If you would like to view the full recording of the live Crowdcast event, you can find it on our YouTube channel. You can also read the full Quarterly Report by Mihai Grigore and Stephanie Dunbar here.
Participants:
- Mihai Grigore – Senior Research Analyst, Messari
- Stephanie Dunbar – Research Specialist, Messari
- Doug Petkanics – Founder & CEO, Livepeer
Mihai Grigore (00:04): My name is Mihai Grigore, I’m an analyst with Messari and today I’m joined by my co-host and colleague at Messari, Stephanie Dunbar. Hi Stephanie, thanks for joining.
Stephanie Dunbar (00:25) Hi everyone, nice to be here with you, co-author this piece with Mihai, and we’re happy to be here speaking with Doug Petkanics from Livepeer, the founder and CEO of Livepeer Inc. Doug if you want to give a little intro about yourself as well.
Doug Petkanics (00:40): Sure thing yeah, thanks for hosting: Messari, Stephanie, Mihai, excited to be here. I’m Doug Petkanics, founder at Livepeer, a video infrastructure protocol that’s been running live in the Ethereum ecosystem for 4 years. Excited to talk to you guys and dive in.
Mihai Grigore (00:57): So for those of you in the audience who are not accustomed with Messari and with our quarterly reports, so, at Messari we do data and research aggregation for open decentralized protocols and Livepeer is one of them and in front of us we have a sample of the recent quarterly report that Stephanie and myself have co-authored on the state of Livepeer Q1/2022, and in a nutshell here, what we do is we aggregate data from different sources. So for this specific case of Livepeer we aggregated data from Dune Analytics and also from The Graph and this whole effort is done together with other teams in Messari, so we had here fantastic help from our data engineers and data scientists putting together and aggregating all the data, all the necessary data to be able to put together all the quarterly metrics that we have been tracking for Livepeer so perhaps in a couple of words, what we have been looking at was at the usage of Livepeer in terms of minutes transcoded, in terms of also demand for the whole service. How much revenue has been generated both on the supply side, also on the demand side, and then we went through tokenomics and further on through the roadmap and notable events for the quarter. Yeah, so before we actually get into, looking at the numbers, we would first like to start with a short intro on what is Livepeer so that we get a bit of a grasp for everybody in the audience. What is the problem that Livepeer is trying to solve? And you can see here on the left-hand side under ‘input’, we do have represented the demand part for Livepeer, which is various applications and developers who need transcoding services. Video transcoding is the process of transforming videos into a format that, for instance, can be watched on our mobile phones and this is often used in live streams for instance. So yeah, from the demand side we come to the, to the middle, to the Livepeer Network which is formed by three main stakeholders, so they are like the main node operators being in charge for the transcoding, then we do have the transcoders actually doing the work and then we do have delegators. So all of them together deliver the transcoded videos that are being, actually output, and delivered to the users via different services, so perhaps here, to start it off Doug, we would be interested in your take on why is transcoding so relevant for the web services nowadays?
Doug Petkanics (04:44): Yeah absolutely, thanks for that overview. Backing up to Livepeer’s mission, it’s actually built the world’s open video infrastructure and open video software. Video is 80% of the traffic on the Internet. It’s the way that we increasingly communicate, get our entertainment, get our education, learn and participate in the world economy, so it’s growing at a tremendous pace and that puts a tremendous burden on the internet infrastructure to actually deliver all of this video around the world, and we saw that during the pandemic, when the European Union actually asked Netflix to reduce the quality of their streaming because it was taking too much bandwidth and too much, putting too much strain on the infrastructure and you saw Zoom kind of buckle under the growing demand and these sorts of things. So video is very important but it’s very burdensome on infrastructure, therefore it’s very expensive, it’s very hard to scale and it’s actually kind of prohibitive for developers that are trying to compete with the big cloud owned platforms, like Google owns YouTube and Amazon owns Twitch. If developers are trying to create new innovative video experiences, they typically can’t afford to scale the infrastructure to create compelling applications and so, what Livepeer does is, it actually lets all of this idle compute capacity and bandwidth around the world and data centers that node operators happen to have, come together to form a video infrastructure network, that’s an alternative to Amazon Web Services or Google Cloud. It’s an independent network, it’s an open marketplace where anyone with compute capacity can put it on and the Livepeer software provides many different elements of the video infrastructure from ingest to origin services, to a player, to everything developers need but what the Livepeer network does, this network of 100 node operators and thousands of transcoders and tens of thousands of GPU devices out there, it provides all the compute that’s necessary to make sure that we can actually deliver video reliably around the world and that compute takes the form of something called transcoding, you mentioned, that’s when, an HD video in 1080P comes into the network, you need to convert it into lower resolution 724p, 480p, 360p, so that it can be watched on phones and smart TV’s and low bandwidth connections without buffering, without disruption and all the stuff happens behind the scenes. Typically as a user, you don’t care about or to know about it, all is that when you press ‘play’ that video starts streaming to you and it doesn’t buffer and cut out and reduce the quality and that’s what our great network of node operators are powering at a fraction of the cost of what you see in other kind of SAS enterprise, live streaming services that typically charge up to $3 dollars an hour per live stream video that’s coming in. Here, the Livepeer network does it for about a tenth of the price and even less in some cases.
