Parallelized EVM - The Future is Now, Old Man

Parallelized EVM: Monad vs Sei

March 4, 2024

In conclusion

Morning lads. Get your cock out of your hands and clean your ears. We are diving into parallelization. This cycle, we have two juicy courses. Modular blockchains - think Papa Celestia, who wants to eat everyone’s lunch - and the second iteration of the old guard (Ethereum), think Parallelized Ethereum Virtual Machines (EVMs) like Monad and Sei Network. 

These are the big upgrades we get to the world computer(s). And obviously, in a game where value runs downhill, everybody wants a piece of the stack.

If you made money on the layer one trade last cycle and managed to keep it, congratulations. If you missed it or bag held into the void, good news! This cycle we are running it back turbo. The layer one trade is great from loads of angles. If you’re trading with size, all of them are relatively liquid. And it is also really simple. The thesis is literally: ‘‘I am buying token X because I believe people will want to transact on chain-X. More users mean more people purchasing block space, and this demand will increase the value of my token.’’

The base asset gets used as the money within an ecosystem and blockchain lets us plebs own the base money. In TradFi, we get stuck with second-class shekels, but in the beautiful on-chain realm, we get our hands on first-rate guap. 

Rule number one. Never fade the simple things. And please, for the love of god - either be in the modular or parallelized EVM trade this cycle (preferably both).

There are lots of things to say, but first, let's get grounded. 

What is the Ethereum Virtual Machine? 

Blockchains are computers. Imagine the EVM as a cloud computer. Node operators all over the world contribute and create a virtual computer capable of computing smart contracts. That is all the EVM is. A cloud computer that executes smart contract transactions. Nerds often toss around the word ‘state.’

State is literally just that - the blockchain's current condition/status/state. The state is updated with each new block, and it is the record of all transactions that the network has deemed valid. All network participants have to agree on each new block; when they do, there is a state transition. The concept of state is crucial, and before we get into the weeds of consensus mechanisms and double spending, we need to take a quick breath and a step back.

A block is data. It contains a list of transactions and smart contract instructions. You can only stuff so much into one block, so you’ve got to keep adding them. The state is the sum result of all agreed-upon transactions added to the chain. It is the pure record of all valid transactions. Here, we can already see the importance of state syncing. Basically, consensus is the method used to reach an agreement between nodes that transactions are valid; therefore, the current state is an accurate and non-fraudulent record of transactions. State underpins the whole decentralized jungle.

Why does state matter? Well, when you work out transactions sequentially, it is pretty easy to harmonize state. But today lads, we are looking at parallelized channels, and updating state gets a little complicated with concurrent execution. 

The Bores of Sequentialism

At first glance, the EVM seems wicked, and in truth, it is. A global computer capable of executing smart contracts that allow you to put your money where your mouth is and do almost anything imaginable on-chain. However, the EVM operates sequentially. It is single-threaded. It works through transactions one after another (sequentially), and this is actually pretty weird. 

Modern computer hardware has multiple cores and is capable of executing multiple work streams simultaneously. Worth noting that almost all upgrades to hardware have centered on increasing core count instead of making each core more powerful. 

Each year, humans apply themselves and some crafty methods to add more cores at a lower cost basis. So, yeah… we are using a piece of tech built of hardware that can process multiple work streams but chooses to execute in a single file. Single-threaded EVMs basically toss out the bulk of human progress. Obviously, this is a disgusting technical oversimplification. Just because something is possible does not mean it is easy. The boys over at Monad have a great diagram highlighting the differences between single-threaded and parallel execution. 

Parallelizing the EVM: What it Means & Why It’s Better

Feast your little peepers on the optimization gains. 

In a sentence, parallelization means multiple transactions can be executed simultaneously, unlocking better performance. The blockchain executes more transactions per second, and gas fees for end users decrease because they are not bidding against every other Tom, Dick and Harry to get their transactions included.

A parallelized EVM blockchain has parallel execution channels, and each of these can have its own fee market. Instead of all ecosystem activity driving up any cost of interaction, there are now supply and demand dynamics based on operation types (execution channels). This is how we can improve resource pricing and create separate fee markets with parallelized EVMs. At this stage, we are starting to converge on the type of tech that can compete with the TradFi ladies and gents. 

