Inside Glamsterdam: How EIP-7732 Redefines Ethereum | terence tsao

Hello. Hi, I'm Terren. I am from Option Labs. I am an Ethereum core developer. I work on Prism.

So today I want to talk about how like EIP7732 EPBS redefine Ethereum. So I want to start the talk first by saying scaling is the primary number one priority right you have seen this multiple times on the internet even in the blog post in the EF that like scaling is the most important thing that we can do today right and then we did that last year we shipped peer does last year December 2025 that changes how we scale data availability and I want to say we're going to do this again this year in 2026 which will fundamentally changes how we scale two things. The first is DA the second is execution. So we're going to scale more this year. So how are we going to scale?

So pier does what we shipped last year scales DA and then it uses eraserial coding polio commitment something that like I would say probably 99% of this audience probably not super familiar but it works and that's all it matters right and um EPBS kind of scale Ethereum differently this is through pipelining and this is something that's so simple that everyone in this audience will be understand at the end of this talk I promise you and yeah so so basically something we'll get through more so before we talk about scaling let's ask today what is the bottleneck right to run a node we don't want to sacrifice decentralization we want everyone to be able to run a node even your internet is 100 megabits per second you have a say like a Raspberry Pi you should be able to run a node right but what's the bottleneck bottl net is that you need to download a block you need to download three things the block the payload and the blobs and those are very bandwidth heavy then after you download them you need to verify them you need to verify the block you need to verify the payload payload here are the execution stuff like basically what the EVM does right bandwidth is the most important part is the most expensive thing compute is the second most expensive thing and here is the Ethereum slot basically what I said in the last slide let's translate this into a diagram right you have 4 seconds to do everything. You have to download the blob. You have to download the block. You need to verify everything. If the block passes the 4 secondond mark, then the block gets reorked.

It becomes a missing slot. The whole 12 seconds is missed. So why is bandwidth a bottleneck today? Just like bandwidth like contradict with scaling, right? Because the more you want to scale, the bigger the block size is.

There's no way around it. And the bigger block size is the more your node has to download everything. And um surprisingly today the consensus part of the block is only five to 10% of the total block size. Consensus here means like the attestations your validator votes but majority of it is this execution part which is the EVM stuff. And then we cannot scale further unless we do something about this right.

And shout out to this side that Ethereum Foundation has made that just shows that the purple part is the execution part. The the green part is the consensus part. And today most of the block size are dominated by the purple part which is the execution part. So here is an Ethereum slot right again another diagram that we have 4 seconds and the max we can do today in the spec is 600 mil execution gas. That's the max.

And then we can do 21 blobs. is about 2.68 megabyte, right? We can exist that maybe but once if we exist that limit that chances are it can go beyond the 4 second mark and then the block can get rewarded or no one else can run a node, right? So now you guys are sharp.

You guys probably are as uh asking the question why only 4 seconds, right? Because between four to 12 seconds there's still eight seconds left. Why can't we do something there? Right? What are the nodes doing there?

The nose is doing something there. The nose is aggregating all the votes together, right? And then but then like can we actually use the second half of the slot for execution for blobs, right? Like wouldn't it be nice to essentially use the whole 12 second for bandwidth for compute such that we can process more transactions, we can process more blobs, right? Now what can we do to get us there?

So EPBS is enshrined proposal builder separation but you can also think about this as execution payload and block separation right the idea is not very new it's originated in 2021 vitalic has written about that before right but the original version was that validators has to be decentralized builders are more sophisticated they have the order flow they they have their special packing algorithm so why don't We separate those two together as two different iteers in protocol such that this is a more efficient market structure right and then later his proposal actually instead of one slot it becomes two slot because you need two uh a tester votes to vote on the second part which is not sustainable. You don't want to increase the slot time. We solved that by having another payload committee such as this smaller size we can fit within 12 seconds. But if you see the later proposal which was proposed in glam sedan we emphasize more about the key benefit being better slug utilization because of scaling. It wasn't so much about like we want to improve the market structure which is still there.

The benefit is still there but the proposal becomes more like the scaling proposal which we end up realizing. EPBS is the least resistant way the most efficient way we can do today to scale the execution to scale the blob and then it was uh finalized as a headliner last uh last year. So what so what so what are we doing here? It's very simple. You separate the block from the payload.

Today block payload are together. You separate them. You have your block which has a commitment to the payload and the payload is a separate network object and then you have some testers validators that votes that payload is uh arrive on time. So here is what an EPBF slot look like. You have a second zero you're proposing the block.

you have second two before it's the second four you were testing for the block and then if there's enough attestations then the builders will reveal its payload between the five to six second mark and then there's another committee that attests for the timeless of the payload at the 7 to9 second mark and as you can see from here kind of right the whole slot is being utilized so what's the benefit here right let's let's recap what the benefit is today because of timing gain and everything most validators most nodes have two seconds to process the block. Now you have between 9 to 12 seconds to process the block. 9 to 12 second to download the blobs, right? Which means that the asertestation you can get a head vote, right? More the chance you see less reward and it's even bonus.

This is good for shorter slot time because when you reduce the slot time for the shorter, you want to utilize the whole slot. This is also good for ZK proving as well because now as a prover you have more time to generate a proof. And this is what the EPBS slot look like, right? Before, remember, we have four seconds and then you have just two brown lines, but now we're utilizing the whole 12 second. And what is really cool about this is that like you can play this tradeoff game, right?

Because blob is slower to download, but it's very very fast to verify versus execution payload. It's slow to download, but it's also slow to compute and verify just that you actually have to process it. So you can kind of like do different deadline. You can say okay as a blob the deadline is the end of a slot but as a payload the deadline is a 9 second march. So we can give 5 seconds to download and then another 3 seconds to verify and most of this will comes from testing to figure out what's the right trade-off is.

So I only had 15 minutes so I didn't really like cover like 90% of the other things. There are a lot of benefits beyond scaling as well such that like what happens when you essentially allow the proposer and the builder to essentially uh sell and buy the block together. What happens if like that whoever proposed the payload their identity is on chain? What can you do? what are the stakeholder impact right as a proposer as a builder or even as a relayer like what changes in the EPSBS model finally um what does EIP7732 enables for us in the future right just now we have this builder on chain you can actually enables a lot more future road map iteration such as like you can do the auction you can do slot auction for example instead of selling the block commitment up front you can just sell the whole slot right you can do MBV burn for example code.

You can even lower the insurance that way. So, yeah, this is all I have for today. I'm around, so feel free to um ask me any questions, but this is coming um this year and hopefully we'll see another big increase in terms of gas limit and the blob. Thank you.