Noah Jelich | Breaking Narratives: How Code Review Beats Marketing Decks | Common S3nse - 2026

[music] Today I'm going to be presenting a few stories about marketing and I hope in a fun way. So um you know the the movie fugazi fugazi it's a wazi it's a wuzzi it's fairy dust and it doesn't exist. These are the famous words of Mark Hannah in regards to the stock markets in regards to investing in stock markets. Um I fundamentally do not agree with this. I I think it's not all fairy dust.

I think there is some alpha to be found and I think there is money to be made by investing according to certain principles. uh and I'm hoping to give you an insight into how to think to find out how uh TLDDR I did a bunch of things currently at EPM but this presentation was written while at cyber capital uh which is the oldest crypto full exposure hedge fund in Europe um and I was reading source like my life depended on it as I was trying to fully analyze potential investments for risk. Uh outside work, I have a toddler and I really love trains, bikes, and hikes. So, let's get straight into it. Today, you will hear the stories of three projects and I will let you make your own conclusions of how these may end.

Before that, a disclaimer. Uh, Cyber is an activist hedge fund and they affect projects they invest in, hopefully guiding them to better outcomes. This presentation is based on the state of these projects in late 2025 and may not reflect current reality. And of course, none of this constitutes investment advice. With the disclaimers out of the way, let's get to it.

So, we have Bitcoin scaling uh and potentially predictions for Ethereum. We have D0ero which is an infra project for Solana let's say and we have the famous hyperlquid hype thing whatever it's going on there. So let's start with a familiar story. So, Bitcoin uh not founded by this guy uh is is the first decentralized cryptocurrency based on free market ideology uh invented in 2008 when an unknown person published a white paper under a pseudonym of Satoshi Nakamoto and the use as a cryptocurrency began in 2009 with the release of the open source implementation. Uh by 2012 2013, Bitcoin was reaching the limits of its throughput.

One block every 10 minutes, one megaby blocks. The block size debate is often framed as a technical disagreement over the single parameter, but that framing is quite incomplete. Uh cyber cyber capital was investing in Bitcoin at the time uh but needed to make a decisions. Um what was visible at that time was a clash between two economic models. On one hand you have the constrained block space which leads to high fees and layered scaling and on the other there is the expanded block space aiming for base layer utility.

But uh more important than the actual economic model is what this exposed about bitcoin governance. How it works in practice. how code can be controlled, how repositories and narrative are malleable. And if you're paying attention to GitHub activity, to mailing lists, and to PRs, uh you could see where this was going. Um I'm having a horrible feeling that something didn't load.

Ah, we have one mystery image for some reason. Um Oh, because I ran out of internet. That's horrible. should not be happening. We're having a bit of a connectivity issue.

So, by the midpoint of the debate, a majority of key contributors to Bitcoin Core were employed by a company called Blockstream. Um, this matters because Blockstream was a for-profit company whose business model centered around second layer solutions like side chains. Those solutions are only valuable if the base layer of Bitcoin is constrained. you get a very clear incentive structure. You know, small blocks mean congestion, congestion means high fees, high fees means demand for L2s and L2s mean monetizable infrastructure.

And this was observable at the time. Uh these employment relationships were public, repo control was visible and the road map direction was changing. So from left to right missing one, we have Greg, Matt, Horge, Mark, Patrick, Warren, Adam, and Peter. of those the first and the last had direct commit access to the Bitcoin core repo and uh and they took advantage of it. Um so as Bitcoin was working on not scaling BTC uh just to clarify Blockstream employed the people in this image right uh and as Blockstream was working on not scaling who were core Bitcoin developers and meanwhile there was another movement the Bitcoin XT client which was trying to increase the block size and it was faced with general oppression.

So uh there were large-scale DDoS attacks according to reports targeting nodes even infra providers. Uh companies like Coinbase were targeted socially and operationally simply for running alternative clients and discussion channels such as Reddit and Bitcoin talk were increasingly moderated along ideological lines and even technical decisions became narrative weapons like uh anti-DDoS protection is now censorship. Um this all culminated in the eventual introduction of segwit segregated witness which is an efficiency increase in terms of block backing. Um while in the original proposal sold as a scaling initiative, it was actually a major blocker for blockstream to execute some of their products uh that provide alternative scale paths to the network. uh it was pushed with a promise attached of additional block size increases but that bump never happened because the PR submitted and discussed stood in review held up by the previously mentioned Blockstream employees.

Eventually the outcome was more or less locked in. Uh fees rose dramatically exceeding $50 per transaction and this was not treated as a failure. Uh this was a validation of the fee market. Right? Meanwhile, support for Bitcoin by businesses like Steam uh was dropped.

Uh users migrated to alternatives. Lightning was blooming and the chain split creating Bitcoin Cash. Um today I fear the same happening with Ethereum if it refuses to scale meaning meaningfully as a base layer. uh and for cyber at the time this meant stopping Bitcoin investment picking higher conviction better scaling fundamentals based bets. So that's the first story.

