加密新熱土與科技未來：Chopping Block Live 激辯

0:00Chopping Block 開場與主持群介紹

0:00歡迎來到 The Chopping Block。每隔幾週，我們五個人就會聚在一起，從產業內部的角度聊聊當下加密圈最熱門的話題。先來快速介紹一下大家。首先是 Tom，DeFi 專家兼迷因大師。大家好！接下來是 Tarun，Gauntlet 的超級大腦和總顧問。Yo。然後是 Arthur Hayes，Maelstrom 的 alpha 仲裁者，TOKEN2049 的傳統嘉賓。Howdy。當然還有 Balaji，比特幣大腦。最後是我，Dragonfly 的 Haseeb，早期加密投資人。不過要提醒大家，這裡說的都不是投資建議、法律建議，甚至不是人生建議。更多揭露請看 choppingblock.xyz。Welcome to The Chopping Block. Every couple weeks, the now five of us get together and give the industry insider's perspective on the crypto topics of the day. So quick intros. First, we got Tom, the DeFi maven and master of memes. Hello, everyone. Next we've got Tarun, the Giga Brain, and Grant Puba at Gauntlet. Yo. And joining us again, a Token 2049 tradition, we've got Arthur Hayes, the arbiter of alpha at Maelstrom. Howdy. And then, of course, we've got Balaji, the Bitcoin Brainiac. And I'm Haseeb, the head hype man at Dragonfly, where early-stage investors in crypto, but I want to caveat that nothing we say here is investment advice, legal advice, or even life advice. Please see choppingblock.xyz for more disclosures.

0:46TOKEN2049 現場氛圍與產業觀察

0:46我們現在人在新加坡的 TOKEN2049。我覺得這已經是全球最大的加密會議了。這裡還能感受到那種熱情的能量。美國的會議現在都變得很機構化，西裝筆挺，這裡則是唯一還能讓人熱血沸騰的零售級大型盛會。你們覺得這次 TOKEN2049 怎麼樣？So we are here at Token 2049 in Singapore. It is, I think at this point, the biggest crypto conference in the world. It's one of the only places where I think you still feel this kind of energy. And it's sort of like the U.S. conferences have gotten very institutional, a lot of suits. This is like the one retail mega conference that still gets the juices flowing. How have you guys felt about Token 2049?

1:09你們這群傢伙來我城市，這週交通超級塞。不過不只我們，這週還有五個會議同時舉辦，像 Milken、F1 也在這週。還有空氣品質超差，像德里一樣，聽說是印尼那邊燒林。其實這裡也感受到機構化的轉變，今年第一次沒有看到包膜的 Lambo，大家都變成熟了，雖然有點小失落。錢還是不夠多，大家都喊窮。不過還有一天，夢想還有機會實現。Lambo 可能明天才會出現。I mean, you motherfuckers come to my city and traffic is insane for, like, you know, a whole week. But it's not just us. It's like, it's everything happening at the same time. You know, there's like five conferences happening. There's like Milken. There's obviously F1 happening this weekend. So it's not just us. Also, the air here is insanely bad for some reason. It's like Delhi. It's bad. Yeah. The worst. Yeah, that's what I've heard. Fires in Indonesia. I feel like the institutional shift—you also feel it here. This is like the first year where there's been no, like, wrapped Lambo out front, and it's like just everyone is growing up a little bit, which makes me a little bit sad. I kind of was looking forward to more kind of jubilation. There's still not enough money. Everyone's broke. There's one more day. You can make your dream happen still. Yeah, that's true. That's true. The Lambos are tomorrow. Yeah.

1:53Balaji：東西方機構與加密產業未來

1:53我有個觀察：加密產業其實在模仿西方機構，但東方機構反而更有韌性、更能存活。西方金融機構反而會被淘汰。只要是手上持有大量 10 年、20 年、30 年美債的，肯定會出事，Arthur 你應該有很多想法吧？You know, I have an observation, by the way, which is I think crypto is sort of emanating Western institutions, but Eastern institutions are more functional, so they will survive. Western institutions, financial institutions won't. Anything that's holding a lot of 10-year, 20-year, 30-year treasuries, I'm sure Arthur has some opinions on that.

2:10他們都完蛋了。They're wrecked.

2:11Network State 遷移：新加坡到哈薩克

2:11說到東南亞的火災，Network State 最近也把總部從東南亞撤出，搬到哈薩克。其實我們已經搬過去了。不過我們在新加坡還是有辦公室，也會在東南亞開新據點，敬請期待。為什麼選哈薩克？哈薩克會成為下一個 El Salvador、UAE、新加坡。Well, I wanted to start with this story. So speaking of fires in Southeast Asia, the network state has recently pulled the plug on its headquarters in Southeast Asia and is now moving to Kazakhstan. Well, so yeah, so we actually already moved to Kazakhstan. Tell us the story of how this all played out. So actually, we actually still have, you know, an office in Singapore, and we're actually going to open up something else in Southeast Asia, so stay tuned on that. But basically, Kazakhstan, just to talk about Kazakhstan, Kazakhstan is going to be the next El Salvador, the next UAE, the next Singapore.

2:44為什麼哈薩克之前沒被注意到？因為有部 Borat 電影，讓大家對哈薩克有錯誤印象。電影只是誇張的刻板印象，現實完全不是那樣。哈薩克其實是「內陸亞洲」，不像大家熟悉的「沿海亞洲」。美國人對亞洲地理的認識，通常是美國打過仗的地方，像越南、韓國、日本，這些國家都在沿海。哈薩克則是超級內陸，靠近俄羅斯和中國，講俄語，不太用英文發聲，所以美國人沒什麼理由去那裡，加上 Borat 電影讓大家以為那裡很糟。And the reason is, so first, why was it off the radar, right? Well, there's a Barat movie, which really misled everybody about what Kazakhstan is. It's not, you know, you see a movie and you think, okay, it's like a caricature of reality, but it's like in the ballpark, you know, but like Jurassic Ballpark, you know, you ballpark approximate it like the DNA in Jurassic Park, but it's actually totally not. And really what Kazakhstan is, is it's land Asia. And you're all familiar with like coastal Asia because you know, the way Americans or Westerners learn geography is America fights wars there. So, like Vietnam, Korea, Japan, Americans know those countries exist, right? And they're on the coast, right? China also, you know, is a part of the Korean War, right? Singapore, recently Crazy Rich Asians, there's a movie, they know Singapore exists, but they're all on the coast. And Kazakhstan is so far inland, and it's adjacent to Russia and China. It speaks Russian, so it doesn't post in English, right? So for all these reasons, you know, Americans don't really have a reason to go there. Plus Barat made it seem like it's not like a great place.

3:43所以哈薩克像是被琥珀封存的寶地。我會說它像新加坡，但有超大空間可以建設。它有像 UAE、新加坡一樣的高治理品質，政府裡的人都很聰明。物理奧林匹亞還跟中國、印度並列第一，這國家很強。哈薩克有 2100 萬人口，跟 14 億人口國家並列，超猛。國土面積是新加坡的 3000 倍，但人口只多三倍。如果你想要 1 萬英畝來蓋 Arthur Hayes City，哈薩克可以辦到，他們也很歡迎你。當地人很想打造創業城市、加密城市。And so because of that, it's like this encased in amber, absolutely amazing country, which is, I would call it Singapore, but with space to build. So it has all of the—quality of governance of a UAE or Singapore were really high IQ people and government. Like they were number one in the physics Olympiads, tying with China and India, which is pretty legit as a country, right? Yeah. Yeah, with 21 million people. They're tying countries are 70x, with 21 million people in Kazakhstan. Yeah, because it is an enormous landmass. Exactly. So they're 3,000 times the size of Singapore, but only 3x the size population-wise. So Kazakhstan is the place that if you want 10,000 acres for Arthur Hayes City, we can do that. Yeah, we should. Right? Every corner. Yeah, you can get it, and they'll want you there, okay? And the people will love it. They want to build startup cities. They want to build crypto cities.

