Epicenter - Learn about Crypto, Blockchain, Ethereum, Bitcoin and Distributed Technologies - Alex Leverington & Julian Zawistowski: Golem – The Worldwide Decentralized Supercomputer

Episode Date: June 14, 2017

With most of the world’s computing power concentrated in Internet companies’ data centers, the Internet has become a very centralized place. The Golem project aims to build the sharing economy of computing power. This “global, open sourced, decentralized supercomputer” will function thanks to the combined power of a massive network of machines, from personal laptops to whole data centers. We’re joined by Julian Zawistowski and Alex Leverington, CEO and P2P engineer at Golem. We discuss how Golem hopes to create a more balanced distribution of computing power for the Internet. Topics covered in this episode: What is Golem and the problem it is addressing What may people and companies use Golem for How Golem compares to other projects like iExec or TrueBit Who are the different participants in the Golem network How Golem executes computations The role Ethereum has to play in Golem How payments work in Golem The economic model for Golem’s decentralized marketplace The project’s current status and roadmap Episode links: Golem Website Golem Blog Pre-Brass Golem 0.6.0 Released! This episode is hosted by Brian Fabian Crain and Sébastien Couture. Show notes and listening options: epicenter.tv/187

This is Epicenter, episode 187 with guests Alex Loverington and Julian Zavistovsky. This episode of Epicenter is brought you by the Ledger NanoS, the hardware wallet which sets the new standard in security and usability. Get it today at ledger wallet.com and use the offer code Epicenter to get 10% off your order. Hi, welcome to Epicenter, the show which talks about the technologies, projects, and startups driving decentralization and the global blockchain revolution. My name is Sebast Sankujiuio. And my name is Brian Kroen. We're here today with Julian Zavisovsi and Alex Labyrinth. I'm sure many of you will have heard of these guys before.

They are working on Golem. Golem is one of the, was one of the quite earlyish Ethereum, what's called ICO today. You know, they did the crowd say, raising about $9, $9 million at some point last year to build a decentralized cloud computing platform. And so we have today here, Julian, who's the CEO and Alex Leverington, who is working on their peer-to-peer networking architecture, and he's also an advisor to the project. I know Alex, because we used to both work from the Ethereum hub in Berlin.

So I've known him for a while. He was before a developer at Ethereum. And also Julian, I've known because he's given a talk before about Golem before their fundraise or even at the meetup in Berlin. And yeah, so super excited to have you guys on. Thank you and thank you for inviting us. Yeah, I mean, you guys have been, I guess, through quite a insane journey to know because you had this early fundraiser, which was already very successful, right? I think you guys raised like $9 million in like half an hour. And since then, it's just this whole crowdfunding thing has gone exponential and exponential.

But Golem has fared very well, right? but there's a lot of attention on the project and price-wise. It's appreciated a lot as well. How much that means about the project? I guess it's always a question. But so how has that journey been for you guys? Yeah, well, actually that wasn't that long time ago when we had our crowdfunding. It was November 2016.

But already, I think, like, things were starting. to go pretty insane at that time. So obviously our crowdfunding took 29 minutes, but the last 10 minutes were only because we were waiting for a fraction of ITER to close the crowdfunding amount. And also, perhaps it would take around 2 minutes if we had not DDOs my ITER wallet for which we are like very sorry until today by recommending like everyone to use

Maitor wallet during crowdfunding that was that was said to big success and and of course like ever since we we feel the responsibility of all the people trusting us during this crowdfunding you're working hard to deliver and and of course this is this is this is This is difficult technology. This is bidding at technology, but I think we are making fair progress. And yeah, personally, I'm still as excited about that as I was half a year ago or even like a few times more.

It's definitely very exciting. It's good to see other projects also sort of growing in the space as well. So my ether wallet, you know, they did a great job and they were really responsive. And so it's that kind of thing where, you know, rising tides raises all ships. And it's also interesting to see people around the world who have, who said they, they bought Gollum tokens because they believe in the project. That's also a little nervous too. So we have a lot of work to do and security is really important in this space. So it's really intense. Yeah, I should also disclaimer at this point that I also put in a little bit of money into the

Gollum fundraiser back then. So I still hold some Gollum tokens. So you actually made it into the crowd fund. I did actually. I did. Actually using my. the wallet too I think oh yeah that's great that's great yeah but maybe we can take even a set before that crowdfunding event and maybe julian can just tell us a bit about how did the original idea of golem come about and how was that project born yeah although i'm not engineer i i always was a kind of techie person and and especially uh uh decentralized computing I always found a very appealing idea, like running Sedi at Home on my desktop back in the late 90s, in college times. And back then I thought that it's something that kind of makes sense to use computers that wouldn't be used otherwise.

