BuildSpeak每日 builder 文摘
今日归档生词本关于
🎙 播客Training Data· 2026 年 5 月 5 日· 4,959 词 · 约 25 分钟

Anthropic's Boris Cherny: Coding's Printing Press Moment

SPACE 播放 / 暂停·←→ 上一句 / 下一句
Speaker 100:01 - 00:10
Okay. I'm excited to introduce our next speaker. Show of hands, who here uses Claude Code? Okay. Show of hands, who here has Claude Code psychosis?
Speaker 100:01 - 00:10
好的。我很兴奋来介绍我们的下一位演讲者。举个手,这里有谁在用 Claude Code?好。再举个手,这里有谁得了 Claude Code psychosis(精神错乱)?
Speaker 100:11 - 00:20
Come on, guys. It's okay. It's okay. My my team lovingly says I have Claude Code psychosis, which may or may not be true. We are delighted to have Boris Cerny with us today.
Speaker 100:11 - 00:20
别这样,大家。没关系,没关系。我的团队半开玩笑地说我有 Claude Code psychosis,这话也许是真的,也许不是。我们今天非常高兴 Boris Cerny 能来到现场。
Speaker 100:20 - 00:45
Boris is the creator, the father of Claude Code, and in the process of doing that, has just had a front row seat to reinventing the modern way of software development. And we're really grateful to you Boris for taking the time to speak with us today. We know that the entirety of software development kinda rests on your shoulders. So thank you for taking an hour of your time to be with us today. And interviewing Boris is Lauren Reether from our team.
Speaker 100:20 - 00:45
Boris 是 Claude Code 的创造者,可以说是它的“父亲”;而在打造它的过程中,他也坐在第一排,亲眼见证了现代软件开发方式的重塑。Boris,我们非常感谢你今天抽时间来和我们交流。我们知道,几乎整个软件开发世界某种程度上都压在你的肩上。所以非常感谢你愿意拿出一个小时来到这里。今天采访 Boris 的,是我们团队的 Lauren Reether。
Speaker 100:45 - 00:46
Thank you.
Speaker 100:45 - 00:46
谢谢。
Speaker 200:52 - 00:54
Getting our chairs. You
Speaker 200:52 - 00:54
我们来把椅子摆好。你——
Speaker 300:56 - 01:01
took my you took my opening line. Usually ask who here uses quad code. Those are a lot of hands. That's awesome.
Speaker 300:56 - 01:01
你把我的开场白抢了,你把我的开场白抢了。我一般都会先问这里有谁在用 quad code。举手的人可真不少,太棒了。
Speaker 201:01 - 01:23
Thank you for joining us, Boris. It's very special to have you here. As a room full of builders, I think you are changing building entirely. And so I'm very curious to explore how you think about the future of software, coding, and what we should spend all of our free time on. But I'll give a tiny bit more background on you so that everyone has a little bit more context.
Speaker 201:01 - 01:23
Boris,感谢你加入我们。你能来到这里真的非常特别。面对这一屋子的 builders(构建者、创造者),我觉得你正在彻底改变“构建”这件事本身。所以我非常想深入聊聊,你是如何看待 software(软件)、coding(编程)的未来,以及我们到底该把所有空闲时间花在什么上面。不过在那之前,我会再稍微补充一点你的背景,让大家有更多上下文。
Speaker 201:23 - 01:44
So beyond creating Cloud Code, Boris is very much an engineer's engineer. You were writing a lot of code through your whole career, writing textbooks about code, including programming in TypeScript. And I think last time we chatted, you hadn't written a single line of code in the last year, or at least so far in 2026, which is quite the change.
Speaker 201:23 - 01:44
所以,除了创造 Cloud Code 之外,Boris 也是非常典型的 engineer's engineer(工程师中的工程师)。你在整个职业生涯中都写了很多代码,也写过关于代码的教材,包括 Programming in TypeScript。并且我记得我们上次聊天时,你说自己在过去一年里一行代码都没写,或者至少在 2026 年到目前为止还没写过,这变化确实很大。
Speaker 301:45 - 01:57
There's also a a little known thing. Back in middle school, I wrote a guide about writing basic for TI 83 plus calculators. And I I just I I searched it. It's actually still on the Internet. It's extremely embarrassing, so please don't search it.
Speaker 301:45 - 01:57
还有一件不太为人所知的小事。早在中学的时候,我写过一份关于如何给 TI 83 plus calculators 编写 BASIC 的指南。我刚刚还搜了一下,它居然现在还在 Internet 上。这事非常尴尬,所以请千万别去搜。
Speaker 301:57 - 01:58
But it exists.
Speaker 301:57 - 01:58
但它确实存在。
Speaker 201:58 - 02:01
We will definitely be finding that. So we're
Speaker 201:58 - 02:01
我们肯定会找到这一点。所以我们——
Speaker 402:01 - 02:04
gonna do I'm start with a few questions here. Maybe we'll
Speaker 402:01 - 02:04
要做的是,我先从这里的几个问题开始。也许我们会——
Speaker 202:04 - 02:14
start with a little bit of the history of Cloud Code, how you started it. And then we're gonna have a lot of audience Q and A for this one. And so start thinking about your questions in the back of your head, and would love to turn it over to you all soon.
Speaker 202:04 - 02:14
先从一点 Cloud Code 的历史讲起,你们当初是怎么开始做它的。然后这一场我们会留很多时间给现场观众做 Q and A(问答)。所以大家现在就可以先在脑子里想想自己的问题,我们也很快就想把时间交给大家。
Speaker 302:15 - 02:25
Yeah. And also, real quick, so for people that use Cloud Code, do people use the CLI mostly? Like, okay, majority CLI? That's a lot. Majority desktop?
Speaker 302:15 - 02:25
对。还有,快速问一下,对于使用 Cloud Code 的人来说,大家主要是用 CLI(命令行界面)吗?比如,好的,大多数是 CLI?那很多啊。那大多数是 desktop(桌面端)?
Speaker 302:26 - 02:32
K. Majority Versus code or JetBrains IDE? Okay. That's actually not a lot. Okay.
Speaker 302:26 - 02:32
好。那大多数是 Versus code 或者 JetBrains IDE(集成开发环境)吗?好。其实那也不算很多。好。
Speaker 302:32 - 02:36
Other? I'm like iOS mostly these days.
Speaker 302:32 - 02:36
其他?我感觉我这些天主要是在做 iOS。
Speaker 402:36 - 02:37
Yeah.
Speaker 402:36 - 02:37
对。
Speaker 302:37 - 02:47
Okay. Cool. Yeah. So I started Cloud Code kind of accidentally in a lot of ways. I joined this team back in late twenty twenty four.
Speaker 302:37 - 02:47
好,很酷。对,我开始做 Cloud Code 在很多方面都算是个意外。我是在 twenty twenty four 年末加入这个团队的。
Speaker 302:47 - 02:58
It was sort of this incubator within Anthropic called Anthropic Labs. And the team kind of served its purpose. We created QuadCode, MCP, and the desktop app. It was a team. It was just a few of us.
