Seby Rubino: Mortgages You NEVER Repay? Resi Labs, AI Real Estate Appraisals, Tokenized Assets | 80

No, I think we're actually going to use AI to replace the BS that the world has been facing, such as a huge use case, right, is real estate where there's a lot of BS. And when we can change the way that people interact with real estate, when we can change the way people interact with markets, when we can change the way people interact with GitHub, right, you create these outcomes that are insanely exciting and it gives you a new hope for what is possible and what the next evolution of technology looks like.

I have this vivid vision of the future where we have billboards all around the country that say get a mortgage that you don't have to pay back. Now imagine that, right? I see unicorn opportunities that are literally just the creation of amazing offers and that is an amazing offering. A mortgage you don't have to pay back. Get live in a house that you have exposure to on the ownership level and you don't have to pay the mortgage on it. Young people can buy houses in America. What? Right. It changes the way that we interact with property as a whole.

So to close this idea, think as big as you possibly can. Think like I don't know if you have to go do shrooms, you got to go smoke a joint, you got to go meditate. I don't care what it is for you to empty your mind of all the preconceived notions that may be positive, negative, or neutral about yourself, but go and empty your mind and then think to yourself, what is the biggest impact? What is the thing that I care about the most in this world? How can I impact it? How can I change its direction for the rest of society?

Welcome to the Ventura Labs podcast. Today we're joined by Sebie of Resi, Subnet 46. He shares the subnet's pivot from real estate data collection to AI appraisals. He critiques the traditional MLS system. He envisions brokerages as tokenized platforms with resi powering appraisals, inspections, title verification. He discusses tokenized shared ownership enabling mortgages without repayments via appreciation and rents. And as always, thinking big and questioning assumptions.

Okay, Sebie of Resi, Subnet 46. This is your second appearance on the podcast. How you doing today, man?

I'm doing great, Grant. It's always good to be here. I thank you. Thank you for having me. A lot of exciting stuff to talk about, but I'll let you captain the ship and roll with it. So, yeah, thanks again, man.

All right, so your subnet, Subnet 46, has been live for a few months now. Let's just start with the history of resi subnet.

Wow. Bit Tensor is quite a nuanced place, right? You have a lot of problems to sort through. There's a lot of benefits to make sure that you grasp. And I think that when we first purchased the subnet, we were just so excited and eager to one be part of the family of BitTensor to be in this community. One two as our team had this real estate data background, we were so excited about real estate data.

As we got to speak with const and as we got to work with BT Labs and as we met with more subnet owners and also customers, right, people in the real estate industry that are dying for this innovation, we've changed our path a bit. So to give the 50,000 foot view on what that looks like, you know, we started out and for the first two months we were doing we had the thesis of let's collect all of the data that we possibly can so that we can then use that for training models in the future, right? It was going to be always the vision was multiple subnets and it's like okay well you can't really build out these other sub subnets without the right data. Let's go get that data to start. That was the wrong assumption.

Thankfully we got there quite fast and now we're on the path to allow our miners to really drive that innovation on where the data is coming from. So right now our competition has changed to an AI appraisal competition right models are submitted on hugging face. We verify the predictions of those appraisals, right? What a property is worth against properties that are already sold like in the last day, right? So, you can't overfit, but we also know the outcome that should be output and we can create a highly valuable feedback loop.

So, we get these AI models that come in, they're extremely good at predicting property value, and then we're able to now move to, you know, we're doing residential now. Then we're going to do commercial and then we're going to inspections. And as you see this start to happen, this allows for narrow intelligence narrow AI models that can do an entire real estate transaction in the real world on the onchain world whatever it is. And I have this new invigoration of excitement about bit tensor about resi and the path that we're on. So, I'm excited to share more about it, but that's really our history and where the hell we're going.

Okay. The current subnet design, as I understand it, is miners submit public models to HuggingFace and then those models are supposed to predict the prices of specific homes. Is this a correct definition of what the subnet's doing?

