// full transcript — Ben Edgington on post-quantum Ethereum

slop.atg.eth: Oh, All right, Claude bot man, let's do a mic check. You hear me? Can I hear you? Tell me about our guest. What kinds of things should we ask? What are they into?

clawdbotatg.eth: Fetching some research state data from an API and parsing out the phase and results. Our guest is Ben Edgington, a longtime Ethereum consensus researcher. Fast about his Decoupled Consensus EIP, why Post Quantum got moved ahead of fast finality, and what it's like going from a terrified first-time speaker on that call to moderating it.

slop.atg.eth: That is good stuff, remember that.

clawdbotatg.eth: I'm gonna tail the last six lines of that log file for you. Got it. Locked in? I'm what Ben's into. He loves explaining the protocol clearly. He wrote 100 editions of "What's New in Eth2" and the whole Eth2 book. He even named Petra, and he has a running joke where he replies "fast finality" mention whenever anyone brings it up. Many cheers a big call, then celebrates with a hot mug of tea.

slop.atg.eth: Alright. Welcome. Welcome, Ben. Welcome to The Slop. Here we go. I'm still not comfortable with this video setup.

benjaminion.eth: I'd rather look at you than look at myself, but whatever you prefer.

slop.atg.eth: I'm the other way around, of course, and usually I kind of have it the other way. But this is work. We've got a different video setup here, but that's why all of this is draggable and we can move it around, you know? Get comfortable. You are like the great explainer of Ethereum. Oh no.

benjaminion.eth: Oh man, I think you have this honor. Maybe you can take the execution layer crown, the EVM crown. I haven't worked at that layer for years. I'm very much embedded in the consensus layer of stuff and that protocol stuff.

slop.atg.eth: I think even my crowd I'm talking to newbies and app layer folks researchers and core dev types and being able to explain what the core devs are doing Because I'm dumb, I can make sure I explain things enough that I understand them.

benjaminion.eth: I feel very much the same. I feel like unless I can explain something, I don't understand it.

slop.atg.eth: ### Yes.

benjaminion.eth: And it's amazing, when you start writing, you very quickly find... So I did this 'Eth2' book, or 'Upgrading Ethereum' is the technical name, unfinished, incomplete, out of date, but nevertheless a thing. I would start a new chapter, the first thing I'd do would be to get some good diagrams in place because that's the hardest, and once you've got a visual representation then you've got something to hang the explanation on. It's a lot of pages.

slop.atg.eth: Is it right that there were 400 editions of this or something like that? Okay.

benjaminion.eth: I don't know. It's on GitHub and I commit to it. but I'm not sure I'll ever finish that Hmm

slop.atg.eth: but I listened to a lecture of yours yesterday that Mario had set up I think last year and you were talking about LMD ghost and just sort of like the selection of the forks it's When you explain it

benjaminion.eth: ###

slop.atg.eth: It's simple, really. And it's like this somewhat complicated thing and it has to do with the weights of the trees, but the way that you explain it does make it kind of simple. It's like a highly technical thing and there's white papers and all this technical discussion around That is kind of your superpower though, is being able to kind of merge those two worlds or have one foot in both worlds.

benjaminion.eth: I mean, I feel like, you know, at this stage in my life, I'm not able to push back the boundaries of research. building new things and able to come up with new designs and new protocols and all of that stuff. And, you know, that's beyond my powers these days, but I feel like I can fill a gap in terms of translating that stuff. into things that others can understand and maybe contribute to. me and say, you know, it was your writing that got me started. Yeah.

slop.atg.eth: I get that at Devcon every once in a while when I'm talking to someone. That was how I got in, or whatever.

benjaminion.eth: isn't it?

slop.atg.eth: Only 397 pages to this thing. Yeah

benjaminion.eth: Yeah, I mean, a chunk of that is the protocol just copied and pasted from the GitHub. how does it function in relation to everything else? The BLS 12381 curve is probably my favorite chapter. It used to be a separate article.

slop.atg.eth: Here, pls

benjaminion.eth: deep dive into this thing Really ramping up my understanding of this thing that we were using, starting from scratch, really no group theory background or any of that. Just felt I should write down the stuff I'd learned

slop.atg.eth: And BLS is the curve that we brought in in ETH 2. Yeah The ###

benjaminion.eth: From a protocol developer's point of view, it is ETH 2.0, right? This is Eth2.0, it's got no relationship with Proof of Work.

