EPISODE 1658

[INTRODUCTION]

[00:00:00] ANNOUNCER: Akash Network is a decentralized cloud computing platform that leverages unused compute capacity around the world. It makes this capacity available to others and provides a decentralized peer-to-peer model for managing and paying for these resources in an online marketplace. Greg Osuri is the CEO of Overclock Labs which created Akash Network. He joins the show today to talk about AOS. 

 This episode is hosted by Lee Atchison. Lee Atchison is a software architect, author and thought leader on cloud computing and application modernization. His best-selling book Architecting for Scale is an essential resource for technical teams looking to maintain high availability and manage risk in their cloud environments. 

Lee is the host of his podcast, Modern Digital Business, produced for people looking to build and grow their digital business. Listen at mdb.fm. Follow Lee at softwarearchitectureinsights.com. And see all his content at leeatchison.com.

[INTERVIEW]

[00:01:12] LA: Greg, welcome to Software Engineering Daily.

[00:01:15] GO: Thank you so much for having me, Lee. 

[00:01:16] LA: First of all, tell me in your own words what is it that Akash does and what value does it bring to an average application developer? 

[00:01:26] GO: In simple terms, Akash is described as what we call a supercloud. A supercloud is a layer that sits on different cloud providers. Be it private. Be it public. Be it hybrid. It doesn't matter. By sitting on top of this, it gives you a simple and easy interface to access different cloud providers. And an interface is very familiar for container developers. It looks and feels like a Docker-composed interface. 

Additional advantage is Akash providers is a marketplace as well. You can actually discover new providers and discover new resources and set the price that you're willing to pay. That translates to enormous discounts. In some cases, for compute loan, it's about 10 times cheaper than what you would otherwise pay on Amazon on-demand. And in cases like GPUs, it's access, right? It gives you access to high-density GPU speed. A100 or H100s that you can't get on-demand on any of the current cloud providers. But Akash is a place to get that. 

And the way Akash makes that happen is by tapping into providers that don't normally have a way to sell their computer to a customer. Akash gives you this software stack, which is Kubernetes-based, to expose your computational resources. Be it CPUs, GPUs, bandwidth, storage and whatnot in a safe and easy manner. And it gives ability for you to conduct billing, charging and all the good stuff that you normally need to do as a cloud provider for free. It's open source. Effectively, anyone that installs Akash controller in their Kubernetes clusters can effectively become a cloud provider on day one. 

[00:03:12] LA: It's not just cloud-based resources. These could be on-premise resources as long as, of course, they're accessible. That can be sold or leased, I should say, is probably the more correct term onto the network and used by other people.

[00:03:28] GO: Yeah. If you have data center or if you have some capacity in a colo sitting somewhere, as long as it has a public IP and good internet, it can become a provider on Akash. Sometimes even resources that may not be in an environment with good internet like your home computers that serving - that are used to play games. Has a beautiful GPU, 4090 Ti, which has 24 GB on-chip memory, which is very valuable for machine learning. Even things of that nature can perform an asynchronous background job on Akash. 

[00:03:59] LA: I think back to what some 20-so years ago, there was a screen saver that you could get that install on your local computer that allowed you to run SETI algorithms looking for extraterrestrial life and part of the SETI program. And that was something that could use the resources on your local computer in a distributed way. It's like that but on a much larger scale. It's for a much larger use applications.

[00:04:25] GO: It's SETI for general purpose compute. 

[00:04:28] LA: Good way to put it. It's a marketplace. And there are other marketplaces out there in marketplace-like things. What I think about compute marketplace, the first one that comes to mind of course is AWS bot instances, which is kind of their way of selling their excess capacity. But this is very different. This is arbitrary people able to sell spare resources they are on an open marketplace to other people. What other differences does it have from traditional marketplaces or traditional cloud providers?

[00:05:01] GO: Yeah. Spot market is for Amazon resources only. This is a spot market for global resources. There a big difference. Akash is considered to be the Airbnb of compute, right? Essentially anyone with compute can provide - can offer their resources to be leased out by anybody that wants. And it's fully open-source. It's the first open-source cloud. A lot of folks that contribute to Akash, we have about 480 contributors, come from all different places and having no incredible ideas. It's maintained and run by an open community. 

Every feature of Akash is discussed in the public. And every item on the road map is approved by the public. It's not some private company sitting and setting the rules. It is public that owns and uses the software actually contributing to the network. And it's permissionless and decentralized. There's not a single company controlling the cloud. It is people that use the network that control. All the way from setting the amount of margins the network needs to attract in order to sustain ourselves. To the deposits people need to pay. All that can be changed with a governance proposal. Anyone can make a proposal to change the network anytime without permission. And as long as community members, the owners of the network approve, it goes through. That gives a lot of confidence for platform builders, right? Especially builders that are building PaaS applications on top of Akash that add additional advantages or additional sort of capabilities and repackage and sell it to the users have complete confidence that their API key is not going to go away in case a board member at the cloud company decides is a good business to go after. 

