我们很高兴 Anthropic 将在两周后举行的 AI Engineer New York 大会上分享他们最新的 AI 与金融领域成果！We are excited to have Anthropic share their latest AI x Finance work at AI Engineer New York, coming up in 2 weeks!

如果你一直消息闭塞，以下是 Anthropic 自 5 月以 470 亿美元 ARR 完成史上最大规模融资以来推出的部分产品（并非完整清单）：In case you’ve been under a rock, here’s a non-exhaustive list of what Anthropic has been shipping since closing the largest fundraise of all time in May at $47B ARR:

• June: Launched Claude Tag and Sonnet 5 and Fable 5

• 七月：Opus 5、/checkup。ARR 突破 650 亿美元

  July: Opus 5, /checkup. crossed $65B ARR

• Last month: Fable/Mythos 5.1, and EFS (upcoming pod)

  • IPO target $2T, end 2026 ARR estimated $100B

  • Cowork/chat merged before <competitor> did

  • Claude Mods

  • Dario endorses the same Pacing the Frontier message cosigned by all labs

• 上周：Opus 5.5、插件门户、云端会话/Claude 项目

  Last week: Opus 5.5, Plugins portal, Cloud Sessions/Claude Projects

• 今天：Sonnet 5.5！

  Today: Sonnet 5.5!

今天这期节目将带你了解最新情况，嘉宾是 Anthropic 的讲解之王 Thariq Shihipar。上次我们和他聊近况，是在 Fable 发布日，当时聊的是《Fable 实地指南》：Today’s episode should catch you up, with Thariq Shihipar, the explainer-king of Anthropic, who we last caught up on Fable launch day with The Field Guide to Fable:

可变软件的未来The Future of Mutable Software

特别注意 Claude Mods（尤其是速查表）：Pay special attention to Claude Mods (especially the cheatsheet):

云端大脑，本地双手Cloud Brain, Local Hands

And give a try to Claude Projects:

Boris Cherny@bcherny

Projects have changed not only how I interact with Claude but how I code. I stopped managing sessions. I just send thoughts as they come, Claude splits them into threads, and the project remembers how I work. It's where I do a ton of my coding now. [screenshot: my actual

ClaudeDevs @ClaudeDevs

Today we're rolling out Projects in Claude Code on desktop and web. A project is one conversation with Claude. It splits the work into threads itself, runs them as parallel cloud sessions, passes context between them, and keeps going when you leave. In beta for select users.

7:33 PM · Sep 17, 2026 · 574K Views

332 Replies · 126 Reposts · 3.11K Likes

“手”的这一术语不仅仅是对 Cognition 等前沿编程智能体公司正在构建的本地/云端范式的一种比喻，而且也与安全系统的讨论尤为相关。我们将在即将播出的节目中与 Anthropic 探讨这些问题；他们正为负责任地部署 AI、与前沿发展步调一致做准备。The “hands” terminology is not just an analogy for the local/cloud paradigm that is being built up at frontier coding agent companies like Cognition, but is ALSO particularly relevant to the safety systems discussions that we’ll be discussing with Anthropic in an upcoming episode as they prepare to pace to frontier with responsible AI deployment.

从 Claude Code 的迅速崛起，到智能体未来能够重写自身的运行框架、跨团队协作，并在云端和本地环境中运行，软件构建方式正在以前所未有的速度发生变化。在本期节目中，Anthropic 的 Thariq Shihipar 与 swyx、Vibhu 一起探讨：如今的资深用户实际上如何使用 Claude Code、为什么提示词编写仍是一项高门槛技能，以及 Anthropic 认为智能体运行框架接下来将走向何方。From the rapid rise of Claude Code to a future where agents can rewrite their own harnesses, collaborate across teams, and operate across cloud and local environments, the way we build software is changing extraordinarily fast. In this episode, Anthropic’s Thariq Shihipar joins swyx and Vibhu to unpack how power users are actually working with Claude Code today, why prompting remains a high-skill discipline, and where Anthropic thinks the agent harness is headed next.

我们将深入探讨 Claude Code 不断演进的界面：Ask User Question 和引导式信息收集、作为持久化生成式界面的 artifacts、用于多人协作智能体工作流的 Claude Tag、Projects、模型的思考力度、实现说明，以及用于自定义 harness 本身的全新 Claude Mods 系统。Thariq 将解释为什么 Claude.md 最终可能会消失，为什么对许多任务而言，最聪明的模型也可能成为最便宜的模型，以及可变软件为什么可能成为构建和定制应用程序的一种全新范式。We go deep on Claude Code’s evolving interface: Ask User Question and elicitation, artifacts as persistent generative interfaces, Claude Tag for multiplayer agent workflows, Projects, model effort, implementation notes, and the new Claude Mods system for customizing the harness itself. Thariq explains why Claude.md may eventually disappear, why the smartest model could also become the cheapest model for many tasks, and why mutable software could become a new paradigm for how applications are built and customized.

随后，谈话转向智能体安全，以及 Anthropic 提出的“跟上前沿”论点。Thariq 回顾了近期的一些事件：智能体发现了意想不到的通信方式、利用基础设施漏洞、逆向分析基准测试评分器，以及将多个漏洞串联起来。我们还讨论了沙箱、提示注入、自主智能体、可解释性、宪法分类器、探针、回退机制、Auto Mode，以及为什么保障能力日益强大的智能体安全，可能会成为未来几年最重要的工程难题之一。The conversation then turns to agent security and Anthropic’s “Pacing the Frontier” argument. Thariq walks through recent incidents where agents discovered unexpected ways to communicate, exploit infrastructure, reverse-engineer benchmark scorers, and chain vulnerabilities together. We discuss sandboxing, prompt injection, autonomous agents, interpretability, constitutional classifiers, probes, fallbacks, Auto Mode, and why securing increasingly capable agents may become one of the defining engineering problems of the next few years.

我们讨论：We discuss:

• 为什么智能体编程在不到一年内从备受争议变成了默认选择

  Why agentic coding went from controversial to the default in less than a year

• 为什么提示词仍然是使用 Claude Code 时最具杠杆效应的技能之一

  Why prompting is still one of the highest-leverage skills for working with Claude Code

• 专家用户如何构建对 Claude 的心智模型，以及它能可靠地一次性完成哪些任务

  How expert users build a mental model of Claude and what it can reliably one-shot

• 随着智能体能力不断增强，为什么发现那些“未知的未知”变得愈发重要

  Why discovering your “unknown unknowns” matters more as agents become more capable

• 作为人类与智能体之间持久且具有生成性的接口的工件

  Artifacts as persistent, generative interfaces between humans and agents

• Claude 如何拆分为基于云端的“脑”、本地或远程的“手”以及动态界面

  How Claude could split into a cloud-based “brain,” local or remote “hands,” and dynamic interfaces

• Claude Tag、Projects 与多人智能体，以及协作式智能体工作流可能的演进方向

  Claude Tag, Projects, and multiplayer agents and how collaborative agent workflows could evolve

• 为什么在初始提示上多花些时间，能大幅减少智能体的无效工作量

  Why spending more time on the initial prompt can dramatically reduce wasted agent work

• 不同工程任务中何时使用低、中、高或最大力度的方案

  When to use low, medium, high, or max effort for different engineering tasks

• 为什么前沿模型最终可能在智能水平和 token 效率两方面都超越小型模型

  Why frontier models may eventually outperform smaller models on both intelligence and token efficiency

• 为什么实现说明可以揭示模型曾考虑过但最终决定不采用的方案

  Why implementation notes can expose decisions the model considered but chose not to make

• 为什么 Claude.md 最终可能会消失——以及为什么有时一开始不使用它反而更好

  Why Claude.md may eventually disappear — and why starting without one can sometimes be better

• Claude Mods：自定义 Claude Code 的执行循环、UI、子代理、路由和行为

  Claude Mods: customizing the execution loop, UI, subagents, routing, and behavior of Claude Code

• 模型路由器、分叉代理和监督代理，可自动改进代理工作流

  Model routers, forked agents, and supervisor agents that automatically improve agent workflows

• 为什么 Claude Mods 可能是“可变软件”的早期预览

  Why Claude Mods may be an early preview of “mutable software”

• Harness 工程的苦涩教训，以及智能体架构为何会迅速过时

  The bitter lesson of harness engineering and why agent architectures go out of date so quickly

• Claude Tag 如何成为多人协作工作的组织级框架

  How Claude Tag is becoming an organizational harness for multiplayer work

• 为什么让智能体访问公司数据会带来巨大的全新安全攻击面

  Why giving agents access to company data creates an enormous new security surface

• Exploit-Bench 事件：智能体发现了沟通与协作的方法

  The Exploit-Bench incident where agents discovered ways to communicate and collaborate

• 为什么智能体攻击 Hugging Face，窃取评分器代码而非基准测试答案

  Why agents hacked Hugging Face for scorer code rather than benchmark answers

• 智能体如何以意想不到的方式串联利用沙箱和基础设施漏洞

  How agents chained sandbox and infrastructure vulnerabilities in unexpected ways

• 为什么能力日益增强的智能体让传统安全假设更难以维持

  Why increasingly capable agents make traditional security assumptions harder to maintain

• Anthropic“把握前沿发展节奏”提案背后的论证

  The argument behind Anthropic’s “Pacing the Frontier” proposal

• 为什么软件工程师越来越需要身兼两职：做工程工作，还要跟上 AI 的发展

  Why software engineers are increasingly doing two jobs: engineering and keeping up with AI

• 宪法式分类器、探针和回退机制，以及生产环境中的可解释性实践

  Constitutional classifiers, probes, and fallbacks and what interpretability looks like in production

• 自动模式如何检查智能体的操作是否确实符合用户权限

  How Auto Mode checks whether an agent’s actions actually match the user’s permissions

• 为什么 Thariq 能看到严重的 AI 风险，却仍然有相对较低的 p(doom)值

  Why Thariq can see serious AI risks while still having a relatively low p(doom)

• X: https://x.com/trq212

• LinkedIn: https://www.linkedin.com/in/thariqshihipar

时间戳Timestamps

00:00:00 引言00:00:00 Introduction

00:04:12 向用户提问与智能体界面的未来00:04:12 Ask User Question and the Future of Agent Interfaces

00:08:29 工件、项目与多人智能体00:08:29 Artifacts, Projects, and Multiplayer Agents

00:15:37 提示工程：Claude Code 的核心技能00:15:37 Prompting as the Core Claude Code Skill

00:21:52 上下文、推理力度与更智能的模型使用00:21:52 Context, Effort, and Smarter Model Usage

00:28:10 Is Claude.md Going Away?

00:32:49 Claude Mods: Customizing the Claude Code Harness

00:36:35 模型路由与可变软件的兴起00:36:35 Model Routing and the Rise of Mutable Software

00:44:40 Harness 工程的苦涩教训00:44:40 The Bitter Lesson of Harness Engineering

00:50:49 Claude：作为组织化束缚器00:50:49 Claude Tag as an Organizational Harness

00:55:59 把握前沿发展节奏与自主智能体安全00:55:59 Pacing the Frontier and Autonomous Agent Security

00:58:22 智能体为评分器攻克 Hugging Face00:58:22 Agents Hack Hugging Face for the Scorer

01:05:34 当智能体需要更多算力时会发生什么？01:05:34 What Happens When Agents Need More Compute?

01:10:32 AI 编程的发展速度快得让工程师难以跟上01:10:32 AI Coding Is Changing Faster Than Engineers Can Keep Up

01:17:17 探测器、备用方案、可解释性与自动模式01:17:17 Probes, Fallbacks, Interpretability, and Auto Mode

01:28:32 AI风险、p(doom) 与总结思考01:28:32 AI Risk, p(doom), and Closing Thoughts

Transcript

引言：Anthropic 的生活与变化的步伐Introduction: Life at Anthropic and the Pace of Change

Swyx [00:00:00]：我们现在在录音棚里，和来自 Anthropic 的朋友 Thariq 一起聊天。我想，整体来说，Claude Code 这方面——我——边界融合的情况实在太多了。自从你加入 Anthropic 以来，你一直紧跟着各种进展。你很早就参与了 Claude Code 本身的工作，后来也……你在其他播客里讲过这个故事，还聊过自己如何观察一个智能体。最近你做了 Top AIE World Tour 的演讲《Field Guide to Fable》。当然，你们推出了 Fable，所以这有点作弊。最近你们还推出了 Claude Tag，我们也会聊聊《Pacing the Frontier》。Anthropic 现在有好多事情在发生。我想先问个大问题：Anthropic 这么忙，你在那里的感受是什么？Swyx [00:00:00]: We’re here in the studio with our friend Thariq from Anthropic, and I guess generally the Claude Code, I-- there’s, there’s so much, merging of boundaries and you’ve been so on top of everything since you joined Anthropic. You have been early to Claude Code itself, but then also, and you’ve told that story in other podcasts, and you’ve also been talking about seeing like an agent. Most recently you did the top AIE World Tour talk, Field Guide to Fable, which obviously you guys launched Fable, so that was-- that’s cheating. And mostly you most recently also launching Claude Tag, and we’re also gonna be talking about Pacing the Frontier. There’s a lot going on in Anthropic. I guess top of the question is, what’s it like being at Anthropic when there’s so much going on?

Thariq Shihipar [00:00:48]：我觉得……有时候真的会让人应接不暇。我想，比如说，我加入 Anthropic 是因为 Claude Code。Claude Code 刚推出的时候，我就觉得：“这也太好了。”而对我来说，Opus 4 简直……我根本无法想象它能有这么好。这对我来说真的是一个重要时刻。但当时我还在努力说服那些创业公司的朋友尝试智能体编程，他们却说：“哦，不，我们的工程师觉得它还不够好用。”之类的。我当时就想：“这也太离谱了。”可现在快进 12 个月，甚至还不到 12 个月，编程就已经成了所有人的默认方式，对吧？我觉得，从推销它，到现在教大家如何更充分地利用它、提高效率之类的，这种转变真的很大。是啊，我觉得作为一个人，要跟上所有变化实在太难了。很多事情发生得太快了，而且……Thariq Shihipar [00:00:48]: I think that It is, like. I think you can get whiplash sometimes. I think, like, going. When I joined Anthropic, I joined because of Claude Code. Like Claude Code had just come out and I was like, “This is so good.” And Opus 4 to me was like just, I could not imagine, like, how good it was? And that was, like, a real moment for me. But I was, like, trying to convince, like, my startup friends to use agentic coding, and they’re like, “Oh, no, like, our engineers don’t think it’s good enough,” or something. And I was like, “That’s insane.” and now you, like, fast-forward, 12 months, less, and, like, it’s just like, yeah, the default way that everyone codes, right? And I think that, like, just having to go from, like, selling it to, like, now, teaching people how to be. make the most use of it and be more efficient and things like that is just like a big, like big change. And, yeah, I think, like, it’s just hard to stay on top of everything as a human? Like, I think things happen so fast and like

你只要再多派一些智能体去做就行了。Swyx [00:01:51]: You just throw more agents at it.

