Akash Network - Economics Special Interest Group (SIG) - Meeting #4 #

Agenda #

• Updates from Cheng regarding Incentive Distribution Pool discussions
• Open up to community for ideas or concerns

Meeting Details #

• Date: Wednesday, May 3, 2023
• Time: 10:00 AM PT (Pacific Time)
• Recording
• Transcript

Participants #

• Adam Wozney
• Andrew Gnatyuk
• Anil Murty
• Boz Menzalji
• cheechyo
• Cheng Wang
• George Pro
• Greg Osuri
• Joao Luna
• Pablo Estrada
• Scott Carruthers
• Scott Hewitson
• Zach Horn

Notes #

Scott (Hewi) opened up conversation by calling to action all members of sig economics in participating in discussions that involve community pool spend. Members of the community are the first line of defense to ask questions and provide feedback for potential proposals to make sure they are aligned with Akash’s long term goals.

Cheng then took over to discuss the creation of Incentive Distribution Pool (IDP) and Sub-accounts. He also discussed the progress of stable payments and that code has started on that.

Noteable Questions (Paraphrased) #

• Question (Hewi): Regarding incentive pools, is the focus solely on AI and GPUs right now or is general compute being included in early discussions?
  • Answer (Greg): It all comes down to liquiduty mismatch. Right now, there is huge demand for GPUs and a supply shortage. This is where we are focusing at the moment because we don’t see the mismatch in other areas.

Action Items #

All are welcome to contribute! Anyone who wants to contribute please visit sig-economics in the Discord Server

Transcript #

This editable transcript was computer generated and might contain errors. People can also change the text after it was created.

Scott Hewitson: All right, GM, everybody. This is the fourth installment of SIG economics. And last last time we talked a little bit about stable payments and Chain, discussed his, what he put in to the to the repo for that.

Scott Hewitson: Greg, I’m not sure if you have any agenda items. There are two things that I’d like to talk about one being The, let’s see the hand. the just in, in general, just as As. proposals come into discussions just, you know, making sure that we have involvement and people asking questions and making sure that we’re Involved there and then also talking about the incentive distribution pool. I know Cheng put a, put it into discussions. He’s not here at now but maybe he’ll be in and can talk about it or you can talk about it if you like is there anything else you want to talk about specifically today?

Greg Osuri: Nothing. I just wanted to see the Yeah, I think we progress on the The stable payments and and And incentive distribution. Right. So I don’t have any specific.

Scott Hewitson: Okay. All right. Well, I guess Unless I’ll wait and see if Cheng comes in and he can talk about what he posted in for the instead of distribution pool. But something I just wanted to talk about in general. So in in discussions, folks have been putting in, you know, proposals for ideas or things. And in my mind, you know, say economics is kind of like the first line of defense and government governance and seeing how we can, how Akt is used in a sustainable way. There was a discussion put in about liquidity incentives or economics pool. I responded to it. So, I think encourage people to check that out and see ad comments. It’s, you know, in my opinion I I believe that.

Scott Hewitson: You know, we should be using Akt to progress the development and use of cloud computing resources at the moment. I know people have different opinions but that’s, those are mine. So, yeah,…

Cheng Wang: If you sorry,…

Scott Hewitson: I’d encourage folks to go in there and add to those discussions. And here’s the man Cheng. What’s going on?

Cheng Wang: sorry I’m really a bunch of meetings running back to back to back all these. Thank you for teeing it up and and managing this always hear how’s everyone?

Scott Hewitson: doing well, yes, I was just saying my two big agenda items were kind of just encouraging folks to get into the discussions and talk about when people have proposals because I see, you know, see economics kind of like first line of defense for You know, not squashing proposals, but just kind of just making asking the right questions and kind of you know, making sure we’re not having our community pool drained and areas where we’re not, you know, seeing the value, I don’t know. But the other the other item was the incentive distribution pool that you posted I think three weeks ago and to discussion. So I didn’t know if that was something you wanted to talk about today or run through

Cheng Wang: yeah, I think really briefly the totally agree in terms of

Cheng Wang: I think it is our prerogative and Judy to really be good stewards of the community pool, right? And in so doing it’s not necessarily poopooing things per se, but again, we have to evaluate things with like a fine tooth comb especially in this particular environment, where, you know, scarcity is real right. And we want to make sure that those resources are being employed and most maximally effective manner possible. And when they are deployed, we’ll send share with the community as far as like the events proposal, right that we were able to get Akt from. We have we’re building out an entire, you know, open access reporting back and not just report right? Where everyone can see what were the estimates. What were the actuals? How were you count for them? How we report on them? So that there’s maximum accountability for the use of funds and how we use them and where we are using them. So that hopefully that creates a model going forward.

