Building Sustainable Open Source: Open Core & Web3 Decentralized Clouds – San Francisco 2025

I feel honored to be the closing speaker here. I wasn't originally scheduled, but we shifted around. I've I don't think I've ever been a closing speaker at an event. Um, sometimes I've been first. So, uh, this is really fun.

Um, I'm going to try to keep within my allocated time, although somebody at my company pointed out I have no time limit because, you know, there's no one pushing me off stage. But, okay. So, I think this this topic for me is one I'm very passionate about. I sometimes do give topics uh co talks that are a lot more AI focused really a lot about like what crypto does for AI what AI does for crypto it's incredibly great um you know I think uh you can hear a lot of things but I think when when we talk about it at ien layer I think it's an interesting angle so I invite people to look at that this conversation though I'm not going to mention AI once which is a little refreshing these days in crypto I'm going to talk about crypto settlement and manifesto for open source builders. Um, can I ask everybody to raise their hands?

Everyone, please, please, please, please. Okay, good. Okay, almost everybody. And then I'm going to ask you to drop your hands if you are uh if you are not in the web 3 space. I am, but Okay.

All right. Good. Okay. So we've got a number of people here who are kind of bridging over um and then a number who are who are in the crypto space. So um I'm going to try to address both groups.

The important reason why I do that is the manifesto for open source builders. I don't really have to go into the details of crypto as much. Um but I have a little description here. It's still pretty pretty brief intro. So, I think it's a good way um to kind of look holistically at what crypto does as a net value ad without getting into proof of work, you know, space time, uh stake, all of that kind of gobblelygook.

Um then the more important part for people outside of crypto is verifiable clouds. um they're verifiable because they use crypto settlement but normally if I wanted to I could just kind of uh simplify crypto settlement even more. Uh that's the bulk of the talk and then there's a couple of wrap-up slides at the end which are just explaining what the implication is for this and I'll talk about two angles. One is the people who build open source. Um and uh you know open source itself gets much better regenerative funding in this model.

And then what why web consumers win. Um uh and a lot of times people talk about consumers with crypto about identity better identity better privacy better data control all of those kind of things. Those are all there too but I'm actually going to present uh a different reason. Um so okay so uh just quickly quickly saying about crypto uh and blockchain there's really only three key things that blockchains uh do that are so critical at the technology level. One is they permissionlessly allow anyone to run the software and you know to essentially have all the state meaning that you got the open source and the open data to be able to verifiably just say hey I'm going to just keep going.

They have livveness meaning that it's essentially impossible to stop the the blockchain from running. That's very important. A lot of times you have nation states uh as well as other actors that try to attack. Uh protocols, you can't send them a cease and desist. They just exist and won't stop running.

And then security. Security is the one I'm going to talk about most. So I'm just very want to kind of say uh accurately what does that mean? And it means that you will not actually revert uh transactions or like commitments that have been written to the chain. Um and so it's sort of like uh uh immutab immutability.

Um so there's three aspects to the security or the settlement. Financial, tech, and social. And um they all are important. Financial is the granddaddy of like Bitcoin and others. You have assets move around.

Tech is going to be a big part of this talk. meaning in verifiable clouds you have a lot of things that you're doing computation data manipulation in communicating data and you want to make sure everything is done very accurately without just trusting somebody who says oh it was done correctly the data is right the computation's right and then social um which is very important uh in blockchain if we did not have the social aspect we could still have even with uh blockchain systems people kind of centrally saying hey you know this group of people who own the protocol in sense of tokens and others they could start being monopolistic behavior as well. So uh crypto has this idea of forking. Uh we've seen this on all of the crypto protocols. There's always a threat.

The community says this isn't working for me anymore. The data is open. The uh running the chains are open and so you just actually create a new chain. Like you know a lot of people at this week at uh SF blockchain week are talking about Ethereum or ETH and ETH is not actually the original Ethereum chain. The original Ethereum chain had ETC which was kind of a copy of uh BTC and that was forked because of a problem that happened on that chain.

