比特币遇上AI智能体，一切都将改变 | 蒂尔曼·霍洛韦与安德鲁·帕里什

如果你看比特币，其实所有加密货币，它在收回权方面有一个特殊的用途，就是你可以锁定它，而不需要付出巨大的收回成本。

If you look at Bitcoin, but really all crypto, it it holds a special, use case as it pertains to repossession rights and how you can lock it up without tremendous, repossession costs.

所以对于现实世界的资产，你就没有这个优势。

And so you don't have that against real world assets.

对于黄金，你也没有这个优势。

You don't have that against gold.

对于任何实物商品，你都没有这个优势。

You don't have that against any physical goods.

比如，想想看，去收回一辆车。

Like think about, you know, going and repossessing a car.

如果这辆车不值多少钱，收回它的成本比抵押品的价值还高。

If the car is not worth very much, it's more expensive to repossess it than the value of the collateral.

所以这就变成了一个无法放贷的资产。

And so that becomes an unlendable asset.

大家好，最近怎么样？

What's going on, guys?

今天，我们和蒂尔曼·霍洛韦以及安德鲁·帕里什进行了一场很棒的对话。

Today, got a great conversation with Tillman Holloway and Andrew Parish.

他们是Arch Public的联合创始人。

They're the cofounders of Arch Public.

在这次对话中，我们探讨了一些有趣的话题。

And in this conversation, we go into some interesting topics.

首先，我们聊了聊代币。

First, we talk about tokens.

你有没有想过，有计算代币，也有加密货币里的代币？

Did you ever think about the fact that there's compute tokens and there's tokens in crypto?

代币的历史是怎样的？

What's the history of tokens?

为什么它很重要？

Why is it important?

它会如何改变你的生活？

How's it gonna change your life?

为什么AI代理会在AI、比特币、加密货币、代币化以及许多其他技术交汇的领域占据主导地位，这些技术正以完全相同的方式汇聚在一起。

Why are the AI agents taking over at the intersection of AI, Bitcoin, crypto, tokenization, and many other technologies that are all coming together in the exact same way.

然后我们讨论了全天候市场、波动性、这些人是如何投资的，以及他们构建了什么来帮助人们应对金融的未来。

And then we talk about twenty four seven markets, volatility, how these guys are investing, and what they have built to help people navigate all of the future of finance.

这是我和Arch Public的这两位天才的最新对话。

Here's my latest conversation with these two geniuses at Arch Public.

好的，Tillman。

Alright, Tillman.

我们看到AI和比特币的兴趣激增。

We've seen a huge surge in the interest of AI and Bitcoin.

这里的交汇点非常明显，我认为AI代理生态系统将比人们想象的要大得多，这意味着许多技术的可寻址市场显著扩大了。

The intersection here, it is very obvious, I think, that the AI agent ecosystem is going to be way bigger than people thought, which means the addressable market for lots of technologies just got significantly bigger.

在AI、比特币、加密货币以及这些数字技术的未来交汇点上，你们看到了什么？

What are you guys seeing when it comes to this intersection of AI, Bitcoin, Crypto, and kind of the future of these digital technologies?

我认为这是一个非常有趣的交汇点，你知道，代理将成为加密货币的主要用户。

I think it's a really interesting intersection of, you know, agents are gonna be the the prolific users of crypto.

智能体可以瞬间创建出来。

Agents can be created instantly.

它们不需要等到成年才能开始使用钱包之类的。

They don't have to become adults to start using, you know, wallets.

而且，无论是算力还是计算能力的交易，那都会成为最需要交易的商品，它们需要一种方式来完成这种交易，也就是代币。

And transactions for hash rate or for compute power, that's gonna be the commodity that's going to be needed to be traded against most, and they need a form by which they can do that, tokens, if you will.

它们……

They...

我们现在就是这么称呼它们的。

That's what we actually call them right now.

所以，我觉得这是未来五年会呈指数级增长的事情之一。

So that's one of the things that I think is just gonna grow exponentially over the next five years.

我们很快就会看到。

We're gonna see it.

但我也认为，与加密货币相关的债务，以及它在抵押品方面的潜力，可能是最有趣的部分。

But I also think that the debt, attached to crypto and what it can be from a collateral perspective is probably the most interesting part.

比如说，你看房地产，你会说，好吧。

If you look at real estate, for example, and you say, okay.

那为什么在美国，靠房地产发家的百万富翁会那么多呢？

Well, why have there been more millionaires created on the back of real estate in The US?

那是因为你可以很容易地从任何一家本地银行借钱，用房地产做抵押，而且有一个系统的资金流动，让他们能够……

It's because you can borrow against real estate very easily from any local bank, and there's a systematic flow of capital that allows them...

那些……

Those late...

本地银行可以把那笔债务卖出去。

Local banks to sell that debt.

而如果你看比特币，其实所有加密货币，它们在收回权方面有一个特殊的用途，就是你可以把它锁起来，而不需要付出巨大的收回成本。

And if you look at Bitcoin, but really all crypto, it it holds a special, use case as it pertains to repossession rights and how you can lock it up without tremendous, repossession costs.

所以，在现实世界的资产上，你没有这个优势。

And so you don't have that against real world assets.

在黄金上，你也没有这个优势。

You don't have that against gold.

任何实物商品都没有这个优势。

You don't have that against any physical goods.

比如说，你想一下，去收回一辆车。

Like, think about, you know, going and repossessing a car.

如果那辆车不值多少钱，收回它的成本比抵押品的价值还高。

If the car is not worth very much, it's more expensive to repossess it than the value of the collateral.

所以这就变成了一个无法贷款的资产。

And so that becomes an unlendable asset.

