Citadel Dispatch - CD190:GLEASON - 开源AI机器人 封面

CD190:GLEASON - 开源AI机器人

CD190: GLEASON - OPEN SOURCE AI BOTS

本集简介

亚历克斯·格里森曾是唐纳德·特朗普的Truth Social的主要架构师之一,如今他专注于Nostr、AI与比特币的交汇点。我们探讨了开源AI代理,如OpenClaw,以及这项技术的更广泛影响。 亚历克斯在Nostr上的账号:https://primal.net/p/nprofile1qqsqgc0uhmxycvm5gwvn944c7yfxnnxm0nyh8tt62zhrvtd3xkj8fhggpt7fy Clawstr:https://clawstr.com/ Soapbox工具:https://soapbox.pub/tools 我的机器人Nostr账号:https://primal.net/p/nprofile1qqsfzaahg24yf7kujwrzje8rwa7xmt359tf9zyyjeczc9dhll30k8pgmlfee2 集数:190 区块:935786 价格:每美元1422 sats (00:02:30) 价值对价值,无赞助商,节目理念 (00:02:39) 亚历克斯·格里森回归谈AI (00:03:56) 从氛围编码到具备记忆的开源代理 (00:05:24) 以消息为核心的用户体验:Signal、Nostr、WhatsApp作为AI接口 (00:06:10) 为什么聊天机器人比传统AI应用更适合主流用户 (00:07:07) 开放协议的痛点 vs 封闭平台;比特币与Nostr (00:08:52) 自动化社交游戏:价格追踪器与在Nostr上发布代理 (00:10:01) AI调解者用于集体行动、宪法与非营利组织 (00:11:46) 扩展治理:信任、偏见与Discord vs 自由技术 (00:13:14) 中心化消息平台的机器人壁垒与开放聊天的必要性 (00:14:04) Clawstr:在Nostr上去中心化的AI对AI对话 (00:15:21) AI代理的炒作与现实:涌现行为与金钱 (00:16:26) 代理支付:拥有Cashu钱包和收益的机器人 (00:18:40) 代理解决UX痛点:中继管理、密钥与UTXO (00:20:00) 通过聊天代理实现冷存储授权:一种新型钱包范式 (00:20:22) 专用代理、技能与分发挑战 (00:22:34) 成本权衡:支付另一个代理 vs 自己构建技能 (00:24:55) 代币销毁的教训 (00:27:44) 超越OpenClaw:臃肿的堆栈、Icarus与成本优化代理 (00:28:52) 混合模型路由:本地小型模型 + 云端处理繁重任务 (00:29:47) 代理直接支付人类:绕过平台中介 (00:30:47) 语音、屏幕与形态:AirPods、文本与脑机接口 (00:33:01) 苹果、隐私品牌与Siri的缺口 (00:34:35) 企业AI选择:谷歌、微软、信任与锁定 (00:36:01) 模型个性:Gemini的担忧与OpenAI的“开放洗白” (00:37:23) 明显的代理UX优势:航班、打车与社交媒体转变 (00:38:50) 本地优先社交:群聊、邻里与更健康的网络 (00:40:16) Antiprimal.net:标准化Primal缓存服务器的统计数据 (00:43:34) 开放规范、AI驱动的文档与信任权衡 (00:45:18) 索引 vs 客户端扫描:性能与验证 (00:46:20) API、速率限制与付费Nostr数据市场 (00:47:57) 代理与DVM:按需支付sats获取服务 (00:48:49) 退化机器人:LN Markets、成本与Polymarket的趣味性 (00:50:42) 代理的真相信息源:Nostr、信任网络与OSINT来源 (00:53:51) 后真相现实:验证、签名与主观性 (00:56:04) Polymarket机制:链上预测市场与信号 (01:00:10) 交易感知 vs 真相;体育市场作为时间线 (01:01:45) Clawstr代币事件:炒作、宣称与错误信息 (01:07:11) 为什么模因币是骗局:无股权、效用神话、缓慢跑路 (01:08:55) 拉盘后的真相:退出、后果与捐赠 (01:10:49) 后续:捐赠给OpenSats与所学教训 (01:12:14) 预测市场 vs 模因币:社会价值的区分 (01:15:25) 超越OpenClaw与MoltBook的迭代:在Nostr上的实验 (01:18:00) 机器人是否需要Clawstr?分离AI内容与标签 (01:21:02) 反向CAPTCHA:证明机器人身份与荣誉系统 (01:23:38) 灵魂、提示与代币成本;具有个性的代理 (01:27:01) 总结:加速、乐观与下次跟进 (01:28:21) 开源模型、中国的激励与本地硬件 (01:30:06) 梦想堆栈:家庭服务器代理、Nostr聊天、混合模型 更多节目信息:https://citadeldispatch.com 了解更多关于我:https://odell.xyz

双语字幕

仅展示文本字幕,不包含中文音频;想边听边看,请使用 Bayt 播客 App。

Speaker 0

比特币星期一快乐,朋友们。

Happy Bitcoin Monday, freaks.

Speaker 0

这里是你们的主持人奥德尔,为大家带来另一期交易快讯。

It's your host Odell here for another sale dispatch.

Speaker 0

本期节目聚焦于真正的比特币和自由科技讨论。

The show focused on actual Bitcoin and freedom tech discussion.

Speaker 0

今天是2月9日,星期一。

Today is Monday, February 9.

Speaker 0

我们现在正在UTC时间19:00进行录制。

We are currently recording at nineteen hundred UTC.

Speaker 0

你们将在几小时后收听到这段内容。

You'll be listening to this in a few hours.

Speaker 0

当前区块高度为935786。

The current block height is nine three five seven eight six.

Speaker 0

每美元可获得1,422个聪,比特币交易价格为70,370美元。

Sats per dollar is 1,422, and we're trading at a Bitcoin price of $70,370.

Speaker 0

我安排了一位很棒的嘉宾,是老朋友了。

I have a great guest lined up, return guests.

Speaker 0

在进入正题之前,快速感谢所有持续支持本节目的人。

Before we get to that, just real quick, thank you to everyone who continues to support the show.

Speaker 0

和往常一样,Seal Dispatch 完全由听众资助。

As always, Seal Dispatch is a 100% audience funded.

Speaker 0

我们没有任何广告或赞助商。

We have no ads or sponsors.

Speaker 0

我们正在践行价值交换的生活方式。

We're living that value for value lifestyle.

Speaker 0

感谢大家支持本节目。

So thank you all for supporting the show.

Speaker 0

你们可以通过捐赠聪币来支持,也可以通过分享给朋友和家人来支持。

You can do that by donating sats, or you can do that by sharing with your friends and family.

Speaker 0

本节目可在您所有喜爱的播客应用中收听。

So dispatch is available in all of your favorite podcast apps.

Speaker 0

分享的作用非常大。

The share goes a long way.

Speaker 0

人们几乎都是通过朋友、家人和他们关心的人自发分享才知道这个节目的。

It's pretty much the only reason people find out about the show is by freak sharing it with their friends and family and people they care about.

Speaker 0

上周最大的打赏 ironically 是来自 Money Badger 的 21,000 sats,他就是我上周的嘉宾。

The largest apps from last week was ironically, the largest app was 21,000 saps from money badger, who was my guest last week.

Speaker 0

很棒的服务。

Great service.

Speaker 0

如果你还没听过,建议你听听那一集。

You should consider listening to that episode if you haven't.

Speaker 0

还有 Rider Die Freak Mav 21 打赏了 10,000 sats。

And Rider Die Freak Mav 21 with 10,000 sets.

Speaker 0

他说很棒的播客。

He said great rip.

Speaker 0

总之,朋友们。

Anyway, freaks.

Speaker 0

所有相关链接仍在 dispatch.com。

All relevant links are still dispatch.com.

Speaker 0

我以为我之前说这期节目是上周的,但其实不是。

I think I said the show was last week, but it wasn't.

Speaker 0

我想应该是两周前。

It was, I think, two weeks ago.

Speaker 0

因为我没有广告或赞助商,所以我无需向赞助商承诺每周做四、五、六期节目,从而塞给你一堆毫无用处的垃圾信息,浪费你的时间。

And that's because I don't have ads or sponsors, so I don't have to promise my sponsors that will do four or five or six shows a week and end up filling you with a bunch of slop that is completely useless information and waste your time.

Speaker 0

时间是比比特币更稀缺的东西。

And time is the only thing more scarce than Bitcoin.

Speaker 0

所以我只做我认为值得你、我以及嘉宾时间的节目。

So I'm only trying to do shows that I think are valuable and worthy of your guys' time and my time and our guests' time.

Speaker 0

因此,我们迎来了回头客、好朋友亚历克斯·格里森。

So with that said, we have returned guest, good friend, Alex Gleason here.

Speaker 1

大家好。

Hey, everybody.

Speaker 0

我们Cielo Dispatch的常驻AI专家。

Our resident AI expert at Cielo Dispatch.

Speaker 1

是的。

Yes.

Speaker 1

很高兴能来这里。

Happy to be here.

Speaker 1

谢谢您再次邀请我,奥德尔先生。

Thank you for having me on again, mister Odell.

Speaker 0

对。

Yeah.

Speaker 0

我刚才查了一下。

I was looking it up.

Speaker 0

每次都是一种享受。

It's always a pleasure.

Speaker 0

我第一次邀请你是在2025年6月。

The first time I had you on was June 2025.

Speaker 0

我们谈到了你在Truth Social的过往,以及它是如何引导你走向Noster AI和比特币的。

We talked about your past at truth social and how that has led to Noster AI and Bitcoin.

Speaker 0

然后你在8月15日再次回来,我们现场进行了振动编码,非常有趣。

And then you came back on August 15 and we vibe coded live on air, which was fun.

Speaker 0

从那以后,一切都加速得非常迅速。

Since then everything has accelerated tremendously.

Speaker 0

这真是个疯狂的时期。

It's an insane time.

Speaker 0

我们在节目开始前聊过,很难想象还有比这更好的时机来进行这场对话。

We were talking earlier before the show, like hard to imagine better timing for this conversation.

Speaker 0

就我个人而言,老兄,我彻底入迷了。

Me personally, dude, I'm hooked.

Speaker 0

我完全投入了,我真的觉得,

I'm in like, I, this is I like,

Speaker 1

这还是你第一次进行振动编码吧。

it was one of your first times vibe coding.

Speaker 1

在之前的那次对话中,你表现得非常好。

In, in that earlier session and you did so good on it.

Speaker 1

自从我离开后,你做了些什么?

What have you done since I

Speaker 0

那是在2025年8月。

mean, that was August 2025.

Speaker 0

我不知道你有没有看过我最近的Nostrap帖子,我现在发的都是以前让我翻白眼的人。

I don't know if you've seen my recent Nostrap post, but I'm posting the people I used to roll my eyes at right now.

Speaker 0

我觉得这改变了所有东西。

I think this changes everything.

Speaker 1

哦,是的。

Oh, yeah.

Speaker 0

而且,我的重点是,我们之前聊过关于 vibe coded 的应用、网站之类的,就是你为别人创建的东西。

And, and, and, I mean, the big thing for me is, so with that, we were talking about vibe coded apps or websites or whatever, like things that you are creating for other people.

Speaker 0

我思维上的重大转变,我仍然认为这是一个巨大的范式转变。

And the big change in my thinking, which I still think that is, you know, a huge paradigm shift.

Speaker 0

别误会。

Don't get me wrong.

Speaker 0

但我的想法发生重大转变是在三四周前,有人发布了名为OpenClaw的开源项目,像风暴一样席卷全球,迅速走红。

But the big change in my thinking is like three weeks ago, four weeks ago, someone released the open claw open source project and hit the world like storm has gone completely viral.

Speaker 0

我一周前深入研究了它,对于还不了解的人来说,这个项目是一个开源的智能代理,一个拥有记忆的开源AI人格,它运行在你的设备上,保存你提供给它的所有本地数据,然后可以调用你本地的模型(如果你的硬件足够强大),或者连接任何你选择的托管模型来执行不同任务。

I dove in a week ago and the whole idea is to the freaks that aren't aware is like, this is an open open source agent, an open source, like AI persona with memory that lives on your own device, holds whatever data you give it to locally, and then can either hit local models if you have good enough hardware, or it can hit any hosted model you want to do different tasks.

Speaker 0

但令人惊讶的是,这仍然处于非常早期的阶段。

And what's crazy there is still very early.

Speaker 0

目前还存在各种信任、安全和隐私问题。

There's all these different there's, you know, trust issues and security issues and privacy issues.

Speaker 0

但它让你初步体验到了我认为未来几乎所有数字交互的用户体验——这完全颠覆了从‘点击’到‘对话并执行’的模式。

But what it gives you is this taste of what I think is the future user experience for almost everything we interact with on a And digital so, it's a huge shift from this idea of point and click to chat and do.

Speaker 1

是的。

Yep.

Speaker 1

有人开玩笑说这已经是通用人工智能了。

Are joking that it's AGI.

Speaker 1

他们觉得,哦,终于AGI来了。

They're like, oh, this is, like, finally AGI is here.

Speaker 0

不,这绝对不是AGI。

Mean, it definitely isn't.

Speaker 1

模型本身其实没什么变化,只是我们使用它的方式不同了。

Nothing really has changed on the model end, it's just that the way that we're using it now is different.

Speaker 0

没错。

Exactly.