So the examples of these financial settlement uh you know if you're in web 3 you recognize all of these. If you're not um then it's not useful for me to go through it as much so I'll just kind of skip but their assets a verifiable cloud and then social I refer to it as the ultimate backs stop to central capture if you have that then you can prevent yourself from ever getting to a point where protocols are going to be uh destructive for the users and society so it's extremely powerful um so now getting to the part of like where does software actually run for a long time we've had it running on this tech stack which starts at the very bottom with real infrastructure. Um and then as you move up, you get to the point where you're instead running um OS virtualization, middleware, software, runtime, the data, the application. Um and that tech stack uh originally while it was all sort of managed by the people who made the software over time it got replaced with uh these three concepts the infrastructure becoming a service through software APIs. So that's things like virtualization and uh storage generalization like Amazon S3 if you're a developer platform as a service and then software as a service and so in crypto infrastructure as a service has a new name decentralized physical infrastructure platform as a service has always remained middleware and then software as a service is a lot of the enduser apps so um when I talk about verifiability and uh crypto settlement I'm just going to say like uh there are chains that just do it by everybody running every transaction.

Very inefficient, but that is actually how the original Bitcoin Ethereum worked. Um and uh then the improvement on that are these moonmath zero knowledge proofs where you can take an infinite amount. You can have weeks and gigabytes sorry gigawatts of uh power running a computation and at the very end it produces a very succinct little uh under megabyte proof. you essentially check the proof and you know in 10 milliseconds that uh weeks of thousands of computers running did exactly the right thing. It even checks for hardware errors so it's not even unscrupulous players.

There's a more optimized cost-effective version of that which is a fraud proof uh and I won't go into it. And then trusted execution environments um I layer we're kind of talking about uh specifically two types of uh cryptoeconomic security. One is those attached to onchain verifiable um uh cloud operations or um hardware software operations and then this other one which is the social settlement layer which is humans deciding something was done correctly. It's a very powerful concept like you can actually commit as a company if you're a ride share for instance that you will never take more than 5% of the of the fee of the ride which is very different like you look at something like Uber it started at 20% take rates moved up to 25 30 40 I was recently in Denver at a crypto conference ETH Denver and Lyft showed how much money went to the driver I was paying for 124 rides the driver was getting 350 which to me was shocking right? 70 75% type take rates.

So things like this, I mean the idea that actually crypto you can't do it with onchain verifiability because you can always change the software, change the fees, but offchain crypto now allows through a new primitive uh which is a whole another talk or talk with me if you want to learn how uh you can have slashing of somebody's committed value based on a human parsible uh statement that they made and then users socially saying wait you violated that. Another great example, uh there's so many more uh that I think people are going to come up with, but imagine a social network. Uh we have decentralized social networks, but they don't have off-chain commitments about their privacy policies and how they're going to monetize users attention. Um but you could have a policy like if we're going to add a new monetization of users, it has to be approved by the users. if we're going to uh you know decrease the privacy of what you're doing and release it to more people um we actually have to get your approval.

If we don't then users will see and say hey this company has billions of dollars that's slashable so we slash it and the data is public and the protocol is public so we then go ahead and run it somewhere else. So guess what happens then a company will not actually do these bad actions because uh they would lose a lot for it. So um you know so we have these words here autonomous and verifiable. You're going to see there's a cloud version of that uh which I've been talking about what software actually runs on the cloud. The service is that whole software as a service platform as a service um idea.

Uh and I said I wouldn't do AI but I should actually say AI or AGI agents can also be inserted under services. Autonomous means that uh the software is unstoppable. It has its own ability to just keep running and verifiable means that when the software runs the results of it are you know have a cryptoeconomic type guarantee that what it's doing is what was uh committed to either through the open- source software or verifiable to English type constitution commitments of companies. So um you know the AVS's like services uh they actually were created by our protocol. It's anybody can create them.

