Secure and Collective Decision Making with Alex Bulkin

Listen to the podcast here.

Check out Alex's work on Adapt.

Nathan: So today with us, we have Alex Bulkin, who's an old friend of mine in this blockchain space, and I'm very glad to be talking with this particular morning. Alex spent this period between August, 2016 and November, 2019 at CoinFund in New York researching and helping deploy significant capital into blockchain based and peer-to-peer infrastructure related assets.

During this time, he has had multiple speaker engagements at high profile events, such as the 2017 token summit and has been invited internationally to mentor organizations and entrepreneurs on the topic of monetary incentives, digital asset investing, and crypto economics. Previously, he spent 14 years as an engineer at Goldman Sachs in the area of high performance financial technology with special interest in reducing human error and other holistic approaches to systems reliability.

What's up, Alex, thanks so much for your time.

Alex: Hey, Nathan, good to be here. Thank you for inviting me.

Nathan: Sure. So you and I go back to 2017, and one of the things I wanted to first ask you was what has been the most striking thing that you didn't expect to learn when you started all this?

Alex: Oh, Whoa. And without preparation to ask this?  Too late. Let's think about it. What is the most striking thing I learned?  Here is one: I learned that people's irrationality can last for many generations.

But when I first saw during the ICO bubble the big run-up of late 2017, it was clear that things are not going to end very well. And then there was like two years of pretty much nothing happening as far as like markets and now it's back and better than ever and all over again. And I just think that it's quite striking that markets can stay this way for so long.  That was probably a big one.

Nathan: Recently. So I've been following you on Twitter for a while. Recently, you went into a rant about governance tokens. How do you think governance tokens fits into the irrationality conversation?

Alex: The whole thing about crypto is we are dead set on making every mistake that people already know not to make in more traditional spaces that includes over leveraging mis-pricing of tail-end risks. And then the governance is another one.

Now governance is a difficult topic because. One is definitely needed. But the other one is how to do a try. It is very tricky. There's there's a handbook of computational social choice, it's a 600 page PDF document that's floating on the internet. If you're issuing a governance token and you haven't read an introduction to that book you're being , how can I put it, you're wrong. It's not, let's not use a stronger term than that.  Because basically voting power, voting dynamics , and then the relationship between the voting and the economic power is very tricky and very easy to take to a place that creates a disaster.

Can I tell you a Russian joke on that? It basically summarizes what I think about governance tokens, and it's very funny. So a Russian traffic cop wins a million dollars in the lottery. And he's asked, so are you going to retire now? And he says, "oh no, I'm just going to buy an intersection in the center of Moscow and the work to my own pleasure." Yeah. So that's my summary of governance tokens.

Nathan: There are a couple of governance tokens out there, some of which are. Masquerading as w both, right? Like you pointed out Maker isn't exactly a governance token.

Alex: Yeah. So those are a lot better. The problem is the space has to teeter in a very narrow and hard to define gray zone. And so for a variety of reasons, including kind of regulatory reasons, people are understandably averse to creating a profit generating token nowadays, and rightly. But then there seems to be a kind of a push to say, Oh, but we don't have to create a profit generating token, we can just create a governance token and use it the same way. And if the markets were rational, I think that'd be an insane idea. But as things stand in the short term it's working, I think, in the long-term it's going to be a disaster. Every project that has done that will probably go to hell, once the initial hype subsides.

Nathan: Do you feel like there is a project that you have seen like that actually has read the handbook of computational social choice?

Alex: Unfortunately so since fall last year, I wasn't following that closely prior to that, no not really. Prior to it, like just the fact that we're still dead set on using voting as the decision making mechanism. And of course it's the simplest one, so it's understandable why we do that. But voting is its own big topic but I haven't seen, I haven't really looked recently. I think my answer to that question would be incomplete.

Nathan: You feel like even  Tezos, which has voting taken into its like main core protocol, do you think that implements any of what you're talking about?

Alex: So yeah, I think again, I haven't been following Tezos recently and I don't know what's been happening there. Here's the problem right? Tezos is a governance protocol can be changed by Tezos' governance. This is one complexity around governance is that you can never use the subsystem A to change subsystem A right? In practice if you look at for example, how the constitutional legitimacy, is given to the US Senate than the US courts, right? That the legitimacy of the governance organ can never be decided upon by the same governance organ. And that's what Tezos is doing.  And that's over the long period of time will eventually be exploited.