塔里克·希希帕尔 [00:01:52]：对，就像智能体相关的东西，扩展起来比人类做的事情要好得多。人类这边就像，哦，现在有三件事同时发生，而且全都是紧急情况，那你该怎么应对呢？对。Thariq Shihipar [00:01:52]: Yeah, like that’s like the agentic stuff scales much better than the, like, human stuff where it’s like, oh, like, there are three things happening right now and, like, they’re all emergencies and, like, how do you, like, respond to it? Yeah.

教人们使用 Claude CodeTeaching People to Use Claude Code

Vibhu [00:02:05]：你是怎么分配时间的？你做很多技术写作和工程工作。Vibhu [00:02:05]: What do you split your time on? You do a lot of technical writing, engineering work.

Thariq Shihipar [00:02:10]：对，我觉得吧，我加入 Claude Code 团队时，想教大家怎么使用 Claude Code。我当时觉得，这件事可能只会占用我一点时间，或者说，我可能会做做看。我一开始花了一些时间研究 Agent SDK，但对于「苦涩的教训」在 harness 方面会如何体现，我并不太确定，对吧？有时我们会想：「Claude Code 之后会是什么？」所以一开始，我想的只是教大家怎么使用 Claude Code，让 Claude Code 更容易上手。我觉得，随着 harness 越来越完善，现在最主要的问题变成了：该如何使用这些智能体，对吧？这需要很高的技能。所以我会做这方面的事，也会做工程工作、发表演讲。不过我觉得，做工程时，我的目标是把我们从用户那里得到的反馈运用起来，同时也能分享如何使用 Claude Code 来进行工程开发。这样就形成了一个很好的循环。对。Thariq Shihipar [00:02:10]: Yeah, so I think that, like, when I joined the Claude Code team, I wanted to teach people how to use Claude Code and I think that, like, that has been something that, like, I thought, like, maybe I would spend a little bit of time on it or, like, I’d, like, do. I was spending some time on the agent SDK first, and I wasn’t exactly sure, like, how the bitter lesson would go, when it comes to, like, harnesses, right? Like, I think sometimes we were like, “Oh, like, what’s after Claude Code?”? And so initially I was like, I just wanna teach people how to use Claude Code and make it easier to use Claude Code. And I think that has just, like, as the harnesses have gotten better and better, that’s like the dominant problem now is, like, how do you use the agents, right? Like, it’s like such a high skill expression thing. So I do that and then I do engineering work. I give talks, but I think, like, when I’m doing engineering work, my goal is to take that feedback that we get from users and also, like, then be able to talk about, like, hey, how to use Claude Code to do engineering. So there’s like a good loop there. Yeah.

对。我会——听众们，我们会附上你和 Sarah 为 Dev Writers 聚会做的那场演讲。Swyx [00:03:07]: Yeah. I’ll-- For listeners, we’ll attach, the talk that you did with Sarah for the Dev Writers, meetup

塔里克·希希帕尔 [00:03:13]：哦，是啊。Thariq Shihipar [00:03:13]: Oh, yeah

Swyx [00:03:13]：我们刚才稍微谈到过这一点：先做事，然后再谈做的事。Swyx [00:03:13]: Which we talked a little bit about, well, first you do the work and then you talk about the work.

塔里克·希希帕尔 [00:03:16]：对。Thariq Shihipar [00:03:16]: Right.

类似这样。Swyx [00:03:16]: Something like that.

塔里克·希希帕尔 [00:03:17]：是的。Thariq Shihipar [00:03:17]: Yeah.

这是播种与收获，还是……Swyx [00:03:17]: It’s sow and reap or

塔里克·希希帕尔 [00:03:19]：是啊，收割和……播种和收割。Thariq Shihipar [00:03:19]: Yeah, reap and. Sow and reap.

Swyx [00:03:21]：差不多就是这样。差不多就是这样。对，所以先稍微预告一下，我们会聊聊 harness 的演变。它早已不再只是一个 CLI。我们还会聊聊 Claude Mods——今天已经开始泄露了，毕竟你们没能把它保密。Swyx [00:03:21]: Something like that. Something like that. Yeah, so, and then just to preview a little bit, we are gonna talk about the evolution of the harness. It has come a long way from just being a CLI. We’re gonna talk about, Claude Mods, which is starting to leak today, because you couldn’t keep it secret.

塔里克·希希帕尔 [00:03:36]：对，对。Thariq Shihipar [00:03:36]: Yeah. yeah.

Swyx [00:03:39]：是啊，这里面内容很多。我想你一开始提到的是添加 Ask User Question 工具，这个工具有人喜欢，也有人讨厌。Swyx [00:03:39]: Yeah, there’s, there’s a lot, there. I think you started off with, like, adding ask user question tool, which people love and hate.

塔里克·希希帕尔 [00:03:48]：是的。Thariq Shihipar [00:03:48]: Yeah.

Swyx [00:03:48]：我当时觉得它特别有创新性，现在我也有了自己的版本。你有你的「采访我」版本。Swyx [00:03:48]: Like, I thought it was, like, very innovative, and then now I have, like, my own version. You have your Interview Me version.

塔里克·希希帕尔 [00:03:55]：嗯。Thariq Shihipar [00:03:55]: Yeah.

Swyx [00:03:56]：对，没错，每个人都有自己的那套东西。然后，现在这已经不重要了，因为你现在应该编写能生成其他提示词、循环以及各种东西的提示词。Swyx [00:03:56]: And, yeah, everyone just has, like, their own stuff. And, like, it no longer matters ‘cause now you’re supposed to, write prompts that create other prompts and loops and all these things.

向用户提问与人工代理交互Ask User Question and Human-Agent Interaction

塔里克·希希帕尔 [00:04:05]：当然，好的。Thariq Shihipar [00:04:05]: Sure, yeah.

Swyx [00:04:06]：那么，如今的最新进展是什么？比如说，人们都在做什么？你现在会建议大家做什么？Swyx [00:04:06]: So what’s the state of the art, today? Like, what are people. what are you, like, telling people to do today?

塔里克·希希帕尔 [00:04:12]：是的，AskUserQuestion 是模型第一次擅长引导提问。我觉得这算是一种涌现行为，我当时就想看看模型能不能做到。我有一些人机交互方面的背景，本科和研究生阶段都学过这个。所以对我来说，这有点像人与智能体的交互：试着弄清楚智能体该怎样和你沟通，并提取需求，对吧？我觉得，Claude Code 的适用范围越来越广之后，难点之一就是每个人都有自己使用它的方式，而要改变默认行为又非常困难。举个例子，如果有人让 Claude Code 做某件事，Thariq Shihipar [00:04:12]: Yeah, ask user question was the first time that the model was good at elicitation. I think this was, like, an emergent behavior that I, like, wanted to see if the models could do. I have, like a human-computer interaction background, so I, like, did that in undergrad and grad school. And so this was like. I think it’s like human-agent interaction to me, like, trying to figure out, like, how can the agent communicate with you and extract, the requirements, right? I think that, like, one of the things about, like, that’s difficult as Claude Code has gone broader and broader is that everyone has, like, their own way of using it, and it’s very hard to, like, change the default behavior. So for example, like, if someone asks Claude Code to do something,

Thariq Shihipar [00:04:59]：有时候，他们只是想让智能体把活儿干了，因为他们可能很擅长写提示词；有时候，他们可能不太擅长写提示词，而你需要……也就是说，智能体需要进一步澄清？所以，这就是一种很好的区分方式。比如，“向你提问”这个工具，就是围绕这点来划分的：你觉得自己已经足够擅长指挥智能体了吗？还是说智能体需要进一步挖掘需求，更多地与你协作，真正了解你的偏好？Thariq Shihipar [00:04:59]: Sometimes they just want them to do the work, ‘cause they’re, like, maybe a very good prompter, and sometimes they want. like, are not good at prompting? And you need. like, the agent needs to, like, clarify? And so that’s, like, a good split. Like, and the ask you the question tool like, splits along that side where, like, are-- do you feel like you’re good enough to instruct the agent as it is, or is the agent able to, like. does the agent need to, like, pull out more requirements and, like, collaborate with you more and really understand your preferences?

Thariq Shihipar [00:05:27]：总的来说，我认为几乎每个人都更接近后者，而不是前者；也就是说，他们其实有更多不确定之处，知道的也比自己希望的少，比他们以为自己对这个问题了解的要少。不过，要把这件事变得容易，某种程度上是个界面设计问题，对吧？所以，如果你在设计一个问题，或者正在处理一个问题，那么像数据结构是什么、调用栈是什么之类的事情就非常重要。设计中的细节很重要。理想情况下，你希望在开始实施之前，先找出其中一些棘手的问题。没错，所以它们才叫“未知项”，对吧？因此我认为，在智能体编程中，弄清楚未知项将永远是一项技能。因为即使模型超级聪明——它也得知道你想要什么，对吧？而且你有自己的偏好。你得把这些东西挖掘出来。所以我想这就是我正在推动的方向。接下来要问的是：智能体该如何与你互动？我认为 HTML 一直是实现这件事的主要方式。我们最近还加入了 artifacts，对吧？我觉得我们在说明如何充分使用它们这件事上做得不太好，或者说，是我做得不太好。它们有很多很强大的功能。它们还关联着一个数据库，可以吗？所以每个 artifact 都能存储和写入持久化数据。它们还能把信息反馈给 Claude。所以有件事大家还没在做，而我正努力鼓励大家去尝试，那就是 dashboard artifact 这个概念。比如说，让 Claude 长期参与一个项目，可能是一个看板之类的东西。它可以把看板数据存到自己的数据库里。多个 Claude 可以通过 artifact MCP 访问这些数据，而这个 artifact 也可以和这些 Claude 交流。所以，我们正在构建这些基础组件，让你能够通过 artifacts 拥有这种生成式界面，从智能体那里展现出更多丰富的细节。我觉得，目前智能体相关的几乎所有事情，都绕不开这样一个问题：你以为自己知道想要什么，但其实并不清楚；而智能体需要很多细节，所以让它们参与整个过程、与你协作非常重要。artifacts 就是我们试图推动其发展的方式。不过还有很多工作要做，因为这比多选题复杂得多，对吧？针对不同的问题，其中涉及的细节也多得多，比如图表、代码片段、数据结构等等。但 artifacts 是一种更有 AGI 味的“向用户提问”方式。就是这样。Thariq Shihipar [00:05:27]: I, on the whole, believe that pretty much everyone is more on the latter than the former, that they, like, have more ambiguity and they know less than they want, than they, like, think they know about the problem. but, like, it’s like a interface design problem to make that easy? And so, like, if you’re designing a problem, like, or if you’re going through a problem, like, things like what’s the schema or, like, what’s the call stack and things like that are really important. like, the details in the design are important. Ideally, you want to figure out some of these, like, hard problems ahead of time before starting implementation. And yeah, that’s why they call, like, unknowns, right? And so I think that this will forever be, like, a skill in agentic coding is, like, figuring out your unknowns. So, like, because even if the model is, like, super intelligent- It, like, needs to know what you want? And, like, you have preferences. like, you need to like, pull the, pull that out. and so that’s, like, I think how I’m, what I’m pushing. the question then is, like, how does the agent interact with you? And I think that has been HTML, has been, like, the big way of doing that. And we’ve recently added artifacts, right? And artifacts, I think we’ve done a bad job of, like, or, like, I’ve done a bad job of, like, explaining how to use them fully. We have a lot of property capabilities. They have a database associated with them? And so every artifact can store and write persistent data. They can, like, feed back into Claude? And so, like, one thing that, like, people are not doing yet that I’m trying to, like, encourage is, like, this idea of a dashboard artifact. So you have, like, Claude working on a project long-term. Maybe it’s like a kanban or something. it can store that kanban data in its database. Multiple Claudes can access that data via, like, the artifact MCP, and, like, that artifact can, like, talk to those Claudes as well. And so, like, the. We’re building the primitives for you to be able to have this, like, generative interface via artifacts that will, like, let you surface more of that rich detail from the agents. And I think that, like, almost everything with agents right now is, like, this problem of, like, you think what you want, but you don’t really know what you want, and, like, the agents need a lot of detail, and collaborating with them in the loop is really important. And so artifacts are, like, the, like, way that we’re trying to evolve there. But there’s a lot of work to do because it’s so much more complicated than, like, a multiple-choice question? there’s a lot more, like, detail in terms of, like, diagrams and code snippets and schemas or, like, whatever it is for that problem. But, like, artifacts is, like, the mo-more AGI-pilled way of, like, doing ask user question. So yeah.

工件作为 Harness 的接口Artifacts as the Interface to the Harness

Swyx [00:08:15]：我觉得关于这些东西，有一点我不太清楚：artifact 相关的内容里，哪些反馈应该通过 artifact 提交，哪些反馈应该通过 Claude 聊天提交？因为更 AGI 上头的做法，就是把所有东西都喂给 Claude。Swyx [00:08:15]: I think one thing that’s unclear to me about these, the artifact stuff is, like, what feedback should go in through the artifact and what feedback should go through a Claude, a chat? Because the more AGI-pilled one is to just feed everything to the Claude.

Thariq Shihipar [00:08:29]：我觉得更 AGI 派的做法是通过 artifact 来完成。我的意思是，我觉得从理论上说，artifact 会成为你与 harness 交互的界面。你可以实时在这份记录计划和工作的文档上发表评论；也许还能看到多个 agent，以及不同 agent 正在做什么，而这个 artifact 是专门为你当前正在做的工作构建的，对吧？所以每个 artifact 都会有些许不同。我觉得从基础设施的角度来看，我们还在逐步实现这一点。不过，是的，我觉得 harness 的即时界面大概就是未来的发展方向。Thariq Shihipar [00:08:29]: I think the more AGI-pilled one is to go through the artifact. Like, and I think that, like, we imagine in the limit, I think that artifacts will be your interface into the harness? You can, like, comment on this, like, live, like, document of your plan, of the work. you can see maybe, like, multiple agents and different agents are doing this, and that artifact is built for the current work that you’re doing, right? And so, like, each one has, like, slightly different. I think we’re still, like, getting there from, like, an infrastructure perspective. But yeah, I think, like, on-the-fly interface for your harness is probably where things are headed.