There are other protocols that are competing with us now. They're all over the place, which is great. Um, and they all together can create this uh autonomous, verified, unstoppable, trustworthy cloud. And uh, as I said, permissionless means it's uh, autonomous um, and sensorless. And then verifiable means that uh the underlying thing why why there's a trust in uh cryptoeconomic protocols is because somebody has put up the trustless thing which is actual monetary value and if they violate conditions that monetary value is removed.

I mean, just imagine, you know, if you actually could have companies or bad actors not just having corporate social responsibility, but uh in fact having a commitment that when they do have something bad happen, they lose a whole lot of money. So, uh it changes very much rational actor behavior. Um so, you know, in terms of when I'm talking about the cloud, um you know, there's a lot of things that you could be doing with this. Um these are examples of kind of software that you need to be running cloud databases, Linux. Um you need watchtowwers to make sure that software is running correctly.

Proofs that are not available right now to software but are available um through this framework are proof of location that you're actually running in other countries. You'd be surprised at how hard that is to actually tell. Um but it is actually solvable problem. Really smart computer scientists have figured it out. These are names of companies that are actually providing these uh proof of identity of actually being a human.

It's actually also possible to and quite valuable to prove that something is run by a machine and not some mechanical Turk. You might actually want to do this for a lot of reasons. I can imagine one you might want to prove that this isn't like slave labor um that is doing these things. It is in fact like an AI. So a lot of good things uh you know are being constructed there.

So um why are open- source builders winning in this framework? um they are winning because when we think about uh proprietary moving to open standards stepping back this is a historical slide won't go through it but if you look at my slides um I at the end show my Twitter handle where these are going to be posted uh networking moved open uh internet was open over time software moved to open- source software private data centers went to something more open uh public clouds but still proprietary on their APIs and uh value capture And then post blockchain. I don't think we've hit the tipping points for any of these yet. These are all Robert Dross estimate of tipping points. There's no I'm not pulling from data.

There's no references here. But I think a lot of things like walled garden and identity that we don't own. Tipping point hasn't happened. Same with data public clouds. So um there's in incredible amount of kind of capture and monopolization uh and therefore control and like uh bad policies that happen in the current public cloud.

We have almost I think those add up to 56 almost uh 70% in three companies and this list here is like you know approaching uh the mid 80s uh very small number and uh the amounts of revenue are massive. Um people don't really track it but we're in the trillions of dollars really. Okay so let me then go to this uh sorry I got the fiveminute warning. I do want to keep it to here and let people get going. Um evolving open source software in relation to how it runs on hardware.

I'm bookending the entire history of open source software or moving to blockchain. Now in both of these you have sort of hardware at a layer uh moving from private to now these things I was talking about decentralized physical infrastructure open core software that's the crux of why it is digital commons and digital open source software that's the model where you develop and build software because you love to do it you're a hacker then eventually people use it you need money you build a company around it you take a little bit of money for it and you keep developing the open- source core that digital software commons that was the pre-cloud open source vision. We started getting billion-dollar open core companies in the early 2000s. Um, but then something happened. The Web 2 public cloud went from being something really helpful to run open-source software that open source creators used, putting value back into the open source core.

A new fork was created. Web 2 cloud started creating their use of open- source cloud where they took the software, wrapped it in a small interface, and then took the value and charged money for it. I show the line of value going back. It is trivial compared to like what open source founding teams actually put back in. And the amount of value that's extracted is not even proportional.

I have to make that arrow bigger because they take a lot more value out because they are competing with the open- source developers themselves. That relationship bookended flips back to the pre-cloud o open source version. We still have the cloud available for open source developers to put out their software to be able to scale elastically everything that you want in the public cloud. But there's not a single public cloud company at the base that's wrapping up all the APIs. These are now uh protocols, decentralized protocols for databases, for compute, for storage, for the massive numbers of different uh software as a service.

And so you don't have a central person who's going to monopolize and take value. And there's no one person who's going to build an open- source version of yours. If it is, it's a competitor uh which is another group of open source developers. It's not a company that has hundreds of billions of dollars of revenue to compete against the poor open source developers. Um so that fork value capture is now basically cryptomonetized open source in the web 3 world and it really regenerates that money back into the open-source ecosystem.