Nathan: Is that due to no division of power?

Alex: Yes. So the no division of powers is a huge thing in crypto.  We just don't seem to be talking about it and understanding it, and division of power is essential. My own design that never. I penciled it into my notebook about two years ago as a kind of the next experiment to try was one with a kind of two-partite governance model, where one part is the economic stakeholders, and the other part is non-economic state stakeholders. Now the non-economic stakeholders, unfortunately, is a difficult topic in crypto because anybody could be an economic stakeholder. So I'm not sure if that's even feasible , but in principle you nominate people to govern the network who do not have an economic interest.

And that makes up one of the governance kind of bodies and then you pay them to make decisions and they can be fired by the economic stakeholders. You can get as complicated as you want a point. My point is not that I know that solution, my point is that there are things to try.

Nathan: Where would you like to see those implemented directly? How can you prove that a non-economic stakeholder has no economic interests.

Alex: Yeah. That's the point, I don't think you can. So there you go. What I just said doesn't work.

Nathan: Lol

Alex: But in truth, like in practice that's been the case in traditional industries and government systems where people make a public statement to that effect and if they are ever proven wrong, they go to jail or get fired. So I don't see why that model would work, except people are so dead set about kind of permissionless and privacy that in this case, I think it works against the space where this model that I think is perfectly reasonable will probably not be particularly popular.

Nathan: The only other governance system I've seen at least even implemented is DAOStack, but DAOStack's whole system is like maybe adopted by a couple of ETH heads. Have you looked into DAOStack at all?

Alex: A long time ago and barely. I've, I'm familiar with the earlier work and I met Matan Field a couple of times, but I never really dug into DAOStack enough to actually evaluate it. I don't think, I don't think any of those have really taken off like DAO Stack and Aragon. I don't know if Colony ever came out. There were a number of very early projects trying to do this, and I don't know if they are anywhere. And I think a lot of the problems there is that by trying to be general, they're missing the market is my guess. Again, not a particularly strong opinion.

Nathan: It looks like Polka Dot has put in some governance that is putting together your council idea.

Alex: Oh, I didn't know that. Again, like I haven't been watching the past few months.

Nathan: Yeah they're they're trying to implement that, although I'm not sure exactly what's going on with DOT either. Most of it's all just futures at the moment. And then it looks like Staker DAO is putting some future of decentralized governance together, something similar with the elected council.

Alex: If that's the case then what that means is that in recent months these ideas that have filtered into the community. The other really interesting topics related to governance of course, is AI. Too futuristic to go into here, but definitely something about.

Nathan: So for the AI conversation , what I'd like to mostly discuss with you, which is which is the work that you're doing on Adapt. And also as it relates to verifiability of content  Do you want to go into this rabbit hole together?

Alex: Yeah, that's, what's one of my favorites. Should I open it or should you, or do you want to go down it?

Nathan: Open it up!

Alex: Okay.  Wow. Okay. Standing in front of the precipice looking down.  Basically. Here's the question, right? Socially speaking, and politically speaking, I think we all agree that humans are no longer to be entrusted with making decisions on behalf of other humans. We see  the really big political kind of deterioration around the world. I not going to go into reasons for why that is or why it came about, it's just that I think that the idea of electing a human to govern other humans is at this point clearly has to be updated.

Now there's many different ways to go but let's consider one. So the one model that I personally like a lot is collective decision-making without the intermediary, without the president, without the Senate, without the judge.

Now how would that work in practice? In practice, it would be a data gathering collation and distribution system that would collect opinions, collect input from everybody in the group that is trying to make a sort of governance decision about itself. It would employ some process which could, potentially include artificial intelligence components or there's some very interesting work on the internet about automatically selecting experts in certain fields to provide input. And it might include pieces related to voting. So for example, one interesting idea is to modify the proposal for the new law until, over 95 people agree or something like that, or over 97 people agreed. So getting as close to consensus as possible.  So let's look at that system from the outside, it would be a data gathering and exchange black box that includes a variety of complex algorithms inside of it. We don't know what the algorithms are we still are to come up with them. But the point is that today, none of this can possibly work. And the reason why it can't even begin to work is that we cannot rely on a system like that to stay impartial because such systems today get operated by people get operated in the data centers, the data is easily accessible, easily changeable algorithms that are easily influenced. So because a computer based decision-making system is so easily influenced by programmers and operators, we can't possibly even begin to rely on it for any kind of governance. And that's a clear point, even before you start asking yourself whether the algorithms are useful or not. So now how would you solve this problem? In order to solve this problem, you have to have a way to trust to perform trusted computation at scale. And by trusted computation here, I mean how can we run an algorithm where we know that no inputs have been altered maliciously and that the algorithm has not been altered maliciously? And that requires data and algorithm protection at a scale that we've never dealt with before. Does that make sense so far? So a lot of the ways that we can do that are the learnings we extract from blockchain because in blockchain, I was talking to my former partner and Jake Brukhman yesterday. He's lists 13 benefits of blockchain, I've listed in one of my blog posts. And I'm sure I'm sure he says because he's an amazingly deep thinker in this space, I'm sure he listed them all really insightfully.