Cheng Wang: For other community projects or just other, you know, community funded things or any initiatives or projects to to really helpfully follow in the footsteps of. So we have fewer

Cheng Wang: Kind of concerned, not concern. But question marks and issues going forward. In terms of How do we do in all that stuff? And we’ll make those templates. Maybe we can make those templates available when people can just replicate them more easily going forward as well and just have that as a, You know, a Google a set of Google things where you know, anyone can go in copy and, and use for themselves. So, that’s one the second thing. IDP accounts. Those are pretty straightforward as Greg. Originally outlined in his Akt, 2.0, proposal If anyone had a chance to read through it or have any questions, Please, definitely raise your hand. If not, I’ll just, you know, run through really quickly, just verbally here. Ultimately, what we currently have now is just one community pool, where some portion, right of network proceeds, go there. Thanks for Lincoln. The, the new implementation for Akt, 2.0, will be creating, you know, a higher level incentive distribution pool. Just call that kind of the holding pool, where all of the in,

00:05:00

Cheng Wang: Like take rates and block rewards and block rewards, etc. Other things will go into, ultimately And as that pool fills, it will be sub-allocated, then, allocator rather into subsequent, I’ll call them. You know, from an accounting perspective, right sub accounts. I’m not really creative with these naming things, I’ll leave that to Greg’s forte there, but those sub-accounts are allocated as outlined previously into certain categories, right? Like developer fun, public goods, fun potentially burning, right? And and also funding

Cheng Wang: Potentially gasless deployment in the future right for just to make the user experience easier and for people who want to test out the the platform where cage, network marketplace to just do. So in a fairly frictionless manner and you know, creating that type of faucet if you will. And I know it’s not truly a faucet that the word does come a little loaded, so take it with a grain of salt but I’m doing that, we’ll definitely help with you know people who are want to try out a caution network lightly and for those who wanted, you know, do stuff on a question. We’re going forward who don’t necessarily have a key to start with or you have that cold start problem there. The those are the type of areas where those the kind of taken funds in these pools can be allocated towards solving, right and greasing the the skids of. So that’s really the IDP thing in terms of setting it up.

Cheng Wang: Putting into production. Of course, all these things will be can be controlled by governance as far as you know, the percentage of allocation and all those types of things. If anyone has questions, please do feel free to comment in that get help discussions thread. I know it’s not many people have or No one has thus far but or you can ask here, but if you’re not comfortable definitely verbalizing here, feel free to jump into get out the discussions and let us know where your feedback is. Any questions on the tooth topics there thus far?

Cheng Wang: All right, onto the third bit, which is a big one, but also part of kind, I guess the not necessarily exactly in the wheelhouse of Akt 2 Dot. Oh, here’s Luna. Akgile per se, but will be liquidity pools, right? And not liquidity pools that you’re thinking of and familiar with within the vein of defy. But within the vein of GPU mining, right? Not giving my excuse me, rephrase that getting GPU capacity and resources on to Akash network. so, that’s something that we’ve talked about lightly and it’s going to be

Cheng Wang: pretty monstrous feature as far as I think, complexity in size that will be discussing. And so I’m putting together documentation a purity discussing that and kind of plain English, right? Not a developer. So, that’ll I think foment a lot of interesting discussion of a tldr where we’re trying to do is similar to

Cheng Wang: You know, I don’t know if everyone hears familiar with defy, I think, you know, and in one way should perform we we’ve all interacted with and in some ways, right? Which is you put in capital into this smart contract, or pool? Whatever it may be, you are given a particular yield for having some asset type in this pool based on demand right and then on the other side people can borrow against it and what have you and there you have, you know, compound V1, right? So what we’re intending to build here is essentially compound V1 for specific GPU types resources and that can expand to other resources in the future as well. But just to kind of start. For example let’s just say you know nvidia a 100 chips right for that’s all the rage right now in additions h100 even rare and harder to come by. But getting some of those chips on to a caution network in a distributed fashion, right? This is kind of the entire ethos of Akash network. You’re being.