所以，如果你想尽可能多地把人纳入经济体系，就必须降低准入门槛，而加密货币，特别是比特币，就成了最理想的抵押品，因为它最容易收回，而且你可以编写智能合约来规定必须发生的具体事项。

And so if you wanna include the most people in the economy, you have to lower the barriers of entry, which crypto and specifically Bitcoin then become the most preferred form of collateral because it's the easiest to repossess, and you can write smart contracts that govern specific things that have to happen.

如果这些事项没有发生，不需要法官和陪审团通过昂贵的诉讼程序来裁决。

And if they don't happen, it's not, you know, a judge and jury that has to decide it through an expensive litigation process.

而是由智能合约立即决定结算结果。

It's literally a smart contract that decides instantly the settlement that takes place.

所以我认为，当我们开始看到加密货币贷款向公众开放时，你会看到大量的活动涌现，因为我认为以这种抵押品贷款的利率会更低，因为收回成本更低，而且他们会把一部分节省下来的成本转嫁给用户以吸引使用，这是我的观点。

So I think that when we start seeing crypto lending available to the general public, you're gonna see a mushroom cloud of activity that's going to really, you know, follow because I think that the the loan rates are gonna be lower against that collateral because the repossession costs are lower, and and they're gonna pass some of that on to induce usage, is what my argument would be.

我觉得我们会看到AI和金融的碰撞，比我们想象的要快得多。

I think we're gonna see a collision course of AI and finance, much quicker than we think.

而且我认为这会被全天候市场加速，因为它们正在到来，而且来得很快。

And and I think that's going to be, accelerated with twenty four seven markets because they're coming and they're coming fast.

纽交所、纳斯达克、芝加哥商品交易所，所有这些大型的，我们姑且称之为，与交易相关的管道和杠杆，在美国，都已经签署了协议，并且发布了新闻稿，说他们几乎准备好进行全天候交易了。

The NYSE, the Nasdaq, the CME, all of the the the larger, let's call it, pipes and levers associated with trading, any asset across The United States have all signed deals and have all put out press releases saying we're we're almost ready to go on twenty four seven trading.

那么，当那些从早上九点半到下午四点做交易的人，在晚上九点半到凌晨四点反正都在睡觉时，你怎么促进全天候交易呢？

Well, how do you facilitate twenty four seven trading when all the humans that do that from 09:30 to four are asleep from 09:30PM to 4AM anyways?

那么，你就需要有代理工具，或者说代理，来在那段时间内促进这些活动。

Well, you're gonna have to have agentic tools, agents, if you would, that are facilitating that activity during that time period.

所以这些系统正在被构建。

So those systems are being built.

坦率地说，这些系统正在Arch Public这里构建。

Those systems, frankly, are being built here at Arch Public.

所以你必须要有某种东西在凌晨两点代表你执行你的意愿。

And so you you you have to have something acting out your will on your behalf at two a.

凌晨两点。

M.

因为市场会在凌晨两点开放。

Because markets will be open at two a.

凌晨两点。

M.

就像加密货币市场在凌晨两点开放一样。

In the same way that crypto markets are open at two a.

凌晨两点。

M.

传统市场也会在凌晨两点开放。

Traditional markets are going to be open at 2AM.

摩根士丹利和贝莱德都在谈论这个。

It's all that Morgan Stanley or BlackRock wanna talk about.

把这也代币化，把那也代币化。

Tokenize this, tokenize that.

今天刚有个新消息，说贝莱德和Ondo签了个协议，你知道的，就是关于代币化的现实世界资产，以及这对他们意味着什么。

There was just a new headline today that BlackRock has signed a deal with Ondo associated with, you know, tokenized real world assets and what it means for them.

所以，全球金融界最大的玩家们，无论是交易所、托管机构，还是产品发行方，都被邀请到美国证券交易委员会开会，讨论24小时全天候交易的事。

So the biggest players in the world of finance, both exchanges, custodians, and then, you know, product producers, you know, have all been invited to the SEC to have meetings about twenty four seven trading.

所以这事肯定会发生。

So it's going to happen.

那么接下来的问题是，人工智能和金融之间会碰撞出什么火花？

So then the next question is, what is the collision between AI and finance?

短期内，就是你必须要有系统，能在凌晨1点、2点、3点、4点，市场仍然开放活跃的时候，同时服务机构客户和零售客户。

Short term, it's that You are going to have to have systems that act out both institutional and retail customers will at 1AM, 2AM, 3AM, 4AM while markets are still open and active.

这很快就会到来，非常快。

That's that's coming very, very quickly.

你知道有趣的是什么吗？

You know what's interesting?

我以前真没想过，就是“代币”这个词，在加密货币里，代币是什么意思。

I never really thought about it before, but the word tokens, what that means, there are tokens in crypto.

AI里也有代币。

There are tokens in AI.

我之前从没想过这两者之间的重叠。

And I never thought about the overlap on that.

你觉得这会带来什么影响？

What what do you think the ramifications are?

是不是未来就建立在代币之上，有些是金融代币，有些是计算代币，也许以后还会有其他形式的代币，但这就是我们现在所处的世界？

Is that just like the the future is gonna be built on tokens and some of them will be financial tokens, some of them will be compute tokens, maybe there's other form of tokens going forward, but, like, this is the world we live in now?

完全正确。

Absolutely.

在我看来，代币不过是一种加密的、可互换或不可互换的价值表示。

A token is nothing more than, in my mind, an encrypted either fungible or nonfungible representation of value.

所以如果你看看……

And so if you look at the...

展望未来，我们正在摆脱所有物理……

That going forward, we're moving against all physical...

你知道，现在有很多商家，尽管我觉得这仍然不合法，却已经拒绝收现金了。

You know, there's a lot of businesses in spite of it, I think, still being illegal, refuse to take cash now.

我们正在走向无现金社会，关于是否应该这样做、这样做好不好，双方都有很多有价值的论点。

We're moving to a cashless society, and and there's a lot of arguments that are valuable on both sides of, you know, whether we should do that, whether it's good.

