# The Sequencer of Tomorrow - Jan Gorzny

- Speakers: Jan Gorzny
- Channel: [ETH Warsaw](https://streameth.org/eth-warsaw)
- Date: 2025-10-07
- Duration: 22:11
- Watch: https://streameth.org/watch/yt-lc6cpJ9x1Og
- YouTube: https://www.youtube.com/watch?v=lc6cpJ9x1Og

## Description

The behavior of rollups is generally well understood, backed by a strong documentation and operational networks. However, as these systems mature, they will add more and more functionality and features. They will do this to add distinctive features, reduce trust assumptions, for efficiency, and many other reasons. These changes differentiate existing rollups and will impact how the rollups of tomorrow will behave. In this talk, we’ll discuss changes to a single rollup component -- the sequencer -- in order to see how it can vary from rollup to rollup. We’ll explore concepts like shared sequencing, based sequencing, and unique sequencer functionality. After this talk, you’ll be primed to talk about all the latest innovations -- and current challenges -- for rollup sequencers.


🧜🏻‍♀️ ETHWarsaw is a series of educational and entertaining events for an active community of blockchain builders, developers and enthusiasts with focus on Ethereum-related tech. Once a year, we organize a large conference and hackathon for the community in the center of the Polish capital with speakers from the best web3 projects and participants from all over the world.

Follow ETHWarsaw on social media for the latest updates!
X (Twitter): https://twitter.com/ETHWarsaw
LinkedIn: https://www.linkedin.com/company/ethwarsaw
Telegram chat: https://t.me/joinethwarsaw
See you all at our events in Warsaw 🙌🏻

## Transcript

our next speaker is Yan gorney and he is the co-founder of zerit he is going to talk to us about the sequencer of tomorrow please step up thank you thank you um nice to be here uh it's uh it's good to be back at eth wara uh I really like it I like the new venue it's uh it's great anyways let's uh let's get into it so uh my name is JN I am a co-founder at zerit and zerit a a zero knowledge um rollup that that just went live um sort of main net in the summer and uh this talk is not necessarily going to be about zerit per se although there's you know a little bit in there um but really to get people here onboarded with what uh sort of buzzwords are going on in the layer 2 sequencer space so if you've been following recent developments in layer twos um a lot of them are trying to differentiate themselves in in different ways right they're they're trying to get their own market share they're trying to innovate in new ways they're trying to do some cool stuff and while they're doing this they're inventing new terms and they're coming up with new Concepts and they're changing how things work and uh that's really confusing to a lot of people um including myself I've been in the space for five years and when people start talking about Bas sequencing I like oh well you know I heard about this idea but what does it really mean uh and things like this so I wanted to give this talk um to sort of clarify some of these terms for some people who are new to the space and to go into some of the details and sort of talk about the things that excite me a little bit about um sequencing and the Innovations there so so uh that's sort of already what this the slide is which is um basically I'm going to talk to you about some of the innovations that have happened in in rups over the last um probably year-ish maybe year and a half some of the ideas are certainly much older than that but um now they're sort of maturing because in the you know past five years rollups came from the sort of idea that oh we could do this too oh my God they're here they're working right that you heard talks now even like people trying to bring it to bitcoin uh Starks versus snarks a whole bunch of stuff like they're really they've matured and so um as they do this they're evolving things are changing and certainly uh it's important that we know how they're changing why they're changing and um you know enable users like yourselves or developers like yourselves depending on what you are to build on or use the appropriate rollup for the things that interest you right if if you care about decentralization maybe there's a rollup that really really does that really well you want to focus on that if you care about other things maybe there's something that you know you want to pay specific attention to right so um you know different languages being used on the network or something like this so I'll talk a little bit about what a rollup is just so that we're all on the same page if you've been in the space for a while you probably know what all of this is but I have to make sure that if you haven't been in the space in a while you know what's going on and I'm just going to talk about some of the the trends that have been going on in this space so in particular we'll look at um decentralized sequencing and shared sequ or base sequencing um shared sequencing different things you can do on the sequencer and then sort of what's still left to come what what would we love to see um I think as an ecosystem and as specific projects over the next couple of years uh to really make rollups sort of your own and and make it useful to what you want to do so first what's a rollup well a rollup is sort of a way to take a blockchain and sort of cram it into another blockchain um what we're going to