Stephanie Dunbar (07:51): Thanks for that overview Doug, and yeah there’s clearly a huge demand for video transcoding services, a huge need across all parts of the internet. I was just curious to hear because it’s such a large problem. Are there other decentralized video transcoding services or networks that are also trying to tackle this problem? And I know that we have the Amazon, and Google. We have some of the other centralized players of the world but if there’s other, decentralized ones as well that are trying to go forward with this mission.
Doug Petkanics (08:22): I think, when you look across the whole decentralized text stack, what makes life here pretty unique is we focus exclusively on video infrastructure letting other developers have these tools accessible to build the applications they want and of course, there’s other great projects that, kind of, circle around this space and do similar things. The Theta Network has their own video application and focuses on, kind of, content delivery transcoding, is kind of a little less considered piece of that. You see other compute type protocols like there’s a great one called Render Protocol which doesn’t focus on video transcoding but kind of has similar mechanisms for applying compute to 3D rendering and these sorts of things so in the Web3 world, we’re kind of all in this together, in a composable, and rising tide lifts all boats ways and excited to see how developers put these things together.
Mihai Grigore (09:21): So in terms of developers and in terms of applications actually using Livepeer, Stephanie and myself have been aggregating numbers on the usage and here perhaps the most important metric to actually track is the number of minutes transcoded and here we can see that over the last six previous quarters we had an all-time-highs including for this quarter which is a 12% increase versus the previous, yeah, according to Q4 2021 and here the question for you Doug is: where does this demand comes mostly from?
Doug Petkanics (10:20): Great yeah, excited to see the upward trend. Of course, it’s always good to kind of surpass the previous quarter. I don’t know if you’re seeing the chart be a little cut off, I’m seeing it, I think the number for Q1 is, was it 31 or 33 million minutes or so? but…
Mihai Grigore (10:35): 33 indeed.
Doug Petkanics (10:35): Yep great so what this means 33 million minutes of video over the quarter, it means that at any given moment there’s a couple hundred live streams of video flowing through the platform that are being transcoded and this doesn’t mean there’s a couple hundred viewers on one stream, it means there’s a couple hundred unique different streams of video coming from the, all the Livepeer users, all the applications that are built on Livepeer, any given moment that are being processed by the node so, who are these users? I think there’s, there’s kind of two different worlds to look at. One world is maybe traditional Web2 streaming applications. These are things that look like Twitch but are maybe built for specific communities there’s an app called Play DJ which is a DJ streaming platform, there’s an app built for artists called Picarto which is an artist streaming platform, there’s a social commerce platform called Korkuma, there’s a radio station that stream 24/7 like The Lot Radio, all leveraging Livepeer for its cost effectiveness and to add transcoding to their stack. What many of these not all have in common is their user generated content apps. Again, things like Twitch where users are the ones that are allowed to, kind of, for free go live and create content, and that means that kind of a scalable infrastructure is really important because you might have hundreds of users streaming at once, you’d have to spin up a lot of servers and operate them if you’re operating them yourselves, whereas if you outsource that to Livepeer, it’s valuable and cost matters especially if you’re letting your users go live for free, you can’t be paying three dollars an hour for each one of those users on their behalf, you wouldn’t scale, you bankrupt yourself and so I’m kind of in the traditional world. Livepeer delivers that impact and that’s great because there’s a lot of demand there, today it’s these dollars and fees into our platform for the node operators help show the viability of the network and improve the product and then, kind of the second camp, I mean, one we’re really building for and excited about is a more Web3 centric creative set of users. So, these are people that are really asking the question: “hey, what’s unique about Web3 and crypto that’s possible now that wasn’t possible in the other world? And, how do we create new experiences around that?” and you see people building decentralized social applications that give users ownership in the platforms and you have community based content moderation and governance policies, things that try and correct a lot of the challenges and frustrations that creators have with YouTube, where they kind of don’t have control over their monetization and they don’t have any say or visibility into whether they’re going to be featured or not featured or de-platformed, right? Same kind of solving the problems you see with platforms like Twitch and Twitter that can be corrected through, kind of community-owned governance and community ownership that you see in Web3, right? You see people experimenting with other primitives like how do NFTs combine with video? How do you do token gating access to content, instead of traditional proprietary DRM type access and so we’re really excited about this world. This quarter was actually a big one where we saw a lot of growth in some hackathons. I think we’re going to see breakout apps there, that dwarf these numbers in the coming, coming year, but today one of the side effects is that world is actually much smaller in terms of minutes contributed here because they’re all experiments, they’re all hacks, the user bases are small but really excited to kind of see some breakout mainstream successes there that crossover and have them be powered by Livepeer.
Stephanie Dunbar (14:37): Super interesting, you’ve been going to tons of hackathons and supporting lots of projects being built on the network this quarter, I think it was 72 this past quarter alone if I recall correctly. I’m just thinking like you mentioned it’s a smaller part of demand right now as those kind of grow and if I understand correctly, Livepeer Inc. is also helping with some of the traditional Web2 applications to join the network and get the video transcoding services and, at what point do you think that perhaps some of these newer projects or web-through-oriented projects will kind of outgrow in demand with some of the other ones you got going on right now that take a significant chunk?