Speaker 302:47 - 02:58
这有点像 Anthropic 内部的一个 incubator(孵化器),叫 Anthropic Labs。这个团队算是完成了它的使命。我们做出了 QuadCode、MCP,还有 desktop app。那就是一个团队而已,只有我们几个人。
Speaker 302:58 - 03:13
So very much like innovation team. We built the thing that we wanted to build. We disbanded the team. Now the team's actually back together for round two. Mike Krueger, who's the chief product officer at Anthropic and used to be one of the founders at Instagram, so he's leading that right now.
Speaker 302:58 - 03:13
所以它很像一个 innovation team(创新团队)。我们把自己想做的东西做出来了,然后就把团队解散了。现在这个团队实际上又重新组起来,开始第二轮了。Mike Krueger 现在在负责这件事,他是 Anthropic 的 chief product officer,之前也是 Instagram 的创始人之一。
Speaker 303:15 - 03:41
So the reason that I started to work on coding is we felt like there was this product overhang. And I'm guessing people here use that word a lot. But we definitely use this word a lot in kind of within the lab. There's this idea that the model can do all this stuff that no product has yet captured. And in late twenty twenty four, when we were looking at coding, the way that we did coding, the state of the art at the time was type ahead.
Speaker 303:15 - 03:41
我开始做 coding(编程)这件事的原因,是我们觉得这里存在一种 product overhang。我猜在场的人应该也经常用这个词。反正我们在 lab 里确实很常说这个词。它的意思是,模型已经能做很多事了,但还没有任何产品真正把这些能力承接下来。到了 2024 年底,我们在看 coding 这个方向时,当时我们做 coding 的方式、或者说当时的 state of the art(最先进水平),还是 type ahead。
Speaker 303:41 - 03:57
As you open your ID and you press tab and you can complete one line at a time. And that was the thing that Sonnet 3.5 enabled for the first time. But the feeling was we could actually go a lot further than that. And the model was almost ready for the next big step. So we don't have to do type ahead anymore.
Speaker 303:41 - 03:57
也就是你打开 IDE,然后按下 tab,一次补全一行代码。而 Sonnet 3.5 第一次真正让这件事成为可能。但我们的感觉是,我们其实还能走得更远。模型几乎已经为下一个重大跃迁准备好了。所以我们不必再做 type ahead 了。
Speaker 303:57 - 04:15
We can just have the agent write all of the code. And so I built it, and it just really didn't work for the first six months. It was not very good. It was barely used it for maybe 10% of my code or something like that. And even after we released QuadCode initially, it was not a hit.
Speaker 303:57 - 04:15
我们完全可以让 agent(智能体)把所有代码都写出来。于是我就把它做了出来,但前六个月它其实非常不好用,基本没起作用。效果很差,我自己可能也就只在大概 10% 的代码里用它之类的。甚至在我们最初发布 QuadCode 之后,它也并没有爆红。
Speaker 304:16 - 04:33
There's a lot of people that used it, but it did not have this exponential growth that it has today. That started with Opus four in May. And I remember that very clearly. That's like when the exponential growth started and then it kind of inflected with every model release. Like it started with Opus four, then 4.5, then 4.6, now 4.7.
Speaker 304:16 - 04:33
用它的人很多,但它并没有出现今天这种 exponential growth(指数级增长)。这种情况是从 5 月的 Opus four 开始的。我对这一点记得非常清楚。差不多就是从那时起,指数级增长开始了,然后每次模型发布都会带来一次拐点。最早是 Opus four,然后是 4.5,再到 4.6,现在是 4.7。
Speaker 304:33 - 04:50
It just kind of keeps inflecting. But essentially, were trying to build this thing that was like pre PMF. And we knew that it wouldn't have PMF for six months because we were building for the next model. And that was the idea pretty much the whole time. And for Anthropic in general, we've always just been very focused.
Speaker 304:33 - 04:50
它就是这样不断出现新的拐点。但本质上,我们当时是在做一个还处于 pre-PMF 的东西。我们知道它在六个月内都不会有 PMF(product-market fit,产品与市场匹配),因为我们其实是在为下一代模型构建产品。这几乎从头到尾都是我们的思路。而且对 Anthropic 来说,整体上我们一直都非常聚焦。
Speaker 304:50 - 05:03
We've always cared about business and enterprise and safety and coding. That's just always been kind of the way that we wanted to build. And so at some point, we kind of knew that we wanted to build a product. We didn't know exactly when. So this kind of ended up being the product bet.
Speaker 304:50 - 05:03
我们一直都很关注 business、enterprise、安全,以及 coding。这基本就是我们一直想采用的构建方式。所以在某个时点,我们其实知道自己想做一个产品,只是不确定具体什么时候做。于是这件事最后就成了那次 product bet(产品押注)。
Speaker 205:04 - 05:20
It's an incredible story, especially that it was an accident. So you've said on the record that you think coding is solved. If this is one of the three best for Anthropic, can you tell us more about what you mean by that and what might still not be solved or what second order problems might come?
Speaker 205:04 - 05:20
这真是个不可思议的故事,尤其是它居然是一次意外。所以你之前在公开场合说过,你认为 coding(编程)这个问题已经解决了。如果这算是 Anthropic 三个最佳案例之一,你能不能再具体说说你这句话是什么意思,以及还有哪些东西可能仍然没有被解决,或者会出现哪些 second order problems(二阶问题)?
Speaker 305:20 - 05:37
All right, I can ask another question for the room. Who writes 100% of their code by hand? Who writes 100% of their code using an agent, like quad code? Okay, who's somewhere in between? Okay, so like 50% solved.
Speaker 305:20 - 05:37
好吧,我可以换个问题问在场各位。谁写的代码是 100% 手写的?谁写的代码是 100% 用 agent(智能代理),比如 quad code,来写的?好,那么,谁是在这两者之间的?好吧,所以大概算是解决了 50%。
Speaker 305:41 - 05:51
I mean, for me it's 100%. The QuadCode code base, you know, it leaks, so, you know, people know. It's pretty simple. It's just like TypeScript and it's React. Like, there's no big secret.
Speaker 305:41 - 05:51
我的意思是,对我来说是 100%。QuadCode 的 code base(代码库),你知道,会泄露一些信息,所以,大家其实也都知道。它非常简单。基本上就是 TypeScript 和 React。没什么大秘密。
Speaker 305:51 - 06:11
There's nothing really complicated. The the reason we picked TypeScript and React is it's very on distribution for the model. So when we started building the code base, the model was not as intelligent as it is today, so the language and the framework mattered a lot. Nowadays, it can write whatever, and it can pick up new languages, new frameworks, it hasn't seen. But back then, you wanted to use something pretty on distribution.