But then, of course, I worked for many years on different things. than a couple of years ago I started working much more with software development. And then we came across Ethereum project very early in an Ethereum project, like early 2014. And then we realized at some point that you could combine the Ethereum network where like for the security and for consensus, every country. computer basically does the same with the other network, which we eventually named Golem network, where every computer does something different.

Because the critical problem with the truly decentralized computing network is how to coordinate all the nodes in a decentralized way. By coordination you can understand a lot of different things, like payments, like security, like how to operate in a trustless way. That was the beginning when we thought that using smart contracts on Ethereum you can actually create a truly decentralized network. Not only in a way that you decentralized computing over nodes, because this is something very basic,

like this is successfully done for a long time, But also how to introduce economy into that system, to give people incentives to contribute their power to the computing power to any project brand on the network, not necessarily like a cool one, like searching for aliens, like we just said at home. So maybe just at this point remind our listeners, what is distributed computing? I mean, we've had other distributed computing projects on the show before, and we can talk about those and how they're similar or different to Gollum later on the show. But give us just a sort of high-level overview of this type of technology and how it exists today

and how it's being used and applied in different fields, industries, et cetera. So I think a sort of quick summary of distributed computing is, I mean, is very much what the term, the term says what it is, and it is what it says. Simply put distributed computing is taking some task or set of tasks and coordinating the separation, the optimal separation of the tasks in the smaller pieces, and then distributing those onto multiple machines. Distributed computing can even be at like a small scale where you have one machine with a lot of CPU cores. And it can also go through the network where you have a data center full of machines

and you distribute the computational tasks across those machines. And what makes distributed computing a little different is that oftentimes the type of task that needs to be executed, the algorithms involved in that task, determine how it can or can't be distributed because in some cases there are tasks that are sequential and they can't be distributed. So distributed computing is coordinating all of that. And for us, the decentralized distributed computing is where that coordination can take place on a peer-to-peer network without a lot of administration or manual coordination from a human operator. What you can do with distributed computing is

execute highly parallel tasks by leveraging massive amounts or in volume of computing units and those computing units execute those tasks in parallel and therefore you sort of distribute the execution to all these nodes. Now, what Gollum is building is the network that will organize the sort of supply and demand of these computing resources and the tools and the marketplaces in order for users to be able to have access to distributing computing networks, if I can sum it up that way. How different is that from how things are done now? Like if I want to use it to distributed computing network today because they're they're out there, how do I do that? So right now today, most software libraries for distributed computing are either very low level or they, if they're high level, they're very application specific.

So one may be for a certain machine algorithm, machine learning algorithm, another may be for a certain image processing algorithm. And what we're doing with Gollum is building tools as well as a network topology that is generalized rather than being very low level and specific to a certain processor. or being application specific at a high level. So it's almost like a middleware framework for distributed computing, together with the transactional aspect, of course. And so what's the goal I'm adding to this? Is it mainly to make it generalized? And why is that such an important thing?

Yeah, well, maybe I was trying. Then you'll add, would you? we do think I'm missing in my answer. So I think there are all these two important things here. Like, what is the economy of the system, and the second is technology. But both boils down into the notion that we want to abstract, we want to abstract computing and using hardware. from using software. So, like, actually you never need your computer,

or you never need your cloud, or your decentralized system. All you need is the result produced by your software. So what you're interested in basically is software and the result of the whatever you want your software to do. Like, everything else is not really necessary. It can be abstracted. So what you want to do is we want to create market for computing power,

but also combined with the market for software. And this way, like, abstract both. And, like, I know that might sound very abstract, put it that way, but I think that at the end of the day, this is exactly what you want to do. Now, what are the benefits? Like, like, if you, of course, if you create, like, one, market which is you can hope that it will work much better than a lot of small fragmented markets as they are as it is at the moment because because they

basically cloud computing market which is not at this at the moment it's not exactly substitute to golem but let's say it's like the close analog to golem so so this market works on a in a little bit wired way because you have like very few, very large players and a lot, a lot, a lot of medium and small players. And I believe that only like adding a market being a kind of middleware, as Alex described, that will make much easier to rent this competing power for anybody around the world, will be like hard, it will be an important distortion of that market. not only in terms of price, but also in terms like who get the job.