Yeah. So, every day we pull 500 or so properties that sold, right? So basically minor submit a model there's a cut off time and then we go pull the data so we can't cheat it either we can't manipulate the incentive mechanism and then we these are 500 properties that sold after the submission window or at least have reported that on public record. We take this data and then we say all right let's plug 500 properties into each of these models and we get a price for each property and then we basically say how close were you AI model to the price in an open manner so we can basically have a great feedback loop but when we go and sell this product to lenders or people that are eye buying or investors or these tokenization platforms we could say don't trust us just verify us, right? Verify on a huge case study essentially, like a clinical trial, per se, how accurate we really are. And, we're on our way.

And in this pivot from the database to the prediction, what was your thoughts on it? Why do you think this was the right choice?

Miners are so impressive, man. The miners on Bit Tensor are some of the smartest I would say people but for the most part you get a person who's messaging you on the discord chat but it's actually a group of people right so these are the smartest teams in the world right bit tensor is a magnet for the highly ambitious highly capable builders and you know on the founding side but also builders on the machine learning and AI engineering across the board the full stack side of things.

I think that we actually you know instead of saying hey let's go scrape a ton of data we go and tell the miners to train a model part of training that model which was always our end goal was to create these you know this intellectual property which are models that only we can productize part of that process is scraping data right so how do you kill two birds with one stone right you basically just skip to the second step and you I have to give credit to const and you know show my appreciation to someone who could be doing a million different things but you know shows up for subnet teams such as resi and says hey man like think about it in this way. If you think about it in this way, you get what you want across the board and you accelerate your road map by months and months and months and yeah, so the team that we've put together is crushing it on building out this incentive mechanism and also right we just brought in a senior product engineer who is now building out and enabling the use case of our API where lenders and tokenization platforms can actually just plug in our appraisals. Now their every their customers or their portfolio investors are going to know, okay, they're allocating capital in the right way and that is that is just exciting to me right this doesn't exist today in the real estate industry.

So, I actually just got off of a space and I and I'll stop rambling after this, but I just got off of a space about what is the actual value of AI? And it's, you know, a lot of people are thinking, oh, well, you know, we're just going to, you know, use AI to replace people. No, I think we're actually going to use AI to replace the BS that the world has been facing, such as a huge use case, right, is real estate, where there's a lot of BS. And when we can change the way that people interact with real estate, when we can change the way people interact with markets, when we can change the way people interact with GitHub, right, you create these outcomes that are insanely exciting and it gives you a new hope for what is possible and what the next evolution of technology looks like.

To touch on the BS to use your wording here, the BS in real estate. One of the things that sticks out to me immediately is the MLS system for listings. Last time we spoke, you were saying the data layer is trying to replace or I understood it as the data you're collecting is trying to replace the MLS and bring that on chain. Why pivot away from this? Why not go sub subnet and go 50/50 for each of these?

Well, that the let's start with the the macro view of this this idea of of the MLS. Let's also narrow it down to understanding that we're talking about a tokenized MLS, right? The world and the MLS, right? Went from a paperbased system. It's a book, right? And that that's why you really needed a realtor to get this book and stay on top of this book and whatever. Then it went digital, right? We saw the rise of Zillow, right? Zillow is basically just reading all of the MLS source sites APIs in the country and then showing it in a really pretty and cute way.

Then we're seeing this trans transformation to tokenization, right? So the third evolution of of the real estate industry is really the third the web three, right? The third the third evolution of technology which is just a new back end. It's a new way to to to aggregate, organize and verify information, right? So now instead of using like almost an Excel spreadsheet or a SQL database, you have this blockchain, right? This ledger of things. All right. So that that's the the macro view. go a little bit more micro.

Our thesis and the thing that we believe that the world is not yet in consensus on is that brokerages, real estate brokerages will turn into tokenization platforms. I think that the it'll actually be real estate brokerages and lending companies will converge together depending on distribution into tokenized real estate platforms. So as that happens, we are developing this tokenized MLS.

This tokenized MLS is basically a consortium, a group of people who have tokenized platforms with trusted data, right? The trust of the data on the MLS today comes from realtors, right? Realtors plug data into the MLS. You can trust it, right? These people's careers are on the line and they're incentivized to make a sale. So they're not going to do something stupid. That's okay.

But what's great is actually automating that cycle so that you don't have to pay somebody 6%. Right? These platforms they take one to 2%. That's great. That is traditional in terms of you know decreasing the cost of you know transacting in an industry. And then we take a ver like onetenth of their cost right anywhere from 0.1% to 0.5% of a transaction for verifying all of the the data and enabling these transactions to to be completed in a scalable manner right if not we're in like this like manual purgatory of tokenized transactions.