Cleaning Up the Hidden Questions

A fair and valid criticism may be:

  • “Why the hell are we as an industry relying on a single-threaded EVM?” 
  • “Why has the EVM become almost absolutely dominant?” 

These are good questions. And, although a discussion of alternative virtual machines is beyond the scope of our investigation, you can go check out our good friend Andy’s AltVM Thesis.

We have to understand the resources in the stack. dApps require devs. And devs are the human capital of the space. Their value is scandalously high, given that we are playing a permissionless game. The crypto native talent pool of devs is familiar with and accustomed to the EVM, and it is why everybody has and always will bend over backwards to attract them. All these scalability gains don’t mean shit unless you’ve got coding wizards churning out applications.

Ethereum has, by leagues, the richest research network of any chain, and the volume of material to lean on is incredible. These coders are kings, and the value proposition of being able to port a dApp from one network to another without change is a bazooka proposition. Solidity specialized devs are the largest addressable market, and parallelized EVMs are a massive vampire attack for EVM liquidity. (Sound familiar to the layer one trade yet?)

Also, let us not forget that most people who made money in crypto got rich buying Ethereum and now have bag-bias. VCs want to invest in projects that align with their cognitive bias, and this knock-on effect is a piece of background music flavoring the entire party in Ethereum-centrism. (If you cannot read between the lines, this means I will be rotating my portfolio into Monad when it launches with the force of a thousand suns.)

Welcome to the Parallelized World: ‘Look at the Strength’

Transactions Per Second (TPS): Roughly 12
Time to Finality: 15 minutes

Sei V2

TPS: 28,300 batched transactions

Time to Finality: 390 milliseconds


TPS: 10,000
Time to Finality: 1 second

The numbers don’t lie. No two ways about it. The upgrade is meteoric. And we kind of already knew that this was going to be the case. Look at the team behind both of these projects. Jay Jog, the cofounder of Sei, was an engineer at Robinhood. Keone Hon, co-founder and CEO of Monad, cut his teeth at Jumping Trading, leading the High-Frequency Trading (HFT) team. 

The team at Monad have spent their careers building trading machines capable of slurping millions of packets a day and responding in single-digit microseconds. People should build what they are good at, and Monad is ludicrously skill-tiered towards execution. Monad is the guy playing an RPG who spends all his skill points on agility.

Here is a good moment to introduce the concept of High-Fidelity DeFi. All of these players want to introduce an experience equivalent to or better than that available in TradFi. This is the kinda shit we are looking at this cycle. Parallelization has one goal. Scale blockchains to the point where they can compete with the big boys. Twitter, PayPal, Visa, and the rest. They have disgusting transaction throughput. Keone has a great insight: the systems are limiting the scale of any application to achieve Product Market Fit (PMF) and grow its user base.

Parallelized EVMs open separate execution channels, create separate fee markets, and, in layman’s terms, eek all the juice available out of the existing hardware.

Source: https://blog.sei.io/sei-v2-the-first-parallelized-evm/

Sei V2

Sei is a layer one with excellent throughput built to enhance the trading experience. Rumor has it that Jay Jog saw Robinhood shut down trading for retail during the GME debacle and decided to build a permissionless, highly liquid capital exchange. Not sure if true, but it's a good story. 

Cutting it short, Ethereum could not support this, and even rollups are limited by the performance of the base layer. Fast forward, and we have Sei launching on mainnet in August 2023. But hold onto your cockles lads. Sei V2 was announced in November, with expected deployment in Q1 this year. Gotta love a team that gets it popping.

Sei V1 is already throwing up impressive numbers with constant activity above 45 transactions per second and block finality at 390 milliseconds. Sei’s current contract language is Rust, and Sei V2 will integrate support for Solidity & CosmWasm. Remember all this EVM parallelization is to absorb liquidity from the existing EVM-compatible chains and monopolize the dev talent pool. And today, we are focusing on the beauty that is Sei V2. 