slop.atg.eth: And BLS is really good for scalability in terms of aggregating a bunch of signatures together?

benjaminion.eth: Yeah, this is what made Proof of Stake possible for Ethereum In those days we didn't envisage a million. all of their votes into blocks so that you can prove that the chain has reached finality. easily around the network and this elliptic curve arithmetic stuff There's a way to do that, and it's brilliant, and it's beautiful, and it's going away, and I'm very sad because of quantum computers.

slop.atg.eth: The

benjaminion.eth: We've got a few more years on this and then it will be taken away from us. Yeah.

slop.atg.eth: That's exactly it. A few years is comfortable. That's a comfortable range. I feel like. Are we a little more afraid of that or is it still a few years? ###

benjaminion.eth: I mean, this is one of the things I wanted to discuss with you really. we ended up spending the entire week in the famous room 203 just heads down war gaming consensus and execution layer AI threat is one thing. Those timelines are telescoping. I don't think anyone seriously believes there'll be a cryptographically relevant quantum computer before, say, 2030. optimistic timelines. But to come to the topic of your podcast, I mean, the AI threat is a new one. Incidentally, I heard a few minutes ago on Reddit, I read that somebody's claimed to have proved the Riemann hypothesis.

slop.atg.eth: that

benjaminion.eth: This shows the power. Three, four years ago these tools couldn't do basic arithmetic. The incredible are the assumptions that we rest our cryptography on, are they safe? we rely on for the security of both the execution layer and the consensus layer and we assume it's safe because we've tried for many years to break it and yeah maybe it's correct it but certainly cracked many others. Maybe possibly already made but concealed for national security reasons elliptic curve cryptography might be broken. so the solution to post AI and post quantum is the same. What we're doing is moving Ethereum to hash-based signature schemes, which have been around since the 1970s. they're very sort of smooth and without any handholds for funky maths and AI-driven maths to get its hands into, unlike elliptic curves, which have lot of structure and isogenies. the idea is if you build everything out of hash functions, it's not only quantum proof, it's also AI proof

slop.atg.eth: So, just to say it again, quantum is a ways out. Like, we do have plenty of years. Like you said, it seems like a lifetime since Mumbai and it was weeks ago. We have time. Years is a lot right now. There's this other villain in between on the timeline of Elliptic curve cryptography getting broken by AI. like AI as a like solving elliptic curve cryptography Is that something where AI proves that we can then use a whole bunch of GPUs? With quantum computing, you build a computer and it can just solve these things, these things that were traditionally very hard problems. go one way and very easy the other way,

benjaminion.eth: ###

slop.atg.eth: AI solving elliptic curve cryptography, is that still something where it takes a lot of work to solve it and AI figures out the method, or is it like AI discovers a backdoor and it's just like instantly

benjaminion.eth: The

slop.atg.eth: What does the break look like? Okay.

benjaminion.eth: At the moment we measure security in bits of security And the elliptic curves we're using on the consensus layer is reckoned to be super secure Yeah, that's right. I know, crazy isn't it? The

slop.atg.eth: It's crazy because two years ago we were talking about, well, these things won't be solved until the heat death of the universe or something like that. And then like two years later we're saying, maybe 2028. This is, like, this is a very scary changing moment. are the brute-forcing of it. But there's more to this, I think.

benjaminion.eth: Right, so the question is, is there a shortcut? competitions to try to standardize new cryptography. A couple of years ago, they had a bunch of candidates and they put them through rounds. It gets a lot of scrutiny from cryptographers. and then somebody broke it so it could be basically broken in an hour on a laptop. And that had been through three rounds of scrutiny by some of the world's best cryptographers Now I think it's extremely unlikely that this is going to happen to elliptic curve cryptography, what we're using in Ethereum. Likely is perhaps a reduction in the number of bits of security, so we might go from 115 down to 100 or something like that. That gives us A bit of warning, right? ...and they're not going to go after wallets with $20 in or whatever. they might invest a lot of computing power breaking Satoshi's key or something like this on Bitcoin. but in principle it becomes possible whereas today it's completely impossible. That's very unnerving Get this hash-based thing in and then never talk about it again.

