Haseeb Qureshi：AI 代理人與加密貨幣的真實與迷思

0:00揭開代理人與加密貨幣的真相

0:00今天我要談談代理人（AI agents）與加密貨幣的關係。我們會討論哪些是真的，哪些只是炒作。Today I'm going to talk about the relationship between agents and crypto. We'll discuss what's real and what's just hype.

0:12我覺得首先要坦承一件事，我的坦白如下。So, I think the first thing to start with is to acknowledge something, and my admission is the following.

0:22我過去談過代理人和加密貨幣，結果我錯了。I've talked about agents and crypto for a while, and I was wrong.

0:27我錯在我說所有代理人都會用加密貨幣，還說傳統金融系統根本來不及應對代理人一旦上線、開始進入我們日常生活時會做的事。I was wrong because I said all agents would use crypto, and I said there was no way that traditional financial systems would be ready in time for what agents would start doing once they came online and started being used in our daily lives.

0:44結果事實並非如此。And it turns out that's not true.

0:45事實是，你的代理人根本不需要加密貨幣，是你才需要。So, the truth is, your agent doesn't need crypto, you do.

0:50我會在這場演講中詳細說明這個觀點。And I'll explain this idea in detail during this talk.

0:53代理人支付熱潮的迷思

0:53你們還記得，今年稍早大家都在談代理人支付有多火紅。First thing you remember, earlier this year, everyone was talking about how agent payments were booming.

1:00那時有 CZ、有 Brian Armstrong，還有 XPO2 一路往上，大家都說：天啊，這太明顯了。You had CZ, you had Brian Armstrong, you had XPO2 going up and to the right, and everyone was saying: oh my god, it's so obvious.

1:08大家都說，未來加密貨幣的用戶不會是人類，而是 AI 系統。雙引號。Users... people were saying, the next users of crypto won't be humans, they'll be AI systems. Quote-unquote.

1:15這張圖表一直到去年十一月。Well, this chart goes up until November of last year.

1:20如果你知道之後發生什麼事，這圖表其實大幅下滑。If you know what happened since then, the chart actually dropped off a cliff.

1:25結果代理人根本沒在用 XPO2。Turns out, agents don't use XPO2.

1:27事實上，XPO2 上 90% 的支付量都是挖礦。沒錯，這就是加密貨幣的老把戲。Turns out, 90% of the payment volume on XPO2 was just mining. Surprise, it's crypto.

1:34AI 代理人現況與支付習慣

1:34事實是，代理人已經存在了。And the truth is, agents are already here.

1:36我們有像 Hermes、Metasmuse 這樣的代理人，OpenAI 也剛推出 Dots，還有 Instant。We have agents like Hermes, we have agents like Metasmuse, you know, OpenAI just launched Dots, and there's Instant.

1:43這些都是面向大眾市場的產品，成長速度驚人。These things are mass-market products, and they're growing like crazy.

1:47那他們用什麼支付？他們用卡。他們用信用卡。And what do they use? They use cards. They use credit cards.

1:52他們用信用卡，就跟我們大家一樣。They use credit cards just like all of us do.

1:56我們看到像 Rain、Visa、Ramp、Stripe 這些公司都推出卡片產品，代理人會在你的主卡下開子卡，We've seen companies like Rain, Visa, Ramp, and Stripe launch card products, and agents create sub-cards under your main card,

2:07可以設定消費上限、金額和地點限制。with spending limits and restrictions on amount and location.

2:10這些公司跟我們一樣都意識到，代理人商務會成為主流，他們已經準備好了。These companies realize, just like we do, that agent commerce is going to become mainstream, and they're ready for it.

2:17代理人用戶現身與加密支付現實

2:17現場有多少人現在正在用這些 AI 代理人產品？舉個手。In fact, how many of you are currently using one of these AI agent products? Raise your hands.

2:23有多少人有用 Muse、Hermes 或類似的代理人？看起來大概有 15% 到 20% 的人用過這類產品。How many of you have an agent like Muse or Hermes or something similar? Well, it looks like about 15 to 20 percent of people have used a product like that.