So yeah that that we're still going after the tokenized MLS wrapping up the idea we have this thesis that we are serving for. And if we know that these brokerages and these lenders have distribution, we're going to be able to serve them, right? We're going to be able to give them our AI models to make their lives easier and make their companies more profitable and then fully onboard them into this tokenized future as we do things like launch our smart contract framework which we've been in development with for some time now and then put the full stack of our appraisal, inspection, and title verification.

Right? that that's really all you need in a real estate transaction is inspection, appraisal, and title verification. Does somebody own the property? And what is it worth? That's it. So yeah, we're we're going to cozy up to these lenders and realtors to to better their life and better their service. And then as the world changes, we're going to help them adopt that change so they can actually still have a job, still create an amazing career, but do it in a more decentralized manner.

I'm not sure I 100% follow you, but this is what I distill that into is you're saying these predictions that you're having the miners make with the models will replace will will combine the inspection and appraisal parts into one and because that would factor in the inspection in the appraisal value that it gives in the prediction.

Yeah. So, let's go through that that thought process. It not only allows me to share our road map but let's go into it. So you have an appraisal incentive mechanism right what resi is today all that is is a model competition. Now how do you make a model competition more robust? You add an agentic layer to that competition. Right? That is our next build.

So, what it looks like is, hey, go and create your own models or use models that are public and available through our competition. Then make an agent and submit that agent with the proper licensing so we can productize it exclusively. And now we have a system where our appraisals as they improve, they will be able to now have a gentic function which will allow them to go deeper in the research. Narrow their focus based on the the input, right? Where is that property? Okay, this agent can go and find very niche specific data and increase their accuracy.

So, it's like how do you make appraisals better before implementing inspections is that adding model plus agent competition. The next step, you just said it was inspections. As we do inspections, then we will be able to then accelerate even more this model plus agent competition. So in the end you have an inspection report that is AI generated and can be fed into the appraisal model and that is actually this is crazy. I don't know even if people will believe me but inspections are not even used today in appraisals. Does that I mean like I'm assuming like you're not a real estate first type of guy, but it just seems logical that you take this inspection, a deep understanding of the property and then you would use that to to price a property. It's not being done today.

So, we not only add on top of the model competition for appraisals, but then we add the inspection. And the marriage of the model, agent, and appraisal or inspection is a 99% accuracy of what a home should sell for or a commercial building. And that is true innovation. That allows for institutional underwriting of loans. That allows for people to actually start taking these tokenized properties seriously. And I mean, I have this this excitement about life now that we have this clear path that nobody could thwart or damage because it now it's just a matter of time before resi powers this industry.

So, you said something there that you're like, I don't think people will believe me. I don't know if people believe me that the inspection happens after the appraisal. So, isn't that crazy? The the process is basically people agree on a price once they have an agreement, then the inspection happens and if the inspection's bad, there may be a renegotiation after the fact. Is this the current flow for how it traditionally works?

Yeah. So, let's go through that transaction process. I think it'll answer also some ancillary questions. Okay. I have a property. I'm selling it. Right? You're the buyer, right? So, you say, "Sebie, uh, cool property. I'm going to offer you 500 grand." If you're going to do this in the traditional way, we set a closing date 90 days, 120 days in the future and which sucks, right? That's a problem within itself. We set that closing date and then we go through the transaction process. In the contract, it states, hey, well, you have, you know, 14 days to do this inspection. That's really stage one.

Once you go to do the inspection, they come back with all these problems and then you come to me and you're like, "Sabbie, we got to negotiate, right?" You know, you have 30 grand worth of fixes that need to be done to this property. That is like things that we need, right? You need a new heating unit. The roof is in shambles and at least needs some repair. And you got this tree in your in your backyard that's destroying all of the piping and it's creating, you know, septic or pl or plumbing issues in the house that we have to remove. These are, you know, a lot of things, right? You know, that are are big costs. And say it's 30 grand, right? Just throwing numbers out there. I will then have to make like a a concession and say, "Okay, uh my $500,000 house for you grant is $470,000."