But one of them of course, is trustless compute.  And so we know from blockchain, how to make trustless compute and we use consensus to to do it. Here's, there's two problems, right? First of all, consensus is expensive. We know that there's some efforts to make it cheaper. We are yet to see if they're successful or not. Your favorite project Avalanche is one of them. I am very curious how that project comes out because I very much respect Emin Gün Sirer and others in his team.  But to my point is that actually, when it comes to trustless, compute, you can sometimes do simpler things.

Sometimes you don't event require consensus to have trustless compute; secure enclaves is a very prominent example of how trustless compute can be possible much cheaper. Of course the security guarantees of enclaves and blockchain are very different and that's granted. And the second thing about trustless computed scale, is of course privacy, because not only do we have to ensure that the data doesn't get altered, we also have to ensure that the data doesn't get revealed to anybody other than the algorithm that needs the data's input.

And the privacy is a little bit harder with blockchain. Again, lots of new efforts there, but in terms of fronting arbitrary algorithms privately I think the only solutions out there that have a near term possibility of being built are those that actually use secure enclaves in the context of blockchain to provide privacy.

So that's the rabbit hole. The rabbit hole is how does governance, AI, blockchain, and the thing I'm working on, all come together to build a brand new generation of systems that are completely impartial in the way they work with and with the guarantee.

Nathan: To give it context. So the Twitter hack happened. There's "God mode" and Twitter on their end. So they can go in and determine what is and not published. And that Twitter hack happened in July. And then with the release of GPT-3 it's arguable that in the near term, we won't be able to determine if content was created by a human or not. So what's the next step in the rabbit hole?

Alex: That kind of lays out the landscape for me. For me looking at this inspires me to build systems for this purpose, right? For the purpose of data verifiability, data integrity, data privacy, along with a sort of general purpose, algorithmic components being protected together, both data and algorithmic components with a really high degree of security.

We haven't really had to look at this problem in this form before. It's not a very well understood problem because the systems that we used to build up until we entered the blockchain age were all really only relying on kind of human promise to in lieu of the security guarantee, but it is now becoming completely clear that's just not enough.  It's not enough for pretty much anything actually we're not talking governance systems alone, right? We are entering an age where extra efficiency of transactions between companies can only be achieved when the company started sharing data with each other.

Efficiency is the game, but of course they can't really share data with each other because it's just such a sensitive zone related to confidentiality that companies wouldn't want to do it. And and across the board, we see either a need to build such systems or in blockchain, we're actually building such systems, just using a very ad hoc approach.

Nathan: What you've said on Twitter is that it will be hard for us to verify content is written by a human, unless we have some sort of content hashing mechanism.

Alex: I think that actually that even the bigger issue that is identity. So how do you create a network of human identity, such that it combines so to say, the un-combinable, right? What is non-combinable here? On one end on one hand, it needs to have very significant privacy guarantees because people who are against, global government identity are rightfully against it, because it's very easy to abuse when it's built naively.  And then the second one is basically it should be sybil proof. And those two things are incredibly hard to combine.

Now, there is the work being done on digital human identity that I think will eventually get to a point where we could have a global identity and it will be actually safe and it will allow us to sign our social media where, and it will ensure that the Russian propaganda bot cannot get one. And when those things happen, I think the information manipulation landscape will shift a little bit, we will be able to breathe a little easier reading information on the internet. But basically, yes, that's the idea. It's like anything that generates content nowadays be it a human being or a digital camera it's desirable if it had an identity to prove that yes, I am a digital camera and not a deep fake yes, I am a human being, not a bot. And having that proof really has to rely on the new generation of secure software of the kind I'm describing.

