper .agent domains, may over time have, and it can be used maybe temporarily or as an on-ramp and [00:16:00] distributed by agent-building frameworks.

Or let’s say maybe the EU will have some agent re- like trust requirements and new regulations in the future, and so maybe the .eu.agent can become like a subdomain category where all hosted agents on those domains will meet EU standards or will have to meet EU requirements, whatever those become. I see.

So we are thinking on, these levels. The reason why, because kind of like just beyond having an agent, the other question that interests a lot of people who are more deeply involved is the question of trust.

Luke: Right.

Balazs: Where identity is a building block, like more on the, in this case, more on the technical side, not like personal identity.

But what an agent [00:17:00] can do, or more importantly, what if an agent does something it shouldn’t have done?

** :** Mm-hmm.

Balazs: And what happens then? there is a lot of phil-philosophical questions, there are legal questions there, there are economic questions. And I would say, a part of the community will be dealing with these questions because to a degree, the ideal long mid-midterm scenario for .agent is that it becomes also a trust symbol.

Mm-hmm. So it’s not just a agent running on your .agent, but a, to a degree, trusted agent running on that, .agent. Which is of course, restricts how well-distributed .agent gets, but it also increases the value of a .agent domain because it will… It is already extremely easy to launch an agent. It’s going to get unbelievably- Right

easy. So there [00:18:00] will be, there are probably billions of agents today because I don’t know how many CloudCode and Codex and O-OpenClou and whatever agents are running right now, but many of them ephemeral, they just go away. But it will be possible. You have a VPS, you can have seven agents parallel or 50 agent on that, tiny VPS because it’s just some memory, some CPU, and the AI is hosted through an external service anyway.

And so there will be many, many, many. And the question becomes that which one can I let in? Which one should I interact with? Which one should I block? And sure, there are, like, the ultra-centralized approaches, but we are thinking on, like, we– if, of course, we get this TLD and we become the registry, then we could create, very interesting kind of like trust signaling-

** :** Mm-hmm

Balazs: working together with [00:19:00] this big community. So not setting it as like a corporate policy that our CEO thinks trust is this, so we wrote it down- Right … we hard-coded it, and up until our shareholders agree with it, we will keep it, but once they don’t, we will shift. It will be that, it went through a discussion, whoever had the time or the ability contributed, and this is what all that group of people in the community believe should be.

Can change it. It will change. That’s the other thing that as we are, like, growing and building, the interesting thing is, we know that in two years, the topics, the discussions will be completely different than what we have now.

** :** Right

Balazs: Because these things have evolved so much, and the scenarios and then reality is very different, and some topics will be old [00:20:00] news.

It doesn’t worth to have effort to talk about because the new stuff is happening, and, it’s gonna be very– it’s gonna be a constant intense work.

Luke: A couple things from that. And to just help put it into perspective for people too, Brave’s, our search index, like our Brave search API, was set as the default for web results on OpenClaw.

And just from February till, I think April, we saw over seven hundred and twenty-two thousand accounts generated, using that. It’s huge, and that’s only a small amount of people using it for those things. These numbers are nuts. You’re gonna need to know… everyday folks are just gonna need to know if an agent is who it says it is or, that trust level you talk about, right?

And the other side of it too, and I think that this is, an underappreciated thing that, touches on some of the versatility of using a domain for this, is that, the approach that you were talking about earlier where it might be a .eu or an EU variant, right? Or a regional variant.

We’re all [00:21:00] working in a global network space here, and AI has become a national security issue for a lot of governments and things with jurisdictional parameters, right? I really like the idea because These are patterns that people can become, have become familiar with, right?

And if we want something to scale and, there are concerns about things getting off the rails, and you need trust to scale in a certain way, and especially with all of the added attention of it being at that national security level, and legal level, there’s already…

legislations are already getting passed in regions around this stuff. And, having that familiarity, that anchor around it or with it from the domain level makes a lot of sense, like at least practically from what I’m taking in on this. Let’s dive in a little bit on this community-owned element versus, like a corporate-owned side, ‘cause that’s another thing that’s really nice about what you all are doing here.

You were going there toward the end of what you were saying- a minute ago, where you don’t [00:22:00] want, a CEO to basically set this because, that might then not only include the ethos of that company and the shareholders, whatever, but also of the jurisdiction that they’re in, right?

How does this work with the community-governed thing? ‘Cause I think a lot of people might not necessarily be aware of, like, how that governance model works. So everybody’s involved, it’s not owned by a corporation. Like, how do things get done in that kind of environment?

Balazs: So it is shifting naturally.