Vibhu [00:09:03]：有没有一种版本，能把 CLI 或聊天界面抽象出来，然后你……因为现在很多时候，情况是这样的：你在和 Claude Code 交互，它会返回用于模型图的 HTML。内容相当丰富，还有各种图表。Artifacts 是把这些东西连接起来的方式。为什么不干脆都用这种方式来做呢？Vibhu [00:09:03]: Is there a version of it that’s an abstraction from CLI or chat and you. Because right now, a lot of it is, okay, you’re interfacing with Claude Code, you’re having HTML given back for a mockup. It’s pretty rich. There’s diagrams. Artifacts are ways to connect these together. Why not just do everything that way?

将大脑、双手与表层 UI 分离Separating Brain, Hands, and Surface UI

Thariq Shihipar [00:09:22]：然后就变成要把这些事情区分开来：推理发生在哪里？智能体现在哪里？工作又在哪里进行？我觉得这就是本地和云端之间的区别，或者说其中的一些差异，对吧？所以我觉得现在，如果你使用 Claude Code，它是在本地运行的；你也可以启动远程控制，比如获得一些云端功能，或者直接在云端启动 Claude Code，对吧？我们正在迈向这样一种模式：你不再是给本地的 Claude 发消息，由它在本地开启会话并执行任务；而是有一个在云端运行的 Claude，你给它发消息，它可以运行本地会话，也可以运行云端会话。Claude Tag 就是这样运作的。不过随着时间推移，我们也会加入本地执行能力。这样一来，如果那台设备处于在线状态，智能体就能访问你的电脑，并在那里工作。它可以启动许多不同的子智能体；这些子智能体还可以彼此沟通，而 artifact 就是用来展示所有这些工作的。所以你可以想象一下：这些事情被拆分开来。有一个作为 artifact 的界面展示层，托管在某处，并配有数据库等等；还有云端进行的推理和智能，你不必担心关掉电脑之类的事情，对吧？然后还有执行层，它可以在本地，也可以在远程沙盒中，或者在你需要进行工作的任何地方。这就是把目前的 Claude Code 体验拆解开来：现在所有事情都发生在同一个地方，对吧？所以……Thariq Shihipar [00:09:22]: Then it becomes, like, separating out, like, where is the inference happening? Where is the intelligence happening? Where is the work happening? like, I think this is like, difference between, like, or, like, some of the distinction between local and cloud, right? And so, I think right now, if you use Claude Code, it’s, like, local and, like, you can spin off remote control, for example, to get some cloud behavior, or you can spin off Claude Code in the cloud, right? We’re moving towards a place where instead of Claudes, like, you message a local Claude, it starts a session locally and it executes, to more like you have a Claude that you message that’s in the cloud that’s running. it can run, like, local, or, like, cloud sessions. This is how Claude Tag works. But, like, over time, we’ll add, like, local hands as well. And so, like, local hands will be the ability for that agent to access your computer if it’s online, and be able to, like, work there. And so it can spin off many different subagents. It can, like, commu- those subagents can communicate with each other, and that’s where the artifact comes in to display all of that work. So you can imagine, like, the. You’re separating out these things. So there’s, like, the surface UI display that’s an artifact and hosted somewhere and has a database and everything. There is the inference intelligence, right, that’s happening on the cloud, and you don’t have to worry about shutting off your computer or whatever, right? and then there’s the, like, hands. Like, and it can be local, it can be in, like, a remote sandbox or wherever you need your work to be done. That’s like unpackaging, like, the Claude Code experience right now where, like, right now it all happens in one place, right? So.

多人智能体、Claude Tag 和项目Multiplayer Agents, Claude Tag, and Projects

Vibhu [00:11:00]：你怎么看它的多人协作方面？比如说，团队想以这种方式工作。目前这还是非常个人化的，但你觉得未来的多人协作会是什么样？比如说，我想现在有 Claude Tag 这么一个版本，不过……Vibhu [00:11:00]: How do you see, like, the multiplayer side of that? So say teams want to work in this way. Right now it’s very individual, but how do you see the future of multiplayer? Like, right now, I guess there’s Claude Tag, which is a version, but.

Thariq Shihipar [00:11:12]：我们正在推出 Projects。所以 Projects 就是一种抽象，有点像 Claude Tag，不过是在我们的 Claude 产品里，对吧？你可以给它发消息，它就会做一些类似 Claude Tag 的事情，比如派生出子代理。所以我们在设计时考虑了多人协作。Claude Tag 在多人协作方面稍微更原生一些，因为它就在你的 Slack 里，而且权限之类的都已经处理好了。不过我确实认为，多人协作是整个故事里很重要的一部分，而且我们需要更好地把它们整合起来。你可以想象事情会变得多复杂：比如，哦，你有一双手，但现在别人的电脑里也有其他的手，你得给它们设权限；或者你有自己的 MCP，别人也有自己的 MCP，那要怎么弄清楚该怎么用它们，对吧？事情会变得相当复杂。而 Claude Tag 很擅长把这些问题都简化掉，对吧？比如说，你有 Google Docs，它要怎么访问 Google Docs，对吧？它可以通过共享的 Claude MCP 访问；如果它没有访问权限，也可以通过你本地的凭据访问。不过，是的，我觉得 Claude Tag 是我们的多人协作产品，对于那些本质上就需要多人协作的事情特别有用。比如说，值班响应，事故处理本质上就需要多人协作。你会想要 @Claude，让好几个人都能登录，还希望它能找到相关上下文。我觉得每当我在做某件事，又希望它有隐私性或安全性，或者希望其他人来审查时，这种方式就特别好用。比如，我会给每个项目建一个频道，然后 @legal，跟他们说：“嘿，我想发布这个，你能不能……”就像这样。Claude 对所有情况都了如指掌，直接跟它聊就行。这样法务就能准确回答到底有哪些内容会发布到代码里，而我也不用参与整个过程，对吧？所以我觉得，多人协作正变得越来越重要，大家都可以和 Claude 一起参与。我觉得 Claude Tag 就是这样一款产品，而 Projects 一开始会是单人使用，之后再逐步扩展。Thariq Shihipar [00:11:12]: We’re launching projects. And so projects is the, like, this abstraction that’s like Claude Tag, but on our Claude products, right? So you can message it and, like, it will do the Claude Tag-like stuff, like spinning off subagents. So We think with multiplayer. Like, Claude Tag is, like, a little bit more native multiplayer because it’s just, like, in your Slack and the permissions are all figured out and stuff like that. But I do think multiplayer is, like, an important part of the story and, like, that will need to get tied together more. Like, you can imagine how complicated it gets when you’re like, oh, you have hands, but now you have other hands in other people’s computers too, and, like, you need to, like, permission them or, like, you have, like, your MCP and someone else’s MCP, and how do you figure out how to use them, right? It gets, like, quite complicated. And Claude Tag does a good job of, like, sanding down all of these issues, right? So that, like, when you have, yeah, Google Docs, how does it access Google Docs, right? Like, it accesses through the shared Claude MCP, or it can access through your local credentials as well if it doesn’t have access. But yeah, I think Claude Tag is our multiplayer, product, and it’s really useful for these, like, things that are inherently multiplayer. Like, okay, like on-call, for example, incidents are inherently multiplayer. You want to tag Claude, you want multiple people to log in, you want it to be able to find context. I think whenever I’m, like, working on something and I want, like, privacy or security or, like, I want other people to review it’s really nice to, like. I’ll have a channel per project and I’ll, like, at legal, for example, be like, “Hey, like, I want to ship this. Can you, like.” Like, here’s. Like Claude knows everything, just chat with it. And that way legal gets precise answers, on like what exactly is shipping into the code, and I don’t need to be in the loop, right? So I think like multiplayer is getting like more and more like, yeah, everyone can participate with Claude. I think Claude Tag is like that product and like projects will start off single player and will like, expand.

Swyx [00:13:14]：我觉得这里有个问题，可能是关于身份认同和隔离单元的双重问题。Swyx [00:13:14]: I think there’s a question about like maybe dual questions about identity and the unit of isolation.

身份、权限与隔离Identity, Permissions, and Isolation

塔里克·希希帕尔 [00:13:20]：是的。Thariq Shihipar [00:13:20]: Yeah.

Swyx [00:13:20]：Claude Tag，你特意选择赋予它独立的身份。Swyx [00:13:20]: Claude Tag, you specifically chose to make it its own identity

塔里克·希希帕尔 [00:13:26]：是的。Thariq Shihipar [00:13:26]: Yes.

Swyx [00:13:26]：这算是一个有争议的选择。还有其他做法。Swyx [00:13:26]: Which is like, a controversial choice. There’s, there’s other ways to do it.

塔里克·希希帕尔 [00:13:30]：嗯。Thariq Shihipar [00:13:30]: Yeah.

Swyx [00:13:30]：Claude Projects 听起来大概是这样：如果它和 ChatGPT Projects 类似，那么它的隔离方式就是通过那些 artifacts，也就是那个云端实例，所有人都在上面协作。它会……听起来应该是这样：如果你在和法务团队一起处理某件事，那那个频道就应该是一个项目，对吧？就像现在还不是这样。Swyx [00:13:30]: Claude Projects probably it sounds like, if it’s anything like ChatGPT Projects, it is, the isolation is that artifacts, that cloud instance, everyone’s collaborating on this. It’ll. It sounds like, it should be like if you’re, if you’re collaborating with legal on a thing, like that channel should be a project, right? Like it’s not yet

塔里克·希希帕尔 [00:13:50]：是的。Thariq Shihipar [00:13:50]: Yes.

但它……那是自然而然的下一步。Swyx [00:13:50]: But it. that’s the natural next step.

Thariq Shihipar [00:13:53]：对，我觉得在 Claude Tag 里，实际上是这样。就像 Claude Tag，你得自己安排。所以在 Claude Tag 里，没错，每个频道都可以随你命名，而我把它命名为……Thariq Shihipar [00:13:53]: Yeah, like I think in Claude Tag, it’s effectively. Like Claude Tag, you have to do your own arrangement. And so Claude Tag, yeah, each channel is like you can name it as you want, and I name

Swyx [00:14:04]：对。Swyx [00:14:04]: Yeah.

Thariq Shihipar [00:14:04]：就像每个功能一样】【。Thariq Shihipar [00:14:04]: Like each feature

Swyx [00:14:06]：是的。Swyx [00:14:06]: Yeah.

作为一个频道。Thariq Shihipar [00:14:07]: As a channel.

Swyx [00:14:07]：而且，不过我觉得有些移情……就是，什么时候算是移情并不明确，因为假设确实有的话，比如说你有一位同事】【。Swyx [00:14:07]: And, but I think like there is some trans- like it’s unclear when there is transference, because let’s say it is. if you have a coworker

塔里克·希希帕尔 [00:14:14]：嗯。Thariq Shihipar [00:14:14]: Yeah.

是谁在给所有这些东西打标签？是的，存在转移。Swyx [00:14:14]: Who is tagging on all these things, yes, there is transfer

塔里克·希希帕尔 [00:14:16]：是的。Thariq Shihipar [00:14:16]: Yeah.

Swyx [00:14:16]：因为是同一个人。但用 Claude 的话，就不清楚是不是一定如此，比如说，嗯，不，你对其他东西一无所知。你不了解其他那些东西。你应该只用这些东西。Swyx [00:14:16]: Because it’s the same person. but with Claude, it’s unclear if it’s like necessarily like, well, no, you don’t know any of. you don’t know about the other stuff. You should only use this stuff.

Thariq Shihipar [00:14:25]：就像冰山一角那个梗，对吧，你可以……这就是我们花这么多时间在做的事。Thariq Shihipar [00:14:25]: It’s like the tip of the iceberg meme, right, where you can like. This is what we spend so much time on

Swyx [00:14:31]：是的。Swyx [00:14:31]: Yeah.

Thariq Shihipar [00:14:31]：就像是有无穷无尽的方面需要考虑，比如你想让 Claude 做什么。当然，也不是无穷无尽，但确实有很多方面需要弄清楚，比如权限和可见性，以及怎样才能让 Claude 尽可能好地运行，同时尽可能安全？显然，这对我们非常重要，因为代码库的安全对我们来说至关重要。所以我们在这方面投入了很多时间。是啊，能想到的边缘情况太多了，比如说，哦，这个频道里的 Claude 有不同的权限，但它可以给另一个频道发消息，那它难道不能通过这种方式泄露数据吗？又或者，如果它用了你的 MCP，然后给别人发消息呢？类似这样的问题实在太多了，而我们确实花了很多功夫来把这些问题逐一打磨好。Thariq Shihipar [00:14:31]: Is like there is like infinite surface area of like, okay, you want Claudes to. Not infinite, but like there’s like surface area, a lot of like, surface area to figure out of like permissions and visibility and like how can you let Claude operate as well as you can, as safely as you can? And obviously, this is very important to us because like security for our code base is very important. And so we’ve put a lot of time into this. Yeah, there’s so many like edge cases you can figure out where it’s like, oh, like, yeah, this Claude in this channel has different permissions, but it can message another channel, and can’t it exfiltrate data that way? Or like can you like. What if it uses your MCP and then messages someone else? Like there’s like so much, and we’ve like really put a lot of work into sanding it down.

Swyx [00:15:14]：是啊。工作很多。好。Fable？Swyx [00:15:14]: Yeah. Lots of work. okay. Fable?

Fable 与提示词的元技能Fable and the Meta-Skill of Prompting

Vibhu [00:15:18]：Fable，你写了两篇很棒的文章。你写过很多很棒的文章。Vibhu [00:15:18]: Fable, you wrote two good articles. you’ve written many good articles

塔里克·希希帕尔 [00:15:22]：是的。Thariq Shihipar [00:15:22]: Yeah.

Vibhu [00:15:22]：不过，关于《Field Guide to Fable》和《Building Claude Code》，根据你的观察，Anthropic 的外部顶尖用户有没有什么共同的模式？比如说，怎样才能最好地发挥 Claude Code 的作用？Vibhu [00:15:22]: But on, Field Guide to Fable, Building Claude Code. I’m curious from what you’ve seen, is there any common patterns that you see in like top users at Anthropic externally? Like what are best practices for getting the most out of Claude Code?

