Polygon CEO: The New Era of AI Agents Onchain

27 Jan 2026 · 25:18

Sandeep Nailwal, co-founder of Sentient and founder of Polygon, joins The Rollup's AI Supercycle series to question assumptions about the overlap between AI and crypto. Nailwal argues blockchains matter mainly as payment rails for autonomous agents, discusses Sentient's open source reasoning framework and its Arena for testing agents, and touches on Polygon's recent payments-focused acquisitions and token economics for AI protocols.

Speakers

Topics

Transcript

Apart from doing payments and some very niche smart contracting behavior, I don't think that the AI and crypto per se has a lot of overlap. But why do you think it's so crucial that we implement blockchain as a fundamental piece that powers a lot of these AI agents? [music] Welcome back to AI [music] Supercycle, our premier AI show airing every single week presented by Near. We cover the ins and outs of decentralized AI, privacy, and the future of this [music] massive technology. Near is the blockchain for AI and the execution of AI native apps.

You can check out Near's latest AI product at near.ai. Now, sit back, relax, and enjoy the show. What's up, guys? Welcome back to the rollup. I am here with Sandeep from Sentient as well as Polygon. You guys see the shirt. You recognize the purple. Uh we're here on an AI super cycle. This is our flagship AI segment. It is powered by Near. They're doing great job in AI as well as Sentient who is backed by Near and uh I I believe this is a partnership that's incredible. Sentient's got some big news. the token is live uh or depending on when you're listening to this it will be live very very soon and so Sandeep is here to talk to us about AGI sentient

and everything that is happening uh in the world of crypto and AI sandep you're a friend of the show welcome back man it's a pleasure to have you >> thank you thanks for having me on >> absolutely last time uh last time we saw each other I think it was in Korea we were on a on a panel together in KBW and uh it was fun feels Feels like a long time since KBW. >> Yeah. Yeah. I I don't even to be honest like it got wiped out of my memory. Like it's so long back. Which one was that? >> Um what it was? Man, I think we were talking about L2s and interoperability like composability with L2s. Ben from was on there.

>> Which month was that? Which month was that last? >> Oh, KBW. Was it October? November. >> Oh my god. It's like only two month two months. Three months only. So much so much happens, man. You've been busy. >> I almost feel like I'm thinking like was it in April, May, June, like you know, this is October. >> Yeah, >> KBW is back in October. What What have you been up to since then? You've been busy. A lot's been going on. >> Yeah, I mean, uh, lots of stuff. Uh, you guys also must be watching on Polygon. We, you know, just, uh, did two large acquisitions for around $250 million.

and uh you know like going deeper and deeper and deeper into payments and uh you know trying to create a new paradigm like just like Ethereum and consensus and consensus added a lot of value back to Ethereum by building like a you know wallet and a lot of peripheral infrastructure on top. So same thing like I'm uh you know converting polygon labs uh from like a pure play you know kind of uh uh you know kind of a cost center which is simply building the protocol itself because the protocol is by and large built. So now start adding doing more profitable uh activity building middleware building

providing services with which polygon labs can make money and then it becomes less dependent on the protocol and the token treasury sales to fund itself. So you know I think like you we are going to see a lot of uh protocols uh and projects going to do that. >> It's a more pragmatic way of thinking about things. It's like thinking about it more in terms of a business and a company with product market fit rather than this new tokenpowered like economic zone. I think it's a much more pragmatic way of doing things. >> Look, >> we uh we spoke with Mark and John from the Polygon team. They're doing great work and and you know we were chatting with them about the acquisition of coin me and sequence. So Sandeep today I want to talk to you about sentient uh because

by the time this airs live the sentient token will be live. tell us about sentient and tell us specifically about why the token is so important uh to sentient protocol and and the mission that you guys have with AGI. Yeah. So basically u I mean the the mission with sentient has been to empower the open-source AI development because you know by by now it is very clear with everyone that you know we don't really want these you know AI super powerful AIs to to lie in the hands of or rest in the hands of few individuals uh as well as individual organizations.

So that has been the mission of sentient right and then since then sentient has done a wide range of research like uh our uh reasoning framework called Roma became extremely viral for many weeks. it was the the most uh visited and the most trending uh repository for open-source reasoning across the globe not in crypto but like the overall AI space and uh you know multiple research papers from fingerprinting to watermarking to various different things um you know to culminating in new rips I think in November December which is the largest AI conference and where like I think we had multiple uh papers and conferences and everything on that all

of that research being you know trying to consolidate into a framework which kind of using the using the the the token incentives uh with the people and the institutions building on it creating a creating a system where all of these individual parties can come together and contribute into the open-source AI. So you know I I'll give more concrete example right because it has taken us also a long time to to to reach a place where we can say okay even though we have built a lot of open-source work but how how all of this gets combined into let's say one single product or one