You have this fundamental guarantees that Akash provides that cloud providers don't provide. And it's cost-effective. The cost is very hard to beat. And the reason for that is it's transparent pricing. A provider and a tenant engages using the Akash blockchain in a peer-to-peer manner. That means I - if you are a provider and I'm a tenant, there's no middleman extracting any fee or providing additional value. It is directly interaction. And whatever you make, you keep minus a small percentage that you agree to pay back to the network for the network to maintain itself. All that being transparent, being open, having the world's first reverse auction mechanism for pricing, be it open source, gives a sense of security, sovereignty and cost-effectiveness that the traditional cloud provider cannot provide. 

[00:07:47] LA: There's nothing centralized. There's no owner of a network that manages all of this. It is all completely distributed across your users' compute environments, the infrastructure to run the company.

[00:08:00] GO: Right. There are 70 different providers last I checked. I think there are two more providers today. It's growing at 20 providers a month I believe now. Soon enough - I mean it's already the largest distributed network in terms of regions. It's got more regions than Amazons and Googles of the world. By end of the year, we anticipate at least 500 to 600 new regions. 

By end of the year, it'll be the largest distributor network in terms of physical distribution. You can imagine the edge computing applications that can be finally enabled on Akash. And Akash is the only place to get on-demand H100s or A100s, which happen to be the most advanced chips in media mix. And H100s are very critical for machine learning. And they're impossible to get on Amazon. 

It's already starting to offer resources that the traditional cloud cannot offer because of the Akash model. And imagine these massive multi-cloud Kubernetes clusters running and operating in a self-operating manner with zero DevOps people. From just a scale standpoint, it's thousands of nodes running by themselves. And so, it's a fascinating technology what the decentralization can bring. And it takes a long time to build such systems. But when you see the systems in action, it hasn't gone down a single time the last 3 years. Only time it stopped producing blocks or it pauses block production is during planned upgrades. Where during upgrades, the big challenges, because it's decentralized, everybody needs to upgrade in a reasonable amount of time. And if they don't have a consensus, the system doesn't process new request. Those are the only times you see degradation in performance. But it doesn't go down at all. That's a very, very reliable piece of software. 

[00:09:49] LA: Walk me through how this works. I've got a general idea but I just want to make sure that the listeners understand how this works. I, as a consumer of compute, need compute resources with a certain level of reliability and availability. I'm running some application on it that needs a certain level of availability. Now I can use any of the resources that are available on the network. Not all of those resources are going to have any sort of guarantee of availability. You yourself said you can use home machines even to be on the network. And they are notoriously not high availability. But, certainly, even companies who are selling excess capacity, they need their own capacity, they take it back. The resources aren't available for the network anymore. In general, the consumers of the resources need higher-level guarantees than your producers typically get. How do you manage that? Who's responsible for that, first of all? Is it the consumer that's responsible for managing that? Or is there something in the network that helps with that? 

[00:10:53] GO: Yeah. The way you look at Akash is, like I said, Airbnb. That means you get a wide variety of providers. All the way from tier 4 data centers. Be it that Equinoxes the world, Lumens of the world. Even Amazon as a matter of fact. Some Amazon servers. You get that level of providers to Joe's Datacenters. Anyone with a compute. 

You get a wide degree of performance and reliability. The process begins with the provider having a data center, having a machine online and registering the machine with Akash Network and then getting audited, which is a very important step. There is a decentralized set of auditors. These are different companies. Overclock Labs is one. There's another company called Moultrie, which is founded by former service members, Department of Defense employees. Because they implemented Akash and DoD and had incredible results and they stopped. They quit DoD to start their own company to do Akash stuff. They are one of the auditors. And you have several other professional auditors that go audit the claims of a provider. 

What that looks like is if I'm Equinox and I say I have you know 100 servers in San Jose in this data center. This data center is tier 4, SoD compliant. HIPAA compliant. All these attributes they advertise as a provider on Akash. The auditors go and validate the attributes. If you have SoD compliance, you got to show documentation. And they post their outcome whether that attribute is validated or not. 

You as a tenant, you define your application in something called an SDL file and you say, "I want an up time of 99.997 minimum. I want my data center to be SoD compliant or HIPAA compliant. I want my data center in San Jose." And then you post that request as a buy order on the blockchain. And blockchain effectively is an order book here. And providers that satisfied these requirements can bid on your order. And if you say I want a minimum two or three auditors, and by the way, these are the auditors I choose or trust, you will only see bids from providers that pass the audit tests. That way you have indemnity and have guarantees in terms of performance and SLAs. Right? 