Speaker 305:51 - 06:11
里面其实没什么真正复杂的东西。我们之所以选择 TypeScript 和 React,是因为它们对 model(模型)来说非常 on distribution(处于训练分布内)。我们刚开始搭这个 code base 的时候,model 还没有今天这么聪明,所以语言和 framework(框架)非常重要。现在的话,它基本上什么都能写,也能学会它没见过的新语言、新框架。但在当时,你会希望用一些比较 on distribution 的东西。
Speaker 306:12 - 06:31
Because of that, I think fairly early, we got to the point where the model just wrote 100% of the code. And for us, this happened sometime in October, November. And so for me today, the model writes 100% of my code. I write somewhere, usually a few dozen PRs every day. There was a day last week I did 150 PRs in a day.
Speaker 306:12 - 06:31
正因为如此,我觉得我们相当早就到了 model 直接写 100% 代码的阶段。对我们来说,这大概发生在 10 月、11 月那段时间。所以对今天的我来说,model 写我 100% 的代码。我每天通常会写,或者说产出几十个 PR(pull request,拉取请求)。上周有一天,我一天做了 150 个 PR。
Speaker 306:31 - 06:41
That was a record. I was just trying to push to see how far I can get it. But yeah, for me, it's just solved. But this is not the case everywhere. There's very big complicated code bases.
Speaker 306:31 - 06:41
那是个纪录。我当时只是想往前推一推,看看我到底能把它推进到多远。不过对我来说,这件事就是已经解决了。但并不是所有地方都是这样。有些 code base 非常庞大,而且非常复杂。
Speaker 306:41 - 06:49
There's kind of weird languages the model's not good at yet. And then, you know, as everyone here knows, it's it's getting there. Usually, the answer is just wait for the next model.
Speaker 306:41 - 06:49
还有一些比较奇怪的语言,model 现在还不太擅长。然后,你知道,在座各位都明白,它正在接近那个点。通常,答案就是等下一个 model。
Speaker 206:49 - 06:55
Can you actually tell us about your personal setup? You walked us through it the other day. It is pretty wild.
Speaker 206:49 - 06:55
你能不能具体讲讲你个人的 setup(工作配置)?前几天你给我们演示过一遍。那真的挺疯狂的。
Speaker 306:55 - 07:05
Yeah. So I shared my personal setup like six months ago or something on on Twitter. And it it's funny. I actually I shared it. I didn't realize that it would be surprising for anyone.
Speaker 306:55 - 07:05
对。所以我大概在六个月前之类的时候,在 Twitter 上分享过我的个人 setup(配置)。很有意思的是,我当时其实只是分享了一下,没意识到这对任何人来说会是件让人惊讶的事。
Speaker 307:05 - 07:07
That was just like the way that I coded.
Speaker 307:05 - 07:07
那就是我写代码的方式。
Speaker 207:08 - 07:09
And it's changed since then.
Speaker 207:08 - 07:09
不过从那以后,它已经变了。
Speaker 307:09 - 07:16
It's changed. And so now, actually most of my work I do from my phone. And so,
Speaker 307:09 - 07:16
变了。所以现在,其实我的大部分工作都是在手机上完成的。所以,
Speaker 507:16 - 07:17
I don't know
Speaker 507:16 - 07:17
我不知道
Speaker 307:17 - 07:33
if you guys won't be able to see this, but I have, So I have the Quad app. And if you open the Quad app, on the left hand side, there's this little code tab. And I just have a bunch of sessions going. You probably can't see it.
Speaker 307:17 - 07:33
你们能不能看清这个,不过我装了 Quad app。然后如果你打开 Quad app,在左边有一个小小的代码 tab(标签页)。我就在里面开着一堆 session(会话)。你们可能看不清。
Speaker 207:33 - 07:34
How many sessions?
Speaker 207:33 - 07:34
有多少个 session(会话)?
Speaker 307:34 - 07:48
Usually, I have maybe five to 10 sessions. And then the sessions usually have a bunch of agents. So I think currently, probably a few 100 agents going. Usually, every night, have a few thousand that are doing deeper work. There's a few ways to manage it.
Speaker 307:34 - 07:48
通常我会有大概五到十个 session(会话)。然后这些 session 里通常又会有一堆 agent(智能体)。所以我觉得目前大概同时在跑几百个 agent。通常每天晚上,还会有几千个 agent 在做更深度的工作。管理它有几种方式。
Speaker 307:49 - 08:09
One is that you ask Claude to use a bunch of sub agents to do work. Actually the thing that I've been finding myself using more and more is loop. So this is slash loop and it's just like the coolest thing. It's like the simplest thing that works. All it is is you have Claude use Cron to schedule a job for some point in the future, and it's a repeat job.
Speaker 307:49 - 08:09
其中一种方式是,你让 Claude 使用一堆 sub agents 来完成工作。其实我最近越来越常用的是 loop。所以这个就是 slash loop,它真的算是最酷的东西之一。它像是那种“最简单但能用”的方案。本质上就是让 Claude 用 Cron 把一个任务安排到未来的某个时间点执行,而且这是一个重复任务。
Speaker 308:09 - 08:27
And it can run every minute, every five minutes, every day, kind of however often you want to schedule it. And at this point, I have, like, dozens of loops that are running for stuff. So I have one that's babysitting my PRs, like fixing CI, auto rebasing. I have another one that keeps CI healthy. So, like, if there's, like, a flaky test or whatever, it'll it'll go and fix it.
Speaker 308:09 - 08:27
它可以每分钟运行一次、每五分钟运行一次、每天运行一次,基本上你想设定多高的频率都可以。到现在,我已经有几十个 loops 在跑各种事情了。所以我有一个会帮我盯着我的 PRs,比如修 CI、自动 rebase。我还有另一个用来保持 CI 健康。所以比如说,如果有 flaky test 之类的问题,它就会去把它修掉。
Speaker 308:28 - 08:46
I have another one that grabs feedback from Twitter and clusters it for me every thirty minutes. So I just have a bunch of these loops running at any time. I feel like loops are the future at this point. If you haven't experimented with it, highly, highly recommend it. And we also just launched routines, which is the same thing but kind of on the server.
Speaker 308:28 - 08:46
我还有另一个会每三十分钟抓取一次 Twitter 上的反馈,并帮我把它们聚类。所以我随时都有一堆这样的 loops 在运行。我现在感觉 loops 就是未来。如果你还没有试过,我强烈、强烈推荐你去实验一下。我们最近还发布了 routines,本质上是同样的东西,只不过更像是跑在 server 端。
Speaker 308:46 - 08:49
So even if you close your laptop, it keeps going.
Speaker 308:46 - 08:49
所以即使你把 laptop 合上,它也会继续运行。
Speaker 208:50 - 09:03
So that's your personal setup. Tell us about what you think teams will look like in the future. How do you extrapolate from all the work you're doing to keep everyone on the team moving forward, understanding the context? Or do you think we need to let go of a lot more to agents to make it work?