4:32這週五 Network State Conference 會有哈薩克代表團來，很多高層部長都會出席。哈薩克現在在國際舞台上很積極。Binance 已經要在那開辦公室，Solana 去年有設特區，Shervin Pishevar、Olaf 也都在那設點。因為哈薩克有我們熟悉的數位基礎設施，還有很棒的加密法規。想蓋資料中心，這裡有空間。德州現在反而在限制，哈薩克則是亞洲能源和建設成本最低的地方，還有全球 40% 鈾礦。這裡數位和實體條件都很強。新加坡雖然很棒，但沒什麼土地。有些國家有地但政府不行。哈薩克很罕見，兩者兼具。In fact, they've got—if you come to Network State Conference on Friday, you're going to see a delegation from Kazakhstan. Many senior ministers have come down. This is Kazakhstan's sort of big thing on the world stage. Binance is now opening an office there. Solana, you know, last year had a special economic zone. Shervin Pishevar just is setting up there. Olaf has come up there. There's going to be a lot of things happening there because it has all the digital stuff that we know. It's got actually a really, really good crypto law. It's, if you want to build data centers, they've got the space. Texas is now banning them, you know, or stopping them. Kazakhstan has the lowest energy and building costs in Asia. Because it's got, like, enormous amounts of—it's like 40% of the world's uranium. So it's got all the digital and all the physical. See, Singapore is amazing, amazing place, but it doesn't have any land, right? Or there's countries that have land, but they don't have great governments. It's very unusual to have both.

5:23Jurisdiction Shopping：選國家像選企業客戶

5:23你離開東南亞時，我記得你在考慮很多不同司法管轄區。你應該是少數真的在壓力下做這種「選國家」決策的人，這過程怎麼樣？其實有超過十個國家搶著要 Network State，我們未來會在全球開很多節點。我們有能力把矽谷模式複製到新地方，尤其現在美國簽證卡死，人才流動會分散到世界各地，就像加密本身一樣去中心化。我們這個社群現場就有 50 本護照。加密身分不等於傳統身分，這是我們下一步。先脫離法幣，再脫離舊身分，拿到第二本護照比買第一間房還重要。So, so, so when you left Southeast Asia, I remember you were sort of looking at many different jurisdictions to go to next, and I think you might be one of the only people in the world who has done this kind of jurisdiction shopping under duress in the way that things played out? What was that like? So, great question. So the thing is, first of all, you know, ten different countries bid for Network School, okay? More than ten, in fact. And we're going to open up nodes all around the world. We've clearly got something where we can print out Silicon Valley in a new place, especially with all the visas being cut off to the U.S. There's this flow of talent that wants to go somewhere, and now it will go everywhere, right? And it'll decentralize just like crypto itself. Just like our community here is whatever, you know, 50 passports in this room. That's something, just like cryptocurrency doesn't map to legacy currencies, crypto identity doesn't map to legacy identities, right? That's the next big step for us, you know. Get out of your currency, then get out of your legacy identity. Get a new passport. You know, second passport is much more important than the first home.

6:28你在選國家時，像不像企業談判？誰來找你？有什麼籌碼？他們開什麼條件？其實這很像企業級銷售，從零售到中小企業、再到大企業、再到小國家、大國家。小國家其實很像企業。這裡有個祕密：大多數國家其實都很小。你把 193 個聯合國會員國人口做個直方圖，會發現一半國家人口不到一千萬，20% 不到一百萬。只有很少數像俄羅斯、中國、印度、美國這種人口過億的。大多數主權國家都很小，像開曼群島才 7 萬人，英屬維京群島 3 萬，比 Palo Alto 還小。他們最大的資產就是國旗。I think I'm more curious, what does, like, horse trading look like when you're shopping for a new jurisdiction? Like, who's coming after you? What are the levers you can pull? Good question. What do they offer? Like, what do the offers look like? Yeah, yeah, yeah. So, so... Without kissing and telling, so to speak, okay? It's actually a lot like— I'm just saying, I'll give— do a little better. They'll like it. Yeah, it's, it's, it's like, it's a lot like enterprise sales, you know? It's like basically as you ascend from retail to SMB to—like enterprise to small governments to large governments. Small governments are actually a lot like enterprises. See, the thing is, this is one of the secrets, ready? Most countries are small countries. Okay, most people have not actually made this graph. Take the 193 UN members. Let's use that as an operational definition of what a country is, okay? There's exceptions and so on, but let's use that. And you make a histogram of their populations, and you'll find that half of them have less than 10 million people, and like 20% have less than 1 million people, right? Only a small fraction of countries by number are the 100 million-plus countries like Russia, China, India, America. Most sovereigns are small, like Cayman Islands. You know the population of Cayman Islands? Who knows the population? 70,000. Yeah, 70,000, yeah, exactly. But it's BVI. Fees like 30K, okay? That's smaller than Palo Alto. So the number one asset that they have is a flag, okay?

7:51我們有超大國際加密網路和社群，但沒實體土地。他們是小國家，有土地和合法性，但沒什麼資本或人脈。這其實很互補。十億美元在十萬或百萬人口的國家很有影響力，換算分別是每人一萬或一千美元。你比想像中有更多談判籌碼。重點是你真的要親自去現場，因為好萊塢的印象通常是錯的，但你不知道正確答案。你必須親自走訪，才能校正認知。And so the complementarity between what we do, we've got these gigantic crypto networks, social networks. We're gigantic international networks, but we have no physicality. They are small states. They have physicality, they have legitimacy, but they don't have anywhere near the capital. Or connections, in many sense of the term, that we have. That's a really great deal to be done, right? A billion dollars goes a long way in a country with 100,000 people or a million people. That's like $1,000 a head, $10,000 a head, right? So you have a lot more leverage than you think. And the key is to—now, here's the tricky part. You actually have to go to these places, and the reason you have to go to them is you have to just assume that Hollywood—Hollywood is wrong, but you don't know what is right. You can't ballpark. Do you see what I'm saying? Yeah, yeah.

8:44好萊塢偏見與國際現實

8:44好萊塢從不拍非西方國家繁榮的電影，除了《瘋狂亞洲富豪》，像《魷魚遊戲》還是韓國人互相廝殺。因為美國人看到別人過得好會覺得不舒服。就算你貼中國城市的炫目影片，大家也只會說那只是裝了 LED，完全不懂。事實是深圳 40 年前根本不存在現在的樣子，他們不是只裝燈，是整個城市都新建的。沒有親自旅行，你根本無法分辨哪些國家真的好、哪些只是表面好。你會遇到四種情況：你以為不好、真的不好；你以為不好、其實很好；你以為好、其實不好；你以為好、真的好。你在評估國家時，這四種都遇過吧？And here's why Hollywood is actually deeply wrong. Essentially, Hollywood doesn't show—and I want to give the ball to you guys, I don't want to monopolize—but Hollywood does not show movies of prosperous non-Western countries. With the exception of crazy rich Asians, that's like, you know, even Squid Game is showing all the South Koreans killing each other or whatever, right? And these are, you know, and the reason is it's depressing for Americans to see other people doing well. So even if you post a video of, like, Chinese glowing cities, they'll be like, Oh my God, they just put LEDs on their cities, right? They don't understand. You'll see this, you know, complete cope on social media. And it neglects the fact that Shenzhen basically didn't exist in its current form 40 years ago. It's like, no, they put LEDs on the cities, they built the cities, right? So the point is that without travel, you can't calibrate as to some countries really are bad, and you have the whole two by two, right? You think it's bad, it is bad. You think it's bad, it is good. You think it's good, it is bad. You think it's good, it is good. Literally the whole two by two is there. And you saw all of those when you were evaluating computers. Yes, that's right.