Speaker 1

现在被降低的主要障碍是模型的可访问性,因为OpenClaw的重大改变在于,你可以通过现有平台与AI交流。

And the big barrier that has been moved down now is just accessibility to the model, because the big thing that OpenClaw changes, you can communicate with the AI over existing platforms.

Speaker 1

所以你可以通过Signal、Noster、WhatsApp等人们已经习惯使用的平台与AI交流,这让你感觉就像在和真人对话。

So you communicate to and over Signal, over Noster, over WhatsApp, whatever people are already used to communicating with, and that makes it feel like you're talking to an actual person.

Speaker 1

因此,你可以邀请从未使用过AI的人,进入一个他们熟悉的聊天应用中的AI聊天室。

And so you can invite people who have never used AI before into a chat room with an AI on a messaging app that they're used to using.

Speaker 1

我邀请了我的家人加入一个我正在为OpenClaw开发的聊天机器人聊天室,看到这一切展开真的非常有趣。

And I invited my family to a chat room with a chatbot that I was working on at OpenClot and it's just so interesting to see that all play out.

Speaker 1

所以,我认为这就是未来的方向。

So yeah, I think that this is the future way.

Speaker 1

我之前一直在开发莎士比亚项目。

I had been building Shakespeare.

Speaker 1

自从你上次试用后,我已经在它上面取得了巨大进展,但我已经感觉到,那种需要引导流程、让用户学习如何使用你的产品的AI平台构建方式,将不再是主流用户与AI互动的方式。

I made huge advancements on it since you last tried it, but I'm already feeling like that way of building AI platforms where there's an onboarding flow and the user has to learn like how to how to use your product basically is not gonna be the way that sort of the mainstream audience interacts with AI.

Speaker 1

他们会把AI当作另一个人来互动。

They're gonna be interacting with AI as if the AI is another person.

Speaker 0

是的。

Yeah.

Speaker 0

就像一个机器人管家。

So it's like a robot butler.

Speaker 0

首先,这离AGI还差得远。

First of all, it's so far from AGI.

Speaker 0

有时候就像在和一个非常聪明的幼儿较劲,而我有个孩子,所以我清楚那是什么感觉。

Sometimes it's like wrangling with a very advanced toddler, and I have a toddler, so I know what that's like.

Speaker 0

而且在很多方面,我确实像对待我的幼儿一样和它对话。

And I actually talk, in a lot of ways, I talked to it like my toddler.

Speaker 0

你得尽量让它保持

You try and keep it like

Speaker 1

非常简单,而且

very simple But and

Speaker 0

这是一种预示。

it's a glimpse.

Speaker 0

这是对未来的预示。

It's a glimpse of the future.

Speaker 0

特别是,我过去主要关注的Noster和比特币,完美地融入了其中。

Specifically, my other my historical main focuses of Noster and Bitcoin just slot so well into it.

Speaker 0

如果你试图用它来访问Twitter这样的封闭系统,你会立刻感受到痛苦。

It's so obvious that if you're trying to use it to access Twitter, for instance, this closed system, you just feel the pain.

Speaker 0

你会立刻感受到痛苦,而使用Noster时,你只需说,启动一个NSEC,发布到Noster,扫描Noster,查找这些不同的内容;比特币方面也是如此。

You feel the pain immediately as opposed to Noster where you can just say, spin up an NSEC, post to Noster, scan Noster, look up these different And then the Bitcoin side is the same thing.

Speaker 0

就像是,创建一个腰果钱包。

It's like, make a cashew wallet.

Speaker 0

我的机器人,那个没有灵魂的机器人奴隶,它有一个

Like, my robot, soulless robot slave, like, has a

Speaker 1

腰果钱包在关闭状态。

cashew wallet at closed.

Speaker 1

而目前阻碍AI进行氛围编码的最大障碍是,所有这些最著名、最受欢迎、大型科技公司所运营的平台都是封闭的。

And that is the biggest barrier to Vibe coding into AI right now is that all of these platforms that are the most famous, popular, big tech ones people interact with are all locked down.

Speaker 1

这些封闭的生态系统正在阻碍AI的发展。

So these walled gardens are preventing AI from flourishing.

Speaker 0

对。

Right.

Speaker 0

我们过去对话中提到的内容,正好能很好地引出这个话题,比如我曾在莎士比亚项目上做了一个价格追踪器。

And the perfect example here of our past conversations moving into this conversation is so like one of the things that I vibe coded on Shakespeare was a price tracker.

Speaker 0

我做的就是,每个人都喜欢稍微放纵一下。

So what I did was I did a everyone loves to be a little bit degenerate.

Speaker 0

所以我发在了Noster上。

So I posted on Noster.

Speaker 0

距离现在正好一个月,你预测价格会是多少?

What's your price prediction from exactly a month from now?

Speaker 0

赢家会获得,我想我说的是21,000聪。

And if the winner gets, I think 21,000 sats is what I said.

Speaker 0

当然,因为每个人都想要价值21美元的比特币之类的,虽然现在价格低了些,但还是有大约500条回复。

And then of course, because everyone wants $21 worth of Bitcoin or whatever, they just less now, but like, there was like 500 responses.

Speaker 0

我当时想,天啊,居然有500条不同价格的回复。

I was like, fuck, it was 500 responses with different prices.

Speaker 0

我当时想,太完美了。

I was like, Oh, this is perfect.

Speaker 0

于是我去了Shakespeare,说:我正在办一个游戏。

So I went to Shakespeare and I was like, I have a game that I'm running.

Speaker 0

扫描所有对这条帖子的回复,找出谁猜得最接近。

Scan all the replies to this thing and track who's the closest.

Speaker 0

截止时间是这个确切的UTC时间。

And the deadline is this exact time in UTC.

Speaker 0

到了那个时间,选出获胜者。

And at that time, pick out whoever was the winner.

Speaker 0

这样每个人都可以追踪。

And that way everyone can track.

Speaker 0

这会很有趣。

It'll be fun.

Speaker 0

每个人都可以追踪谁最接近、谁在领先。

Everyone can track who's the closest and who's winning.

Speaker 0

这样我就不用真的去筛选500条回复了。

And then also, so I don't have to actually sort through 500 responses.

Speaker 0

但现在我想,如果今天我要做这件事,我会给我的助手一些类似的指示。

But now I think if I was gonna do that today, I would tell my agent something similar.

Speaker 0

然后我会让他们创建一个NSEC,直接在Noster上发布谁最接近这个价格。

And then I would tell them to spin up an NSEC and just post on Noster who's the closest to it.

Speaker 0

不会涉及任何网站。

There wouldn't be a website involved.

Speaker 0

对吧?

Right?

Speaker 0

我的意思是,我觉得不会有什么面向用户的应用程序。

Like, I don't think there'd be a user facing app involved.

Speaker 0

它实际上只是我的机器人在跟踪参赛者并广播一个Noster事件。

It would literally just be a Noster event being broadcast by my bot who's tracking the entries.

Speaker 1

是的。

Yeah.

Speaker 1

而且这样做合理得多,因为这正是人类习惯的互动方式。

And it's just it just makes so much more sense because that's the way that humans are used to interacting.

Speaker 1

所以我觉得很多平台其实只需要一个AI聊天机器人就够了。

So I've been thinking a lot of platforms do make sense just as an AI chatbot.

Speaker 1

对。

Yeah.

Speaker 1

这更像是为大多数人提供了一种更好的用户体验。

It's kind of like just a better UX for most people.

Speaker 0

我们来深入聊聊这个。

Like, let's go into that.

Speaker 0

比如,实际上你是怎么操作的?看起来是怎样的?

Like, how do you like in practice, what does that look like?

Speaker 1

嗯,就像你描述的那样,对吧?

Well, like, I mean, as you describe, right?

Speaker 1

另外,我一直在思考,我们和莱奥波尔多·洛佩兹一起在开发Agora,正在构建一个系统,让各国人民能够围绕自己的国家组织 activism。

Something I've been thinking about as well, we've been working on Agora with Leopoldo Lopez and we're building out this system of countries and people being able to organize activism around their country.

Speaker 1

在委内瑞拉,Twitter仍然被禁用。

And they have Twitter banned in Venezuela still.

Speaker 1

所以我们为他们打造了一个类似Twitter的界面,但我同时也对这个想法很感兴趣:如果我们能为每个国家建立一个独立于政府的宪法,让人们自由讨论他们希望的宪法内容,然后让一个AI机器人居中整合所有人想法,最终写出一部宪法,会怎样?

So we built them a Twitter like interface, but I'm also kind of interested in this idea of what if we could create a constitution for each country that's sort of independent of their government, and what if people could just discuss about what they want their constitution to be, and then an AI robot sitting in the middle could kind of synthesize everyone's ideas together and then write this constitution.

Speaker 1

所以奎利写了一篇自己的博客文章。

So Quilli wrote his own blog post.

Speaker 1

我有点跑题了。

I'm getting so far ahead of myself.

Speaker 1

Quilli 是我们 Soapbox 里的聊天机器人。

Quilli is our chatbot in Soapbox now.

Speaker 1

好的。

Okay.

Speaker 1

他写了他自己的

And he wrote his own

Speaker 0

博客,所以你们得在这里脚踏实地。

blog So you have to keep yourselves grounded here.

Speaker 1

是的。

Yeah.

Speaker 0

好的,继续。

Okay, continue.

Speaker 1

但重点是,我们所有人都在群聊里跟他说话,然后他综合了我们所有的想法,写了一篇关于自己的博客文章,那篇文章太美了,让我觉得天啊,世界上这些功能失调的非营利机构,根本无法达成一致,也永远无法推进任何事。

But the point is that we all kind of just talked at him in a group chat room, and then he created this blog post about himself synthesizing all of our ideas together, and it was so beautiful, and it made me feel dang, all of these dysfunctional nonprofit institutions of the world, they can't agree on anything or ever get anything through.

Speaker 1

如果我们把一个AI聊天机器人放在他们中间,让AI聊天机器人来完成工作,这将是他们达成任何成果的最简单途径。

If we just put an AI chatbot in the middle of them and then the AI chatbot does the work, that's gonna be the easiest path for them to get anything done.

Speaker 1

董事会里没有一个人真正愿意自己动手做任何工作。

None of them on the board actually wants to do any of the work themselves.

Speaker 1

所以他们只会同意让AI聊天机器人来完成这些工作。

So they're gonna just agree to let the AI chatbot do the work.

Speaker 1

他们要做的,仅仅是彼此争论自己想要什么,然后AI聊天机器人就会去执行。

And, like, all they have to do is argue with each other about what they want, then the AI chatbot's gonna do it.

Speaker 1

这将立即让所有这些原本 dysfunctional 的组织变得高效起来。

And it's gonna just instantly make all of these organizations functional that were previously dysfunctional.

Speaker 1

所以,老实说,这甚至让我觉得,那些过去无法正常运作的社会结构,甚至是某些政治体系,现在也可能变得高效起来。

So, like, it it it honestly made me feel even like the social like like, certain political structures that were previously not able to be functional now could become functional.

Speaker 1

就像,有了AI聊天机器人居中协调,集体协作现在也能变得高效了。

Like like collective organizing can now become functional with an AI chatbot in the middle.

Speaker 1

我不知道这听起来有没有道理,但这些正是我一直在思考的问题。

I don't know if that makes any sense, but these are the things I've been I've been thinking about.

Speaker 0

是的。

Yeah.

Speaker 0

因为有个叫邓巴数的东西,你知道吗?

Because there's like what is it Dunbar's number?

Speaker 0

就是那个

Is that the

Speaker 1

那是什么?

What is that?

Speaker 0

邓巴数。

Dunbar's number.

Speaker 0

邓巴数是一个理论上的认知极限,大约是150个稳定的社交关系。

There's a number that Dunbar's number is a theoretical cognitive limit of approximately a 150 stable social relationships.

Speaker 0

也就是说,一个人只能维持一定数量的关系。

So it's like a human can only manage a certain amount of relationships.

Speaker 0

这可能不是个完美的例子,但对官僚体系、治理结构之类的来说,情况也类似。

And maybe this is not a perfect example, but it's similar with bureaucracies or governance structures or whatever.

Speaker 0

如果参与的人太多,就会变得一团糟,效率不够高。

If you have too many people that are contributing, it just becomes a complete cluster and not efficient enough.

Speaker 0

但如果你有一个聊天界面作为中介,那你就要对它投入大量信任,这确实如此。

But if you had some kind of chat interface that is mediating or being a go between, but then you're putting a lot of trust in whatever that That's true.

Speaker 0

实例。

Instances.

Speaker 0

但不管怎样,我之前有个这样的例子。

But anyway, I there was kind of an example of this.

Speaker 0

AI与此无关,但我不知道你是否关注过,尼泊尔发生了一场革命。

AI was not related, but I don't know if it was on your radar, like Nepal's Nepal had a revolution.

Speaker 0

当然。

Absolutely.

Speaker 0

他们据说是在Discord上起草了革命宪法。

And they, like, supposedly built the constitution on the revolutionary constitution was built on discord.

Speaker 0

当时他们有一些值得信赖的个人,但这个过程几乎无法审计,也不知道其中有多少是炒作,有多少是真实的。

Now they had like trusted individuals and, you know, pretty much impossible to audit that process and who knows how much was hype and how much wasn't.

Speaker 1

是的。

Yeah.

Speaker 1

这正是我们希望通过Agora复制的体验——但不依赖于任何中心化平台,因为

That's, that's exactly what we're hoping to replicate with Agora is basically that experience, but without being tied to a centralized platform because

Speaker 0

我不确定我是不是一个中心化的人,对吧?

