Alien: Decentralized Human Identity and Agent Trust for the Age of AI

追加: 2026-05-09

[00:00:00]I believe that the future of um proving something online is never absolute. It's always a probability. So, are you human or not is a probabilistic [music] question.

[00:00:16]>> Carill, thanks for joining me.

[00:00:18]>> Thank you [music] for having me.

[00:00:19]>> Maybe we start off by you explaining what Alien is.

[00:00:22]>> Alien is a um decentralized unique identity platform. And unique identity here is very important in a sense that um usually when when people when you talk about digital identity they kind of like dismiss that word. Um but unique is meaning that one human uh equals to one account. So you can think about this as like an SSN um in the age of AI um online that is not connected to any specific country and um that is um a scalable new blockchain um that is um growing virally by users inviting each other and vouching for each other and verifying for each each other and we have just launched a few days ago. We've been in stealth and building this whole thing. So mainet is live today.

[00:01:08]>> Awesome. Congrats. All right. Well, I I think there's a lot of parts of that that I would love to dig into, but maybe first like why like how did you come to build this? Why why are you building this? Like I mean, tell us about your background and motivation.

[00:01:20]>> Right. Well, I I started coding at 10 years old and and um I I've been as a teenager hiking around like different kind of products and and and small things. And at one point um I was a youngest employee of VK. VK is kind of the Facebook office of Easter Europe, one of the largest social networks. And um back then when I was teenager when I was a teenager there was a thing um um called social graph API and like API that somehow was also available on on Facebook before that but then on on VK was widely widely um kind of accessible and very simple to use. And so I was building different apps and uh bots u on top of this platform to kind of propagate um different messages and try to like create some awareness um on the social network. Um when you get like uh access to people messages and and and not to messages but to to send them messages and and and like um being able to um notify a lot of people. um you could create a lot of like um information kind of like networks built on top of social networks. And so um I've seen what's possible back then um in terms of like um how you can create really easily and manage a lot of bots just by by being a single person. Um and um I also was a part of the largest social network uh really scalable. I was building a bunch of social apps. um uh and um mobile social apps for teenagers.

[00:02:54]And um one of the things that always kind of um I saw is uh the reputation um on the on these platforms that exist uh but actually does not exist. And the reason why it doesn't exist is because you constantly need to recreate the social graph. You constantly need to recreate the identity layer. Um and uh internet has is like the perfect kind of like protocol for uh sending information over the like large uh um distances for a lot of people. Um [snorts] does not really allow you to um coordinate people and that is that is that is quite weird.

[00:03:30]That's what I what I what I noticed after moving actually here. Um the uh um as a like a immigrant having seen a lot of uh um kind of like uh things for example the um visa situations for for immigrants is is quite difficult and and uh all of that has always somehow connected has been connected for me to one single problem which is um identity and identity is usually not a consumer product. So people don't think about identity but if you look at like internet identity has never been built and embedded inside of the protocol [snorts] um while we have HTTPS and and HTTP and um other protocols. So the uh um that kind of became like the um initial um spark for that we need to solve this otherwise the a lot of companies are suffering. Um there is a centralized power inside of the corporations, governments are suppressing like in Russia governments are suppressing uh people with censorship and like um um a lot of different things like that. So um and then the interesting thing is like when I started building this AI became a thing um and um it became clear that um it will be very very soon if you include the um compound effects of AI that uh we will not be able to distinguish like anything online and so that internet theory became kind of like a myth that's everybody started to think more and more >> um so um and currently I guess um with Nana Banana it's kind of like already clear what's going to happen with videos soon and stuff like that. So um every single like information that is being shared online um is actually coming from a single point of truth which is someone who is writing it either like AI or a person and um being able to cryptographically prove that um and sign different messages kind of creates this trust framework for for the internet.

[00:05:33]And so um that is kind of like u what Ellen is all about.

[00:05:38]>> Yeah. Yeah. That there's a there's I guess a lot of interesting things in there to pack. I guess what strikes me is like yeah like it makes sense.