9:46所以你真的要親自跑一圈，實地考察。哪些國家是你以為好、真的好？哪些是你以為不好、其實很好？其實西方國家就是大家以為好、但其實不怎樣的例子。雖然他們也有來邀請 Network State，但我說的是，他們其實已經是「下降世界」了。所謂第一世界、第三世界這些詞都過時了，我現在只看「上升世界」和「下降世界」。第一世界就像繼承家產卻自己蓋不起來的富二代。So, and that just means you just have to put a lot of miles, and you actually have to see it on the ground. So who are the folks who were—you think it's good and is good, and who are the folks you thought it was bad and they were good? Well, you know, who's bad? I'll say who's—who do people think is good and is bad is Western countries. They were Western countries also. But they actually bid on your network. Did they want— No, no, no. What I mean is, like, they are—they're in a bizarre... You know what it is? The terms first world and third world are obsolete. I just think about it in terms of—descending world and ascending world, right? So the quote first world is like the fail son who's, like, inherited something that they couldn't build, right?

10:27小國家、大影響力與加密社群槓桿

10:27你們一直想套我說出國家名單，但我不會那麼容易講出來。其實只要人口少於一千萬的國家，像這樣的國家有一百個，十億美元在這些地方影響力超大。你可以用 Twitter 追蹤數來校正，像 Arthur 你的追蹤數，跟這些國家總統、立委比一比，AI 一天就能跑出來。你去那個國家，可能就是他們今年最大新聞。甚至在美國，科技圈大咖的影響力比國會議員還大。如果你真的支持某個政治人物，幫他轉推，他會把你當大咖。你只要真的住在那裡，參與社群，你的槓桿比想像中大很多。And they're like, Balaji, we're really trying to pry a list of countries out of you, and you are not. Yeah, yeah. Come on, man. Who did you sing KTV with? How much baijiu did you drink? Zero KTV, absolutely. I'm happy to. But it's not that much like enterprise sales. Yeah, yeah. What I'll say is countries under 10 million people, which is like 100 countries, really, like, you know, a billion dollars goes a long way. Long way, right? And so, I mean, one of the things—here's another way to calibrate it, okay? Take your Twitter following, you know, Arthur, your Twitter following. Make a histogram, and just, you know, AI can do this for you in a day. Take any country, take its legislators, its president, and so on, and compare how many followers they have to how many followers you have, right? Your visit to that jurisdiction will be the number one news event of their year. Okay. I realize even for congressmen and senators in America, tech guys actually have much more distribution than you think. If you just took a liking to a certain politician and just wanted to retweet everything they did, like, I mean, obviously you have to, like, it's something where you actually believe in it, right? But you would be a very important constituent for them. You know, you should live in the jurisdiction and so on and so forth, right? So. Our, you know, like, basically we radically underestimate our leverage, is the short version.

11:47上升世界、後共產經濟與科技崇拜

11:47所以 Balaji 給我們一個實驗：去看誰在轉推他，就能猜出他在談哪些國家。不過他沒那麼容易讓你猜到。明年這時候我們每人都要「買」一個政治人物？不是啦，重點是你是在對國家做資本投資。所謂上升世界的國家，很多是後共產國家，或像拉美是後毒品經濟，阿根廷則是後通膨。美國最近很愛玩「後經濟」迷因，這些國家則是「後反經濟」——他們對共產、凱因斯主義有抗體，不用再從頭解釋什麼叫努力才有成果。這些國家的人都準備好要做生意。最棒的是，你去這些國家，會看到 Elon 的壁畫，科技大咖的壁畫。他們不是要打壓你，反而會幫你立雕像。Arthur，只要你真的投入，他們真的會幫你立雕像。你要搬去哈薩克嗎？我已經幾個月前就設點了。So wait, Balaji has given us an experiment. We have to go look at who's retweeting him to figure out the list of countries. Exactly. He won't say the countries, but the data is— I didn't make it that easy for you. This time next year, we're each buying a politician. No, no, no, it's not— very clear. You're not buying a politician. Okay, I got you. Thing is, you're actually making a capital investment in the country, right? So that's the thing is, the ascending world countries are ascending world because they actually have a— many of them are post-communist, or they're like in Latin America, they're post, I don't know what you'd call it, post-drugs. You know, right? Like what Kelly is saying, right? Post-inflation in the case of Argentina, right? It's like everyone in the U.S. keeps making post-economic memes, given some of the recent... Yeah, yeah. And now we're talking about the post-drug economy. Well, actually, I don't know if we're at the post-drug. Okay, it's actually interesting. It's like post-economic normally means, you know, okay, you're wealthy enough, you don't have to worry about things, right? These guys are like sort of post-anti-economic. Whatever communism was, whatever Keynesianism is, is like anti-economics. So the great thing about them is they have immunity, they have antibodies to all the communist Keynesian stuff. I love it so much. I don't have to explain it from scratch. We don't have to debate about whether, you know, effort leads to results or whatever. So that whole part of the world, all those people. They're ready to do business. And here's the thing, you know the best part? You go to these countries, and there's multiple of them. They have murals of Elon, okay? They have murals of all the tech guys. They're not, like, trying to shoot at them or— you know, there's a saying, a prophet is without honor in his own home, okay? They love tech, okay? Arthur, they'll build you a statue, okay? I'm serious. They really will. If you find a country and you basically adopt it, right? I mean, you have to actually commit. That's the thing. So are you moving to Kazakhstan? Huh? Yeah, I've already set up there. We set up there months ago.

13:40科技富豪出走潮與全球遷徙

13:40你在哈薩克有你的雕像嗎？我從來沒要求過要有雕像啦。你會不會想要騎馬的雕像？對啊，Arthur 的雕像一定很威風，像希臘神一樣。這只是個比喻啦。重點是，有些地方會歡迎你，不像舊金山現在有人對 Sam Altman 家丟汽油彈。某些國家會根據你是誰來決定要不要歡迎你。Do you have—is there a statue of you in Kazakhstan? Yeah. I never asked for a statue of me. I'm kind of—how about you on a horse? Statues. Yeah, he's got to be on a horse. Arthur will have a great statue. It'll be like a Greek god kind of statue. So that's why I recommend, yeah, like holding a... Yeah, something like that. That's right. But, but, you know, I just mean there's a figure of speech. The point being, they will welcome you as opposed to throw Molotov cocktails at Sam Altman's house, which they're doing in San Francisco, right? So, like, and to be very clear about this, some countries will welcome you depending on who you are.

14:11現在有很多藍州（美國民主黨州）覺得億萬富翁不該存在，但也有很多小國家很想要這些科技億萬富翁。結果就是，藍州的億萬富翁會消失，因為他們會通過各種稅收法案。這也是為什麼 Larry Page、Sergey Brin、Zuck、Larry Ellison、Elon、Thiel、Bezos、Ken Griffin 都離開了藍州。這些州會失去稅基、失去億萬富翁、失去上億資產的人。那他們去哪？其實世界比美國大多了。Yeah, look, there's a bunch of blue states that think billionaires should not exist. Okay, and there's a bunch of small states that want those tech billionaires there. And so they're both going to get what they want, kind of, which is blue billionaires will not exist. Okay, you will not be able to be a billionaire in a blue state because they'll pass these taxes and so on and so on. That's why Larry Page is out. That's why Sergey Brin is out. That's why Zuck is out. That's why Larry Ellison is out. That's why Elon is out. Thiel is out. Bezos is out, right? Ken Griffin is out. All the amount of blue states, right? So those all just go to zero and lose their tax base and lose their billionaires and they lose their centimillionaires and so on. And where do they go? Okay, they can go within America, but actually the world is a lot bigger than that.