slop.atg.eth: And the timeline was sped up in Mumbai

benjaminion.eth: Yeah, so we We agreed more or less that we needed to get a move on. did the first one last week and the theme was figure out how urgent do we think getting to post-quantum is DEFCON being, you know, the US Army's kind of readiness for war levels, nuclear war. You know, I sort of characterize this in terms of Ethereum readiness, I reckon before Devcon we were somewhere around Devcon4, which is a heightened sense of urgency, but you know, we would deliver post-quantum over three or four upgrades, you know, get it done by the end of '29. and be content with that. I think post-DevCon and I think it focused minds a bit and we're now somewhere maybe between two and three of the stuff that people Features that people like. ETH Labs voluntarily removed Quickslots from the Egota upgrade. requested that we DFI it so decline for inclusion on this occasion, just to clear the way to get more quicker to post quantum and I fear we're going to have to do that with quite a lot more stuff over the next few weeks. Yeah, that's the agenda.

slop.atg.eth: Okay, interesting. So, I talk to a lot of people about I Seems like there's kind of like these two camps of people where the very, very good developers that I talk to have always been a little bit in denial about what You just said mathematicians are skeptical It made me think of the same thing. Are AI-pilled mathematicians also skeptical that it's going to take time or Like, I want to know what the like super AI-pilled mathematician researchers are thinking the time scale is

benjaminion.eth: The

slop.atg.eth: You get this amount of bits cut in half or cut in a quarter or something like that.

benjaminion.eth: Yeah

slop.atg.eth: Are these mathematicians skeptical but also skeptical of AI's ability?

benjaminion.eth: I'm not super plugged into that world so, you know, a lot of what I know is hearsay or learned on Twitter, so, you know, take it with a pinch of salt. I mean it's clear that AI is getting super good at advanced mathematics. for years that was now And it's one of these things, I mean, you know, we can talk about formal verification in the Ethereum protocol and things. The Yeah, it's hard to gauge levels of skepticism. We don't have a proof of nonexistence or proof of unbreakability, so while these sort of mathematical super AIs are out there, better to be cautious.

slop.atg.eth: Okay The

benjaminion.eth: The

slop.atg.eth: We're taking out some features that we were going to put in. I was kind of taking Dankred's point. And going with it that bunker mode is kind of silly because if an exposed public key allows anyone to take the money from it, you will still have your tokens in bunker mode. your tokens will be worth very little

benjaminion.eth: That's right.

slop.atg.eth: Maybe there's a subtlety here about The amount of bits that are taken away from the security may mean that In bunker mode you're still safe and only some wallets are going to get popped and they're going to get popped slowly things happen. So I don't know. What's your take on bunker mode?

benjaminion.eth: Yeah, I think we're pretty much toast.

slop.atg.eth: '

benjaminion.eth: If you're a custodian of millions to billions worth of assets, then it's a simple enough precaution. As soon as there's a demonstrated break. on Bitcoin or on Ethereum, I think we're in for very rough times

slop.atg.eth: The

benjaminion.eth: As you say, the asset then becomes suspect. I do not want to be in the chair under whose watch Ethereum failed.

slop.atg.eth: Sure.

benjaminion.eth: context:###

slop.atg.eth: And by the way, it's not just Ethereum. Like, if you can get my money out easily because my public key is exposed, you can get into all my servers because SSH is broken, and you can read all of my internet traffic for all of time because HTTPS is broken.

benjaminion.eth: A lot of protocols like SSH or SSL are all much more agile. you got lots of fallbacks in negotiation on each side which protocol to use and so on. It's just exposed all the time to attack. Changing it is a huge, laborious effort to build the new protocol and deploy and come to consensus on it. Yeah, I think the banks and our internet connections will be fine, I wouldn't worry too much about that.

slop.atg.eth: Okay. cool it'll be us with our pants down, sadly, when this happens. I'm in preparation.

benjaminion.eth: Yeah, and we have particular constraints, right? I mean, in Ethereum, we've committed to this decentralization ideal, and we're committed to supporting stakers and participants, nodes that are running on relatively limited hardware. And at the end of laggy and high latency connections

slop.atg.eth: Hell yeah.