The second story this is my personal research um presents itself as a fundamental upgrade to blockchain networking. Um, it sounds great when somebody tells it to you, you know, reduce latency across validator networks uh by using private backbone. For those unfamiliar, private backbone, imagine literally private optic p cables, private submarine cables that you're renting out and similar to provide you the best latency there is. Uh and this would improve global distribution because normally you have all the validators somewhat centralized based on the existing low latency paths between data centers and all that for a tiny little fee of pi 5% on validators. Um on top of that, this quite amazing company already had market fit um at least seem to have market fit because they had doubledigit uh percentage adoption by Solana validators despite taking a 5% fee.

Um and it seemed like a great investment. So the question was is it a great investment? Um and the answer may surprise you. So the first step for me was a basic infra and market sanity check. Uh for reference for a competitor so to say I took Megaport.

Megaport is an existing global backbone provider with very low latency routing over 11,000 access point at the time of uh of uh research uh and a huge number of interconnections. Megaport operates at a much larger scale than double zero uh about 100 times more access points which means like 10,000 more connections right uh 10,000 times more connections and despite this double zero was valued at this huge premium compared to the incumbent right uh so we had 1.5 billion versus uh well at the time prior to when we were looking at it as an investment and um spoiler alert skipped it, it was similarly valued. Um so the question was why is it so valuable other than being blockchain something something and uh to answer that I took a look at I we also talked to the team but uh I took a look at the actual routes that they provide. Um, zero quite often claims that they provide an amazing uh boost on the Europe to Latin America route because they're they use one of the only um submarine cables which is not yet part of the public internet and as such they gained the edge of uh like 50% nearly from 20124 millisecond latency which is very very significant when you're doing anything high frequency.

This is thanks to Ella link which is a relatively new transatlantic cable connecting Europe and South America funded by the EU. Thank you very much. Uh so so to check this uh whether this is a very valid claim that they provide such a massive latency advantage, I uh I pretended to want to run a validator. And uh from my super cool investment fund email, I I asked Latitude and Vulture, which are two prominent hosts in Sao Paulo, Brazil, uh for a little bit of information on their provided infrastructure to provide me some trace routes to known servers in Europe uh and also their backbone provider spec sheets. Um what I found was on one hand um these providers already had access to lolink and on the other hand it was being used uh which the trace routes did prove.

Uh so talking to zero more about this they said that sure but they are renting a single dedicated line and these let's say public access things from latitude or whatever providers uh may get congested during let's say uh crucial time um price change events uh where you may not be able to get the performance which they're saying they would guarantee Uh however this statement is generally weakened by the more recent consolidation of Megaport and Latitude uh where Megaport acquired Latitude which is a major Solana host which means that Latitude would likely have access to the Megaport private backbone uh in the back end which means these premium routes become a bit more standardized or offerable by latitude. Uh the other thing I did was a little bit of social engineering. Um I reached out to many a Solana validator and I asked very simple thing. Are you using zero and why or why not? And I got the weirdest answer.

Um hey Noah, just got your email about double zero. Would love to chat about it. Do cyber capital run a validator in not this point but researching it. Yeah. So basically they take plus 5% they bribe you they bribe you that is why people are using double zero.

Nobody's using it because of the technological advantage because of the amazing performance. The answer was they sort of bribe you with the percentage of the stake and that's why they had the amazing market fit because this was at the time not documented. It has since been documented of course because they've been somewhat called out on it. Um because they were more or less providing a stake boost to validators who would sign in and thus they were not really making a profit other than whatever they were earning from their own stake. Um and their delegation fund is huge.

So what does this mean for0 as an investment? There are still some things where0ero has a potential edge such as a multiccast within uh within virtual networks where you can get a massive latency improvement by not having to route packets and by having a very application specific implementation. But in practice they have to subsidize adoption and capture market share which requires a massive amount of capital because nobody's really uh signing up for this willingly. While on the other hand they uh they have the threat of Megaport just replicating this overnight because they have 10,000 times more connections uh they could use and leverage. So this was a very fun research uh a company we decided not to invest in and thanks to that we missed on the out on the amazing minus 90% gains we could have had um and finally there's hyperl hyperlid is very complicated um but so I cannot make such a clear yes or no but there were a few interesting bits and pieces of hyperlquids functioning which I did find out which are generally not very well described nor documented in any marketing materials.

Uh so that's why I would like to just present these so we have an idea of what to look for. Um so Hyperlid is marketed as a next generation trading network. It's extremely fast. It's decentralized. It's fully onchain.

It even says it in a funky way like den. I cannot spell anymore. Um, and uh, do you think that's true? Do do you think it has, I don't know, globally distributed validators? Do you think it has permissionless participation?

And do you think that execution on Hyperlid is transparent? Because you may you may be a little bit surprised. Uh so researching hyperlquid the first thing was go to the reper right I I am the guy that reads the source code and finds the dirt in the source what source code there is no source code here what uh so instead of a readable implementation what I found was uh binary uh minimal documentation how to run a node some tips and tricks um without being able to review the code you cannot verify this consensus logic the execution logic nor the data flows. Uh so this was a massive problem. Um meanwhile there have been some reverse engineering attempts on hyperlquid nodes but come on this is not how life should be.