14:52哈薩克？還有像 Thiel 這種人會去烏拉圭、巴拉圭。拉丁美洲現在很有趣。你有遇到拉美的買家嗎？這地方很大啦。就說有可能吧，不要再問了，換話題。Kazakhstan. No, no, he also mentioned Teal, so Uruguay, right? Paraguay and Uruguay. Well, yeah, of course. Okay, so Latin America, Latin America, very interesting. Did you have Latin American bidders? What's that? Were there any bidders who were from Latin America? What can I say about that? It's a big place. Come on. Let's say, let's say it's potentially. Potentially. Yeah, come on. Okay, okay, we're not getting into that. You know he likes to do. Let's move on. Let's move on. Okay, okay.

15:23OpenAI 數學突破與加密危機

15:23說到歸零，昨天 OpenAI 公布了一個超大消息。他們一次釋出了大約 700 篇 AI 寫的數學論文，產生了新的數學成果。有些已經用 Lean 形式化，有些還只是 AI 亂湊的東西。裡面有幾個超級重大的結果，其中一個被認為是近百年最重要的成果之一，叫做 quasi-Riemann hypothesis。還有關於矩陣乘法、快速傅立葉轉換等，特別是在漸進式改進上有很奇怪的突破。這讓很多人都嚇壞了。So speaking of going to zero, one of the big announcements yesterday was this giant drop by OpenAI. OpenAI released something like 700 AI-written manuscripts producing new mathematical results. That they, some of which they had formalized in Lean, some of which they just kind of had this, I don't know, AI slop that they published. There was a couple of really massive results in this drop, one of which is considered to be one of the most significant results in the last hundred years, which is known as the quasi-Riemann hypothesis. There was some stuff about matrix multiplication. There was some stuff about fast Fourier transforms, particularly weird results, I think, in improvement in asymptotics. It's freaked a lot of people out.

16:04大家會怕，是因為數學家可能要被淘汰了，或是他們的工作型態會徹底改變。而我們加密圈也不能置身事外，因為區塊鏈本來就靠數學。昨天 Ethereum Foundation 的密碼學家 Justin Drake 就發文說，現在是 crypto 進入『bunker mode』的時候了。And part of the reason why it's freaked a lot of people out is that, of course, there's this idea that, okay, mathematicians are maybe now either obsolete or they're very quickly facing a complete change in what their job is going to look like. And we're not immune to what's happening in mathematics because, of course, crypto is dependent on mathematics. And so yesterday, Justin Drake, one of the cryptographers at the Ethereum Foundation, he posted that it is time for crypto to go into bunker mode.

16:32什麼是 bunker mode？他的意思是，數學現在進入超速發展階段，原本要五年十年才會有的進展，現在幾天就出現了。我們以前用的密碼學演算法會慢慢換，像從 SHA-2 換到 Ethereum 的 SHA-3，20 年前還用 MD5。這些演算法會隨時間演進，但以前的速度很慢，現在數學的進展快到我們跟不上。Now, what is bunker mode? Bunker mode, in his parlance, is like, look, mathematics is now going into overdrive. Things that took, you know, five, ten years are now happening in days. And we don't know what's going to change. Normally, a lot of the cryptography that we use in blockchains, they change over time, right? We used to use SHA-2, now we use SHA-3 in Ethereum. You know, back in the day, we used MD5 20 years ago, and we move across cryptographic algorithms over time. But the pace at which we move is much slower. And the pace at which this math is now happening.

17:06這次 OpenAI 釋出 700 個新成果，結果跟密碼學有關的竟然是零。這很奇怪，因為密碼學應該是數學裡經濟價值最高的領域，卻一個都沒公開。這很可能代表他們其實有發現什麼，但沒公佈，可能在做負責任揭露，或覺得太危險需要另想辦法。We had 700 drops in this OpenAI drop that—sorry, 700 results that came in this OpenAI drop. There were zero related to cryptography. That's really weird. It's really weird because probably the most economically significant part of math is cryptography, and yet there was not a single result in cryptography, which probably implies that there were some results. But they didn't publish them because they may be doing some responsible disclosure, or they may just be saying, look, this is so potentially damaging that we need to do some totally different paradigm of how we deal with this.

17:39加密社群的應對與風險思考

17:39所以 Justin Drake 建議大家現在要 freeze，什麼都不要動，把所有有價值的資產都放到從沒簽過名的地址，因為我們不知道會發生什麼事。區塊鏈底層很多數學問題，像離散對數問題，是公開金鑰密碼學的基礎，但我們其實沒證明這些對傳統電腦真的很難。雖然很多聰明人試過都沒破解，但數學界最近很多原本認為很難的東西也被攻破了。以前 crypto 的進展速度，未來可能不適用。So Justin Drake's recommendation is that, look, freeze. Do not move. Take all of your assets that are of any size and put them into addresses that have never signed any transaction because we don't know what's coming. And if—a lot of the mathematics that undergirds blockchains, like the discrete log problem, which is the foundation of public key cryptography, if it turns out that, you know, we don't have any proofs that this stuff is hard for classical computers. We're pretty sure it is, because a lot of smart people have tried and they haven't been able to prove anything. But that's true of a lot of stuff that's now falling in mathematics. And so the kind of pace that we used for crypto in the past may not be the right pace to think about what's happening going forward.

18:22他的建議是，現在可能進入一個關鍵時期，大家都該有點警覺。你們怎麼看 Justin Drake 這個警告？其實很多人，包括 Vitalik，都覺得這有點太誇張。首先，這些新突破大多不是實用的演算法，很多只是理論上的存在證明，跟真的能攻擊還有很大差距。And so his recommendation is this: we're now maybe going through a critical period, and everybody should be a little bit scared. How do you guys react to this warning from Justin Drake? I think there's, like—Two things that are a little bit—it's also, by the way, a lot of people had responses to this, including Vitalik and others that were like, This might be a little extreme. I think the first thing to note is that most of the advances that people found are not, like, practical algorithms. They're generally, like—Okay, there's something that you learn in intro CS class that, like, you can't multiply two numbers faster than a certain amount of time. That's clearly wrong, but it also assumes that you have, like, infinite memory. So I think there's a sense in which the practical algorithms that break this stuff versus, like, the theory of, like, oh, there exists an attack are actually quite—there's still quite a big gap.

19:11但如果真的有實用的攻擊方法，像 OpenAI 這種有無限算力的機構才有可能做到，一般攻擊者沒那麼多資源。你有沒有看過他們這些突破到底花多少算力？他們說平均三小時，但加起來也才 2,100 小時，沒那麼誇張。重點是，就算明天 ECDSA 被破解，問題是你要花多少時間才能偽造一個簽名？就算有攻擊方法，也還是有 proof of work 的門檻。假設要五天才能破解，那才真的要擔心。但我們怎麼會知道？誰會來公告說 ECDSA 壞了，大家快跑？But that's that, like practical when there's hundreds of dollars at stake. But OpenAI has infinite compute resources to put at these things, whereas if you're an attacker, you don't. We don't know how much—I haven't gone through this—but, like, how much compute did it cost for them to do each one of these discoveries? Have they published that? I mean, they said three hours. They said three hours on average. On average, yeah. But that's a... So, but you can now add up the 2,100 hours, right? So it's not, like, that crazy. I think the main—or $21,000. So the main thing I think is worth thinking about is, like, okay, even if ECDSA breaks tomorrow, the question is, like, can I forge a signature in X amount of time? Like, there is some proof of work even with I know an attack, right? So let's say it takes five days. Yeah, yeah. I think then you have to be a lot more worried. But we're not going to know. How are we going to know? Who's going to tell us, Hey, guess what, everybody? ECDSA is broken. What the fuck do you do now?