But also I think that the fact that how we want software to be used on top of Golem gives a very interesting new distribution channel for its software developers. And on top of that, of course, you have all the economies of scale by which we do. we hope that the price will be lower than in traditional system. But of course, this is not like the, this is one of the promises, but I don't think this is the most important part of it. And the second, you have, you have all the advantages and disadvantages of the centralized system. And I believe that in this particular case, like advantages are much more important than disadvantages.

So you described so the staff. right. The bottom layer we have the different distributed computing networks and then Gowlem is somewhere in the middle there between those networks and the applications. Would Ghalm then enable sort of new types of apps or new types of business models? Maybe describe at the higher level in the stack what types of new applications could emerge if Gullum would be successful. Sure. So I would say that to start with, we are not addressing new applications. You're addressing new ways of using applications that are already here.

So this is also why I think this is a really valid business model because we just, you know, look around. We see a lot of software that could benefit from running on software on Gullum. And this is what we hope to see on Gullum in a few years or a few quarters, perhaps, in some of these cases. At the end, at the end of the day, I would say that the most important difference is in a business model for the software developer. So now usually in a task we are in a software we are aiming to is it works like that, that you get the software and then you deploy it somewhere. Or you've got hardware, you deploy software on your hardware and then you are seeking for users to use the software to trans on your hardware. And what we want to do is that user gets software and then rents using the software along with renting the hardware that the software is running on. And I think this is a really a game changer here.

Also, you should think about open source software and community-driven integrations. or by deploying them on GOLM, you solve both the problem you want to solve with the software itself, but also like the usability problem, like the problem of how to deploy that on a large network.

Perhaps Alex would like to comment on that as well. Yeah, yeah, I think that there may be some new apps, but I think we're going to see new apps, new apps because we're seeing more peer-to-peer networks. I can't necessarily say that those new apps are going to be something that is made possible because of Gollum, but more because people learn how to leverage the peer-to-peer networks. But I do think, like to what Julian is saying, that we will see applications function

differently. So some of the stuff is going to be under the hood. And I think that's the same case with, with like Ethereum that we may see some interesting new apps, but they're going to be like sort of recreations of what we already have, you know, only powered by Gallum or powered by Ethereum. And the economics, you know, for example, crowd funds, that's something Ethereum enabled. And so we're not sure, you know, exactly what Gollum will enable, but we think it's going to have to do with the software developers and the providers providing their resources to the internet. It seems that at this time, there's sort of a lot of projects in the space that are doing things that are quite similar or somewhat similar to Gollum. So our listeners will be familiar with episodes we did with with iExec, with True Bit, and then perhaps a little bit more further away from Gollum,

but to some extent still looking to do the same thing, projects like Microsoft Bletchley and what they're doing there with Criplets and Town Cryer, maybe more focused on the distributing, on the distributed computing projects like Iexec and Truebit. Can you perhaps talk about how Gollum is similar or different to those two projects? So I exec and Truebit, I guess those are, I think those are two very different projects. From our understanding of Iexec, it's sort of a gateway to leveraging computing

resources. I think their software will evolve. So as we'll Gallum, so it's really

hard to say where these two will stand long term. But right now, Gollum is focused on integration

with Ethereum together with this sort of decentralized peer-to-peer network. So any

peers can join or leave the compute network. Whereas with I-exec, there's sort of a kind of a guardian system there. What we're doing is we're building up this peer-to-peer network that's stable, given any peers coming and joining, and then we can layer more application-specific networks on top of that. There's been some mention of I-Exiexie- exec being used for DAPS and GOLOM not.

But I think GOLM will be used for DAPS as well. Somebody will come up with some silly little thing where it launches a service on a Gullum node that then kicks off a smart contract, that then launches another service or some silly thing like that. So really, I think it's going to be up to the developers and how they leverage our transaction framework. And with Truebit, so Trubit is a really solid piece of technology, I think.