Nathan: So you're, you've said that Adapt is being built as a system for rapid deployment of secure enterprise and mobile data management components. You say its primary goal is to offer companies a possibility of a deep security upgrade, basically in a stack by rewriting small critical components with a new framework. Do you want to go that a little?

Alex: Yeah, I'd  love to talk about Adapt. I don't very much because I'm basically spending most of my time writing code. But to introduce, Adapt a little bit, it's my answer to what a software development stack would look like for a system like that.

The security goal is to protect data and algorithms in a much stronger ways than we've been doing so far, to do it cheaply, to use a variety of environments where the component can run and to serve as a kind of a building toolkit for decentralized and operationalist systems.

And I'm talking I'm looking at the enterprise space, not at crypto very much right now, because I think frankly, crypto is a bit of a mess. People, I think look to economic properties of a project ahead of technical properties which is understandable, but unfortunately what that means is most of the thinking crypto is incredibly short-term.  And when I say that I'm building something that's for decentralized system, but it's not a blockchain, the people are like are you DeFi are you not DeFi, are you layer or layer two. I'm like look, here's what I got. It's like off to the side a little bit in terms of what it's doing, but it's in the same general ethos of building, the trusted computation systems.  But that really doesn't evoke a response. But I found that in the enterprise world, their response is actually much more warm and it's warm because people in business actually are having trouble sleeping today with all the data that they're storing, sometimes by needs sometimes by choice, it's a little bit hard for them to ascertain for themselves that data is not gonna be stolen. And that data is not going to be abused or like in  Twitter's case, that data is not going to be accessed by some, teenager in Florida. Frankly, I expected them to be Russian, but okay, I was wrong.

And intuitively business people understand that when security companies come to them and say, yeah, we can verify the security of your system, that the security companies are actually not increasing significantly the amount of protection that the systems have, and the reason why is because the traditional model of security is not operator-less, it basically means that when you build your systems, there's still some employee that you have inside of your company that has access. And because of that and that doesn't fall into the purview of the security consultants is helping you make your decision, they can introduce you to a fate of your employees, but we all know how bad people are at using these types of security-sensitive kind of work flows. And that kind of means that they don't really know if something bad is going to happen, and we almost no way to guarantee that it won't. So when I come to them and say, "Hey by the way, the system that I will help you build, isn't going to have an operator, isn't going to have a single person who's able to look at the data. They're like, "Oh, That makes so much sense."

Nathan: So what you're talking about instead of a hardware, like an Intel SGX, you're talking about a software version.

Alex: No, it would

Nathan: be using Intel SGX if it

Alex: were available as a means of actually implementing the protections, but the point is that wherever it is, it will use the best protections available. So I think the main advancement there is not from no SGX to SGX, it's from a system which has an operator catchall that says and if anything else happens, an operator can come and change this to a system that is designed for an operator-less operations, so say. It's a system that says if somebody needs to do something else, they actually must, they can't avoid asking a number of permissions and getting themselves allowed by a number of people to access whatever component that they need to access in this particular case. So it's not operator-less in that things can't be changed, it's that the system imposes such a strong isolation that you can't just dig into it, you have to go through, its predefined workflow in order to get to it. Did you see what I mean?

Are you using snarks in

Nathan: some way, how are you isolating these?

Again, Intel SGX is

Alex: the easiest. So imagine a process that's running inside of a secure enclave that has a very thin API in a Ethereum or in blockchain terms, you can think of that as a list of transactions it allows you to run some transactions you can just throw out and then some transactions you have to get authorized to run, and there might be a bunch of kind of multisig models to authorize people to do different things. And one of the transactions is, "Hey, can you run this code?" So instead of having a blockchain that allows you to load any smart contract you want, which I think creates a tremendously large surface of attack which I think is one of the weaker aspects of the Ethereum system, you have a very narrow surface of attack and in order to run more code inside of that container to, debug or fix the problem, or  upgrade or institute a new process or whatever, you have to go to the same transaction API. You would have to say, "Oh, I want to add this new module to the system as an operator, but I can't just do it, I have to say, Oh, and by the way, these five guys have to agree to let me do that. So I will submit a proposal to introduce a module and they can examine it and they can approve it"

And

Nathan: now I'm going to loop back, and do the callback. How does, what you just suggested fit into your governance model? As proposed in the handbook of computational social choice?