20:04其實如果 ECDSA 被破解，傳統金融系統會先爆炸，因為他們用的密碼學更爛。至少 crypto 是公開的，大家會看到攻擊發生，有點像礦坑裡的金絲雀。但 bunker mode 這種建議也怪怪的，什麼叫轉到沒簽過名的地址？如果連簽名都能被反推，那有沒有簽過根本沒差，連已知密文都沒用。真正該怕的是 AI 讓攻擊變得實用化，而不是只是理論上存在攻擊，但計算量大到 10 的 80 次方那種。I would say also the rest of the financial system is broken first. To be— yes, to be sure. Like, it's not just crypto. We have much shittier cryptography in most of TradFi, right? And so, like... At least cryptos in the open, you'll see an attack. There has to be some public sort of canary in the coal mine. But I think, like, the bunker mode thing is also a little bit weird, because it's like, okay, go to an address you never signed from. Well, if they already broke the signature, they found a reverse, you know, one-time thing to supposedly one-way function, you're fucked anyway. Like, who cares whether it's never signed before? Like even the known ciphertext is not going to help you attack. So I think the whole thing is like, yes, there is something to be afraid of. But the thing to be afraid of is more that someone figures out how to make these things practical quickly with AI versus like, okay, they found the theoretical attack, but it's like 10 to the 80th, the constants in front of everything are 10 to the 80th.

21:00人性、數學與區塊鏈社會契約

21:00我們昨天也聊到，人類很愛追求數學的優雅答案，而不是最有效率的答案。這也是現在出現最大哲學分歧的地方。我有幾個想法。首先，這種 AI 一次丟一堆結果的做法，其實很像 Ramanujan 筆記本那種數學期刊，大家要去驗證、理解這些結果。第二點，我覺得 crypto 也許是第一個被打擊的，但也是最有可能存活的。因為全球只有中國和 crypto 真的把網路當一回事，一個有防火牆，一個有區塊鏈，都是網路上的防禦結構。We were talking about that yesterday, right? It's like— I have several. Go ahead. You're going to finish. We were talking about this yesterday, right? Like, there is this thing about humans in that they've over-indexed on elegance in the answer, right? They want to make math have an elegant answer versus just like, what's the most efficient answer. And I think that's the thing that we probably have. That's the biggest philosophical difference, it seems to be coming up. So I have several thoughts on all of this. The first is— This kind of drop of results is similar. You know, there's actually a journal called Ramanujan's Notebooks, where people go through Ramanujan's old results and try to prove them and so on. And this is kind of similar to that, where you just get these results dropped, and then you have to try to figure out, you know, how to make sense of them, how you relate to things, number one. Number two is, I think crypto, in the fullness of time, Maybe it's the first hit, but it's also the one that survives. And I've been thinking about this for years, which is, you know, there's only really two large groups in the world that take the internet seriously, which are China and crypto, because of the Great Firewall and the blockchain, respectively. Those are defensive structures on the internet.

22:08像美國五角大廈、OPM 這些 Web2 系統一直被駭，只有 Google、Facebook 這種有自己 token 的公司還能自保。Web3 至少有 bug bounty，駭客攻擊會反映在幣價上，但系統不會直接死掉，因為還有社群共識。這就帶到下一點，也許我們需要一個新的區塊鏈社會契約，例如大家用 Ethereum 時要在鏈上同意，如果某個 hash function 或簽名演算法被攻破，礦工可以把狀態倒回一週前，所有交易都回溯，大家事先同意這種 on-chain 條款，這樣就不會有爭議。Proof-of-stake 鏈可以做長距離回溯，其他鏈也可以有不同做法，這是區塊鏈的彈性。The Pentagon, by contrast, gets hacked over and over and over again. OPM gets hacked. All Web2 systems. With maybe the exception of Google and Facebook owners that have the tokens to insulate them, are all going to get hacked. And Web3 systems, by contrast, have at least a live bug bounty, where if they get hacked partially or what have you, then their price drops. But they don't totally die, and the reason is they actually also have a social component. So that gets the next point. One thing I think we might have to do is a new social smart contract among blockchains. Which is basically something as follows. It's like, you use Ethereum, and by using Ethereum, you may have to assent to something on-chain that says if there's a hack of this, that, or the other hash function or, you know, digital signature or what have you, then the miners can rewind state to, like, something a week ago or something like that. And all transactions are reversed, and everybody has consented to that as part of the terms of service on chain. So there's no controversy as to whether or not you could rewind the chain, okay? Proof-of-stake chains can do that, right? They can do long-range unwinding, okay? And then other chains may decide not to do that, but that's a great thing about blockchains. You can have many different types of things like this, right?

23:25但現在鏈之間的依賴太多了，像跨鏈橋一堆，真的要回溯很難。比特幣這種狀態單純的鏈還有可能，但現在這種回溯會造成巨大損失。而且我們根本還沒準備好新的架構。雖然損失可以量化，像金融危機一樣，總有人會賠錢，但這條鏈可能會有 coin holder 投票決定要不要回溯。不過如果駭客已經拿到很多幣，投票也會被操控。你可以設計各種機制，像只用 30 天前的快照等等。Crypto 最擅長這種危機處理，傳統系統完全沒這能力。I mean, that was more possible. Let's say five, ten years ago, right? Now there's so many cross-chain dependencies and so many other systems. Like if you're rewinding all these bridges— It depends which chain it is. Sure, sure. If it's like a Bitcoin style, like there's not a lot of state, there's not a lot of complexity, then yeah, that's doable. But like I think the reality is that any such unwind would be so damaging. And the other thing too is, like, we're really not ready for a new architecture. Well, it's damaging, but, like, it's always quantifiable how much money is lost, right? It's like some amount, you know, a crash in price. Like, okay, whatever, people lose whatever million dollars, even a billion dollars. But then the thing, I mean, that's like, I don't know, like the financial crisis of 2008, right? The people will lose money when financial things happen. It's not going to be zero loss. It's just going to turn out that that chain—and people might have, there might be a coin holder vote, which legitimates whether that chain rewinding happens, right? But because maybe it's only 100 million. Post-hack, you can't do a coin holder vote anymore because the coins themselves are... So that's what you have to contemplate, that. Is it like the DAO attacker, where they've got 30% of the coins and then they can corrupt it, you know, or maybe it's only the coins... From 30 days ago, Convo and so on and so forth, and your old, you know, keys work. There's many ways you can do this as you start thinking through, like, the verticality of this. And crypto is, if there's one thing we're good at, it's this kind of thing, right? This is the thing we are good at, right? And legacy systems are not good at this, right?

24:54反科技浪潮與未來社會矛盾

24:54我認為接下來會有越來越多人走向『反科技』，這種情緒就像 Ted Kaczynski（包裹炸彈客）現象。反科技已經變成新一代的反資本主義，像病毒一樣進化，讓右派也能接受，不再只是共產主義。大家會說，科技搞出一堆亂七八糟的東西，不如回歸原始生活。這種浪漫情懷跟工業革命時期很像，因為社會還沒進入穩定狀態，很多人就會反對變革，這也是共產主義的一部分。未來會有很多人只想要實體硬幣、離線生活、當現代穴居人。但現實是，現代供應鏈太數位化了，真的很難完全脫離數位世界。Now, the other thing I think is going to happen is this is going to accelerate the—Ted Kaczynski-ification of a large segment of the population, where there's a growing anti-tech movement. I mean, I think anti-tech is the new anti-capitalism, and it's morphed just enough evolutionarily. It's like a virus that's adapted just enough that it appeals enough to the right that it doesn't see it as exactly communist, but it's kind of a morph on communism. Where it says, look, there's all these crazy things that are happening with the emergence of all these technologies. Let's just throw it all away and go back to living in the woods like men were supposed to, you know? And that's like a, right? That's a, that's a, there's a, it's very similar to the same romantic sentiment that accompanied the Industrial Revolution for many of the same reasons, because people were in the middle of the disruption, everything was changing. And they weren't yet in a stable state, and so a big part of communism was the rejection of all this change, and there was a rightist aspect of that as well. Anyway, point is, a lot of people are going to be like, I just want physical coins. I just want to be offline everything. I just want to be a caveman again. Why does my digital... And the problem is that, yeah, you can potentially do that. It's pretty hard to do that in our modern world. So much of the supply chain is going to be digital.