However, Trubit, I think, is more aligned with something like the Ethereum VM or some kind of concise environment, like an EVM or proof of work, where you expect every machine to do the same piece of work. for validation. And Gollum, I think, will be similar, and at some point, it might even make sense to implement TrueBit, but we're focused on building a sort of framework and layer for applications to interconnect

and perform distributed computations. Whereas I suppose TrueBit is focused on a validation algorithm, and iExx is focused on a sort of gateway for leveraging compute resources. Okay, yeah. So on the Trubit side, we're talking more about validating complex computations outside of Ethereum, but not necessarily computations that are meant to be highly parallel, whereas on the Gollum and perhaps Iexec side, mostly those projects are geared towards providing the tools and creating the marketplace

for distributed computing resources. Yeah, somewhat. And I think Trubit is really good for like one EVM, so one piece of software, and all the nodes run the same software. Whereas what we're doing with Gollum is that every node is going to, run and execute the software that it chooses and the requesters are going to choose which software they're executing. So Truebit we sort of might fit inside that, but we're building the whole stack. Yeah, yeah, exactly. Just giving some head up to how we see the whole system

working. The verification is often has to be like integration specific and for some use cases for some integrations like Truebit might be great to the verification. For others it will not work. So I would I would rather say that that Trubid is in fact potentially really like synergistic we call them not competitive at all. Let's take a break to talk about the Ledger NanoS, the new flagship hardware wallet by Ledger. I'll pass it over to the Ledger's CTO, Nicola Baca, who can tell you all about Ledger's security features and SDK. The Ledger NanoS is a personal security device based on a secure element, a screen and button, so that you can verify everything that is done on device and make sure that you are really doing what you wanted to do.

Compared to our previous solutions, this device is based on the latest generation secure element, the ST-31 from ST-Migrant. The SC-31 is using a secure armcore, which means that you can have the same ease of development that you would have on a generic microcontroller, but benefit from the security features of a secure element. Security features include an application firewall at the lowest level that lets you protect applications from each other, which means that you can load multiple applications on the hardware wallet, even post-issuance, and you as a developer will be able to leverage these features to load these features to load these. your own application without our authorization and without any kind of authorization from the vendor. We will be providing this device with an open SDK that let you do anything you want with this device. We provide sample applications for cryptocurrencies, different cryptocurrencies, so Bitcoin, Ethereum. We will also provide Fido Authenticator and you will be free to add everything you like. For example, you could add some secure messaging, some encrypted chat,

And you'll see that the solution is quite powerful and very easy to develop with. The NanoS sets the new standard in hardware wallet security and usability. You can get yours today at ledgerWallot.com. And when you do, be sure to use the offer code Epicenter to get 10% off your first order. We'd like to thank Ledger for their support of Epicenter. Let's dive a little bit deeper into how Golem works technically. So first of all, Golem's network, there's different parties that do different things in the Gollum Network. Can you just run us through, you know, who are the different actors and what are their roles?

The different actors are we have the providers who have the compute resources. We have the requester, which is kind of like the end user. That's like, you know, the guy ordering a taxi and the providers, the driver. And then we have the application. We have the software developer. and the software developer implement software that's going to be run by the providers and

the requester is going to utilize that software as a part of dispatching the request to the network maybe we can run through concrete example rendering images or 3D computations often mentioned as a use case for Golem so let's say I'm an architect

you know, I'm designing some building have this really heavy computational task. I want to send this to the gold network to be rendered and executed. So what steps would I go through on my end? And then what would happen on the side of the network to deliver that service to me? At the moment, perhaps, Frederick is the best example because that's the one which is working at the moment. So maybe maybe the best would be just the sort of. how Golem works right now, so how the alpha version works like now, from the user

perspective. So basically, of course, you prepare your work in your software you used to the work and then you, in case of Blender, you write it to the blender file, perhaps with some more files used as resources to do the job. up and then once you've got that already, then you start using Gullim. So you create a task, you add this file with any additional resources that are to be used. You can of course change some parameters, like if this is like an animation, how many frames, which frames do you want to render, what is the resolution, those kinds of things. You can also decide on the price.

This is a little bit tricky, but boils down at the end on setting the price for a benchmark hour of computing. So this is what you said. This is not exactly how it works. But at the end of the day, this is acid. That way. And then you just send it to the nodes that technically to the notes that accept your work. like maybe maybe maybe maybe maybe alex will want to elaborate more on the on the protocol but what happens then of course is the computing of the chunks of this of this part on different

computers on and by the way at the moment like you decide how this is how this is distributed so that might be very straightforward if you have like 100 frames you may want like every frame to be computed very different note if you have uh you can also like random one frame splitting that into six, 12, 10, 20, whatever tiles. And then, and then you gather back to results and you have your work done. Underneath, of course, when this is completed, a Curium transaction happens with a transfer of D&T from you to the other nodes. this is still works in a rather basic way

one of the priority efforts we are in at the moment is how to make that very much complicated in a different ways depending on the use case so I think this is like very high level Alex yes that's high level and a little low level too so yeah right I think right now the current I mean the current interface we have under development is that you drag and drop the render file,

the blender file into Gollum, and then say, oh, this is a blender file that I want to render, and then set the parameters, like Julian said, price and everything, and then hit go. I don't know if the button says go, but maybe...