And after that you - again, the choice is yours whether you want to - you get presented with a bunch of options that pass all the requirements and you choose to create a lease. Once you create a lease with that provider, the interaction then becomes peer-to-peer direct. What that means is the Akash blockchain is only there to facilitate the payments and facilitate the lifetime of a release. But your interaction is directly with the machine, with the provider. There's no performance degradation or any abstraction on top that's extracting value that you not know of. And it's fully open source. 

Akash is very unique in a way because it does have capabilities to filter out providers in a granular manner. And also, on the provider side. What are the chances that safety of providers is important? There's moderation APIs where the provider optionally chooses to use and that filters out unwanted type of workloads. And that API hooks up to federal databases and all kinds of good things that tradition Cloud providers use that most people don't know. But a lot of people use these databases to check the content. You don't want nefarious actors using your data center, right? Then you're in trouble. 

We have all that sort of enterprise-grade capabilities Akash offers. Most crypto networks don't really look at these aspects of it. But Akash is built different. It is built for the enterprise. Safety is paramount both on the provider and the tenant side. But also, keeping it censorship resistant and keeping it open is also important. But having the control, whether you're a provider or tenant, is also very important. That's the incredible balance Akash plays really well.

[00:14:58] LA: It's very much an enterprise-grade compute environment is the type of environment you're trying to provide. Users request the level of performance they require. And it's a rational expectation that that's what they get.

[00:15:14] GO: Absolutely. It's very enterprise already. I mean, we have few enterprise healthcare companies in fact that deal with HIPAA-compliant data choosing Akash. Obviously, DoD has implemented Akash internally. They're not using the network but they're using a fork of Akash. Because Akash is a glorified multi-cloud management system. You can actually use it for free. It's a very permissive license. 

There's a lot of implementation of Akash that people use privately, publicly. And cool thing is if you have a private version of Akash, you can connect to the public version of Akash. And that way you can also have interoperability between different networks. You can use Akash in any way, or form, or shape you see fit. And that's what makes it very, very exciting.

[00:15:53] LA: Cool. Cool. Let's talk about your typical customer then, which I know that there's never one typical customer. But is your typical customer large enterprise? Or is it more startup or mid-tier companies? And let's start with the customer that's the consumer of resources. And we'll talk about the providers later.

[00:16:14] GO: Yeah. Tenant, we call them.

[00:16:15] LA: Tenant. Okay.

[00:16:16] GO: Yeah. A typical tenant is a funded startup. That's our typical company. It's mostly machine learning. Spending anywhere $50 an hour to $100 to $200 an hour. $50 an hour would look something like 10 H100 machines, right? Roughly. And so, usually these companies are not large enough to make large deals that can get them the supply they need. 

Because today, if you're a machine learning company, the options are very minimal for you to get on-demand GPUs. Amazon will not give it to you and unless you do a multi-million dollar, multi-year lease with them. No cloud provider will give it to you. Be it Lambda. Be it Lambda. Be it Korvi. Be any of the GPU clouds we talk about that's. Just the demand is so high and supply is so low. That's just the nature. Akash is the only place to get it. 

Even Akash, we're running like 95% utilization for some of the high-end chips. But, unfortunately, Akash has this amazing way to attract suppliers that most cloud providers don't have. I'll get into that in a bit. But supply is coming in every month but it's gone pretty much the moment it gets. 

A lot of the users on Akash, there's really no other option to go. And there was an article recently in Semaphore, I believe, that did a very good op-ed on AI being saved by crypto. For the first time, we actually had a positive note about crypto. But in that, they interviewed a bunch of users on Akash. One of the good stories was this Columbia student who was a researcher and he wanted to get some compute for A100s on Amazon. He couldn't. And he came to Akash and he could get it at the price that he wanted. 

And so, there's a lot of University kids. There's a lot of researchers. There are a lot of startup companies that are simply not served by the larger providers or typical customers today. We don't have large enterprises in the sense the typical enterprises that consider large yet. But we're seeing more on the supplies side. In fact, large enterprises more than the demand side. Because large enterprises tend to have large investments with data centers. And they also tend to have a lot of underutilization. 

We're seeing billion-dollar companies starting to supply to Akash. I mean, Foundry is a big company. It's a public company. DCG. Owned by Digital Currency Group. That's a big provider on Akash. But we're also starting to see similar, in fact, larger companies on the supply side. I think a lot of these companies are also realizing Akash could be used on the demand side as well. But we are yet to see a big implementation. 