[00:05:45]There's there's this there's this internetwide problem of identity um in that it's now easy to create you know this like kind of civil problem across all um uh uh network systems. Um, and it's also like a very acute problem in distributed consensus >> and it's interesting that like the solution is is a form of distributed consensus. So like what brought you to like was it just like you're like I want to solve identity. Uh there does likely need to be a blockchain at the core of this or like how you know talk talk about like that fundamental architectural decision.

[00:06:24]>> Yeah, great question. So the um the interesting thing is like if you look at all the projects that are claiming to be doing something about identity um they're usually in two camps like one is um extremely technical not thinking about distribution at all um and just like purely a the promise is like you can own your data um which is a cool promise but actually people do not consumers it's an abstract consumers don't really care um >> um if they cared about this then there would be already a solution and whereas as we see like Facebook owning all data, people are fine with that. So um and the the other camp is um kind of like uh centralized solutions. Um so the the the centralized solutions is like actually both for the companies that trying to create like an ID system like stripe or uh governments like u a lot of governments are actually introducing CBDCs and CBDCs are directly connected to like centralized digital identity. So those two camps are happening. Um and I would actually would not even think about the first one because they have just zero users. I haven't seen any any product like that that that um but if you look at the second cam with CBDC's um the um the interesting thing is like it's an inevitable future um if we don't do anything about it. Um because governments have a tendency to like total control. Um uh I think right now there are like 49 or like I forgot the extra number I think something like 50 countries already have implemented it.

[00:07:56][snorts] Um and um the uh the only kind of solution is not to say like we don't need that or like we will not use that uh but actually to build an alternative.

[00:08:06]So, um the um coming back to your question about the this the being the solution to be a a a consensus like the um the um looking inside of the solutions and and thinking about like how can we make this not just technically possible but also how can we distribute it to a lot of people. It became clear that the solution um definitely needs to be decentralized. Um and now the question is like should you build it on top of some other decentralized networks or you should build your own and um if you think about decentralized networks um they all have a lot of problems like uh speed is one of them or you can also think about them as like countries and governments with their own kind of rules. So when you build on top of them you inherit a lot of kind of like um foundational decisions that have been made. [snorts] Um and uh um the other interesting thing is like building a solution like that if you actually can enable it to grow to a lot of people. Um the value that this whole network would create for for the people who will own that value because in essentialist networks um more people own the value than an essentialized network.

[00:09:20]So the um in case that like a lot of people are asking us like what if you build this on top of Ethereum like if you build this on top of Ethereum then um all the >> ETH holders >> ETH holders or like uh which is like a minority actually they have they did an ICO where they have minted like 70% of the total supply. Right now it's like I think still 60 or 50% of the total supply of the whole ETH is being held by those ICO like a few thousand people. um they will become richer because Ethereum will definitely become richer because there's going to be like u more um kind of like uh uh value created for that. So um but fundamentally if you want to build a system that tries to be the most credible neutral and and like trustless um the um only solution is to for everyone to use it is to give everyone the value that they produce. And that became kind of the initial kind of like insight for us that we can actually do that. And so we use identity um inside and we use the unique identity which is as some other people call proof of personhood or proof of human um on the consensus level. And the the way we think about this is um every consensus mechanism is actually a um a voting machine. Um proof of work people vote with CPUs. Proof of stake people vote with dollars. Um and but they all dismiss their simple fact that actually the whole system is built usually on trust. It's not all purely cryptography and math. Um Bitcoin white paper has the word honest I think um u 20 times or something like that. Um and uh he also he he always talks about like that the system is for honest participants. Um so honesty is very important. If honesty didn't exist, Bitcoin would not exist.

[00:11:10]So the um um for us um the the system works because we understood that uh proof of work and proof of stake are kind of proxies to the idea of voting system which is one human one vote. And uh when you think about one human, one vote and if you actually can do that, you dismiss the blockchain trilemma >> um which people kind of like things these days as one of the um um blockers that you can't actually surpass like specific um limitations and it allows you to scale the system. So um thinking about >> because like mo like most attacks on distributed consensus have some grounding in in civil based.