Uh the second thing I noticed was the node topology. What I mean by that is okay um you see this thing the second thing for lowest latency run the node in Tokyo Japan and uh to avoid jailing it is recommend to achieve 200 millisecond two-way latency to at least onethird of validators by stake and also notice validator may connect up to two non-validator peers by specifying their IPs as sentry IPs. So I I think it's pretty clear what we have as a network structure. Um we have a circle of 28 nodes that are all connected to each other. These are the validators.

They're each connected to at most two relayers. So there's at most 60 relays and nobody can externally access this network structure. Uh not very easily, right? They would need to be a validator. Um and of course the whole Tokyo Japan thing was very very interesting.

Um so to find out I went to Japan. Um and uh there I managed to make some connections because if you are at all familiar with Hyperlid you know the team is a bit cryptic. Uh and the answers to why why it's in Japan is it's very interesting. Uh Jeff Jeff apparently lives in Japan or his wife is from Japan. So that's that's the reason why Hyperlid is in Japan.

The other reason may be the the low latency between the data centers in Japan. But I I needed the real data. So uh in Japan I met some people I can give a shout out to to Au from Maria uh who are a validator on uh on Hyperlquid who were so generous to provide me an actual data export from their node which takes 64 GB of RAM to run which was a lot cheaper back then as well as to uh Fractton Ventures and Hyper Quant for connecting me to the various validators and giving me some tips on how to extract this data and actually uh analyze it since again everything is closed source right and uh what I found was predictable so the validators the stake and recently produced blocks were highly centralized in just a couple asns as are uh let's say servers server farms uh by names uh mostly in the Amazon data server and you want to guess where these They're all in the same data center cluster about 20 mters across located in Japan. Why? Because the consensus rules effectively require it.

And um the one guy running a node in London gets jailed like every day. I think he's just a bit of a cuck at this point. Um and yeah. Uh so this means hyperlquid is not in a strict sense decentralized. It's something you can drop a single bomb on which is not generally a thing you can do to Bitcoin.

Uh so you have that. Uh the other thing was permissionless participation. This has since been fixed. Now you can permissionlessly validate on Hyperlid though there's still like six validator slots that are locked to Hyperlid Foundation or something like that. So um yeah but a bit better and the final piece is execution.

So the key detail here is nons behavior. Uh on Ethereum nonses are strictly sequential. You cannot skip them. Uh like if you submit six and seven you cannot have six not execute and seven happen. Right?

uh on hyperlquids nonsenses are only strictly increasing and transactions can be invalidated in some very special ways. Uh what does this mean in practice? Um you can have you can submit about 50 to 100 nons at once within like two milliseconds and then only 30 of them get executed uh because they only need to be increasing. Uh in practice this and of course in the documentation they recommend to cancel uh transactions by submitting a new transaction which is a no op not by running a cancellation of an order. Now this may seem like just an okay optimization but let's go back to the network structure.

[snorts] Where is the mempool? Who has access to mempool? He had 60 re layers, 30 validators and the mele happens somewhere between these re layers and the validators. So those are the parties that know the state of the mele. No one else.

You cannot really read it because it's all very closed, right? Um and the way nonsense work is such that you don't create an order then that order sits there then somebody fills that order. The way it works is that you create an order and you can just sort of delete it by it not existing which means that this order sits in the mele before getting executed. And in short this means that there is a data symmetry within hyperlquid because the validators and the relayers are the only ones that really know what the orders are. They are the only ones with the true state of the order book.

nobody outside them because at any point there are thousands of orders that are sitting in the mele waiting to either be executed, matched in the mele or cancelled by a noop settling on chain and then being executed and that is non-transparent execution uh invisible. So yeah, in conclusion, uh I hope this was interesting and educational. Uh and uh if you would like to learn more, feel free to ping me on Telegram or LinkedIn. Thank you very much.

Oh yeah, I I shaved my hair. any questions? I forgot that exist. Uh I think his wife is there. I I think that's the thing.

Uh it's either that but also when you look at the data center clusters, Japan has a very good one in terms of in the latency between data centers like when you look at the graphs of that the Japan one uh uh the I I think it's Southeast Asia East or something like that. Uh it has like only seven or eight milliseconds between the clusters while the other ones go up to like 2030. So it has the best local switching gear. Not to just, you know, blame his wife or whatever. Any other questions?

Go on. Yeah. Yeah. Yeah. That's why I said it's a bit more of a complex story.

This is the dirt I found. But this again, you can have a very good thesis, but still the market behaves a different way. This is not a very strong thesis. It has usability. That's the thing.

it is just not doing it at a decentralized way. But half of the L1's or L2s are not very decentralized anyway. You know, uh it more so opens your eyes maybe to how much or how little decentralization is valued. No. Any other questions?

Thank you very much.