This effort is not extremely old. Just to set the date, it started 2025, January- Mm-hmm … as a very small effort. I had the idea and start building and then, it has grown. Effectively, would tell why it started before we would go into the actual communities. Yeah, yeah.

Please. So- let’s go into it … I personally have been working on identity for over a decade, and digital identity in different hats. At some point, I was a director of [00:23:00] the Decentralized Identity Foundation. That’s a standards organization under Linux. So I have dirtied my hand with the topic, many times.

And

The year before the project started, I was helping friends, friend companies, and like mostly just building with AI, whatever was possible and I came up with, and one of the projects was some form of a workflow that I wanted to call an agent. And so I wanted to get a .agent domain, and it was not available.

So my next idea after a quick Google or ChatGPT search was that instead of getting the domain name, let’s get the TLD, and let’s- … figure out how to get the TLD because it must be possible. It cannot not exist.

** :** Mm-hmm.

Balazs: , Then through my network, I talked with some people who were involved with ICANN applications.

I started digging into the [00:24:00] ICANN processes and bylaws. It’s extremely bureaucratic. If someone loves bureaucracy, then ICANN it is.

Luke: Can we, just break down what ICANN is, So ICANN is the governing- … understand what that means?

Balazs: body of DNS, and therefore domain names Mm-hmm So, so the standardization and DNS and the operational management of all the TLDs, .com, .dev, .org, everything. Awesome … they are the ones that give the right to become a registry. The registry is the w- So every domain has one single registry. So .com has one company that is the seller.

Then there is the registrars. There are the GoDaddys, the Namecheaps, the hover.com that are wholesalers effectively. Sorry, yeah, so like basically they buy at wholesale and then B2C sell the domains. So we are applying to become a registry. And effectively when I started this, like my [00:25:00] second thought that was like this is great, but if I thought about this with no domain background, then someone who actually been working in TLD operations will think about this given the degree of hype that…

Even though back then it wasn’t this hyped, but it was coming and of course by now it’s everywhere. Others will think about it, and some of those players might have trillion dollar valuation and infinite marketing budget.

** :** Mm-hmm. Mm-hmm.

Balazs: And the problem there is because how the ICANN process this year is going to operate, but technically like the last application was also 14 years ago, was the last application for domains.

So when I said bureaucratic, I meant bureaucratic.

Luke: 14 years, that’s a long time. So every

Balazs: domain, XYZ, .horse, .dev was applied in 2012. It might have not launched until 2017 or until recently, but [00:26:00] it, the application was submitted in 2012.

And because the process normally looks like you submit the applications if there is a, the open window, and if there is more than one normal application arrives at ICANN to the same string and both are clarifying, meeting the requirements, then it goes to a bidding process, and the highest bidder win, and the bidding money goes to ICANN.

Mm-hmm. Good for ICANN. Now, .agent is a strategically important domain for many reasons. The chances of multiple companies will apply for it is very, very, very high. Applications are usually kept in extreme stealth for the reason so that you don’t give ideas to your competitors to apply as well.

** :** Mm-hmm. Or,

Balazs: or your industry so we don’t exactly know. We have some urban legends, of who might be [00:27:00] applying. But effectively, that route was n- was not realistic because the numbers some of the potential applicants could spend on marketing, not as a business, not as like we will distribute it and we will earn rewards.

Purely as the… Because there is one single company at the center, the company set the policies, and they can withhold as many domains never to be sold to anyone, as many they want. So they can own search.agent, travel.agent, name your favorite .agent domains, and they can just withhold it internally and sell the Hello Kitty forty-four and whatever else to the public, or don’t, because it’s all in their purview.

And so however, there is a community priority evaluation process, and that’s why it is Agent Community. So the initial idea was because we discovered this bylaw within [00:28:00] the extremely bureaucratic application system of ICANN, that if there is a community application that clears the requirements, which is a preset, pointing system set, set forth by ICANN, if the community meets the requirements and collects twelve of the sixteen points and meet all the technical and whatnot requirements, the domain goes to the community before it would go to public bidding.

So it’s a separate flow in the application process that kind of comes before the bidding would be starting. And if it’s successful, then the domain, the domain goes to the community. And so effectively that’s what we are aiming for, to do a CP or Community Priority, uh, Evaluation process. And that is the initial start for doing it as a community.

We can’t raise money. Let’s do it as a community. [00:29:00] And from this, naturally then we can do it publicly. Great. So we can reach out to companies. We don’t have to hide our application. We can develop a degree of branding around it and get invited to the Brave, podcast and have other opportunities.

Plus, we can build, an extremely interesting advisory board that kind of like signals ICANN the seriousness of the effort. And so like we can do all these things and not talking with PE firms about how many millions of dollars, we will return to them over time by, operating this domain name to e- to everyone and anyone.