[00:11:55]>> Exactly. That's the the the the general problem of decentralized systems. Uh with decentralized voting, you need to know that the person did involve more than one type. This is why um the core innovation of of Bitcoin exists. So the um and uh um yeah and so the uh that's kind of like >> all right. So then it becomes similarly a question of like there's a lot of lots on the back end like can you do this? So what what convinced what convinced you that you could cryptographically uh ensure that you could pull together a system that cryptographically ensured one person was one >> account.

[00:12:32]>> Well it's important to say that we still haven't reached a billion users. So we don't know if we actually succeeded but uh supposedly that I have thought about this for like a really really um a lot of uh sleepless nights. Uh I would say that um I started to think about this as a hacker from the side of a hacker like um if you're thinking about security systems you >> adversarial. Yeah.

[00:12:55]>> Yeah. You think about like how would you hack it and so how would you not allow others to hack it? Um, and so the uh um I I think the one of the interesting things is like people know a lot of a lot of ways uh on how to verify someone.

[00:13:15]Um but none of those solutions actually um help you to solve all of the things at the same time which is being decentralized, being able to have a simple UX um being able to verify 8 billion people and um >> like an example would be like go down somewhere with your passport and like >> right example of that is is that or like uh biometric device or like um [snorts] and privacy is very important as well and or or Um there is like a system called web of trust or um some other versions is like yeah verify with your passport or with any like government identity. So the government identity problem is like you have a trust to to a specific kind of party. So we cannot just focus on that um because at some point when you become too big governments will be able to um create fake identities which you don't want to happen. Um and um the uh what what I realized is that uh there is no ideal solution. Um and but one of the things that I think a lot of founders dismiss these days is like thinking about hard problems and thinking about like that it's actually if there is a hard problem um it's actually might be curious to think about how you can solve it rather than like dismiss it like impossible you know. And so this is the path we chose. Um and we found a way to think about this in a in a kind of um an analytical way where um we could combine multiple factors of uh verification um all those known like biometrics like social graphs and web of trust like verifiable credentials um to a single kind of like chain of events. And if you combine those um you can um attack one single kind of factor u but if you combine those together it's kind of becomes really really expensive to attack um so it's I mean like a lot of things in crypto [clears throat] central because it becomes this probabilistic >> yes >> uh approach where it's you may not have perfectly you know a proof that that's one person but it's it's get you can get it to where it's approaching approaching impossible I would say it's impossible possible to say with certainty 100% um ever because we live in a in a high entropy world. Um so um yeah the um uh but it's uh it's always about the cost and uh I believe that the future of um proving something online is never absolute. It's always a probability. So are you human or not is a probabilistic question.

[00:15:54]>> Yeah.

[00:15:55]>> Okay. And what what got you to confidence that you'd come up with a system that probabilistically, you know, resolved that question?

[00:16:04]>> Yeah. Um we thought about like step by step like okay there is a biometrics for example um the um biometrics um uh there are multiple ways of doing biometrics like if you if you say you if you if you take your phone and take a front camera of your phone and scan your face there is a you you actually can um create a uniqueness but on a small cluster of people. Um there's some other ways uh of like hacking this but the um uh then the question becomes like okay if you if you can do this for a small cluster of people how can you make it for for bigger cluster of people and and uh to make it for a bigger cluster of people you need to take the whole humanity and clusterize it somehow. How can you deterministically clusterize the humanity? Um well um there are actually multiple ways of doing that. One of the things that most people have um and uh u it's quite deterministic and you cannot change it uh is your birth date. Um now the question is okay if you cluster people based on your birth date how can you make sure that the birthday that people um tell you is actually uh the the right birth date? Um um so and we found the answer for that. And then the question is like okay if you um clusterize all the people and you know that they they confirmed the right uh date, how can you make sure that they uh did not use um different faces um of their friends for different dates uh birth dates. So they can like create multiple combinations and have basically unlimited uh kind of like uh and we found the answer for that. Actually the answer for that is is a social graph um which can um your friends can confirm that this is actually you. Um and there's like a game of theory incentive system inside for that. Um and then in the end like we have like thought about this and like okay um can we make this as simple so people actually would do this and uh it would be like low friction um and it would be quite easy to onboard other people and so um and uh and we found the answer for that. Yeah, because you're already I mean you're already on a cost continuum where like what the one end of the spectrum is like going and paying someone to verify everybody and so like that's this is always this constraint of your system is that you need to onboard everybody and do it in a cost uh not cost way. Um, so it's very I mean it sounds like the steps were just very um just finding bottlenecks and >> yeah it's kind of like it's kind of like a um step by step uh approach. Um and I wouldn't say that I I would I wouldn't say that this approach is kind of like as a perfectionist um engineer I I would like 100% love it. Uh there there are a lot of like things that I don't like about it. uh but I like talking to a lot of people um I don't think there is there's like another solution that would um uh kind of attribute to the um um uh levels of um our um requirements >> and then you but then you also note that like you're not entirely sure until it's at scale because there's you know this clustering all all these all these things are are are a function of number of people on the system and collisions or the like, right?