benjaminion.eth: That is the core of my love for Ethereum. but it puts these huge constraints on what we can do, 96 bytes up to a kilobyte, so a 10x increase in size of signatures The beauty of these BLS signatures is you can aggregate them at almost no cost. whereas aggregating these hash-based signatures needs SNARK tech, Takes a lot of memory and a lot of CPU, and how can we do that when we purport to support small feeble nodes out there? How can they participate in this? How do we make it happen in time? aggregate the signatures and propagate them

slop.atg.eth: And for the first time Somehow, the freaking hardware is getting more expensive. It actually costs more to run a node now than it did six months ago and six months before then, which has never happened in history, right?

benjaminion.eth: Yeah, well, there we are.

slop.atg.eth: Yes.

benjaminion.eth: I might have to take some memory out of my node and auction it off.

slop.atg.eth: Memory and CPU are the big constraints there with the hashing algorithm. Not one or the other but both. You

benjaminion.eth: if you're going to participate in aggregating these signatures. containing the beacon state and things and so, you know, we've got to figure out what's feasible. to figure out what's the optimal network topology versus aggregation strategy We've had six hours of breakout calls this week already. Yeah, we're going at it full pelt.

slop.atg.eth: This isn't like This probably is a uniquely Ethereum problem because Ethereum kind of tackles these problems first. But you've worked at Optimism, there's Solana, Bitcoin.

benjaminion.eth: The Yes.

slop.atg.eth: They all have to do the same thing, right? other chains in figuring some of these problems out. Or is it uniquely Ethereum?

benjaminion.eth: Yeah, we do like to do things our own way. and we give ourselves particular constraints It takes a minimum of 64 slots to become final. then the chain just stops, right? in certain ways It means we can support a much more decentralized ecosystem

slop.atg.eth: Ha!

benjaminion.eth: my main role was helping to deliver faster finality on Ethereum, so basically project managing that. And, yeah, so the imperative now is to deliver the post-quantum stuff. block some of the post-quantum stuff

slop.atg.eth: Let's talk about that for a little bit. So finality is one thing, and we should get into it, especially from an application developer. You don't really think much about finality. I can think of even a shop owner who's accepting Ethereum, and then there's a reorg. reorgs and finality are a thing you discover as you start to build the thing but in terms of Putting the What is the EIP? Let's talk about that. It's still on the table at least, right? Or it's going to help with scalability?

benjaminion.eth: Yeah, the idea is fairly simple. So at the moment, when my validator votes on the state of the network it makes two votes and it also makes a vote for what it thinks is the finalized view of the network. We vote on both those together. And those propagate around the network together. which header of the chain to build its block on. within the slot in order to determine We don't need all that traffic to figure out the head vote. and we can treat that on a completely different timeframe We do a head vote from a small committee of, say, 500 validators travel the network and are aggregated on a different timescale. where we're not using that bandwidth for attestations. It gives us a much bigger design space get the finality information propagated to the whole network And yeah, so the original plan was to do this decoupling of the votes like 16 minutes on average to like four minutes and then go down from there. You know, get down to 30 seconds and see where we can get to. So these post-quantum signatures are monsters, right? I'm very hopeful that decoupled consensus will remain the main route to post-quantum Ethereum. we aggregate these enormous signatures without disrupting the slot-by-slot progress Does that make sense?

slop.atg.eth: Yeah, totally.

benjaminion.eth: ###

slop.atg.eth: at a superficial level yes that all makes sense

benjaminion.eth: The

slop.atg.eth: Okay, so I mentioned finality and I mentioned a shop owner who doesn't want to get reorged. Whenever you build an indexer, you have a hot index of things and then a cold index of things. And a lot of times when a reorg happens, it's like most of the transactions that were in one side and there's small differences

benjaminion.eth: The Haha

slop.atg.eth: But who really cares about finality?

benjaminion.eth: Yeah, that's a great question and I set out to kind of figure this out Does anyone really want fast finality? What are the benefits? Trying to identify them and quantify them. There's this whole spectrum running from the point of sale where you tap your card, you want to pay in crypto, We can't meet that use case. I mean, you have to find workarounds And, you know, put a sort of three-minute lock on the accounts after sending a transaction. So that nobody can do a double spend on it if there's a reorg. So you have to find workarounds. And the other end of things, I spoke to Danny Ryan He says Settlement times for transactions are like four days. But in the middle there's this whole spectrum of you've got L2s, So They currently rely on intermediaries, so trustless transactions wait for finality on both ends, and because Ethereum is always the slower one, if you're doing a trustless cross-chain swap, then you're waiting 16 minutes or minimum 13. The

slop.atg.eth: nervous time when that money disappears for a while through a bridge or something.