Alex: So the data protection. So if you have a really complex system with a lot of data coming in and going out and with a lot of complexity of algorithms running inside of it, and a lot of needs to upgrade, update, and evolve, and at the same time it's extremely sensitive and should remain neutral, then the only process which allows you to do this is a process where the operator is an extremely limited role. So if I am to build a voting system using Adapt, I would build a voting system that would keep all the data private and it would never be possible for an operator to go inside and look because an operator will say "I want to do this," and then it will have to basically, it would have to basically get a vote of of all the participants to allow the operator to do that.

Nathan: But in this case, you're suggesting that all all parties are known or in a Federation of some sort.

Alex: In some sense, I think that's very likely to be the case. It doesn't have to be, but I think that we in crypto really underestimate federated models. And there's a reason why, it's because, we operate on the kind of the permissionless open model, but in truth there's a ton of ton of zones where that's not even necessary, like in the enterprise systems open and permissionless is a taboo and rightly, so federated operation is an important thing that I'm also trying to cover here.

Nathan: How

Alex: does this

Nathan: relate to your conversation about identity?

Alex: Same thing your identity has to be stored in exactly that type of system.

Nathan: So where in this case is Adapt, storing identity?

Alex: It depends on what identity information looks like, which we don't know yet, but Adapt is not just a one node kind of database that collects data and runs algorithms, Adapt is actually a network of such containers. So you can have the interaction. So just simplifying, if this was like a KYC identity network let's say, then there would be a network of KYC providers having their keys for the inside of their Adapt containers. And you have your own adapt container that stores your private key and your identity information signed by the KYC provider.

And then as you go and use your identity for anything voting, buying security tokens, or whatever, you would submit the part of your identity's, Merkle tree as you desire. And it will come signed by the KYC provider. And that would be submitted to an Adapt container of the system you're talking to. So for example, if it's a security tokens network then it would be you submitting, it would be your adapt container talking to whatever component is hosting the security token.

Nathan: Where does the token communicate with Adapt?

Alex: Two possibilities, right? Either the token is hosted by, Ethereum or I dunno, Polka Dot, in which case your Adapt identity container would be sending a blockchain transaction. Or your security token is running in an adapt container as a kind of a simplified one node or mirrored nodes blockchain system. An easy example here would be your open banking. So if your bank allows you to buy stock and gives you an API point to buy stock then your Adapt container could be talking to an Adapt container on the side of the bank that contains your, let's say stock portfolio.

Nathan: Okay. So the bank has to spin up like a Docker image of Adapt?

Alex: So for

Nathan: example, Adapt is actually

Alex: like Docker. Docker is a great kind of comparison here. Adapt is a containerized system. And  yeah, you'll basically spin up a node of Adapt. You load it up with software  that does what you want it to do, and then users would use their local Adapt containers that contain their keys and identity data to talk to the backhand Adapt container that contains whatever business logic it is contains, in the case of a bank, bank accounts.

I got it.

Nathan: And you said that the enterprises that you've been speaking to seem to have a better grasp of this than the blockchain freaks...

Hey, I didn't

Alex: use the word freak. That's the one where they did see

Nathan: you

The ETH heads. Where do you see enterprise fitting this in right now?

Alex: Oh, a ton of places. I think open banking, GDPR, enterprise data storage and the price, data sharing enterprise peer to peer trading by the way is a big thing, so we can talk about that. The same reason why we love DeFi so much is why enterprise also wants to DeFi, but they  don't want Ethereum DeFi because the tail end risks on that are very significant, and compliance is really important.  So what enterprise ones is DeFi with, backdoors for regulators, complain as much as you want about that, as well as certain very clear kind of governance components.  So if enterprise wants DeFi, then what enterprise is going to have to build is some kind of a peer to peer asset exchange DEX, if you will. Why do you want DEXes? Well because there's reasons why we don't like exchanges and it's not just that they're expensive, they might not even be. It's that exchanges are really slow to move. And that's that I think is the biggest reason why enterprise will build their own DEXes before too long. It's because, if you have a stock exchange it's not very likely to trade, with other kind of instruments, then there's such a fragmentation in the kind of a centralized exchange space in traditional finance.

Think about the kinds of assets you want to trade. It's not just stock options, futures, and commodities, it could be real estate shares, it could be IP shares. It could be carbon credits. It could be service rights, airline miles, all kinds of stuff shipment capacity trading could be a thing.

