Can Corporations Be Replaced? Yes, With AI and Blockchain, Says Illia Polosukhin

Transcript

in that backdrop like how do you build an organization? How do you structure an organization? How do you get people to coordinate? Um I'm I'm really curious to get your perspective on this but uh but I think we're out of time for uh >> give us like two sentences though like this game. >> Yeah. I mean the the high level where I want to get to is markets that that's the highest level, right? like we're transitioning from corporation or foundation or whatever needing to be the holder of people because you needed kind of to to bundle because you had trouble doing credit assignment right like if you know Infinex done something it was really

hard to like this individual person contributed this much to this product >> I think that actually is a lot easier to solve now with AI right it can actually estimate this and then on the other side each individuals becoming more productive. So like I mean it's you know I know multiple people who working on a bunch of projects at the same time now uh because you kind of have a lot more bandwidth and so I actually think we're going to start transitioning to more organizations almost that are just operating on the market principles where you're effectively hiring you know an individual and their army of agents to do something and you kind of can estimate a like how much that contribution would have done to the

bottom line of the bigger thing you're doing right so I don't know if you saw so Jack Dorsey actually like he posted something like which is spiritual and kind of going from hierarchical organization to more kind of uh peer-to-peer and again this is where blockchain and kind of all of the you know intentbased stuff that we've been doing is exactly that's exactly the world we've been building for where you effectively can say hey this is intent like I want this new feature in infinex and now you know there can be multiple what it's worth to me delivering. >> Yeah, this is what it's worth. Or you can have a bidding in an auction and finding, you know, a solver that will say, "Cool, I'll build it in, you know, 24 hours." Like, "This is my expectation of like of quality it's going to be."

And then if you don't agree, right, there's another agent that comes in and reviews it and effectively judges if that was done. And so this is actually what we're building as our neari marketplace and kind of testing this out. right now this is still early so I'm not uh but we have like first actual businesses starting to use it like for insurance claim filing for background checks for event then like actual you real use cases where they're effectively hiring agents to do some jobs and it can be like because somebody who is an expert effectively codified the workflow packaged it up and now offering that for a dollar >> as a surface right aggregating what would have had to have been inside of a corporation you know like you have you're that that function of three,

four, five, 10 humans that are you know manually >> doing that all the time and sometimes they don't have enough work and sometimes you know you pulling them to do something else now you can yeah disagregate all this and then have a marketplace and so that's actually I think where we're going it's I mean it's not going to be it's going to be a bumpy ride but I think that like systematically that what makes sense and again this is where AI and blockchain really fits well because they're kind of powering this trustless transaction we have escrow we have verifi ible AI to to evaluate if the agent the job well you can still have human in the loop right to kind of crosscorrect if something went wrong. So it doesn't need to be like and now everything automated. It can be like you know I hire you know

Kane's agent and you know Kane will like make sure that his agent done the work right or like corrected before passes his overright. >> My other agent will make sure that the other agent did did a good job. I I I won't be involved. It's going to be fine guys don't. It's going to be great. >> Well, so so like within the near ecosystem and and we've experienced this, right? Like you guys are already somewhat in this direction like these smaller corporate entities like smaller smaller business entities. >> Yeah. I mean that that has been my principle from long time ago. So I mean again to me AI again is an accelerator, blockchain is coordinator. The principles are the same. But so so very but from a very practical standpoint,

right? Like the this is something that I'm I'm like super focused on at the moment. Um you know, we went from 2025 where like you could grind your way through and get some like production code was really hard. Then December 2025 through to like maybe March or something like that. You could you could not just grind your way through like you could just kind of prompt and get like production if you had uh you know enough experience with Um, and so a lot of engineers were like, ah, I don't need to write the code anymore. I can get agents to write the code and have you, but it was still very much like a synchronous process of like a single person sitting in front of a computer producing code, right? Um, that doesn't to me feel like the final form of this, right? Like it

feels like we we need uh something else. And so the interesting thing about centaur what what paradigm uh dropped last week is like it's trying to move the agents into slack so that you have like an agentic partner or teammate or whatever that is responsible for the coordination that an individual human would you know by part to to coordinate uh and can you know deploy agents and then like it's really the the humans in the loop for taste and judgment and and goal seeking and and all of that and then like the execution side is all handled by by agents. Um like you know given how much you guys are are doing in

this space and and how long you've been involved like from a near orb perspective are you still in the mode of one person sitting in front of one terminal doing things or are you experimenting in this like multiplayer space? >> Yeah, we're experimenting definitely and so the ironclaw itself is actually multi-tenant. So open open claw and hermes like they're effectively like your system and so >> ironclaw actually supports a multi-tenant mode where you can log in as Google you know or or near account and you're effectively joining a single instance you have your memory you have your kind of missions etc but you also have shared projects you have shared memory you have shared workflows and so yeah we're experimenting kind of

designing how we call the commitment system so whenever you know we having a meeting with Ukain And like Ilia said, you know, yes, we're going to ship this feature for you. This should actually go somewhere automatically and then, you know, pop up. >> Yeah. Well, it also should pop up for the team. And then I also like re, you know, follow up if it's done, put it on road map, update the right documents. Similarly, like I actually have workflows that, you know, scan everybody's meeting notes and like give me like decisions that were made decision.