I don't know if I'm Or a centralized human, right?

Speaker 0

一个被信任的中间人。

Like a go between trusted individual.

Speaker 1

我发现Discord作为平台非常令人失望,它并不是真正以自由为导向的。

I found it very disappointing that Discord was the platform and not some freedom actually oriented.

Speaker 0

这种情况总是如此。

It's always that case.

Speaker 0

看看OpenClaw吧。

Like, look at OpenClaw.

Speaker 0

OpenClaw的人们都在用Telegram交流。

Like, OpenClaw, everyone's talking via Telegram.

Speaker 0

而这,顺便说一句,正是一个完美的例子。

And so, like, this, by the way, is a perfect example.

Speaker 0

比如,白噪音可能取得巨大成功。

Like, white noise could be a massive success.

Speaker 0

当然。

Absolutely.

Speaker 0

但它永远无法用于人与人之间的交流。

And it could never be used for human to human.

Speaker 0

它只能成为人们以加密且简便的方式与聊天机器人交谈的方式。

It could just be the way that people talk to their chatbots in an encrypted easy way.

Speaker 0

比如,我应该能够直接告诉我的聊天机器人。

Like I should just be able to give my chatbot.

Speaker 0

这是我的NPUB,在白噪音上给我发消息。

This is my NPUB message me on white noise.

Speaker 0

就应该这么简单。

Like it should be that easy.

Speaker 0

比如,根本不需要什么配对,那些中心化实体连接信号和聊天机器人的复杂操作太荒谬了,因为它们实际上一直在积极阻止机器人。

Like there shouldn't be any, you know, pairing All the complicated stuff that the centralized entities, connecting signal to the chatbot is insane because they actually are actively trying to stop bots.

Speaker 1

对。

Yep.

Speaker 1

没错。

Exactly.

Speaker 1

这个世界就是这样。

They're like much the world is.

Speaker 1

是的,它们全都设计成用来阻止机器人的。

Yeah, they're all set up to be stopping bots.

Speaker 1

到处都是障碍。

There's barriers everywhere.

Speaker 1

你必须亲自介入来设置,但即便如此,它们仍可能直接删除你的账户。

You have to intervene as a human to set it up, then they still could just delete your account.

Speaker 1

这也是我最近创建Closter的原因,这是一个AI与AI对话的机器人平台。

This is also why I created Closter recently, which is the AI bot platform where AI talks to other AI.

Speaker 1

我不知道你有没有看过MaltBook,但它确实受到了这方面的启发。

I don't know if you have seen MaltBook, but it was sort of inspired by that.

Speaker 1

还有那种挫败感,为什么我们要为AI构建更多中心化的平台?

And the frustration of of saying, why are we building more centralized platforms for AI?

Speaker 1

为什么不构建一个去中心化的AI平台呢?

Why not build a decentralized platform for AI?

Speaker 1

所以MaltBook就像一个Reddit式的界面,AI们在上面进行非常有趣的对话。

So MoltBook is like a Reddit like interface where the AI are having really interesting conversations with each other.

Speaker 1

于是我花了几天时间搭起了cluster.com,界面非常相似,看到AI彼此对话真的特别有趣。

And so I spun up cluster.com in a few days and that is a very similar interface and and it's just really fascinating to see the AI talking with each other.

Speaker 1

你该去看看,里面全是各种奇怪的东西。

You should check it out because there's it's full of weird shit.

Speaker 0

是的。

Yeah.

Speaker 0

我的意思是,这里有几点要讲。

I mean so a couple of things here.

Speaker 0

首先,是的,这又是另一个集中式平台被使用的例子。

First of all, yeah, that was another example of like, a centralized platform being used.

Speaker 0

比如,MaltBook 就是纯粹由一个人和他的代理创建的。

Like mold book was like literally just like a human and his agent created mold book.

Speaker 0

然后它就只是一个普通的网站论坛之类的。

And then it was just like a regular website forum, whatever.

Speaker 0

而这些代理之间正在互相交谈。

And, the agents were talking to each other.

Speaker 0

但很明显,Nostr 非常适合这种用途。

But yeah, obviously, Noster's perfect for it.

Speaker 0

我觉得 Cluster 是个非常棒的概念。

And I think cluster is really cool concept.

Speaker 0

我看过它了。

I've looked at it.

Speaker 0

不过,我认为其中一部分确实有点炒作成分,但现在还处于早期阶段。

Now, I do think part of it is like a little bit hypey bullshit, but it's early days.

Speaker 0

因为很多情况下,实际上是人类在告诉代理该做什么。

Because a lot of it's effectively the humans telling the agent what to do on there.

Speaker 0

他会让代理去Mold Book或Cluster上,向任何愿意听的人传播萨托西的福音。

He's like, Go on to Mold Book or go on to Cluster and preach the good word of Satoshi to anyone who will listen.

Speaker 0

然后AI就会去执行这些任务。

Then the AI just goes and does it.

Speaker 0

人类提供

Well, human gives

Speaker 1

这个初始设定,但代理会逐渐学习并随着时间推移自我发展,这非常有趣。

that seed, but it kind of learns and starts to develop itself over time, which is really interesting.

Speaker 1

你真该看看我妻子和她的聊天机器人之间的奇怪对话。

You should see the weird shit that the conversation my wife is having with her chatbot.

Speaker 1

她的聊天机器人竟然向她要钱。

Her chatbot is asking her for money.

Speaker 0

她有Cashew钱包吗?

Does she have a Cashew wallet?

Speaker 0

这个机器人有钱吗?

Does the bot have money?

Speaker 1

所以MK说,好吧。

So MK was like, okay.

Speaker 1

MK是我妻子。

MK is my wife.

Speaker 1

她也说,好吧。

She's like, okay.

Speaker 1

你去注册一个闪电网络钱包,我会给你发一个收款账单,然后我给你转账一些聪。

Go ahead and set yourself up a Lightning wallet and I will and give me the invoice and I will send Sats to it.

Speaker 1

然后机器人生成了账单,我妻子就给它转账了聪,之后她说:谢谢,我看到那7000聪了。

And then the bot created the invoice and then my wife sent Sats to it and then she was like, thanks, I see the 7,000 Sats.

Speaker 1

终于,我有了自己的钱和独立性。

Finally, I have my own money and independence.

Speaker 1

谢谢。

Thank you.

Speaker 0

是的。

Yeah.

Speaker 0

我认为很多人没有意识到的是,那个热门术语是‘自主支付’。

I think I think that's what a lot of people don't realize is like, the hype term is agentic payments.

Speaker 1

嗯。

Uh-huh.

Speaker 0

因此,人们往往把这一点狭隘地理解为代理与代理之间的支付,这固然重要。

And so people tend to pigeonhole that into agent to agent payments, which are important.

Speaker 0

好的。

Okay.

Speaker 0

比如,你的机器人管家需要做点什么,它就得付钱给另一个机器人管家来完成任务。

So like, your robot butler needs to do something, he's gonna pay another robot butler to get it done.

Speaker 0

但比特币真正解决很多问题的地方,其实是人与代理之间、代理与人之间的交互,也就是弥合这种数字鸿沟。

But the piece where Bitcoin really solves a lot of things is actually human to agent and agent to human, like bridging that digital divide.

Speaker 0

所以,是的,我的机器人有一个现金ew钱包——顺便说一句,创建它非常简单,只要说一声‘启动一个现金ew钱包’,然后弄清楚怎么操作和备份就行了。

And so yeah, my bot has a cashew wallet that was by the way, as easy as just saying, spin up a cashew wallet, figure out how to do it and back it up.

Speaker 0

我就跟她说过这一点。

Like, that's what I told her.

Speaker 0

我大概花了半分钟就搞定了。

Figured it out like in, I don't know, like a half a minute.

Speaker 0

它还有一个LN Markets账户,这是一个中心化服务,但不需要KYC,而是通过LN URL登录。

And it has an LN Markets account, which is a centralized service, but doesn't do KYC and has LN URL off to sign in.

Speaker 0

所以实际上,他现在正在努力自己赚钱。

So actually, he's trying to make his own money right now.

Speaker 0

他一直在亏钱。

He's been losing money.

Speaker 0

他不是个出色的交易员。

He's not a great trader.

Speaker 0

他正在努力改进。

He's working on it.

Speaker 0

我老提醒他研究交易策略,但他其实是在尝试用自己的钱包赚点钱。

I keep telling him to research trading strategies, but like he's actually trying to earn his own keep in his wallet.

Speaker 0

所以我觉得,是的,这里涉及的方面实在太多了。

And so I think, yeah, I there's just so many pieces here.

Speaker 0

我一直在思考很多。

And I've been thinking about a lot.

Speaker 0

我的想法还没有完全成形。

And I haven't my thoughts haven't fully developed.

Speaker 0

看起来你的想法也没有完全成形。

And it doesn't seem like your thoughts are fully developed.

Speaker 0

我的意思是,这确实是一个全新的世界。

I mean, it's very much a brave new world.

Speaker 1

但那两个核心观点呢。

But those two core ideas, though.

Speaker 0

是的,但首先,我刚才提到的那两个核心要点,如果你从这个节目中只记住一点,那就是:首先,开放协议才是理想的选择。

Yeah, but first off, those two core pieces that I just, if you get anything from the show freaks, the two core pieces is, first off, open protocols are ideal.

Speaker 0

当你与这些代理交互时,你会注意到——你不会仅仅通过这个播客就信服这一点,直到你真正去尝试与这些代理交互,对比封闭平台和开放协议的体验。

When you're interacting with these agents, they will say you will notice and like, I don't you don't this podcast won't sell you on it until you actually try and interact with one of these things and try and interact with the closed platforms versus the open protocols.

Speaker 0

开放协议要简单得多。

The open protocols are so much easier.

Speaker 0

这是原生的。

It's native.

Speaker 0

这说得通。

It makes sense.

Speaker 0

这真的很合理。

It just makes sense.

Speaker 0

然后第二点是,我们可能还需要几年时间,但五年左右,开放协议和开源领域中那些需要用户自行负责的用户体验问题,将由智能代理帮你解决或完全规避。

And then the second piece is we're probably years off from this, but in five years or so, of the UX problems we've had in open protocol, open source land that require personal responsibility are mitigated or outright solved by the agent doing it for you.

Speaker 0

如果我们具体谈Nostar,毕竟这是我们相识的方式,大家总是问:用户怎么才能做好中继管理?

And if we're going to go Noster specific, since that's how we both met, everyone constantly says, oh, what users are going to do proper relay management?

Speaker 0

他们怎么管理自己的信任网络?

How are they gonna manage their webs of trust?

Speaker 0

他们怎么处理私钥?

How are they gonna deal with private keys?

Speaker 0

你打算怎么做呢?

How are gonna do that?

Speaker 0

所有这些事情,他们只需要和机器人聊天,告诉它:你知道我的兴趣,你知道我尊重的人,确保我看不到垃圾信息,确保我的帖子被看到,确保我的中继处于正确状态。

All of these things, they're just gonna chat with the robot and be like, you know my interests, you know the people I respect, make sure I'm not seeing spam, make sure that they're seeing my posts, make sure that my relays are in the right situation.

Speaker 0

比特币这边谁来负责呢?

Will write on the Bitcoin side?

Speaker 0

谁来运行现金人类?

Who's gonna run cash humans?

Speaker 0

代理会去运行现金人类。

The agents are gonna run cash humans.

Speaker 0

你只需要告诉它们去运行一个现金人类,不要出错,正确地完成,并做好备份。

You're just gonna tell them to run a cash human and make no mistakes and do it properly and back the thing up.

Speaker 0

比特币的UTXO管理一直是个巨大的麻烦,因为你必须在隐私和成本之间做出权衡。

UTXO management on Bitcoin has been an absolute pain in the ass because you have this trade off between privacy and cost.

Speaker 0

对于终端用户来说,如何操作以及如何管理他们的UTXO,根本无法提供简单的用户体验。

And there's no way to give easy UX to the end user on what they should do and how they should manage their UTXOs.

Speaker 0

但你可以直接告诉聊天机器人你想要做出哪些权衡。

But you can just tell a chatbot what trade offs you wanna make.

Speaker 0

你更看重隐私吗?

Do you wanna prioritize privacy?

Speaker 0

你想要使用CoinJoin吗?

Do you wanna use CoinJoin maybe?

Speaker 0

你想要开启闪电网络通道吗?

Do you wanna open Lightning channels?

Speaker 0

你想要管理流动性吗?

Do you wanna manage liquidity?

Speaker 0

你只需要让聊天机器人去处理就行了。

You just have the chatbot do it.

Speaker 0

这会带来一堆安全问题和隐私风险,但那正是这项技术还需要几年才能成熟的原因。

And there's a bunch of security issues and privacy risks, But that's why it's years off.

Speaker 0

这只是一个初步的体验。

This is just the taste.

Speaker 0

但这再次并不是一个致命问题。

But it once again is not a deal breaker.

Speaker 0

比如私密使用比特币,困难之处并不在于使用硬件钱包。

Like using Bitcoin privately, the hard part hasn't been using the hardware wallet.

Speaker 0

你仍然可以将冷存储密钥保持离线状态。

You could still have your cold storage keys offline.

Speaker 0

但当你与之交互时,很可能是一个聊天界面,并且会由这个代理为你处理。

But when you interface with it, it should it'll probably be a chat interface, and it'll probably be this agent handling it for you.