So a lot of a conference like this on funding the comeds is going to talk about community contribution, donations, you know, sort of grant-making and I love that. I think open source should be funded in many different ways. This is talking about a systemic profit- driven regular capitalism way that we get much better open source and much better outcomes for open source developers. So it's very exciting to me and that's why I'm, you know, it actually is my big mission for why I'm in crypto. I have never been here for the decentralized token value in the finance sense.

So I do want to point out this is a really big thing. This is a single screen developer cheat sheet. This is a small subset of the APIs and software that you can get on Google. And uh I would probably have to go through one by one in order to say a percentage in appreciable and I think majority and maybe actually most of these software packages are renamed packaged open- source open core software. Um and when you go on the website and look at it, this goes on and on and on.

So all of those are actually kind of value extraction from open source. Um okay. So now um you know what I want to say is like why is this better um building this way? I actually argue it's better for fundraising. If you look at how much uh how quickly the velocity for investors on getting returns in the space is in a year, year and a half you get a couple up rounds.

you already released to the public markets. In crypto, it's called an initial coin offering versus initial uh public offering. It's the done decentralized around the world in one instant. Um you have a lot of buildup before the time. Great ability to kind of market and reward your uh entire consumer base.

You have your users ready to go. So I actually think in the end this is better for investors. This is going to be the faster action for money going in and then cycling the money for future um startups in the decentralized space. Better for building. You de-risk and you're not competing against the actual platform you're deploying software on.

Um you also have the ability that uh you have global versus walled gardens of things like data and identity. So you actually start off new protocols without having them empty. You have a network effect in web 3 to base your protocol on. Uh second to last slide and then um better for outcomes, more ways to monetize. Uh you have actually the full freedom of tokens and the crypto economy um is much more flexible than the capitalism construct.

It's a great coordination echi uh mechanism but it was created 4 or 500 years ago and so it's kind of crazy that we're still using it exactly in that form. Uh in a sense crypto is sort of uh creating a much better version of capitalism. Much more freedom and innovation can happen. Web two consumers. This is the point I wanted to say that you know the consumer superpowers.

Of course, people focus on the data and other things and those are really important. Uh you bring all of your data with you. If you're an Uber driver, you go to any new thing like the Uber the example in Southeast Asia of somebody creating Uber with a 5% take rate 0 to 5%. uh you can now actually take all your reputation assets from Uber and Lyft and any other rid share service you are and you can import them into new crypto protocols and that is something dramatically new uh so drivers can trust uh riders riders can trust drivers uh very different um and uh you you actually get choice this one permissionless algorithms it's a little bit esoteric but I actually think given that your information diet or your attention is a really precious asset you are kind of you think what you consume in a lot of ways. And so imagine a world where um what you're looking at in YouTube or social networks or others you choose.

Maybe you actually just put as an LL LLM or maybe you have your favorite influence uh spiritual coach whatever and they give you a moderation algorithm that you choose and you may choose a different moderation algorithm to start the morning and a different one to wind down at night and a different one when you're hanging out with your friends. So you know we should have open exposed choosable algorithms for what we're actually consuming on these platforms. The final one is that um there's a reason why capitalism has been the dominant mechanism and uh it it encourages competition and a lot of alternatives and kind of a meritocracy. It eventually collapses to monopolies in a lot of places at scale but like the beginning part is really good. So by having these types of open fast innovation, we actually see for users new features and improvements with higher velocity, monopolized economies stop creating new features because they actually don't want change.

They want to protect if there's a good new thing coming up. Who remembers how many times Facebook has bought things, right? They bought WhatsApp, they bought Instagram. They do that because, you know, they actually do not want to compete against these people. if they need something, they buy it and then they kind of freeze it for the most part unless they have to like with Snapchat add uh you know ephemeral images and messages things like that.