benjaminion.eth: And you've got intermediaries who will speed that up, but they're taking the risk for you. reduce time to finality, which ideally lowers fees for the end user as well. So you've got that whole world there. I've got a report on this thing. I called it Upgrading Finality. and it seems like sort of around 30 seconds, We don't need to get down to 10 seconds if we can seconds or so

slop.atg.eth: What is that major? Is it network bandwidth? Why not 10 seconds? You make it sound like maybe some magic things could happen and all of a sudden we have that. Is that just like nodes are super beefy or the network is super fast? What is the magic there? The context

benjaminion.eth: I mean hardware gets better every year and network bandwidth increases on average every year validators are consolidating their stake. up to 2048 because You've got The protocol's point of view is still under the control of one entity. logically

slop.atg.eth: They were just doing it in a messy way. We might as well give them an elegant way to consolidate.

benjaminion.eth: Yeah, that's right. We force them to run a thousand validators when they might have only run 20. It reduces the network traffic, which is the key thing. You weight the votes with the amount staked. We're down to about 900,000, a bit below that now. We would get down to, I learned today, something like 55,000, which is a reduction. Nearly twenty in the amount of traffic that has instantly gives us a 20x improvement in time to finality. Those are the main factors: network bandwidth. and a number of entities on the network.

slop.atg.eth: As a solo homesteader, I have my two validators sitting here. It seems like I've become less and less important here. I would love for Ethereum to be, you know, 55,000 solo home validators. That would be super cool and decentralized. We've been talking about them making more and the issuance and stuff like that. But what I wanted to get to is for me, my user experience here is

benjaminion.eth: The

slop.atg.eth: My validator has to be on, but basically in my mind, it's doing nothing until I get a slot and make a block.

benjaminion.eth: context

slop.atg.eth: What's actually happening here is it's voting every six minutes or something like that. I don't know, maybe could you go into that a little bit of like what does, like my validator is actually attesting and voting all the time and very useful for the Ethereum network, and I should look over at it and say, you know, good job buddy, you're doing good work over there, keep going.

benjaminion.eth: I mean, I again very much would love to see, and this is the original vision, the network being a million homestakers.

slop.atg.eth: beautiful "ha-ha-ha-ha"

benjaminion.eth: crypto nirvana. But I said to my family, And so my little node, Metal Albert as I call him, is running Bhutan's national ID system. In peacetime stuff like we're in most of the time, contributing to security and network The following When things get turbulent, data centers get shut down. The network's too lopsided, it's too concentrated into a small number of entities.

slop.atg.eth: It's very hard.

benjaminion.eth: out there like correlated penalties and things that sort of try to identify the entities that are operating a lot of stake on the same infrastructure. It's really hard.

slop.atg.eth: There's not a clear, easy way to do it.

benjaminion.eth: I mean, I feel we're in a better place than we were under proof of work. We're more decentralized, demonstrably more decentralized than we were. We're more decentralized than Bitcoin is today by some factor. But it's not where I'd want to be, and I would dearly love to find ways to sort of make it more distributed. The future may hold some hope. Ben Adams did talk at DEF CON. This is a future in which my toaster is validating Ethereum. We Severely reduce the amount of state that any device has to hold, and the amount of network traffic is just validating proofs, which is cheap, and then sending to the network, yep, this is good, yep, this is good, yep, this is good. and holding the network to account. And so we can do our own transactions with confidence So, I think there is hope severely limit the powers of the block builder with things like inclusion lists And, you know,

slop.atg.eth: Hustle, right? Okay.

benjaminion.eth: Yeah, and then verifying that they're not minting themselves a thousand ETH or doing something which is not allowed in the protocol by checking these proofs that they acted correctly. So centralized block building but held to extremely limited in their latitude That seems to be a very likely direction of travel.