Speaker 208:50 - 09:03
所以这说的是你个人的工作配置。那谈谈你认为未来团队会是什么样子吧。你如何把自己为了让团队里每个人都持续推进、都理解上下文所做的这些工作,外推到未来?还是说,你觉得为了让这一切真正运转起来,我们需要把更多事情交给 agents?
Speaker 309:05 - 09:27
I think so. I'd say it's so hard to make predictions, but I'm here to make predictions, so I'll try to make some. I feel like the way that things are going is generally there's going be a lot more generalists than there are today. And today when we talk about generalists, I think largely we're talking about people that are still engineers. So they're still writing code, but maybe they're kind of product engineers.
Speaker 309:05 - 09:27
我觉得是的。要我说,做预测其实很难,但我既然坐在这里就是来做预测的,那我就试着做一些。我感觉事情的发展方向总体上会是:generalists 会比今天多得多。而今天当我们谈 generalists 时,我觉得很大程度上说的仍然是工程师。所以他们仍然在写代码,只不过也许更偏 product engineers。
Speaker 309:27 - 09:52
Maybe when we say generalists, it's like they do iOS and web and server, for example. That's like a generalist in engineering. But I think the thing that we're going to start to see a lot more of is generalists that are cross disciplinary. So this is engineers that are really good at product engineering, but also really great at design, or really great at product and data science and engineering. I don't know, it's something that we're starting to see on our team.
Speaker 309:27 - 09:52
也许当我们说 generalists 时,指的是比如他们同时做 iOS、web 和 server。这算是工程领域里的 generalist。但我觉得接下来我们会越来越多看到的,是跨学科的 generalists。也就是说,这些工程师既非常擅长 product engineering,同时也非常擅长 design,或者非常擅长 product、data science 和 engineering。怎么说呢,这已经是我们团队里开始出现的趋势了。
Speaker 309:52 - 10:22
So actually, like a lot of people on the Cloud Code team are generalists across disciplines. Everyone on our team codes. So our engineering manager, our product manager, our designers, our data scientists, our finance guy, our user researcher, every single person on our team writes code. And so, you know, like there's specialists in something, but now also everyone's just coding. And, you know, I'm seeing some nods, but I bet also it's actually not that surprising to people in this room because I bet you're seeing the same things.
Speaker 309:52 - 10:22
所以其实,Cloud Code 团队里很多人都是跨学科的 generalists。我们团队里的每个人都会写代码。所以我们的 engineering manager、product manager、designers、data scientists、finance guy、user researcher,团队里的每一个人都写代码。所以你知道,虽然大家各自都有某方面的专长,但现在同时每个人也都在写代码。我看到有些人在点头,不过我想这对这个房间里的人来说其实也未必那么令人意外,因为我猜你们也看到了同样的情况。
Speaker 210:24 - 10:44
I'll have one more thread of questions and we'll open up to the audience. So we talked a bit about what's changing with coding. I'm curious about what you see changing in the world of software or software products. I think as we see AI making writing code 10 or 100 x cheaper, what happens to the value of the products that are produced with software? Do we have a SaaSpocalypse on our hands?
Speaker 210:24 - 10:44
我再问最后一组问题,然后我们就开放给观众提问。前面我们聊了一些 coding 方面正在发生的变化。我很好奇,在 software 或 software products 的世界里,你看到哪些东西也在变化。我想,随着 AI 让写代码的成本降低到原来的 10 倍甚至 100 倍,会发生什么?那些用 software 生产出来的产品,它们的价值会怎样变化?我们是不是正在迎来一场 “SaaSpocalypse”?
Speaker 210:44 - 10:47
How do you think this plays out? And again, you're gonna have make another prediction.
Speaker 210:44 - 10:47
你觉得这会如何展开?还有,这次你又得再做一个预测了。
Speaker 310:48 - 11:09
The SaaSpocalypse question is my favorite question then. I think there's two things that are going to happen, and I don't think either of them is the thing that people have been talking about. Think one is Is anyone here an Acquired listener, like the Acquired podcast? Yeah. It's like the best podcast.
Speaker 310:48 - 11:09
那我最喜欢的就是这个 SaaSpocalypse 的问题了。我觉得会发生两件事,而且我认为这两件事都不是人们一直在讨论的那些。先说第一件:这里有人听 Acquired 吗,就是 Acquired podcast?有啊。那真的是最好的 podcast 之一。
Speaker 311:10 - 11:41
I actually I I got to do an unplugged with them the other week, and I I just I I felt like I got to, like, meet my heroes because they're they're just like the hosts are the best. So they have this idea of seven powers and and this is this is like Hamilton. He kind of wrote he wrote a book about this and this is kind of the seven motes in business. And I think what's gonna happen is because of AI, some of these motes are gonna get more important and some are gonna get less important. And so, like, for example, one that gets less important is switching costs because you can just use the model and you can kind of port from one thing to a different thing.
Speaker 311:10 - 11:41
前阵子我其实还和他们一起做了一场 unplugged,我当时真的有种见到偶像的感觉,因为他们——两位 host——真的都特别棒。他们有一个叫 seven powers 的框架,这个概念来自 Hamilton。他写过一本关于这个的书,可以把它理解为商业里的七种 moat(护城河)。我觉得接下来会发生的是:因为 AI,这些 moat 里有些会变得更重要,有些会变得没那么重要。比如说,没那么重要的一种就是 switching costs(转换成本),因为你可以直接用 model(模型),把一个东西比较容易地 port(迁移)到另一个东西上。
Speaker 311:41 - 12:05
Another one that gets less important is process power because for companies whose moat is like a workflows and process and things like this, Quad is getting really good at figuring out process. And especially with 4.7, it can just hill climb anything. So if you give it a target and you tell it to iterate until it's done, it'll just do it. I think this is the first model like that. So I think these are gonna get less important, but I think the previous moats actually still matter.
Speaker 311:41 - 12:05
另一个会变得没那么重要的是 process power,因为有些公司的 moat 就在于 workflow(工作流)、process(流程)这类东西,而 Quad 现在在理解和推演 process 方面已经变得非常厉害了。尤其是到了 4.7,它几乎可以对任何事情做 hill climb(爬山式迭代优化)。所以如果你给它一个目标,再告诉它不断 iterate(迭代)直到完成,它就会一直做下去。我觉得这是第一个具备这种能力的 model。所以我认为这些东西的重要性会下降,但我也认为,之前那些 moat 实际上仍然重要。
Speaker 312:05 - 12:29
So this is like network effects, scale economies, cornered resources, things like that. These are not really changing with AI. I think the second thing is if you look at the number of startups today or maybe in the past ten years, I think the number of startups in the next ten years that are just going to disrupt everything is going to increase 10x. Because right now, you can be a tiny startup. You could build a thing that's as valuable as a large company.