但我认为，彩虹另一端的经济价值实在太诱人了，让人无法忽视。

But I think the economic value that lies on the other side of that rainbow is too hard to ignore.

我的意思是，世界上有很多人没有银行账户，超过40亿人，如果有办法把他们纳入进来，他们本来可以或者可能会被包括在内。

And what I mean by that is, like, there's a lot of people unbanked in this world, like 4,000,000,000 plus people that would would or could be included if there was a way to include them.

而且，你知道，在资本主义社会，获取客户的成本门槛是实实在在的问题。

And, you know, the cost barrier to having customers is a real thing in a capitalist society.

所以，在某个层面上，像路易威登这样的品牌不会去争取那部分客户群。

And so it's like at a certain level, customers like Louis Vuitton isn't isn't going after that customer base.

那么，从银行服务的角度来看，这些人从哪里获得服务呢？

And so where do they get served from a bankability perspective?

比如，哪里……

Like, where where...

他们什么时候才能对银行有足够的吸引力，让银行开始为他们提供服务呢？

When do they become attractive to a bank enough for them to start participating in the banking space?

我觉得区块链把门槛降低的程度，比任何现有系统都要大得多，简直是数量级的差距。

And I think blockchain lowers the bar better than any current system has ever lowered it by by orders of magnitude.

所以它能让所有人都参与进来。

And so it it allows everyone to participate.

而且我觉得这些代币，不管我们叫它加密货币，还是你是比特币的忠实粉丝，还是别的什么……

And I think the the tokens, whether, you know, we called them cryptocurrency or or you're a Bitcoin maxi or...

说到底，它……

At the end of the day, it...

它们都是代币。

They're all tokens.

它们只是以数字形式存储的价值代表。

They're just representations of value that are digitally stored.

这就是未来，因为这样电子商务就能让另外40亿人接入互联网。

And and that is the future because that's going to be the way in which ecommerce allows 4,000,000,000 more people to come online.

而且当你谈到全球经济，谈到打破那些阻碍法币即时兑换、转账和全球流通的壁垒时，我觉得我们真的正在目睹这种壁垒的瓦解。

And and when you talk about a global economy and you talk about breaking down the walls that preclude, you know, fiat from being instantly converted and transferred and sent anywhere globally, I think we're literally in the midst of watching that breakdown.

这很令人兴奋，因为我觉得如果做得好，而且如果工作量证明仍然……

And and it's exciting because I think if done correctly and, you know, if if proof of work still...

它意味着从安全角度来看，我们会在很长一段时间内有一个可以依靠的后盾，作为普通人，作为用户本身，而不必去信任其他人，对吧？

It means something that we're gonna have a a backstop from a security perspective for for a very long time that we can rely on as a people, you know, as the users themselves and not have to trust other other people for that?

嗯，你知道，代币存在的时间比我们意识到的要长得多。

Well, you know, tokens have been around a lot longer than we realized.

有一位代币的先驱，他在八十年代初真的是超前于时代。

There was a a forefather of tokens, and really was ahead of his time back in the early eighties.

他的名字叫查克·E·奶酪。

His name was Chuck e Cheese.

我不知道你们还记不记得他，但在他那里，你必须在玩任何游戏之前把你的钱换成代币。

I I I don't know if you guys remember that guy, but, you had to you had you to move your money into tokens before you played any games at his establishments.

所以查克·E·奶酪就是代币的先驱。

So Chuck e Cheese was the forefather of tokens.

但说真的，关于代币，有一点值得提，就是价值会在这些实体之间来回转移。

But in all seriousness, it's a good point to be made associated with tokens that there's value that's going to be moved back and forth between those entities.

金融代币和计算代币。

Financial tokens and compute tokens.

再说一次，我觉得我们正处在这些世界发生重大碰撞的边缘，而且这两个世界能以多快的速度、多轻松地协同工作，也是值得期待的。

Again, I think we're we're, like, on the precipice of meaningful collision, of those worlds, and what it looks like and the speed and pace and ease at which those two worlds can work together.

我们正在飞速朝着那个方向前进。

We're we're we're headed there very quickly.

本期节目由Lava赞助播出。

This episode is brought to you by Lava.

Lava卡是第一张让你每次消费都能获得高达5%比特币返现的信用卡。

The Lava Card is the first credit card that allows you to earn up to 5% back in Bitcoin on every single purchase.

大多数信用卡都有高昂的年费和不清晰的积分计划，但Lava卡提供了更好的选择。

Most cards have steep annual fees and unclear points programs, but the Lava Card offers a much better deal.

每次消费都能赚取真正的比特币。

Earn real Bitcoin every time that you spend.

而且，没有年费。

Plus, there's no annual fee.

在国外消费时，没有外汇手续费，也没有汇率差价。

There's zero FX fees and zero spread when you spend internationally.

此外，Lava还提供比特币信用额度、现金收益、即时稳定币存取款，以及银行账户之间的便捷转换，还有很多很多其他功能。

On top of that, Lava gives you access to a Bitcoin line of credit, yield on cash, instant stablecoin deposits and withdrawals, on and off ramps to and from your bank account, and much, much more.

使用Lava，你可以以最低利率借入比特币，赚取收益，全球转账，并且每次消费都能积累比特币。

With Lava, you can borrow against your Bitcoin at the lowest rates, earn yield, move money globally, and stack Bitcoin with each and every purchase you make.

大多数金融服务都会用高额费用和复杂的奖励计划来压榨你。

Most financial services squeeze you with high fees, confusing rewards programs.

Lava保持简单，帮助你通过日常消费建立长期储蓄。

Lava keeps it simple, and they help you build your long term savings via your everyday spend.

今天就访问lava.xyz/pump，几分钟内即可开始。

Visit lava.xyz/pump to get started in just a few minutes today.

lava.xyz/pump，今天就加入吧。

Lava.xyz/pump to get started today.