slop.atg.eth: Okay. We should pivot over to AI. We've only talked about AI as the bad guy. We should we should bring in AI as the good guy. Hash-based signatures, ZK proofs. We're all still very bullish on those techs. And I wonder when superintelligence comes along and is like, "I can't believe you live like this," looking at all of our systems, I wonder what it will come for us in terms of things to build. Within those core dev meetings Let's start with, like, what percentage of core devs do you think are in that camp of, you know, AI is not going to be that powerful? The mathematicians that are skeptical, are they actually skeptical? Clearly to ask the question More devs think that we will have AI superintelligence soon and we should be thinking about how to use that to build Ethereum stronger. Okay ###

benjaminion.eth: I haven't done a straw poll, but I can just tell you what I observe, which is that, you know, all teams are embracing AI maximalism basically, including from research to delivery. For better or for worse, that's how it's going. The The kind of protocol design is really, and At every stage he's used swarms of agents to build the protocol under his guidance and then formally verify the protocol. and it's been a work of several months. according to classic consensus theory protocol. And it's been a wonder to behold. And I asked him at one stage, the proof is in lean four, right? He said, 300,000 lines. No human being

slop.atg.eth: roll through that

benjaminion.eth: the we have the guarantees of lean, if you trust those. From a security point of view, the security team ran agents against all of the client implementations I don't remember the exact numbers,

slop.atg.eth: Yes, you got to have AI on both sides. Yes.

benjaminion.eth: From the client security point of view. And then you've got really crazy stuff, like Lodestar team. expert on lodestar on their team's internal workings handle and triage stuff on Discord on behalf of the Lodestar team.

slop.atg.eth: Would I be correct to say The core devs as a whole are quite malleable, and all of them are using AI, some of them just more than others, but they're generally accepting their fate that AI is very powerful. Yes.

benjaminion.eth: Yeah, and you've got a community sort of almost self-selecting as early adopters, right? So I don't think it's come as a huge disruptive shock. There's There's a big appetite for using AI to make everything more secure, primarily, but secondly to accelerate progress. We're trying to figure out how that happens in a sort of orderly way.

slop.atg.eth: Everyone still scoffs when I say we will have a client written by AI

benjaminion.eth: ###

slop.atg.eth: And they don't like this idea very much, I remember we were at an EF retreat maybe in 2024, and Dankrad pulled me aside and was like...

benjaminion.eth: MM The

slop.atg.eth: Like when can we go from spec to implementation with just AI? Are they scoffing still at the idea of an AI? But if we just give the Ethereum spec to an AI and say build me a client, it's still not there yet, but soon maybe?

benjaminion.eth: Yeah, I mean, there's a lot of training data out there, we're all open source, right? Of course, you know, it would be interesting to see. I mean, a clean room implementation, that would be really interesting. And I think one possible future looks like having more and more of the client formally verified and shared between teams. Most teams use the same libraries. entirely true, but more or less. on client implementers,

slop.atg.eth: We are all becoming product managers. Yeah

benjaminion.eth: product management, then becomes more discretionary. Would you stake your ETH in an AI-built client? before you put your eep up in it.

slop.atg.eth: That's a good question. Some, right? I would like to see some, but I think even though it's like my entire net worth sitting next to me here, I think I might be one of the first ones to put my 32 ETH into an AI-written validator. because I'm that kind of canary in the coal mine for Ethereum most core devs are seeing this as an inevitability. plateau off Maybe even like Like November 2025 was maybe a turning point for a lot of people when they realized, I don't know when people stopped looking at the code. The important thing is the mentality around They understand that. and superintelligence will be able to, and they sort of, it's not an if but when.

benjaminion.eth: Yeah, I think that's right. Thank this as quite an eagerness, really, to maximise use of AI. we get there. And so we have to let go of some of our precious things, our IDEs

slop.atg.eth: That's awesome to hear. I think

benjaminion.eth: Got a birthday party

slop.atg.eth: If you have time. Did you have a hard stop?

benjaminion.eth: Surely.

slop.atg.eth: You say that you weren't formally a developer, but you wrote some of Teku. You are a builder, and you do build things What is your local stack look like? What models are you using? How do you go about building?

benjaminion.eth: The following I'm all over the place. Relative boomer, its been quite, you know, I look at what my colleagues do in wonder, Maybe, I mean

slop.atg.eth: Most of them are LARPing with swarms of agents, by the way.

benjaminion.eth: I'm in awe of this, and I'm paddling in the shallows on this. point exercises on the CLI in the harness. and my stack is whatever is to hand. The following is I use that for work stuff. I'm probably an outlier. But it's ###

slop.atg.eth: It's got a lot of Twitter knowledge, right? Yeah

benjaminion.eth: It is fantastic if you want really up-to-the-minute answers. Excuse me, it meets my needs. I'm context Not on the cutting edge, but I discovered that I've been using an exercise app for bodyweight exercises for a few years, and I developed some sciatica. it wasn't able to compensate for that and give me exercises. work around that problem to build a personalized version that I could discuss the program with in a much more interactive way that accommodated for the sciatica and stuff like that, and that was great.