[00:19:04] LA: That makes sense. Large enterprises providing resources to a funded startup is the typical model there. 

[00:19:13] GO: Typical model.

[00:19:12] LA: Good. Good. And AI being a primary use case. But it also sounds like things like HIPAA-compliant and other regulatory-compliant restrictions are also kind of a niche that work well with Akash as well.

[00:19:28] GO: Yeah. There's one interesting use case that I thought was very fascinating. It's an enterprise company. They are a healthcare data management company with a twist. The problem they're trying to solve is medical records interoperability. The way they're solving that is by providing, like you and me are users, a place to store our healthcare data with the keys that we control. 

We have guarantees that no one accesses the data except the ones that we agree. And you can be sponsored by the insurance company but the keys are with you. You have complete control and sovereignty when it comes to your data. That way, there's less liability on the insurance company. Liability is very expensive for insurance companies and healthcare companies in general. They pay a lot more for liability than compute. 

By de-risking, by removing liability and guaranteeing Akash the sovereignty in the end because it's controlled by a key, there is enterprise play there. There are interesting use cases like that we're seeing from large enterprises that, "Hey, if I can reduce my liability, it's a great opportunity for me." 

But innovators like that are looking at Akash but we haven't - obviously, healthcare is one of those highly liable industries. We don't see that type of use case in other industries. But it's interesting to note. But most of our - I would say 70%-plus of the revenue today Akash makes comes from machine learning and AI. 

[00:20:57] LA: One thing else that you do that is unusual for a startup, I'll just say it that way, is you post all your financial and resource stats on your website. And when I was doing research for this, I came across your stats dashboard and started looking at it. And I thought it was rather interesting to see that level of detail on your website. Can you tell me - I'm talking about the stash.akash.network. What am I seeing when I look at this dashboard? 

[00:21:24] GO: Well, Akash is fully open-source. Every aspect of Akash is open to the public including the data. And that's the best part of Akash. Because you need to know exactly how much it's being used. That way you have a better idea. First thing you see is obviously the active leases. Active leases are active applications that are running on Akash. 

Right now, about 1,700 applications that are running actively or 1,800 applications. And it's growing really fast. You see the growth. Active providers are 76. I said 70 because I guess a bunch of people came today. You have your daily spent. How much money is being spent on Akash? How much different - what blocks there are? You can pretty much know everything you need to know about Akash network just by going to the stats dashboard. 

Accountability and radical transparency is key to Akash's success. The reason why we are able to get enormous success with attracting contributors and capital at the same time is the radical transparency system. Building decentralized systems is extremely hard. Because there is no central entity that is responsible for a single aspect - that's why when you have a community, the foundation for a community is transparency. Not only the code but also the data needs to be transparent in order for this to work. That's why Akash is radically transparent. 

[00:22:51] LA: It makes perfect sense. And I'll tell you, even other companies who claim to be transparent, who claim to be peer-to-peer don't have this level of transparency. And so, it's really neat to see. And anyone who's listening to this, I highly recommend. Or is thinking of using Akash, take a look at this dashboard and you can see the level of transparency that Greg is referring to. And it's stats.akash.network. 

Looking at it, a couple things that come to mind is, first of all, still the numbers aren't huge. You're a growing startup. You're just new in the space and growing. But I did notice that there's a big change in your growth curve just a few months ago. I think it was in October if I'm reading the graph correctly, where you had a huge hockey stick upwards in growth. What happened in that time period that caused that hockey stick?

[00:23:42] GO: GPUs came online.

[00:23:43] LA: Ah. Makes sense. Okay.

[00:23:45] GO: Until then, Akash was computer-only. And GPUs were in the Testnet for about a year. Testnet is essentially a private version that's limited in terms of its use. Because decentralized software is very hard to upgrade once you put it out there, because talking about thousands of nodes that needs to be upgraded every time you push an update. Coordination cost is very, very high in decentralized software. That's why there's quite a lot of emphasis on stability pre-launch. 

GPUs were in Testnet for about a year before they made it to Mainnet. And Mainnet obviously now is live. And we're still in like very early stages. I would say the next two months' revenue is going to be at least triple or quadruple based on what we're seeing in terms of supply. We are supply constraint now more than demand constraint. 

As you can see, the GPU supply is also not that high. Because a lot of the GPUs on Akash are high-density GPUs that are a very high demand. But there's an incentive program that's going online. It should be approved by the community this week. It looks like we have all the votes that we need to pass that. That's offering about $5 million in incentives to providers to put supply on Akash as an input to the larger incentive budget. We have currently about 4, 50 million dollars in incentive budget that's on-chain. that'll be used to incentivize supply. You're going to see quite a lot of GPUs flood Akash Network in the next few months starting off with about H100s that are coming online next week. These are very expensive rentals, for example. 