Source: https://blog.sei.io/sei-v2-the-first-parallelized-evm/  

Big Ticket Upgrades

- Backwards compatibility (all EVM smart contracts can be deployed without changes)
- Optimistic parallelization
- SeiDB
- Interoperability between Sei and V2

Backwards compatibility means that anything live on Ethereum can be literally copied and pasted to Sei V2. Sei does this by having nodes import Geth (the most popular Ethereum client). This has the additional advantage of allowing Ethereum tooling to be used, aka wallets like Rabby & the dreaded fox. 

SeiDB is a big upgrade to storage, optimizing disk usage. SeiDB reduces state bloat and increases storage efficiency while powering faster consensus since disk access requirement is reduced. The interoperability feature is very cool. CosmWasm and EVM contracts will be able to communicate with each other, and devs can access native tokens. This ease of access for devs is an excellent play for the talent pool. In short, Sei is getting a major overhaul, and there will be plenty more horses under the hood.   

Optimistic Parallelization: Niggles of the EVM 

Optimistic parallelization deserves a little bit more attention. Both Sei V2 and Monad introduce optimistic parallelization. Why? It is a constraint forced upon them by the EVM. Currently, there exist two ways to introduce parallel execution. You either have state access lists, or you go the optimistic route.

The Solana Virtual Machine (SVM) has state access lists. Every transaction includes information about where it is going to read and write. Nodes understand which transactions are non-conflicting and execute them in parallel. But here’s the rub with the EVM. Access lists have to be specified upfront, and introducing them would break EVM compatibility, and then you’ve lost all the gains of building an EVM-compatible chain in the first place. Nodes receiving EVM transactions have no idea what state changes will occur or what storage slots will be used. But this does not matter because they spit out one transaction after another.

To parallelize an EVM the optimum route in existence today is an optimistic approach. Basically, run all the transactions, and if something conflicts, you run it again. Worth mentioning that state access lists place a heavier technical burden on developers. There is more of an onus on Sol devs than Eth devs. But with frameworks like Anchor on Solana, the reality is that much of this complication has already been abstracted away.

Left Curve Definition: Optimistic parallelization relies on trial and error and if there are conflicting transactions nodes just run them again.

Source: https://twitter.com/monad_xyz/status/1745866042579042459/photo/1


Monad is the other big player in the parallelized race and has already started teasing the release of its testnet. It is an absolute beast out the gate with the team understanding the importance of building a community, and the guys building this are ex-wizards from Jump Trading’s HFT team. They have the memes and the technical chops. Not to mention Monad’s seed round was absolutely filthy.

Source: https://twitter.com/monad_xyz/status/1625591020070002689 

Bruv, just look at this list. Haseeb and Burniske have better eyes for value than most and look at that in the corner. The man, the myth, the legend himself. Señor Cobie. Crypto-natives are going to bid this shit to infinity. (no disrespect to Hasu or 0xMaki) but few exist in the space who are more revered. The closest equivalent this cycle has gotta be Ansem.

Back to Monad. This baby purportedly has the capacity to pump out 10,000 transactions per second with 1 second finality. How does Monad deliver this performance? It introduces its own list of upgrades:

- MonadBFT
- Deferred Execution
- Parallel Execution
- MonadDb

Notice how a lot of these gains come from optimizing data storage. When talking about Monad, you’re going to hear the word ‘‘pipelining’’ all the time. For my fellow left curvers, all it means is utilizing all of the system’s resources efficiently through introducing stages of work- stuff gets scheduled so that every ounce of muscle in the system can get used.

MonadBFT is the consensus mechanism that uses a two-phase approach to agree on transaction ordering. It is a proof-of-stake model, and Monad’s docs say this:

‘MonadBFT is a high-performance consensus mechanism for achieving agreement about the transaction ordering under partially synchronous conditions in the presence of Byzantine actors.  It is a derivative of HotStuff with the improvement proposed in Jolteon/DiemBFT/Fast-HotStuff, which is the reduction from three rounds to two rounds by utilizing quadratic communication complexity in the event of leader timeout.’

That’s some big-brain shit. For my more simple brothers it's a consensus mechanism with high performance that guarantees transaction ordering even if bad guys try and play some tricks. It is more efficient at reaching consensus than the current existing EVMs. 