Speaker 312:05 - 12:29
比如 network effects(网络效应)、scale economies(规模经济)、cornered resources(独占资源)之类。这些东西其实不会因为 AI 而发生根本变化。第二点是,如果你看看今天,或者说过去十年里的 startup 数量,我认为接下来十年里,那种会颠覆一切的 startup 数量会增加 10 倍。因为现在,你完全可以是一家很小的 startup,却做出一个价值能和大公司相当的东西。
Speaker 312:29 - 12:39
And you can actually compete head to head. Because the large company has to evolve their business process. They have to evolve the way they work. They have to retrain everyone to use technology. They're going face a lot of internal resistance to that.
Speaker 312:29 - 12:39
而且你实际上可以和大公司正面竞争。因为大公司必须去调整自己的 business process(业务流程),必须改变自己的工作方式,还得重新培训所有人来使用新技术。他们在内部会面临非常多的阻力。
Speaker 312:40 - 12:52
But, you know, no one here has that problem. If you're starting fresh, then you can kinda build with AI natively from the ground up. So I don't know. I I think it's the best time to build. It's the best time to be a start up.
Speaker 312:40 - 12:52
但你知道,在场各位没有这个问题。如果你是从零开始,那你就可以从一开始就以 AI-native 的方式自底向上地构建。所以我也说不好,但我确实觉得,现在是创业最好的时机,是做 startup 最好的时机。
Speaker 312:52 - 12:54
It's there's so much disruption coming.
Speaker 312:52 - 12:54
现在有太多颠覆性的变化即将到来。
Speaker 212:54 - 13:03
So there is hope for us after all. Thank you, Boris. I would love to open up to audience questions if anyone has anything they would like to ask. Dan?
Speaker 212:54 - 13:03
所以归根结底,我们还是有希望的。谢谢你,Boris。如果有人有想问的问题,我很想现在把时间开放给观众提问。Dan?
Speaker 513:08 - 13:24
Hi. Yeah, I'm curious. You said that you built six months before there was product market fit. But now, given that the models are good enough, how much do you attribute the success of QuadCode to the model versus product decisions and the feel of the product?
Speaker 513:08 - 13:24
嗨。对,我很好奇。你说过,在出现 product-market fit(产品市场契合)之前,你们先做了六个月。但现在,既然这些 model(模型)已经足够好了,你认为 QuadCode 的成功有多大程度应归因于 model,又有多大程度归因于产品决策以及产品的整体体验?
Speaker 313:25 - 13:35
I think it's probably a mix. Yeah, I think it's a mix. I think if you asked maybe a year ago, the ratio was maybe something like fiftyfifty. Maybe, I don't know, if you asked me six months ago, the mix would be fiftyfifty.
Speaker 313:25 - 13:35
我觉得大概是混合因素。对,我觉得是个组合。我想如果你一年前问我,这个比例可能大概是五五开。也许,我也不确定,如果你六个月前问我,这个组合也会是五五开。
Speaker 513:35 - 13:36
What about in two years?
Speaker 513:35 - 13:36
那两年后呢?
Speaker 313:37 - 13:39
Oh, two years, I don't know, dude. We plan in like, we plan one week out.
Speaker 313:37 - 13:39
哦,两年后,我不知道啊,老兄。我们的规划基本上就是——我们通常只往后一周做计划。
Speaker 513:39 - 13:41
Six months, sometime in the future.
Speaker 513:39 - 13:41
六个月吧,未来某个时候。
Speaker 313:42 - 14:13
And by the way, I think the reason it was fiftyfifty is, you know, I did YC back in the day. I was like the first hire at a YC company and I did a bunch of startups. And in startups, the thing that they drill into you, and especially in YC over and over, is build something people love. And so it doesn't matter what the product is, it doesn't matter the model and all this stuff, you still in the end have to build a thing that people love. And I think that's why the product matters is we pay so much attention to the little details so that as you use it all day, it's a really great experience.
Speaker 313:42 - 14:13
顺便说一下,我觉得之所以是五五开,是因为——你知道,我当年参加过 YC。我曾是一个 YC 公司里的第一个 hire(招聘员工),也做过一堆 startups(初创公司)。而在 startups 里,他们反复给你灌输的一点,尤其是在 YC,一遍又一遍强调的,就是 build something people love(做出人们真正喜爱的东西)。所以不管产品是什么,不管 model 是什么,诸如此类,到最后你还是得做出一个人们喜爱的东西。我觉得这就是为什么产品很重要:我们会非常关注那些细小的细节,这样当你整天使用它时,整体体验才会真的非常出色。
Speaker 314:14 - 14:30
I think as the model has gotten better, the harness kind of gets less important. And I like, I think that we're thinking about right now is like, how do we evolve the harness? So like, how do we make loops more of a first class thing? How do we make it easier to run a lot of agents? Know, beside, know, like sub agents is one idea.
Speaker 314:14 - 14:30
我认为,随着 model 变得越来越好,harness 这类外层框架的重要性某种程度上会降低。而且我觉得,我们现在正在思考的是,如何让 harness 继续演进?比如,怎样让 loop(循环)变成更一等公民(first class)的能力?怎样让同时运行大量 agent 更容易?比如说,sub agents 就是一个思路。
Speaker 314:30 - 14:52
There's a bunch more stuff that we're cooking. But I think in a year, the model will be much better aligned. And so all the safety mechanisms that we have today around prompt injection and kind of static verification of commands and permission modes, human in the loop, all this kind of stuff, it's just gonna be less important cause the model will just do the right thing. So yeah, that's my prediction.
Speaker 314:30 - 14:52
我们还在推进很多别的东西。但我觉得一年之后,model 会对齐得好得多。所以,我们今天围绕 prompt injection、对命令进行某种静态验证、permission modes、human in the loop 等等这类安全机制,到时候都会变得没那么重要,因为 model 本身就会直接做对的事。所以,嗯,这就是我的预测。
Speaker 214:52 - 14:53
Thanks. Thank you.
Speaker 214:52 - 14:53
谢谢。谢谢你。
Speaker 614:55 - 14:55
You want
Speaker 614:55 - 14:55
你想
Speaker 214:55 - 14:56
to toss the box, Dan?
Speaker 214:55 - 14:56
把盒子递给 Dan 吗?
Speaker 414:59 - 15:30
Great. To zoom out a little bit from software, I think Cloud Code did a cultural change a few months ago where it democratized building software. You can see shop owners building their own software for themselves or even programming microcontrollers to control the light when someone opens the door. Do you see in the future building software becoming a skill like I know Microsoft Office? So it's a thing that everybody can do, not just people in the tech industry?
Speaker 414:59 - 15:30
很好。稍微从 software 这个话题往外拉远一点,我觉得几个月前 Cloud Code 带来了一次文化层面的变化,它让构建 software 这件事民主化了。你现在可以看到店主为自己搭建自己的 software,甚至给 microcontrollers(微控制器)编程,让有人开门时灯自动亮起。你是否认为,未来构建 software 会变成一种像“我会用 Microsoft Office”这样的技能?也就是说,它会成为每个人都能做的事,而不只是 tech industry 里的人才会做?