And if you really attempt to do that, you're going to have the kind of composability and the kind of a fast implementation write once, run, everywhere model that we have in DeFi. And so that's where enterprise, I think can make a huge use of a system like Adapt because enterprise doesn't need the blockchain for this.

This is really the key insight here. Enterprise does not need blockchain because enterprise does not need permissionless. Enterprise does not need open-end enterprise does not need a hundred percent available. So the availability guarantee can come on the legal paper, just fine.

Nathan: Okay, so you have you that, that makes a lot of sense. That's why a lot of projects, especially like Avalanche has been developing the kind of build your own DEX. They've been spinning up,   they have a couple of grants open for it. Have you spoken specifically to enterprises and they're like, yeah, if you build me X, I'd like to use it for Y?

Alex: In some cases, I'm not going to reveal all the details. These are early stage conversations and I don't want to name the companies.  But in general as compared to even six months ago, lately I've been really getting not people jumping on board, but people are like, "Oh, I see what you're saying, I think that's makes a lot of sense. Whereas six months ago, people are like, I don't even understand what this is and why. And if you think about it, I think the reason for that is not really blockchain per se. I don't really call myself like a crypto graduate when I talked to enterprise people, I think the reason for that is that hacks are coming so quickly and have such devastating effects that the consciousness of what the next 50 years in the data security would have to look like is increasing very steeply right now.

Nathan: Is there something else you'd like to cover with Adapt? Otherwise I have a couple more questions.

Alex: No, this is great. No, I think that's pretty much covered it.

Nathan: Want to talk about so one of the the points that Jake was making about the 13 blockchain commandments is, do you want to put your opinion on value creation from blockchain outside of a trustless execution? Do you want to talk anything about Mattereum?

Alex: Mattereum is amazing, I adore Vinay. So you know how it is, right? If you are looking one step in the future, you're a genius. If you're looking two steps in the future, you're an idiot. The Vinay is looking maybe three, four steps in the future. And so being early to market is very difficult. It's unfortunately, one of the reasons why I really dislike the VC space is that essentially the incentives that investors have makes them a little bit extractive in their thinking because they don't put the money into two steps in the future thinkers, because rightfully they determined that two steps in the future thinkers are actually early to market.

And so it's very hard for projects like that to survive. But I think there ought to be some way, and I still don't know what it would be, but I really hope we find some way to incentivize people to support two steps in the future efforts. So Mattereum is clearly one of those Mattereum is basically saying the only way we can gain the efficiency that we need to gain as the global societies is by tracking physical objects in a better way because we're wasting too much resources, making too many things and throwing things out just because we can't find a where the thing is. We can increase our efficiency, 30 to 50%, if we could only incentivize people to build sturdy equipment and then incentivize other people to use it for many years and transfer in between themselves rather than buying a new one.

And as far as like how it connects to blockchain while it's the same reason that I'm building Adapt is because data protection in a system like that. And in the case of Vinay, I think openness and permissionless is a little bit important. These things matter a lot for large scale  efficiency building systems, because you really have to prevent extractive behaviors. And in order to do that, the system has to be very well-protected.

Yeah. I've followed Vinay's work for a while as well as you, and the ability to track carbon credits is critical especially as it relates to climate change or, how our phone was made under what slave conditions  your batteries were made or whatever.

Yeah. Yeah. I think the solution to how overwhelmed we are with the amount of data we have to digest today is to create even more data.  And I'm joking and not because because I think if you could make data, I don't want to use the word verifiable because there's no way to make data verifiable, but at least traceable, if you could make data traceable and well protected from being mis-trained mislabeled and lost is how you get to that level of efficiencies.

Nathan: He's been railing against the direct mislabeling of capitalist consumer products, which is where we get all of these problems from.

Alex: Correct, exactly. It's basically, both me and Vinay are in the same business. We're in the business off removing the excuses which people use to do the wrong thing.

Really? That's all we're doing for me.  If you come to a company, if you go to Facebook and you say Facebook, why aren't you protecting your data? Facebook is going to shrug and say, "why how is it even possible to protect the data in such a strong way?

If Adapt out there, you can point it and say, "Hey, here, this is an operator-less system that allows you to make a technological guarantee of data use to your users, why aren't you using it? In Vinay's case, it's the same exact thing. It's "why are you mislabeling your product? Cause my, the material I got was mislabeled" right? If you remove the excuses, things will get much better. Same for governance, same for opinion polling, for example,  why is our opinion polling so bad? Why do we get such a bad vibe from social media?