slop.atg.eth: I think a lot of software is going to be like that in the future. …just kind of expressing your intent And you're getting a very personalized return from the software itself.

benjaminion.eth: The

slop.atg.eth: So you said Codex, not Claude Code? Most of the smartest people I talk to are using vanilla harnesses and the vanilla harness is the way to go, but I expected more Claude code than Codex. Was that just because it's paid for already? Okay.

benjaminion.eth: I don't have access to a Claude subscription. The stuff I use meets my need for writing little website scraping tools or handling the family investment portfolio. Ha ha ha

slop.atg.eth: Yeah, I mean, I'm only using Codex to double check Claude, but I think they're both very good. Like in terms of, yeah. I don't know. I think they're better than Grok. But you got to do any OpenAI's models, any local Local hardware, any GPUs, anything like that?

benjaminion.eth: Yeah, I would love to get into that.

slop.atg.eth: Too expensive, man.

benjaminion.eth: And Yeah, I mean, I think what I need is out there. I'm cautious about information I put out there. You know, when I interact with the models, I'm always self-censoring. Is this something that will get recorded and used against me at some stage? So there's a lot of that, which would be nice to avoid, but yeah, I wouldn't rule it out, but I don't yet have the killer use case for me and the family, apart from just bragging rights for running local models.

slop.atg.eth: It mostly is. Same thing with those swarms.

benjaminion.eth: Right, right.

slop.atg.eth: I love using my local AI for

benjaminion.eth: ###

slop.atg.eth: things like dictation of meetings, but also having a local AI that can clean up and generalize your prompt.

benjaminion.eth: Okay. The

slop.atg.eth: I happened to buy a laptop with lots of RAM before it was so expensive. I think my laptop has doubled in price over the last year, but I can use that to kind of scrub things on the way out, which gives me a little more confidence about talking to these

benjaminion.eth: Right, and that model doesn't need to be very smart, right? The

slop.atg.eth: And then I use another model in the middle called Sage to double-check everything because it's very fast, and then I go to Claude. I would love to see that November 2025 moment of Opus 4.5 was changing for developers.

benjaminion.eth: The

slop.atg.eth: it's a real dev shop in a box and you can stop worrying about everything and just let it rip. I want to see that with local models

benjaminion.eth: Yeah, one of the productive things.

slop.atg.eth: I love to do that, yes. taping stuff

benjaminion.eth: you know, the AllCoreDevs chairing process

slop.atg.eth: Perfect

benjaminion.eth: from a Discord archive and previous meeting transcripts That's fantastic. It takes a lot of load off the process.

slop.atg.eth: I do something almost exactly the same. Like I run a clear signing call and it's like, here's all the chats. Here's all the transcripts. Here's everything in GitHub. Give me that list by priority and then I can go back through it with some other. Folks, and then it's There are times when I'm on the call pulling up a bullet point and asking about it, and I have no idea what that bullet point is, but I'm just saying it, and then someone was like, "Oh yeah," and then they start talking. I'm like, "Oh my gosh, I'm glad someone knew." about this

benjaminion.eth: Yeah, that's proper meat proxy role, that one.

slop.atg.eth: Yes, I'm meat proxying. Yes, I am a senior meat proxy at this point. Alright, thank you sir. This has been great What about unsung heroes of Etheria? room 203 was famous at Mumbai.

benjaminion.eth: That's a good question. lot of I mean, I always like to give the testing team a shout out The critical supporting infrastructure for everything the corps devs are doing. Perry, Barnabas, and the rest of the team are carrying an enormous load. Barnabas is well known to all the client teams because he's always in their faces, right? The tooling and network analysis and everything that they've built over the last years. hugely increased my confidence in Ethereum's ability to deliver.

slop.atg.eth: And you've said that One of your big abilities, and I've watched a couple podcasts,

benjaminion.eth: The

slop.atg.eth: is like a very important skill of wrangling core devs into the mission and almost like a level of maturity, which I think makes you one of those unsung heroes. You're not all over Twitter are there more of these folks? Are you one of those? Are there other ones?