Speaker 315:30 - 15:45
Oh my God. Yes, yes, yes. I think it's going be even more than that. I think it's going to be, I don't know, it's going to be a skill like, yeah, like I know how to send a text message. I I I think, you know, like, I I read my my two genres are essentially sci fi and tech history.
Speaker 315:30 - 15:45
天哪。会,会,会。我觉得甚至还不止如此。我觉得它会变成一种——我不知道——会变成一种技能,就像,嗯,就像“我会发 text message”一样。我我我觉得,你知道,我平时主要读的其实就是两类东西:sci fi 和 tech history。
Speaker 315:45 - 16:04
This is what I read a lot of. And I think in tech history, there's one thing which I think to me is the clearest parallel for what's happening right now. And this is in the fourteen hundreds, the printing press in Europe. And what happened was before the printing press, essentially 10 of the European population was literate. They knew how to read and write.
Speaker 315:45 - 16:04
这是我读得很多的内容。而我觉得,在 tech history 里,有一件事在我看来是当下正在发生的事情最清晰的类比。那就是 1400 年代的欧洲印刷术。之前发生的情况是,在 printing press 出现之前,欧洲人口里本质上只有 10% 是识字的,也就是他们会读会写。
Speaker 316:05 - 16:29
They were often employed by kings and lords that were not literate. And their was to read and write, and this is not something that everyone knew how to do. The printing press was invented, then there were two more presses. And in the fifty years after the first printing press, there was more literature published in Europe than in the thousand years before. And over the same period, the cost of literature, cost of a book went down like a 100 x.
Speaker 316:05 - 16:29
他们过去常常受雇于不识字的国王和领主。他们的职责就是读和写,而这并不是人人都会的技能。后来印刷机被发明出来,接着又出现了更多印刷机。在第一台印刷机出现后的五十年里,Europe 出版的文献比此前一千年还要多。而在同一时期,文献的成本、一本书的价格,大约下降了 100 倍。
Speaker 316:29 - 16:50
And then, you know, it took a couple hundred years because, you know, learning to read and write is hard. You need education systems and government and everyone can't be working on farms and so on. But over the next few hundred years, literacy globally went up to like 70%. And so now we can all read and write and you don't need a degree in reading and writing to know how to read and write. Although still there are professional writers and that is a thing that you can do.
Speaker 316:29 - 16:50
然后,你知道,这又花了几百年,因为,识字会写本来就很难。你需要教育体系、需要政府,而且也不能所有人都一直在农场干活,等等。但在接下来的几百年里,全球识字率上升到了大约 70%。所以现在我们都能读会写了,你也不需要有“读写专业”的学位才知道怎么读写。当然,职业作家依然存在,而且那仍然是一条可以走的职业道路。
Speaker 316:50 - 17:15
So I think the thing that's about to happen, and it's going to be much faster than fifty years, is software will be a thing that is fully democratized that anyone can do. And there's a lot of corollaries to this. So for example, let's say you're writing accounting software. The best person to write accounting software, I think maybe even today, is not an engineer, it's a really good accountant. Because they know the domain really well and coding is the easy part.
Speaker 316:50 - 17:15
所以我认为,接下来即将发生的事,而且速度会比五十年快得多,就是 software 将会成为一种被彻底 democratized(民主化、普及化)的东西,任何人都能做。这还会带来很多推论。比如说,假设你在写 accounting software,我认为写 accounting software 最合适的人,也许甚至在今天,都不是 engineer,而是一个非常优秀的 accountant。因为他们对这个领域懂得非常透彻,而 coding 反而是容易的部分。
Speaker 317:15 - 17:20
It's knowing the domain that's the hard part. And I think this is just obviously the future.
Speaker 317:15 - 17:20
真正难的是理解这个领域本身。而我觉得,这显然就是未来。
Speaker 717:22 - 17:39
So one of the things Greg said was that you guys are living in the future a little bit because you get to have access to the models and the agents. Cloud Code is an internal tool before you release it. Is the gap between where you guys are in engineering and the rest of the world, is that a month? Is it three months? Is it six months?
Speaker 717:22 - 17:39
Greg 刚才说的一点是,你们这些人算是有点生活在未来里,因为你们能够接触到这些 models 和 agents。Cloud Code 在发布之前也是一个内部工具。你们在 engineering 上所处的位置,和世界上其他人的差距,大概有多大?是一个月?三个月?还是六个月?
Speaker 717:39 - 17:42
And is that gap getting bigger or smaller over time?
Speaker 717:39 - 17:42
而且这个差距会随着时间推移变大还是变小?
Speaker 317:43 - 17:59
Yeah. So internally, we use the same models everyone else does. For us, the dogfooding is really, really important. So we use the thing that everyone else here does. We use a little bit of Mythos to try it, and then we use a lot of Opus 4.7 to dogfood it and to write most of our code.
Speaker 317:43 - 17:59
会的。我们内部使用的,其实和其他所有人用的是同样的 models。对我们来说,dogfooding(内部试用自家产品)非常非常重要。所以我们用的就是在场其他人也在用的东西。我们会稍微用一点 Mythos 来试一试,然后会大量使用 Opus 4.7 来做 dogfood,并用它来编写我们大部分代码。
Speaker 318:00 - 18:19
I think on the model side, there isn't really a gap. It's pretty much mythos, and that will become some version of some descendant of that will become available at some point to everyone. I think on the product side, there's probably a far larger gap. And that's just related to us changing all of our processes. Like if you talk to people at Anthropic, we use Quad for literally everything.
Speaker 318:00 - 18:19
我觉得在 model 这一侧,其实并不存在什么真正的差距。基本上就是 mythos,而它的某个版本、某个后继分支,最终总会在某个时候向所有人开放。我认为在产品这一侧,差距可能要大得多。这只是因为我们把自己的所有流程都改掉了。比如如果你去和 Anthropic 的人聊,你会发现我们几乎什么都用 Quad。
Speaker 318:20 - 18:38
And our quads are talking all day. As my quads are coding in a loop, they will communicate over Slack to talk to other people's quads that are also running in a loop to kind of figure out unknowns. We have no more manually written code anywhere at the company. All of the SQL is written by models. Everything is just built by the models.
Speaker 318:20 - 18:38
而且我们的 quads 全天都在交流。我的 quads 在循环中写代码时,会通过 Slack 与其他人的、同样在循环运行的 quads 沟通,以便某种程度上弄清那些未知问题。我们公司里已经没有任何手写代码了。所有 SQL 都是由 models 编写的。所有东西基本上都是由 models 构建出来的。
Speaker 318:38 - 19:06
So I think actually the place that we're ahead is not the technology, because the same technology available to us is available to everyone here. Because fundamentally, we are building a platform. And so for us, it's really important that developers can use the same thing that we're using and that we dog food everything that we put out there. But I think there's actually a far bigger weed in the organizational structure and organizational process. And this is a place where hopefully we can talk about it in places like this, and everyone can learn from it and also evolve.