[00:19:26]>> Well, there are there are possible collisions. We we uh social graph actually helps you to solve collisions.

[00:19:31]There are like a lot of corner cases when you think about scalability of 28 billion people and how to make it auto automized because like a lot of like corner cases in centralized systems are being solved manually by people. Um, so here it's just going to be um also people, but they're going to be um in a like uh in a decentralized network being able to kind of like vote for their decisions.

[00:19:54]>> Um the um the uh the important thing here is also like if you actually can technically do this like what would be the motivation of people doing that? And so as I coming back to the initial conversation um this value that is being produced um inside of the network is being represented in in in the form of an alen coin. An alen coin is being given to people for verifying who they are. Um network becomes more mature. The lencoin value um becomes more mature mature based on um madov's law. Um if you're familiar with madov's law. So the um uh and over time >> wait so the the the the value of the network is the square of the >> right logarithm uh >> of the number of participants >> right so it grows >> yeah it grows exponentially >> um and um and we actually can see that in in form of like Facebook shares uh that is being distributed to shareholders with more users are coming to Facebook um obviously there is more utility and we need to build this more utility inside of the inside of the system. Um so >> and then so all right so this is like a separate thing now now you you have you feel like you've gotten confidence and that you have a method for verifying people uh and you and you think that this should this sort of system works best and is if it lives in the comments effectively and is owned by all the participants and now you have like a cryptoeconomic system to design on your hands like how did you approach that I mean this is a whole different type of uh engineering and thinking about problem solving in that it's perhaps perhaps emergent, right?

[00:21:40]>> Yeah. Yeah. Well, cryp like cryptoeconomics is actually very interesting. Um it's very it's very um hard these days to go against the status quo. Um and I think um one of the things we did is like we did um a lot of unique decisions that are um not approved by the current like market conditions and how like the coins should be launched.

[00:22:06]The thing is a lot of people are launching coins, they become rich um but their protocols are ultimately never reached the scale um to be as useful as we want the protocol to be. And so if we um if we haven't seen that then the question is like then probably something is wrong. Um and people are not asking that question like people just doing what what they what others are doing and uh >> there's intention with people care like perhaps the thing that made them rich got people to care about the network in the first place. So you're like balancing.

[00:22:37]>> That's right. You need to balance it.

[00:22:39]>> Yeah. The uh the um the the main problem that we we saw is that most of the networks out there are um giving more actually I think all of them are giving value most of the value to early adopters. That is a very limited amount of people. Um for example, if you take Bitcoin, um the uh people who have mined Bitcoin in the first like four years of existing of the network >> are going to be the and some of them are already the richest people in the world.

[00:23:14]>> Yeah.

[00:23:15]>> And your children or me who was like born like 5 to 10 years later than I when I could actually buy some Bitcoin um >> are never going to be able to have that opportunity. So that is not I believe like quite fair but at the same time it creates this huge kind of like um belief for early adopters and being able to to propagate the u yeah it's well the the the the distribution for the network to maximize utility at scale is like misaligned with um the actual utility of like like the the early adopters are higher utility in terms of getting the network off the ground >> but then suddenly not anymore once it once it reaches some threshold. So it's like so you're you're threading this needle of like how do I figure out >> what I want to be on the other side while al while not risking no one caring.