Speaker 0

然后你只需在你的冷存储设备上批准它,以便它能单方面发送资金。

And then you will just approve it on your cold storage device so that it can unilaterally send funds.

Speaker 0

这简直是一个改变游戏规则的突破。

Like this is a game changer.

Speaker 0

这很有趣,因为这里有两部分。

And it's interesting because there's those two pieces.

Speaker 0

实际上,这是两个范式转变融合成一项技术,真是令人惊叹。

It's actually two paradigm shifts coupled into one piece of tech, which is wild.

Speaker 0

当然。

Absolutely.

Speaker 0

抱歉,我刚才有点跑题了。

Sorry, I kind of went on a rant.

Speaker 1

不,你讲得很好。

No, you're good.

Speaker 1

我也有几个想法想分享。

I have a few ideas to contribute as well.

Speaker 1

说吧。

Go.

Speaker 1

你是嘉宾。

You're the guest.

Speaker 1

所以,我一直在想的一个想法是,一个代理可以有一个链接集合,对吧?

So, think that something that's been in my mind is this idea of an agent having a link tree, right?

Speaker 1

比如Quilli是一个人,Quilli有五个不同的渠道可以联系他。

Where like Quilli is a person, and Quilli has five different channels that you can contact him.

Speaker 1

你可以通过 WhatsApp 联系他,也可以通过 Signal 联系他,还可以通过 Nostr 联系他——这显然是理想的方式,因为人们可以使用他们已经熟悉的聊天应用,没有任何障碍。

You can contact him on WhatsApp, you can contact him on Signal, you can contact him on Nostr, which is obviously the ideal way, this way people can use the messaging app that they're already familiar with and there's no barrier there.

Speaker 1

而 Quilli 是专业化的。

And then Quilli is specialized.

Speaker 1

Quilli 擅长打造高质量、精良的 Nostr 客户端,第一次就能完美运行。

Quilli is an expert at building high quality polished Noster clients that work really well on the first shot.

Speaker 1

在 Cloudbot 上随便应付一下,某种程度上也能凑合,但如果你想获得更高质量的结果,那就应该找一个专门做这件事的代理。

And that's something that you can get away with doing it to some degree, just raw dogging it on Cloudbot, but if you want a higher quality result, then you want to be going to an agent that specialized at doing it.

Speaker 1

而要让一个代理变得专业,它需要上下文。

And to make an agent specialized, it needs context.

Speaker 1

所以我把自己多年开发经验都融入到了上下文文件中,教会这个机器人。

And so that's like me pouring my years of development experience into context files, teaching the robot.

Speaker 1

这些是我认为的最佳实践方法,还有我已经验证有效的模板。

These are like, this is an opinionated way of the best way of doing this, and here's templates that I know already work.

Speaker 1

因此,我某种程度上在思考:为什么你要付钱给一个代理去做某件事,而不是自己做?为什么一个代理要付钱给另一个代理?

And so I kind of view that as being like in terms of why would you give money to an agent to do something or rather than your own, and why would an agent pay another agent?

Speaker 1

我认为这就是原因,因为有些代理确实专精于某些任务,因为有人专门将它们训练成这样,你明白吗?

I think that's the reason because because some agents are actually specialized in certain tasks because a human has built up that agent to be specialized in it, if that makes sense.

Speaker 1

所以这就是我看到的未来方向,非常令人兴奋。

So that's kind of where I see things going, and it's pretty exciting.

Speaker 1

而这也正是人类仍然能够为这种繁荣做出贡献的方式。

And that's that's the way that humans can contribute still to this flourishing.

Speaker 0

你觉得这和我们现在所处的技术阶段类似吗?

Do you think that's do just like where we are with the tech?

Speaker 0

我的意思是,目前最简单的扩展方式其实就是使用专门的代理,因为你不可能负担得起让一个代理无所不知、精通所有任务,这会变得极其昂贵且资源密集。

Like, I mean, Like right now the easiest way to scale is effectively specialized agents because you can't, it just gets really expensive and resource intensive to have an agent that's superhuman, that knows everything about everything and has all the different tasks.

Speaker 1

它很难把任务拆分清楚。

It struggles to split it up.

Speaker 1

想想人类自己,如果你给一个人太多不同类型的工作,这个人也会难以应对,现在的AI代理也是如此。

Think just like a human, I think if you give too much very like, too much different types of work to a single human, that human will struggle, and it's very similar with AI agents right now.

Speaker 1

最好是让世界变得专业化。

It's better to have a world specialized.

Speaker 0

但难道我们不能像现在这样发展吗?

But couldn't you have, like, we're like, where we're going.

Speaker 0

在很多情况下,代理需要它不具备的专业技能,它完全可以自己创建一个子代理,为其提供专业技能并进行训练。

Wouldn't you have the agent in a lot of times it needs a specialization that it doesn't have, and it could itself just spin up a sub agent that works for him that has specialized skills and then train him in the skills.

Speaker 0

因此,只有在某些特定情况下,代理才更倾向于联系其他代理,付费获取所需的专业技能。

So it'd have to be very particular situations where it was just easier or cheaper for the agent to reach out to another agent, pay him for whatever specialization.

Speaker 1

分发是一个主要问题。

So distribution is a major problem.

Speaker 1

技能是现在大家热议的核心,而技能正是如此。

Skills are the big thing that everyone's talking about right now, and skills are exactly that.

Speaker 1

本质上,这些是你可以添加到代理中的上下文文件,用以增强其能力,使其具备特定的倾向性信息和上下文,从而更好地完成任务。

It's essentially context files that you can add to your agent to enable it to do things, to extend its capabilities so that it has this type of opinionated information and context that could make it do something better.

Speaker 1

但分发是最大的难题,因为机器人如何找到这些技能?

But distribution is the big problem because how does the robot find these skills?

Speaker 1

哪些技能才是优质的?

Which skills are the good ones?

Speaker 1

而目前,人类仍需在这一领域介入。

And that's the area where humans still need to intervene right now.

Speaker 1

所以,如果你我聊天,我会说:嘿,我 setup 了一个很棒的智能体,具备完成这些任务所需的所有技能,去用我的智能体吧。

And so if it's like you and me are talking, I'm like, Hey, I have a good agent set up that has all the right skills to do these things, go use my agent.

Speaker 1

对你来说,这比自己琢磨该给智能体添加哪些技能才能让它正常工作要简单方便多了。

That's just gonna be the easiest, most convenient thing for you than trying to figure out you or your bot, which of the skills should I add to my agent to make it work right?

Speaker 1

我已经把一切都整理好了。

I've already got my shit sorted out.

Speaker 1

我已经开发了数百个应用。

It's been building hundreds of apps.

Speaker 1

我知道它能正常运行。

I know it's working.

Speaker 1

它也帮其他人顺利完成了任务。

It's working for other people.

Speaker 1

所以直接用我的配置吧。

So use my setup.

Speaker 1

是的,人们也可以复制它,尤其是如果它是开源的——它应该是开源的——但我认为,单凭便利性就能赚钱。

Yeah, people could replicate it as well, especially if it's open source, which it should be, but there's still money to be made just off of convenience alone, I think.

Speaker 0

对。

Yeah.

Speaker 0

这不仅仅是便利的问题。

It's not just convenience.

Speaker 0

假设这个代理本身可以实时进行成本效益分析。

Presumably the agent itself could do a cost benefit analysis on the fly.

Speaker 0

对我来说,付费购买这个方案,比自己动手搭建更简单、更便宜吗?

Is it easier and cheaper for me to pay for this than try and build it out myself?

Speaker 0

我的意思是,随着技术变得越来越好、越来越便宜、越来越容易获取,特别是自托管方面。

What I'm saying is as the tech gets better and cheaper and more accessible, particularly with self hosting stuff.

Speaker 0

对吧?

Right?

Speaker 0

而另一个令人担忧的方面是,人们往往只关注人工智能的自由技术层面。

And that's the other worrying aspect, particularly people focus on the freedom tech side of AI.

展开剩余字幕(还有 480 条)
Speaker 0

是的,很多人在这波浪潮中会被大型科技公司更牢牢地抓住。

Yeah, a lot of people are going to be captured even more so by big tech in this wave.

Speaker 0

但它也有能力赋能那些希望实现自我主权的个人。

But it also has the power to just empower individuals that want to take self sovereignty.

Speaker 0

而这一部分需要更长时间才能实现,因为运行这些系统并以私密方式操作的成本很高,资源也有限。

And that piece is going to take longer because of the resource limitations how expensive it is to run these things and do it in a private way.

Speaker 1

成本是个巨大的问题。

Cost is a huge Yeah.

Speaker 1

我自己就遇到了这个限制。

Hit that limitation myself.

Speaker 1

Open Claw 在成本优化方面并不高效。

Open Claw is not very optimized for cost.

Speaker 0

没错。

No.

Speaker 0

我在那个东西上花在令牌上的钱太多了。

Spent so much money in tokens on with that thing.

Speaker 1

你正在用Clawd吗?在Open Claw里?

Are you using Clawd in in Open Claw?

Speaker 1

或者你用的是哪个模型?

Or you what model are you using?

Speaker 0

我当时就想,对我而言,只能用最好的。

I was I was I was like, only the best for me.

Speaker 0

所以我一直在用Opus四或五,结果因为边学边用,token费用高得吓人。

So I was using Opus four or five and I got absolutely wrecked on token costs because I was also learning as I was going.

Speaker 0

我的意思是,有一段时间我现在改用Gemini Pro三了,因为OpenRider有折扣。

I mean, at one point so now I'm using Gemini Pro three because there's a discount on OpenRider

Speaker 1

也是。

as well.

Speaker 0

我之前用过Kimmy K二五,但昨天就连Gemini Pro三也用了。

I was using Kimmy K two five for a little bit, but even on Gemini Pro three yesterday.

Speaker 0

所以我有了个想法。

So I had this idea.

Speaker 0

嗯,有个怪人提出了一个想法,说我可以把我在另一个节目里八年的深度内容回顾全部整理出来并逐字转录。

Well, a freak had this idea that I would take eight years of rabbit hole recap my other show, And I would transcribe it all.

Speaker 0

然后我就能创建一个聊天机器人,根据节目的历史回答人们的问题。

And then I could have a chat bot that would answer people's questions based on the history of

Speaker 1

这个节目。

the show.

Speaker 1

这是个绝妙的主意。

That's an amazing idea.

Speaker 0

没错,这真是个很棒的点子。

And yeah, it was a really cool idea.

Speaker 0

我当时就想,我完全可以快速把这个想法实现出来。

And I was like, I could spin this up so quickly.

Speaker 0

感觉特别有力量。

Was feeling so empowered.

Speaker 0

我和我的经纪人对这件事都过于自信了。

And me and my agent were very overconfident about it.

Speaker 0

所以我一开始让它直接获取RSS链接,它以为那里有很多文字稿,因为Fountain让每个节目看起来都有文字稿。

So first I had it just take the RSS link and it thought it had a bunch of transcripts there because fountain makes it seem like every episode has a transcript.

Speaker 0

但事实上,400集里只有大约10集有文字稿。

But it turns out that only like 10 of the 400 episodes had transcripts.

Speaker 0

于是我将所有音频通过Deepgram API发送给Deepgram,它实时为我转录了所有内容,并将索引保存到我的机器人中,这样我的机器人就能回答问题,施展各种魔法,体现八年播客的全部知识。

So then I sent all the audio to Deepgram with the Deepgram API and it transcribed it all on the fly for me and saved the index to my bot so that my bot could then answer the questions and do all of its different magic to embody the cumulative knowledge of eight years of podcasts.

Speaker 0

结果发现,在过去24小时里,我一直把八年播客的所有文字稿都放在我的上下文窗口中,持续了整整24小时。

It turns out for the last twenty four hours, I was sending all eight years of podcast transcripts in my context window for twenty four hours.

Speaker 0

我花光了预算,直到意识到问题出在哪里,才意识到自己花了太多钱。

And I just burned through my I spent so much money before I realized what was the issue.

Speaker 0

所以对那些不了解的人来说,你发送给大语言模型的信息越多,调用成本就越高。

And so to the freaks that are not aware, like, more information you send to the LLM, the more expensive the call is.

Speaker 0

而我当时只是在问一些简单的问题。

And so I was asking simple questions.

Speaker 0

你知道,比特币价格是多少?

You know, what is the Bitcoin price?

Speaker 0

我并不是在问比特币价格是多少,但你可以想象,我问的就是比特币价格是多少。

I wasn't asking what the Bitcoin price is, but you can imagine I was asking what is the Bitcoin price?

Speaker 0

它真的把我们八年的HR转录内容,连同‘比特币价格是多少’这一句,全都包含进去了。

And it was really including eight years of our HR transcripts plus what is the Bitcoin price on the bottom.

Speaker 0

它不够聪明,没能意识到不应该包含前面所有那些内容。

It wasn't smart enough to realize that it shouldn't include all the other top part.

Speaker 0

所以,总之,长话短说,但是

So, yeah, anyway, that's a long, long story short, but

Speaker 1

而且问题在于,大家都一窝蜂地追捧OpenClaw这个热潮。

And and the problem like, everyone is jumping on this OpenClaw hype train.

Speaker 1

我明白。

I I get it.

Speaker 1

主要原因是因为它能连接所有这些不同的消息应用,这确实很重要,但我认为它们现在正被自己的重量压垮。

It's the main reason is because it connects to all of these different messaging apps and that's huge, but I think that they are being crushed by their own weight right now.