Speaker 318:38 - 19:06
所以我认为,实际上我们领先的地方并不是 technology(技术),因为对我们可用的同样技术,在这里对每个人也都可用。因为从根本上说,我们是在构建一个 platform(平台)。所以对我们来说,非常重要的一点是,developers(开发者)能够使用和我们自己相同的东西,并且我们会对外推出的所有东西都先自己 dog food(内部先用)。但我觉得,真正大得多的杂草其实在 organizational structure(组织结构)和 organizational process(组织流程)里。我希望我们能在这样的场合讨论这件事,让每个人都能从中学习,也一起演进。
Speaker 219:06 - 19:11
Yeah. And I think that's one of the advantages startups have. It's so much easier to start there. Jared?
Speaker 219:06 - 19:11
对。我觉得这确实是 startups(初创公司)拥有的优势之一。从那里起步要容易得多。Jared?
Speaker 819:12 - 19:23
Yeah. Last time we talked, I think you'd mentioned we talked a little bit about multi agent. And I was very encoded at the time at a prior Sequoia event. And you mentioned that there were some things going down the pipeline. There's a thing you're talking you're thinking about.
Speaker 819:12 - 19:23
对。我们上次聊的时候,我记得你提到过,我们也稍微谈了一下 multi agent(多 agent)。而我当时在之前一次 Sequoia 活动上,对这个还非常“encoded”。你提到说,有一些东西正在推进中,有个你们正在谈、正在考虑的方向。
Speaker 819:23 - 19:29
Now, obviously, there's slash batch. There's slash loop. There's sub teams. There's teams. Can you speak some to either at
Speaker 819:23 - 19:29
现在,显然已经有 /batch,有 /loop,有 sub teams,也有 teams。你能不能谈谈,不管是从
Speaker 319:29 - 19:30
the model level and at
Speaker 319:29 - 19:30
model level(模型层)还是从
Speaker 819:30 - 19:49
the harness level, how you're injecting priors in the harness level, how the objective function is changing at the model level to make this experience around delegating work, spinning up agents better. Because so much of the work is parallelizable. You can go do so many things so much faster. And I feel like I have to overlay my own intuition for when to paralyze things rather than the model understanding that you can spin up tens of agents for something.
Speaker 819:30 - 19:49
harness level(编排层)来看,你们是如何在 harness level 注入 priors(先验),以及在 model level 上 objective function(目标函数)是如何变化的,从而让这种委派工作、拉起 agents 的体验变得更好。因为这里面有太多工作都是可以并行化的。你可以同时去做很多事情,速度快得多。但我感觉,我现在还是得叠加我自己的直觉,来判断什么时候该把事情并行化,而不是由 model 自己理解:其实你完全可以为某件事一下子拉起几十个 agents。
Speaker 319:50 - 20:08
Yeah, I mean on product side it really just comes down to prompting. That's how it is. So we tweak prompts to help the model do stuff in parallel more. But also honestly, as the model gets better it just naturally does this. And so something like Loop, I found actually 4.7, it just starts doing, which is really cool.
Speaker 319:50 - 20:08
对,我是说,在 product(产品)这一侧,归根结底其实就是 prompting(提示词设计)。现在基本就是这样。所以我们会调整 prompts,帮助 model 更多地并行处理事情。但说实话,随着 model 变得更强,它也会很自然地做到这一点。所以像 Loop 这样的东西,我发现其实到了 4.7,它就会自己开始这么做了,这很酷。
Speaker 320:08 - 20:23
It does something like, I'll tell it, go pull this data query. And it's like, hey, I noticed that the data is changing over time. I'll start a Loop, and I'll give you a report every thirty minutes. And I'm like, great, can you send it to me over Slack? And then it uses the Slack MCP to do that.
Speaker 320:08 - 20:23
它会做这样的事:我告诉它,去拉取这个 data query(数据查询)。然后它会说,嘿,我注意到这份数据会随时间变化。我会启动一个 Loop(循环),每三十分钟给你一份报告。然后我会说,太好了,你能通过 Slack 发给我吗?接着它就会用 Slack MCP 去完成这件事。
Speaker 320:23 - 20:37
So I think actually over time, it's not on users to figure out how to hold the tools better. And if that's the case, it's actually a product design problem and like, I'm not doing a good job. It's really on the model to do this stuff better, and on us kind of prompting it so it naturally does this.
Speaker 320:23 - 20:37
所以我觉得,实际上随着时间推移,不该由用户来琢磨怎么把这些工具用得更好。如果是这样的话,那其实是个产品设计问题,说明我做得还不够好。真正应该把这些事情做得更好的是 model(模型),而我们要做的是适当地 prompting(提示)它,让它自然而然地这样做。
Speaker 920:40 - 20:43
So right now it seems like a lot of
Speaker 920:40 - 20:43
所以现在看起来,很多
Speaker 320:43 - 20:43
us
Speaker 320:43 - 20:43
我们
Speaker 920:43 - 21:22
use Claude or Codex or these tools in the cloud to do a lot of our computing, but then there are some very vocal advocates of have your AI be local, and I could imagine over time as open way models and other things catch up that this could be more of a possibility for people to get really high quality coding assistance. So I'm curious your vision of, say, the next years or something like that, do you see the trajectory of everyone still really relying on the cloud centralized compute? Or is there a pivot to, oh, we all just have our local agents that we can rely on and they don't get throttled and other benefits?
Speaker 920:43 - 21:22
都会使用 Claude、Codex 或者这些 cloud(云端)里的工具来完成大量计算,但也有一些声音非常大的倡导者主张让你的 AI 保持 local(本地化)。而且我可以想象,随着时间推移,随着 open way models 和其他东西逐渐赶上,这对人们来说会越来越成为一种可能,让大家获得高质量的编程辅助。所以我很好奇你对未来几年之类的愿景:你觉得发展轨迹会是所有人依然非常依赖 cloud 里的中心化算力吗?还是会转向“哦,我们每个人都有自己可以依赖的 local agents(本地 agent),它们不会被限流,而且还有其他好处”?
Speaker 321:26 - 21:39
I don't know. There's maybe a few ways to answer that. I think maybe the most fundamental way to answer that is it doesn't matter. Because I think now we're getting to the point where the model is just able to figure it out. So I think by a couple of years from now, the model is just going to be doing all the code.
Speaker 321:26 - 21:39
我不知道。也许有几种回答方式。我想,也许最根本的回答是:这不重要。因为我觉得现在我们正走到这样一个点上——model 已经能够自己把这些问题搞定了。所以我觉得再过几年,model 基本上就会完成所有代码工作。
Speaker 321:39 - 21:50
It's going be starting the agents. It's gonna be building the environments. And so if it decides, actually, I'll use local models to do this, that's what it'll do. I don't think these will be decisions that we are making as engineers anymore.