I think Akash will start offering H100s at around $2, which is the cheapest ever you can find. For comparison, Amazon offers H100, if you're lucky, for $12 an hour. Akash is doing the same exact chip for $2 an hour. It's a very attractive price point.

[00:25:45] LA: Yeah. Definitely. Definitely. My follow-on question, which is mostly answered now, is, is that growth - that hockey stick growth sustainable? And I guess the answer to that really is, as long as supply is available, it's sustainable. And so, you talked about in the near term, let's say 3 to 6 months, processes for getting more supply and growing your network. And I can easily see the hockey stick going on from there. How do you keep this going long-term with that sort of a growth rate? Or a sustainable high growth rate? 

[00:26:21] GO: Right. By optimizing the channels that give us the growth. The growth is coming from different integrations that we have worked very hard to get from platform services, right? PaaS's or platform as a service companies. The recent spark in growth came from this company called Spireon, which is a platform as a service. And they are smaller deployments, fast deployments. And they've given incredible amount of active leases, if you look at that curve. 

Every month, we have one integration going live. This month, it was Agora Labs, which does Jupyter Notebooks as a service essentially. It's great for college researchers where you want a collab on your machine learning code or big data code. You can spin up a Jupyter Notebook on-demand on Akash using Agora Labs for relatively low cost. It's very challenging apparently. Because today the only option is Google Collab. And it's very slow. Because they give you cheap GPUs. Because big GPUs are very hard to come by. 

With Akash, you don't have a problem. And on the supply side, when the demand is fairly high - right now, we have 95% utilization for A100s. That makes a case for suppliers to have that some degree of guarantee that, "Hey, you can actually supply and get paid with utilization even without incentives." Right? 

Right now, our utilization is without incentives. We're experimenting with incentives to attract big players. Because a lot of the companies with huge amounts of supply don't normally talk to you unless they see 10 to 15-million-dollar potential outcome annually. I mean these are large companies. 

To attract these large companies, you need these on-chain incentives. But the incentives will enable the integration. And post-integration is the utilization. To sustain, we need ecosystem and distribution. And that's what we're focused on right now.

[00:28:23] LA: Cool. One of the problems with a democratic distributed management model, like what you have right now with the Akash, is it can be slow at making decisions. The more democratic your environment is, the less centralized control, the longer it takes for consensus to be created and all that sort of stuff. How do you deal with that? 

[00:28:49] GO: Good question. Open systems have high coordination cost. The decisions are slow. They're communal. But they give a degree of accountability that you don't normally get with a close company, right? The way I look at it, open versus closed as such. With closed, you can go fast. With open, you can go far. That's why any sufficiently important technology that has wide adoption is always open. Starting off with the internet. With operating systems, right? 

What Linux did, do operating system purification, Unix couldn't even think about. What internet did, closed systems couldn't come close. What Kubernetes did, VMware couldn't even come close. Open source always goes so much more further than closed source can but it's so much slower then closed source could. 

That's why we created efficient mechanisms. And that tends to be a huge area that we all live on. Any open-source maintainer will tell you that Akash has pieces that we took from Kubernetes, pieces we took from lot of very highly-functioning groups. And we added incentives on top of it. 

With Kubernetes today, we have a contributor churn problem. Contributors don't stick around a lot. When you don't have people that - because they have different careers. And most of them are coming and doing part-time work more than full-time work. Except for companies like Red Hat and Google that have full-time employees. But for most, others are just part-time work. Instead of having these part-time roles, have a sustainability fund, like a public goods fund that funds developers' lifestyles to work on Akash full-time. 

We have about $10 million right now in the sustainability fund. And that's increasing. I think adding about $2 million a month or something. And that's funded by taking a portion of the hosting fees and so on so forth. And that fund, actually anyone with ideas can come to a special interest group. There's a special interest group for each aspect of Akash Network. Be it providers. Be it clients. Be it blockchain. Be it - all different aspects of Akash has their own group and propose a feature, write a proposal, get funded by the community and implement that way. 

We have eight different companies now working on Akash that keeps sustaining. As long as there is money in the fund, you should be able to build something and dive something. 

By decentralizing decisions by creating smaller actionable groups, and somehow those groups all tie together with a larger group. Really focusing on empowerment than enforcement I guess is a way to look at how to move fast in decentralized systems. And a little bit of magic, too. It's very hard to say what exactly worked. Can we repeat that? What we did for Akash? With others? I don't know the answer to the question. And it also has to do with with culture that you said. 