Doug Petkanics (15:14): We’re ready for social summer. I think that’s the hashtag going around I think, it’s been really exciting we’ve been working on Livepeer for over five years now and in the early days there were very few developers that were actually thinking at the application layer, that were actually creating great consumer experiences because a lot of the infrastructure and tooling like wasn’t quite ready yet, and there was so much opportunity to still build out the primitives the infrastructure, the financial primitives, the development platforms, right? Now, in the last year kind of is catalyzed by the acceleration of gaming adopting blockchain and NFTs as being this kind of breakthrough opportunities for creators to better connect with, monetize their art, their creations, their content. Now, you’re seeing this interesting mainstream crossover and a lot of the infrastructure platforms are in place so Livepeer can clearly power live streaming, Arweave and IPFs/Filecoin can provide reliable kind of storage and applications surveying, The Graph is a great indexing platform, there’s all of these primitives are in place where developers can create scalable kind of social applications, media applications and we’re seeing things like, kind of, the DSO network launch and Lens protocol come to their mainnet very soon, which are social protocols that are actually catalyzing a lot of developer attention on these mainstream applications, so, the social summer concept is something that has been thrown around that I think if you just look at the next three to six months, I think you’ll see these, these valuable proof of concept social apps actually take the next step, do something innovative that changes the game a little bit and, attract a million users and that will serve as inspiration for a whole bunch of development that can only grow, grow from there. Eventually we’ll see a Tik Tok scale application that leverages crypto assets and tokens and NFTs because why wouldn’t you? That creates such a better kind of experience for the creators themselves that they should flock to it.
Mihai Grigore (17:29): So yeah, with the event of the “social summer” as you, as you call it, you essentially say that the market is going to continue to grow and looking at the numbers for the previous quarters and looking at this tremendous growth, I’m wondering whether the growth comes actually from a growing market or from starting from a very, very small base, starting from very close to zero and then increasing quarter over quarter, what do you think is the perspective here?
Doug Petkanics (18:07): In the, kind of in that web two-ish world, I think you’ll see kind of a steady more predictable growth motion just through kind of increasing adoption of the software and those apps themselves growing, but I think in the Web3 world which I said, is really what we’re building for, it’s really where Livepeer has the opportunity to be the leader in a new and growing market as opposed to just a competitor in a pretty crowded mature market. Video has been around for 70 plus years and there are many service providers there. In the Web3 world we’re really well positioned to be the video infrastructure and ensuring 72 hacks last quarter were built on us, many more hopefully this quarter, those couple breakout applications that cross over. I think we’ll, you’ll see kind of a step function growth where you’ll see single applications delivering a higher order of magnitude than even this total streaming volume but it’s hard to predict exactly when that happens, but once that happens you’ll see kind of a step function growth from there. So, hard to make a minutes prediction and in fact you may even see due to like leaning into Web3 world and kind of a focus on this opportunity which has lower volume today, you may even see less growth in the next quarter than previously but we’re true believers building for the future and believe that those breakout Web3 applications will really be the thing they catalyze, not just minutes to live your network but actually pretty impactful changes to the world and the way that people consume their social media and their media.
Mihai Grigore (19:47): Yeah, that makes total sense so we’re looking forward to seeing demand continue to grow over the next quarters as well and perhaps shifting gears towards supply. Here in this chart we see that Livepeer offers staking rewards and on the x-axis we see pretty much the timeline since Livepeer was started roughly four years ago, and then on the y-axis we have been plotting, the percentage of stake and we can pretty much see and unfortunately this is not shown correctly, let me try to see whether, yeah, now it is being shown correctly. So on the y-axis we have the percentage of Livepeer token being staked and we can see that the percentage started from zero in May 2018, and then increased, gradually surpassing a threshold of 50% and going towards 70% in the summer of 2020. And then since then, we see a reduction in the percentage and everything converts towards the threshold of 50% which we have been discussing in the report, it’s the so-called equilibrium we hold and then since then, we see that pretty much the percentage stays roughly at that level. Perhaps, yeah, it is worthwhile to explain a bit how this actually works. So, whenever the percentage goes above 50% there is a tiny reduction in the inflation, in the staking rewards, that brings actually that percentage slowly back, and then whenever the percentage is below 50% there is a slight increase in the supply and here, what we would like you to tell us is how did you come up 50%? Why hasn’t it been 60% or 70%? Did you have any rule in mind when you set this one?