[00:24:08]>> So the the the um um I think it's pretty clear in terms of how to create a new store of value and Bitcoin and Salana and Ethereum have shown that like you it's it's it's very understandable how to create store of value um and how to make it grow over time just need to shrink the supply over time. Um the um but uh to be able to do something more than that um you need to think about how to give more value to bigger number of people and how to um kind of like have the inflation rates um uh sustainable over time. And one of the things that alien does that don't I haven't seen any other network to do is because we have the unique identity so we know actually how many participants there are in the network we can uh we can have a deterministic um supply um based on uh pegged basically to the um amount of users in the network and we can do this in a automated way. So there is no kind of like central authority that is like issuing this currency um so it's all a protocol based currency.

[00:25:16]Um, okay. I guess switching gears slightly like there, you know, in your early days of working on social networks, like the identity probably arises based on just this bot problem.

[00:25:28]And now we have AI and and the entire thing has gotten way worse and crazier and and and it's even more complicated in that these these a these a agents feel a little bit more like real entities as opposed to just wrote bots.

[00:25:45]And so like how like when you think about you know like I guess tie this back a little bit to like what's going on in AI and like what are you thinking about when you're building alien relative to agents being everywhere? Are they participants in the system? Are they more things? Are they just something you need to defend against as a more aggressive civil or you know what?

[00:26:05]>> Yeah. Well, first of all, I haven't seen yet a actual agent that I would be able to say, "Yeah, this is like a smart thing that can do a lot of things." So, >> there's another level to the I guess the touring. There's a lot of people talking about that and I haven't seen a pro >> but but we could but at least like you could it feels much more possible than >> right it's going to be it's going to it's going to happen. Um we're definitely on the trend of like uh over um kind of like um um hyping the the term. Um but um the uh interesting thing about agents is um we can apply the scales of um apply the compound effects of of and and and scalability of um uh AI and how the it it moves um and um how quickly it becomes like smarter and smarter um to to agents and um think that agents are actually um just uh kind of like uh software um engineering tools um with intelligence inside.

[00:27:14]>> Um so over time we're definitely going to be able to um build those tools that are um becomes much more autonomous um and uh can do many many more things um um that humans cannot do. And the um um when that when that's going to happen um I think the world is not ready because again I I've seen what's possible with like a with bots that one person can do actually uh Russia is still using Kremlin bots for propagating a lot of misinformation. Um and those and them and they are using um real people behind those. Um so imagine that you now don't need people because people are a problem. uh you can just do this on your own like you can one agent that gonna me manage like thousand agents that's each going to manage another manage another like thousand agents right so um and agents are definitely going to be um part of the society um and um we definitely will start treating them as some other entities in our society um um and uh saying that like this is that is kind of like um thinking against like the technology and progress. Um but how can we apply what we know about trust and and and and cryptography um to those systems to be able to propagate trust and actually be able to use those uh agents um to um help humans rather than like um being uh to treat them like in a bad way. So um I believe that alien is actually the solution to that um because um human identity agent identity um agent identity is actually much easier um in some ways in some ways it's harder but uh the um you can link people people's identities and and agent identities um and now you have a cryptographic proof that some specific agent is acting on behalf of a real human which is I think very very valuable um because Now, um, as a website owner, you you will not be able to um um you you'll be welcomed to to to get to get an agent from a real human because it's kind of like it's almost like an human coming to your website um rather than a random bot that tries to like scam you or something like that.

[00:29:42]So, persistent reputation that can be inheritable to agents is um another concept. And uh giving permissions, custom permissions to agents um for like what data they can use of yours or what um like money um access they have um um and and different things like that is another kind of like layer on top of it.

[00:30:06]>> Um I guess again switching your slide.