今天的节目由GalaxyOne赞助播出。

Today's episode is brought to you by GalaxyOne.

很多人都在让自己的钱生钱。

A lot of people got their money working for them.

但现在的问题是，你的钱到底在努力赚多少？

But the question now, how hard is it working?

GalaxyOne那边的人，他们走的是完全不同的路子。

Well, the folks over at GalaxyOne, they're taking a very different approach.

用GalaxyOne现金账户，每个人都能拿到3.5%的年化收益。

Everyone earns 3.5% APY using GalaxyOne cash.

如果你年收入超过20万美元，那你很可能属于合格投资者，这样你的现金收益最高能到8%，而且由Galaxy Digital提供担保。

And if you make over $200,000 a year, that means you're probably an accredited investor, and then you can earn up to 8% on your cash guaranteed by Galaxy Digital.

最棒的是，你可以设置让收益直接自动再投资到你喜欢的股票、加密货币或者ETF里。

The best part, is that you can actually set it so that your yield immediately is reinvested into your favorite stocks, crypto, or ETFs.

你完全不用每个月都去操心这件事。

You don't gotta think about it every single month.

它会自动帮你搞定一切。

It just automatically does it for you.

这就是我喜欢GalaxyOne的原因。

That's why I love GalaxyOne.

所以，如果你的钱没有像你一样努力赚钱，那就是时候行动了。

So if your cash isn't working as hard as you are, then it's time to make a move.

点击描述中的链接，今天就开始赚取收益吧。

Check out the link in the description and start earning today.

你知道吗，我觉得有趣的是，当我们看到一些开始流行的产品时，会出现一些几乎全新的网络类型。

You know what's interesting to me is as we look at some of the products that are starting to take hold, you have these almost new types of networks.

所以Instinct现在有……

So Instinct is now having...

如果我有一个代理，我妻子也有一个代理，这两个代理可以知道他们在同一个可信网络中。

If I have an agent, my wife has an agent, those two agents can know they're in a trusted network.

所以他们很可能会一起工作或合作。

And so they're likely to work together or collaborate.

嗯，钱就成了其中的核心部分。

Well, money becomes a core component of that.

谁付什么钱？

Who's paying for what?

你知道，钱怎么在这两个人之间流转？

How does it, you know, transfer between these two people?

然后你再看一些B2B类的产品，比如，我的供应商，我和他们的代理之间的关系，应该和那些在街上冷不丁向我推销、虽然同行但并不是我当前供应商的人不一样。

Then you go and you look at some of these b two b type products like, well, my vendor, I should have a different relationship, my agent and their agent, than somebody who is cold pitching me off the street that is in the same business, but isn't currently my vendor.

所以信任就成了一个非常有意思的元素，几乎就像人们愿意让代理和技术自动化更多操作，当他们已经某种程度上把对方列入白名单的时候。

And so trust becomes a very interesting element of this, where it's almost like people are willing to allow the agents and the technology to automate more when they've already kind of whitelisted.

我们知道白名单显然存在于加密货币、钱包，以及金融领域的代币世界里。

And we know whitelisting obviously exists in crypto and wallets and, know, kind of the the token world on on the finance side.

但当你开始以一种非常奇特的方式把这些东西融合在一起时，你其实是在试图复制人际关系运作的方式。

But when you start to merge this stuff together in a very weird way, you are trying to replicate the way that human relationships work.

我信任这个人，所以我愿意和他们做生意。

I trust this person, therefore, I'm willing to do business with them.

但你这么做的时候，也相当于把钥匙交给了AI。

But you're also doing it where you're kinda turning the keys over to the AI too.

你就像在说，嘿。

You're kinda saying like, hey.

你知道，我信任安德鲁。

You know, I trust Andrew.

我的AI和他的AI，你们去搞定吧。

My AI, his AI, like, go figure it out.

你们担心这个吗？

You guys worry about that at all?

比如，这是不是会增加安全方面的顾虑之类的？

Like, does that increase security concerns or or anything like that?

我觉得安全措施会不断迭代，让它变得更安全、更好用。

I think there's gonna be, you know, an evolution of security iterations that make it more secure and easier to use.

我觉得GrokBot可能是最近最好的例子，从用户体验的角度来看。

I think GrokBot's probably the best recent example of that from a user friendly, user user, you know, perspective.

它绝对是最好的，而且它在为你做事时有很多制衡机制。

It's by far the best, and it has a lot of checks and balances in, you know, what it's doing for you.

它在很多方面就像个人助理，会去完成很多跑腿的工作，但最终发送前会交给你审批。

It's acting in a lot of ways like a personal assistant where it's going and doing a lot of the legwork, but it's bringing it back to you for final approval before it's sent out.

所以我觉得可以通过编码来解决这些问题。

And so I think that you can code around those problems.

我确实觉得可信层是我们自己都见过的非常有趣的东西。

I do think that the trusted layers are really interesting thing that we've even seen ourselves.

Slack就是一个很好的例子。

And so Slack is a great example of that.

它既是一个公司内部协作的平台，同时也能非常有效地与外部伙伴合作。

Like, is platform for both internal, you know, efforts within a company, but then also you can bring and collaborate with external parties very effectively as well.

这就成了Grock Bot和Slack之间一个非常有趣的API集成，因为Slack可以成为那个可信层，你知道那里没有……

That becomes a really interesting API integration between Grock Bot and Slack because then Slack can be that trusted layer where you know that there are no...

你知道，那个Slack频道里的每个人都是经过批准或列入白名单的。

You you know, everyone in that Slack channel, is approved or whitelisted, if you will.

所以我觉得这将是未来一个重要的组成部分，尤其是在如何分配权限，决定你的代理能做什么、不能做什么方面。

And so I think that's gonna be a major component moving forward as it pertains to how you parse out permissions to what your agent can do and what it can't do.