So the takeaway message on that one very very simple message with competition you innovate with monopolies you stagnate and uh I think that's the final concept there. Um I actually have a list of this so if you grab the slides these are my information diet of some of the things that have led me to this thinking as I've gone in the crypto space. Um a lot of great ideas. If you're not familiar with them you know take a screenshot go read about them. the majority you can either read open like Carl Fogle's book he started writing in the early 2000s fully common um common good funded uh you know every single version now you can download the PDF you can download or read it on HTML you can also go ahead and you can buy a very inexpensive printed copy if you're if uh if you're down for that um okay and so takeaways we got those takeaways okay you can read more of my thinking on this there's a fulllength article about in uh magazine ETH investor club which does uh you know periodical every 6 months or so.

uh great articles from Kevin Owaki in that art in that issue as well um who was talking here as well as Tim Beo who's uh you know one of the defining people in the Ethereum Foundation and uh then also if you want to read more about what autonomous verifiable means there's a great blog article that u um really smart guy in our team wrote and then if you want to capture my Twitter um I talk about stuff like this a decent amount and uh you can look at this in other slide presentations I drop videos of things that I I do and panels there as well. Um and then if you actually want to talk with us in our protocol, um we currently have about uh 40 live um SAS and pass protocols that are building and deployed. Uh so they're actually accessible for developers. There's uh you know we're passing 200 that are in the pipeline of people who are deploying soon. And if you look at the open source world, I really think that number goes up into the thousands.

I will also say like we are not a very extractive protocol. The only thing that we do is we allow the decentralized trust security commitments to be made. We actually don't compete in any of these markets in the cloud, it is fully meant to be built by open source developers. And you know, like I said, you know, uh if somebody thinks we're doing the wrong thing and we're being too competitive to our own space, they can fork us. Okay, so that's it.

Thank you all very much for your attention. No questions, right? Or are we doing anything? Uh, if anybody has a question that they think would be kind of useful to share for the whole audience, feel free. Also, totally fine.

You know, we're getting to the end of the day, so but I'll just give a moment. Sometimes people do think a little bit. Oh, I sorry you were by the light. I couldn't see you. Yep.

All the way in the back. Can you do me a favor just so I know? Can you say your name and uh you know where you coming into this conference from? I can repeat it for you. Okay, that's great.

It's really good. Uh Al Morris P network. We're also a computing protocol similar to layer. What do you think of that? There does seem to be an issue in this industry with shelling kind of effects around things.

So, uh, in the same way that like Bitcoin kind of moved over to Ethereum for wrapped Bitcoin, do you worry at all about, uh, the effects of monopolization actually, like we're probably going to end up in a duopoly of compute networks at some point at best or a monopoly actually. It should it should head that way eventually. Yeah. I mean this is one thing that you know even when it comes to monetary systems I sometimes worry about crypto in general because you know if you're ever read death the first 5,000 years um they have a great observation there that on average a money system in a society has had an average lifespan uh average or median I don't remember of 80 years. So we see a few random examples like the Roman Empire or you know British Empire or now American Empire and you see currencies and monetary systems that last hundreds of years albeit with you know different levels of inflation but most actually eventually hyperinflate or the government's overthrown or you know you even have things like in India where they just basically turn off an entire cash system and then move to a new cash currency.

So, it's both very exciting to have like, you know, uh to have a money system that's meant to last forever, but it also sort of points out there could be problems. Um, as far as duopolies, um, you know, I think the the moneyiness aspect is probably the place where there's some kind of feeling on risk on the um, blockchains. I think for stuff that's running on a blockchain, it's very you're very able to kind of mute over and run other places. So, uh, at least for Ethereum, like the EVM has propagated itself off into the Cosmos ecosystem. In Salana, they have an EVM compatibility layer in Avalanche.