00:10:00

Cheng Wang: Will be distributed. And permissionless and incentivizing those chips to come on to a cost network in a way where they can yield or earn some percentage of costs, right? Or just earn Akt yield in order to have those resources available. So that when people come in and want to deploy, the providers will be kind of locked in stocked, right? And the shelves will be will be at least have product there for people to use and experiment with whether production testing playing around hobby stuff, Whatever the case may be to have that available and so we’re putting together some documentation behind that in addition to some modeling to understand what the marketplace currently looks like. As far as pricing goes in order to back into what we think, may or may not be probably right a A reasonable yields right on a monthly basis and will probably also implement or implement for discussion rather.

Cheng Wang: A bonding curve, right? To say If I have an nvidia A100 chip that’s, you know, 40 gigabyte, 80 gigaby, Whatever it is, I put onto a caution network if I put it on and committed for seven days, the yield on it will be lower than if I committed to 14 days and lower than if I committed for like let’s say 30 days, right? So those are just some high level. Thoughts and numbers and I’m throwing out there as far as the bonding curve goes in terms of keeping that supply on available and steady or users and akash network. And then that I think that’ll be a critical component to bootstrapping that effort as the, you know, GPU feature continues to be flushed out and tested. And once it’s ready for Mainnet to go live, this will be a very important component of it to, to follow on. Does anybody have questions on this item or Greg or anyone? If you want to add on to it, please feel free. Also.

Cheng Wang: Fire me.

Scott Hewitson: Yeah, I just got it. I mean I know. So GPUs is the focus right now, you know, we’re doing testnet for it. It’s it. You know, AI is big. Big narrative, big focus, the initial, you know, so this is for everyone say This is the provider incentives. Pool is what we’re talking about here, correct. Yeah,…

Cheng Wang: That’s right, that’s right.

Scott Hewitson: and Yeah.

Cheng Wang: So it’s Exactly.

Scott Hewitson: And so so you know, we’re focusing on GPUs. Do we? I know and I know this is super early and you know, we’ve only had a few internal discussions about it. But are we looking at doing Are wanting to do like pools for just like normal compute to bring providers on, or is that, or we just thinking for GPUs at the moment?

Greg Osuri: So, the Tension. Of. To Sol. Liquidity mismatch. We have.

Greg Osuri: a lot of times we hear,

Greg Osuri: Turn. The demand. for a cautious that people come to Akash expecting a certain resources, Those resources are not available. So we will incentiv. Any type of configuration depending on demand? If we see quite a lot of demand, For nodes, right? So nodes.

Greg Osuri: Cannot. And be required. You know, certain amount of disks this require certain cap. IP addresses and and certain type of CPUs, right?

Greg Osuri: Understand and stud. and side, especially looking at That are not fulfilled. On the demand side. And based on that will be able to propose what pools we should incentivize. Thanks. it’ll be definitely demand when we think, Validated. Quite a lot of demand from a GP. Point. Because Because ai is just. of like, You know, crazy tape.

Scott Hewitson: On fire. Yeah.

Greg Osuri: At. Right. And there’s just a supply shortage for GPUs right now. so, We’re, you know. No. A weekly basis I get at least two. People asking. Specifically. 100%. So, Exactly. What Use minute incentivize right, but with compute, it’s such diverse. Also needs right, from at least from what people ask. Hard like pin down. What I’ve

Greg Osuri: It inside of ice, right? So that’s how we’re looking at it. A100s are easier. Much more validated and h100 school. So it’s so it’s, it’s Rescue. Isis. What we’re trying? Orders, we incentivized certain type of configuration. and then, Don’t have the demand to fulfill the Is now just giving away free tokens, right? Which Absolutely want to avoid. Um, you know the poll the whole goal of incentivizing these pools, essential that there’s Supply enough supply to be able to attract.

00:15:00

Scott Hewitson: That makes sense.

Cheng Wang: Well said Yeah and that will be out in the coming weeks as far as the next steps for that particular thing. Provider incentives we really know next steps and unless you know anyone wants to to ask any questions now but soon to follow, I know work has already started on the first Akt 2.0 PRT that was openly discussed and submitted which is stable payments and take rates so code is ready, being written for that. And so hopefully there will be something very soon that we can share. So yeah, I’m you know, you really awaiting that as well and as everyone here. So once we get that alive and running and look forward to continuing to follow that on, with the subsequent components of Akt to point out.

Scott Hewitson: Great, thanks for those updates. Does anyone have I mean, I guess at this point did we can kind of open it up and let’s Cheng or Greg. Do you have anything else you want to add any other items that are related cigar economics that you want to add?