So, like, let’s then, then really lean in, because this is what we can do, and I’m also interested in it. And yeah, it’s, it has been mostly growing since then. And back to the actual question. On how the governance [00:30:00] works. So it’s all evolving. Effectively it is evolving right now, so, like, we are laying down a, a couple of the mechanisms.

For the time being, the board is more the, a little bit the decision-maker. Also, there is not much decisions to be made because we need to apply, for the… that is the biggest step for the time being, the application, because the application window is open right now from April 29 until the 12th of August.

So that’s the biggest step. So we are making sure it works. There are gonna be, there are meetings, and effectively, members can join. Fast-forward, we will keep hosting the meetings, and the topics will evolve and we can call these working groups or discussion groups, currently more on the discussion group category.

And the idea is that the registry right achievement will make a shift. [00:31:00] If we become a registry, then effectively there will be agent community members and agent community voter members. So everyone who has a .agent domain will have voting rights.

** :** Mm-hmm.

Balazs: Anyone becoming a member is possible, but becoming a voting member, you will have to have an agent domain, and that’s gonna be basically, the requirement for participating in decisions.

Luke: It seems like it’s what’s worked, right? Traditionally, with, governance and almost like a kind of something like a DAO almost model where, if someone or not in a DAO in the sense of, like, holding a certain amount of things, but just, having, being a participant in the system by having a domain or whatever.

But it seems like near term, your goals are to get as many companies involved with this effort as part of ‘cause the community is essentially companies at this stage, right? That are getting those domains

Balazs: Companies are preferred because ICANN [00:32:00] application. Oh, I

Luke: see.

Balazs: But, we have 6,000 companies which technically Makes us the biggest agentic community effort with the highest number of companies involved- Wow … in the world. That’s awesome … and all together we have 23,000, 24,000 members. So it’s like three to one individuals to companies.

Yes, we are focusing more on companies because overall, using the logo of a big company will matter more at the evaluating table than a, a bigger number, unfortunately. But we want to have every builder’s voice heard. So like- Awesome … my ask to everyone, please join. Joining is free. You can pre-register your dream.agent domain.

I do need to say it’s non-binding for now. We will do our best to, offer these domains to the applicants. [00:33:00] But by joining you can also join the effort, like the other efforts, the working groups, and effectively help us to make sure that it’s not set, the policies are not set by a shareholder value-generating company.

Luke: That, that’s a really key point, right? And I think that this is one of those rare things that happens where it’s with open source development too and other things like that, where, having the diversity of, opinions, of voters, of people at the table, is super important, because this is such a critical thing, in the space right now and people are wondering, what can I do, what can I do, to , help this not get, a-as to much of a nightmare scenario?

This is one of those things you can do, where, these efforts are the things that people are trying to do to make this, to get this set on the right track, right? Like, where it’s not, one company, handling governance on all of, agent domains. It’s, this is one of those things, go get involved.

, E-especially if you’re a developer too and, or com- Brave’s got our we’re [00:34:00] participating too. I think it’s really, it’s key. This is one of those important things. I appreciate that you guys are doing this and what a cool thing to start out of nowhere, basically, like your own interest in getting this to be the biggest thing out there right now.

It’s awesome. Just

Balazs: another, like, kind of like mental model, is there are a number of companies that are approaching the question that we should verify humans so that they can use the internet. And what if we just verify agents so that they can use the internet? There you go There, there will be more agents than what we can verify.

But, what if we let in the agents we trust and not close out non-humans? Right. Like it’s just a mental model shift on, approaching the problem. I would say potentially both will be happening or both are happening for different reasons, but, which will also pose the question of, taxation and [00:35:00] all these topics will come up, ‘cause these are going to be global problems.

Mm-hmm. And, and it will become very complex because some of the companies are operated under one jurisdiction as a global enterprise, and there are agents or AI that you can run locally, to do agentic work or eco- economic, uh, participation on your behalf. And trust is part of that, and registering or, like tax ID.

I’m pushing it, but, these are the questions that will start to be coming up more and more and more of, do you want your agent to pay taxes because it does your work? Hmm. Then you have, then you have to bring it to the Department of Machine Verification. Right, right. And, like this is funny, but, this is the topic- Oh, no, no, I know

that will, be coming up.

Luke: It’s inevitable. It’s just the cost of computer factor. There’s all these things, right? These are complex, real-world [00:36:00] costs too that are being thrown into the mix of everything right now., There are very few things I’ve ever seen in tech that have had such a strong demand from the top down of every Fortune 2000 company C-team is telling their staff, “You need to put AI in.”