So now we proceed with an updated contract of $470,000 to purchase the property. Mind you, you're getting a mortgage. So now the appraiser comes in because you're not getting the appraisal. I mean, you don't care. The bank is getting the appraisal. And they're not saying, "What is this property going to sell for?" They don't care about that. They're saying, "What is this? What can we liquidate this property for in the event that this person defaults?" Right? what if you stop paying your mortgage and you get foreclosed on, what can we get rid of this liability for this asset for and that is what the appra how the appraisal works today. It's a real big problem, right? Because it doesn't account for uh it actually gives you the short end of the stick. How? I'll explain.

If you if they're saying, "Well, we're not going after what somebody will pay for on the market. We're going after what we can liquidate the property for." Uh, we appraised this property at $450,000 and uh you have to pay a down payment of 20%. So, mind you, I can only get a you can only get a loan for $450,000. And when we clip this up, I'll make sure to put all the math so that everybody can follow. right now the property that you're buying for 470 qualifies for a loan of 450. Uh of that 450 I have to bring 20%. Right? So I have to bring 90K or sorry you have to bring 90K to the closing table plus the 20 grand that the appraisal missed. So to for you to buy this $500,000 property, you have to come out of pocket 110 or $110,000.

So then you have you pay your mortgage and then you know we do the title check. We make sure that you act I actually own the property. I don't have any leans. There's no issues. I don't owe any mechanics or handymen money. And uh then we go to the closing table and that's it. So appraisals are not what we even think they are. They don't use the full resource pool and and structurally they're they're bad, right? And then it's it's a means to a mortgage. Mortgages are bad, right? I I uh to to round this idea out, I have this like vivid vision of the future where we have billboards all around the country that say get a mortgage that you don't have to pay back. Now imagine that, right? I I see unicorn opportunities that are literally just create the creation of amazing offers and that is an amazing offering. A mortgage you don't have to pay back. Get live in a house that you have exposure to on the ownership level and you don't have to pay the a mortgage on it. Young people can buy houses in America. What? Right. It changes the way that we interact with with property as a whole. So that's the the full scope of a transaction. The and also where it falls apart.

Mortgage you don't have to pay back. I'm going to circle back to that one. I'm not I'm not going to ask you immediately about that. So inherently the best appraisals from these AI models will eventually factor in the inspections. Yes. Is how you view this.

Yes. Eventually. keyword, right? First model competition, then model plus agent competition. We believe that this it's this this way we can actually reach state-of-the-art, which is 95% accuracy. This is better than Zillow. This is on par with human appraisals today. Remember, we do it instantly, right? And no one has to go there, right? A seller or a buyer can just take a video and and images and we have it. And then adding the inspection into that appraisal model and agent is going to get us to 99% accuracy.

Okay. So, this 95% number, what are you basing it off of? And how do you determine this is state-of-the-art?

Yes. Because appraisals today are used traditional for the most part by mortgage companies. These are direct lenders or institutional lenders. Their risk profiles are on this premise of a 5% inaccuracy, right? They know that the property is not going to be perfectly assessed in value. And yeah that so the 5% comes from there and our our goal of 99% and not 100% is the deviation of crazy people in a crazy hot market just coming in and and paying way over price in a bidding war. So I don't think anyone ever will hit 100%. But 95 is where the industry lives today. And that costs 500 bucks, right? And it takes like a week to two weeks. Imagine your competitive edge if you can be at 99% and you're only ever interacting with the seller, right? You don't have to send anybody out there. You don't have to send any inspector out there. You don't have to do anything. It's just a it enables a peer-to-peer exchange of real estate or a you know a lender to peeerto-peer which some people just can't get around.

And this accuracy is on the amount that banks and lenders are willing to give you for a specific property?

Well, that's how it's used today. What we're going for is the onmarket value of the property, right? This isn't the liquidatable value, right? The $450,000 from $470,000 example. But it is rather if we sold it on market through an agent, what could we get for this property? That is a more accurate proxy of value because especially when you tokenize a property and say there is an exit event, you're not selling in a distressed manner, right? you're selling it for the highest possible value, the optimal value on the market. And that is why we're taking the direction we're taking, right? Cuz that's our use case.