Speaker 321:39 - 21:50
它会启动 agents(agent)。它会搭建 environments(环境)。所以如果它判断,实际上,我会用 local models(本地模型)来做这件事,那它就会这么做。我不认为这些还会是我们作为工程师来做的决策。
Speaker 221:50 - 21:55
We have time for a couple more questions so I can toss this out. Jamie?
Speaker 221:50 - 21:55
我们还有时间再问几个问题,所以我把这个抛出来。Jamie?
Speaker 321:57 - 21:58
Nice to hear.
Speaker 321:57 - 21:58
很高兴听到这个。
Speaker 221:58 - 21:59
Thank you.
Speaker 221:58 - 21:59
谢谢。
Speaker 1021:59 - 22:22
It feels like one of the great decisions with Cloud Code was making use of the fact that a lot of developers' tools and workflows are local. But that isn't necessarily always the case for sort of general knowledge work with cloud tools. I'm curious how you're thinking about this with Cowork, of how do you give Cowork enough access to the tools that we use to be powerful the same way that Cloud Code is for developers?
Speaker 1021:59 - 22:22
感觉 Cloud Code 的一个特别好的决策,是利用了这样一个事实:很多开发者的工具和工作流都在本地。但对于更广义、使用 cloud tools(云端工具)的知识型工作来说,情况不一定总是这样。我很好奇你们在做 Cowork 时是怎么考虑这件事的:你们如何让 Cowork 获得足够多的工具访问权限,从而像 Cloud Code 对开发者那样强大?
Speaker 322:22 - 22:41
Yeah, that's a really great question. I know when I was at a big company, took five years moving all the environments to remote. It's just so much work, especially at a big scale. But for knowledge work, largely, it's there already with Salesforce and Docs and things like that. For us, it's always just the simplest answer.
Speaker 322:22 - 22:41
是的,这是个非常好的问题。我知道我自己以前在一家大公司时,光是把所有环境迁到 remote(远程)就花了五年时间。这工作量实在太大了,尤其是在大规模情况下。不过对知识型工作来说,很大程度上,这些东西本来就已经在那里了,比如 Salesforce、Docs 之类。对我们来说,答案一直都是最简单的那个。
Speaker 322:41 - 22:52
It's just MCP. So the same MCP connector that you have in Quad AI, you hook up Salesforce, you hook up Google Docs, Google Calendar, and then a co worker can use that. Quad CLA can use it. Quad Code everywhere can use it.
Speaker 322:41 - 22:52
就是 MCP。也就是说,你在 Quad AI 里用的同一个 MCP connector(连接器),接上 Salesforce,接上 Google Docs、Google Calendar,然后 co worker 就能使用它。Quad CLA 可以用它。Quad Code 在各处都可以用它。
Speaker 1022:54 - 23:01
SPEAKER For the systems that don't have MCPs, do you think that's where computer use is going to be a big opportunity?
Speaker 1022:54 - 23:01
SPEAKER 对于那些没有 MCP 的系统,你觉得 computer use(计算机使用)会不会是一个很大的机会?
Speaker 323:01 - 23:16
SPEAKER Yeah. I think computer use is kind of a catchall. So I think currently, as far as I know, I think Anthropic is pretty far ahead on computers. And so if you use it through co work, it's quite good. So it's able to use pretty much any piece of software that you have on your computer.
Speaker 323:01 - 23:16
SPEAKER 会的。我觉得 computer use 有点像一个统称。所以就我目前所知,我认为 Anthropic 在 computers 这方面是相当领先的。所以如果你通过 co work 来使用它,效果其实很好。它基本上能够使用你电脑上的几乎任何一款软件。
Speaker 323:16 - 23:32
It's very slow, but it does it quite well now, especially with 4.7. Yeah. But I think otherwise, MCP is kind of the answer. It's and, you know, all this stuff just doesn't matter that much. It could be MCP, CLIs, APIs, just some sort of programmatic access because the the model doesn't care.
Speaker 323:16 - 23:32
它非常慢,但现在做得确实相当不错,尤其是配合 4.7 的时候。对,是这样。但除此之外,我觉得 MCP 大概就是答案。而且,你知道,这些东西其实没那么重要。可以是 MCP、CLIs、APIs,只要是某种 programmatic access(程序化访问)就行,因为 model(模型)并不在乎。
Speaker 323:32 - 23:34
It's to to the model is just tokens.
Speaker 323:32 - 23:34
因为对 model 来说,那都只是 tokens(token)而已。
Speaker 223:35 - 23:43
Alright. We have time for one more question. Ryan? Sean, do want to toss the Thank you.
Speaker 223:35 - 23:43
好的。我们还有时间再回答一个问题。Ryan?Sean,你想把这个问题抛过去吗?谢谢。
Speaker 623:44 - 24:02
You've kind of alluded to this, but if like some time ago you saw the product overhang and thought to build a product that would then become more interesting once models got better, could you just talk even in vague terms about the shape of a product you built today that you think becomes much more interesting as models get better in six months to a year?
Speaker 623:44 - 24:02
你其实已经隐约提到过这一点了,但如果说在一段时间之前,你看到了 product overhang(产品悬置/产品机会积压),并因此想要构建一个产品,而这个产品会在 models(模型)变得更强之后变得更有意思,你能不能哪怕用比较模糊的方式谈谈:如果是你今天构建的某种产品形态,你会认为它会在未来六个月到一年里,随着 models 变得更强而变得更加有趣?
Speaker 324:03 - 24:21
Yeah, quad design, think, is a really good example. It's pretty good today. It's going to get a lot better. There's also a few things that we're cooking up for quad code that are going be landing over the coming weeks, so you'll see those. And then I think loop and batch and things like this around massively paralyzing agents, that's gonna get better.
Speaker 324:03 - 24:21
嗯,我觉得 quad design 是一个非常好的例子。它今天已经相当不错了,而且还会变得好得多。另外,我们也在为 quad code 筹备一些东西,接下来几周会陆续上线,所以你们会看到这些。然后我觉得,像 loop、batch 以及这类围绕大规模并行化 agents(智能体)的东西,也会变得更好。
Speaker 324:22 - 24:24
I think computer use is another good one.
Speaker 324:22 - 24:24
我觉得 computer use 也是另一个很好的例子。
Speaker 224:25 - 24:30
Alright. Boris, thank you so much for joining us. I think we'll be here for a little longer if anyone has questions.
Speaker 224:25 - 24:30
好的。Boris,非常感谢你来参加。我想如果还有人有问题,我们还会在这里再待一会儿。
Speaker 324:32 - 24:33
Thanks, guys.
Speaker 324:32 - 24:33
谢谢,各位。
原文 ↗https://www.youtube.com/playlist?list=PLOhHNjZItNnMm5tdW61JpnyxeYH5NDDx8
BuildSpeak — 关于本项目BUILT IN PUBLIC · 跟随 builders 而非 influencers