do is we're going to take some blockchain and we're going to say that it state will occasionally be anchored to another rollup typically uh and say that at some point you know the underlying roll up the layer one like ethereum will have some sort of say in what what uh the the rollup network is considered as final and so essentially we'll take the second blockchain the layer two build it on top of ethereum the layer one and say that the layer two is only final if the layer one sort of see some things happen and so what we can do is separate the the computation of the layer 2 blockchain off of the layer one itself and we can build layer two blocks and occasionally just post the State updates for those layer two blocks onto the layer one so typically this works through the process of um sequencing the layer two blocks so we we collect some transactions we build layer two blocks often they look like layer one blocks because right now everyone really wants um sort of evm compatibility or equivalence this is you know kind of a good goal because you have a lot lot of e developers so if you can just have their smart contract Support over and everything works it saves costly audits everything just works out of the box it's really nice but as we'll see there might be other ways we want to do that but anyway you build these layer two blocks they look like layer one blocks occasionally you write them onto the layer one itself and then after some conditions are met you say okay they're final because the L1 says they're final and in an optimistic rollup this might be something like a time oute that hasn't been challenged um so you post a layer two block and no one says you've you've made stuff up in that block or cheated about the transactions or in a zero knowledge rollup you could do something like um you know provide a validity proof that says actually the block was constructed and I definitely didn't cheap because here's a mathematical proof but either way at the end of the day your layer one sort of has some say in what um is considered final today though we're going to talk about not the finality part which is now more or less solved although you know there's still work to be done and and there's probably edge cases on on what happens if things go down or things like this um but we're going to talk about the sequencing part so how do we actually build a layer two blocks uh and and what does that mean and what can we do differently and you know what are the implications of changing this this piece of code that is actually common to every rollup regardless of if it's an optimistic rollup or is a knowledge rollup or something else entirely so to be clear we'll talk about the sequence of the build blocks and it can take inputs from two places in particular it can take trans transactions either from sort of the layer 2 Network itself so you can configure your wallet like metamask and say I'm going to send my transaction directly to the layer 2 uh and I'll I'll pay for fees and all that but typically you need to get some assets onto the layer 2 itself and you do that through a bridge or from the underlying layer one and so you can also have transactions which we'll call deposit transactions that come from the layer one not always a deposit per se but um this this word has become conflated with this so from the sequencer uh sorry oh no okay from the uh from the sequencer uh we'll build blocks and occasionally we'll do some other stuff with the blocks so we come up with a block which you know people have different models of what a block is in a layer two if you look at l2b and you sort of look through all the the rollups some people have blocks that look like ethereum blocks others have blocks that contain only one transaction so they they're kind of different but at the end of the day you have something that looks like blocks and you might also get the sequencer to um publish some data like the um the transactions it's going to put in the blocks or the batches so that uh if you don't want to wait for a finality mechanism you have some commitment that actually this is what the blocks will look like assuming the sequencer is honest it's called a soft commitment it's the first Arrow downwards there afterwards if you do wait for the finality you get the finality so you can't go back once you get the hard finality but the sequencer generally gives some nice heads up so that people can be assured that actually if everything goes right the transaction is going to sort of execute and we'll be good so it doesn't have a huge job basically it just needs to take transactions it needs to produce blocks this looks like gu right pre uh proof of stake where you only had one client uh you just had get something like gu and in fact a lot of the sequencers look like this but they're augmented with some other things to say get deposit transactions from from the layer one to sort of parse those blocks and build that but that's all they do they build blocks that's that's the key thing and now we can change them and do other things with them so uh oh I already already talked about this I'm sorry uh so sequencers do exactly that the sequence I'm going to skip this slide maybe we'll catch up on some time anyways so how can we change them well we can change them in a couple ways one thing we can do is say well we have multiple networks uh layer 2 networks you have like 30 right now that are currently live on L2 L2 beat I think um