Doug Petkanics (22:23): Yeah that’s a good question so what we know yeah, transport yourself back to the crypto world and crypto economic thinking five and a half years ago before there was a ton of examples in DeFi protocols and yield farming and studies of incentives and what not and we actually—so many networks were beginning to launch or publish white papers and they were creating an issuance schedule. Bitcoin being the initial example that said well, there are 21 million bitcoin all time and exactly what the issuance or release schedule of that bitcoin is going to be in each block and based on happenings. And then you saw in other protocols come along and create curves and say, “Ok, there’s going to be an inflation curve and we’re going to have this much release for the first couple years and this much release for the second”, and they kind of innovated on that a little bit and we did something kind of innovative. We said, “Well really what we want to be doing is, we want to be incentivizing people to participate in this network and to provide video encoding capacity and to be staking their tokens to do quality assurance and route work on this network and we always want that incentive to be there, regardless of, kind of, how much the network has been used or matured.” There’s always going to be a balance between how many fees there are to earn versus how much inflation. Inflationary Livepeer token needs to be earned by these known operators to incent these things if fees are really high maybe you need less inflation, if these are really low maybe you need a lot of inflation and it’s impossible to sit here and predict what that curve should look like for the next 100 years, instead what we should do is we should just target something that self-corrects, to try and ensure there’s always about half of the tokens participating and staked on the network and you ask the question why 50?, why not 60? why not 70? I’ll be honest, the 50% is somewhat arbitrary, it’s not, it’s also acknowledged, like, the community through governance could adjust that if they learn over time that it should be something else that’s being targeted but, what are the reasons? Ok, you want a bunch of tokens staked and participating and that’s locked up for a period of time so it’s long-term committed, you also want some token to be liquid so that it allows people to enter and exit the system, to cover their costs, to acquire token, to begin participating in Livepeer, and you want to leave some room for kind of experimental alternative uses of the token, like if people want to create additional protocols on top of Livepeer around governance, or DAOs, or signaling, or whatever the case may be, and so the 50% number felt loosely correct, like a good starting point but I’d say it’s worked out pretty well and I wouldn’t say there’s been some major learning that it should be adjusted to be higher or lower. The more interesting thing is really like when you look at this chart it’s not surprising that it took a while to hit 50% because of the way the token was released and people began to participate and learn about it over the first couple of years and during that year and a half the inflation rate rose up every day, so the incentives to participate in stake were very high, in fact I think it was nearly a 100% annual, kind of, yield or rate of return on staking, so you double your token by staking it on the network over the course of the year, um, big incentive to participate in stake, right? And so you saw the participation rise up to 70% and then as soon as you pass that threshold and, May or June of 2019, while the inflation rate started to tick back down and actually got kind of close to zero and as you got. It was 2, 3, 4 percent annual return and people found, oh, there are better opportunities for me to get better returns in the DeFi ecosystem for example, let me un-stake and, maybe leave it to go chase those but as they did that really if you look at the past year or so, we found this equilibrium around 50% participation and you’ve seen the inflation rate go up for a couple of days, and down for a couple days, and up for a couple of days, down for a couple of days and it’s really settled, it’s been going up for a little while and it’s settled around this eight and a half percent, kind of inflation for the entire network but because that only goes to the 50% or so that are staked at participating. I think I did the calculation today it’s about an 18% kind of yield on your LPT if you stake, which I think is pretty appealing and competitive, has led a lot of people to get involved to stake, to stay long-term participate in network and yeah, maybe we’ve found that equilibrium for the current volume of fees that are earned by node operators. I think as that increases by many orders of magnitude maybe the inflation will come down a little bit as people chase that but for now it seems healthy.
Stephanie Dunbar (27:35): Thanks, Doug. And you see on that chart as well, if you head over to mid-February 2022, the sharp drop, so I guess shifting gears a little bit if you want to talk about what caused that sharp drop the migration to Arbitrum, what that kind of meant overall for the community the whole transition period.
Doug Petkanics (27:59): Yeah, so one of the most painful and challenging things that we’ve dealt with in Livepeer is congestion on the Ethereum network we deployed on the Ethereum network four years ago, gas fees for those who are familiar with gas were about 1 gwei which is a 100 in order of magnitude of 100 to 200 times lower than they are today and so the network actually became very expensive for our node operators and our token holding delegators to stake and participate, in fact only those that had a lot of stake could kind of profitably do the protocol transactions every day to generate new Livepeer token and so for the last year it’s become a huge priority to solve this, so we made some protocol updates to give relief but really the big win was to migrate a lot of the in protocol actions off of main Ethereum, onto an Ethereum secured Layer 2 called Arbitrum, and in January we kind of completed that migration the community did a great job through the governance and the testing of the test-nets and what not to get ready for this. We completed that migration and actually were able to pretty quickly have most of the node operators migrated over to Arbitrum and bring all of the stake that was staked towards them, kind of along with them. It took a couple of days to get everyone over and that’s why you see kind of a quick drop and then a recovery, but, almost all of the LPT has migrated over to Arbitrum at this point and is actively participating in the network and the good news is it actually reduced the fees for node operators and token holding delegators by something like 97%, so you used to kind of only be able to run a node and earn LPT every day if you were going to be making more than a couple hundred dollars worth of LPT and rewards for doing that required a lot of stake and doing a lot of work on the network. Whereas now as long as you’re kind of able to earn more than a couple of dollars a day you can, you can profitably participate in the network and that’s a huge unlock that helps kind of the health of the supply side.
Stephanie Dunbar (30:18): That’s great so that must have been like a significant chunk of node operators are now able to reach profitability given those fees, would you say that it’s gone from like let’s say 20% to a 100% something like that?
Doug Petkanics (30:31): Yeah, that’s a good question. I don’t have the number right in front of me for that, so I’m going to ballpark it but it had gotten so bad and kind of the tail end of the Ethereum era there, that there was something like only 17 node operators or 12 node operators or something that could like every day consistently call reward. Now the others could still compete to earn these fees and were, fighting to profitably transcode video and we’re doing okay that way, but they weren’t able to become owners in the Livepeer network, they were able to earn Livepeer token every day as was designed and then with migration to Arbitrum, I know, immediately it was something like 60+ of the node operators had the opportunity to be, be profitable and I, I don’t have the number off the top of my head but I wouldn’t be surprised if there’s a set of 100, I wouldn’t be surprised if the majority of them can participate profitably on Arbitrum if they’re there and actually running infrastructure and are competing to transcode video and not just kind of an idle node.