26:08所以你會看到很多人對這些科技憤怒，這也是我為什麼會在這裡。東方社會比較有社會凝聚力，能處理這種變局。西方像美國有三億支槍，社會張力很大。Tom，你怎麼看這些事？So you're going to have this rage at— that's why, like, that's why I'm out here, right? Like, the East will deal with it better because it's got a better social fabric. In the West, you know, America's got 300 million guns. Anyway, Tom, what's your take on this?

26:21行動與信號：AI 安全與加密社群反應

26:21我覺得遇到這種事或 AI 安全問題，重點是看行動而不是嘴巴。上 Bloomberg 講話很簡單，真正有決心的人會像那個 Anthropic 的安全研究員一樣，退休去鄉下寫詩。這才是真正的行動。對這種 crypto bunker mode，我會更在意看到大戶真的把資產移到新協議、真的上鏈操作。如果看到交易所、協議、大戶真的這樣做，那才是真的信號。Yeah, I think when it comes to anything like this or, you know, AI safety, I think looking at actions is, like, usually a lot more instructive than looking at words. It's very cheap to go on Bloomberg and, you know, get some eyeballs. It's another thing to, you know, I think of the anthropic safety researcher that retired about a year ago and said he's going to go live in the countryside in England and, like, write poetry. And I respect that, and to my mind, that's kind of more showing action. And so... For this, you know, I think I take more worry when I see sort of large whales actually enacting this protocol and sort of moving on-chain. And so I think if we see exchanges, protocols, large holders actually doing this, then you're actually, like, putting action behind your words. And then I think there's a lot more belief to it.

27:03但很多這種危機都會落入 Goldilocks 危機（不冷不熱剛剛好）的陷阱。像 Q Day（量子電腦攻擊日）一樣，大家以為會慢慢有預警，結果可能一下子從『需要量子電腦』變成『五天內用商用硬體就能破解 ECDSA』，再過三個月搞不好你用 MacBook 就能做。這才是真的可怕。But I think a lot of these things kind of fall victim to this Goldilocks idea. I mean, same with these, like, really elegant proofs of, well, kind of like with Q Day, it's so far out, we're going to kind of very gradual sort of heads up when, in reality, we've gone from you need a quantum computer to maybe break ECDSA, which is to it's, you know, five days with a, you know, commercial-grade hardware to, you know, maybe three months from now, you can do it on your MacBook. And I think there's something scary I saw a couple hours ago, Matt Green, who's a famous—

27:31加密末日論與 AI、量子威脅

27:31有位來自 Johns Hopkins 的密碼學家在推特上說，他覺得公開金鑰密碼學可能已經走到盡頭了。當然，說說很簡單，但也許他以後真的都會把東西用明文公開吧。其實他是在那個討論串裡跟另一位密碼學家吵架。社群媒體總是把事情放大到最聳動的程度。我覺得比較有趣的是，Vitalik 還特別寫了一篇長文，說其實不要太聽 Justin Drake 的話。A cryptographer from Johns Hopkins tweeted that he thinks maybe public-private key cryptography is over. Of course, talk is cheap, but hey, maybe he's going to put everything in clear text from now on. To be fair, he was fighting with another cryptographer in that thread about this. Social media amplifies things to be as shocking as possible. I think it's more funny that Vitalik had to write this long post, basically saying, actually, don't listen to Justin Drake.

28:02我覺得這有點像加密圈的末日論。那些說這種話的人，其實就是在聽 Justin Drake 的觀點。其實我覺得他寫這篇文章很好，提出這個觀點也很好，大家可以討論、反駁這個很重要的議題。我對 Justin 沒有意見，我覺得他很聰明。加密末日論這種想法很常見。其實連 Justin Drake 也不是說這一定會發生，他只是說，這種事情的機率從 0.1% 提高到 5%。而 5% 的機率已經高到你應該要買保險，因為這種結果可能會讓整個加密產業歸零。Yeah, I think this is kind of the crypto equivalent to doomerism, which is that, you know, the people who are saying this, listen to Justin Drake. That's very good. Like, the reality is that—I think it's good for him to write it. It's very good to float it. It's very good to argue and counter-argue this very important point. I had nothing against Justin on this, but I thought it's clever. Crypto doomerism is very— Yeah, and look, the reality is that even Justin Drake is not claiming that this will happen, but he's saying that, look, the probability went from like 0.1% to like 5%. And 5% is enough that, like, you should probably get insurance against a 5% outcome that results in wiping out all of crypto.

28:38AI、密碼學與數學極限

28:38我想講一件事，這幾年前我跟 Yudkowsky 爭論過，其實 AI 能做到的事情是有硬性上限的。你可以證明像亂流或混沌現象，根本無法預測太遠，因為計算上有極限，精度有限。某些密碼學系統也是這樣，有數學上的『盒子』可以把 AI 關在裡面，它們突破不了，甚至有些還有物理極限。很多人把 AI 當成無所不能，什麼都能破解，但除非我們的數學定理推導有錯，否則其實不是這樣。One thing I do want to talk about is there are, and I was arguing this with Yudkowsky years ago, there are hard bounds on what AIs can do. You can prove that turbulent or chaotic phenomena, you can't forecast them out. There are computational limits on how many time steps out you can forecast them given finite precision arithmetic. And then, depending on where certain theorems are, there are cryptographic systems that you also can't forecast out. So there are mathematical boxes that you can put these AIs in that they can't break out of, mathematical and in some cases physical limits. A lot of people treat them as being completely omnipotent; they can crack anything. But unless there's an error in our derivation of a theorem, which is possible, right?

29:28但重點是，公開金鑰密碼學其實沒有真正的『困難性證明』。我們沒有證明說它一定安全。有人會說，假設 P 不等於 NP，或假設 Green-Mont 假說成立，那就有證明。當然可以這樣假設，但其實還是有很多不確定性。像之前有一堆中國的後量子簽章演算法被提出來，結果有研究者用 AI 一次性把全部破解了。所以就算不談量子電腦，光是後量子演算法，很多可能明年、後年就被破解。這也代表我們現在考慮的後量子轉換，演算法的強度證據其實很弱，除非是純粹 hash-based 的演算法。But the point is, for public key cryptography, we have no hardness proofs. Like, we don't actually have any proof that— Well, we have. Assume P is not equal to NP. Assume Green-Mont hypothesis, then you have a proof. Sure, sure. But I think also, I mean, there was a bunch of these Chinese post-quantum signature algorithms that were candidates for post-quantum algorithms, and some researcher basically disproved every single one in a single AI-fueled data drop. So, like, the reality is that even, okay, you know, we're not even talking about post-quantum, right? Because quantum is a lot of engineering problems, not just scientific problems. But even going to post-quantum, it's like, okay, well, a lot of these algorithms might just get busted over the next year or two, in which case that also speaks to waiting longer for a post-quantum transition if— The algorithms that we're considering right now, we have very weak evidence of how robust they are, unless they're purely hash-based algorithms.