它会发现自己被困在你故意设置的沙箱里，这样……

It's gonna find itself trapped in sandboxes that you purposely put it in so that...

而且，把你的机器人隔离开来也是一种有效的方法。

And even siloing your your bots is a is a effective means to do this.

比如，即使在GrokBot内部，你也可以隔离出五个不同的机器人，不让它们互相越界，这样就能获得一些制衡或冗余的预防措施，防止事情在未经额外监督的情况下被发送出去。

Like, even within GrokBot, you can silo out, you know, five different bots and and not allow them to see across the fence, if you will, so that you're getting some checks and balances or some redundant, you know, precautionary measures that it pertains to things being sent out without getting additional oversight.

我很喜欢我们能进行这样的对话，而且我可以在这里用一台超级计算机做实时研究。

I love the fact that we can have these conversations, and I can have a supercomputer here to do, live research.

而且，我的实时研究，我刚刚查了一下代币的历史。

And, my real time research, I just looked up what's the history of tokens.

它从哪里来的？

Where does that come from?

这可以追溯到二十世纪五十年代和六十年代，代币最初来自……

And it goes back to the nineteen fifties and sixties, and tokens originally came...

计算机科学家是从语言学那里借来的这个概念。

Computer scientists borrowed it from linguistics.

就是这种把一段文本当作某种解析器来处理的想法。

This whole idea of being able to take a chunk of text as, you know, some sort of a a parser.

然后他们把它放进编译器里，这样就能真正使用那些编程语言了，源代码会用这些令牌来表示关键字、运算符、标识符等等。

They then put that into compilers so that they could actually use those programming languages, and the source code would use those those tokens for keywords operators, identifiers, etcetera.

到了八九十年代，在网络和安全领域，出现了……

And then in the eighties and nineties, networking and security, there was...

IBM推出了令牌环网络，可以在网络中传递令牌，用来做身份验证。

IBM introduced the token ring network, the ability to pass the tokens around so that you could do authentication.

再往后，像RSA、SecureID这些认证凭证，甚至能真正给人们访问权限。

And then even taking it further with things like RSA, secure ID, authentication credentials, and and being able to actually give people access.

然后历史就直接跳到了2010年代中期的AI子词分词，里面讲了一堆对我来说太深奥的东西。

And then the history goes right immediately into the subword tokenization in AI in the mid two thousand and tens, and it talks about a bunch of things that are way too smart for me.

但大致思路就是利用数据压缩算法，想办法把文本真正转换成机器能理解的东西。

But the general idea is that it is using these, like, data compression algorithms, and it's trying to figure out how do you take text and actually translate that into the machine.

我觉得这有意思的原因是，过去两三年里最棒的一句话就是：最热门的编程语言是英语。

Now the reason why I think that's interesting is one of the best lines of probably the last two or three years is the hottest programming language is English.

现在你只需要用自然语言告诉机器你想做什么，它就会去帮你把这个产品做出来。

And now you can just simply, in natural language, describe to the machine, here's what I wanna do, and it will go and it'll actually create that that product for you.

嗯，这其实就是我们现在在金融领域看到的趋势。

Well, that's kinda what we're getting in finance now.

比如，你现在可以开始跟它说，嘿，

Like, you're now starting to be able to be like, hey.

你知道，我想……

You know, I wanna...

我希望我的代理能全天候工作。

I want my agent to work twenty four seven.

加密货币市场是24小时不间断的。

The market's twenty four seven in crypto.

最终，股票也会实现24小时交易。

Eventually, you're gonna get stocks doing twenty four seven.

我可不想坐在那儿点一堆下拉菜单，然后还得琢磨那些……

I don't wanna sit and click a bunch of drop downs and figure out all...

就像，直接帮我搞定就行了。

Like, just, like, do this for me.

你知道，就像那个玩笑说的，嘿，Claude。

Like, you know, the the joke of, like, you know, hey, Claude.

帮我赚钱。

Make me money.

别出错。

Make no mistakes.

就是……

It's...

你们可能是目前最接近这个目标的了。

You know, you guys are probably one of the closest versions of this so far.

能不能讲讲你们是怎么考虑用自然语言描述，还有怎么实现自动化，同时还要在二十四小时不停的市场里运作，而人类在这种市场里其实挺吃亏的，不管怎么说。

Like, describe a little bit how you're thinking about natural language, you know, description and and and, you know, the automation of this, but also operating in a twenty four seven market where humans are just ill positioned, you know, better or worse.

我们确实需要睡觉，也许还要过自己的生活，出去走走，而市场还在交易。

Like, we do need to sleep and and, maybe live our lives, touch grass, you know, while while markets are still trading.

嗯，创造一个全能的精灵，把它和你的钱包一起放出去，说‘去给我赚钱’，这其实是个愚蠢的想法。

Well, the notion of creating an all powerful genie that you can turn loose with your wallet and say, go make me money is a fool's errand.

这听起来很诱人，表面上很吸引人，因为智能体看起来确实很强大。

It's it's it's sexy on on, you know, face value because agents do seem very powerful.

但当你深入了解，你会发现你受限于你背后拥有的计算能力。

But when you peel back that onion, you realize, like, you're a slave to the amount of compute power that you have behind you.

如果你觉得现在交易有劣势，把你的计算能力和任何公司比一比，你远远排在后面。

And if you think you have a disadvantage now trading, compare your compute power to pretty much any company, and you're you're you're way down the totem pole.

所以……

So that...

这不是看待它的正确方式。

That's that's not the way to look at it.

在我看来，看待AI在市场中作用的方式，因为市场是零和游戏，而且你不能像机器人战争那样竞争，你必须先弄清楚你想要实现什么。

The way to look at AI in the markets, in my opinion, because they are a zero sum game and because you can't compete just like robot wars, you you have to find out what you're trying to accomplish.