Stephanie Dunbar (31:36): Great thanks, Doug.
Mihai Grigore (31:42): Perhaps looking at the sharp rebound in the percentage of LPT late, so after February 14th we see that rebound back to 48 and very, very close to 50% and we were wondering whether that went organically or whether there was any intervention from the community side, how did that go?
Doug Petkanics (32:09): Intervention is actually an interesting word because it, I don’t know, it implies kind of, proactively taking a forceful stance in order to intervene and will something to happen. I wouldn’t say that there was anything that needed to be like immediately reacted to that wasn’t expected. I would say there is proactive communication and we’ve had so many community calls, blog posts, our Discord channel, our node operator community, we call them orchestrators. The orchestrator community is incredible and like even far more active than the core team in terms of communication and sharing information and so I don’t think there was an intervention but I think it was pretty clear to everyone that, “Hey!, those rewards that you have been earning, like you want to keep earning them, you need to do this transaction to migrate to Ethereum, and by the way, as long as you installed the latest node software and upgraded other than doing kind of one manual transaction, it should be pretty seamless and shouldn’t affect your operations too much.” I know they had, the community did a great job and actually had a little bit of trouble running on Arbitrum because running Arbitrum nodes was way more difficult than running Ethereum nodes for some reason or the hosting providers like Infura and Alchemy were expensive or rate limited or kind of tough for an independent operator to access profitably, so the community actually came together and solved this problem and jointly ran a node and formed a little coordination around it even though they’re competing with one another for fees, they’re actually more of a tight-knit community working to kind of make this network usable so that was really inspiring to see.
Mihai Grigore (33:54): Yeah absolutely and also this is a testimonial that the staking reward mechanism actually works and works on its own and this is a beautiful exemplification of that. Perhaps to show very quickly what happened in terms of token inflation. We were able to see that inflation in March so, roughly let’s say two weeks after the migration increased from 0.65% to 0.8%, so this was like from those small gradual increases in order to bring the percentage of stake, Livepeer token yeah back from 32% to 50% and pretty much this is what we were able to measure in terms of increase in inflation and we were wondering that indeed the mechanism, the reward mechanism seemed to have worked flawlessly for the whole migration. Are there any other qualitative insights that you experienced during the migration that maybe we haven’t seen in the numbers so far?
Doug Petkanics (35:22): Yeah, one thing, this is outside of the numbers, and kind of the stake mechanics but one thing that you do see is these layer 2 like Arbitrum and Optimism and Polygon actually provide like tremendous benefits to the users that are actively participating in the protocols and create opportunities for them to participate profitably, become owners in them at low cost just like in the early days of Ethereum. One thing I would highlight is that there’s still a ways to go before each of these ecosystems actually has the full composability and functionality and features that layer one Ethereum has, so for example to get your tokens out of Arbitrum back to Ethereum, there’s a seven day waiting period for the bridge which is just based a security mechanism of this, this layer 2, and why do people need to go back? Well, there’s a lot of like liquidity through DeFi for tokens on layer 1 Ethereum and there’s protocols where that liquidity can exist on layer two but maybe it’s not there yet or exchanges allow you to deposit and withdraw token from layer one on Ethereum but haven’t yet added support for kind of layer two deposits them with withdrawals so, these layer 2 are great for kind of participating profitably and being committed to the protocols and I think they’ll become even more usable and full-featured as the ecosystems around them get supported.
Mihai Grigore (37:00): Yeah so, speaking of all the operational matters when migrating from Ethereum to Arbitrum, what we’ve done in this chart is, we took all the rounds in Q1, so starting January 1st to March 31st and we had in total like 100 such rounds and it was like 50 on Ethereum and then 50 on Arbitrum, so you see on the x-axis actually the number of the rounds and on the y-axis we have been plotted the moved stake. This is actually a logarithmic chart because we weren’t able to actually visualize, like the big difference from before versus after, so we had to apply logarithm to the values and pretty much we can see that, the amount of move stakes, so the amount of activity increased a lot after the migration and what we have been marking highlighting on the right hand side was the so-called moving rush, so essentially in the second half of February, so between February 15th and March 1st there was a sort of, moving rush and pretty much we’ve been trying to see whether afterwards things will settle down and if they settle down, at what rate would that be? So what we did is, we’ve tried to calculate yeah, the moved stake that actually happened in March and we wanted to compare it with what happened before and the best comparison we were able to find is to compare March actually with January the full months and the percentage that we, that we found, so when comparing March to January was actually 60 times. So there was like 60 times more moved stake in March like post-Arbitrum migration than actually before that, and we were wondering whether employing such a methodology is accurate first of all, so we’ve been trying to see whether if we compute it for one and a half months, since February 15 to April 1st, we might actually compare apples and bananas, so we didn’t want to do that. If you were to actually compare like before versus after, would that be another methodology that you would employ here?