I think another 70 or in production or something like this somewhere around 100 a little over 100 in production or currently live and um maybe actually you don't want to be always changing your endpoints to some of these things right maybe what you want to do is say I'm going to just send all my transactions to one spot and maybe it'll you know go on it on one network or maybe it'll go on another Network depending on something else more or less you can do that you can have a sequencer that's responsible for building blocks for two networks simultaneously or more um for now I think two is is a good starting point um there probably will be more in practice uh this is the notion of a super chain that the optimism people are sort of pushing forward so unsurprisingly here the colors are sort of um red and blue red for optimism blue for Bas both of which have the same underlying infrastructure and code um and both of which are moving towards a super chain where they'll have a single sequencer that builds both blocks or at least that's the idea that's what they've been writing about the implementations might differ but that's certainly the goal and what this means is some really cool stuff um first of all they have really easy time developing their Network because they don't need to build all of this from scratch they just Fork optimism and say hey we're on board with you guys um you know you're going to use your sequencer at some point we'll make sure we're always compatible and everything just works our blocks will look like your blocks etc etc and what this means is that in the future they could actually get some really nice perks if you have transactions that originate on Bas and then go to optimism and they have a shared sequencer well the sequencer can sort of be instructed or coded in such a way that you make sure that those cross chain messages are actually always in the same block for example so you can enforce atomicity and this is really cool because now you can do you know things that you probably can't do if you have different sequences you need to talk to because latency is existing in other places that isn't existing here um you could also reduce me in some way between these networks you also have really nice liveness in the sense that if one network sort of is up the other one is also always going to be up right you can share things like Escape patches and things like this and the structor coding it and actually running it also gets very very easy you can just point your metam master a single sequencer and everything just sort of works the downside is maybe on base they don't want to do this but maybe you know your your next chain on the super chain wants to extract me in this case the sequencer is working for a number of chains so it doesn't necessarily have any incentives to sort of bias your particular me concerns now maybe you could bribe it or do something but that's certainly not the goal of the super chain um maybe other ones will do it differently and the other downside is uh if you have very disjoint Networks you're going to have a hard time sort of um sequencing them because the sequencer sort of benefits some shared infrastructure you start changing what what the networks look like um you have to do a lot of work to the sequencer in which case maybe it doesn't make sense to have a shared One to begin with so you can have shared stuff that's pretty cool uh you can also do other things you can say I don't want to be responsible necessarily for ordering all of my transactions um as a sequencer because I am unreliable and what you maybe trust more than a sequencer which may be centrally oper operated is the underlying layer one itself like ethereum and so you can say actually ethereum you come up with the blocks for my layer to and as you come up with them I will build my Layer Two up and if you're clever about this you can still get performance gains everything can still be cheaper all of this but what you really really get out of this which is very nice um is the ability to have uh very strong liveness guarantees as long as your layer one is operational your Layer Two is also operational and if you do this in in particular ways you don't need to worry about sort of emergency functionality you don't need an escape PCH if your network is down because well your Network's never down either ethereum's down which is bad for all of us or both networks are always up and that's really really cool so um one team that's looking at this and they they've started doing this is Tao if you've seen them uh they have a smart contract where you can propose layer two blocks and it builds it that way there's um sort of some caveat here with the pluses uh they say Simplicity and cost I don't know right now that either of those are strictly true um Tao is one of the more expensive zi knowledge U ZK rollups to operate according to elter beat but in principle it's got a really nice design uh mechanism and it should go down once we have some really good you know proposer Builder separation type things in order to prevent uh sort of wasted execution of proofs or block generation and things like this but you can certainly do this and uh what this also means is that your sequencer can be sort of decentralized because if just calling a smart contract saying take this block and put in my L2 if you don't restrict who can call that function in the smart contract boom you have decentralized sequencing and