Okay. Now, how are mortgages not going to be paid back? How is how are these billboards going to say a mortgage you don't have to pay back?

Short answer with Resi. That's that's the innovation of Resi right there. But let's talk about how that happens. This is probably the the element I'm most excited for. So, I'm going to talk about two things. How it's possible and then how we're going to even build upon that that inevitable future. First, how is it possible in real estate? It's in it's worth noting that you can actually there's only a few ways to make money in real estate. appreciation, rent, and interest, right? You can sell the the the service of shelter or the product of shelter and you get rent. You can lend somebody money so that they can have the experience of shelter and you get interest, right? This is a mortgage. And then you also have the fact that you can buy a house and say you buy it outright and you it appreciates in value.

Right now we live in a world where you go to a bank and they're making money on interest and the owner is making money on appreciation. That's how people in the USA especially my family is part of this as a I have an uncle who's a real estate developer. This is how they they created wealth, right? You buy a property with a mortgage. The mortgage interest rate is less than the appreciating appreciation of the property and then that's the arbitrage that you are that you are taking on, right? And especially if you add rents on top of that, you now have appreciation plus income that pays off your mortgage. And oh my god, is that a wealth creation engine.

How we are enabling this is if you can tokenize a property, you can have access to liquidity, right? People want a a piece of this asset and you can actually have joint partners, right? These are like your it's like a doing a a house as a joint venture and there's a managing director, the person who lives there and then there are shareholders, right? It's like a solo venture that has a bunch of angel investors. That's how we see properties in the future and and you can buy those tokens back, right? As people sell them, you can make contracts that say, "Hey, well, this wallet can actually have the option to buy these tokens at this price on a schedule, right? Say, for example, if I get into a house and then I get a big bonus, I can buy those tokens back even if they're not available. A lot of edge cases that allow you to put more money into your house.

But you create an environment where you pay a down payment, the taxes, the utilities, and now you live in a house for the highest capital efficiency that the world has ever experienced. And that is a breakthrough. If when you could change the way that people can access property and access the appreciation of property, holy cow, it is pivotal for for the economy in my opinion and the the basically simulations that we've run.

Let's add to this grant. If we have a platform of people and you have a bunch of tokens that are on an illquid market, there's no liquidity pool. That would be very capital inefficient, right? To have a liquidity pool for every property. How do you access liquidity in this situation? How do you access leverage, right? How do you emulate the the way that capital markets interface with property?

The answer is you create a pool of capital. This is really where resi starts to explode in value. Ready for this? If we pull together, say $100 million in USDC, USDT, right? This goes back, I think we talked about in the past, resi finance, USDT and USDC in a pool. And we as Resi have the smart contract standard and the Oracle. So, we know the value of a property. And this is auto updating all the time. We're able to say, hey, well, if you're using our smart contracts and those smart contracts are validated by our AI models and agents, you can borrow against this pool or sorry, you you can access liquidity and borrow against your real estate token from this pool of liquidity.

There you can do fixed rate loans. There you can do variable rate loans depending on the pool. But now if I have $100,000, I can access $50,000. From that $50,000, I can now access $25,000. From that 25, I could do 12. From that 12, I could do six. And I can loop this up with the understanding that property is not going to collapse by 50% in value. and my 100,000 in exposure turns into $200,000 in exposure. Holy cow, I'm not going to get liquidated. I've doubled my portfolio allocation and whether it's the rent or whether it's my active income can basically pay that loan. That's insane. That is where I think that we we not only change the world but we we give people and and like the institutional side of things we get them involved because everybody wants exposure to real estate and the the if if liquidated they get real estate exposure right so it's like a bank today I'm going to give you money I'm going to give you USDC and if you get if you default I'm going to take your property same thing same thing that they're used to. Appraisal, inspection, title verification, same thing they're used to. Now, it's just happening on happening on this third database, which is crypto blockchain. And man, I don't know if you could tell, but I'm amped up about the direction that we're in every day because I've I've always wanted my my LinkedIn has always said this. I want to create technology that that enhances the human experience. It is resi in a nutshell.

Okay. So this tokenized shared ownership of assets is the way to have mortgages you don't pay back. Do you do you view these shareholders similar to an example is opening a leveraged loan on crypto. How they pay a funding rate to keep the exposure to that asset. Is that how the mortgage still gets paid back by these shareholders with the equivalent of a funding rate when you're borrowing against the property asset?