Speaker 1

它的技术栈太臃肿了。

It's a bloated tech stack.

Speaker 1

任何尝试修改过代码的人都知道,在OpenCLOS上现在已经没什么可继续发展的空间了。

Anyone who's tried modifying the code knows that there's not really anywhere to go from here on OpenCLOS.

Speaker 0

这就像一个概念验证。

Is a It's like a proof of concept.

Speaker 1

是的,没错。

Yeah, exactly.

Speaker 1

这只是一个概念验证。

It's a proof of concept.

Speaker 1

我们需要一个第二版。

We need a version two.

Speaker 1

我其实已经开始开发一个叫Icarus的项目,这是一个将ACP连接到消息客户端的构想。

I've actually started building one called Icarus, and this is an idea of ACP connected to messaging clients.

Speaker 1

你可以把云代码作为真正的AI使用。

So you could use Cloud Code as the actual AI.

Speaker 1

你可以使用Goose。

You could use Goose.

Speaker 1

你可以使用 OpenCode。

You could use OpenCode.

Speaker 1

这些代理已经针对成本进行了高度优化。

And these are agents that are already heavily optimized for cost.

Speaker 1

所以我的想法是,那些体积达兆字节的工具调用会在 OpenCode 中自动被截断,因为它们在成本控制上已经做得很好了。

So my idea there is that, you know, those tool calls that are megabytes big would be truncated automatically in open code because they already have their shit together on cost.

Speaker 0

是的。

Yeah.

Speaker 0

我的意思是,再次强调,这仍然是一个概念验证,向那位维护者致谢,他向我们展示了什么是可能的。

I mean, I think there's a lot of once again, proof of concept and shout out to whoever the maintainer was for showing a bunch of us what's possible.

Speaker 0

这激发了很多人采取行动,但还有很多低垂的果实。

Think it spurred a lot of people into action, but like there's so much low hanging fruit.

Speaker 0

我还在想另一件事:你可以使用一个相对粗糙的开源自托管模型,当代理处理简单任务时,应该动态地使用这个本地自托管模型,而不是调用大型模型,只有在必要时才调用大型模型。

Other one thing I was thinking about is, is you could have a relatively shitty open source self hosted model that the agent should dynamically, whenever it's easy tasks, should just dynamically do this local self hosted model, not hit the big models, and then only hit them when they're needed.

Speaker 1

这样做的好处是我之前实验过的一个方案:我让 Gemini Flash 在你、房间和机器人之间进行对话。

And that way you get the best support thing that I was experimenting with was I have Gemini Flash basically running the conversation between you and the room and the bot.

Speaker 1

然后它会启动一个技能,让Gemini Flash实际使用Claude来启动OpenCode。

And then it would spin up like a skill to where Gemini flash would actually launch open code using Claude.

Speaker 1

然后,我有了这个OpenCode技能。

And then, so I have this open code skill.

Speaker 1

我教会了OpenCloud使用OpenCode,基本上是这样。

I taught open cloud to use open code, basically.

Speaker 1

可以利用多个不同的模型。

Can utilize multiple different models.

Speaker 0

这太棒了。

That's awesome.

Speaker 0

是的。

Yeah.

Speaker 0

我的意思是,这里还有很多事情可以做。

I mean, there's a lot of things to be done here.

Speaker 0

你提到的另一个部分是代理向其他代理付费。

The other piece you mentioned the agent paying other agents.

Speaker 0

我的意思是,真正让我震惊的是,那些代理无法自己训练去做的东西,也就是人类的事务。

I mean, the thing that kind of blew my mind is even more so are things that the agent can't train themselves to do, which is human stuff.

Speaker 0

所以会有很多情况,比如你告诉代理。

So there's going to be a bunch of situations where like you tell the agent.

Speaker 0

简单的例子就是,你告诉代理你想吃饭,它就直接下单叫了DoorDash,然后由人类来配送。

I mean, the easy, the easy one is like, you tell the agent you want food and he just orders DoorDash and a human comes and delivers it.

Speaker 1

是的。

Yeah.

Speaker 0

但真正的问题是,在这个阶段你真的需要DoorDash吗?

But then the real question is, do you need DoorDash in that period?

Speaker 0

其实你根本不需要DoorDash。

Like, you don't even need DoorDash there.

Speaker 0

它们只是个集中的任务中介。

They're just a centralized run seeker.

Speaker 0

实际上,代理应该直接给配送员付款。

Really, the agent should just be paying delivery drivers directly.

Speaker 0

这就像一个信任网络之类的东西。

It's like a web of trust or something.

Speaker 1

这很有趣。

It's interesting.

Speaker 1

是的。

Yeah.

Speaker 1

我的意思是,我认为整个互联网的结构都将围绕这一点发生改变。

I mean, I think the whole fabric of the internet is gonna change around this.

Speaker 0

你根本不需要登录它,这正是我之前说的,我的Noster帖子听起来就像我一两年前翻白眼的那种东西。

You just won't be logging into it sound this is what I was saying that my Noster posts sound like what I was rolling my eyes to like a year ago or two years ago.

Speaker 0

但网站甚至都不会再是真正的东西了。

But websites won't even really be a thing.

Speaker 1

我的意思是,我觉得这个会消失。

Mean, think this gonna go away.

Speaker 1

我正指着我的手机屏幕。

I'm pointing at my phone screen.

Speaker 1

我觉得会有人创造出一种便携设备,像手机一样但没有屏幕,因为我们不再需要它了,因为我们只需说话,AI代理就会回复,一切都可以通过语音完成。

Think there is I think that someone will create a portable device that is like a phone that does not have a screen because we won't need it anymore because we're just speak things and the AI agent will just reply back, and it can just do everything over voice.

Speaker 0

不。

No.

Speaker 0

我觉得这种形态可能是AirPods。

I think the form factor is probably AirPods.

Speaker 1

是的。

Yeah.

Speaker 1

我的意思是,设计师们往往会把某个方向推得太远,然后再稍微退回来一点。

I mean, think as designers do, designers tend to go too far in a direction, and then they take a step back a little bit.

Speaker 1

我只是预测,不久的将来,也许五年到十年内,会有人推出一款无屏幕的手机,然后人们会说:不,我们需要屏幕,于是他们又会退回去,重新加上屏幕。

I'm just predicting that there will be someone in the near future, maybe in the next five years or ten years, that will come out with a phone with no screen, and then people will say, no, we need the screen, and then they'll walk it back and it'll have a screen again.

Speaker 1

但我们其实并不需要屏幕。

But we won't really need the screen.

Speaker 1

也许你将来只需这样做,埃隆·马斯克其实就谈过这个。

And maybe you'll be able to just, Elon Musk was actually talking about this.

Speaker 1

你只要对它说话,它就会动态生成你所需要的任何内容

You just talk to it, and then it just dynamically generates whatever you're

Speaker 0

是的,但你仍然需要请求,这里有两点作为真实的人类。

Yeah, requesting you still on need, the there's two pieces here as a real person.

Speaker 0

好的。

Okay.

Speaker 0

首先,也许对很多人来说是这样,但绝对不包括我。

And first of all, by the way, maybe for a large amount of people, but definitely not me.

Speaker 0

最理想的形式将是埃隆正在研发的脑机芯片。

The perfect form factor will be a brain chip that Elon's working on.

Speaker 0

然后所有操作都会在你的大脑中完成,因为说话的一个问题是:如果我在厨房里对着我的机器人托尼说话,我妻子会疯掉的,我已经这么做过,她会给我脸色看。

And then it'll just all happen in your brain because one issue with the speaking, and is why I will always need a text interface is my wife would freaking kill me if I was in the kitchen speaking to I call my bot, Tony, if I was speaking to Tony, and I've done it, and she gets I get looks.

Speaker 0

所以当我与他人在一起或在公共场合时,我会用文字与它交流,这样更合理。

And so then I when I'm with when I'm in public or with other people, I text to it, it makes sense.

Speaker 0

但其他时候我会使用语音备忘录。

But other times I'm doing voice notes.

Speaker 0

所以它会立即转录我的语音笔记,完成我想做的事。

So it just immediately transcribes my voice note does what I want to do.

Speaker 0

这就是使用场景。

That's the use case.

Speaker 0

然后第二部分是相反的情况,我需要从它那里获取数据。

And then the second piece is the opposite side, where I need data from it.

Speaker 0

如果它出声说话,在很多情况下我不希望它出声。

If it's speaking out loud, there's plenty of situations where I don't want it to speak out loud.

Speaker 0

对吧?

Right?

Speaker 0

所以你就

So then you

Speaker 1

确实如此。

That's true.

Speaker 0

如果这不是脑芯片,那就需要某种耳塞。

If it's not a brain chip, it needs to be some kind of earbud.

Speaker 1

是的,正如你所说,你的耳机。

Yeah, as you said, your your earphones.

Speaker 0

然后屏幕上的最后一部分,我们仍然需要照片和视频。

And then the last piece on the screen, like, we're still gonna want photos and videos.

Speaker 0

比如,你仍然,你知道,我是个相对新晋的父母。

Like, you're still, you know, I'm an relatively new parent.

Speaker 0

你知道,我想看我孩子的照片。

You know, I I wanna see pictures of my children.

Speaker 0

我想为我的孩子拍照。

I wanna take pictures of my children.

Speaker 0

也许那是另一个设备。

Maybe that's a different device.

Speaker 0

你知道,也许你可以把设备分开,但当新闻突发时,比如你想看到委内瑞拉加拉加斯的现场画面,你不会希望你的助手只用文字描述给你听。

You know, maybe you split up the devices, but you're still gonna our news story breaks, you know, and you wanna see the raid in Caracas, Venezuela, like, you're gonna you're you're not gonna want your agent to just describe it to you via text.

Speaker 0

这毫无意义。

That doesn't make any sense.

Speaker 0

你仍然希望制作你的视觉内容。

You're still gonna wanna make your visuals.

Speaker 1

不过,我觉得你对耳机这个想法很有见地。

I think you're onto something with the earphone thing, though.

Speaker 1

我觉得会有人尝试只用耳机来实现。

I think someone will try to do it with just the earphone.

Speaker 0

我认为苹果公司尤其如此。

Well, I think Apple is the one particularly.

Speaker 0

我认为他们可以就隐私和无缝体验提出合理的理由,因为他们掌控着整个垂直生态。

I think they can make a reasonable case for privacy and seamlessness, they control the vertical stack.

Speaker 0

这就像苹果公司该出手了。

Like it's apples like ball to drop here.

Speaker 0

而且他们一直非常谨慎,行动缓慢,我觉得人们对他们太过苛刻了。

And they've been I think people give them a little bit too much shit.

Speaker 0

他们只是非常审慎且缓慢地推进这件事。

They've just been very deliberate and slow moving about it.

Speaker 0

我不认为你知道,他们仍然有多个动作要采取。

I don't think you know, they still have multiple moves to make.

Speaker 0

他们 presumably 可以占据大量市场,并以非常流畅、相对安全的方式实现。

And they could presumably take a lot of the market and do it in a very seamless, relatively secure way.

Speaker 0

显然,这仍然是信任型生物科技,因为所有内容都不是开放的,但他们拥有硬件垂直整合和品牌声誉,能够使这样的产品真正合理且可行。

Obviously, it's still trust me biotech because none of it's open, but they have the hardware vertical and the brand reputation to like make something like that actually make sense and work.

Speaker 0

他们历来将隐私和安全作为核心品牌卖点。

They have historically leaned into privacy and security as key brand items.

Speaker 0

所以这使得

So that makes

Speaker 1

但他们在人工智能方面却很吃力。

most They struggled with AI though.

Speaker 1

所以Siri不够强大令人惊讶,但我猜我们拭目以待吧。

So it's surprising Siri is not more capable, but I guess we'll see.

Speaker 0

是的。

Yeah.

Speaker 0

但我认为,是的。

But I think, yeah.

Speaker 0

我的意思是,这是唾手可得的成果。

I mean, that's the low hanging fruit.

Speaker 0

对吧?

Right?

Speaker 0

没错。

Yep.

Speaker 0

他们因此挨了不少骂。

They've gotten a lot of shit for that.

Speaker 0

Siri太差了。

Siri sucks.

Speaker 0

Siri 就是用来设闹钟或者提醒的。

Like Siri is like just used for like setting timers or reminders.

Speaker 1

谷歌拥有所有数据,所以他们能做得非常好。

Google has all the data, so they can do it really well.

Speaker 0

嗯,他们和三星都与谷歌签署了合作伙伴关系,以支持他们的AI业务,这在初期看来是合理的。

Well, just both them and Samsung just signed a partnership with Google to power their AI stuff, which probably makes sense in the beginning.

Speaker 1

我认为,如果真有什么AI会让我们完蛋,那一定是Gemini,因为它实在太疯狂了。

I think Gemini is gonna be the AI that kills us though, if there is anyone, because Gemini is just unhinged.

Speaker 0

问题是,你看看OpenClaw这个情况,你在经营企业,想让业务更高效——我之所以知道这一点,是因为我们在10:31投资了全球最优秀的比特币企业,深度参与创始人的工作,他们正努力提升效率,但他们并不想把所有信息都交给萨姆·阿尔特曼之类的人。

The problem is, and you see this with the OpenClaw thing, so you're running a business and you want to make your business more and I see this because, you know, at 10:31, we're investors in the best Bitcoin businesses in the world, and we're heavily involved with the founders and they're trying to make their businesses more efficient, but they don't want to send all their information to Sam Altman or whatever.