Yeah, well, at the moment, to drag and drop. not in a not in a not in a not in a drag and drop is still on the

on the development yeah yeah like I said but ultimately this is like a drag drop and drop yeah and so

I think it's important to note though that other software applications in the future using Gollum they may integrate Golm directly

into their software so you would just inside the software you would click render and then have a little checkbox that says render with gollum and then hit okay and then you'll get the gallum parameters and then finish and then the you know the sort of application will drive the process instead of external software but right now it's it's much easier and more effective for us to develop the software as a separate application and and i guess that's also one of the

because one of the big obstacles with this right is going to be whole like cryptocurrency buying G&T first like all of that is like not something that's going to be I think achievable for mainstream users anytime soon but then if you have Golem integrated into the website that I presume really you can pay with your credit card to the application developer and then behind the scenes you know he kind of loads a G&T in the application and pace based on behalf of the user did you guys see a playing kind of of like that?

Not exactly, I think. That depends on developer, I believe. But I rather think what Alex described is, it's like, Gullim running somewhere in a system tray as a kind of Damon that you, that does your services for you, like, provides Gollum services to your applications. And then this is not really visible to the user if he uses any kind of render software, any other software that uses Gullum. exactly happens with Golly. But then as for the tokens part and if this is like dealt by the

software developer or the user, our thinking is that both should be possible but basically we should not delegate hard think to the software developer but we just should make the hard to think easy. And then and then and then at some point this This is what I described before, that just user pays for using the software over the centralized network and the network itself using this goal and that works seamlessly somewhere in a system. And I'm not very much worried about how people will get tokens. I think that we will witness like a massive progress with usability of DAPS tokens. over the next couple of quarters

or maybe even over the next couple of months. Basically, this is what this industry needs to take off one of the things this industry needs to take off. This is not a call it a specific problem. And a lot of people are working on making that simpler. And this is going to happen. Because there are so many options, we're also going to try and build in that usability

to where there's something consistent. So a sort of like Facebook Connect or PayPal, but more for like these native applications that need to interface with with tokens. But we do, you know, we do have to sort of progress along with the industry and help where we can, making it as easy to use as possible. Now one important question I want to dive into and I remember when you gave to talk in Berlin, Julian, that was also one of the questions. a lot of people asked is a question about, you know, verifying work because there's something actually interesting in which a Golem is similar to a proof of work in Bitcoin, right? So in a way you share also this computational problem with all the different miners, all do their work, and the unique feature in Bitcoin, of course, is

because if the hash is you can verify the work very, way cheaply and very efficiently, and you don't actually have to kind of rerun the work to make sure they did it correctly. But now if you talk about, you know, rendering these 100 frames, we send out to different people. Like, how do I actually make sure that what is sent back is, you know, the work done properly as opposed to, you know, maybe half-assed or not at all? Or, you know, in some way that's not, you know, according to specification. Like, how can you ensure that in GOLM? So there are several versions of Gullum. And in the first version, we will be working on tasks like rendering, where we can linearly split up the task and sort of uniformly split the workload of the tasks uniformly across the peers.

And when we do that, there's a couple of ways that we can get validated work. and one of them is to add redundancy to where some peers execute random pieces an additional time and they don't know that that's what they're doing. So basically you have peers zero through nine and peers one and five work on the same piece of work and they don't know that they're working on the same piece of work. And then if they don't know, if they don't match, then there's this sort of implicit arbitration. So it's sort of like a preemptive true bit together with kind of a staking system. Now, what we have right now is we have a reputation system together with this redundancy.

Slowly we will progress towards stronger validation algorithms, which basically have punishing mechanisms similar to what Vitalik's described. with this slasher protocol. It's also important to point out that we will need to have different validation algorithms over time because other computational tasks like machine learning, it may be possible to get the results in the deterministic way, but the funny thing is, is like with machine learning, you actually sort of have to do another test in order to test the quality of the results. And with machine learning, the quality may be easy to cheat.