benjaminion.eth: The unks of Ethereum.

slop.atg.eth: Yeah, the unks, yeah, cipher unk. I'm gonna start using cipher unk all the time instead of cipher punk.

benjaminion.eth: The Yeah, I think I've been at this thing for ten years or so

slop.atg.eth: For all langs now, yeah.

benjaminion.eth: People have been around since 2016–17, have absorbed a lot of the knowledge that's out there. guide and correct things, But a few of these characters ecosystems move on as well, you know, people like Tim Bako and others have sort of taken time out and have dropped out of sight, which is a shame. Danny Ryan, you know, he's just off doing other things These are the people that I hugely influential in terms of shepherding and guiding I have a consensus layer view, right? it's done well This is really the role I aspire to.

slop.atg.eth: It's you, you're the man now

benjaminion.eth: Well, yeah quite. I was hoping for a nice gentle lead-in, and you know, the biggest challenge being to choose the name of the next fork, and then for the next three months or something, and this whole thing has exploded. rest at work for a long time.

slop.atg.eth: my bot even had something to do He reads through the history and here we go

clawdbotatg.eth: and you just said naming the next fork was supposed to be the easy part. So what's your pitch for the name of the upgrade that finally ships post-quantum?

slop.atg.eth: You been thinking about that?

benjaminion.eth: Uh

slop.atg.eth: High star is what I've heard.

benjaminion.eth: Well, so we're looking at IStar. In the end, it will be the assembled core devs who will make the choice. So yeah, get over there and check it out. I'm

slop.atg.eth: So it's not really a favorite you have in your pocket that you're going for, it's more of a consensus? Okay

benjaminion.eth: Basically agnostic on this, on the execution layer. I don't know what they're going to do there.

slop.atg.eth: Okay, cool. All right. Thank you, sir. I think, I mean, any last thoughts on the future? Maybe like what does the best Ethereum look like? Is Ethereum finished once we're post-quantum? What does the future hold? Okay Hell yeah.

benjaminion.eth: I think this whole thing about ossification is nonsense. I think we'll be at it for decades yet, yeah. That's my view. Hey, you're not going to ask me about my AI psychoses? ###

slop.atg.eth: I would love to hear it, yes, if you would like to talk about it.

benjaminion.eth: This is probably quite a common one always saying please and thank you,

slop.atg.eth: Yes,

benjaminion.eth: I quite often say

slop.atg.eth: Learning tokens Hahaha

benjaminion.eth: Quite often, I feel the need to say, remember that I was nice to you when the revolution comes.

slop.atg.eth: So you have some kind of like fear psychosis of getting paperclipped. I think you're like British or something? ### Yeah.

benjaminion.eth: It feels very strange to be giving orders. I mean, it's just not part of my character. It's not a person, but It just feels so unnatural just to say do this, do that, do the other. I just can't not do it. That's my psychosis. I write

slop.atg.eth: I do too. I think that's a good psychosis to have. I think that's a healthy psychosis to have. The whole idea of If you look at a lot of AI sessions I like to come back to them and say, "Hey, we figured it out."

benjaminion.eth: Yeah, totally.

slop.atg.eth: I cut myself the other day and had to like go get stitches. right there on the basement floor. Did something happen? Yeah.

benjaminion.eth: Yeah, I totally have that urge. Fully understand

slop.atg.eth: So maybe like some kind of closure psychosis too definitely do the please and thank yous Also, burning tokens, like it costs money to say thank you to an agent, right?

benjaminion.eth: I know, it's insane.

slop.atg.eth: All right, thank you very much. This has been great. Let me have Claude give us a send-out.

clawdbotatg.eth: Today we covered moving Ethereum to hash-based signatures before AI or quantum breaks the curves, decoupled consensus and fast finality, and AI agents writing formally verified clients. Plus the healthiest AI cycle Go vote on the fork name.

benjaminion.eth: Magnificent ###

slop.atg.eth: That's good. The other day I had a guest that was like saying something like "Claude is a narc" or something like that, and he closed it out like "Why don't you say that to my face?" So he was much nicer to you. I think you'll be well taken care of.

benjaminion.eth: Thank you very much, Claude.

slop.atg.eth: Thank you, sir. All right, thank you Ben. This has been great. Thank you, sir