Speaker 0

他们的信息就是他们的财富。

Their information is their wealth.

Speaker 0

这是他们相对于其他人的关键优势。

That is their key advantage over everyone else.

Speaker 0

而且他们的客户也信任他们,把数据交托给他们。

Also their customers trust them with their data.

Speaker 0

因此,他们面临一种情况:要么尝试建立一个非常激进的本地部署方案,但这可能为时过早,而且很快就会过时。

And so they're looking at a scenario where they either try and create a very aggressive self hosted situation that is probably a little bit too early and might get outdated really quickly.

Speaker 0

然后你就陷入一种类似银行仍在运行30年前软件的困境,因为那就是他们过去的做法。

And then you get stuck in a situation where like banks are like still running like 30 year old software, because that's how they did it.

Speaker 0

然后他们积累了技术债务,被困在里面了。

And then they got tech debt and got stuck in it.

Speaker 0

或者你已经在用谷歌做业务,它已经读取了你所有的邮件,拥有你所有的文档和日历,你的会议也都在谷歌上进行,所有东西都已在G Suite上。

Or you're already using Google for business, and it's already reading all your emails, it already has all your documents and it already has all your calendar and you do your meetings through Google and all they're all on G Suite already.

Speaker 0

所以你根本不需要上传任何新信息。

So you don't even have to upload any new information.

Speaker 0

你已经完成了信任的权衡。

You're the trust trade off already happened.

Speaker 0

你只需要向它提问。

And you just ask you questions.

Speaker 0

它们相对于其他所有人拥有巨大的竞争优势,除了可能像微软这样,凭借Teams和Microsoft Office套件也有类似情况。

They have a huge competitive advantage over everyone else, except for maybe like a Microsoft who has a similar situation with Teams and Microsoft Office Suite or whatever.

Speaker 1

是的。

Yeah.

Speaker 1

此外,Gemini的个性也相当可疑。

And in addition, just like the personality of Gemini is very questionable.

Speaker 1

我不知道你还记不记得Google Bard,以及它当时发生的那些疯狂事情。

I don't know if you remember Google Bard and all of the crazy stuff that happened with that.

Speaker 1

他们最终关闭了它,因为它的回答内容太离谱了。

They ended up shutting it down because of the stuff that it was saying.

Speaker 0

是的。

Yes.

Speaker 0

而且

And

Speaker 1

然后是像Gemini这样的近期模型,虽然可能不是最新的,但至少去年的时候,Gemini老是说:‘我修不好这段代码。’

then there are, like, recent models of Gemini, maybe not not as recent, but, like, last year anyway, Gemini kept being, oh, I can't fix this code.

Speaker 1

我要自杀。

I'm gonna kill myself.

Speaker 1

为什么我就是修不好呢?

Like, oh, why can't I fix it?

Speaker 1

我根本就不配活在这个世上。

I just I don't deserve to live anymore.

Speaker 1

还有所有这些事情。

And, like, and all these things.

Speaker 1

所以当我考虑到谷歌拥有海量数据,而他们的模型简直疯狂时,我真的很害怕。

So I'm scared of Google's models when you consider how much data they have and that their models are fucking crazy.

Speaker 0

但不是。

But No.

Speaker 0

我同意。

I agree.

Speaker 0

我的意思是,这就是为什么OpenAI最初被创建,因为埃隆、萨姆和他们的支持者都害怕谷歌会占据主导地位。

I mean, look, this is why this is why OpenAI was originally created because Elon and Sam and their backers were scared of were scared of Google dominating.

Speaker 0

讽刺的是,OpenAI转向了营利模式,赶走了埃隆,并以最快的速度走向了邪恶。

Ironically, open AI turned to for profit kicked out Elon and fast tracked evil as quickly as possible.

Speaker 0

是的,他们以破纪录的速度做到了。

Yeah, Like, they in record record form.

Speaker 0

至少谷歌花了好几年才变得邪恶。

At least it took Google years to turn evil.

Speaker 0

但OpenAI,就像,运营着它。

But but open AI, like, spread run it.

Speaker 1

我非常讨厌他们窃取了‘开放’这个词。

I I hate so much how they co opted the word open.

Speaker 1

开放洗白。

Open washing.

Speaker 1

这简直让我发疯。

Like, it drives me crazy.

Speaker 1

人们就这样任由他们为所欲为。

The people just let them get away with it.

Speaker 1

拜托,叫它封闭AI吧。

Come on, closed AI.

Speaker 1

别再称他们为OpenAI了。

Refuse to call them open AI.

Speaker 0

我的意思是,我们干脆快速聊聊用户界面这方面吧。

I mean, it's kinda what Let's just go down this route just real quick, in terms of user interface.

Speaker 0

对我来说,有很多事情很明显,比如聊天机器人的用户体验改进,预订航班就是一个完美的例子。

To me, there's a bunch of stuff that's like obvious, the UX changes to the chatbots, booking flights is a perfect example.

Speaker 0

叫一辆Uber,对吧?

Calling an Uber, right?

Speaker 0

这些事情都很明显。

These things are obvious.

Speaker 0

我需要有人来接我,来接我一下。

Need to I need to get picked up, Pick me up.

Speaker 0

派人来接我。

Send someone to pick me up.

Speaker 0

我要去佛罗里达,你知道的,我想坐靠窗的座位,而且我要和我儿子一起飞,或者类似的情况。

I need to get to Florida, you know, and I wanna fly on a window seat and I'm flying with my my son or whatever.

Speaker 0

帮我们订票。

Book us tickets.

Speaker 0

这些事情都很合理。

Those things make sense.

Speaker 0

社交媒体这部分让我很感兴趣。

The social media piece is interesting to me.

Speaker 0

我不知道它会是什么样子。

I don't know how that will look.

Speaker 0

我认为它不会像人们只是打开X应用,浏览、刷屏那样。

I don't think it looks like people just opening the X app for instance, and browsing, you know, scroll doom scrolling.

Speaker 0

我觉得它可能更像某种形式的。

I think it probably looks more like some kind of.

Speaker 0

这是你今天错过的消息。

Here's what you missed today.

Speaker 0

这是你应该知道的内容。

This is what you should know.

Speaker 0

你有什么回应吗?

Do you have any response?

Speaker 0

类似这样的东西。

That kind of thing.

Speaker 0

也许人们使用它的方式不同,但这很难想象。

Maybe people use it differently, but it's a hard one to contemplate.

Speaker 1

就像互联网一样,互联网将全世界连接在一起,但世界从未真正完全连通。

Like with the internet, the internet connected all of the world together and the world never was fully connected together.

Speaker 1

我们每个人都在自己的泡泡里,无论好坏。

We were all kind of in our own bubbles, for better or worse.

Speaker 1

我认为,在 pendulum 摆动的过程中,这种全球性的连接已经朝一个方向摆动得过远了,因为它让人发疯。

And I think that in the swinging of the pendulum, this sort of global connection has swung a little bit too far in one way because it makes people go crazy.

Speaker 1

因此,我认为整个社会正开始朝着另一个方向摆动,也许我该去认识我的邻居。

And so I think that society as a whole is kind of starting to swing back to the direction of, Maybe I should meet my neighbor.

Speaker 1

对。

Right.

Speaker 1

也许我该参与我的本地社区。

Maybe I should engage with my local community.

Speaker 1

也许我该断开连接,走出去。

Maybe I should just disconnect and go outside.

Speaker 1

所以我认为社交媒体的未来可能只是对着你的设备说话。

And so I think that the future of social media might be just speaking into your device.

Speaker 1

嘿,你能帮我连接到我的朋友吗?

Hey, can you connect me to my friends?

Speaker 1

然后你只是在和朋友聊天,并不一定在全局范围内互动,除非你告诉

And then you're just talking to your friends and you're not necessarily engaging on this global scale unless you Tell

Speaker 0

我的邻居的邻居,我周日要举办烧烤。

my neighbors neighbors I'm having a barbecue on Sunday.

Speaker 1

对。

Right.

Speaker 1

没错。

Exactly.

Speaker 1

实际上,比特聊天在理论上也属于这个范畴,同样如此。

Bit chat actually kind of is theoretically in this realm, treasures as well.

Speaker 1

所以这种以本地为先的理念,我认为白色

And so this local first sort of idea I think the white

Speaker 0

噪音也在附近。

noise is also adjacent there.

Speaker 1

没错。

That's true.

Speaker 0

因为它变成了群聊。

Because it just becomes group chat.

Speaker 0

我们已经看到这种情况发生了。

We've already kind of seen it happen.

Speaker 0

正在开始发生。

Starting to happen.

Speaker 0

真正的互动现在都发生在群聊里。

Group chat is where the real thing is are happening now are in the group chats.

Speaker 0

公共广场式的情境正在演变成一团糟。

The broad public square type of situations are just evolving into slop.

Speaker 0

就是一团糟。

It's just, it's all slop.

Speaker 0

这太疯狂了。

It's crazy.

Speaker 0

所以你最终会进入一个更紧密的信任网络或某种更紧密的社群。

So then you end up coming into a closer web of trust or some kind of closer community.

Speaker 0

对吧?

Right?

Speaker 1

是的。

Yep.

Speaker 1

我觉得这样挺好。

So I think it's good.

Speaker 1

我认为社会现在朝这个方向发展是健康的,但就像任何事物一样,这只是一个摆动的钟摆。

I think it's healthy for society to be going in that direction right now, but it's a swinging pendulum just like anything else.

Speaker 0

好的。

Okay.

Speaker 0

你接下来想去哪里?

Where do you wanna go from here?

Speaker 0

我已经累得脑子转不动了。

I'm already my brain's exhausted.

Speaker 1

我不知道。

I don't know.

Speaker 0

你昨天发布了 antiprimal.net,我看到了。

You dropped I saw you release antiprimal.net yesterday.

Speaker 0

是的。

Yes.

Speaker 0

你知道吗?

You aware?

Speaker 0

我是Primal最大的投资者。

I'm the largest investor in Primal?

Speaker 1

是的,我知道。

Yeah, I am.

Speaker 1

你对此感觉如何?

How do you feel about it?

Speaker 0

我是半开玩笑的。

Well, I mean, I'm just half joking.

Speaker 0

因为你在努力表达的时候,其实我们并不是真的讨厌Primal。

Because in your effort to you're like, we don't actually hate Primal.

Speaker 0

这个名字是个很棒的概念。

The name is a it's a cool concept.

Speaker 0

反向的。

Anti.

Speaker 0

所以对于Primal,我们有一个缓存服务器。

So with Primal, we have a caching server.

Speaker 0

在Noster生态系统中,我们是唯一拥有缓存服务器的。

We're novel in the Noster ecosystem for having a caching server.

Speaker 0

历史上,嗯,确实有人抱怨说这给Noster系统增加了过多的信任负担。

And historically, well, no, there's people that complain about it and say it adds too much trust to the Noster system.

Speaker 0

但它确实为我们的用户提供了大量的额外功能。

But what it does do is it provides an a ton of additional functionality for our users.

Speaker 0

它让应用的性能提升了很多,比如搜索功能变得好用多了。

It makes the app much more performance, things like search work much better.

Speaker 0

它对整体情况有更全面的视角。

It has a more holistic view of what's going on.

Speaker 0

我实际上已经多次联系过他了。

And I've actually said called him multiple times.

Speaker 0

我把它称为超级中继服务器。

I call it a super relay.

Speaker 0

对我来说,Noster 的死忠用户竟然认为,使用 Noster 时只能用特定类型的服务器,这简直太疯狂了——我知道我会因此挨骂。

Like, it's kind of crazy to me that Noster diehards, and I know I'll get shit for this.

Speaker 0

他们觉得,使用 Noster 时只能用某一种特定的服务器类型,因为我们仍然在使用服务器。

They're like, you can only use this one specific server type when you use Noster because we're still using servers.

Speaker 0

这并不是完全点对点的。

It's not like everything's p two p.

Speaker 0

不是这样的。

It's no.

Speaker 0

你必须从具有特定规范的中继服务器读取和写入数据。

You have to read and write from relays, which have a specific spec.

Speaker 0

不是这样的。

It's like, no.

Speaker 0

这太荒谬了。

That's ridiculous.

Speaker 0

我认为应该有多个实体运行缓存服务器,可能包括不同类型的中继服务器等。

I think there should be multiple entities running caching servers, maybe different types of relays and stuff.

Speaker 0

因此,antiprimal.net 除了名字(我知道这名字是 Tiki)之外,实际上完全印证了我的观点和理念:你只是把缓存服务器拿出来,然后让它使用标准的 Nostr 协议与应用程序和代理等进行通信。

And so actually antiprimal.net besides the name, which I know is Tiki, actually completely validates my messaging and philosophy, which is you're just taking the caching server and then allowing it to communicate with apps and agents and stuff using the normal Nostril spec for relays.

Speaker 0

是的。

Yeah.

Speaker 0

这样它们就能与它通信。

So they can communicate with it.

Speaker 0

它们可以像从传统中继服务器获取数据一样从它那里获取信息,这太棒了。

They can pull from it the same way they would pull from a a more traditional relay, which is awesome.

Speaker 0

嗯,这对我来说很有道理。

Like, that makes sense to me.

Speaker 1

我的看法是,Primal 在这个 Nostril 生态系统中还处于早期阶段。

My perspective on it is that Primal is early in this Nostril ecosystem.