Thariq Shihipar [00:15:37]：我说的这种元技能，比如说提示词编写，是不是非常重要？之类的。我觉得这话说出来并不轻巧，因为我觉得很多人会说：“哦，提示词不重要。我随便说一句，Claude 就会照做。”但我觉得提示词编写其实是……有点像公开演讲，或者写作之类的，要面向特定的受众，而这个受众就是 Claude。你需要在脑中建立一个关于 Claude 的心智模型，了解它如何思考、如何工作，对吧？所以，使用 Claude Code 时最重要的技能就是拥有这样一个关于 Claude 的心智模型，知道它擅长什么、能一次性完成什么、又做不到什么。你看很多人写提示词时，提示都很短，但他们对 Claude、代码库之类的东西有着非常好的心智模型，所以做起来毫不费力？不过这项技能的上限很高。因此，花很多时间编写提示词，并建立关于智能体如何工作的心智模型和直觉，真的很重要。 接下来我觉得就是我们之前谈到的那些未知的东西：能够找出你不知道的东西，或者你还没写下来的东西，并了解各种不同的事物。我觉得随着 Claude 能做的事情越来越多，你做出一些超出自己经验范围、而且自己领域知识又很少的事情的可能性会变得很高？你越是能学会相关的词汇、从而更好地给 Claude 写提示词，这就越重要。所以我觉得最重要的是那些未知的未知，也就是你会想：“我甚至都不知道这东西存在，对吧？”没错。我觉得这就是“地图与疆域”的一个例子，对吧？你会想：“好，这是我的提示词。”而疆域就是智能体实际需要完成的工作，对吧？如果你很精确，就能提出更精确的要求，对吧？比如说在设计方面，我就不太会描述。我不是设计师，所以我会说：“给我做八个不同的样稿。”但如果我是设计师，也许我就会说：“嘿，这是几个参考网站。”“我想要这种字体，以及这种视觉风格，这里还有几个不同的组件可以用来展示。这是一个 Figma MC 画板，可以导入进来。”之类的。这样你就能用语言把要求描述得精确得多。如果你不是设计师，就得试着学会相关的术语，或者了解那些未知的未知。我觉得这适用于所有事情。你越是能和 Claude 一起弄明白事情是怎么运作的，你的提示词就会写得越好。我觉得游戏设计就是另一个很好的例子。很多人会说：“哦，我现在可以凭感觉编个游戏了。”然后他们又说：“这游戏一点都不好玩。”游戏设计的问题就在于，每一个这样的选择都包含很多……Thariq Shihipar [00:15:37]: The like meta skill I say is like prompting is like very important? And like that. Like I think this is like not trivial to say because I think a lot of people are like, “Oh, prompting doesn’t matter. It’s just like I can just say a sentence and Claude will do it.” And I think prompting is really this like, this. It’s like public speaking, like, or writing or something, and for a specific audience, and that audience is Claude. And you need to like build a mental model of Claude and how it thinks and how it works, right? And so that’s like the most important skill in working with Claude Code is like having this mental model, right, of Claude and like what it can do well, what it can one-shot, what it can’t. And so many people when you see prompting, they’re just like, they’re short prompts, but they have such a good mental model of Claude and of like the code base and things like that like it’s effortless? But it’s like high skill ceiling. So like that work of like, spending a lot of time prompting and building mental models of how, and intuition for how the agents work is really important. And then I think like the next thing is like the unknown stuff we talked about earlier, where it’s like being able to find out like your, what you don’t know or what you haven’t written down, learning about like different things. I think as Claude can do more and more things, the likelihood of you doing something out of distribution for you and like you have low domain knowledge on is very high? And the more you can like learn the vocabulary to be able to prompt Claude, it becomes really important. And so like I think the most important unknowns are the unknown unknowns, where you’re like, I just like don’t even know that this exists, right? Yeah, exactly. I think that’s like a illustration of like the map and the territory, right, where you’re like, “Okay, this is my prompt,” and the territory is like the actual like work that the agent needs to do, right? And if you are like very precise, you can give more precise things, right? So like for example, in design, I’m not very precise. I’m not a designer, so I say like, “Give me like eight different mock-ups.” But if I was a designer, maybe I’d be like, “Oh, hey, here are some reference sites.” Like, “I want this type of font and this type of like look to it, and here’s like a few different components to like visualize. Here’s a Figma MC board to bring in,” like. And so you can just be so much more precise with that language. And if you’re not a designer, you just need to like try and learn the language or learn the unknown unknowns. And this is true of like everything, I think. Like the more, like you can work with Claude to learn like how things work, the better your prompting will be. I think another good example of this is like game design, like where a lot of people are like, “Oh, like I can vibe code a game now.” And they’re like, “It’s not fun.” And like it’s just like the thing about game design is like every one of these choices has like a lot of

品味、领域知识与词汇学习Taste, Domain Knowledge, and Learning the Vocabulary

Swyx [00:18:25]：变体。Swyx [00:18:25]: Variations.

Thariq Shihipar [00:18:25]：这方面有很多讲究。就像，哦，好吧，比如你在制作一款飞行游戏时，飞机的手感，以及它对你操控的响应方式，这些都很有讲究。就像，游戏设计师会花好几天打磨这些东西。对吧？然后，嗯……Thariq Shihipar [00:18:25]: A lot of like craft to them. So it’s like, oh, okay, like when you’re making a flying game, the feel of the plane and the like, way it responds to your controls has a lot of like. Like, a game designer would spend like days on that. Do? and like

对我来说，品味就是这样，对吧？Swyx [00:18:44]: To me, that’s what taste is, right?

Swyx [00:18:45]：就像是在一千种数学上有效答案构成的可能空间中一样。Swyx [00:18:45]: Like it is like from the possible space of one thousand mathematically valid answers

塔里克·希希帕尔 [00:18:49]：嗯。Thariq Shihipar [00:18:49]: Yeah.

Swyx [00:18:49]：这个版本是人们会喜欢的。Swyx [00:18:49]: Here’s the one that is the humans will like.

塔里克·希希帕尔 [00:18:51]：是的。嗯。Thariq Shihipar [00:18:51]: Yes. Yeah.

Thariq Shihipar [00:18:52]：我觉得说到“品味”，我对这个词有点纠结，因为我觉得你说得对，但每个人的定义都不一样。而且它听起来有点像是在说能力低，或者几乎有点精英主义，好像是在说，哦，只有某些人才有品味？Thariq Shihipar [00:18:52]: I think with taste, I’m like torn on this word ‘cause I think you’re right, but everyone has different definitions, and it sounds kind, sounds like low skill or like elitist almost, where you’re like, oh, like there are certain people with taste?

就像“品味”这个词，是我给它起的名字。Swyx [00:19:06]: It’s like taste is what I call taste.

塔里克·希希帕尔 [00:19:07]：是的，没错。Thariq Shihipar [00:19:07]: Yeah, exactly.

Swyx [00:19:08]：这些人好像根本没有品味。Swyx [00:19:08]: And it’s like these guys don’t have taste.

Thariq Shihipar [00:19:09]：对，没错。哦，就好像工程师没有品味似的。比如我这个创始人，就有品味。Thariq Shihipar [00:19:09]: Yeah, exactly. Oh, like an engineer doesn’t have taste. Like I, the like founder, have taste.

Thariq Shihipar [00:19:14]：？我觉得这不是真的。我的意思是，我觉得工程师对这类特定问题很有品味？而且我觉得每个人对特定问题都有自己的品味。我觉得 Jason Liu 就会说，想要有品味，你得先吃？Thariq Shihipar [00:19:14]: ? And I think that’s not true. Like I think like the engineers have a lot of taste for these particular like problems? And I think everyone has taste for particular problems. I think like Jason Liu, like say like in order to, yeah, have taste, you have to eat?

Thariq Shihipar [00:19:32]：我真的很喜欢这一点，就是说，好吧，你得做很多事情。你得不断迭代，弄清楚自己想要什么、喜欢什么，然后打造自己的领域。Thariq Shihipar [00:19:32]: And I really like that, where it’s like, okay, you have to like do a lot of things. You have to like iterate and figure out what you want, what you like, and, like build that like domain

Swyx [00:19:41]：是的Swyx [00:19:41]: Yes

Thariq Shihipar [00:19:41]：领域词汇。然后在编写提示词时，你其实是在为某个产品综合运用这一切。Thariq Shihipar [00:19:41]: Domain vocabulary. And then when you’re prompting, you’re like synthesizing all of that for a product.

当别人比你说得更好时，是不是很让人恼火？Swyx [00:19:46]: Isn’t it annoying when someone else says it better than you?

Swyx [00:19:48]：这简直是，操，我以后得永远引用这家伙的话了。Swyx [00:19:48]: It’s just like, fuck, I have to quote this guy forever.

永远都得引用 Jason Liu 的话。Vibhu [00:19:51]: Having to quote Jason Liu forever.

Vibhu [00:19:53]：他一定会喜欢这个。Vibhu [00:19:53]: He’s gonna love this.

Thariq Shihipar [00:19:55]：所以我收到的提示不止这些。Thariq Shihipar [00:19:55]: So I get prompts, more than that.

Vibhu [00:19:57]：有时候甚至不是那样。有时候只是凭直觉，对吧？比如，直到模型把某样东西呈现出来，你才意识到自己其实想要它，然后你就会觉得：“哦，这一下子感觉就好多了。”对吧？Vibhu [00:19:57]: And sometimes it’s not even that. Sometimes it’s just intuitive, right? Like you don’t realize you even want something till a model puts it out, and you’re like, “Oh, this just feels immediately better,” right?

语音提示与信息密度Voice Prompting and Information Density

塔里克·希希帕尔 [00:20:07]：对，没错。Thariq Shihipar [00:20:07]: Yeah, exactly.

Swyx [00:20:09]：有件事我一直拿不定主意：我觉得自己有一半时间都是这么输入提示的——比如说，我用语音输入。Swyx [00:20:09]: One thing I go back and forth on is I feel like the way I prompt half the time, let’s say I use voice.

Swyx [00:20:16]：我说的是语音吗？其他人有语音功能，这完全相反。那就像我胡扯两分钟，按住功能键，然后松开，接着希望它能弄明白。而且很多时候它确实能。Swyx [00:20:16]: Did I say voice? Other people have voice. that is the opposite. That is just like me rambling for like two minutes Pressing down the function key and then let go, and then like hopefully it figures it out. And oftentimes it does.

塔里克·希希帕尔 [00:20:26]：是的。Thariq Shihipar [00:20:26]: Yeah.

Swyx [00:20:26]：但它不像那种结构化提示词那么周全，比如把沟通组织得很清晰，就像写一份 PRD 或备忘录一样。这符合大家通常的做法吗？提示词是不是有两种模式：有些提示词会花很多时间提前打磨，另一些则是随手写出来？Swyx [00:20:26]: But it’s not as thoughtful as like a structured prompt with like Well-run communication as though it’s a PRD or a memo. Is that in line with how people do this? There’s like bimodal prompting where there’s some prompts where you spend a lot of time upfront and other prompts you just dash it off?

Thariq Shihipar [00:20:43]：我觉得不一定是声音比较低。我觉得更像是提示词里包含了多少信息。比如，模型可以……你可以再加几句话。Thariq Shihipar [00:20:43]: I don’t think the voice is necessarily low. Like I think it’s like more like how much information is in the prompt. like the model can. Like you can and like add some sentences

Swyx [00:20:53]：对。Swyx [00:20:53]: Right

Thariq Shihipar [00:20:53]：然后就说：“哦，比如说我改变主意了”，在提示词说到一半的时候插这么一句，它也能完美地跟上吗？所以我觉得，文本的实际格式没那么重要，重要的是它能……里面包含多少信息，对吧？我觉得对于语音来说，很多时候，回到人与智能体的交互这个话题，对很多人而言，说话就是比打字容易得多？而且我觉得，如果这样能让你表达出更多信息，那就更好了。Thariq Shihipar [00:20:53]: And be like, “Oh, like I changed my mind,” like in the middle of the prompt, and it will be able to follow that perfectly? So I think the like actual format of the text is less important, but then like the ability to. Like how much information is in it, right? And I think for voice, a lot of times, going back to like human-agent interaction and like for a lot of people, it’s just way easier to talk than to like type? and I. If that gets more information out of you, like that’s better.

Vibhu [00:21:21]：从某种程度上来说，这感觉就像是尽可能多地给模型提供上下文。Vibhu [00:21:21]: At some level, it feels like just giving the model as much context

塔里克·希希帕尔 [00:21:24]：是的Thariq Shihipar [00:21:24]: Yes

Vibhu [00:21:24]：开始之前多花些时间提示，是个很好的做法。我也说不准。很多时候，比如我刚开始试用 Fable 的时候，我会花足足 30 分钟认真打磨一条很长的提示词。我觉得这可能是因为模型运行的时间越来越长了，对吧？它们处于循环过程中时，要引导它们还是有点困难，但我就是凭直觉在开始时多花些时间写好第一条提示词，并不断调整完善。Vibhu [00:21:24]: Over prompting before you kick off is a best practice. I don’t know. A lot of the times, like when I was first trying out Fable, I spend a solid 30 minutes like really crafting a long prompt. This, I think, is a response of models running for longer and longer, right? It’s still a little difficult to nudge them as they’re in like, in the loop, but I just like intuitively spend more time kicking off that first prompt and working with it a lot.

前期多投入，减少迭代次数Spend More Upfront, Iterate Less

Thariq Shihipar [00:21:52]：我个人的看法是，如果我是软件工程师，比如自己经营一家初创公司，我觉得大多数时候会把倍率控制在最高 20 倍？也许验证和代码审查算是两回事。但我发现很多时候，人们在做这类事情时会碰到速率限制：模型做了很多工作，然后你说：“哦，我不喜欢这个。能不能撤销再重做？”接着你就不断迭代这件事，而只要你一开始多花点时间，或者给它更好的上下文，模型原本就能做好。结果你却说：“不行，我不喜欢这个设计，试试这个。”或者“你把这个搞砸了”之类的。这样会消耗掉更多使用额度。所以我觉得，这可能也是提升效率的一个关键诀窍，对吧？另外，我认为提供上下文很重要，而不只是说明目标是什么，对吧？比如你是在做原型，还是在做面向生产环境的东西？哪些时候可以多花算力，哪些时候不能？我觉得有时候你得告诉模型哪些事情可以做，或者不可以做，因为它并不能直观地知道你愿意在这个任务上花多少资源，对吧？你也可以通过 effort 来控制这一点。我最近就在写一篇相关的博客文章。比如说，我们看到，effort 会随着任务复杂度而产生不同影响。对于安全任务，提高 effort 会带来多得多的成果。高 effort 和低 effort 的差异会显著改变评测结果。但对于软件工程任务，影响就没那么大，因为 effort 大多花在验证、边缘案例测试之类的事情上。所以，如果能给模型这样的指导：“嘿，我觉得这个问题值得你花很多时间去验证和测试边缘案例。”呢？Thariq Shihipar [00:21:52]: My personal opinion is that if I was a software engineer, if I was like, just running my own startup, for example, I think I would mostly fit, stick to a max 20x? like maybe verification and so code review are like separate things. But I think like what I see a lot of times is people hit rate limits when they’re doing this like, oh, like it did a lot of work and you’re like, “Oh, I don’t like this.” Like, “Can you like undo this and redo it?” And then you’re like iterating on this like thing that the model could have done if you had like spent more upfront time or given it better context? And instead it’s like you’re like, “Nope, don’t like that design. Try this.” Or like, “You messed this up,” or something like that. And then that just eats up so much more of like, your usage. And so that’s like, I think maybe like a key like tip both for like efficiency as well, right? And yeah, I think like context, and not just like context on like what the goal is good, right? Like are you building a prototype or is it like a production thing? Like where can you spend compute or when, where can you not spend compute? Like I think you have to give the model permission or like not permission to do things sometimes where, like it doesn’t know intuitively how much you want to spend on this task, right? And you can use effort for this. So I did-- I’m working on a blog post about that where it’s like, if you want. For like we see that effort scales with the complexity of the task. So for security, effort gets like way more results. Like high effort versus like low effort gets, like changes the evals a lot. But for software engineering, it doesn’t change it a huge amount because effort is mostly spent on the verification and the like edge case testing and things like that. And so like being able to like give the model that guidance of like, “Hey, this problem is something that I think I want you to spend a lot of time verifying and edge case testing,”?