Scott Hewitson: No. Okay.

Greg Osuri: Believes table payments. also currently in Right, so people is code.

Greg Osuri: Operating stable requirements to come. At least to. To complete Mid-tune. Hello about 45 days or so, we should see some stable payments code in desperately. So that’s moving forward along. I take income is going to be the first implement. I believe. so, You know, 45 days before in the next 60 days, will be able to see some taking comments. Some stable payments as we are approaching GPUs

Greg Osuri: Some back of the envelope estimation. A GPU demand, what we have in our pipeline, there’s no guarantee will be able to fulfill all this demand, but just what we have from Pip. It looks like about requests for about. Hundred. Right. so, the translator roughly You know. About about 1,500. Hour in. Early revenue.

Greg Osuri: Just by fulfilling this 500. Hundred. Dollar 30 cents an hour. So just just, you know,

Greg Osuri: Yeah. That’s, that’s Our thousand dollars, right? Yeah. that’s, Premise as well. So

Greg Osuri: Getting. There being an aggregated, document for all the changes. or there is

Greg Osuri: There is a link. For a Kitty 2.0, I don’t know. Had a chance to. But of course it’s going to be part of the documentation, right? Yeah.

Greg Osuri: Economics. Interesting. What kind of GPUs do you know Aquin

Cheng Wang: Yeah. All are welcome. Honestly. If they have resources where we’re happy to re-engage or engage new folks that we’ve been we work with historically, right? So I think everything’s fair game and I’m not familiar with it, but if you’ve talked with Greg or Boz or others on this call, that would be fantastic to qualify understand what they have. So as an example, charity service, who is a, you know, they’re doing kind of proof of concept thing, for their own, internal use case, for Akash network, they spun up a, you know, very small provider initially. And over the course of time, with they can at the end of the ballot or campaign, we’ve been running with kava, we we’ve helped them kind of scale up. More resource introduce more resources to their provider in order to host some of these workloads. So that’s definitely something. We would love to see, you know, what kind of resources, economics, for example has and and you know what the economics are. And if it makes sense to to bring on to the

Joao Luna: Yeah. So for context, while I was at kubecon, they they were ones of that that recognize the cash. So we kind of approached each other and we talked a little bit about about it. How it was in the past. Just overall talking about the Akash network. And then I mentioned the GPUs, they were kind of like they are not publicly announcing that they have GPUs, it’s kind of a niche A portion of the infrastructure but they do provide those in like a partnership way. I don’t know how expensive they are. I I try to ask the guy that was there couldn’t really give an answer, but maybe something that you might want to revisit something to explore.

00:20:00

Greg Osuri: He has good point, I’m absolutely. I mean, we’re religious for them. Yeah, we definitely loved. It how? Could see from there website. They do have You know. Words, right? So they looks like it.

Greg Osuri: Let’s see.

Greg Osuri: Like a good a100, but GPU. They don’t see how much on chip memory. And they don’t talk about how much it costs, or We got to see. Yeah. In, we’ll take any. kindness at this point, you know,

Scott Hewitson: Great does anyone else have any other, any items they’d like to bring up or any other questions or in things they want to address, MMM, radio silence.

Cheng Wang: All good on my end, honestly.

Scott Hewitson: Yeah, no. I mean we don’t need to take the whole time. So you know, this is Good good discussion. It seems like we’re making some good progress, you know, Akt 2.0 is a big undertaking, so chunking it out as nice for sure.

Cheng Wang: I did have a quick calls action for I guess people here who are not in the Overclock Labs core team. If there’s anything else that you would like to see us discussed more of less of differently. Please feedback is always welcome. Let us know in this form or any others. Right discord Github if you want to. I know it’s sometimes it’s just you know you want to hear in and just simply be a fly on the one. Listen, that’s totally okay too. But if you do have a strongly held opinion, we would love to know.

Scott Hewitson: Absolutely. Alrighty. Well with that, I’m gonna stop the recording and then we can wrap things up. This go and we’ll see everybody in the Metaverse as Adam was says.

Cheng Wang: Thank you, Scott. Thanks everybody.

Greg Osuri: Thank you.

Joao Luna: Thank you.

Cheng Wang: See you next time. And yeah.

Scott Hewitson: Yeah.

Andrew Gnatyuk: And bye, guys.

Cheng Wang: Bye. See Andrew by

Meeting ended after 00:22:54 👋