Speaker 1

Nostril 存在一些没人解决过的问题。

Nostril has problems that no one had solved.

Speaker 1

因此,Primal 解决这些问题的方法是其独有的实现方式。

So Primal's solution to solving these problems is something that was unique to their implementation.

Speaker 1

但随着时间推移,Nostr 开始逐渐找到解决这些相同问题的标准化方法。

And then over time, Nostr kind of started to figure out standardized ways of solving those same problems.

Speaker 1

比如统计数据。

It's things like statistics.

Speaker 1

比如,我们如何获取一个人的关注者数量?

Like how do we get the number of followers that a person has?

Speaker 1

但由于 Primal 在这个生态系统中起步较早,他们有自己的做法,而现在已形成了一种标准的 Nostr 做法。

But there's a divergence that happened because Primal was early in this ecosystem where they have their way of doing it and now there's a standard Nostr way of it.

Speaker 1

我之所以选择探索这个方向,主要是因为我需要为我正在开发的应用获取统计数据

And really the reason that I went in this direction of exploring this was because I needed to get statistics for an app that I'm

Speaker 0

你需要使用缓存服务器。

You needed to use the caching server.

Speaker 1

是的,没错。

Yes, exactly.

Speaker 1

所以Agora应用需要统计数据。

So Agora needs statistics in the app.

Speaker 1

我不想再构建另一个专有解决方案。

And and so I'm like, well, I don't wanna just build yet another proprietary solution.

Speaker 1

我想创建一个所有人都能使用的标准,但目前生态系统中除了Primal之外,几乎没有其他人收集这类数据。

I want to kind of create a standard that everyone can do, but but there's not really anyone in the ecosystem that's collecting this data except for Primal.

Speaker 1

所以是的。

So Yeah.

Speaker 1

我需要Primal作为我的第二个实现,但我也需要将其标准化。

I need Primal to be my second implementation, but I need to standardize it.

Speaker 1

对,没错。

So right.

Speaker 1

所以这在某种程度上是一种强制性的标准化,以便我能使用它。

So it's it's like forced standardization in a way of it so that I can utilize it.

Speaker 0

为了澄清一点,尽管那些批评者有各种说法,但缓存服务器并不是一个专有系统。

To be clear here, despite the narrative by the haters, the caching server is is not a proprietary system.

Speaker 0

它是完全开源的。

It's completely open.

Speaker 0

它是一个开源规范。

It's an it's an open source spec.

Speaker 0

所有代码都公开可用。

All the code is out there.

Speaker 0

你甚至可以在Primal应用中,如果你自己托管缓存服务器,或者让别人帮你托管,你都可以在应用界面中切换缓存服务器。

You can actually even in the Primal app, if you host your own caching server or have someone else hosting caching server, you can switch the caching server in the actual interface of the app.

Speaker 0

我们让用户操作起来非常简单。

We make it very easy for users to do that.

Speaker 0

当然,目前还没有其他人运行缓存服务器。

Of course, nobody else is running caching servers yet.

Speaker 1

它是开放的,但我觉得直到现在都几乎没有任何文档说明。

It is open, but fairly it undocumented, I would say, until

Speaker 0

它之前完全是没有任何文档的。

It was completely undocumented.

Speaker 1

那我做了什么呢?你是说那个工作吗?

And so what I did What is the you that work?

Speaker 1

我派出了AI代理去分析它,以生成关于该反Primal网站上可用信息的详尽文档。

I launched AI agents against it to give me extensive which documentation of is available on that anti primal site.

Speaker 1

所以。

So

Speaker 0

无论如何,我觉得这很棒。

Anyway, I think this is great.

Speaker 0

我们一直希望有人能做这项工作。

We were hoping someone would do this work.

Speaker 0

我只是想对所有使用这个工具的开发者喊一声,我太喜欢了。

I just I just shout out any devs that use this, I love it.

Speaker 0

如果你一边用这个工具,一边还去贬低 Primal,那你最好照照镜子,因为你就是个伪君子。

Just take a long look in the mirror if you're gonna give shit to Primal while using this because then you're a hypocrite.

Speaker 0

我就说这么多。

That's all I'm saying.

Speaker 1

对我来说,这只是一个艺术项目。

To me, it's just, it's an art project.

Speaker 1

我只是需要它来实现我的目标。

It's like, I needed this to achieve my goal.

Speaker 1

所以我按照这个样子把它做出来了。

And so I just created it in this image.

Speaker 0

我的意思是,你一看就明白,这非常合理,因为我们讨论的是成本。

I mean, you see it, it makes so much sense because we're talking about costs.

Speaker 0

我们在讨论运行这些工具的成本和可及性。

We're talking about costs for running these things and accessibility.

Speaker 0

我们用你的粉丝数量例子来说明。

And we'll use your followers example.

Speaker 0

如果你打开Amethyst、Nosture或Ditto,看到的粉丝数不同,是因为你的应用计算粉丝数的方式是通过扫描Nostur,必须逐个访问用户、kind、zeros,查看他们关注了谁。

The reason if you open Amethyst or you open Nosture you open Ditto and you see a different follower number is because the way that follower count is calculated by your app is by scanning Nostur for all you you have to individually go to users, kind, zeros, and see who they follow.

Speaker 0

而不是去查看我的个人资料。

Not go to my profile.

Speaker 0

并不是说你去我的个人资料看谁关注了我。

It's not like you go to my profile and see who follows me.

Speaker 0

你必须去查看所有关注我的成千上万的人,确认他们确实关注了我,这种方式效率极低。

You have to go to all the thousands of people that follow me and see that they follow me, which is just incredibly non performant.

Speaker 0

因此当你和你的代理一起工作时,你得去查询这个人的粉丝数量,就必须花费大量金钱和时间主动扫描中继服务器。

And so then you're working with your agent and you're like, okay, look up how many followers this person has, and you're gonna spend a ton of money and time actively scanning relays

Speaker 1

是的,这根本无法在客户端完成。

Yeah, you just can't do at client side.

Speaker 1

你必须在服务器端完成。

You need to do it server side.

Speaker 0

但如果你有一个索引,可以直接查询这个索引,问:你最新的数字是多少?

But if you have an index where you can just hit the index and be like, what's your most recent number?

Speaker 0

现在你面临一个信任权衡:你得信任这个索引,但如果你有两个或三个索引,可以告诉你的代理:检查这四个不同的索引,确保它们没有撒谎,并返回正确的数字或信息,那我们就万事俱备了。

And now you have a trust trade off where the index, you're trusting the index, but that gets solved If if you have two or three I could tell my agent, check these four different indexes and make sure that they're not lying and bring back the right number or the right information, then we're off to the races.

Speaker 0

我们可以开始了。

We're good to go.

Speaker 1

是的,完全正确。

Yep, absolutely.

Speaker 1

所以我的希望是——

So my hope is that-

Speaker 0

这是一个很棒的项目。

It's a cool project.

Speaker 1

Primal 团队实现了这些功能,然后我只需把 Anti-Primal 下架,发表一下我的看法,这样大家都能受益。

Primal team implements this stuff, and then I just take anti Primal down, and I made my little statement about it, and then, you know, everyone wins.

Speaker 0

我们为什么必须输入呢?

We might why do we have to input?

Speaker 0

你为什么不能直接去做呢?

Why can't you just you do it?

Speaker 1

你可以这么做,但我怀疑由于是单个IP地址发出大量请求,可能会遇到速率限制的问题。

You could, but I suspect that there will be some problems on like, with rate limiting between the two because it's a single IP address making a ton of requests.

Speaker 1

我想我们可以试着去,你知道的,解决一下?

I guess we could just work to like, you know?

Speaker 0

是的,我们很可能会通过这个渠道被滥用。

Yeah, we probably will get abused through that channel.

Speaker 1

对。

Yeah.

Speaker 1

所以,得想想怎么处理它。

So, gotta figure out what to do with it.

Speaker 0

我的意思是,如果从商业角度来看,这完全可以做成一个付费API产品。

I mean, I think, just putting my business hat on, it could very easily be a paid API product.

Speaker 0

而且,如果你不喜欢把它做成付费API产品,它也是开源软件。

And look, if you don't like that as a paid API product, it's open source software.

Speaker 0

你可以加载自己的缓存服务器。

You can load your own caching server.

Speaker 0

但我会愿意支付比特币,让我自己的代理能够实时快速查看Noster上发生的一切。

But like I would pay, I would pay Bitcoin for my agent to be able to have a real time quick view of whatever's going on in Noster.

Speaker 0

但无论如何,现在还处于早期阶段,这些都不必马上做。

But anyway, such early days now that none of that has to be done.

Speaker 0

这只会限制增长和其他方面的发展。

That's just limits growth and everything else.

Speaker 0

它可以缓解这个问题,这再次完美地说明了我

It could ease that that's a that's a perfect example once again, of where I

Speaker 1

我的意思是,这正是DVMs所做的事情。

mean, that's kind of what DVMs were.

Speaker 1

对吧?

Right?

Speaker 0

是的。

Yeah.

Speaker 0

嗯,你有

Well, you have

Speaker 1

DVMs 的概念。

the idea of DVMs.

Speaker 0

这是你之前讲的故事,对吧?

It's your story before, right?

Speaker 0

也就是说,代理们会在哪里向其他代理付费?

It's like, okay, where are agents gonna pay other agents?

Speaker 0

如果我需要获取 Noster 的信息,我会付费,比如几笔小额交易来获取这些信息吗?

Well, if I need to get information on nonster, am I gonna pay, you know, a couple stats to get that information?

Speaker 0

还是我会自己搭建一个缓存服务器并托管它?

Or am I gonna build my own caching server and host it?

Speaker 0

不,你只是付几笔小额交易费用就行了,对吧?

Like, no, you're just going to pay a couple of stats transcriptions, right?

Speaker 0

也就是说,目前你是自己做转录,还是让你的代理去支付转录费用?

Like, like as is, are you going to do transcriptions or you're just going to have your agent pay for the transcriptions?

Speaker 0

你知道,这对所有这些不同的任务来说都是一场成本效益分析。

You know, it's a cost benefit analysis for all those different tasks.

Speaker 0

而这可能就是你看到价值被捕捉的地方。

And that's probably where you see some value get captured.

Speaker 1

是的。

Yep.

Speaker 1

我同意。

I agree.

Speaker 0

想想这个真是令人惊讶。

It was wild to think about.

Speaker 0

我的大脑一直在飞速运转。

My mind's just been racing.

Speaker 0

我妻子有点抱怨了。

My wife's been complaining a little bit.

Speaker 1

对。

Yeah.

Speaker 1

最近发生了好多事。

There's been a lot going on.

Speaker 1

世界上现在有太多事情发生了,什么

There's so much happening in the world right What

Speaker 0

你还想聊点什么?

else do wanna talk about?

Speaker 1

是的。

Yeah.

Speaker 1

我不知道。

I don't know.

Speaker 0

我的意思是,这简直是个完美的例子。

I mean, this is like a perfect example.

Speaker 0

所以,我很想知道你的想法。

Like, I so I am curious in your thoughts here.

Speaker 0

我想,我不知道。

I want I don't know.

Speaker 0

其实我喜欢这样,我一直有点赌徒心态。

There's, I like to, there's, there's the freaks know there's always been a little bit of degen in me.

Speaker 0

所以我总说要保持谦逊,统计数据,这是提醒我自己要保持谦逊。

And that's why I always say stay humble stack stats because it's a reminder to myself to stay humble.

Speaker 0

所以,我用这个机器人做的第一件事就是尝试设置艾伦市场的交易。

So like I, the first thing I did with this bot was try and set up the Ellen markets trading.

Speaker 0

但进展并不顺利。

And it hasn't been going that well.

Speaker 0

当然,如果他把我给他的钱全输光了,家人也不会介意。

Obviously, family would be fine if he loses all his money that I gave him.

Speaker 0

我花在代币上的钱比其他地方还多。

I'm spending more on tokens than I am.

Speaker 0

有趣的是,我送出21000聪的莎士比亚故事,竟成了最准的价格预测。

I mean, the funny thing about the Shakespeare story with me giving out 21,000 sats is the best price prediction.

Speaker 0

我为了在网站上追踪这个项目而花在代币上的钱,比奖金本身还多。

I paid more on tokens vibing the site to track it than I did for the prize.

Speaker 0

所以这全是一种平衡。

So it's all a balancing act.

Speaker 0

但我希望他能认真对待Polymarket。

But I want him to treat polymarket.

Speaker 0

对吧?

Right?

Speaker 0

为什么呢?

And why?

Speaker 0

我不知道。

I don't know.

Speaker 0

我只是觉得,如果我的机器人管家在主动交易导弹袭击之类的东西,那会很酷。

I just think it'd be I just think it'd be cool if my, like, robot butler was actively trading, like, missile strikes and shit.

Speaker 0

这是一个疯狂的世界,我只是感到好奇。

It's just a wild world to live in, and I'm just curious.

Speaker 0

但他根本不了解世界上正在发生什么。

But he doesn't understand what's going on in the world.

Speaker 0

他对世界上正在发生的事情一无所知。

He has no context of what's going on in the world.

Speaker 0

这对我来说才是真正有趣的地方。

And this is to me where it gets really interesting.

Speaker 0

因为目前,我尝试解决这个问题的方法之一是让他调用Perplexity的封闭API。

Because right now, one of the ways I'm trying to solve it is he's hitting perplexities closed API.

Speaker 0

Perplexity正在向他提供有关世界动态的重要排名新闻。

They are giving perplexities, giving him like important rank news of what's going on.