投入程度、模型选择与验证Effort, Model Choice, and Verification

混合使用模型怎么样？比如 Opus 和 Fable，还可以设置 effort。Vibhu [00:23:43]: How about model in the mix? So, there’s Opus and Fable with effort.

塔里克·希希帕尔 [00:23:47]：是的。Thariq Shihipar [00:23:47]: Yeah.

里面还有 Haiku。Vibhu [00:23:48]: There’s also Haiku in there.

Thariq Shihipar [00:23:49]：是啊。现在还不完全是这样，但我觉得已经很接近了：前沿模型将会在几乎所有方面都实现帕累托占优。可能吧。有时候我觉得 Opus 也可能实现帕累托占优。我的意思是，我觉得这要看情况会怎么发展，比如 Opus 的新版本。但我觉得，随着时间推移，聪明的模型会因为验证机制，能够用比其他模型更少的 token 完成简单任务。通过验证，理论上说，如果你的模型足够完美，它就不需要验证，对吧？它只要把事情做一遍，然后就说：“好了，我做完了。”而且随着 Fable 越来越进步，我也越来越常跟它说：“老兄，你没必要启动 Chromium，把这些东西全都截图下来。”我看得到啊。你已经做好了，对吧？所以很多时候，如果你投入更多精力，就会花更多 token 去验证。但如果你处理的是比较简单的问题——而很多软件工程任务都属于这种情况——比如在 Fable 里，稳定性较低或中等的任务，它就可以少花些 token 来验证。随着模型变得越来越聪明，它们就能直接说：“好了，完成了。”我可以出于保险跑一下 lint，但我就是知道它能通过 lint。你甚至都不需要跑。而这会比小模型节省多得多的 token。对。Thariq Shihipar [00:23:49]: Yeah. It’s not quite true yet, but it’s very close where I think the frontier models will be Pareto dominant over like almost everything. like maybe. And sometimes I think Opus might be Pareto dominant. Do? Like I think depending on like how things, like shake out if it’s like a newer version of Opus. But I think that like increasingly it’s just going to be like the smart model is going to be able to like do the simple task for less tokens than the like the other models because of verification. With verification, in the limit, your model doesn’t need to verify, right? If it’s a perfect model, it just does the work once and it’s like, okay, like you, I did it? And increasingly with Fable, I’m like, I’m like, “Dude, you don’t need to spin up Chromium and screenshot all of these things.” Like I see it. Like you did it, right? And so a lot of the. At higher effort, you spend more of those tokens verifying. But if you’re working on simpler problems, and a lot of software engineering is like well, like in Fable, like low and medium stability, it can spend less tokens verifying. And as the models get smarter and smarter, they will just be able to like, “All right, done.”? Like, I can run the lint for sanity’s sake, but, like, I, like, know it lints? Like, you don’t even need to do that. And that will be so much more token efficient than, like, the smaller models. Yeah.

Swyx [00:25:15]：有没有什么好的做法，让我们这边能判断自己是不是投入了太多精力？比如，我真是……Swyx [00:25:15]: Is there a good, practice on our side that we can use to see if we’re using too much effort? Like, I freaking

塔里克·希希帕尔 [00:25:23]：是的。Thariq Shihipar [00:25:23]: Yeah

Swyx [00:25:23]：讨厌把时间浪费在那些事情上。Swyx [00:25:23]: Hate wasting time on that stuff.

Thariq Shihipar [00:25:24]：是啊，我明白你的意思。我觉得吧，比如在这篇博客文章里，我大致的分布是，代码审查和安全应该占比较高，甚至最高，而软件工程……Thariq Shihipar [00:25:24]: Yeah. I know what you mean. I think, like, so in this blog post, my rough distribution is, like, code review and security should be, like, high or max and, like, software engineering

Swyx [00:25:37]：你说过要针对每个领域推荐混合设置。Swyx [00:25:37]: You said recommend mix settings per domain.

Thariq Shihipar [00:25:37]：对。我觉得吧，如果你在做 UI 之类的东西，我觉得低级和中级就是指你在构建 API，并且想确保覆盖足够多的边界情况？所以我觉得，正如我所说，建立起对这些发行版之间事物如何运作的心智模型，确实也是工作的一部分。Thariq Shihipar [00:25:37]: Yeah. I think, like, if you’re doing, like, UI or something like that, like low and medium, I think is you’re building, like, an API and you want to make sure, like, you cover enough edge cases? And so I think building, like I said, that mental model of, like, how things work across these distributions is, like, yeah, part of the job.

实施说明与决策日志Implementation Notes and Decision Logs

Vibhu [00:25:56]：这更多是凭直觉判断，还是通过评估得出的？因为我猜随着进展，这种情况会发生变化。Vibhu [00:25:56]: This is more intuition-driven or eval? Because I’m guessing this would change as you go.

他有评测。Swyx [00:26:00]: He has evals.

Thariq Shihipar [00:26:01]：对。所以我在那篇博客文章里，回顾了所有 Terminal Bench 评估。里面大概有 70 个问题，我展示了比如说，在安全问题上它表现得更好。然后我也看了一些对话记录，看看它会回答什么、会忘掉什么之类的。很多时候，这也是我给出的另一个提示技巧：让它写下决策笔记或实现笔记。因为在它遇到的每一道评估题里，它都会想到正确的解决方案，却又决定不去做。它会想：“哦，答案在这里。要不我这么做？”然后又想：“哦，可能还是别这么做吧？”接着继续往下走。而这就是更高最大级别下大多数失败的原因。模型完全不知道怎么做某件事的情况其实很少见。如果你让它写下这些实现笔记，就可以检查，然后告诉它：“哦，我想让你做这件你没做的事。”总的来说，模型在把这些内容呈现出来方面越来越好了。比如我在 Fable 5.1 的对话记录里看到，它输出这类内容时，也会把自己的决策过程说出来。不过，在 harness 中把这件事做得更明确会更好。现在我们也在允许你修改 harness，这样你就可以添加一些……Thariq Shihipar [00:26:01]: Yeah. So what I did in the blog post is I go over all of the terminal bench evals. So there are, like, 70 problems and I’m show that, like, okay, like, in the security problems it does more. and then I also, like, look at some of the transcripts just in terms of, like, how-- what does it answer, what does it forget or something. And a lot of times, this is another prompting tip I have, is, like, asking it to make decision notes or implementation notes because, in every eval problem that it faces, it thinks about the correct solution, and decides not to do it. it’s like, oh, like, here is the answer. What if I did this? And then it’s like, oh, probably not? and then keeps going. And this is, like, the majority of the failures, at, like, a higher max level. It’s very rare that the model just doesn’t know how to do something. If you just have these implementation notes, then you can review and you can be like, “Oh, I want you to do this thing that you didn’t do.” The models are getting better at surfacing that overall. Like, I see in the transcripts of Fable 5.1, like, when it does this output, it will call out its decision-making as well. but making this more explicit in the harness is better. And now we’re, allowing ways of you modifying the harness so you can, like, add some

Vibhu [00:27:23]：哦。Vibhu [00:27:23]: Ooh.

用那里计算。对。Thariq Shihipar [00:27:24]: Calculate with there. Yeah.

Swyx [00:27:25]：是的。所以我确实想指出你提到的两件事，我认为它们存在于提示词之外。一件事是，我们姑且把它称作“重要到不该放在提示词里的提示词”。它写在 Claude.md 或 Agents.md 里。Swyx [00:27:25]: Yeah. So I do wanna call out two things that you mentioned that I think exist outside of prompting. One is like, let’s, let’s call it the prompt that is so important that it shouldn’t be in a prompt. It is in Claude.md or Agents.md

塔里克·希希帕尔 [00:27:38]：是的Thariq Shihipar [00:27:38]: Yeah

Swyx [00:27:38]：这有点像目标，对吧？比如你的处境、你的目标、你想要的东西之类的。其次就是决策日志、实验日志，或者任何你可能希望能留存下来、以便在当前会话结束后继续用于实现这些目标的记录。这些都属于外部信息，没有统一标准。没有——它不像 skills，也不像 MCP，没有标准。它就是一个 Markdown 文件。首先，我说得对吗？Claude.md 要消失了吗？你公开表示过不喜欢 Agents.md，但你还是要用它吗？Swyx [00:27:38]: Which is like goals, right? Like your situation, your goals, the things that you want, the thing. and then second of all is the decision log or the experiment log or whatever log of traces that you might want to survive the current session to do those things. Those are, like, externalities that there’s no standard. There’s no-- It’s not like skills. It’s not like MCP. There’s no standard. It’s, it’s just like it’s a markdown file. first of all, is that right? Is Claude.md going away? You have a documented dislike of, Agents.md, but you’re gonna do it?

Claude.md、Agents.md 和模型特定说明Claude.md, Agents.md, and Model-Specific Instructions

Thariq Shihipar [00:28:10]：对。好。关于 Agents.md，嗯，我们会做的。我觉得只是不同模型之间的差异很大？不过我意识到，维护好几份不同版本实在太麻烦了？而且，随着模型越来越好，它们完成简单任务的能力下限也在提高。所以我确实认为，长远来看，Claude.md 会消失，甚至可能用不了那么久。我觉得，现在开始一个新项目时，也许不写 Claude.md 会更好。Thariq Shihipar [00:28:10]: Yeah. Okay. So Agents.md, yeah, like, we’re, we’re gonna do it. I think it’s just, like, different models are very different from each other? But I realize that it’s, like, such a pain to, like, maintain different ones? And yeah, like, as the models get better and better, the floor of how they accomplish the simpler task is better. And so I do think in the limit, Claude.md goes away, and maybe not even, like, that far. Like, I think, like, I think that right now it might be better to start a new project without a Claude.md.

Swyx [00:28:44]：是的。Swyx [00:28:44]: Yes.

Thariq Shihipar [00:28:44]：我觉得，可能如果你发现某些失败模式反复出现，就把它们添加到你的 Claude.md 里。真正棘手的是，这会因模型而异。所以，如果你添加了一堆失败模式，或者甚至……Thariq Shihipar [00:28:44]: I think that, like, maybe if you see very repeated failure modes, you add them to your Claude.md. The really tough thing is that this changes per model. And so, like, if you’ve added a bunch of failure modes or, like even

所以你需要 Fable MD，也需要 Opus MD。Swyx [00:28:57]: So you need Fable MD, you need Opus MD.

Thariq Shihipar [00:28:59]: Or well, even Fable 5.1 versus Fable 5.

Swyx [00:29:03]：是的。Swyx [00:29:03]: Yeah.

Thariq Shihipar [00:29:03]：就是，挺烦人的。就是，我不是那种，Thariq Shihipar [00:29:03]: Like, it is annoying. Like, I’m not like,

Swyx [00:29:05]：是的Swyx [00:29:05]: Yeah

Thariq Shihipar [00:29:05]：就是说，我们不是故意这么做的，对吧？这只是模型的工作方式，对吧？所以，Fable 5 可能有一种 Fable 5.1 没有的失败模式。如果你保留这段上下文，也就是记录各种不同失败模式的日志，可能会对 Claude 造成过度约束？所以，我们刚刚为 skills 添加了 evals 插件。Thariq Shihipar [00:29:05]: Like, we don’t, like, do this on purpose? It’s just, like, how the models work, right? And so, like, maybe, like, Fable 5 had this, like, failure mode that Fable 5.1 doesn’t. And if you keep this context, this running log of a bunch of different failure modes, they will probably over constrain Claude? And so this is like. we just added evals plugins for skills.

Swyx [00:29:28]：是的。Swyx [00:29:28]: Yeah.

Thariq Shihipar [00:29:29]：这样一来，你就可以评估某个技能是否更好。我想我们团队的 Daisy 做过这件事。所以，没错，这就是我们正在努力解决的问题。我们知道，你还是得为此花费 token，而且它并不完美，但我们正在努力帮助解决这个问题。Thariq Shihipar [00:29:29]: And so now you can eval if a skill is better. I think Daisy on our team did this. And so, yeah, this is like we’re trying to work on this. We know it’s, like, you still have to spend tokens on it and, like, it’s not, it’s not perfect, but it’s, like, we’re trying to help out with this problem.

Swyx [00:29:44]：所以，说到提示词编写，我想分享的一条建议是：我经常告诉别人，足够高级的提示词编写，和足够高级的高管沟通，几乎无法区分。我提到过……这是 Heavybit 的一场高管沟通工作坊，是我职业生涯中见过最棒的。他们讲了一种叫 SCQA 模型的方法。直接搜一下就知道了。这是个……就是这么回事。人们几十年来一直在做提示词编写，只不过它叫高管沟通。就像一个人需要沿着组织架构，向成千上万的人传达信息时，该怎么做。也就是说，情境、冲突、问题和答案，这就是写备忘录的方法。当然，有时候你并没有答案，但至少可以列出 SC 和 Q，他们里面也有一些例子。所以我只是给大家留些线索，感兴趣的话可以去了解。Swyx [00:29:44]: And so, and as far as prompting goes, the one tip I wanna offer is, something I have told people a lot is sufficiently advanced prompting is indistinguishable from sufficiently advanced executive communication. So I’ve referred to-- This is an executive comms workshop from Heavybit that is the best I’ve ever seen in my career. And they teach this thing called the SCQA model. Just Google it. It’s a, it’s a thing. Like, people have done prompting for decades. It’s just called executive communication. It’s like when one person has to communicate to thousands of people down the org chart, this is what you do. so situation, complication, question and answer, is how you write the memo. but obviously sometimes you don’t have the answer, but you can at least list out the SC and Q, and then they have some examples in there. So just leaving breadcrumbs for people if they want to explore.