It's because data's not traceable because there's too many excuses that you can say, wow, there's no way I can make it any better. Guess what, they will be a way to make it better and we better make it fucking better.

Nathan: Otherwise we have very serious societal issues at hand.

Alex: Exactly.

Nathan: Would you like  to go over a little bit about where you are with Adapt and what your next steps are, where people can learn more about it?

Alex: I'm pretty early. I'm working on the next version prototype of the system right now. Where I am is I basically pre-seed I can show you a prototype of the compiler. I can show you the prototype of the container. I still have to package it, I still have to make it secure. I still have to give it some decent user interface. I still have to port it to a couple of different environments, et cetera, et cetera. This is a ton of work. I'm currently bootstrapping I've received virtually no money from anybody yet, because I basically think I'd rather have something to show people and get to a point where there are clear signs from the market that this is the thing before I start raising money. I'm very responsible about raising money.

A lot of people are saying why don't you do an ICO? And truth is I could cause there's a couple of cases I'm not going to go into for why a token would be useful, but I'm just not going to do it. I think ICO or kind of yield farming incentives are so short-term oriented and I want this project to survive, for 30 years, and so I'm not going to do it. But yeah, I think in the fall I'll probably do a round assuming things go well.

Nathan: Which of the most important building blocks from blockchain, do you find the most useful? It sounds like we've covered consensus. It sounds like we've covered trustless execution. Do you feel like zk-snarks fall in this category as well?

Alex: Zk-snarks are a specific answer to a very general problem of zero knowledge computation and zero knowledge computation is announcer to a specific problem of trusted computation.

So in some sense when you name a specific technology like that, you're probably going a little bit too low level on the space.  In general, when it comes to zero knowledge proofs, homomorphic encryption, and the all kinds of really advanced cryptography stuff out there, I think that's the future, frankly, but I think we are quite far away from where these things will become like a commodity on every phone. Where a few apps on your phone use the zk-snarks, every day. I think that time is coming. I think between a secure enclaves and advanced cryptography, you have an amazingly promising tool sets for building really powerful systems. But I think it will all come together in the next five to 10 years, not necessarily right now. So for example, let me give you a curious example of something that I've been waiting to see, and frankly haven't yet is secured GPUs. So the question here is that basically, if you want it to build AI models that can not be maliciously affected.

It's not enough for you to just make sure that they get neutral training data. You also have to compute them in such a way that you know, that the computation is correct. And those computations happen in GPUs usually. And I don't know if secure GPUs exists. It's basically the same idea as a secure enclave, but for matrix computations for vectorized computations where, the computation is performed on an encrypted data and there's a protocol where they can talk to each other, but the GPU also uses encrypted data as input and output. Somebody who is more in that space can, I would love to be corrected on that. But I don't think it's there yet. Then it's really essential.

Nathan: Coda does incentivized snark generation. They have a snark bid process and also low cost storage and validation for mobile phones.

Alex: So that's similar. That's similar. I haven't looked at CODA recently. I've looked at the CODA about a year and a half ago when they were first coming out. I think back then, I don't know what Coda is now even, but back then the whole idea of succinct blockchains I think was fascinating.

But the trick there is, of course is how do you make them both succinct, private, and smart? And so I know that I've seen some proposals for smart contracts inside succinct blockchains, but I don't know if any of them actually came out or succeeded building anything. But I think a succinct blockchain in the form where they were coming out in 2017 and 18, where all they were trying to be as a kind of a replacement for Bitcoin, I don't think that's really where we are today. I don't think we need that beyond like a very niche market. The other problem with privacy currencies, this is a bit of a tangent is the difficulty of creating liquidity for them because centralized exchanges don't want to touch them for regulatory reasons.

And they're very hard to connect to decentralized exchanges because they're so different. So an atomic swap between Ethereum and Coda. for example, I don't think that it's ever going to happen.

Unless Andreessen deems it so on Nathan: Coinbase.

Alex: Yeah, exactly. But  it's difficult. It's this an actually, I dunno if I might interject there is a way to create liquidity for these guys actually through, SGX-based peer-to-peer atomic swaps. And they're not really atomic in the sense of a pure kind of blockchain-to-blockchain atomic swaps, but because  if you can run a swap inside of a secure enclave, I think you can trade, CODA for Ethereum in a decentralized way. This was one of the use cases for Adapt that I was thinking about, I might still do it. I know there's some people looking at SGX based peer-to-peer trading outside outside of my zone of knowledge, but I've heard about it.