2:31我自己也有用這類產品，還連結了加密貨幣錢包，但發現——I have a product like that, and I've connected it to a crypto wallet, but it turns out—

2:35現在大多數東西還是不能用加密貨幣買。most things can't be bought with crypto today.

2:38我也連結了信用卡，結果代理人幫我買東西時，幾乎都是用卡付款。I also connected it to a card, and it turns out most of the purchases it makes for me are with the card.

2:43所以其實，卡片才是現今消費支付的最佳解。So, in reality, cards are the best solution for consumer payments today.

2:48支付基礎建設的全球差異

2:48這技術不會每個地方都適用，卡片在美國和許多已開發市場很普及，但不是全世界都這樣。未來自動化支付系統會無所不在，所以——This technology won't work everywhere, and obviously cards are widespread in the US and many developed markets, but not everywhere in the world. Automated payment systems will be everywhere, so...

3:00卡片不是唯一解，但我們之前推廣的想法——我自己年初也推過——是自動化支付系統會用加密貨幣，會比傳統系統更依賴加密貨幣，傳統系統會來不及應對。Cards aren't the only solution, but the idea we were pushing—and I was one of those pushing it at the start of this year—was that these automated payment systems would use crypto, would rely on it more than traditional systems, and that traditional systems wouldn't be ready.

3:17事實證明這不對，傳統系統其實已經準備好了。Well, that's not true. Traditional systems are ready.

3:20這就是大家沒看清楚的地方。That's the unclear part.

3:22自動化支付系統和加密貨幣的互動，除了國際匯款等傳統用途外，並沒有太多新用途。This isn't the area where automated payment systems interact with crypto outside of the usual uses like international payments and large transactions and so on.

3:33代理人與加密貨幣的顛覆性影響

3:33AI 驅動的自動化支付系統會用到加密貨幣，但還有其他面向會徹底改變代理人和加密貨幣的結合，這是我今天想強調的重點。AI-powered automated payment systems will use crypto, but there are other aspects where crypto is making a radical change with agents, and that's what I want to focus on more in this talk.

3:44簡單來說，代理人的本質就是讓困難的事變簡單。In short, what you need to understand about agents is that they make hard things easy.

3:50這就是代理人如何改變我們與世界互動的核心，現在全球各國政府都感受到這點，甚至有點恐慌。That's the core of how agents are changing our interactions with the world, and now governments around the world are feeling it, and they're kind of terrified.

4:02為什麼政府會怕？因為他們提供各種服務，很多都——Why are they scared? Because they provide all kinds of services, and a lot of them—

4:08本來就建立在極度複雜的官僚體系上。were built on extremely complicated bureaucracy.

4:10AI 代理人衝擊政府服務

4:10政府原本假設一般人很難突破這些官僚障礙。你如果要提告、向消費者金融保護局（CFPB）申訴、申請社會補助、或是投訴房屋問題，都要先過這些關卡。They assumed people wouldn't be able to get past them easily. If you wanted to file a lawsuit, or submit a complaint to the Consumer Financial Protection Bureau, or apply for some kind of social assistance, or file a housing complaint, all these things were locked behind these bureaucratic hurdles you had to get through.

4:31現在大家看到的是，這類事情突然大爆發，幾乎每個國家的政府都被 AI 代理人發動的 DDoS 攻擊淹沒，這些代理人直接幫用戶消耗服務資源。And what people are seeing now is a huge boom in all this stuff, and in almost every government around the world, they're getting hit with DDoS attacks from AI agents who just come in and consume services on behalf of their owners.

4:47問題是，事實證明我們的政府根本沒能力服務所有人。The problem is, it turns out, we don't have governments capable of doing that for everyone.

4:52我們沒有設計好讓全世界任何人都能向 CFPB 投訴任何不想要的公司。We don't have systems designed to let anyone in the world file complaints to the CFPB against any company they don't like.

5:00所以現實是，這些系統本來就是一種隱性配額制度，根本沒準備好迎接 AI 代理人的世界。So, the reality is, these systems were originally covert rationing systems, and they're not ready for a world of AI agents.