30:25AI 與物理世界、量子金錢新想像

30:25未來很多東西會有一種『防駭』的物理元件。AI 在數位世界很強，但在實體世界還沒那麼強，它必須真的去操控機器人或無人機才行。所以很多事情會變成線下儀式化的活動，AI 反而讓線下變得更有價值。Proof of human、實體見面這些會變得更重要。我想補充一點，大部分加密圈的人聽到量子電腦，都會覺得 ECDSA、整數分解這些會被破解。但其實有一種用量子態做的金錢機制，長得很像區塊鏈，因為量子態不能被複製，所以也有防雙花的特性。這些理論 1970 年代就有了，最近十年才真的能用。所以大家總說量子電腦會毀掉加密，但其實也可能讓我們有更難複製的物理金錢，甚至能用量子態隨時產生私鑰，這比現有密碼學還安全。所以如果量子電腦真的來了，也可能成為新一代金錢系統，大家可以樂觀一點。There's going to be something where there's like a hack-proof physical component to many things. Like, you know, basically that, you know, what AI is like. Very strong in the digital world, but not yet really strong offline. It has to actually animate a robot, animate a drone to go and do something, right? So, like, a lot of things are going to be ritualized offline kinds of things. Like AI makes the offline more valuable. Proof of human, you know, like in-person things become more valuable. Something that, you know, I think a lot. But I would add one thing, which is— You know, most people in crypto, they hear, okay, there's a quantum computer. It's going to break ECDSA. It's going to break integer factorization, whatever. But there are known money schemes that look like blockchains that use quantum states where you basically can't clone them, and you get a very blockchain-like no double spend properties. In fact, these were, like, kind of theorized in the 1970s, and only in the last, like, 10 years have people, like— Make them kind of usable. And so there are—everyone's always like, quantum computer, crypto is over. But there's also, like, quantum computer, now I can have these much harder-to-copy physical states that you can generate secret keys on demand from. And that stuff is actually much more secure than kind of our existing cryptography system. So I think you should be a little more optimistic that, like, if we have the quantum computer to break stuff. There's also the quantum computer to be the new money system.

31:53AI 節點與鏈上攻防新模式

31:53你剛剛讓我想到，未來很多區塊鏈節點可能會直接內建 AI 模型。為什麼？因為不只是拿算力去挖礦，還會用 token 預算讓 AI 不斷嘗試駭進自己鏈，提前找出漏洞。只要防守方的 token 預算比攻擊方多 10 倍，防守方又更在意，這樣就可以用加密貨幣買 AI token，讓 AI 幫忙防禦。你會看到未來 Bitcoin node、Ethereum node 都會內建 AI 模型。這其實跟你們 Flop 在做的事情有點像。So something actually you just kind of prompted thinking. A lot of blockchains may start to actually bundle an AI model with the node. Why? Because they'll throw some of the compute not just at mining, but at tokens to keep trying to hack the chain itself, right? Because you have a token budget for defense to try to find a compromise before the attacker does, and in general, it's not 100%. Obviously, you need to have better prompting and so on. But if the token budget for the defender is greater than the token budget of the attacker by 10x, and the defender cares more, right, which it would, you take some of that crypto, you spend it on AI tokens, you spend some of the compute on AI tokens, you start bundling AI models with your nodes, right, with your Bitcoin node, with your Ethereum node, and so on. So this actually rhymes with what you're working on with Flop.

32:39我們 Flop 比較像是 summit layer，不是直接綁節點，但這不是我們要解決的問題。我們想做的是：Flop Network 就像是 AI agent 的食物，AI 的支付網路。我們的理論是，應該有一種貨幣能直接兌換算力，這才是你會想持有它的原因，就像你持有美元是因為可以買食物。也許這種貨幣會自我強化，用算力來強化自己。你可以想像區塊鏈直接綁定矩陣運算。當然，形式驗證只能做到某個程度，如果有新數學出現還是沒辦法。但不管怎樣，未來可能會有開放式模型註冊機制，就像高頻交易一樣，只要有新模型上架，區塊鏈就會用 token 預算去檢查每一行程式碼，不斷自我駭入和修補，搶在攻擊者之前。通常防守方會更在意，這是一種可能的新模式。未來的 agentic blockchain 可能跟大家想像的完全不同，重點不是花錢，而是防禦。Yeah, we're more like a summit layer, but we're not, like, bundling nodes. But, I mean, this isn't the problem we're trying to solve. Yeah, trying to be more a little bit less. Cryptography and more like, what the fuck are they going to spend? What are you doing? Can you explain? I don't—I'm not up to date. What are you— So basically, it's the Flop Network is, we call it food for your AI agent. So what is the payment network of AI? Our theory is it should be whatever some currency that we create that's directly convertible into compute, and that is why you would hold it. Similarly, we hold dollars because we can buy food. So maybe it's self-preferential where it uses that compute to fortify itself. Sure. Yeah, I think it's the other way around. It's you, you know, bundle a blockchain with your matrix multiplication. And basically, like, formal verification—I mean, well, formal verification only gets you so far if there's, like, new math or what have you. Exactly. But even still, whatever the new model is, you know, you have some open router-style model registry. Like, it's almost like high-frequency trading, right? As soon as that new model is listed, your blockchain has a token budget, it hits that. It asks questions of every single line of the codebase. You're getting new prompts, obviously, that's also an important kind of thing. And it's constantly trying to hack itself and patch itself before an attacker can kind of do that. And probably the defender cares more. So that's one possible model. Agentic blockchain may actually be something very different than people think. It's not spending, but defending.

34:06L2 專案關閉潮與區塊空間現實

34:06最後想聊一下最近一波專案關閉潮。上週 Blast 這個 L2（Blur 推出後誕生的），還有跟 Pudgy Penguins 相關的 Abstract 也宣布關閉。每次這種專案倒閉，大家都很生氣，說自己在農點數，怎麼突然就沒了？但這其實反映時代變了。以前我們都說區塊空間無限，想開新鏈就開，成本超低，這些東西會永遠存在，永不關閉。結果現在這些專案突然變得很『地氣』，有現實的經濟壓力。像 Abstract 團隊就說，他們花了幾千萬美元在這個 L2 上，錢都花去哪了？不是在 sequencer，而是在 indexer、整合、各種雜項。現在區塊空間好像開始集中化，可能我們蓋太多了，這對未來有什麼意義？Okay, last story I want to touch on is there's been another spat of recent shutdowns. So just in the last week, we had Blast, the L2 that was—came about after the launch of Blur, and then Abstract, which is associated with Pudgy Penguins and that whole ecosystem, also announced that they were shutting down. Naturally, whenever these things shut down, you've got a lot of people getting mad, saying, Oh, you know, I was farming this. I had points. You know, why is this thing—why are you doing this to me? But I think it's a sign of the times that we used to tell ourselves in crypto, it's like, okay, well, we have infinite block space. We can create new ones on demand. It's so cheap. These things live on forever. They never shut down. And now all of a sudden, these things seem surprisingly terrestrial. They have these economic concerns of, oh, you know, the abstract guys were saying, you know, we spent tens of millions of dollars running this L2. And, you know, where's those $10 million going? It's not in the sequencers. It's like on the indexers, it's on the integrations, it's on this, on that. So how should we think about this moment with now it seems like block space is consolidating, maybe we overbuilt, and what does that mean for the future?