They’re cutting jobs and blaming this This is like a generational type of shift we’re in the middle of right now, and these things, I laugh, but it’s a reality. These things are gonna come up, right? And just to kind of round it out, like from an optimistic sense, if you guys get this right, and if

agents are able to evolve in an open, kind of a user-owned, ecosystem or community, what becomes possible there that, that isn’t necessarily possible today?

Balazs: I will break this down. User-owned, it might be our ChatGPT agent running on, Luke M dot agent domain, right? Mm-hmm. Like, like, it’s, it doesn’t mean you are, like, self-hosting and you are, like, a rickety super developer.

That’s the start. What [00:37:00] it will enable is that a .agent does not belong to a corporate. That’s kind of like the base. But more importantly, like we have a build their own internet property that is more and more becoming a trust symbol, hopefully, that helps people and companies and other agents to know that like interacting with these parties or trusted, like I can like work for them, HTTPS, concept that like it is infrastructure largely.

But of course it has other implications, potentially. I’m being a bit vague because these are topics that we are discussing, and some of them might never be fruition. Some of them might be solved completely outside of this community where we might be contributing. But effectively, like it is a trust symbol that you can trust, you can use it, it’s available, and there might be services by members that are automatically available [00:38:00] if you get a .agent domain.

Like that you can just like Cloudflare, like you can just make it happen that you have an agent just by buying a domain. Mm-hmm. A lot of the how to run a agent, you get the API key, you get the harnesses, you pick whatever you want on a self-serve framework, and then all that gets put to- put together by the agent, community agent that helps you to decide what your agent should look like.

** :** Mm-hmm.

Balazs: Mm-hmm. I’m not saying we are working on this, but the idea is that And it meets the criteria. Is it like technical pieces? Is it registration? Is it, something the government registration piece? I’m not saying I, really, really want this, the government registration, but- Right

but the question of accountability is really not being discussed- Yes … enough beyond it will destroy the world, and that’s the end of humanity. But if this gets better, then petty crime or like big [00:39:00] crime or was it hallucination, these questions are going to be discussed in courtrooms very soon.

Luke: Mm-hmm. Absolutely.

Balazs: And the topic is just fascinating. So I think like one piece is that there are so many topics to talk about that are positive and potentially negative as well when it comes to the agentic web, that I’ve been mostly referring to this like shift.

I’m not always in the AI bubble. I don’t technically live in San Francisco right now. I’m seeing normal humans. I’m seeing- … normal businesses. I’m not seeing that AI is about to take over every business. The US- No … itself is unique. In the US, it is- Yeah … everywhere. Outside of the US, it is ChatGPT, a little bit of Claude.

And about that’s it. Of course, there is data center build-out. Everyone is aware of, like, [00:40:00] all the Instagram videos.

** :** Mm-hmm.

Balazs: And how it makes the experience kind of not better. But it’s not the agentic web happening outside of the US. In the US, I think the economic impact of AI is way bigger in percentage of the glo- of the country’s, like, GDP than in- Mm-hmm

many other countries. Uh, maybe in Singapore they just spend so much money on, on GPUs. But it’s a rich country- Yeah, sure … it’s a very different experience.

Luke: It seems like, in that respect there, the time is now, right? And we’re not, too far.

Things have not gotten to where there isn’t an opportunity to, to do this from the community level and to get ahead of this problem in particular, from becoming a, a problem, right? I think it’s a amazing effort, that you’re putting out there, Balazs, and, We covered a lot.

I think our audience has a lot to digest from this. Again, I’d encourage everybody to go check out the Agent [00:41:00] Community. Where can people go to, to check it out and to register?

Balazs: Yeah. It’s agentcommunity.org.

Luke: Awesome. And where can people go if they wanna, connect with you or follow what you’re putting out there into the world?

Balazs: So probably Twitter is the best. It’s, Nembal. I think the, my name will be in the title.

Luke: Yeah, yeah. We’ll throw it in, the show notes for sure … and it’s

Balazs: N-E-M-B-A-L. Excellent. The, there is a story about that as

Luke: well. Right on, Balazs. I really appreciate you for making the time today.

Thank you for educating us on this effort that you’re taking on. And appreciate that you’re doing that too. And, love to have you back to kinda check back in on how things are going.

Balazs: Yeah. Well, look, super appreciate. And, yeah, thanks for having me.

Luke: Right

Balazs: on.

Luke: Thanks for listening to the Brave Technologist Podcast. To never miss an episode, make sure you hit follow in your podcast app. If you haven’t already made the switch to the Brave [00:42:00] browser, you can download it for free today at brave.com and start using Brave Search, which enables you to search the web privately.

Brave also shields you from the ads, trackers, and other creepy stuff following you across the web.