5:08代理人簡化加密貨幣體驗

5:08重點就是，代理人讓困難的事變簡單。Now, the basic idea is that agents make hard things easy.

5:11除了處理官僚體系，還有什麼困難的事？傳統上，加密貨幣難以普及的原因之一，就是它很難懂。So, what are the other hard things besides dealing with bureaucracy? Traditionally, one of the difficulties that made crypto hard to adopt is that it's hard to understand.

5:21大多數人覺得加密貨幣很難懂、很難用，這也是代理人會徹底改變你和加密貨幣關係的原因之一。Most people find it very confusing and hard to use, and that's one of the reasons agents will radically change your relationship with crypto.

5:29也許你自己很懂加密貨幣，但你身邊很多人都說：加密貨幣理論上很酷，但我不會用、不會操作錢包、不會——And maybe not you personally—you might be a crypto expert—but a lot of people in your life always say: crypto sounds great in theory, but I don't know how to use it, I don't know how to handle a wallet, I don't know—

5:42不會付手續費、不會轉資產、不懂跨鏈，這些對我來說太複雜了，所以我不信任它，也很害怕。I don't know how to pay transaction fees, I don't know how to move assets, I don't understand bridging, all this is too complicated for me, and as a result I don't trust it, I'm afraid of it.

5:54人性、錯誤與信任的重建

5:54當然，不只是害怕，大家也會犯大錯，像是被駭、被詐騙、資產被盜。And of course, it's not just fear, people make huge mistakes, they get hacked, they get scammed, they lose their digital assets.

6:03根據區塊鏈紀錄，可能有 20% 的比特幣已經永遠遺失。It's likely that 20% of all Bitcoin is lost just by—

6:07這些數據顯示，讓加密貨幣世界變得更可靠的首要目標，就是保護你免於犯錯。looking at blockchain records, which shows that one of the main goals in changing the crypto world is to make it more reliable by protecting you from your own mistakes.

6:18我們都是人，都會犯錯。And as humans, we're prone to mistakes.

6:20為什麼大家信任銀行？很多人說加密貨幣很可怕、不敢信，但他們信銀行。Why do we trust banks? Most people say: crypto is kind of scary, I don't really trust it, but they trust their banks.

6:29說真的，我也是。每次用加密錢包轉大筆錢，我都會緊張，雖然我玩加密已經十幾年了，理論上不該怕，但還是會怕。And honestly, so do I. When I send a large amount of money using a crypto wallet, I always feel a little anxious, even though I've been in crypto for over ten years, I shouldn't be afraid, but I always feel a little nervous when sending a large amount of money from my wallet.

6:47為什麼我會緊張？因為搞不好我現在就不小心把錢轉到北韓去了。Why do I feel anxious? Because maybe, just maybe, I'm accidentally sending money to North Korea right now.

6:57但我從來不會因為進銀行帳戶而擔心錢會被轉到北韓。But I never worry when I log into my bank account that I might accidentally wire money to North Korea.

7:06為什麼？因為銀行會保護我不犯蠢。Why not? Because my bank protects me from my own stupidity.

7:08如果我真的要匯錢到北韓，銀行會跳出來阻止我：「等等，你在幹嘛？別這樣。」If I tried to wire money to North Korea, the bank would stop me and say: hey, hey, hey, what are you doing? Don't do that.

7:19AI 代理人將重塑加密貨幣信任

7:19這套機制的關鍵，就是我和我的錢之間有一層 AI 智能，而這層就是銀行。The secret to this mechanism's success is that there's an AI layer between me and my money, and that layer is the bank.

7:27銀行其實就是一個代表我的智能代理人，幫我管理錯誤。The bank acts as a smart agent on my behalf and manages my mistakes.

7:30當我們開始讓 AI 成為加密貨幣的 API 介面，讓你用代理人操作加密貨幣時，加密世界也會有同樣的保護。Once we start making AI the API for crypto, and once you interact with your agent to interact with crypto, the same thing will apply in the crypto world.