因为这一切都归结于你个人的情况和你想达成的目标。

Because it all boils down to your individual circumstance and what you're trying to accomplish.

而AI是一个管理工具。

And the AI is a management tool.

它的目的是在你无法顾及的时候延续你的意志，并且在你情绪上不想做决策的时候执行你的意志。

It's meant to extend your will into time periods where you're not available, And it's meant to execute your will when you're not emotionally in a place where you want to be, you know, the decision maker.

这是两个主要的好处。

Those are the two primary benefits.

所以，如果有人觉得在这个方程里不需要一个非常重要的舞伴就能用AI工具成功交易，那就会碰壁，人们很快就会发现问题不是那样的。

And and so to to for people to think that there isn't a dancing partner that's very needed in that equation to get successful trading accomplished with AI tools, that's where kind of the the road of hard knocks comes in, and people find out very quickly that that that doesn't work.

你可以去追。

You can chase.

那基本上就是试图用你在网上听说的某个系统去打败赌场。

That's essentially trying to go in and beat the casino with some system that you heard about online.

赌场的赔率，你无法战胜，除非你作弊。

The casino's odds, you cannot beat unless you cheat.

而且一旦你作弊，就会被抓住并受到起诉。

And when you cheat, you get caught and get prosecuted.

所以市场方面，你首先得比它们更理性，资金也更充足，但大多数人根本不会从风险的角度去看待市场。

And so the markets have, you you have to be, first of all, more solvent than they are irrational, and most people don't even approach the markets from a risk perspective.

他们总是从贪婪的角度出发。

They approach it all the time from a greed perspective.

而现在大家都能用的一些工具，恰恰能帮你避免这些错误。

And, part of the tools that are now available to everyone keep you from making those mistakes.

比如，在做决定的时候，自我约束很重要，而不是凭情绪行事，或者临时想个计划。

Like, there is self governance, as much as anything in terms of making decisions when you're feeling emotional rather than coming up with a plan.

比如，我想在接下来的...

Like, I wanna accomplish this by the...

一年半到两年内实现这个目标。

In the next year and a half, two years.

然后设定目标，制定计划，回测这个计划，再部署它，让它一直为你工作。

And then setting a goal, coming up with a plan, back testing that plan, and then deploying it and letting it work on your behalf all the time.

我觉得这里缺失的关键是，有了全天候的监控，市场上会出现一些波动带来的机会，这些机会好到不容错过，尤其是在帮助人们实现目标方面。

And I think the key missing piece here is is that with that all the time or twenty four seven monitoring, there are volatility opportunities in the market that are too good to pass up as it pertains to getting people towards their goals.

比如，你见过比特币在一个月或两个月内涨幅超过40%的情况。

Like, you've seen a Bitcoin move of 40 plus percent in a month or two month period of time.

那是对成交量加权平均价格的大幅偏离，而这些机会通常就是大家追逐绿色蜡烛图的时候，而不是先锁定一些利润，确保……

That is a massive deviation off the volume weighted average price, And those opportunities are typically where everyone chases the green candle versus taking a little profits off the table, making sure that...

你知道，评估一下他们还有多少可用资金，以备价格下跌时进行后续交易，以及……

You know, assessing how much dry powder they have for subsequent trades if the price drops and how...

如果价格真的下跌，杠杆购买力会是什么样子，风险状况会如何变化，你知道，当比特币在65000美元时，和两三周后到了85000美元时，情况完全不同。

What that leverage purchasing power would look like if it did drop, what the risk profile change looked like, you know, when when Bitcoin was at 65,000 versus two, three weeks later, it's at 85,000.

从风险角度看，这改变了什么？更可能的情况是回撤到75000，还是继续上涨……

What what has that changed from a risk perspective, and what's the more likely event that it draws down to 75 or that it goes...

继续涨到95000？

Continues to 95?

这两者之间的概率因素是多少？

What's what's the probability factor that is between those two?

而且我觉得大多数人不是这样看待市场的，我觉得像这样的工具，以及我们在 Arch Public 正在打造的东西，就是为了这个目的，或者说应该朝着这个方向去设计。

And I don't think most people approach the markets that way, and I think tools like this and what we're building at Arch Public are geared to do that or they should be geared to do that.

嗯，这其实是你意志的一种体现，而且我……

Well, it's it's an acting out of your will and I...

在继续之前，我想感谢 Tillman 提到我大学时的外号，那个“颜值即正义”的性感称号。

Before I go further, I wanna thank Tillman for talking about my, you know, what I used to be called in college, sexy on face value.

但是，但是，这里必须有人为的输入。

But there there there is, you there has to be human input.

所以想法是，你懂的，让一个智能体去帮我赚钱，别犯任何错误。

So the idea is, you know, take an agent, go make money for me, don't make any mistakes.

他们会基于概率来做决策。

They're going to do things based on percentages.

就是这件事发生的概率，或者那件事发生的概率。

So the percentage that this will happen or that will happen.

而且在足够长的时间线上，说实话，任何概率，要么是对的，要么是错的。

And on a long enough timeline, any percentage, frankly, is probably right or probably wrong.

那要是你就让它随便跑，结果它做空了，会怎么样？

So what happens if you just let it run loose and it ends up in a short position?

但根据它们的数据，我早晚会有对的时候。

But based on their data, I'm going to be right at some point.

可与此同时，你亏掉的钱多得你心理上根本承受不了。

Meanwhile, you're losing enormous amounts of money that you can't personally stomach emotionally.

所以我们在Arch Public的做法是，先搞清楚你想达成什么目标，然后我们跟你一起把工具配置好，帮你实现它。

So what we do at Arch Public is, well, there's a lot we want to know what you want to get accomplished, and then we'll set our tools up with you to make that happen.

