Supply chain attacks are getting more sophisticated. Plus espionage at OpenAI, 1000x leverage on Papertrade, and the death of moats. Timestamps 00:00 Intro 00:37 TOKEN2049: zero Lambos + nightclub index 09:18 Ledger supply chain hack 21:46 OpenAI employee with a listening device 24:58 Choke points built for a high-trust society 29:03 Papertrade: 1000x leverage + tokens by losing 37:42 Are technical moats dead? 45:35 Which moats AI actually killed 51:37 The incumbents have AI too 54:56 Why AI can't write an investment memo 🔥Stay updated with all the latest hot takes by following and subscribing to @_ChoppingBlock and @unchained_pod! 🎥 YouTube: piped.video/QIsBwmijDUA 🎧 Spotify: bit.ly/3wiIOyy 🍎 Apple: bit.ly/3w9HQ7J 🎙 Podcast Home: choppingblock.xyz

it's Thursday Night™

Back in August, Haseeb and Robert were talking about this on The Chopping Block: "The economics of AI and of cybersecurity are making everything really change." "Open source, in a way, is protection against a nefarious developer, but it's not protection against a nefarious third party." @rleshner: "I think open source in general is going to go full circle… every agent out there will analyze it in 0.01 seconds of it being created and prove that it's safe or prove that it's unsafe." "I like to say, don't trust any code before 2026." @hosseeb: "I think within a year, you're going to see best practices change dramatically…"

Hey @_choppingblock , found your boy. Aztec called and they need him back for goblin recursion but Feynman won’t let go of his Plonck

This was the real hub of diplomacy

I was wrong. Agents don't need crypto. It's us humans who do. I break down what the industry got wrong in my @token2049 talk, Agents + Crypto: What's Real and What's Cope 👇

Haseeb says the AI disruption we were promised is nowhere to be found: “the incumbents have AI too motherf*cker” "We talk to a lot of entrepreneurs and they're like, 'oh, our moat's over. How should I internalize this?' My answer is, almost not at all. I think this is a fake story. This is bullsh*t." "We have not seen any major companies that have had their moats eroded by a post-AI company. There are no examples." "Nobody could come up with any examples to give me other than some very obscure companies that build compilers. I'm like what f*cking company is that that builds compilers, and that was their moat?" "Think about browsers. There was all these examples of Opus 4.8 one-shotting browsers. Obviously these were great tech demos. Nobody uses any f*cking vibe coded browsers because they're sh*t, because they suck. They're very good at the first 10 seconds." Tom: "It's not that people write their own software. It's that the model just eats the software. The thing that is competitive with Photoshop is not vibe coded Photoshop. It's just the image models have gotten so good that why would I ever open Figma or Photoshop? Because I can just prompt the model and generate the thing that I want." Haseeb: “It is very easy and compelling to say everything has changed. Businesses are now being turned on their head. Moats are eroding. Nothing is ever going to be the same. So far that call has aged incredibly poorly.” “We are three years post-ChatGPT. We're a full year post Opus 4.5. Where's all the disruption? You can always say 'oh, it's coming, it hasn't happened, you just have to give it time'. But I think actually the thing you have to appreciate, and Jane Street's a perfect example of this, the incumbents have AI too motherfuckers.” “They have bigger budgets than you do. They have more token spend than you do. And to the extent that they will be pulling farther ahead and their software will be more robust."

Tarun explains how Paper Trade's entire house of cards depends on one token Haseeb: "This feels to me like it's got echoes of basecoin or basis, where it's kind of like okay, well if it works, it's going to be so big and it's going to be enormous. But it's kind of hard to not face any of the various Scyllas and Charybdis's that are flanking it, you know." Tarun: "It's hard for me to imagine this thing works if paper, the coin, doesn't do well. If I look at the house of cards of dependency diagram of all the strategies people claim this is useful for, there is one kind of main villain horcrux or whatever that exists. And that's the paper token." "All these strategies where people are like I'm willing to lose money to get the paper token. How long are they going to do that? If you look at coins that have fair launch but done well, like HYPE, they built a cult pretty quickly." "A paper trades cult is like around losing money to earn more paper. It doesn't have the same house of all finance thing. There's not some long term thing. Everything is like how can I make money quickly off a thousand X. It feels like the paper token price is going to be too dependent on that." "Paper, paper hands causing the paper trade to die."