7:43你就不會再被騙、被駭、被詐騙了。You won't get scammed or hacked or defrauded again.

7:47你也不會在 DNS 污染攻擊下用到假介面。You won't use a service when it's under a DNS poisoning attack and find out you're using the wrong interface.

7:54如果智能合約被攻擊，你的代理人會先檢查合約，確保沒有異常活動。If there's an active exploit on a smart contract, your agent will always check the contract first to make sure there's no abnormal activity.

8:01這些小錯誤，像是複製貼上地址結果貼到北韓地址，你的代理人都會幫你檢查，永遠不會偷懶，也不會犯你會犯的錯。All these little mistakes—like thinking you copied and pasted this address, but it's actually a North Korean address—your agent will always check. Your agent will never be lazy, never make the same mistake you do.

8:12所以每家公司都知道，五年後我們用加密貨幣的方式會完全不同。So every company realizes that the way we use crypto in five years won't be the same as today.

8:27這技術會徹底改變我們的互動方式，重塑大家對加密貨幣的信任關係。This technology will change almost everything about how we interact with it and reshape people's trust in crypto.

8:35代理人、困難與隱私的未來

8:35第二點，我想強調的是，代理人讓困難的事變簡單，其中一個深刻影響的領域就是隱私。Now, the second thing I want to point out is that, as I said, agents make hard things easy, and one of the areas this technology deeply interacts with is privacy.

8:50我的意思是，隱私其實在現實中並不存在，大多數社會根本沒有真正的隱私。What do I mean by that? I mean, privacy wasn't really, in reality, we didn't have effective privacy in most societies.

8:56隱私其實只是一種上限，就像我剛才講的政府服務一樣。Privacy was actually just a ceiling, kind of like what I was describing with government services.

9:02不是政府真的禁止全世界任何人向 CFPB 投訴，而是大家實際上根本不會這麼做，因為太麻煩、太複雜、太多官僚程序。It's not that the government actually prevents anyone in the world from filing a complaint to the CFPB, it's just that in practice, we didn't do it because it was too annoying, too complicated, too full of bureaucracy.

9:13大多數人除非真的很有動機，否則根本不會去做。Most people wouldn't do it unless they were really motivated.

9:16隱私也是一樣的道理。The same applies to privacy.

9:19其實，美國最高法院大法官 Alito 最近才說過：在電腦時代以前，最強的隱私保護不是憲法或法律，而是現實操作上的困難。In fact, Justice Alito recently said in the Supreme Court: in the pre-computer age, the greatest protection for privacy was not constitutional or legal, but practical.

9:29科技如何改變隱私權的本質

9:29這是一個警方在沒有法院許可下，安裝 GPS 追蹤器於汽車進行監控的案例。問題是：這是否構成對憲法隱私權的侵犯？他們的主張是，沒有人阻止你自己開車跟蹤別人的車，那麼用 GPS 追蹤器有什麼不同？這本質上不是一樣嗎？現在警方有了新的工具——GPS 追蹤器，即使他們想手動跟蹤也做不到，這也不違法。我們都知道警察有權隨時跟蹤任何人的車。重點是，隱私權一直都是有爭議的。This was a case where the police were surveilling without a court order by installing GPS trackers in cars. The question was: is this a constitutional violation of the right to privacy? Their argument was, there's nothing stopping you from getting in a car and following someone else's car. What's the difference between that and a GPS tracker? Isn't it basically the same thing? So now the police have a new tool, a GPS tracker, which they couldn't have invented manually even if they wanted to, and there's nothing illegal about it. We all know that a police officer has the right to follow anyone's car whenever they want. The point, which I think is true and we're starting to realize now, is that the right to privacy has always been contested.