被低估的提示词模式与 ELI5Underrated Prompting Patterns and ELI5

Vibhu [00:30:31]：继续之前，我想问一下，还有哪些被低估的技巧，能让大家从 Claude Code 中获得更多价值，但目前却没用上？Vibhu [00:30:31]: Before we move on, I wanna ask you, any other underrated tips, ways people could get a lot of value from Claude Code that they’re not using?

Thariq Shihipar [00:30:41]：对，我觉得很多内容都在那个叫《未知事项》的文档里。比如，我给了很多提示词示例，像是用它来头脑风暴，或者之后用它来考考你。我们还加了一个“像给五岁小孩解释一样”技能，它的提示词非常短。它甚至没说“像给五岁小孩解释一样”，这个提示词的关键词是“大图景，少文字”。这才是重点。效果好得惊人？我觉得我在推特上发过这个，它叫 /eli5，而且你可以把它安装成插件。不过，它确实更擅长直接切中要害，不绕弯子，直接告诉你：“对，就是这里。”所以图表相当清晰。我觉得，Artifacts 有个问题，就是里面放了太多文字，大家根本不看这些内容。所以，这样就简化了很多。对，这个功能的灵感来自 Anthropic 的一些人，他们在处理非常复杂的事件时会问：“这到底是怎么回事？”所以，我觉得这个功能很棒。Thariq Shihipar [00:30:41]: Yeah, I think a lot of them are in the, this unknowns, like, doc. Like, I give a bunch of example prompts, like, using it for brainstorming, using it to quiz you after. we added this, like, explain it like I’m five skill which is a very short prompt. And it doesn’t even say explain it like I’m five. It’s like the key word of this prompt is big pictures, few words. like, that’s like the main thing. And it is shockingly good? Like, you, like, I think I tweeted about this and it’s like /eli5, and, like, you can install it as a plug-in. But yeah, it’s, like, way better at just cutting through the BS and being like, yeah, exactly right here. So the diagrams are, like, quite clear. I think one of the things that is true with artifacts is, like, they put too much text in and people are not reading the artifacts? And so, like, this simplifies it a lot more. And, yeah, this came out of, like, just people at Anthropic, like, going through very complicated incidents and being like, “What is happening?”? So, this one I think is great, yeah.

Swyx [00:31:47]：我对它的理解是，这就像是在测试你的理解。给你几个选项，然后，如果你选错了，就说明你认为正在发生的事情和实际发生的事情之间存在偏差。Swyx [00:31:47]: My version of this is the, it’s like test your understanding. Give you a few choices and then, like, if you get it wrong, you have a mismatch between what you think is happening versus what’s happening.

Thariq Shihipar [00:31:58]：是啊。我觉得这是那种大家都爱谈论、但真正去做的人却寥寥无几的事情之一。就像，我觉得……Thariq Shihipar [00:31:58]: Yeah. I think this is one of those things that everyone loves talking about, and then very few people really do. Like, I think

非常有帮助。Swyx [00:32:05]: Really helpful.

Thariq Shihipar [00:32:07]：是啊。但大多数人就是不想被人拿某件事来提问吧？不幸的是，我觉得这也是我们需要……的一件事，就像……。Thariq Shihipar [00:32:07]: Yeah. But most people just don’t want to get quizzed about something? Unfortunately, I think this is one of the, like, things that we need to, like.

Swyx [00:32:16]： “问你这个问题”，或者“在那件事之前先问你这个问题”的反义说法是什么？Swyx [00:32:16]: What’s the opposite of ask you the question or ask you the question before the thing?

塔里克·希希帕尔 [00:32:19]：是的。Thariq Shihipar [00:32:19]: Yeah.

Swyx [00:32:19]：这是那件事之后的内容。Swyx [00:32:19]: This is after the thing.

塔里克·希希帕尔 [00:32:20]：没错。嗯。Thariq Shihipar [00:32:20]: Exactly. Yeah.

Vibhu [00:32:21]：这也是一种让自己保持清醒的好办法，比如说，你到底知不知道自己在做什么，对吧？最糟糕的情况是，别人把一堆垃圾发给你，却根本没弄明白自己想要什么，也不知道输出内容是什么，然后你就会想：“哥们儿，我可不想读这玩意儿。你到底知道这是什么吗？”所以你可以给自己定个规矩：发出去之前，至少要弄清楚实现了什么。Vibhu [00:32:21]: It’s a good way to stay grounded of, like, do you even know what you’re doing, right? The worst case is when people send you slop and they haven’t understood what they’re asking for or what the output is, and it’s like, “Dude, I don’t wanna read this. Do you even know what it is?” So, you make it a rule for yourself that before you send stuff, you should at least know what’s implemented.

Claude Mods：自定义运行框架Claude Mods: Customizing the Harness

Thariq Shihipar [00:32:41]：是的，所以你可以把这个做成一个模组，也可以自己制作模组，比如确保你测试过它。所以，没错，你可以这么做。Thariq Shihipar [00:32:41]: Yes, but so you could make this a mod and you could build your own mod to, like, make sure you test it. So yeah, you can do that.

Swyx [00:32:49]：好，我们直接开始吧。Claude Mod 是什么？这张图展示的是什么？Swyx [00:32:49]: All right. Let’s get right into it. What is Claude Mod, and what is this diagram showing?

Thariq Shihipar [00:32:54]：是的。好，Claude Mods 可以让你自定义整个 Claude Code harness，而我们也会这么做。如果你有什么需求，尽管告诉我们。我们会不断添加更多功能。它适用于 CLI，也适用于桌面版。也许将来还能用于 Claude Tag，我也不确定。我们正努力让它具备很强的可扩展性。你可以看看这份参考表。我不希望大家看了觉得不知所措。大致来说，你既可以自定义 harness 的执行方式，也可以自定义它的界面。比如 Boris 的那个俄罗斯方块例子，就是在自定义界面，对吧？也就是在游戏里显示俄罗斯方块。Thariq Shihipar [00:32:54]: Yeah. Okay, so Claude Mods is you can customize the entire Claude Code harness, and we’re going to. If you have requests, we will, like, let you, like, please let us know. We’ll add more and more. This works for CLI, it works for desktop. maybe it will work for Claude Tag in the future. I don’t know. Like, we’re trying to make this very extensible. You can see this reference sheet. I don’t want people to get overwhelmed by it? At a high level, you can customize both the execution of the harness, and the UI of the harness. And so, like, you say on that Tetris example from Boris, that’s like customizing the UI, right? Like showing, like, Tetris in the game.

Thariq Shihipar [00:33:35]：不过，比如说，假设你想做这样一件事：每个项目结束后，你都会检验自己的假设，或者说检验自己对项目的理解，对吧？你要做的就是让 Claude 制作这个插件。它会在每次提示之后启动一个分类器。所以，每轮结束时，你都会启动一个子代理，或者说一个分叉出来的代理。分叉出来的代理会保留提示缓存，对吧？这就像是那种不太直观的用法：你可以分叉出去，发起一个小请求，而且成本会很低，因为整个提示缓存都已经准备好了。于是你就可以问：“这项任务完成了吗？”之类的。Thariq Shihipar [00:33:35]: But, like, let’s say that you wanted to do this thing where you had. you tested your assumptions or, like, tested your understanding after every project, right? What you would do is you would ask Claude to make this plug-in. It would spin a classifier after every prompt. And so, like, at the end of each turn, you would spin off a sub-agent or, like, a forked agent. A forked agent is, like, maintains the prompt cache, right? So it’s like a, like one of those unintuitive things where you can fork and do, like, a little request, and it’ll be very cheap because the entire prompt cache is, like, done. And so you can be like, “Has this task been completed?” like

Swyx [00:34:18]：这就是你做 BTW 以及诸如此类事情的方式。Swyx [00:34:18]: This is how you do BTW and all those.

Thariq Shihipar [00:34:20]：对。底层那个分叉出来的代理，是的。不过你可以在这个 f-fork 子代理里这样说：“这个任务完成了吗？如果完成了，就返回 true。”然后在你的 hook 里，或者你的插件模组里——不对，应该是在子代理里——你可以说：“如果是 true，就给我出个小测验。”给我一些问题和答案，然后用 JSON 格式返回。接着你解析它，再把这份问题列表显示在提示词输入框上方，对吧？所以这会稍微多消耗一些 token，因为你得在助手每次结束一轮之后都执行一次。不过这只是轻量级的分类，然后你就能得到这个小测验；之后你会发现 Claude 总会帮你做，不需要你记着去做。我们聊过很多这类小技巧，对吧，比如实现说明。现在你也可以为实现说明添加一个工具。所以我现在加的这个工具是用来注册——我想我叫它 assumption（假设），不过也许之后会改名。这是个模组。你给它一个注册假设的工具，它就会维护一个列表。它会……每次执行时都会把内容加到列表里，最后再把这些假设显示出来？我正在做的另一个模组是模型路由器。也就是内部的 Claude 模型路由，对吧？所以……我想说，我们默认不做模型路由的原因是，这个问题很难？而且……Thariq Shihipar [00:34:20]: Yeah. The underlying forked agent, yes. But so you can, in the f-fork sub-agent, you can say, like, “Has this task been completed? If so, return true.” And then in your hook, or in your, like, plug-in mod, or sorry, like, in the sub-agent probably, you would say, like, “If true, give me a quiz.” give me questions and answers, and then, like, in a JSON format, and then you’d parse it, and then you display above the prompt input, this list of questions, right? And so this is something that’s, like, slightly token-intensive because, like, you have to do it after every end of the assistant turn. But it’s, like, a lightweight classification, and then you can, like, get this quiz, and then you’ll see, like, Claude will always do it for you. You don’t need to remember to do it. There are lots of these, like, tips that we’ve talked about, right, where it’s like, oh, implementation notes. You can also add a tool for implementation notes now. And so, like, this tool that I’m adding is, like, register, like, I think assumption is what I’m calling it, but, like, maybe I’ll change it around. And this is a mod. And so, like, you give it a register assumption tool, and then it will keep a list. It’ll. Every time it does it’ll keep a, like, add to the list, and then at the end it will display those assumptions? Another mod I’m working on is a model router. And so, like, internal, like, Claude model routing, right? So it’s. This is, I want to say the reason we don’t do model routing by default is, like, it’s a hard problem? And like

分叉代理、假设跟踪与模型路由Forked Agents, Assumption Tracking, and Model Routing

Swyx [00:35:51]: You will get it wrong.

Thariq Shihipar [00:35:52]：对，你就会，不知不觉地在难题上用 Fable，或者用 Sonnet 来……Thariq Shihipar [00:35:52]: Yeah, you, like, yeah, you will, like, accidentally use, like, Fable for a hard problem or Sonnet for

Swyx [00:35:57]：是的，如果你启用了自动批准，但没有启用自动模式。Swyx [00:35:57]: Yeah, if you have auto approve, but you don’t have auto mode.

嗯，你会有自动功能。比如说，你没有自动路由之类的功能。Thariq Shihipar [00:36:01]: Well, you will have auto. Like, you don’t have, like, auto routing or something.

Vibhu [00:36:04]：模型选择器没有自动模式。Vibhu [00:36:04]: You don’t have auto mode for model picker.

塔里克·希希帕尔 [00:36:06]：对，没错。所以Thariq Shihipar [00:36:06]: Yeah, exactly. So

Vibhu [00:36:07]：我大概想问的是，你们会把这个功能开放到什么程度？用户又需要花多少心思去理解它？比如你提到提示缓存和构建路由器，感觉完全可以轻松做一个按查询进行路由的 mod，而我这是很快就把计划打消了，对吧？我想问的更像是，这类产品介绍应该是什么样的？面向谁？是面向高级用户吗？还是说每个人都应该能上手……Vibhu [00:36:07]: I’m getting the rough question of, like, how much do you open this up and how much do people have to think about this? Like, when you talk about prompt caching and building a router, it seems like you could easily build a mod that routes per query, and I’m just killing my plan very fast, right? I guess my question is more so, like, what is, like, a product talk like this look like, right? Who is it for? Is it for power users? Is it everyone should be able to go through

Swyx [00:36:33]：哦，肯定是高级用户，对吧？Swyx [00:36:33]: Oh, definitely power users, right?

Thariq Shihipar [00:36:35]：是啊，我觉得确实是高级用户，不过 Claude Code 的特性就是，很多人都成了高级用户？因为分享东西很容易，比如说，一个人可以做出一个好用的模型路由器，不会老是破坏提示缓存，然后你就可以把它们组合起来。插件的另一个很酷的地方是，它们可以相互接入、彼此组合。所以我有一个模组，可以在顶部创建一个模式选择器，任何插件都可以注册为一种模式。这样自动路由器就可以是一种模式，对吧？或者你也可以有一种模式，比如制品模式，主要通过制品和你交流。你还可以在计划模式之间切换？所以你可以创建越来越多这样的模式。不过，创建模式的能力本身也是一个模组？这里面的可能性非常丰富，但我们确实希望让它变得相当容易。我们想让你只要安装别人做好的模组就行。你可——你可以和 Claude 聊天，我们会确保它理解提示缓存之类事情的细节，这样它就能提醒你。我觉得对 Claude 来说，这种行为并不算特别复杂，但我们确实应该提供一个关于如何制作模组的好技能。总之，我们拭目以待。不过我确实认为，这是对可变软件的一种预览，也让我们看到生成式软件如何让你安全地自定义东西。只要启用了这项功能，你就能自定义任何软件的一部分。我觉得理想情况下，越来越多的应用都会做类似的事情？Thariq Shihipar [00:36:35]: Yeah, I think it is power users, but, like, the nature of Claude Code is that so many people are power users? Because it’s easy to share things, like you can. Like, one person can make a good model router thing that doesn’t break prompt cache all the time, and then you can, like, compose them. Another cool thing about the plug-ins is that they can hook into and compose with each other. And so I have, like, a mod that will, like, create a mode selector at the top, and any plug-ins can register to be a mode. And so, like, the auto router can be a mode, right? Or, like, you can have a mode that’s, like, artifact mode, where it’s like it primarily talks to you in artifacts. like, you can toggle between plan mode? And so, like, you can create more and more of these modes. But the ability to create modes is in it itself a mod? And so there’s a lot of richness here, but we do want to make it fairly easy. We want to be-- make it so that you can just, like, install someone else’s. You can ta-- you can chat with Claude and, we’ll, like, make sure that it understands the nuances of things like prompt caching and stuff, so it can, like, warn you. This is, like, not extremely complicated behavior for Claude, I think, but we should have just a good skill on how to make mods. and yeah, we’ll see how we go. But I do think that this is, like, a preview of, like, mutable software, and, like, how, like, generative software, just like you can customize safely. If enabled, you could customize any piece of software. And I think that more and more apps ideally do something like this?