permissionless sequencing which is really really cool um the downside is you might not get me again because you know maybe your block will be included by the layer one maybe it won't maybe it'll take someone else's depending on if there's a wasted effort in rebuilding these blocks uh and again you're probably going to get constraint sequencing depending on uh how this is built out and what needs to be done by the L1 to actually build the L2 blocks but this is really really cool there is also related Notions of um sort of pre-confirmation and sort of a lot of cross-chain atomic swaps that can come with base sequencing if everything sort of works but first we got to make sure that we have a really really strong um Contender for a base sequencer that's actually you know cheap and useful and people work on this it's really exciting but you know certainly it's um still a fairly New Concept just like um shed sequencing it's not quite sort of full mainstream develop there there's stuff to be done still um but if you want to take a simpler approach and you still want to have a centralized sequencing you can also change things right um one easy way you can do something is change the language on your layer 2 right solidity is really cool everyone loves solidity if you're on ethereum probably but maybe actually you want to use CYO star I did this right uh maybe you want to use C++ maybe you want to use Java don't think anyone wants to use Java but like you could do anything you want and once you start controlling the sequencer and actually changing your VM you can change a lot not just the language one thing we're doing at circuit is adding this notion of sequence or level security where actually what we're going to do is we're going to have um a system in place where we can look at layer 2 transactions as they come into the layer 2 and if they seem to be problematic we'll put them in quarantine and so what we're aiming to do is change how the VM Works in order to prevent malicious transactions from hitting our chain and this affords everyone who used this chain either as a user or as adap developer um an extra layer of security they get for free so at zerit we can look at the transactions figure out if they're bad if they're bad we'll not put them on the chain or at least not put them in the chain right away and that means people's funds are more secure um of course there are considerations for this uh for now it it starts with a little bit of centralization we need to make sure that we're on the same page about what good and bad is and there needs to be good performance but uh certainly There's an opportunity to save people lots of funds right even if adap you know doesn't do their their sort of homework and get an audit or you know has bad engineering practices we could guard funds within that dap which is really cool now I'm not saying do that please if you're deploying on circuit or anyone else or anywhere else get an audit you know build your dap as securely as possible but of course we can help if you're if you don't taking a step back um you know the Holy Grail of most of you know this ecosystem is decentralization base sequencing is one way to do that but there's other ways you can do this um people like uh the team at think at espresso are looking at other ways to sort of more traditionally introduce consensus mechanisms into um a sequencing system so instead of having a single SE sequencer that say we run or someone else runs you could ship the code on GitHub and have anyone be able to pick it up and run it and then they talk to each other under some specific consensus mechanisms and if that's performant enough and if that makes you know the right economic sense you can also um do this in a way that doesn't necessarily rely on the layer one to produce the um the layer two blocks and so in this case decentralized sequencing is actually just more General than base sequencing so you know base sequencing is decentralized sequencing but decentralized sequencing isn't necessarily uh base sequencing probably you know there will be some strong ties there but if you do this you get strong liveness guarantees because if you're a sequencer operator ever goes offline you just spin up a node yourself and you say Network's back up we're good right and as long as there's no barriers to entry and everything runs and you don't need specialized Hardware this is also quite nice nice the downside is um people have been thinking about this for a while it's very complex because you're introducing consensus back into the system which case sort of looks sort of like an L1 in which case looks kind of slow because you need security so there's stuff to be done stuff to work on but people are working on it people much smarter than me and uh I'm really excited to see it come to the future but you can do a whole lot more than that too sorry I don't know if the mic's been on the whole time I hope so uh you can do other things with the sequencer too uh we talked a little about me um you could build a rollup where your goal is to take as much me off of your sort of network users and put it into your own Pockets I don't know if you'd want to do that I don't know if the ecosystem would want you to do that but certainly you could build me extraction into it and then maybe you could pass that back to your users in the form of Cheaper fees or offsetting costs for publishing data to your L1 or wherever you store it but you could