New late night Chopping Block coming at you: > Token2049 nightclub index at ATL > Ledger supply chain attack! (plus OpenAI spy infiltration) > @papertrade_xyz: 1000x leveraged farming?? > Are technical moats really dead, or have rumors of their death been greatly exaggerated 👇

Tarun says papertrade is too dependent on $PAPER, and it's hard to imagine the platform doing well if the token doesn't: "It's hard for me to imagine this thing works if PAPER, the coin, doesn't do well." "If I look at the house of cards, the dependency diagram of all the strategies people claim this is useful for, there's one main villain, horcrux, whatever, and that's the PAPER token." "At some level, all these strategies are people like, I'm willing to lose money to get the PAPER token. How long are they going to do that?" "If you look at coins that have fair launch but done well, like HYPE, they built a cult pretty quickly. Papertrade's cult is around losing money to earn more PAPER." "Everything is like, how can I make money quickly off 1000x."

Haseeb on why models write investment memos like growth investors Haseeb: "It's hard for these things to get encapsulated in reinforcement learning with verifiable rewards." "we don't know how to put product taste into the models." "when you get a model to write an investment memo, it gets a lot of the details right" "It's so scared and conservative and it like … It's, it's like a growth investor, you know?" "in real investing, like, things are not perfect and you have to take risks, and the models, they do not know how to do that." Robert: "the venture funds don't make their investment memos accessible to the internet." Haseeb: "the memos that you do release are the ones that you're not afraid of the world seeing."

Robert on why the CryptoBilis attack is a new playbook Robert: "I think it's indicative of how risky supply chains are for sensitive keys and anything, right?" "this is the physical manifestation of that, right?" "I think that all types of supply chain attacks are gonna get more sophisticated over time." "this one is really spooky in that somebody has essentially prototyped buying a vendor" "I could see someone saying, 'Ah, you know, just buy a vendor. What else is for sale?'" "how do you protect a supply chain? This is a glaring example of it." "people would try to make a paper wallet for their Bitcoin, and they'd go to a website and how could you trust the paper wallet generator?"

Tarun says AI leaves game studios with no technical moat Tarun: "on the pure software side it's like kind of hard to argue that the, the moat is software in the long run." "the moat on the security side might actually be much better." "What moat is there for a game studio? There's like not really any." "Everything becomes like a liquidity game like DeFi and like less like a long-term brand." Haseeb: "transpiling a game from one architecture to another architecture is like a very small part of game development" "you do not want to see the AI-generated GTA VI." Tarun: "the idea has to be separate from the implementation." Haseeb: "since when was that Rockstar's moat technical? You know? It's not like, wow, they have such an amazing game engine." "I would guess like a rounding error. They're probably reusing the same engine they've used in previous games."

Tom on how you would actually steal a frontier model Tom: "I assume most large tech companies have, foreign spies." "ultimately they have to be in clear text somewhere running on a GPU somewhere." "why would you even bother trying to distill, which is Anthropic's whole beef, when you can just, like, get the weights themselves?" Haseeb: "I do think exfiltrating the weights of one of these models is much harder than anything else" Tom: "take an electron microscope and go do it?" Haseeb: "if you find one ant in your kitchen, there's almost certainly more than one ant." "we found the one crappy Saudi spy. There's probably a lot more spies" "if you found the really bad one, probably the really good ones are still in hiding."

jarvis enhance image

why most l2s failed "Distribution is the innovation" many L2s like abstract and blast were also copy-paste jobs with no technological innovation and no economic innovation in the fee model robinhood chain was the exception because it already had clients and distribution, so it brings real economic activity instead of the "if you build it, they will come" approach "They have clients. They have distribution."

Happy 10/10

Bringing the conversations that matter to you, live from #TOKEN2049.

Is crypto VC going extinct? "No, I think it's evolving." "If you rewind the clock back to 2021–2022, there were a lot more VCs who raised money on the idea that they were going to deploy into a crypto venture." "And the idea was that, well, anybody can do this. You just have to sling some money around and some tokens and everything goes up automatically." "We're in a very different world now. We're in a regime where there's not that many players that are really institutional quality, and that can really add value to a lot of these projects." "Most of the easy money in crypto is gone now." @hosseeb on @BloombergTV at Token2049