Doug Petkanics (39:59): Good question, first of all, I just want to say this is so cool that Messari was able to create not only like just, like a quarterly report just like a company would do where they have to provide all their own data but like Messari was able to independently create this report and draw these insights totally from like open data accessible on a blockchain to anyone. It’s just like a really powerful testament to kind of blockchains, open data sets, Web3, and you’re able to come up with, like, insights like this, right? I think the methodology is sound, I’m not surprised by this dramatic increase because back on Ethereum, there were incentives for you to move your stake around, move it from one node to another to potentially chase yield, except that those incentives were obliterated especially for kind of smaller stakers by the high fees that you have to pay in order to do the transactions, so, great you can earn a couple more Livepeer token over the course of a month or two if you take the staking action, except if it’s going to cost you $200 dollars in Ethereum to do a couple of transactions it doesn’t make financial sense. Whereas as soon as you, kind of enter Arbitrum in this low-cost world now you have the incentive in the world to move stake around, to activate high performing nerves to chase fees to act in your own self-interest and the interest of the network at large. So it’s great to see such an increase, I’m sure some of this was due to kind of that pent-up demand and initial action when this became available, I wouldn’t expect sort of that pace to necessarily maintain in some respects, though I’d imagine if you compare like a May this month to let’s say, a couple of months from now, I think now because staking becomes cost-effective for all different magnitudes of operators, there’s a lot of product work that the community and team is working on, to actually encourage more stake movement, doing things like showcasing to the community like which nodes earn the best yield in terms of the combination of fees and LPT rewards or surfacing like new nodes that are high performing and are actually sharing outsized portions of the fees that they’re earning relative to a node that you may be staked on and that would encourage people to move stake to activate and route more work to new nodes and make this a more global reliable decentralized network, and so I think there’s a lot of product opportunity now that, things are cost-effective on Arbitrum again.
Stephanie Dunbar (42:41): Yeah definitely really powerful that you can move, stake around and, stake it to the most effective nodes and kind of help with the security and the strength of the network in that way. I was wondering if you could tell us a little bit more about the verification procedures, the fast verification procedure as well as the off-chain dispute resolutions and how that will also help in the future tie into work most effective nodes and kind of help identify some that may not be operating as effectively.
Doug Petkanics (43:11) Yeah, so in Livepeer, if you’re going to send your video out to be encoded by a random node on an anonymous network, then you kind of want some assurances that they’re actually sending you back the correct encodings of the video, they’re not sending you blank video or malicious video or something and so in the initial Livepeer white paper and first version of the protocol there was this verification game and algorithm that was leveraged called Truebit which could actually give you with perfect confidence on a deterministic computation that it was ultimately done correctly with enough economic security and that was great for the first version, the proof of concept, but it kind of only works for these deterministic computations that are done by CPUs and a lot of the transcoding on the Livepeer network has evolved to be done by something called a GPU or a graphical processing unit and that’s, actually it’s not deterministic meaning that the same video may be encoded slightly differently each time depending on a number of conditions on the hardware used. So you actually need new verification algorithms and the this is an open research area at varying levels of completion but the approach is called “fast plus full verification” and the idea is that as broadcaster you can very quickly check the results that come back to you from nodes on the network and you can get a high confidence that this is likely to be correct. If you have that confidence you can use that, use that video and send it to your users and if for some reason you don’t have that confidence, you can actually kind of quickly fall back to have another node on the network do the work or multiple networks do the work and compare it and again get that high probabilistic guarantee and then, it needs to be coupled with a full verification process which is more expensive, it’s slower, but it can give you that sort of ultimate guarantee that the work that you thought was done incorrectly was and that’s the thing where you can apply like an engine economic penalty to, you can slash a node’s stake and the stake that they have is kind of what secures their work no one wants to lose hundreds of thousands of dollars by maliciously missing coding your video for most use cases, and so where this is at right now, you can follow along with the research and our research channel the fast verification, it has been live in various pilots on the network. There’s an upcoming release which rolls it out more broadly. Full verification, there’s a lot of research on different approaches that are going to layer that on, and so that’s like a kind of high item on our network team’s roadmap in the coming quarter.
Stephanie Dunbar (46:20): Great thanks so much for going into that, I think it’ll be really interesting to see when all the pieces come together, and I guess just touching on a couple other roadmap items, you mentioned in a blog post a couple of months ago I believe working on overall latency, decreasing overall latency for the network and, getting up to the same level as some of the Web2 competitors like Amazon or Google, and I guess I was just wondering what, what level of speed has to be reached in order to compete with these established players and, what will it take to get there? I guess how far off are we from getting that level of low latency with the Livepeer network?
Doug Petkanics (47:06): Yeah good question, this is a kind of a complex one because there’s so many different types of protocols for different use cases in the video space. So your typical broadcast that people have become accustomed to for many years, the one-to-many streaming over the internet for large events can have anywhere from like 10 to 30 seconds of latency and up if you’re, if you’re watching the live stream of the Olympics or the Super Bowl or something, you’re typically seeing that level of latency and, the Livepeer network has always been competitive on that typically targeting 8 to 12 seconds for the conventional, kind of, HLS AVR delivery. Now there’s kind of low latency broadcast which is emerging via new protocols like LLHLS that gets you kind of just 2 to 4 seconds of latency and actually this is kind of not a released production feature, it’s still kind of an alpha or a beta but on Livepeer.com you can actually click a box and get access to kind of a lower latency protocol that brings the broadcast down to that two to four seconds leveraging some of the software called Livepeer Catalyst that resulted from a, actually small acquisition that the Livepeer team did last year of a video technology company called Mist Server and so the software is enabling lower latencies and then when you get to things like real time like this chat that we’re having here where there’s, there’s only a couple hundred milliseconds latency for a real-time communication, that’s kind of a different video stack and that actually doesn’t leverage transcoding, it makes use of kind of different capabilities of the network that I certainly think are in scope for the vision of Livepeer being the world’s open video infrastructure, but kind of require a different set of technology and, you’re targeting different use cases and, and whatnot, then what kind of the main focus the Livepeer has been today.