Yes. In the example I just explained where you're accessing liquidity from this pool of capital. Yes, 100%. The funding rate is basically just the interest rate on capital, right? So if there can be fixed rate or there could be variable rate in this poolled resi finance future. The other side of it goes back to what I said before, which was you can make money in three ways. One of those ways is appreciation. So now you have this decentralized group of investors, whether they're family offices, high net worth individuals or low, middle, high high class income earners. And instead of them making interest on these property tokens that they own, they can at the in the worst case in the minimum case make appreciation so that when that property is revalued through oracle and the prop price increases, boom, they can make money through there on the auction. Kind of like how NFTs are sold, right? There's like an auction. Hey, I would put a price on my assets and what I'm willing to sell it for.

But if you look at these tokenization platforms, we actually just partnered with HoneyShares and they have tokenized a ton of Airbnb properties, right? We're also doing the same type of thing with estate protocol.com, right? This is on arbitron and most of the tokenization happening today is with Airbnb and that provides a yield. So not only do you get appreciation of these assets, this this property token, but you also get the rental yield. So what's the ideal financial strategy here? Let me go and and these yields are like 10 to 20%. Right? So keep that in mind.

The ideal financial strategy in this on these platforms and why I'm so excited to just get this out the door is I'm going to buy a ton of these Airbnb rental token tokens. Then I'm going to borrow against them to get say 2x exposure and then with that rental income I'm just going to pay the interest rate and increase my size and keep doing that. And you can actually do this on a pro programmable level at a you know 50% loan to value meaning the price would have to drop 50% before I get liquidated which has never happened in the history of real estate right even in 2008 prices dropped like 30% and that was like oh my god the world's ending so yeah it it's insane it's insane that the somebody gets access to housing investors get access access to liquidity. And in in the ideal case of commercial and Airbnb type properties or just rental properties for that matter, you have these insane financial strategies that can create wealth on their own.

Is this not just bringing REITs on chain?

Uh no. So a REIT is basically just like a a consortium of a bunch of properties that a centralized entity invests in. So they go and they find these deals. There there are some similarities. You get exposure to real estate. You're doing it in an easy way, right? Cuz you just buy REIT token and then they're taking that that capital and then they're buying property and it's almost like a a DAT, right? It's like a rat, you know, a REIT. It's a a real estate treasure. asset treasury. So yeah, there's a lot of similarities but at the in the end we're actually giving more power to the people to decide what they want to invest in and then we're also giving the power to buyers and sellers to access liquidity on for their property.

Think about it as a seller of a commercial property. Okay, maybe I have a a loan on this deal, but what if I want to access additional liquidity so we can do another deal, right? An opportunity comes up. Well, I can basically sell a part of my deal that I have today. People can buy that, say 10% of the value. Now, I take that 10% of the value and I go buy another deal, right? So this can either be the entire financing stack for a deal a property or it can be just a piece of the financing stack of a property. So so accessibility and composability are ways that real estate has not or luxuries real estate has not had access to and I'm excited to to give that to the world.

This idea of tokenized asset ownership with the managing partner taking the physical usage of the thing. How does this scale to other things say a car or other assets?

Yeah. Well, if you let's focus on the question that you're asking before I go into like the how our tech actually applies to that too. A car is a depreciating asset, right? The way there's a couple ways that this actually works with vehicles. I actually kind of want to do this myself. I'm a bit frugal when it comes to personal spending. But there are groups, and you see this all the time in Miami, that will go and buy a Lamborghini, right? And they'll split it up. They'll buy like like 10 super cars and then they'll split it up against, you know, 250 people and sell a membership to access those cars, right? So, any day of the week you you go and you pick a car out or whatever and you can do it a certain time certain amount of times per month. Same with private private airplanes, right? You can you can have a jet that you co-own with a ton of people where you pay for a seat or you pay for the whole thing but in a fractionalized way. How do you how do you how do you do that without you know tokenization is it's really hard right it's a lot of coordination just on organizing everything. So so that is how it applies to say logistics or transportation because when you ask about cars you're also asking about planes.

Then how does our technology