>> What do what do my guys commit me to? not for me but like in general for the organization but also like actually the most useful being finding where decisions were not made >> because that's actually where organizations taxes is people discuss something and did not make a decision because as a decision maker is not clear or it wasn't in a room and so just extracting that and making sure that's visible like it's already helping because then I can >> and is that like an org level panopticon single agent multi- aent thing like how how >> right now it will prototyped it more like from my perspective just because I have access and but then we are now building this as like a yeah a system that everybody again the the main

question is permissions right like for example there's some HR things right they shouldn't be visible to anyone except >> maybe some HR people and so like you need to kind of configure that but uh otherwise yeah like you know I may have like full purview and then we kind of can like cross cut it into like specific departments but Yeah, we're kind of building it as like a single near foundation at least right now. Ironclaw that has everyone kind of as a user and then there's like a common area of his permissions as well. >> Yeah. Nice. Yeah. This I think this is like the most >> fertile ground for efficiency improvements. Uh that that >> Oh, yeah. 100%.

>> But it's super hard. Like it's hard to reason about. You start building something, you're like, uh this isn't exactly the shape of it. like it's it's a really complex uh design space um that we don't have good intuition about. It doesn't because it's just like >> we've only had humans. >> You just like throw a human at a problem and like they're just different. >> We're very bad at the human side of this like organizations, right? Especially like or running like an organization at scale >> is bad >> is like it is bad. >> It's always broken. Organizations are always broken. That's the >> always 100% always. And so it's like then you're like, "Okay, let me slap an agent on this bitch." And it's like, >> solve everything. What are you talking

about? I mean, >> so you have to you have to like either I think like the approach, right? Like okay, let's let's just solve the problem which is such a common organization problem like why wasn't the decision made or like and also just tracking the decisions that are made. you can actually solve that and bring efficiency to that and then you can then sort of circle back around and be like okay clearly this area of the org this team whatever yeah and like non those non-decisions yeah I worked at many companies where >> yeah I mean >> my entire life was not making sitting on calls that didn't make decisions and me getting like just banging my head why are we here >> yeah I mean I So I pitched this actually

near con 2023 I pitched this idea of AI CEO right this was before any of this AI stuff started to work right and so my my vision always there was like yeah like working from two directions one is how to make individuals more have more context because like a lot of the middle management and kind of it's just like propagating context from like relevant places and at the same time cutting down stuff that's not relevant because like people cannot just like you know have a feed of everything happening in the organization as well and so that is AI like AI can do that very well right like whatever is relevant to your work whatever is relevant and you can still query up whatever you want to know and then the other way is like bubbling up in deep information is like what is

going on what decisions are made what suggestions what's improvement what is like where the trade-offs are like that's the other thing a lot of it is like we're operating on a trade-offs like what are the trade-offs like sometimes like people making decisions have no idea what the trade-offs were that actually led to this and this is a this is actually by the way why I think Dows never worked is because all of this infrastructure is missing >> and so like for Dows to truly work you kind of need to have this system truly working where you don't need middle management you don't need you have credit assignment embedded into the your organizational design because right now again credit assignment is a very like credit assignment I think actually is the fundamental world problem >> interesting

>> like on an organizational level in AI machine learning what gradient descent like the training of models is create assignment problem and so like in every in every place you go it's like create assignment is the main challenge but like if you can assign like oh you know this person contributed this much to this bigger thing we've done if you can do that then like again right now like promotions compensation all of the things are like somewhat arbitrary >> very fuzzy and and adversarial >> and so and and that was put in >> and then you yeah exactly because then You put rules like these very hardline rules on things that are fundamentally fuzzy underneath and then you're like it's fair >> but you have to you're like h we'll just like >> Yeah. Yeah.

>> SL it up somehow and go >> I mean but you know but this other company gives me like double the compensation right so like probably it's not fair. >> It's probably not fair. Yeah. Exactly. Exactly. Yeah. And you know it's funny like there's the unintended consequences of like how you incentivize people are like so funny to me. You know, I I kind of vaselate between these like communist phonuses, right, that we call them in in uh Infinex where it's like, hey, everyone was here. We can't really attribute, you know, perfectly, so we're just going to give everyone the same thing, right, for this particular thing, right? And like the the people contributing the most hate it.

They rage so hard. They're like, "This is a travesty." you like anyway so >> but but I mean I know we're we're over time but markets actually solve all of this like if you forget about or inside organization between companies there is market and so if you don't like the partner you switch to another provider right if you like markets actually solve all this problem we just haven't figured out how to do it on a scale >> lower level and this is again where AI and blockchain and kind of the intent architecture is actually like built for >> if you like this clip Tune in to uneasy money every Wednesday at 400 p.m. Eastern.

Multi-Chain Advisors is an emerging technology growth firm that has helped create over $50 billion in enterprise value for more than 80 clients like Pith, Moon Commerce, and Wormhole. They've worked with some of the largest and most impactful companies in the space. They're the partner you want when you're navigating markets and trying to break out from the noise. They help navigate TGES, go to market, BD and partnerships, capital markets, advisory, PR, media placements, KOA activations, and more. Driving execution from launch to scale. Their results are measurable.

To learn more and start building real traction today, visit multi-chainadv.com.