A lot of times, a culture comes from the top. Building a community is so powerful that, when it works, it becomes remote. Or when it doesn't, it becomes a liability. Or there is no community I suppose if it doesn't work. 

But certainly, culture in a way that community becomes a moat that drives you further is what makes something more successful or not. And a lot of incredible software I've seen, Docker failed because they couldn't create a community to sustain themselves even if they had great product.

[00:32:16] LA: Right. And it sounds like you're actually taking a unique approach there to other open source such as Docker and other open source examples, too, where the general philosophy of open source often also includes nonprofit. The idea being that we need donations to keep this going. Donations of time. Donations of resources. Whatever. But that's how we maintain this nonprofit, decentralized open- source platform. 

Usually, in a lot of those cases, look at things like Redis for instance. You get usurped by one or two larger companies and become controlled by those entities. But it sounds like your approach is different. I don't know if you're legally nonprofit. I'm not talking about that. But you embrace bringing money into the system and using that to pay and fund the building of your system. You're a real company that has cash flow, inbound, outbound, to build and maintain the network. Yet you're still a distributed open-source platform.

[00:33:23] GO: Right. Overclock Labs is the company that created Akash Network. And it's cash flow positive. It has 20 employees in the US fulltime. We build everyday open-source software. And we're doing very well. And the way we do that is through decentralization. And there are seven other companies that are similar to Akash or Overclock. Each of them have their own profit value capture mechanism. These are incorporations. These are for-profit corporations. Tight these are not C corporations. These are not 503c or any of the nonprofits. We pay taxes like normal companies. 

Creating sustainability has been a big passion for me. I've been in the space for a very long time. Very early contributors to Kubernetes' ecosystem. 2013, 2014-time frame. I've seen the rise and fall of Docker. I've seen so many amazing examples of incredible products that couldn't figure out how to sustain themselves. And more often than not, a misguided approach to altruism or the value system where they want to be open and free but really have no options because the foundation is nonprofit. 

That becomes like - yeah, it's cool for a bit. But if you want to pay bills and if you want to have employees that can contribute to their vision, that pay bills and mortgages, you need to think a little bit beyond the good of your heart and think about how to sustain a business. And more people that you help out and have vested interest in making Akash successful or your open source successful are the best ways to support your product. 

And the way the decentralization and the tokenization has given so many different tools for open source developers that were not available during the days of Docker or Kubernetes that Akash is able to take advantage of and create a very sustainable model. 

I mean, Silverlinings is an incredible publication that wrote a piece recently on sustainability using Kubernetes off Akash. Incredible way they really angled it and did an in-depth case study. I recommend people to read that. But, yeah, you have to remove your biases aside and think of - yeah, crypto has got all the like NFTs and DeFi stuff that's - whatever. But cases like Akash where we don't deal with any of the speculative aspect of crypto but actual functioning aspect of crypto is worth taking a look I believe. 

[00:35:50] LA: Yeah. You've separated the concept of nonprofit from open source or from a community-driven open source. You've separated those concepts out, which I think is actually rather unique. You don't see a lot of companies or a lot of projects that have done that approach. That's cool. That's very much - like you say, it's something that wasn't really possible before blockchain really became a viable way of transacting business. Otherwise, you always have some sort of centralized system you have to deal with. But you've actually been able to remove that completely from your equation. There literally is nothing centralized. Yet you have an organization that's making money and making other people money but still is decentralized.

[00:36:36] GO: Correct. There are a lot of people that use Akash, make money. And that's the best part of it. And we're not forced to take the typical route that open source companies are forced to, which is open core, they call. But open core, if you look at it, sure, there are - can make a case saying that, "Look, open source software is great. But in order to go to enterprise, you need support, indemnity and a whole lot of things. And identity, and Integrations and all these noncore product work that is valuable to enterprise so they're going to charge money." Right? 

I get it. There's a case to be made. But, again, the incentive misalignment is so great once you start - as a company start focusing on making money using these alternative mechanisms. Be it support or be it additional features. You got to be focused on making money. It's very hard to split focus on staying too decentralized and working on the open source software. 

I've seen so many companies. I mean, HashiCorp, for example. Recent move with their licensing change that they had. I worked on HashiCorp Terraform since 2013, 2020 timeframe. I mean, I was fixing a lot of bugs and a lot of work I did because it was MIT license. Because it was permissive. Or Apache. I don't remember. But it was a permissive license. And I felt a sense of ownership for the software because I was a contributor. 

But now, HashiCorp as a corporation has fiduciary duties to increase its shareholders' value, right? And it's going to focus on activities that's going to do so. Because it's set up that way. Whereas Akash doesn't. It's not an organization. It's a network. An open network. There is no company called Akash Network. And it is a network owned by the people that built it. 