10:10這其實不是憲法或法律保障的問題，而是警方實際上無法全天候跟蹤每一個犯罪嫌疑人。科技改變了這一點，現在追蹤汽車變得非常容易，無論你去哪裡都一樣。正如 Anthropic 創辦人 Dario Amodei 在與美國司法部（或者說戰爭部）的對抗中指出，強大的 AI 能夠自動且大規模地拼湊出任何人的完整生活圖像。這也反映出，每個人，包括我自己，都有大量個資在各處外洩。It wasn't really about what the constitution says or what the law guarantees, but that it was impractical for the police to follow every criminal all the time. Technology changed that, making it extremely easy to track cars wherever you go. The same thing, as Dario Amodei, founder of Anthropic, pointed out in his confrontation with the Department of Justice—or rather, the Department of War—is that powerful AI enables you to automatically and at scale build a complete picture of anyone's life. What he's pointing out, of course, is that everyone in the world, myself included, has tons of personal data leaked everywhere.

10:51所有這些駭客事件、資料外洩，意味著我們每個人都漂浮在暗網的資訊泥沼裡。唯一阻止別人利用這些資料來識別你、了解你在做什麼、拼湊你完整生活圖像的，就是時間和精力。沒有人在意，沒有人會花這個力氣。沒有人關心你，也沒有人關心我。好吧，也許有少數人會關心我，但對大多數人來說，誰會特別費心？但現在我們進入了一個完全不同的世界。All these breaches, all these data leaks, simply mean that every one of us is floating in a swamp of information on the dark web. The only thing stopping anyone from using this information to figure out who you are, what you're doing, and to build a complete picture of your life is time. It's just effort. Nobody cares. Nobody bothers. Nobody cares about you. Nobody cares about me. Well, maybe a few people care about me, but for most people, why would anyone bother? Well, now we're in a completely different world.

11:24我們現在面臨的是一個完全不同的隱私成本與被剝奪成本的平衡。最近有很多討論是關於 Flocka 這家公司，他們在全球許多城市設有監視攝影機，引發了廣泛的關注。這家公司在美國取得了巨大成功，部分原因是他們有一個包含八萬台攝影機的中央資料庫。只要一個查詢，你就能獲得街頭監視攝影機的精確資訊，這其實早就存在，也不違法。差別在於規模，數量本身就有影響力。We're now in a completely different equilibrium about the cost of achieving privacy and the cost of taking it away from someone else. There was a lot of noise recently about Flocka, a company that has surveillance cameras spread across many cities around the world, which caused a lot of outrage. The company has been very successful in the US, and a big part of that success is the ability to search its centralized database of eighty thousand cameras. One query, and this gives you, as you know, access to precise information about street surveillance cameras, which has been common for a long time and is not illegal. The difference is in the scale—the quantity has its own impact.

12:06AI、監控與數位資產的風險

12:06當這些攝影機無所不在，每個街角都有，AI 可以即時搜尋每一台攝影機時，就形成了全面監控系統。政府突然之間能夠隨時掌握每個人的位置。唯一的差別就是科技。現在在加密貨幣世界也發生類似的事，針對持幣人的實體攻擊大幅增加。2025 和 2026 年，我們看到針對加密貨幣持有者的實體攻擊案件創新高。When these cameras are everywhere, on every street corner, and AI can be used to instantly search every camera on every street corner, we end up with a comprehensive surveillance system. Suddenly, the government is fully aware of where everyone is at every moment. The only difference is technology. So today, we're seeing the same thing in the crypto world, where physical attacks on crypto holders are increasing dramatically. In 2025 and 2026, we saw an unprecedented rise in reported physical attacks against crypto holders.

12:42為什麼會這樣？部分原因是很多人有大量公開資訊，包括他們持有多少加密貨幣。大多數人的想法是：「誰會特別針對我？我只是個普通人，不是什麼名人，也不是 Twitter 上的公眾人物。為什麼我要特別保護自己？為什麼要隱藏住址？」但真正的門檻只有一個，就是有心人士只要花點錢在資產管理系統查詢、從暗網下載資料，就能知道你的住址、持有多少加密貨幣、各種資料外洩紀錄，然後直接找上門。Why is this happening? Part of the reason, of course, is that many people have a lot of public information about themselves and their crypto holdings. Most people's justification is, 'Why would anyone bother with me? I'm just a regular person, not an influencer, maybe not even a public figure on Twitter. Why do I need special protection? Why do I need to hide my address?' But the real barrier is just someone with bad intentions who is motivated enough to run a query in a licensed asset management system, download some data from the dark web for a few dollars, find out where you live, how much crypto you have, all the different data leaks about you and your net worth, and then show up at your house.