These validation problems depend on the use case in the application. Proof of work is something where everybody sees the same data, and it's also self-validating. You just check the hash, and everybody in the world doesn't need some custom software to verify the proof of work output. With Ethereum, it's the same exact thing. Another important thing about Ethereum is that Ethereum operates on, shared data. And so when some Ethereum transaction goes through, every node in the world needs to see the result. With Gollum, however, what we're doing is we have one requester, and he sends out this work to, you know, 32, 64, 128 peers. And only the requester needs to see the output at the end of

the day. And so because our situation is a little different, we don't have this shared data, like shared state like Bitcoin and Ethereum, we can build on different kinds of validation algorithms. We can do some of the things like with hashes and such, but like I said, it's a little different environment. So we can do validation through redundancy, not because there isn't a shared state, but because the output only is needed by the requester.

Yeah, I would add to that that also one of the things we are seriously considering that not working on yet, but considering is introducing some kind of voluntary identification so that you can voluntarily identify yourself in a way, commit to some rule of conduct, and then users might opt out to use only a tier of providers, for example, with this kind of self-imposed identification, which in fact would create a situation a little bit similar to the public cloud, where you don't have any other way of being sure that this is okay apart from trusting the cloud. the cloud provider because he's a cloud provider

you know. Yeah, and I think another thing to point out that's interesting about this validation algorithms and Gollum, and maybe there's a question later on this that I'm sort of preempting. But there's the question of privacy, right? So with Gollum,

you're going to send out your request to some random providers and maybe they're going to look they're going to decrypt your input and then take your work and then go and render it themselves and do something with it. Hopefully we can make that difficult enough that, you know, they don't do that. But the thing is, is it's a public decentralized peer-to-peer network. It'll be more secure than BitTorrent. You know, we'll have encrypted traffic so your internet service provider can't see what's going on.

But ultimately, the providers can actually see the work. So to Julian's point about having named provider sets where there's some trust mechanism with that individual provider, this allows for situations where Gollum can route work to providers that need to be pre-selected ahead of time. And I know that's not important to the validation algorithm, but what's important is that we have to consider, with the, the validation algorithm, we also have to consider privacy because there are software situations, medical records and things like that, where they just can't run the tasks on a, you know, on public nodes. So you guys did your crowd sale on, on Ethereum. And so I was wondering, why is it that you need the Ethereum network to run Gallum? Are you tied to the, theorem network? Could you potentially also use, I've used some other payment mechanism,

you know, perhaps such as Bitcoin, or is it possible in the future that you would use some other smart contract enabled blockchain to run the Gullum network? So there's a couple of answers to that question. But up front, because of the transaction framework and application registry that we have, we need. We need a sort of a global state system with consensus for the transactions. And we're building on the EVM and we're building with solidity. Ethereum is viable, it's in production, there's a team supporting its development. And I mean it's been running for a couple years.

So yeah, maybe we could look at other chains or something but it would be a huge distraction. and we don't see how it would add any value to the project. As far as why something like Ethereum, the answer to that is with Ethereum, the accounts can sort of talk to each other. So with our transaction framework in the later releases, we won't have this in brass, but when a transaction completes,

some other event can happen on the chain. And that's really important for providers and software developers, like for billing systems and things like that. And it's also important for a transaction framework so that software developers can make it where when their software is run on the Gallum network, maybe they get some percentage of the GNT as a part of their software being run. or maybe they need to, you know, have a set of authorized peers or something like that.

And the only way to do that in a decentralized way is to have something like Ethereum where we can have certain pieces of logic executed flawlessly in a global state. I just point out real quick, with Bitcoin, they have the, the script hashes, but one script hash can't trigger another script hash. And this is like the only way that they could possibly add that is they have to then add this sort of gas cost like mechanism that Ethereum has. And so at a technical basis, it's easier to just build on Ethereum because if we want to build on a chain that doesn't have the EVM like Ethereum does and the gas mechanism,

then we would essentially have to rebuild the same thing as Ethereum. And like Gollum is complex enough as it is, so we don't want to rebuild the blockchain. And why rebuild a blockchain when you've got one that works? Like you said, I mean, it's in production, there's applications running on it. There's a roadmap. There's a team there.

I mean, it totally makes sense to be building it on Ethereum. So let's talk about economics a little bit. So as we mentioned, there's a Gollum network token, GNT that is the unit of account, the unit of value with which you pay for computing resources. And so in this, and I believe there's a decentralized marketplace as well. Essentially, yeah. Okay. So in this marketplace, you have buyers that are looking to buy distributed computing resources and sellers that are selling access to the resources,

the unused resources on their computer to those buyers. What are the types of, aside from supply and demand that may affect the price of a computing resource, what are the other parameters that might affect price? I'm thinking of things like geography, latency, the power of the computing resources, maybe like the rate of success, of the resource in question.