35:09其實很簡單，價格跌了，人就走了。你去看這些垃圾幣的 K 線圖就知道。去年 Marquee（新加坡夜店）一個平日包場要 50 萬新幣，去年整週都被訂滿，今年一場都沒人訂。去年都是 shitcoin 專案拿投資人錢去包場、請 DJ。今年完全沒了。說真的，創業很難，我其實很同情這些團隊，他們能做到這麼大已經很厲害了，很多人連零都到不了。給他們一點掌聲。第二個觀點是，你可以用 Carlota Perez 的理論來看，或 Gartner hype cycle。Perez 的研究從鐵路時代就有，幾乎每個新市場都是一場金融泡沫，激勵大家去尋找下一個 Amazon。Web1 是找 Amazon，Web2 是找 Facebook、Twitter。這些花費和 L2 投資，最後帶來了 Solana、NEAR、穩定幣等等。如果你回頭看 Web1 投資的錢，跟今天 Amazon、Google 的市值比起來根本是小數點後的誤差。每一次失敗都遠遠被成功抵銷。這是一場全球平行搜尋，找出能讓區塊鏈擴容的技術。五年前這是大事，現在大家覺得理所當然，Ethereum、Solana 都能隨便塞交易。L2 其實是在找 base、Robinhood Chain 這種答案。你投資組合全壓在這上面，長遠來看是對的。I mean, when number goes down, people leave. Period. End of sentence. It's all about the price. Like, look at the chart of any of these dogshit coins. Well, they actually didn't. Did they have a coin? They didn't TGE, right? I think they did, yeah. Yeah. Oh, I see. Of course. You were saying no one booked out Marquee. Yeah, so Marquee last year was, I think, half a million Singapore dollars on, like, a random weekday minimum spend, and it was fully booked all week last year. This year, not a single booking. And it was all shitcoin projects spending investor money on marquee and, you know, average DJs. Really? Really? Wow. So first I want to say, it's really hard to build a company, so I feel bad for these guys. They got very far. It's like many people die at zero. They don't even get to tens of millions or anything. So huge sympathy for them, props for them. The second thing is— You can think of, you know, Carlota Perez's work on, you know, or if you heard the Gartner hype cycle, right? Carlota Perez shows that goes back all the way to railroads and earlier. Basically, you can think of almost every new market as a financial bubble that incentivizes the search for, like, Web 1.0 was the search for Amazon, right? And— Then, like Web 2.0 was Amazon, Google, Web 1.0. Amazon Web 2.0 was a search for Facebook and Twitter, right? And so all of that spend and L2s and so on got you Solana, it got you NEAR, it got you stablecoins, it got you all this stuff. And if you go back and look at all the money that was invested in Web 1.0, it's like a rounding error relative to Amazon or to Google's market cap today, even... Like inflation-adjusted and discount rate-adjusted and so on and so forth. Every failure was worth the successes by a lot, right? So it's like this giant parallel search of the world for what technology would scale blockchains. That was a huge deal five years ago, and now we take that for granted that you can just blast transactions on both Ethereum and Solana for the most part, right? That's worth it. The L2s were a search for base and Robinhood chain. So you're, you know, you're 100% L on your portfolio. It is in the greater good. Yeah. Okay. Fair enough.

37:13分散、多元信任與 L2 的外部性

37:13回到剛剛講的 hack 鏈問題，其實這只會讓大家更分散。數學上至少有三種做法：第一種是『山頂人』模式，線上最小化，很多人會這樣做。第二種是信任政府，比如中國政府，直接重設一切、幫你跑鏈。L2 就像中國的鬼城，蓋太多最後死掉。第三種是多元分散，你同時信任多條鏈、實體黃金、新加坡幣等等。如果真的被 hack，你最多只損失 10%。因為現在你要重新評估投資組合，思考這個資產有沒有可能因為被 hack 歸零。就算有 10 倍上漲空間，也有 100% 歸零風險，所以不能 all in，必須主動分散。也許最後我們只能回到區塊鏈用帳號密碼登入，像 Facebook OAuth 一樣。當然這也會被 hack。There are things actually on your earlier point, there's one other thing in terms of these hack chains. It's just going to mean diversification. I think there's, like, with the math problem thing from earlier, I just want to make one remark. There's at least three approaches you can take. The first is go like mountain man physicality, right? Minimum necessary online, and some people are going to do that, okay? The second approach is actually to trust a government, like the Chinese government, that'll just— Reset whatever, and we'll run the chains for you. Yeah, yeah, exactly. The L2s are the ghost cities that die and then... Yes. So basically, you know, you can trust fiat, you can trust physical, or you can trust multiple, right? So fiat is like you're trusting, like, the Chinese government or some government. Physical is you're just going totally offline. You're only trusting, like, the physical world. And then multiple is... You're diversified across multiple chains and maybe physical gold and maybe like Singapore dollars and so on and so forth. So if there is a hack, you're only losing like 10% or something like that, right? Because now you have to re-underwrite your portfolio with, hey, what is the probability that this actually literally goes to zero due to a hack? Then even if it's got the upside of 10x, it's got the possibility of going down 100%. So therefore, I don't put it all in the best asset. Instead, I intentionally diversify it on this way. Maybe the answer is we just got to go to, like, username-password logins for blockchains, like OAuth with Facebook. Well, that would itself be hacked.

38:41L2 泡沫、鬼城與創新

38:41Tarun：這就像中國的鬼城，蓋太多，泡沫、詐騙一堆，最後開發者都歸零。但區塊鏈世界，L2 至少可以直接關掉 sequencer，整個城鎮瞬間消失，效率更高。其實這些 L2 讓 Ethereum 進步了，雖然不是每個 L2 都有貢獻，很多根本沒帶來技術或經濟創新，只是複製貼上，最後當然會死。但像 Robinhood Chain，靠的是分發（distribution），這才是創新。其他 L2 只是『蓋了就會有人來』，結果根本沒人。整個 L2 模型本來就是一種假設，Solana 其實已經證明這個假設不成立，Ethereum 也做出回應。你需要全球搜尋，找到像 Anatoly 這種能寫出 Solana 的人，否則就只能靠 L2。你必須實驗，看看哪種模式會成功。Tarun: Yeah, it's like the Chinese ghost cities. Like, there are too many that were overbuilt. There was a bubble. Then there was kind of— Tons of scams. All the devs zero out, except, except instead with the blockchain, I can just shut with L2 at least. I can say bye-bye sequencer and the entire town gets demolished. True, true. Right? Although it's like a more efficient version of that. But look, all those L2s were good because they kind of made Ethereum improve. You know, it was, it was like— Not all. Not all L2s improved. Some. I'm gonna say a lot did not. What I mean by that is the— How much money did you lose on L2s? Zero. Zero. Okay, there we go. Very noble of you to point out the nice externality of L2s failing. Yeah, it was a positive externality. Look, I mean, look, it's just— hey, not knocking. He's a good investor, right? Yeah, that's true. Yeah, I will say, no, look, I do think there is a sense in which it's similar to Chinese ghost cities in that Ethereum really did encourage the creation of L2s, right? This was, in a way, kind of loosely centrally planned, in Vitalik and the EF being like, L2s is the solution. Everyone make an L2. And, like, almost certainly, it was almost like hyperstitioned that there was going to be demand for all these L2s and that they would be cheap and efficient to run. It turns out they're not. It turns out, like, there's so many middlemen that kind of take little pieces when you launch an L2 of pay us for indexing, pay us for being included on CoinMarketCap and DeFi Llama and blah, blah, blah, blah. Oracles. And Oracle, that's right, that's right, that's right. And so, like, to be relevant as an L2 is actually quite expensive, and that probably explained a lot of the capital destruction beyond just— But also, like, there are a lot of L2s, like— Abstract and blast that we're just like, we copy-pasted. We didn't really change much of the L2. We provided no technological innovation, provided no economic innovation in terms of the fee model. Okay, what do you expect is going to happen? I'm just copy-pasting a bunch of potential. But for Robinhood Chain, that was the right answer. We don't need technological— Yeah, but they have distribution. Exactly. Distribution. Distribution is the innovation. But they're bringing— That's bringing economic activity. These other ones were like, hey, if you build it, they will come. Like have the party at Marquee and maybe we'll have some clients. No. Well, look, the entire Elto model was a thesis that was like, you know, I mean, like Solana disproved that thesis in many ways, and then Ethereum responded. But, but like, you needed essentially like a global search, a global draft to find the Anatoly who could— Code as well as, you know, Vitalik, right? And, you know, that was like, otherwise it would have—you actually needed that example of an L1 that could actually scale that then forced it to not be L2s. So you kind of just need to run the experiment and see what version was going to work. Okay, fair enough. On that note, we're at time. We got to wrap. Thank you, everybody, and we'll see you around in Singapore.