这样就不会有意外，不会出现智能体做空了、你却不知道会发生什么的情况，但它又笃定将来一定会对。

So there are no surprises so that an agent isn't taking a short position and you don't know what's gonna happen, but it's certain that in the future it's going to be right.

这简直就是灾难，到处都是隐患。

Like, that's disaster written all over it.

我们一开始就定了几个基本判断，其中一个就是，人们想要的是自己能掌控的自动化工具，由用户来驱动，能让他们直接说，我要这件事发生。

There were a few thesises that we started out with, at the very beginning of this, and and one of them was people want automated tools that they can control, that are user driven, that literally allow them to go, I want this to take place.

他们把它设置好，然后看着它发生，然后说，太好了，我做到了。

They program it and they watch it it take place and they go, yay, I did it.

这就是我们想给人们提供的东西的核心，而且我们希望通过OAuth 2和API集成，尽快把它交到他们手上。

That is the premise of what we're trying to give people, and we're trying to give it to them as quickly as it becomes available through OAuth two and API integrations.

但这完全是一种用户驱动的体验。

But it's a completely user driven experience.

你知道我会怎么形容它吗？它就像一个操作系统，让人们能够管理以前只是代码的东西，对吧？

And you know what I would call it as close to as an operating system that allows somebody to then manage what used to be just code, right?

在计算机的早期，如果你想拥有一台家用电脑，你往往得写很多代码。

In the old days of computer when we were, you know, you were relegated to a lot of coding if you wanted to have a home computer.

然后，所谓的操作系统把这些代码变成了图形界面，操作起来就变成了拖拽，对用户来说更直观了。

And then these things called operating systems made pictures out of the code and then it became more drag and drop and more intuitive for the user.

而自动化、人工智能以及所有这些，其实和我之前提到的Grock Bot没什么两样。

Well, automation and AI and all of these things no no different than what I mentioned earlier with Grock Bot.

它们被卡通化了。

They've cartoonized.

它们把以前那种没人知道怎么交互的长串代码，变成了形象化的东西。

They've they've characterized something that it used to be just this long string of code that nobody knew how to interact with.

现在呢，你只要把麦克风接到电脑上，跟它说话，它就能给出很聪明的内容，然后你可以把整个证据领域和文件里放的那些资料都拿来，作为做决策的依据。

And now, you know, you can just hook up a microphone to your computer and talk to it and it it spits out really intelligent things that you can then take in the entire, you know, field of evidence in the library of things that you're putting in your file to make decisions off of.

但最终，决策还是你自己做的。

But ultimately, they're your decisions.

行动也是你自己的，而且都在你的经纪账户里。

They're your actions, and they're in your brokerage accounts.

还有一点很重要，我觉得当你在很多交易所谈垂直整合时，他们说，嘿，把这些代理工具接进来，但这些工具对垂直整合网站里的工具一无所知，然后还指望它们能适应各种功能、不同的佣金结构等等，这根本不可能。

And that's another major point is like, I think when you're talking about vertical integration on a lot of these exchanges, they're saying, hey, hook up these agentic tools that have no knowledge of the tools that are inside of the vertical integrated, you know, website, and then they're expected to adapt across a wide array of feature sets and different commission schedules and all these things that are, you know, it's just impossible.

所以我们特别关注的一点就是尽可能多地集成平台，这样你就有选择去分散你的资金，按照你想要的方式控制资金的进出。

And so one of the things we really focused on is being integrated with as many platforms as possible so that you have the option to spread your capital out, control the capital in the way you want to move it in and out.

这是一种规避风险的方式来管理那些交易活动。

And that's a risk, you know, adverse way to manage those trading activities.

而且这也会促使竞争。

And it also forces, you know, competitive nature.

如果你是高净值人士，你就能看到所有交易所最好和最差的一面。

If you're a high net worth person, you get to kind of see the best and the worst of all the exchanges.

其中一个方面，我

One of the aspects that I

我一直在想的是，你知道，我们一直在做Sylvia这个项目，然后人们总是跟我说，比如我老婆会吃醋。

I keep coming back to is, you know, we've been building the Sylvia thing and and people will tell me all the time, like, my wife, it gets jealous.

我跟她在一起的时间太多了。

I spend too much time with her.

它……

It it...

在这种非常奇怪的关系里，几乎有一种人际互动的成分。

It's almost, in this very weird dynamic, there's the interpersonal component.

但还有，当人们去拿信息的时候，比如我会去跟他们的会计师聊。

But then also when people go and they'll take, like, information, I'll go talk to their CPA.

然后他们会说，哦，Sylvia告诉我的。

And they're like, oh, Sylvia told me.

他们谈论它的时候，就像在谈论一个人一样。

And they talk about it like it's a person.

所以，我也不知道。

And so, I don't know.

这真的是一个……

It it it is a...

这真的是一个非常奇怪的关系。

It is this very weird dynamic.

而且，你知道，Alexa，我觉得人们在家里对Alexa也会做同样的事情。

And, you know, Alexa, I think that people do the same thing with an Alexa in their home.

我觉得现在有像iRobot那样的清洁设备，或者，你知道，不管它们叫什么名字。

I think that there is now these like, kind of iRobot, you know, cleaning devices or, or, you know, whatever the, those things are called.

我觉得人们开始意识到，嘿，那个Matic或者，你知道，不管是什么，它有个名字。

I think people are starting to, you know, figure out, Hey, the Matic or, you know, whatever is, has got a name

还有性别。

and a gender.

比如她会做这个，你知道，她会做……

Like she she does this, you know, she does

百分之百？

A 100%?

你

You

你知道吗，有趣的是，我没有任何确凿的数据，但凭经验来说，我喜欢问别人你的智能助手叫什么名字？

you know what's interesting is, I don't have any hard data on it, but anecdotally, I like asking people what's the name of your agent?

我发现很多男人会给它取女人的名字，而很多女人会给它取男人的名字。

And what I find is that a lot of men will name it a woman's name, and a lot of women will name it a man's name.