Overclock happens to be one of the biggest token holders and so several other companies that are doing - have enormous amounts of tokens that all have a vested interest in making sure Akash is successful. And protocol set up in a way that there is no single entity that can extract value or go against the values of its users, right? And every time there needs to be a change, anyone can make a proposal and anyone can change it as long as they have enough votes of its users. 

That open model where you still do have some way to capture value but that's more distributed and that's more even and ensures that equality from the ground up from its design is what decentralization brings to something like Akash. 

[00:39:05] LA: Makes a lot of sense. You brought up your token. You actually have what you call the AKT token. Do you want to talk a little bit about what that is and how you use that internally? 

[00:39:15] GO: Sure. Akash is what we call an appchain. An appchain is essentially a blockchain network that serves a single application. It's very different to what Bitcoin or Ethereum looks like. Ethereum, for example, is a blockchain where you can have multiple DApps hosted on the blockchain, which essentially run in a deterministic fashion in something what we call smart contracts. 

Each of these smart contracts again have their own tokens and have their own functionality. Whereas Akash is a blockchain that only serves Akash Network. And the consensus mechanism that secures this blockchain. Because a blockchain is effectively a database anyone can write to, the question becomes who can write in a way that is temper proof on not tampering other records? That's why you need a consensus mechanism. 

The consensus mechanism is called proof of stake, which essentially says that anyone with a token can vote on this consensus round. And the data with the highest number of votes get into the blockchain. Each token has one vote. The primary purpose of AKT token is to secure the blockchain. Without which there is no blockchain. 

And the way it works is by set of validators, which are essentially gigantic machines, then maintain the ledger. Vote on each block, each six seconds with their tokens that they have power for. And you as a token holder, you delegate your tokens to these validators. There are about 100 validators. Because you as a token holder are not online all the time. And your tokens are not voting unless you're online. These validators of the systems that are online that you can delegate to. 

The consensus mechanism we call in Akash is delegated proof of stake, where I as a token holder, I delegate my tokens to a validator A. And that validator A will vote on my behalf. Delegation is a process where I still maintain the ownership of the token. That means I have a private key that tells what I need to do with the token. If I need to delegate, or un-delegate, or send the tokens, or use the tokens to pay for compute and whatnot, I have complete control over. I'm not transferring the tokens.

And the incentive for me to delegate, delegation also means I'm locking up my tokens for about 21 days, is a portion, a dividend being paid back to me. And the dividend is essentially a portion of the hosting fee that's collected by the network that pays stakers like Akash and a small portion of the commission to the validator. 

Effectively, it's a self-sustaining system. When tenants pay for hosting, which, by the way, they use Akash tokens to pay for as well. There's that secondary use case for Akash, which is a payment token. But they pay using Akash tokens, a portion of that goes to these delegators to secure the network. The more people use the network, the better security the system has. 

Akash tokens are also fixed supply. More demand there is, which is driven by usage, the higher the price goes. And the higher the price goes, the more secure the network is. Because the cost of attack of the network is effectively the cumulative value of the tokens that are delegated. If anyone that has more tokens than what's been delegated can attack the network. It's very important that the cost of attack is significantly higher for the network than a state actor can afford. That's how you look at security, and consensus, and the value of the tokens and whatnot. And the token is also used to, well, incentivize providers. 

Think of an economy, right? The way economies work is, well, you have fundamental value exchange. But beyond that, you have something called subsidies. Subsidies are given. In the United States, you had corn subsidies that were given to corn industries that effectively led to what we have when it comes to sugar. Subsidies can be a good thing or a bad thing. But in most cases, like specialized economies, like Akash, they're generally a good thing. Because you can incentivize the behavior that you want to. 

And then these tokens are also used for paying for development. There's this public goods fund that anyone can make a proposal to and whatnot. There's a lot of activity around AKT tokens that happens in the Akash economy with the foundation of security being this primary purpose.

[00:43:51] LA: Cool. Thank you. We're just about out of time here. But I wondering if we could end with just one question, two variations. The question is what's next for Akash? And the second part of that is what's next for you? 

[00:44:04] GO: Yeah. Next big thing - I mean, major capabilities for Akash that we are really striving towards is parity with the cloud. Right now, we have cloud obviously being Amazon, Googles and Microsofts are significantly more capable than Akash can offer. That includes automation, observability. All the way to manage services, right? Today, use a cloud not for computer but for the services that you get. 