So what goes into basically like if I'm providing a computing resources to the network, how do I compare to others, for instance, on what parameters? Yeah, I think we will know it once we are in production because only then we will see

what computers do we really have in the network. We can have some idea of what is in small or medium data centers we accept to join. But apart from that, we don't really know what hardware specifically will join the network. And of course, the other level of complexity is added by GPU computing, which is a little bit more tricky than CPU computing, but which we are going to support for sure. you're working on it. I believe, like having said that, I believe that market like Gullim

should eventually convert with the price to the original cost of computing. So this is like a little bit like like perfect competition situation. Of course like this is this will not be like a single price because although I think like the hardware price, is quite similar around the globe, then electricity price for sure is not identical. Latency, of course, is not identical and latency might be an issue. I don't think this will be much of an issue in a very early integrations with high throughput computing, but yeah, even that that might be an issue and later on definitely latency will be an issue. So where you're geographically will be important not only because cost of electricity but also because of latins.

And then, of course, on top of it, we believe that running your hardware with Goldman will be much more efficient in an upkeep effort needed than anything else also like marketing effort needed. But at first, like, the cost of labor will also play role. So that was kind of a long way of telling you, we don't really know, we don't know yet, but the price should be quite low because we want to create highly competitive market. Okay. So on your website, there's a video where you're sort of explaining the high level vision of Gollum and one of the things you mention is this idea that you want to create a network where anybody can participate and there's sort of this underlying feeling that okay right now like we

send our distributed computing executions to software companies like Amazon and Google and it would be ideal if that could be a bit more distributed and not so centralized there. However, it seems that if Gollum is to be successful that companies like that companies like Google, like Amazon and others with massive amounts of computing resources might also want to participate in these networks to make money, right? And with just the, what would happen if they were just to come in and flood the market and bring the price down, how would I be competitive to, you know, with them? Is that something that's desirable? Are there mechanisms in place to prevent this sort of thing? Yeah, if AWS or Google,

join Gollum, that would be great, and they should, I believe. And in fact, very early in the project, like, two years ago and more, we thought about Gollum mostly as an oversharing computing network. So we thought about Gollum as running predominantly on home computers, on people's desktop. and laptops, maybe sometimes a little bit more sophisticated like gaming computers or rendering machines in someone's cellars. But still the assumption was that this is basically directed to individuals in a computer-sharing mechanism.

And although we still believe in this idea and this is something that keeps motivating us and we want to create technology that will make it possible, I think that we realized that actually this technology is of much interest of small and medium data centers. Also, and not only miners afraid of proof of stake in Ethereum, but also like just regular data centers. And if you think about it, it makes sense not only because they have unused capacity, which is somehow obvious, they have, but also because like true edge in in running data center now, as far as I understand that, small or medium data center is, it's not because you have like some hardware, but because you have some hardware and you know how to put it into use.

So you have already some software, your software or your client software that needs a lot of a lot of computing. And in fact, your ad is with this software, not necessarily with the hardware, which you need for this software. So perhaps we can give them even greater ad, but just giving them more hardware. Yeah, I think there's also one possibility,

which is definitely on the roadmap, is that if Gollum is easier to use than those other systems, pricing is important, but those systems are also still kind of hard to use, and the billing systems are not fun. It's also important to remember that the larger players have tiered pricing. So while the majority of their clients may pay one price,

some of their clients who have high computing needs, for example, with Amazon reserved the instances, even when they're not using them because the price is lower. And now they've basically reserved something that they may not use, but it's still an advantage because it's cheaper than if they use it on demand. And the similar thing happens with the larger data centers and probably with Google too. My point here is that there is some median market price, and the larger players have discounted services for large.

clients. Those larger clients who are committed to that, if they're paying less than the market price for on-demand computing, then they may actually run Gollum on their nodes, even though they're on AWS or software or Google or whatever. And those data center companies can't tell them you can't run Gollum. You know, they can't go and block Gollum on their firewall. So the thing is, is in those environments, Gollum is unstoppable once it gets launched. And there's an incentive to actually run Gollum. So the real question is, are people going to want to run it? Is there enough incentive? And I think we've talked to a lot of providers and they want to see our software running. Universities want to use it so that the universities can lease out their idle compute hardware.

And it'll be really interesting to see how large the non-AWS, non-Google, non-Aazir market is. It'll also be interesting for us to see how many people run Gollum on those environments when they have discounted pricing. So can you guys tell us a bit about the current status of the project? I know you had a release schedule and I think brass release is supposed to be coming up soon. what can we expect then and when is the time when it will come out? It's coming soon, but we are not giving any dates any longer. Like we really believe back in November 2016 that six months is double. But obviously it was not.