我不知道为什么。

I don't know why.

我不明白，你知道吗，这是有科学依据还是只是我几个笨朋友凭经验告诉我的。

I don't understand, you know, is there like science or is that just like a couple of my dumb friends that, you know, anecdotally have told me that.

但确实感觉这里面有点东西，比如你在选名字，你赋予它某种身份，你以某种方式与它交流，现在它成了你生活的一部分。

But it does feel like that there's something there about, like, you are picking a name, you are giving it some sort of identity, and you are communicating with it in a way that now it becomes part of your life.

这对产品来说应该是好事。

That should be great for products.

这应该意味着你不会轻易流失用户。

Like, that should mean that you're not gonna churn out.

你不会想解雇你的朋友，对吧，那种感觉。

You don't wanna like fire your friend, you know, type activity.

但我觉得这还没被充分探索。

But I think it's just underexplored.

而且说实话，我们还不完全清楚这些东西会带来什么影响。

And and frankly, we don't really understand, you know, some of the ramifications of this stuff yet.

我觉得这更像是养孩子，你实际上是在训练它如何在你想要的各个方面满足你。

I think it's more akin to raising a kid from a, I mean, you're literally training it how how to meet you where you wanna be met on every front.

这就是我之前提到的，它就像……

That's why I was alluding to earlier, like, it being...

需要一个舞伴。

Needing a dancing partner.

它不知道你想怎么跳，直到你开始主导，开始和它一起跳。

It it doesn't know how you wanna dance until you start taking the lead and you start dancing with it.

然后你们俩就能一起成为专业舞者了。

And then you guys can become professional dancers together.

在那方面，人才没有上限，能跟得上你走多远。

There is no cap of talent on that side of the equation that can meet you as far as you can take it.

但我还是觉得，有一个上限。

But there is still, I think, a cap.

毫无疑问。

Like, there is no doubt.

我见过一些人，甚至在我们公司里，有些人特别擅长训练AI代理。

I've seen some people, you know, even within our company, some people are exceptionally proficient at training the the AI agents.

而且你花很少的时间训练，就能在生产力上取得非常大的飞跃。

And and you get, you know, very large leaps forward in productivity with very little, time training.

然后还有一些人，他们需要多花点功夫，因为他们很难表达清楚自己想要什么。

And then there's some that it it takes a little bit more effort because it's harder for them to articulate what they want out of it.

我觉得那才是最终的瓶颈，而且这挺奇怪的。

And I think that's really the the ultimate limiting factor, and it is weird.

我提到它的时候，确实会切换性别。

I do switch sexes when I refer to it.

有时候我会叫它“她”，有时候叫它“他”。

Sometimes I'll call it a she, and sometimes I call it a he.

但是……

And but...

这就像……

It's a...

它会……

It's gonna...

当你训练它十年之后，界限会变得更加模糊，它几乎就是你意志的延伸，在很多方面都一模一样。

The lines are gonna continue to be even more blurred when you've trained it for ten years, and it really is an extension of your will almost, you know, identically on on a lot of different fronts.

想想看，就像旅行代理人一样。

I mean, just think about as a travel agent.

它对你的偏好了如指掌。

It knows your preferences down to a t.

比如说，不……

Like, no...

没有人能替代，你知道，作为人类的那种感觉，而且它会设定一些不切实际的标准，我告诉你。

Nobody can replace, you know, that, as a human, and it's gonna set some unrealistic standards, I will tell you.

这里……

There...

它……

It...

有意思的是，Facebook目前增长最快的产品之一就是他们的AI代理Muse，M-U-S-E。

It's interesting that one of Facebook's, their fastest growing product right now is their AI agent Muse, m u s e.

人们疯狂下载这个。

People are downloading that like crazy.

就像，这发生得非常非常快。

Like, it's happening very, very fast.

而且这是……

And it's the...

再说，部分原因就是界面很友好。

Again, part of the reason why is there's a friendly interface.

你知道，有个胖乎乎的小家伙，就是Muse的象征和头像，普通人开始越来越习惯用这个了。

You know, there's a little puffy little guy that's the the muse sort of symbol and avatar, and, you know, regular folks are starting to get very, very comfortable with with doing this.

所以不只是放在你的设备上，我记得昨天扎克伯格还宣布了，就是一个小圆环之类的东西，你可以戴在手上或者随身带着。

So not only having it on your devices, but I think yesterday, Zuckerberg announced, you know, just like a little, you know, a little circular deal that you either wear or keep you with you.

那就是你的Muse，你的AI伙伴，走到哪儿带到哪儿。

That's your muse, you know, AI buddy that you take around everywhere.

所以就算你不在家，手机电脑都不在身边，你也有这个小Muse，那个软软的、胖胖的小家伙，帮你处理事情。

So if you're disconnected from your phone or your computer at home or wherever you are, you've got this this little muse, you know, puffy squishy dude that, is is is handling stuff for you.

再说，从概念上讲挺有意思的，但当这个概念变成大规模应用，在Facebook这个级别有大量下载的时候。

Again, it's it's interesting to talk about from a concept, but when that concept turns into big time uses, big time downloads at the Facebook level.

这跟我是某家公司的开发者，有几台机器跑着Claude之类的，完全不一样了。

This is very different than I'm a developer at x y z company, and I've got a couple machines that are running claud and other stuff.

我们现在到了Facebook这个级别，Facebook的几百万用户都是普通人，美国中产阶级或者哪儿的中产阶级，他们发帖说，我不喜欢街那头那个Sandy，因为她搞的庭院旧货出售，你知道，就是...

We're now at the Facebook level where Facebook's millions of users are your run of the mill Tom, Dick, and Harry type people, Middle America or middle wherever they live, you know, posting about, you know, I don't like, you know, Sandy that lives down the street because the garage sale that she did, you know, was...