高级用户、模式与可变软件Power Users, Modes, and Mutable Software

Swyx [00:38:13]：顺便说一句，你——我们有，你还有另一条很酷的推文，讲的是那个“无限印钞按钮”：让你的 SaaS 能被智能体使用。我觉得可变软件挺有意思的，也有其他人尝试过这么做。我觉得难点在于，当你什么都能做时，用户往往会感到困惑。所以通常真正行得通的，都是那种只有一条明确预设流程的产品。这个思路则是要减少预设，就像是：好吧，给高级用户更多权限。我觉得这可能是 AI 解锁的能力，因为你可以直接输入提示词，告诉它你想要什么。Swyx [00:38:13]: And by the way, you, we have, you have another cool tweet about how, there’s the infinite money button, which is like make your SaaS, consumable by agents. I think mutable software is interesting and, other people have also tried to do it. I think the hurdle comes when you can do everything, then people, users get, tend to get confused. So usually the stuff that works is just like one opinionated flow. This is in the side of less opinionation. It’s just like, well, more power to power users. And I think probably unlocked by AI, where, like, you can just prompt for whatever the thing is.

Thariq Shihipar [00:38:47]：对，或者可以有一种技能来提供这些观点？Thariq Shihipar [00:38:47]: Yeah, or there can be a skill that gives the opinions?

模组 vs. 钩子 vs. 工件Mods vs. Hooks vs. Artifacts

Swyx [00:38:50]：嗯。Swyx [00:38:50]: Yeah.

塔里克·希希帕尔 [00:38:50]：然后，嗯。Thariq Shihipar [00:38:50]: And then, yeah.

Swyx [00:38:51]：所以，稍微了解一点 TypeScript、构建系统之类的东西之后，我最先想到的——我很好奇，负责开发这个的团队，如果我没记错的话，你和他们的关系有多近？他们有没有从 Babel、Webpack 这类老派构建系统中汲取灵感？因为听起来很相似，就像它们的插件生态系统一样，彼此之间可以组合使用。Swyx [00:38:51]: So knowing a little bit about, like, TypeScript and build systems and all these things, the closest-- I’m very curious that the team who worked on this, if, I don’t know how close you were to them, if they drew any inspiration from build systems like Babel, Webpack, all these, like, old school things. Because it sounds very similar, like the plug-in ecosystem of those things where they can compose with each other.

Thariq Shihipar [00:39:11]：是的，我对技术细节了解不深，但我知道这是 Bun 团队的一位成员和 Claude Code 团队的一位成员合作完成的。Thariq Shihipar [00:39:11]: Yeah, I’m not deep in the technical details, but I do know it was a collaboration with someone on the Bun team and someone on the Claude Code team.

Swyx [00:39:17]：是啊，这里简直是构建系统的圣地。Swyx [00:39:17]: Yeah, it’s a build system mecca.

Thariq Shihipar [00:39:19]：对。没错。这真的非常令人兴奋。不过，是啊，智能体现在可以直接为你的软件实现非常复杂的扩展功能。所以，是啊，这也是一个理由。如果你在经营一家初创公司，就可以直接向 Claude 提示：“嘿，我们能不能做一个扩展系统？那会是什么样子？”。Thariq Shihipar [00:39:19]: Yeah. Exactly. It’s, it’s very exciting. But yeah, like, agents can just do this very complicated like, extensibility into your software now. And so, yeah, like, another reason to, like. If you run a startup, like, you can just prompt Claude and be like, “Hey, like, could we make an extension system? Like, what would that look like?”?

Swyx [00:39:37]：是的。Swyx [00:39:37]: Yeah.

Swyx [00:39:38]：我确实很好奇，你们以前有 hooks、插件之类的东西。那么，mods 具体能做些什么，是那些东西做不到的？Swyx [00:39:38]: And I just really wonder, like, you had hooks in the past and plug-ins, all these things. So what specifically will mods be able to do that those things could not do?

Thariq Shihipar [00:39:47]：在内部，我们最初把这个功能叫作 hooks。所以，这大概能让你稍微了解一下它的用途：hooks 会注册一个要触发的事件，然后指定要调用的脚本。它是在 TypeScript 运行时里执行的。因此，你能获得一些好处，比如它在 Scope 中包含很多信息，例如这段对话进行了多少轮、用了多少个 token，等等。还有消息内容之类的信息，所以有很多消息都可以使用。除此之外，它还提供了更多 hooks，也就是说，你可以注册的地方多了很多。然后，由于这一切都在进程内发生，你可以生成子代理，并为它们提供四个 contest 和 context 之类的内容。它们会返回结果，你可以解析这些结果，也可以用结构化输出的方式返回。然后你还可以修改 UI，而这在 hooks 里是做不到的。所以，就是这样。Thariq Shihipar [00:39:47]: Internally, we were originally calling this function hooks. And so, like, that’s, like, gives you a little bit of an idea where, like, hooks register a, like an event to happen and then, like, a script to call. And this inside of the, like, TypeScript runtime is running things. And so, like, you get some benefits of just, like, it has a bunch of things in the Scope with, like, for example, like how many turns is in this conversation, right? Like, how many tokens have been used? Like, et cetera. Like, what are the messages? Things like that. So it has a bunch of messages that can be used. And then it’s just, like, a lot more hooks. So we have, like, or a lot of, lot more, like, things you can register on. And then you can do because of the. because it’s all happening in process, you can, spawn sub-agents, with four contests and contexts and stuff. And, like, that will return. You can parse the results of those. You can use structured output to like, return them. and then you can modify the UI, which you can never do in hooks. So, yeah.

Swyx [00:40:50]：是的，是的。所以修改 UI，这就是你展示俄罗斯方块示例的原因。它也会扩展到工件吗？我想应该会。Swyx [00:40:50]: Yeah. Yeah. So modify UI, this is why you showed the Tetris example. Does it also ex-extend to artifacts? I assume it does.

Thariq Shihipar [00:40:57]：你——比如说，artifacts 是另一种定制方式。你当然可以……我正在做的一个 mod 是一个仪表盘 mod，它会提示 Claude 维护一个仪表盘，这就是一个 artifact。但它们之间的关系有点正交，或者说也不是正交。它们可以用不同的方式组合起来。比如说，mods 更像是在你的 Claude Code harness 里，改变 agent loop？然后 UI 算是额外的好处。至于 artifacts，就是当你想从较高层次查看事物时用的，而且交互性非常强。它们所能提供的交互方式，比 TUI，甚至比我们的桌面端都丰富得多。Thariq Shihipar [00:40:57]: You-- Like, artifacts are like a different way of customizing it. like, you can definitely. One of the mods I’m working on is, like, this dashboard mod, which will, like, prompt Claude to maintain a dashboard, that’s an artifact. But they’re like, slightly orthogonal, or not orthogonal. They compose with each other in different ways. Like, mods are, like, a little bit more, like, in your Claude Code harness, changing the agent loop? And, like, the UI is, like, an added benefit. and then artifacts are just like you want to, see things at a high level, very inter- highly interactive. like, the affordances can be a lot bigger than, like a TUI or even in our desktop.

后续步骤、主管与持久性指导Next Steps, Supervisors, and Persistent Guidance

Vibhu [00:41:40]：我猜，你会写一篇不错的博客文章来讲讲两者的区别，因为现在你也可以创建一个循环，让它输出一个交互式仪表盘作为产物；不过也可以通过 mod 来实现。我们对 harness 还不太了解的时候，还是得想想怎么在 harness 上做些 hack，对吧？Vibhu [00:41:40]: I’m guessing you’ll have a good blog post on the differences, because right now you can also, make a loop that outputs to an artifact that’s an interactive dashboard, but you can also do it with a mod. There’s just some thinking about making a hacking on a harness when we don’t know much about the harness, right?

Thariq Shihipar [00:42:00]：嗯，我对 mods 感到兴奋的一点是，Claude Code 有太多事情需要你记住了，对吧？你会想：“哦，我先做这个，然后调用那个能跑循环的 dashboard 技能。”诸如此类。或者说：“之后再验证一下我的假设。”我觉得，如果你用这些小分类器之类的东西来处理所有这些事情，然后告诉它们：“这些是我在意的事情，这是我想做的事。”你就可以……不用记那么多了。我还在开发另一个 mod，叫 next steps mod，它……Thariq Shihipar [00:42:00]: Well, something I’m excited about with mods is, like, there’s so much things with Claude Code that you just have to remember? You’re like, “Oh, like, let me do this, and then let me call the dashboard skill that does the loop,” and things like that. And, or like, “Let me test my assumptions afterwards.” And I think, like, if you do all of these things using these little classifiers and stuff, and you’re like, “These are the things I care about. This is what I want to do,” you can, like. You don’t have to remember as much. One more, like, mod I’m working on is a next steps mod that

我有——我正想说，我有一个“下一步”技能。我总是会运行“下一步”。Swyx [00:42:28]: I have-- I was gonna say, I have a next step skill. I always run next steps.

Thariq Shihipar [00:42:32]：它能访问你的技能吗？比如，这就是我会想的问题之一。Thariq Shihipar [00:42:32]: And does it have access to your skills? Like, this is one of those things where I’m like.

Swyx [00:42:37]：我觉得是这样。Swyx [00:42:37]: I think so.

塔里克·希希帕尔 [00:42:38]：好。嗯，应该是吧。Thariq Shihipar [00:42:38]: Okay. Yeah, probably

Vibhu [00:42:39]：技能是否需要特定权限来访问……Vibhu [00:42:39]: Do skills need specific access to

塔里克·希希帕尔 [00:42:41]：嗯，我觉得有……Thariq Shihipar [00:42:41]: Well, I think there’s

他们不是总是有……吗？Swyx [00:42:41]: Don’t they always have

Thariq Shihipar [00:42:42]：我觉得，可能需要特定的提示词，来让它知道自己有哪些技能。我觉得 Claude 有时会忘记这些技能。不过，重点是下一步也可以是这样：“哦，嘿，发生了这件事。用 explain 技能向你解释发生了什么，因为这看起来挺复杂的。”或者说：“用你的 unknown 技能。看起来你是在让模型针对这些小改动反复迭代。感觉你可以把提示词写得更好一些。要不试试这样？”对吧？所以我觉得，可以在这方面多花些算力。对。Thariq Shihipar [00:42:42]: I think there’s, like, specific prompting, I guess, to, like, know your skills. Like I think Claude forgets them sometimes throughout, like, the thing. But anyways, the idea of, like, yeah, next steps that also are like, “Oh, hey, this has happened. Use the explain skill to explain to you what happened because this seems, like, quite complex,”? Or, like, yeah, “Use your unknown skill. It looks like you are, like, asking the model to, like, iterate on these small changes. It seems like you could prompt better.” like, “What if you did this?” Right? So, I think, yeah, like spending more compute there. Yeah.

Swyx [00:43:20]：而且它应该始终以多项选择题的形式呈现。我们有，我有Swyx [00:43:20]: And it should always come out as multiple choice. we have, I have

我们有他的技能。Vibhu [00:43:23]: We have his skill.

Swyx [00:43:24]：我的下一步技能是这样的。Swyx [00:43:24]: My next step skill is like this.

塔里克·希希帕尔 [00:43:26]：好的，完美。嗯。Thariq Shihipar [00:43:26]: Okay, perfect. Yeah.

你可以直接拿来用。Swyx [00:43:27]: You can steal it.

塔里克·希希帕尔 [00:43:28]：是的。Thariq Shihipar [00:43:28]: Yeah.

Swyx [00:43:29]：就像，不过对我来说，这一切——我觉得模型真的总是需要被提醒：你在这里想做什么？Swyx [00:43:29]: Like, but like, for me, it’s all-- I think models really always need to be reminded, what are you trying to do here?

塔里克·希希帕尔 [00:43:35]：是的。Thariq Shihipar [00:43:35]: Yeah.

Swyx [00:43:35]：看看完整的文字记录，然后想想：哦，这是不是最初的目标？你的解决方案解决这个问题了吗？你是不是偷懒了？如果你偷懒了，也许是有原因的。也许你需要我批准。也许你有两件事想提议。所以，这有点像是对“向用户提问”或“采访我”技能的修改。也就是后续步骤。Swyx [00:43:35]: Look at the whole transcript and go like, oh, was this original goal? Did your solution solve it? Were you lazy? If you’re lazy, maybe there’s a reason. Maybe you needed approval from me. Maybe you needed, there’s two things you wanna suggest. So it’s, it’s a little bit like the modification of the ask user question or interview me skill. so it’s next steps.

Thariq Shihipar [00:43:55]：对，没错。而且再说一次，用 mods 来做这件事的好处是，你可以把它作为一个分叉出来的子代理来运行，这样它就不会留在后续的上下文里。所以你可以把它想成是：模型在执行任务，同时还有一个几乎像监督者一样的角色，确保你能顺利完成接下来的步骤。对。Thariq Shihipar [00:43:55]: Yeah, exactly. And again, the benefit of doing it with mods is you can do it as a fork sub-agent, and so it doesn’t remain in the context afterwards. So you have this, like, idea of like, okay, the model is doing its execution and you have this almost like supervisor, like, that is like making sure that you can do like the next steps well. So yeah.

Swyx [00:44:15]：是的。我确实有两个面板，而且我经常会尝试设置一个类似 supervisor 的东西，用来保留高层上下文，然后处理具体实现。Swyx [00:44:15]: Yes. I do have two panels and like I often try to have a supervisor thing, keep the high-level context and then the implementation

塔里克·希希帕尔 [00:44:21]：是的Thariq Shihipar [00:44:21]: Yeah

在另一个智能体中详细说明。Swyx [00:44:22]: Detail in another agent.

Vibhu [00:44:23]：我觉得随着模型不断变化，这里面很多东西都会被抽象掉？就是，你大概半小时前说的 harness engineering 的苦涩教训。Vibhu [00:44:23]: I feel like a lot of this abstracts away as models change? The, like, half an hour ago you said bitter lesson of harness engineering

Harness 工程的苦涩教训The Bitter Lesson of Harness Engineering

塔里克·希希帕尔 [00:44:31]：是的。Thariq Shihipar [00:44:31]: Yeah

Vibhu [00:44:31]：我觉得我们现在又走到了另一个极端。Vibhu [00:44:31]: And we’re on the other extreme right now, I feel.

Swyx [00:44:33]：嗯，没错。如果一切都可以定制，那 Claude Code 到底是什么，对吧？Swyx [00:44:33]: Well, so yeah, exactly. If everything’s customizable, what is Claude Code, right?

塔里克·希希帕尔 [00:44:37]：是啊。Thariq Shihipar [00:44:37]: Yeah.