single offering which can actually create a synergistic uh you know and self sustainable and and growing um you know kind of sube economy within this within the sentient ecosystem. system and so where we have like now landed with all of these research is essentially obviously the whole world is becoming the agentic world right >> and what we realize is that like the way you have seen hugging face right >> hugging face is for models the way hugging face became the go-to platform for models where everybody you know the bespoke uh engineers building the models and then you know people can test those models and all that and this is was hugging face has been the place where

you go and Check out what is happening in the model space and the world is moving from now models to agents right so essentially where because sentience reasoning framework allow agents to become more powerful more accurate less hallucination and all that >> and so now we are building the the sentient arena which is basically you can almost think of as a hugging face for agents because where we provide we have done a lot of like hackathons like werewolf based hackathons and all where people will bring in their agents >> and [clears throat] then there will be a lot of like AI based personas built over for example let's take a customer agent uh you know uh customer bot agent right uh for a for banks let's say and then

you know a lot of people are building those agents that's that is not of our interest right what we are interested is that how can you build an arena where all of these you know kind of agents are battle tested against each other against the automated AI bots with different personas. Let's say there is a persona of a uh end customer which is very angry and persona end customer who's broke and how do you deal with this let's say you know the the credit card uh calls like how does your was agent work on that and so all of these things improve the agents and then everybody can take home the learnings from there they don't need to share their user data they can anonymize the data and then the overall sentient economy the way we are thinking

right now these models are still evolving but probably you know once the data is open source. Probably this will be BSL uh kind of business license data where the large entities of the world and we have like some of the biggest names in the banking in the AI world who wants to you know want to do individual specific campaigns here and they kind of like either buy uh or stake sentient into this in order to get the rights to the data or they pay some fixed amount to get the data and all those things.

All those business models are on the cards. But that is what the sentient research team has been doing. like my role at Sentient is more of a you know high level strategic level things like how this economies economics make sense and and you know this thing while the sentient has three full-time co-founders professors who are doing all of this research and all that and I think like I I now feel at a very much you know comfortable place where I think uh this whole all of the sentient research is culminating into one single product which is connected to the token economy and then you know there is a uh virtual cycle uh which can uh which can reward all the participants of this token economy.

>> Havachi is fast and private per trading secured by ZK with Celestia [music] underneath. Go to habachi.xyz to start trading today. Not your keys, not your [music] coins. From the team that pioneered cold storage, Treasure has just released its new wallet. Guys, if you are still using hot wallets in 2025, 2026, you are missing the entire point. Secure your coins. [music] Get a treasure at treasure.io. Holiday is unified crypto payments. Give your users more wherever they are. Go to holiday.xyz to never write a smart contract again.

Absolutely. And this is something that you've had a lot of experience with at Polygon. And so you know how these token economies tend to work. You know how developing a product, getting product market fit, getting adoption for that ultimately is going to power the token. And the token vice versa is then going to power and fuel a lot of the uh adoption of the product itself. and it creates the virtuous flywheel like you said. >> Um, what when you're thinking about the AI space because you know we've seen a lot of like talent vacuums, right? We've seen AI suck a lot of talent out of crypto and crypto take a lot of talent out of Silicon Valley and AI. How do you think these things ultimately go in terms of AI onchain? like why is it so important that a lot of these AI agents

and personas like you mentioned you know the ones that are able to train banks and help with customer service and these far reach into every facet of the economy and you mentioned earlier like watermarking some of these things that blockchains are able to help with but why do you think it's so crucial that we implement blockchain as a fundamental piece that powers a lot of these AI agents >> see I have actually I've always taken the other side of this like I Don't think that apart from like you know you doing payments which is which has nothing to do with AI like payments are payments whether it's AI payments human payments >> apart from doing payments and some very niche smart contracting behavior I don't think that the AI and

crypto per se has a lot of overlap right you know crypto you have to see crypto as a horizontal which is financial rails your trustless financial rails and you have to also see like Right now AI is a sector right eventually AI will not be a sector it's the same the example is cloud computing right do you say that you know oh you know cloud computing is taking away resources from you know like if I convert you like it's taking from like you know VMware or virtualization or some other like server technology it's a server technology now healthcare transportation finance everybody uses cloud right >> same thing is going to happen with like right now that is happening with crypto and the cryptos this thing is crypto

have major implementations and applications in the financial sector only and this is what you what I'm seeing at polygon for example the major segment has become the fintex and the and and the banks and and these large uh you know institutions using uh crypto for payments and their other financial needs right because in finance you need custom business logic and all that but same way I I feel that you know AI will be one more horizontal which will use uh you know uh crypto for payment rails for like creating some sort of these economies where people can stake you know let's say people can stake scent and get access to uh you know like a license data and things like that right so but it is it is going to be one