Nathan: I think what we should do in the next month after a lot of these like Avalanche is launching soon CODA is launching soon. We should double back around  and see what the current state is.

Alex: Yeah. That would be very interesting.

Nathan: Is there anything else you'd like to cover?

Alex: Personally,  think I'm. I'm good. I know you want to talk about Urbit.

Urbit seems likeNathan: a red pill. OS open source operating system based on planets and stars. I haven't dug too deep into that.

I don't know if I can do Urbit justiceAlex: by introducing it. But the way that I understand it, it's actually not dissimilar from Adapt in that it's, it gives you a private well-defined data container to use and to run applications on.  And it's genius on many levels, but it takes a while to get from zero to 60 on understanding what it is and why it's there. And that's why it's it's kinda like a fringe project. That's been, kinda, moving along at an amazingly steady pace in the past five years, but it actually started earlier than that.

And the ethos of Urbit is very similar but applied it to a slightly different space, similar to Adapt I mean. It's applied to the space of social media and social interaction where your Urbit container in that case is your private platform on which things run.

So in some sense, it's like a cell phone in that you can load up applications from a kind of an app store and run them, but it's a server-side component. if the ideal place to run your Urbit is on a private server, I think, what are we going to see in the near future is private servers as a consumer hardware product.

So if you think about your, let's say Wifi router, that already has a pretty powerful CPU and it's already has a hard disk. And so all it needs to have is a basic operating system environment, and then it can run your Urbit and it can run your Adapt container can run your blockchain node.

And I know that some people are actually trying to do that right now build devices exactly like that. And I think Urbit is really fundamentally well suited to that type of an environment where you have your Urbit identity, you have your Urbit node, it has very predictable behavior in that can't be effected by anybody but you.

So in that work of such nodes can be used to build a new generation of social media apps that do give you the right level of sovereignty as compared to Facebook, Twitter, TikTok, and whatever other extractive social media, moguls do out there.

Nathan: Yeah. Where do you see them in terms of development? Don't they have a 0.10 release that they've just been working on?

Alex: They are fascinatingly far along Urbit is a functional system. It doesn't have that many apps on it yet, but they're being built very actively. The learning curve to build an Urbit is extremely steep. It's a very different system from anything. They have a really strong community.

They have a lot of support from some very smart people. And I think they'll get to a place where there's a big enough community of developers, where you're going to see your Urbit based social media pretty soon. And that would be a fascinating development because when people start switching off of Facebook and Twitter, like I remember a couple of attempts.

I remember Peepeth as an Ethereum-based kind of incentivized Twitter. I don't know where it is now, I don't know expect that it's particularly successful because it's very hard to bootstrap a social media system. But I think because Urbit has such a strong community, I think they have a real chance.

Nathan: Can you explain what a planet and a star is in Urbit?

Alex: Yeah, but it's irrelevant. Yeah. That idea is that your address space where one address, a planet, is your personal ID, is separated into what they call ships. Planets combined into stars where stars is 65,536 addresses and stars combined into galaxies that are, 256 stars. And all it is a really neat way to implement a kind of a DNS system, where in order to find an address that you need to talk to, you go to it's star first it's galaxy, then it's star, and then to it it's just a way to organize the address space.

Nathan: Got it. Yeah, because I see some of the, I see planets selling for 0.1 ETH somewhere.

Alex: 0.1 ETH? That puts the implied value of a star in hundreds of thousands. That's a lot of money for a star. I don't know how fast they're selling though. Urbit has a future. It's very clear that there's a clear sense of distinction between the short-term thinking and in crypto and the longterm thinking and longterm thinking moves very slowly. Think about MakerDao. They took four years to build the first version of Maker. And that's because they were trying to do the right thing and build for the long-term. Urbit took like 10 years to get the place where it is now. And that's because it was building for the long-term. They rewrote the entire system three times.

And I've been working on A now for about three years much more actively recently. And that's because I know that what I want to make out of it as a long-term project. And then there are short term projects that just come out with, a simple system and make a ton of money. I'm not saying it's bad, but definitely not long term.

Nathan: Anything else you want to cover?

Alex: Yeah.  Very cool talking to you. And yeah feel free to send people my way if they have questions or thoughts.

Nathan: All right, Alex. Thanks so much for your time. We appreciate it.

Alex: Yes, thank you Nathan.