Mihai Grigore (49:13): Perhaps when looking at the migration to Arbitrum and we see that it went well and everything is fine, light latency is there, I have been thinking of actually stimulating the demand of, from other developers who build applications on other chains, have been thinking of enabling that part of demand?
Doug Petkanics (49:44): Yeah absolutely, like that was one of the big themes that became clear in 2021. This truly is a multi-chain world if you will there’s no way that you can build a social app or a gaming app or something more consumer friendly on Ethereum where you’re competing for gas block space with DeFi transactions that are willing to pay a thousand dollars for a transaction, well no one’s willing to pay a thousand dollars to like a Tweet for example and so a lot of applications are being built on Polygon, on Solana, on some of the kind of emergent challenger L1 if you will, Avalanche…etc. And the good news is Livepeer, as a video infrastructure, doesn’t require the broadcaster, if you will, to be on-chain and so they don’t have to be on Ethereum or Arbitrum in order to use Livepeer. It’s still accessible to them especially through a gateway like livepeer.com but what we’ve actually started doing is developing SDKs software development kits for developers on these other chains and the first one that kind of people can access is a Polygon SDK for minting video NFTs leveraging the Livepeer network to do that and that’s been really well received at hackathons, a lot of the hackathon projects have built on it and I think you’ll probably see that sort of approach for other chains, as well as thinking how to actually be a little more native to those chains where they maybe don’t need to go through a gateway, they can deposit their own native asset to pay for the infrastructure and build directly.
Mihai Grigore (51:20) Yeah so speaking of SDKs and hackathons, so this is essentially enabling a better life for developers to be able to build on top of Livepeer, could you perhaps tell us what are your favorite projects that you might have seen during the past hackathons taking off on Livepeer?
Doug Petkanics (51:45) Sure yeah, a couple come to mind one is called Huddle01. They’ve participated in a number of hackathons and continued to iterate and I think are in a really well positioned spot to actually be the sort of the Zoom of Web3 if you will. You can use them for conferencing and then you can stream those events or those conferences using Livepeer to a broader audience, you can record the content, play it back later and I think there’s a great opportunity for communities to use it to actually organize their archive of content in a Web3 native way, so check out Huddle01. I would look at an app that’s been building on the Lens protocol called Iris and I think they have another app called Tempra, you think of this as kind of experimenting Web3 social and creating a feed something that looks like, a little bit like Twitter that can embed video and other types of content, so, I think that’s exciting. There’s an app called Beam which is, think of it a little bit like a Patreon for creators publishing and streaming and creating and pushing out their content. So these are a few exciting ones that come to mind and we’ve been hosting a series of Twitter spaces I think typically on Friday afternoon, so if you follow the Livepeer Twitter account, each week we try and showcase some applications building on Livepeer in Twitter spaces.
Stephanie Dunbar (53:13): Nice I think we lost Mihai, hopefully he’ll be back in a second. I think he lost connection, but I’ll take over for now. Curious to know are there any other applications being built leveraging Livepeer that can be like a decentralized alternative to Crowdcast and then we can start hosting these kinds of talks using Livepeer. I think as Ryan mentioned on the last community caller, last call with you eating our own dog food so to speak and, getting to leverage Web3 protocols instead of traditional Web2 ones?
Doug Petkanics (53:49): Yeah I’d encourage you to check out Huddle01 for that I think, I think you can do the job because we could be having a conversation just like this amongst ourselves but you can also be streaming it to a broader audience that’s when you kind of need to leverage those different technologies to have that, have that wide reach and I know we’ve used it for a number of internal meetings it’s gotten better and better every single week and I don’t think we should be in a world where all of us are just archiving all of our content to YouTube and hosting all of our calls through, through Zoom necessarily. I think we should be using the kind of open source, openly accessible web free native tools that will be built in the Web3, features that mesh with the way that we’re using this new wave of the internet.
Stephanie Dunbar (54:34): We’ll definitely check that out, looking forward to that as well like seeing more Web3 things take over in all the things that we do in this industry, and then I guess, talking about other things that Livepeer can do moving forward or different services the network can offer this upcoming quarter you’re planning to have object recognition and scene classification hit the scene and I’d love to hear a little bit more from you about like some of the powerful use cases that can open and also, how it will affect the network, like how things will change adding additional services of it above and beyond video transcoding?