Deferred Execution is one of Monad’s truly influential breakthroughs, where the agreement on the transaction order is separated from executing the transactions. Monad nodes agree on ordering decoupled from execution. These wizards have pipelined consensus-executions staging, and severing this link allows the network to agree without having to wait for execution. The next set of transactions can then be processed, and the scaling gains start flowing in.

Parallel Execution should be pretty self-explanatory at this stage, but we are going to delve into MonadDb with a little care. This Parallel Access Database unlocks high-performance state access with minimal requirements, and Keone has a great analogy explaining it. Parallel performance requires effective state access, and bottlenecks form for single-threaded EVMs when they access state data. 

Keone uses the example of cooks retrieving ingredients from a special and complicated cellar to cook a feast. Only one cook accompanied by the ‘wizened maester’ can go into the cellar at one time (state information from storage).

The more blockchains scale, the more horrible this bottleneck becomes, as you can have all the execution in the world, but it makes no difference without an efficient method for state access. MonadDb is this custom database that allows multiple cooks to retrieve ingredients in tandem, and MonadDb powers concurrent reads and writes, literally obliterating the state access bottleneck. Fucking gg, eh. 

Monad Oh Sweet Monad

Monad has full backwards compatibility, and anything deployed on Ethereum is a copy-and-paste job. The general thesis is to build a global state that really cranks up the revs and bleeds the tank dry, maximizing bandwidth and node processing power. The wild thing about Monad is that although the chain processes in parallel, the strategic execution arrives at the end state as if the transactions had been executed one after another. There is literally zero difference between Monad blocks and Ethereum blocks. 

Bear Case 

When are we going to let go of the EVM? You have a mad amount of work going into an imperfect virtual machine. Ethereum’s dominance is increasingly hindering the space, and somebody ideally needs to kidnap a small town. Teach them all to code in a new language and spin up a new VM. This is the solution. 

Any VCs should genuinely look into this, pay for education, and have a clause where the poor souls have to churn out code in the new language. Hello, indentured labor! Making crypto great again! EVM upgrades are pouring capital and talent to work backwards from an imperfect start.

Another viable argument, and hear me out on this. Why do we need new layer ones when we can just build loads of rollups? Especially with the alternative data availability (DA) solutions on the market. Just spin up loads of general-purpose layer twos using Celestia or EigenDA, have them settle on Ethereum, and let people like Axelar iron out the cross-chain issues. All the bag holders win. Rollups are kind of a dirty form of parallelization with execution environments operating in parallel anyway.

One central concession has to be made. Rollups focus on execution, whereas a parallelized EVM upgrades the virtual machine itself. It is difficult to argue against a technological shift that leverages modern advances in computer hardware and creates separate fee markets. Parallel execution will likely become table stakes in the coming years. 

Sei Vs. Monad

It ain’t about being right. It is about making money. Truth be told, nobody knows what will happen, and you’ve got to wait for the things to go live before making any genuine judgments on performance. 

What follows is a mere humble opinion, but the Sei experience (in its current form) is nothing to write home about, and it already looks overvalued (not that that means anything in the current meta). Sei V2 can do everything it likes, but going head-to-head with Monad is kind of a non-starter. Happy to be wrong, but you’ve got an ex-Robinhood guy squaring off with the elite of Jump Trading backed by CT OGs. People bid what people want to bid, and gun to my head and maximum portfolio rotation mandatory. I would dump everything into Monad. 

Monad is going to surf the bull market fever perfectly. It’s already juicing everything in preparation and will catch the largest parallel EVM bid. Just don’t baghold into the next cycle. VCs are still raising for layer ones and will continue to do so because it is easy to dump on us retail plebs. There’s a lesson in there.

Welcome to the parallelized EVM future, fuckos. 

Opening MetaMask...
Confirm connection in the extension

The current connected wallet does not hold a LARP. To get access to the Meal Deal please connect a wallet which holds a LARP. Alternatively, visit Opensea to purchase one or visit Join the Meal Deal to purchase a subscription

Table of contents