do this um people have talked about this I don't know if any teams any actually working on this but it would be really cool if if it did exist uh I don't know if there's a there's a good Market appetite for it but be a nice technical challenge um you could do other things like Fair sequencing base sequencing allows you the opportunity to always have sort of strong liveness concerns about being able to sequence layer two blocks whenever you want but what if you want to be fair what if you want to do the opposite of an me boost and say I definitely don't want someone to be able to sort of craft blocks in such a way or submit transactions in such a way that they can profit unfairly you know maybe you can put your sequencer behind a zero knowledge rollup or sorry a Zer knowledge proof framework that says um you know the SE quencing algorithm is always doing this and that algorithm is fair therefore the sequencer is always fair I don't know if that's easy I think it'd be really cool people could do it of course if you also start looking at where the transactions come and go from the sequencer you could talk to other chains shared sequencer affords you the ability to do uh Atomic swaps between networks that you can use sequence but that doesn't mean your inputs sort of only need to be tying to one layer one maybe you could also read from different layer ones uh and in which case you know you're still anchoring on one of them or maybe you're not I don't know uh but maybe you can do some kind of maybe not always Atomic cross-chain messages but some cases maybe you can get some benefits um right so if you do some messages from binance and others from ethereum you know maybe there's some way you can make them really tight and you know there's some good use cases for that I don't know and then the last thing that probably really needs to be talked about is still the notion of what happens if your sequencer offline um you don't want your network to go down so if you're not a base sequencer you probably still need to do something uh and in the past sort of we call these things skap patches or more sort of accurately these days um forced inclusion mechanism so if your network sequencer goes down you still want to be able to sort of update your L2 State especially as it sort of implies the ability to get you your funds out of the network if the network is down but there's also a whole lot of other stuff you can do so you can change the block size you can work about work with pre-c confirmations you can do a whole bunch of of other stuff in short the the sequence of tomorrow is going to to be very different than the one we're seeing today right now it's all about evm equivalence and a couple little buzzwords that sort of say I'm doing this a little bit differently or I want to do something you know cool um but they're they're still immature there's a lot of work that needs to be done for most of these networks shared sequencing for example um has been talked about but I don't think the atomic ability uh to send cross chain messages is there um so when you start looking at these these buzzwords I want you to be cautious of what they are and what they're promising to be because there's often a very big uh discrepancy and so Step One is figuring out what they should be and then step two is figuring out what they actually can do and then step three if you've got time or you got a research budget or you know you want to hack something together is figing out what what they should do for you in the future and then building it out because that's also really really cool when we have um hundreds of of rollups having you know 80% of them do exactly the same thing probably isn't going to be that interesting you know you're probably going to need 20 or 30 of those and beyond that sort of kind of useless uh so do cool new stuff with your sequencer take some bold risks you can do stuff where only your community needs to be really really in invested in it because you don't need to Target all of the ethereum uh ecosystem share and that can still result in some very big pieces of the pie for you but also doing something really cool and interesting that some people can find very very valuable and useful so in short you know take a look at these things keep building them and keep paying attention to what they are and what they should be thank you we have time for one question while we get our next speaker set up who's got a question for Yan questions yes I will hand you my mic uh does sequencer should work only with the same EVMS or it could be uh um sorry uh repeat the question did you say shared sequencers only work with evm was the same like evm like style like if it's like e evm or I can just do sequence in between like salana soie do hyper liquid kind of stuff on the engineering point of view I don't see why you couldn't change to support multiple VMS uh obviously the more different they are the more engineering work that will take uh and maybe there are some properties that will conflict across certain things I can't think of any in principle but uh you know I think it's mostly an engineering function so if if you want to spend the time figuring out what it means to make an atomic swap between a salon of VM and a theum VM and then put that in a sequencer I don't see any fundamental reasons why that wouldn't work just Tech just yeah great okay so thank you thank you so much please give another big round of applause for Yan