Doug Petkanics (55:09) Yeah so these are, I’ll elaborate on what these are, they’re interesting because there are other forms of compute that you want to do on video and you have this live stream flowing through transcoding is one type of compute that you want to use to transform the video, but you also might want to apply these machine learning algorithms to do things like detect what type of video, is it, is it likely to be violent content or adult content or copyrighted content because for your application you may need to flag that or filter that out or take whatever actions are appropriate. It’s very expensive to do that and to leverage proprietary services at nine dollars an hour to do that detection and mesh that with humans and stuff is challenging but the Livepeer network is really well suited to do it because we have these tens of thousands of GPUs that are already encoding the video and can do these computations at the same time, and so these are things that kind of been communicating on the roadmap for a long time, they’re available in kind of pilot form like we’re working with design partners, we’re leveraging them in their applications, learning from that. I don’t have a date for when they’re like productized and kind of accessible to all because what we’re learning kind of informs how you productize it, something like that kind of scene classification: is this adult content? is a little bit more straightforward. You can see how everyone can leverage that and our node operators will be able to kind of earn more money for providing those services while transcoding. Something like object detection, really depends on what type of object are you trying to detect, do you train a model for that specifically for a specific use case so for example, I thought it would be really cool if our network could automatically identify what PFP NFT was being shown on the screen and if it could automatically identify that, it could let you hover over it during a video to see what NFT it is, how much it last sold, for what collection it’s part of, who the owner is and you can see how Web3 community is using a tool like Crowdcast, where they’re using those profile pics of their avatar would really leverage that, but to do that you have to like train a model for on all the NFTs and like productize it in that way and is that useful?Is there a market? This is all open-source software and an open project I’d love to see someone like dive in and take that on, but that looks a little different than doing object recognition for say like an e-commerce shopping application where you want to identify what products people are holding up and make them clickable so they can link through to, to buy the products and those sorts of things so those things we work with users on one-off use cases and try and determine what they need and then we can productize from there.
Mihai Grigore (58:01): Are you perhaps thinking of object recognition and scene classification so the machine learning type of applications to also be monetized in the future? So, currently if you’re looking in terms of purely transcoding there’s like transcoding fees. Are you thinking of having a business model behind this, let’s say new features and these new use cases?
Doug Petkanics (58:34): Yeah it’s like, it’s a business model for the network. So the the abstraction of the Livepeer network is that nodes can advertise what service they’re offering and how much they’re charging for that service and then how they want to verify the work was performed correctly and so right now kind of the, the two services you can advertise are GPU transcoding and CPU transcoding right? They’re both transcoding it’s like one, but you can imagine nodes could advertise the capability of transcoding and scene classification, and content detection, and even expanding to some of the things we talked about before like firewall traversal for real-time conversations with this stack or origin services where we’re hosting serve your video out to your users or to your CDN so you don’t have to and this becomes a marketplace where node operators depending on how much bandwidth they have or what type of hardware they have for transcode can actually like offer a menu of services and then the users can take advantage of the the services they need at the most cost effective and reliable rates and that’s how it kind of expands the market that’s addressable for the Livepeer network over time as well and then of course businesses can build products and services around the super powerful open network to, productize even further in a way that makes it easily accessible to the users.
Mihai Grigore (59:59): Yeah that that makes total sense which brings us also to the end of our questions Doug and, is there anything else you would like to leave the audience and the Livepeer community with for the next quarter or like something exciting that is in the pipeline for you guys?
Doug Petkanics (1:00:22): Well, I’m really excited about Livepeer and everything that we have coming up as you can tell. One thing we didn’t even mention today which is crazy, is all the turmoil and mayhem that’s gone on the broader, kind of cryptocurrency ecosystem. I shared a note to our community about this and we share it here. One of the nice things about Livepeer is that we’ve been around for a long time, we’ve built through many cycles like this it almost feels normal and conventional and we have such like substance to fall back to where we power this very relatable use case video streaming for a huge market, that’s only growing and so we’re not this crypto speculative financial primitive that where there might be some negativity around those use cases for the next months or a year or what not we’re excited about video you can tell. We’re pretty deep into video technology and we’ll keep building through these cycles we’re well capitalized to do so, and we think we can be a big part of video for the next wave of the internet so that’s what I’m excited about that’s what keeps us in our team energized during these tough times for all, but yeah we’ll all come out the other side even stronger than before, I’m sure.
Mihai Grigore (1:01:34): Absolutely, so infrastructure is definitely here to stay it solves like a real-world problem, right, video transcoding, so there’s something that there is a need for, and the service provides definitely a value added to the users and this is also what excites Steph and myself to work on and to cover Web3 infrastructure protocols. We’d like to thank you a lot for this fantastic conversation. We enjoyed it a lot and we thought that the answers were spot on, and with that we’d like to thank you to the audience and to the community. I would like to leave you with, yeah, Messari.io/research, this is where we publish all our research, so we do besides quarterly reports, we have initiation of coverage, we also have ecosystem overviews and everything you find over there. Also, we are very active in Twitter at Messari Crypto and we do have a Newsletter that is read by very many people saying, if you’d like to sign up for that newsletter, you pretty much get an article and many links sent to you every, every day. All right with that, thanks a lot and have a great rest of the day and hopefully a non-eventful weekend ahead.
Stephanie Dunbar (1:03:16): Thanks Doug.
Doug Petkanics (1:03:16): Thanks Stephanie.
Stephanie Dunbar (1:03:23): Nice to talk to you, bye everyone.