So, there's many kind of options even if they're not using the original um protocol. And again, you know, I kind of fall back to on the social settlement side. It may be hard for societies to like harder for societies to take like multi-t trillion dollar monetary assets and fork them although in reality their data and everything is around um to do that. Uh but definitely when you're looking at like layer 1 protocols or others the history of seeing L1 forks on Bitcoin on Ethereum um we've even seen on Ethereum there was one time uh a protocol that launched very bad founder he did a few bad things and the community just you know created a fork protocol and took away all of his tokens um and uh you know that that's a powerful thing. So the construct of the social settlement on ien layer uh the you know the new cryptoeconomic primitive anybody can use it um allows you to not have to go through the painful L1 forking where the slashing kind of takes potentially months or like the etc ETH fork took well over a year year and a half to settle what was the valid chain that everybody was going to work on.

Um but this new cryptoeconomic primitive um allows you in about two weeks maybe maximum three that the slashing and the canonical uh fork is decided and uh you know so that actually speeds things up. Sorry I kind of riff a lot on it. Um did that cover the duopolies or is there still yeah pretty good answer I think uh at least I see where you're headed. Uh one other small question which would be around the collateralization of workloads especially in like compute networks. Do you see that as a limiting factor?

Is there some advantage silver lining kind of thing that we're missing? Did you say collateralization or the what? Uh collateralization for compute workloads. So if you have to stake for every compute workload that basically doubles the collateral required for the system to operate properly. That's a great that's a great question.

Um, so the reason why just for people who aren't familiar with why collateralization can cause an issue is because um, for somebody like in the non-crypto regular world, they kind of get collateral for free and they get it at scale when they're like kind of ready for monopoly and network capture. Um, you know, just taking I don't mean to con constantly be a punching bag on, you know, uh, turn AWS into punching bag, but it is the biggest. So, let me just talk about it in that sense. So you know people kind of trust that there's great SLAs's around data around privacy um and uh around performance and they are many many times violated like we've had long um outages um and there's not actually compensation made and there's a reputation hit but there's not really a lot of things to do if you've already built on Amazon. So um collateralizing is powerful in crypto because now like in filecoin you know our filecoin foundation's great sponsor of funding the commons and juven created this whole uh conference I think originally.

Um so that's great because now you know the SLA is backed with real losses for somebody who doesn't do it and we know that's a lot more incentivizing to any business uh entrepreneur. The challenge is that like assets, you know, if you have a asset like uh US dollars, you know, right now the zero risk um kind of uh uh return and T- bills or T- bonds is like whatever four 5%. Um and so you would expect then that people would say collateralizing and putting a million dollars down behind your storage commitments is also costing you 50,000 a year and maybe higher because if you're doing it in a crypto token and it's volatile, maybe you need 10%. So there's an aspect of our protocol which has actually now proliferated. It's a DeFi category called restaking.

In reststaking, you use the same capital asset. Um you don't rehypothecate. It's not like uh something called liquid staking where you issue a new version of the same token. But the key thing that you do in crypto for the slashing uh where somebody does something wrong and then once you prove they've done it wrong, you essentially take their assets away. typically burn it, but you can also redistribute it to damaged parties.

Um, and we allow you to put, um, kind of stacked multiple slashing conditions on the same collateral. And so, uh, of course you have to be careful. There's some information theoretic stuff you have to do so that you don't get into kind of spirals. So, you want to spread it out. And, uh, there's also kind of reuse of some of the slashing can be put back in to keep coverage for other things.

Um but you know it is actually um it's not that hard like that is tractable and so then you can actually have assets that instead of having one slashing condition and therefore having to have like uh 10% return on one protocol uh if you have 10 or 20 different uh uses of that same capital for uh collateral then you know simple math is just divide it. So you know 10% divided by 20 says now you're covering half a percent. And so half a percent is much better on a million dollars of collateral. That's 5,000 a year. And so that's no longer your major cost of goods.

So this capital efficiency, in my opinion, is really necessary uh to kind of combat the monopolization of a reputation-based uh non non-cost capital in the regular web 2 world. Short questions, long answers. I'm not doing this well. Okay, we're going to take a new question, then we'll come back. Okay, thank you.

Oh no, we're out of time. Okay, just come up and talk to me afterwards. Thank you though. Those are two great questions. Thank you.

[Applause] [Music]