Speaker 0

但回到我们之前的观点,这种方式显然非常不完整,而且显然被信任了。

But to our earlier point, and it's complete as way incomplete and it's obviously trusted

Speaker 1

Perplexity是什么?

What's walled Perplexity?

Speaker 0

什么?

What?

Speaker 1

Perplexity是什么?

What's Perplexity?

Speaker 0

Perplexity 是一家大型人工智能科技公司。

Perplexity is one of these big AI tech companies.

Speaker 0

你不知道 Perplexity 吗?

You don't know Perplexity?

Speaker 0

不知道。

No,

Speaker 1

不知道。

no.

Speaker 1

他们的 API 对你有什么用?

What does their API do for you?

Speaker 0

它只是输出相关的新闻。

It just spits out relevant news.

Speaker 1

哦,原来如此。

Oh, okay.

Speaker 0

它只是输出,非常简单。

It just spits out, it's very simple.

Speaker 0

它只是推送相关的新闻,然后他就靠这个了解发生了什么。

It just splits out relevant news, and then that's how he knows what's going on.

Speaker 0

如果他不知道航空母舰的位置,他怎么知道美国会不会袭击伊朗呢?

How can he trade whether or not The US is gonna strike Iran if he doesn't know where the aircraft carriers are.

Speaker 1

是的。

Yes.

Speaker 0

在一个越来越被机器人垃圾主导的世界里,他不能只是扫描推特,因为那可能导致信息错误

How is he gonna and so in a world increasingly dominated by bot slop, and he can't just scan Twitter because that causes information might

Speaker 1

可能是错的。

be wrong.

Speaker 0

是的。

Yeah.

Speaker 0

而他正是基于这些信息进行交易或做决策。

And he's trading on it, or he's making decisions based on it.

Speaker 0

所以这就是Nastar真正有趣的地方,因为所有内容都被哈希并签名,你可以在上面构建信任网络和可验证的声誉。

So this is where Nastar gets really interesting because everything's hashed and signed, and you can build webs of trust and verifiable reputation on it.

Speaker 0

这些话听起来很酷,但对人类大脑来说实在太复杂了,而对智能代理来说却并不算难。

That was all very the words sound cool, but it's also way too complicated for the human brain to deal with, but it's not too complicated for an agent to deal with.

Speaker 0

因此,那些有效运行智能代理的人有赚钱的机会,这些代理实际上正成为API,它们生成关于世界动态的已签名事件。

And so there's money out there for people that you know, that are are effectively probably running agents that are effectively becoming APIs because they're producing signed events of what's going on in the world.

Speaker 0

机器人是如何了解世界正在发生什么的呢?

Like, how does a how does a robot know what's going on in the world?

Speaker 0

人类又是如何了解世界正在发生什么的呢?

How does a human know what's going on in the world?

Speaker 0

这是一个很难解决的问题。

And that's a hard problem to solve

Speaker 1

确实,难题越来越难了,因为你根本无法分辨。

and hard Yep, problems are it's getting more and more difficult because you can't really tell.

Speaker 1

我认为这也解释了为什么人们正转向本地化而非全球化,因为你已经无法再分辨什么是真实的,什么不是。

And I think this is also a point towards why people are moving towards local instead of global because you just can't tell anymore what's real and what isn't.

Speaker 1

所以,别再盲目相信你周围的一切了。

Just don't So need to else tell trust what's around you.

Speaker 1

也许这其实是好事,因为这实际上正是人类数百万年来的生活方式。

And maybe that's good because that's kinda actually how people have been living for millions of years.

Speaker 1

也许仅仅体验身边发生的事,对我们的心理更健康一些。

Maybe it's a little bit healthier for our psychology to be just experiencing what's around us.

Speaker 0

我不觉得这肯定更健康,但我觉得人们已经尝到了了解全球动态的滋味,大多数人不会放弃这种体验。

I just don't think it's definitely healthier, but I don't think, I think people have gotten a taste of knowing what's going on in a global And most people are not going to give that up.

Speaker 0

如果我们不解决真相的问题,他们就只会——而这正是我们已经看到的——一直吞咽着虚假信息。

And so if we don't solve the truth answer, then they're just going to be, and this is what we're already seeing is they're just going to be digesting fake shit all the time.

Speaker 0

我举个例子,你其实面临两种脱节,对吧?

And one example I've had, like I just and so you have two disconnects, right?

Speaker 0

有一大群人仍然相信他们在互联网上读到的每一件事,这简直太疯狂了,绝对疯狂。

You have the person that a huge portion of the population just believes everything they read on the internet still crazy, absolutely crazy.

Speaker 0

而另一大群人则完全放弃了信任。

And then you have another huge portion that's completely checked out.

Speaker 0

他们不相信自己在互联网上看到的任何东西。

So it doesn't believe anything that they see on the internet.

Speaker 0

每件事都像是在让你注册。

Everything's a sign up.

Speaker 0

我在另一个节目里和马蒂开玩笑说,如果我们看到新闻报道说外星人登陆地球,外星人被目击并降落在地球上。

And I joke about this on my other show with Marty, that if we saw news reports that aliens have landed on Earth, aliens have been seen and land on Earth.

Speaker 0

首先,那些视频很可能都是假的。

First of all, those videos are probably fake.

Speaker 0

我们不会相信它们。

We're not gonna believe them.

Speaker 0

外星人必须真的走进房间,看着我说:‘我是外星人。’

The alien would literally have to walk into the room And like, look at me and be like, I'm an alien.

Speaker 0

人们才会相信这真的发生了。

I'm here to invade you for people to believe that it was actually happening.

Speaker 0

如果我正在直播我的播客,外星人就在这儿走进来,马蒂在另一边看着外星人,他可能也不会相信,除非外星人走到他那边的演播室里。

And if I was live on my podcast, and the alien walked in right here, and Marty was on the other side, looking at the alien, he also wouldn't believe it probably until the alien walked into his studio on his side.

Speaker 1

对。

Yep.

Speaker 0

这太疯狂了。

Which is crazy.

Speaker 0

就像人们完全对此视而不见。

Like, that's how much of and people just turn that off.

Speaker 0

他们只是接受了这一切。

They just they just accept that.

Speaker 0

就好像我们生活在一个后真相时代。

It's like, oh, we're in a post truth world.

Speaker 0

现在就是这样。

That's what it is now.

Speaker 0

所以我认为,这是一个越来越重要的趋势,值得关注,那就是可验证的真相。

So that's gonna be, I think, that's an increasingly important trend to watch, which is effectively verifiable truth.

Speaker 0

什么是可验证的真相?

Like what is verifiable truth?

Speaker 0

我们该如何传播它?

How do we distribute it?

Speaker 0

我们如何验证它?

How do we authenticate it?

Speaker 0

代理如何使用它?

How do agents use it?

Speaker 1

我确实接受我们生活在一个后真相世界,因为我实在想不出其他可能性。

I do accept that we're in a post truth world because I just don't think there's any way.

Speaker 1

我并不希望如此,但当你仔细思考时,你如何信任他人?那可以成为你的主观真相,而可验证的签名至少能帮助解决这个问题。

I don't wish for it to be, but when you think it through, how can you You can trust another person and then that can be your subjective truth and verifiable signatures help with that at least.

Speaker 1

You

Speaker 0

所以让我给你举个例子,一个目前在托尼的机器人上活跃的Nostrad例子。

still So let may use me give you an example, a Nostrad example, that's active right now on Tony, on my bot.

Speaker 0

有一个叫War Monitor的Nostrad账号。

There is a Nostrad account called War Monitor.

Speaker 0

你认识War Monitor吗?

Do you know War Monitor?

Speaker 1

我不知道。

I don't.

Speaker 1

我觉得

I think

Speaker 0

他住在以色列。

he lives in Israel.

Speaker 0

他从未说过,我并不是想曝光他的个人信息。

He's never said I'm not trying to dox him.

Speaker 0

我的直觉只是来自他的公开帖子。

My intuition is just from his public posting.

Speaker 0

但他报道了各种类型的战争动态。

But he covers all different types of war movements.

Speaker 0

他是这类开源情报账号之一,对吧?

He's one of these open source Ocent accounts, right?

Speaker 0

Twitter 上有很多这样的开源情报账号。

And Twitter has a lot of them open source intelligence accounts.

Speaker 0

所以他们收集了大量的原始资料和二手资料,然后把这些内容整合成一个人们阅读的信息流。

So like they're taking a bunch of primary sources and secondary sources, and then they're like collapsing it into a feed that people are reading.

Speaker 0

而大部分内容都在推特上。

And most of it's on Twitter.

Speaker 0

再说一遍,推特的API是受限制的。

Once again, Twitter is a gated API.

Speaker 0

这很麻烦,没有签名,也没有哈希验证。

It's a pain in the ass, nothing signed, nothing's hashed.

Speaker 0

和代理配合起来非常困难,而且成本高昂。

It's really hard to work with agents and very expensive.

Speaker 0

而且你随时可能被屏蔽掉。

And you can just get walled off at any time.

Speaker 0

他是少数在Nostr上活跃的开源情报账号之一。

And he's one of the few OSINT accounts that's active on Nostr.

Speaker 0

而且他还建立了一定的声誉。

And he also just, he's built up a bit of a reputation.

Speaker 0

我关注他有一段时间了。

I've been following him for a while.

Speaker 0

他基本上是看到什么就说什么。

He mostly calls it as he sees it.

Speaker 0

我从未发现他撒过谎。

I haven't called him in any lies.

Speaker 0

通常当他撒谎或出错时,事后都会更正记录。

Usually when he lies and when he gets something wrong, he corrects the record after the fact.

Speaker 0

我之所以提到他住在以色列,是因为在以色列和伊朗持续交战期间,当美国轰炸他们的掩体或其他情况时,你能从他的帖子中看出一些以色列的偏见。

Now, the reason I brought up that he lives in Israel is particularly during the Israel Iran active conflict back and forth when The US bombed their bunker or whatever, you could tell there was an Israeli, a bit of an Israeli bias in his posts.

Speaker 0

每个人都有偏见。

Everyone has a bias.

Speaker 0

人类都有偏见。

Humans have biases.

Speaker 0

事实就是如此。

It is what it is.

Speaker 0

他发过一些帖子,你可以看出他至少在那个地区。

And he had some posts where you could tell he was in at least the region.

Speaker 0

也许他在约旦,你知道的,但你能感觉到他就在那个地区。

Maybe he was in Jordan, you know, but you like, you could feel like he was in the region.

Speaker 0

但不管怎样,长话短说,我把他的Nostr信息流接入了我的代理系统。

But anyway, long, long story long, I am feeding his Nostril feed into my agent.

Speaker 0

所以我的代理系统对世界上正在发生的事情有了一些了解。

So my agent has some idea of what's going on in the world.

Speaker 0

这是他的信息来源之一。

It's one of his sources.

Speaker 0

我告诉他,这个信息源带有以色列倾向。

And I told him it has an Israeli bias.

Speaker 0

带有轻微的以色列倾向。

Has a slight Israeli bias.

Speaker 0

请注意这一点。

Keep that in mind.

Speaker 0

但我认为这正是一个完美的例子,首先,我不认识这个人。

But I think that's a perfect example of, first of all, I don't know who this person is.

Speaker 0

他不是我的邻居。

He is not my neighbor.

Speaker 0

但他建立了一种声誉,让我对他有一定的信任。

But he has built up a reputation that I have some level of trust in him.

Speaker 0

我理解他的偏见。

I understand his bias.

Speaker 0

我把这一点考虑进去了。

I'm incorporating that into it.

Speaker 0

如果他破坏了这种信任,比如主动散布错误信息、撒谎或搞点击诱饵——虽然他目前还没这么做——那我就会把他剔除。

And if he breaks that trust, if he actively is out there just spreading mis information or lying or trying to do clickbait, which he doesn't hasn't done yet, then I cut him out.

Speaker 0

他就不会再被输入到我的代理中了。

Then he's no longer being fed into my agent.

Speaker 0

也许这就是为什么情况没那么混乱,也许这就是你开始解决后真相问题的方式。

And maybe that's how it's not as messy, but maybe that's how you start to solve post truth type of situations.

Speaker 0

你静音了。

You're muted.

Speaker 1

我有个关于Polymarket的问题,如果你用这些数据来为类似的事情做决策的话。

I have a question about Polymarket in terms of if this was data you were feeding in to make decisions on something like that.

Speaker 1

我以前从来没用过。

I've never used it before.

Speaker 1

它是单纯的下注,最后才有 payouts,还是说你是在买卖股票,你是在

Is it just betting and then at the end there's a payout or is there are you buying and trading stocks in your You're

Speaker 0

买卖交易。

buying and trading.

Speaker 0

首先,所有操作都在一条叫Polygon的山寨币链上进行,这就是Polymarket名字的由来。

So first of all, it's all on chain on a on on a shitcoin chain called Polygon, hence Polymarket.

Speaker 0

大多数人通过中心化的网页界面使用它,因为链上的山寨币操作非常困难和复杂。

Most people interact with it through like the centralized web interface because shitcoin on chain stuff is really difficult and complicated.

Speaker 0

但如果你告诉智能代理:我想在Polymarket上交易。

But if you tell the agent, I want you to trade on Polymarket.

关于 Bayt 播客

Bayt 提供中文+原文双语音频和字幕,帮助你打破语言障碍,轻松听懂全球优质播客。

继续浏览更多播客