[00:30:08]So like like AI you know is changing a lot of how software gets written. You're running a lot of software. What does that look like internally in Alien? Is it like kind of up to the people on your team to decide what they want to use? Do you have like an imperative that people start using more AI based coding tools or what's the >> Oh, that's a very interesting topic. We are currently like one of my um um kind of like first employees um one of my closest actually friends um um is uh um constantly pushing me every day that I'm not using AI enough. Um and what we started doing is we started um thinking about how we how can we create like systems and specifications for the protocol for the products that we're building um using AI to be able to have a smaller team count um and and be able to build more more efficiently um than big companies. I don't think that a lot of people are doing that. I talked to a lot of people and I haven't seen uh many of them doing anything like that. The interesting fact here is as a developer and as as a non-developer, as a product manager, you both can use those tools.

[00:31:17]Um, and I don't think that a lot of product managers have thought about this. Um, >> well, I mean, because you're both you both are a team of developer, both developers and have a party product developer facing, right? So, I think imagining thinking through how development is is happening is important on both sides.

[00:31:35]>> Yeah. Um, and uh do you think that um I guess does that does that inform like how you're making decisions about how people might integrate? Like are you trying to uh think through what what is going to happen with the next generation developers, how they're going to integrate and and build on top of your network?

[00:31:56]>> Uh yeah, we definitely think about that.

[00:31:59]Um I we we actually we're like experimenting a lot with like different things that we build inside for ourselves and then we're thinking how can others um utilize that. Um and if there's something like >> it's like if you find it useful >> that's probably makes sense certainly as well.

[00:32:20]>> Yeah. Yeah. Um yeah I don't think we have I I I have too many things to share here. How do how do you balance um building building I mean because it's like I think that like what I what I can hear is that there's there's a lot of things that like you know the the current generation models are very good off the shelf like they're well represented in the data set. It can go spit out like a one-off CRUD application fairly quickly. There are things where it's much harder to get it to do what you want. it probably can, but um the decision-m around what what to build um is kind of lost on on the AI by default because a lot of a lot of what a actual human doing is pruning what not to do before they start writing. And so that means that like you could spend a lot of time writing a bunch of harnesses for your internal coding practices, but you also could just wait to see if the next generation's better and like like is that a b is that something you think about in terms of balancing like how much how much to like how many how many cycles to spend on internal tools.

[00:33:23]>> Yeah. Um well there are some fundamental things that I think are um not going to change very soon with like models like for example memory um is uh the um is the part that everyone dismisses right now but they will need to start thinking about this more because this is a fundamental thing. Um and we for example thought about this because um different things. But the uh um the um the interesting thing I feel like is if you look at um Oppus 4.5 um it is a extremely intelligent model that you actually don't use enough as a developer or product manager at all.

[00:34:10]like um it it is very nuanced like how can you use it um to make it actually to be able to utilize like at least 80% of what what it can do. So a lot of people actually don't don't need that much of intelligence for for the for the regular tasks. Um the um um but I think it's quite clear when you're when you're in a space um and and you um talk to a lot of people and you like try different products and you see what's going on. It's kind of you can extrapolate um in terms of what's going to happen in the next like six to 12 months. Um and then um the question is like um what's going to happen after?

[00:34:51]And then this is like usually a guess and and that is the most interesting question. Um and if if because if that is going to happen then you can build for this right now and you'll be ready in 12 months rather than like if it's going to happen in 6 months you're kind of already late.

[00:35:03]>> Yeah, that makes sense. So kind of like you do have to bet around lineup side you know just like timing risk and all this stuff. Um what so what what's like what's next for Alien? What should what should we be looking forward to?

[00:35:15]>> There are a lot of cool things that are uh coming soon. Uh well, first of all, we just launched the mainet. Uh it's phase zero. Uh there are multiple phases about how we are um rolling out the network. So phase one is coming um then phase two early like middle next year um this year 2026. Um then uh the uh uh alien coin market so um is u kind of like being rolled out in the process right now. Um some people can already buy it, sell it. Um and uh the um we have given an integration with Salana.

[00:35:51]So proof of personhood exists on Salana and we are actually creating a lot of a lot of different really cool projects and products as like um use cases inside of the Alen app. um that are coming soon and we are going to be releasing them like every week or so. Um and uh just scaling and and and and more news around this.

[00:36:14]>> Awesome. Well, that all sounds really exciting. So, I don't know. This has been really fascinating. Thank you for joining me.

[00:36:21]>> Thank you so much for having us.