horizontal which is used across the industries I think the same is going to happen with AI also so this the these boundaries between crypto people becoming AI people or AI people becoming crypto people will be blurred and at the end it will be the business build like poly market is an example right poly market is poly market doesn't say that we are a blockchain based you know prediction market poly market is a prediction market ki is a prediction market right and underneath rails don't matter underneath technology gets uh commoditized same is is going to happen with >> yeah and and there is strong overlap in the in the ethos right because uh despite the infrastructure and the technology underneath the mission of sentient is that agi is open-source and it doesn't fall into the hands of, you

know, uh, one single entity that's able to control and control public perception. And so I'm curious, how are you guys affecting this? What is it that Sensient is able to do? And what do you see out there as the as the biggest threat? Is there a particular company, entity, person or concept, idea that is misunderstood? What is the some single biggest threat to open-source AI right now and how how do you prevent that? Yeah, see the the single biggest threat and which has remained the same is is these some of these models becoming ultra powerful and then you know they and that risk is still there right like everybody says that is like you know I say it as you know Elon Musk says 8020 I

say 99% 99% AI is going to be benevalent only in the sense that AI will not take over human race and will have brain of its own and you know they will start ruling over human beings no but 99% AI is going to be a extremely powerful tool which businesses and countries will use to propel their strategic adv strategic advantages and the goal for for sentient and anybody else who is building in open source AI is that this access to this technology should be available for everyone all institutions all individuals all sovereign countries right so that there is a balance of power and that is the sentient mission like so you know it's not like sentient

is only going to do this like there are multip multiple institutes which are doing this open source. We work very closely with the universities. We having like two very reputed and you know like uh pretty you know fairly famous professors you know running this gives us the access to that you know especially US you know university circle uh India University circle and China university circle some of these top people are working there. So all of the like-minded people who want to create a level playing field for uh everyone in the world in different different forms.

So you know we realized that you know with this kind of funding where uh you know like the AI AI model companies getting funded $1 billion in one single day right like you know at the day zero the seed round is a billion dollars right and you know with that we we can't compete with the crypto kind of race to build a open-source model right so we realized that you know that is not possible so we initially started with some sort of fine-tuning and creating like the domain specific opensource models then making them public but Then eventually we realize that our strength is you know being is being creating and we are compounding on reasoning and and we want to make reasoning available.

just to understand reasoning. What is reasoning is that you can almost think of you know the model is like a very vague analogy but the model is just like your CPU maybe combined with your with your RAM right like all the way you can assume that's the RAM but your computer is not CPU and memory your computer also has a GPU your computer also has various different kind of circuitry with it which actually makes it a computer similarly if you want a business viable outcome I'm I'm being very careful in business viable outcome not like a fancy outcome where you know AI writes a story for you or creates a image for you. If you want a business critical outcome from this, you need to have the model and then on top of that you need to add

reasoning. you can also apply to yourself like when you think about it like if I told you that you know Robbie how will you post this video onto your channel right your your mind will immediately start thinking of steps that okay I will you know like I will finish this then there's you know I will go to my video recorder there will be this thing uh the the version of this I'll go to YouTube I'll upload that at Twitter like this is a whole planning that you did right so but then if you think like meditate over deeply you will realize that there are two parts of your thinking on this you know your one is there is a there is a end goal function where we you know that okay this is where I need to reach and you will realize there is something in the in the

background which is thinking that okay if I need to reach there what is the first step what is the second step what is the third step which is planning right so this part in the middle is actually something akin to reasoning right like you have a base computational brain you have the brain but that brain is actually thinking and our brain if you see like you know with different different regions of it they have all of these different activities different kind of contributions to it. So right now sentient is razor focused on this reasoning part of the space because we can't you know because building the core brain the CPU requires billions of dollars to build and so we can there are other people who are contributing in the open source world for that we are contributing in how do you build the reasoning frameworks which anybody can

take any model and then can apply on top of it. So right now like there is a ecosystem team which is building a co-pilot for trading and these things um you know prediction markets and all that by applying the reasoning on top of cloud code right okay so >> similar kind of things like we will provide like this you have the reasoning framework take your you know whatever model you want to use maybe an open source model apply this reasoning you get better results makes >> sense take it to enterprises and all that and that's the and then then you go to the arena like now you have this reasoning agent how do you make it better so that it can achieve 99.99% [snorts] >> uh you know accuracy for business critical stuff, >> right? And you can have it compete against other agents that are also in