So we are really advanced and we did really a lot over the last six months. months but still we don't feel that this is ready for all these yet. What we are aiming for at the moment is is having beta. What we have like those are just names but what we have at the moment we call alpha because still it's not stable enough and uix is not good enough for that to be to be like usable in like anything close to real use. This is still like for testing mostly. But we hope that we will be able to put together really,

really soon the version that we intend to name beta. So this will be Brass-Goldem beta version. And actually, this software should be of much use for our first potential user group. And actually, the plan is now to start with the onboarding of the first first. early adopters with this better version so make that in not really in a production but in a production like environment to have like real users over a real network which we will probably subsidize a little bit just to make sure that the user experience for the user is great and and this this is basically

how we want to create some scale to to finish all the testing and all the that we need to secure the network before the release. And by release here, I mean switching to the Ethereum Mainnet and GNT for transactions. So what's coming soon is, I hope, this battle release. Sometime after that, production release, like Brass Golem 1.4. oh yeah but of course like the only true answer is like this is ready when it's ready and yeah and as for the progress we made over the last six six months if you we can tell you more about that but this like a little technical so yeah maybe we are running quite late

but maybe Alex you can just give us a three three minutes like you know highlight things of of the technical progress you guys have made Just at a real high level, somewhat technical, we replaced, in the process of replacing a QT interface, QT-based, which is like sort of, you know, native user interface from Python code, with Electron. And Electron is like this sort of web-based user interface for desktop apps. GitHub uses it now and a couple of other large projects. Mist is built with that. But it's a little easier to do development changes and design changes. We have also spent quite a bit of time on the storage layer.

A lot of storage projects that we're looking at are content delivery systems, and we really need something for like these sort of on-demand compute swarms. We're evaluating those different storage systems. subsystems and understanding like how they're going to fit into Gollum in the future and sort of curating it to something that's going to work effectively for brass. Also working on specifications so that we can have effective security audits. We'll also, we've drafted some ideas for smart contracts for the providers. I mean, GNT is really this sort of token for machine to machine transactions.

You have somebody's computer, you know, who knows, could be in Montana, and you have these other computers all over the world. The providers may be in a data center and they don't want to put their private keys on their provider machine. So we're looking at how to do the payment system in a way that providers can also sort of have remote, remote management of the of the tokens.

And the other thing is we're integrating the dev P2P library, which is the one used by Ethereum, for the underlying peer-to-peer network that Gollum is going to use in Brass. And this is so that we could scale up discovery a little easier. And that's going well. We were supposed to do some of that a couple of releases later, but found that it's a little more time-effective to do that now rather than later, which is cool because hopefully it'll mean, well,

I don't want to talk anything about timelines, but the releases that come shortly after Brass, hopefully maybe they'll have more functionality than what we had in the roadmap. But no no promises until we get it done. Cool, I'm sure people are excited to hear that. Thanks so much for the update. Now I have one last question before we up up. So Julian, can you share like to what extent has a Golem community formed with people

who aren't just, you know, interested in the price or own some tokens, but you know actually are interested in building projects on Golem? and do you have some exciting things happening in that area? Yeah, definitely. So, of course, like the community is, of course, like, most the crypto community at the moment. So people who are very, very fascinated about, about the technology, but coming from crypto world.

Also, some token holders. And this is great community, but not necessarily like, that they will be using Gullum for doing any computing. But we see more and more projects coming and some of them really, really great that are interested in building on Gullum. And we see that both from rendering industry, so the first niche we want to work with,

but also from data centers industry. from all things that have something like high-thribute computing. So we had already some very early talks with companies processing larger modes of financial data, larger modes of satellite data. So I would say that there is more and more interest of how Golem can be used. I'm pretty optimistic about that. that if only we are able to deliver technology that others can use, they will use it. Cool.

Well, thanks so much for coming on, guys. It was a pleasure talking to you and learning a bit more about what's going on with GOLM at a moment. Thank you very much. That was pleasure. Yeah, thank you as well. And thanks so much for our listener for once again tuning in. We are going to be back next week with another episode. then of course this part this show is part of Let's Talk Bitcoin so you find this show and

all the shows on Let's TalkBitCon.com and if you want to support the show what you can do that does this a huge favor is leaving an iTunes review and so new people can find the show. Thanks so much and we look forward to being back next week.