Next big thing Akash have on the road map is managed services. Ability to plug in a Redis that's managed or a Postgres that's managed with someone else. The approach we're taking here is a little different from your traditional cloud. As you know, most managed services on Amazon are open-source software. Redis, or ElasticDB, or any of these services. Essentially, Postgres and Redis. Having a layer that's scalable on top of that's hosted by Amazon. 

Instead of like Amazon extracting all the value, we want the creator of this open-source software get some degree of value. That's the focus on the creator. We want to essentially create a marketplace just not for compute but also for services, so that open-source developers can have a sustainable lifestyle. Because a lot of them, I see, open-source developers, be it creators or Docker, be it - I've seen so many examples where the creators themselves have no way to sustain their lifestyle, yet they want to obviously build their software. And that's unfair. 

The companies that take maximum advantage of this open-source software realm is really companies like Amazon. Very rarely you'll see a Kubernetes that came from Google, right? Very, very rarely you'll see such contributions. But if you think about it, the amount of contributions that Amazon made compared to the value they're able to extract from Amazon is completely off the charts. And they do some open source but not close to the value they're able to extract. 

We want to level the playing field as open-source contributors. And that's a big thing for us. But growth has been incredible. I mean, revenue numbers have been amazing. We're a supply constraint. A big unlock of incentives are going to unlock providers. And we want to be at least 500 providers by end of the year. End of next year, we want to have at least 1000 providers. What we call that is what happens when you're as capable as Amazon, including manage services, including automation aspects of it, and you have about a thousand providers, globally distributed? The goal is to get to 10 millisecond latency for 98% of active internet population. 

We want have providers pretty much anywhere you can think of. And the kind of stuff you can do with such low latency network is beyond imagination. We call that innovation beyond imagination. What can you do when you have such a vast network that taps into not only primary providers? But I don't see a reason why we should stop at just data centers. We have to go into homes. We have enormous amounts of GPU supply sitting in homes, like PlayStations and Xboxes that you barely use, has very powerful GPUs. Imagine what we can do when that GPU is coming to market. And Akash, we want to do that in the next two years. 

[00:47:22] LA: Where are you headed? Where are you personally headed with Akash or the next generation beyond? 

[00:47:28] GO: Yeah. Think about my role in this system, one of the things about - if you look at my GitHub. I've written some libraries that haven't changed in 7 years and they still work really well. Mitch Hashimoto, who's the founder of HashiCorp, uses my libraries a lot. I mean, HashiCorp, Nomad use my libraries, Kubernetes command lines and AWS command lines and whatnot, is because I love creating permanent software. As software developers, we don't enjoy the permanency that engineers and other fields enjoy. 

I love old-functioning stuff. Because there's a sense of legacy and a sense of permanence that this stuff has. For example, I use a film camera. And the camera that I - my favorite camera is built in 1965 and works amazingly. And if something goes wrong with it, I can open it and clean myself. There's a sense of satisfaction in simple permanent, long-lasting technologies. And I want to create something like. And I thought Akash is definitely something that's going to be one of these legacies, right? It's going to outlive Overclock Labs. It's going to outlive me. And I've created a culture I feel in the company. And that culture is propagated across Akash Network is to think very long term and not be forced into short-term benefits or short-term thinking.

And one of the reasons we have our investors also aligned on this mission, right? And Overclock Labs hasn't raised a lot of money, fortunately, because the moment you raise that billion-dollar valuation like Docker did, you're now forced to think short-term or force to align with the investors more than your users. And that fundamentally is problematic for open-source software. And we want to create something that outlives any of these models. And Akash gives you the permanency. And that's where I see myself. I don't see myself retiring anytime soon. I have next 12, 15 - I mean, I don't know how long. But I'm going to continue as long as my cognitive capability allows me to do so. 

And our co-founders and team members, a lot of times you'll see our team, formal employees, actually leave the company and start their own companies but still work on Akash. And that kind of legacy I think is amazing. We can work anytime we want. 480 contributors are not working full-time. But the moment someone decides to work on it, still get paid from the protocol. We created that system that allows anybody to come in and out. 

We're never going to move away from Akash no matter what. No matter how many things come and go and how many interests in life is all about, up and down, right? Life has other - I'm having a child now. And when I started Akash, I didn't have a child. Life is different when it comes to your time and permissiveness. Creating a culture that - 

[00:50:19] LA: But if it's sustainable, it stays around.

[00:50:21] GO: Yeah, stays around. It's a marathon. Not a sprint. Really take that into heart so that we can work on this long-term.

[00:50:27] LA: Makes sense. Thank you very much. This has been a great conversation, Greg. thank you so much for your time. And thank you for joining me in Software Engineering Daily.

[00:50:36] GO: This was wonderful. I had a good time.

[END]