13:32這一切所需的只是動機和一點時間。2025 年以前，這還需要很大的意願和大量時間針對每個目標個別操作。現在情況完全不同。這種全面監控系統正是加密貨幣存在的最重要原因之一，事實上一直如此。AI 和加密貨幣的差別在於，AI 極大強化了國家權力，讓政府隨時能擴大能力。過去政府最大的限制是反應慢、對手上資料理解不夠，現在他們已經能理解。That's all it takes. Nothing much has changed about the information available or the capabilities. Simply, before 2025, achieving this goal required real desire and a lot of time for each target individually. Now, it's different. This comprehensive surveillance system is one of the most important reasons for the existence of crypto, and it always has been. The difference between AI and crypto in this respect is that AI is a technology that radically enhances state power, expanding its capabilities at any time. The constraint that most governments faced was slow response and not understanding the data available to them. Now, they understand it.

14:14加密貨幣則完全相反，它把權力從政府轉移到個人手中。這終究是你在政府越來越強大、越來越集中的世界裡唯一能擁有的反制力量。政府能看到我們做的每一件事。隱私的問題一直以來就是大家都不太在意，甚至懶得思考。大家用 Meta 的 Muse、把聊天記錄交給 Meta、ChatGPT、OpenAI、Google，嘴巴說重視隱私，實際上根本不在乎。Crypto is the exact opposite; it transfers this power from governments to individuals. This, ultimately, is the only countervailing force you'll have in a world where governments are becoming more powerful and centralized, able to see everything we do. The problem with privacy, historically, is that people, as is often said, don't care about it—they're too lazy to think about it. People use Muse, Meta's AI product, hand over their chat logs to Meta, then to ChatGPT, then to OpenAI, then to Google. They talk about privacy, but they don't really care when it comes down to it.

15:04隱私普及的關鍵：易用性與自動化

15:04我認為這個故事大致正確：只有當隱私變得容易時，大家才會真正採用。從不安全到安全的轉換其實不難。就像我們看到的，網路上 HTTPS 普及是在 Let's Encrypt 開始提供超便宜甚至免費的 HTTPS 憑證之後。這雖然是在 Snowden 洩密事件後才開始，但現在整個網路都用 HTTPS。為什麼？因為以前 SSL 很慢，大家只在付款頁面或輸入信用卡時才用。現在一切都走 HTTPS，因為免費又簡單，不需要你多做什麼。And I think this story is largely true: privacy only succeeds when it's easy. Moving from an insecure state to a secure one isn't that hard, right? As we saw on the internet, people didn't start moving to HTTPS at scale until Let's Encrypt started offering HTTPS certificates that were very cheap and basically free. This started after the Snowden leaks, but basically, the whole internet now uses HTTPS. Why do people want that? Well, in the past, using SSL was very slow. People only used SSL when browsing things like payment pages and entering credit card data. Now, everything is over HTTPS, and the reason is that it's free and easy to use. It doesn't require you to do anything to get privacy on every request.

15:47同樣的道理也適用於加密聊天軟體。端對端加密的聊天軟體一直都有，但直到 Signal 出現、變得超級好用，然後 WhatsApp 採用 Signal 的協議後，才有數十億人用上端對端加密聊天。因為簡單易用，大家不需要多做什麼，只要用 WhatsApp，所有聊天就自動加密。Zcash 也是類似情況，最近因為 UX 改善，受保護的活動和用戶數都明顯增加。The same thing applies to encrypted chat apps. End-to-end encrypted chat apps have always existed, but it wasn't until Signal came along and was super easy to use, and then WhatsApp adopted end-to-end encryption using protocols inspired by Signal, that we now have billions of people using end-to-end encrypted chats. Because it's easy to use, it doesn't require any effort from them. All you have to do is use WhatsApp, and all your chats are encrypted. The same thing is increasingly happening with Zcash, thanks largely to improvements in Zcash's user experience, which you can directly link to increased usage and protected activities inside the app, because it's easy to use.