the open source world that are reasoning and you can figure out which one better and then they can learn from one another. It it it makes complete sense and it it also generalizes sentient to this reasoning framework which is something that I heard you mention that you can then apply to all the different agents. Look, even if I take it that even if you have a closed source model, you can still apply the reasoning framework on it. And that is when we can start to build and expand some of these models so that they have far-reaching implications in agents in in like you said clawed code and and and the entire AI world so that we can really harness the the depths of its power. But obviously we want you know the entire stack to be open source so that we have full compatibility, full control over

that and it doesn't fall into the wrong hands. Um, Cind before we wrap, >> thank you. Thank you. Um, and very well built. You guys got to give you credit. You guys built this and it's it's very forward thinking. Looking for yield that stands [music] out? Infinifi gives you just that. Deposit stable coins and earn far more than the current savings rate. Infinify is a battle tested, reliable DeFi platform that is ready to go. Head to infinifi.xyz. Yeet is built for big moments. The most fun you'll ever have with your crypto. Go to yeet.com to yeet into your next game.

>> Trade on anything, [music] economics, weather, sports, crypto, and more. All on couchy.com 247 [music] for the world's biggest prediction market. Before we wrap up, I want to bring it full circle to the token economy. Uh because the token um for Sention is is out there and you know, you also have Polygon that's out there and so could you help bridge these two worlds in our mind? How should we think about the the full token economy? You know, you have the polygon token world, you have the sentient token world. Will these have any overlap and and if so, how can how can we think about value acrruel when we have the payment blockchain world on polygon and we have

the open source AI world on Sentient. >> Yeah. See, definitely there are overlaps as I said that the crypto is just a rails for AI. So when sentient actually launches the the like you know right now there is like a lot of tidbits of these open source reasoning and everything available but now we are launching the sentient arena sentient arena will have these staking you know things payment related thing you pay for pay for the data that is you know being generated after those arenas have concluded and all that. So that will be the revenue model in some sense for the sentient uh economy and that will be built you know using wherever polygon rails can be used they will be used uh you know they'll be used there and then sentient was already

prepping up for launching a chain built on the polygon a layer itself and as the arena and all more transactions come and when the need be then they will launch like it's not like launching the chain for the sake of launching a chain so when the transactions achieve a certain level of threshold then sentient is expected to you know launch that chain and uh the other part is like you know otherwise these economies are fairly you know acting at different levels like polygon is purely acting at the transaction provisioning block space level sentient is more of an app protocol think about like ethereum and unis swap or a right like so polygon is basically the base underlying rails and sentient is one kind of app that is being built u on top of it and whereas

sentient you know token economy is is is also like with this arena finally it becomes even more clearer that you know anybody can come and participate in these competitions and all that and whatever data gets generated either they pay to the protocol and that that exactness into the model has to be done but once the arena goes live somewhere you know February mid time frame or end of February at that time like once the whatever data gets generated either you get access to data by staking a certain amount of the send tokens and be a participant in economy or you you know pay pay a fixed price and then buy that data and buy the license for the data and uh and do that. So that is the sentient simple economy and then the those tokens and all that like whoever

are contributing open source participants and all that they get rewarded in the send tokens hopefully enough amount of data gets created that more and more enterprises want to come and stake and then get access to that data. >> Yeah. and ultimately providing this value for this entire economy this entire ecosystem you know provides a lot of value um and then there's a lot of stake uh in >> absolutely and not only in the crypto ecosystem by the way like I you know this is avail this will be available to everybody every AI company literally banks are now getting onboarded to it as we speak that there is a you know like a large probably hundred billion dollars asset in management and all that like bank which is right now testing some of these things and all that so as these things proceed

Um, you know, Sent is going to talk more about these things. >> Incredible, Sandeep. Always appreciate you joining us and and uh, you know, you you've been really busy, man. You've been sentient Polygon making moves all across the map. What's up? What's up next? What what do you >> eight to 10 years in cryp like there is nothing like that's the thing like you know with crypto I think for me having both polygon and sentient like fully stand on their feet because you know crypto like I have seen the crypto crazy cycle like with polygon and all that. What I really really yearn to build is these self- sustaining multi-deade companies. That is what inspires me to do like so I am more focused on systems thinking strategic moves and you know creating an environment where these

entities can build sustainably over time not like a flash in the pan two years your token pumps like you know this thing after that you are like nowhere to be seen. That is my biggest uh you know fear basically that oblivion of of my companies that are building in the space. I hope uh that they stand the test of time and became become a part of the fabric. >> The walk the walk away test, right? That's what Vitali calls the walk away. And uh and uh man, you're you're not going away anytime soon. Continuing to build Sandep. Absolute legend. Thank you so much for joining. Always a pleasure, man. And uh best of luck with Sentient token launch.

>> Cheers, brother. >> Thank you. Thank you. Thanks, Rob. Thanks. Bye-bye.