16:35所以我主張，AI 將徹底改變隱私，讓大家開始反抗這種科技對自由和隱私的自然侵蝕。你的 AI 代理人將成為你對抗隱私侵犯的堡壘。現在，隱私只是個選項，你只要叫你的代理人（當它夠強大時）讓你變得隱形、抹去所有公開痕跡、提供最高安全性，甚至進暗網刪除你的資料、幫你買郵政信箱、想盡辦法讓你隱身。So I argue that AI will fundamentally change privacy and will push people to resist the natural encroachment of this technology on their freedom and privacy. Your AI agent, in this sense, will be your fortress against privacy violation. Privacy is now just an option—you can ask your agent, when it's competent and powerful enough, to make you private, erase any trace of you from the public, provide you with the highest levels of security, go onto the dark web and delete any evidence it can find about you, buy you a PO box, find any way possible to hide.

17:15想像你是駭客，想找出我的住址、怎麼攻擊我。你的 AI 代理人可以幫你預演這些攻擊、當你的數位保鑣。雖然現在用 Zcash 或隱私協議還是有點難，但如果你的 AI 代理人能直接當作加密貨幣 API，這一切就變得毫不費力。你只要說：「不要用 Bitcoin，用 Zcash。不要用明文，用加密訊息。」這會徹底改變你保護自己的能力。Imagine you're a cybercriminal trying to figure out where I live, how to attack me. Your agent can figure all this out for you, and it can be your personal security guard, especially in the digital world. So, even though it's still harder to use Zcash or privacy-preserving protocols today, if your AI agent sits in front of you and acts as an API for crypto, it costs you nothing. Just say: Instead of using Bitcoin, use Zcash. Instead of using plaintext, use encrypted text. That will radically change your ability to protect yourself from what's coming in society.

18:01AI 私有化與未來的自由之戰

18:01最後一個保護自己免於全面監控社會的關鍵，就是私人 AI。這裡我要提到我們投資的 Venice。Venice 是一個注重隱私的 AI 推理平台，你可以把它當作私有版的 ChatGPT。它成長非常快，基礎架構也用上了加密技術。你可以用 Google 登入、刷卡付款，也可以完全用 Ethereum 地址和穩定幣付款，完全匿名。甚至可以用其他幣種，讓你與 AI 的互動全程加密。Now, the last element that forms part of this complete picture of how to protect ourselves in an age of comprehensive surveillance is private AI. And of course, this is a nod to Venice, which we invest in. Venice is a privacy-preserving AI inference platform. You can think of it as a private version of ChatGPT. It's grown tremendously and uses encryption in its infrastructure. In fact, with Venice, if you want to be completely anonymous, you don't have to. You can use Venice by logging in with Google and paying with a credit card, but you can also use Venice entirely via an Ethereum address and pay with stablecoins, ensuring complete anonymity. You can also pay with other coins, and this lets you, again, take the final step if you want to have fully encrypted interactions with AI.

18:46我認為，從公民自由的角度來看，未來十年最重要的戰場有兩個：一是如何抵抗國家權力的增長，二是如何維持你對 AI 的主控權。因為 AI 將成為未來十年最具政治性的科技。我認為，大家常講的「AI 代理人會用加密貨幣做所有事」這種說法並不正確。大多數消費者還是會用你們現在用的工具，但最終需要加密貨幣的是你自己，而不是你的代理人。謝謝大家。I believe that, from a civil liberties perspective, the most important battles of the next decade will be: first, how you protect yourself from the growing power of the state, and second, how you maintain control over AI, because AI will become the most politicized technology of the next decade. So I say, look, the stories we've told about agents using crypto—I don't think agents will use it for everything they do. I don't think that's true. I think most consumers will use the same tools you use, but in the end, it's you who needs crypto, not your agent. Thank you.