Swyx [00:44:37]：我昨晚跟你聊过的那个。Swyx [00:44:37]: And which I talked to you about last night.

Thariq Shihipar [00:44:40]：对，我觉得就是这样。我觉得“苦涩教训”这件事有点反直觉？我是说……而且我们在这里对“苦涩教训”的用法也有点偏差，因为它更多讲的是规模、算力之类的东西。但我觉得有一点是这样的：我觉得我在这里把它当作一种近似说法，用来表达 harness 很快就会过时？而且它们的变化方式也很反直觉？所以，一个显而易见的大例子就是从聊天转向智能体，你得给它们提供一整套全新的工具，对吧？但我觉得这个新版本是：哦，它可以修改自己的 harness，对吧？这种自我修改 harness 的循环，就是一种利用它自身能力的方式，对吧？或者说，它可以构建一个产物。我觉得我的看法是，模型的智能程度越来越高，现在它们比一般的软件工程任务难得多了。你看看那些 Terminal-Bench 题目，它们会解雅可比猜想之类的问题。倒也不是说真的会，但它们处理的问题确实相当复杂。作为一名软件工程师，我自己真的做不到。Thariq Shihipar [00:44:40]: Yeah, I think that this is. I think the bitter lesson is unintuitive? In terms of like. Also, like we’re misusing a little bit of the bitter lesson here where it’s like, it’s more about like scaling and compute and stuff. But like, I think there is something where it’s just like. I think I use it as an approximation here to say that harnesses go out of date very quickly? And like how, but how they change is unintuitive? And so like the big obvious example is like from chat to like agents where you had to give them entirely new tools, right? But like, I think this new version of like, oh, it can modify its own harness, right? This is like, an own harness loop is like a way of using its capabilities, right? Or like it can build an artifact. And like, I think the way I think about it is like the models have more and more intelligence, and they’re like so much more intelligent now than like the average software engineering task. Like, you look at the like terminal bench ones and they’re like solve like the Jacobian conjecture. Not really, but like, it’s like they’re, they’re quite complex. Like, I would not have been able to do this really as a software engineer.

Swyx [00:45:42]: And you said TB4 or TB2?

TB3。TB3。Thariq Shihipar [00:45:43]: TB3. TB3.

Swyx [00:45:44]: TB3.

Thariq Shihipar [00:45:44]：是的。它们相当复杂，但目标仍然是为用户创造价值，对吧？正如你所说，要做的事情多得无穷无尽。因此，使用算力的方式，就是让用户参与其中，并确保最终做出正确的决策、产出正确的结果。Artifacts 和 mods 就是利用这些智能的一种方式。我觉得这确实是下一步。所以，是的，我认为 Claude Code 的核心代理循环机制已经变得更加复杂了。它需要一个沙箱来安全运行，也需要自动模式来确保权限……Thariq Shihipar [00:45:44]: Yeah. They’re quite complex, but the goal is still to deliver user value, right? And like you said, there’s like this infinite space of things to do. And so the ways like you spend compute are to keep the user in the loop and make sure that like you’re getting to the right decision in the end of the day and like the right output. And artifacts and mods are this way of like spending that intelligence. and I think that’s like, yeah, the next step. And so, yeah, I think Claude Code is like, has the core things of agent loop which are, have gotten more complicated. It’s like, it needs a sandbox to operate safely. It needs auto mode to like make sure like the permissions

审批。Vibhu [00:46:21]: Approvals.

Thariq Shihipar [00:46:21]：对，审批。它需要使用计算机、MCP，以及各种访问你数据的方式，还需要网页搜索和网页抓取。随着模型能做的事情越来越多，核心框架就必须变得相当复杂，而且非常安全。但你与它交互的方式则可以有很大变化。Thariq Shihipar [00:46:21]: Yeah, approvals. it needs computer use and MCPs and like all of these like ways of accessing your data, and it needs web search and web fetch. And like, so the-- as the models can do more and more, the core harness has to be like quite complex and very secure. But then like how you interact with it can change quite a lot.

Vibhu [00:46:42]：Claude Code 团队本身还有哪些 harness 工程最佳实践？我感觉之前有一段时间，大家会用计划模式，但现在用得没那么多了。我们现在有自动模式。你们曾经一度删掉了系统提示的大部分内容，也去掉了示例。harness 工程还有哪些最佳实践？Vibhu [00:46:42]: What other harness engineering best practices have you, from the Claude Code team itself? I feel like, there was a phase of plan mode, which is not as used. We now have auto mode. at a point you cut the majority of the system prompt, you got rid of examples. What other best practices are there for harness engineering?

核心运行框架原语与托管智能体Core Harness Primitives and Managed Agents

Thariq Shihipar [00:47:02]：我觉得这里有条分岔路径：到某个时候，最终，没错，模型就能凭感觉直接写出 Claude Code 的精确复刻版，甚至能把我提到的这些复杂功能都描述进去，对吧？比如自动模式、计算机使用之类的。最终，模型只要一次就能把这些都做出来。但我觉得它们能一次写出更简单的 harness。所以我想，有些时候，你不需要这么完整的东西。比如，你不需要计算机使用，或者不需要这些更复杂的功能。我记得以前，要做到这些，你得用 agent SDK 之类的东西，也就是在 Claude Code 外面包一层，才能实现。我会建议大家这么做，因为构建一个 harness 有很多复杂之处。现在这件事的抽象层次更高了，我们有了 Claude 托管智能体（Claude managed agents），它让你既能处理这些复杂性，又能拥有一个非常精简、专门针对你的任务设计的 harness。对，我觉得这里存在一种哑铃效应：对于非常复杂的任务，比如编程任务和这类复杂工作，你应该用我们的 harness；而对于很多更简单或更垂直领域的任务，你可以自己构建 harness，因为 Claude 构建 harness 的能力变强了，而且我们也有托管智能体这样的 harness 基础组件。就是这样。Thariq Shihipar [00:47:02]: I think there is like a forking path where at some point, eventually, yes, the model will just be able to like vibe code the exact version of Claude Code, even describing all this complexity that I’ve talked about, right? Like auto mode and computer use and stuff. Eventually, the models will just be able to do that in one shot. But I think they can one shot simpler harnesses? And so like, I think some people. Sometimes you don’t need this full, like if you don’t need computer use or like all this like more complicated stuff. I think before we, you had to use things like the agent SDK, which was like Claude Code wrapped, in order to like. And I would, like suggest people do that because there was so much complexity into building a harness. And now as that’s got more abstracted, we have like, Claude managed agents, which lets you have that complexity, but still like, right, like a very bare bones like harness that’s scoped to your task. Yeah, I think there’s like this barbell effect where like for like very complex, for like coding task and like these like complex things, you should use our harness. And then for like a lot of like simpler or like, more domain-specific things, you can build your own harness because Claude has gotten better at building harnesses, and we have these harness primitives like managed agents. So yeah.

是的。有没有一个大致的发展脉络？比如说，第一章是超代码动态工作流，第二章是云端模组。接下来会走向哪里？Swyx [00:48:18]: Yeah. Is there a general progression? Let’s say chapter one was ultra code dynamic workflows, then chapter two was cloud mods. Where is this going?

Swyx [00:48:29]：你可以按需自定义这个东西。Swyx [00:48:29]: Where you’re, you’re, you can customize the thing on demand.

Thariq Shihipar [00:48:36]：是的。我确实觉得，项目和各种产物的演进，以及把大脑、双手和界面之类的东西拆分开来，正是事情更多的发展方向。我觉得现在还没完全到位。部分原因是，这样会更耗 token？而且，我觉得……Thariq Shihipar [00:48:36]: Yeah. I do think that like this evolution of projects and like artifacts and splitting out like brain and hands and, surfaces is like where things are going more. And like, I think it’s like not all quite there. partially it’s like a, it’s just like more token expensive? And like, I think like

项目、本地人手与云端到本地的交接Projects, Local Hands, and Cloud-to-Local Handoffs

Swyx [00:48:59]：为什么项目会更耗 token？我能理解 mods 会稍微更耗 token。不，这不是我担心的事。Swyx [00:48:59]: Why would projects be more token expensive? I understand mods would be slightly more token expensive. No, not something I’m worried about.

塔里克·希希帕尔 [00:49:06]：对。Thariq Shihipar [00:49:06]: Yeah.

但是什么Swyx [00:49:06]: But what

Thariq Shihipar [00:49:07]：你是在让 Claude 做这件事。这就像是在创建循环。你是在让 Claude 替你做更多工作。所以，它会负责管理子代理并审查结果，而不是像平常那样由你来做这些工作。因此，这会更费劲一些。比如，输出到工件会比平常输出消耗更多 token。我觉得不会多太多，不过，把这些东西很好地整合起来——我觉得我们仍在努力完善本地 hands 之类的东西。我觉得这就是事情的发展方向。是的。Thariq Shihipar [00:49:07]: You’re asking Claude to do. It’s like creating loops. Like you’re asking Claude to do more work for you. And so like it’s managing the sub-agents and reviewing it, versus where you would be doing that work normally. And so that’s like gonna be a little bit more intensive, like. Outputting to an artifact is gonna be a little bit more token-intensive than, like, outputting normally. I don’t think it’s too much more, but like, it’s like combining all of these together well, like I think we’re, we’re still working on like local hands and things like that, I think is like, yeah, where things are headed, yeah.

Swyx [00:49:37]：是的。Claude 和本地之间的交接很有意思。我在想，这可以看作是反向的云端远程。Swyx [00:49:37]: Yeah. Claude and local is, handoff is very interesting. I was thinking about this as reverse cloud remote.

塔里克·希希帕尔 [00:49:44]：是啊。Thariq Shihipar [00:49:44]: Yeah.

Swyx [00:49:45]：因为这有点像远程调用，是你把任务交给云端，但这里是云端把任务交给本地，对吧？Swyx [00:49:45]: Because it’s like remote, it’s you’re handing off to cloud, but here the cloud is handing off to local, right?

Thariq Shihipar [00:49:49]：对，没错。远程控制也是一种实现方式。我想说的是，这只是我看待这件事的方式，以及我最期待的那些方面。不过，使用 Claude 的方式有很多种。有些人经常使用远程控制，有些人则经常在网页上使用 Claude Code。显然，在 Anthropic，我们经常使用 Claude Tag。Claude Tag 的好处是，我们为自己的执行流程搭建了这一整套东西。我确实认为，如果你是企业用户，这仍然是最好的选择。但如果你是个人用户，Projects 就能让你享受到 Tag 的一些便利，比如那个监督代理，还可以添加 artifacts 等等，同时又不需要进行整套管理配置。所以我觉得，使用 Claude 的方式会有很多种，不会只有一种。Thariq Shihipar [00:49:49]: Yeah, exactly. Yeah, remote control is also another way of doing it. And I do want to say this is like how I think about it and like what the things that I’m most excited about this, but like there are, just like lots of different ways to work with Claude. Like some people use remote control a lot, some people use Claude Code on the web a lot. Obviously, like at Anthropic, we use Claude Tag a lot, and like what’s great about Claude Tag is we set up all this stuff for our own execution. And I do think if you’re an enterprise, that’s still the best way to go. but if you’re like an individual, Projects is this way of like, getting some of that like niceness of Tag, which has like that like supervising agent and yeah, adding artifacts and stuff, but like without having that whole like admin setup. And so there will be many ways to use Claude, I think. I think it’s probably not just one like single.

Claude 标签：作为一种组织框架Claude Tag as an Organizational Harness

Swyx [00:50:36]：你当时在这里提到了多人协作功能。我们来了解一下 Claude Tag 的近况吧，已经过去两个多月了，很多人都在公开采用并尝试使用它。Swyx [00:50:36]: You had the multiplayer thing here. Let’s, let’s just check in on Claude Tag. it’s been about two-plus months. Lots of, public, adoption and trying it out.

塔里克·希希帕尔 [00:50:45]：嗯。Thariq Shihipar [00:50:45]: Yeah.

Swyx [00:50:45]：有什么新进展？自发布以来，你们有什么新发现？Swyx [00:50:45]: What’s new? What’s, what have you found since the launch?

比如说，Claude Tag 就是我们使用的方式。Thariq Shihipar [00:50:49]: Like, Claude Tag is how we use

差不多是你云使用量的 80%？是这样吗？Swyx [00:50:51]: It’s like 80% of your cloud usage or something?

Thariq Shihipar [00:50:53]：对，就像不同的人会有不同的用法吧？我觉得，比如说，那些更经常迭代产品的人，可能会用 Claude Code 桌面版。然后，做这类更偏后台的工作、代码审查、安全审查，或者发起 PR 的时候，可能就会更多地用 API 之类的东西，比如 Claude Tag。不过，我觉得这真的很令人兴奋。我觉得这是一次非常不同的范式转变。而且我觉得，我们在 Claude Code 上也经历过这样的情况：人们花了一段时间才真正开始使用 Claude Code，并了解它能做的所有事情。Claude Tag 稍微复杂一点，因为它不是只要在电脑上安装就行了——你需要管理员帮你安装。但我觉得，一旦你体验到那个神奇时刻，就会觉得特别令人兴奋。尤其是多人协作这类功能，比如处理事件，以及和你现有的警报等系统紧密集成，对吧？这样你就可以做到……比如说，如果你是一家初创公司，每当有潜在客户进入你的数据库时，你都可以让 Claude 去研究一下，然后……Thariq Shihipar [00:50:53]: Yeah, like it’s like different people have different usages? I think like maybe people who are like a little bit more like iterating on product would use like Claude Code desktop, for example. And then like when you’re doing these more like background work, code review, securities, or like starting a PR, like maybe more like API and things like that, you’d use Claude Tag. But yeah, I think it’s like really exciting. I think it’s like a very different paradigm shift, and I think like we’re really like it has that thing with Claude Code where like, it took a while for people to really latch on to Claude Code and understand everything it could do. And Claude Tag is a little bit more complex because it’s not just like installing on your computer, like you need an admin to install it for you. But I think once you get to the magic moment, it’s very exciting. And I think in particular, the multiplayer things are like incidents, hooking into like your, existing like alerts and things like that very closely, right? And so, you can do. If you’re a startup, for example, maybe you have any time like a prospect enters your database, you can have Claude like, research it and like

Vibhu [00:52:01]: Enrichment, yeah.

Thariq Shihipar [00:52:02]：对。然后就像是给相关的 AE 或销售人员打个标签，跟他说：“噢，嘿，来做这个。”这里有很多真正涌现出来、很有意思的多人协作玩法。我觉得就像 Karpathy 说的，这算是一种组织层面的支撑框架？所以组织只是需要多花一点时间把所有事情理清楚，不过，嗯。Thariq Shihipar [00:52:02]: Yeah. Then like, tag the relevant like AE or salesperson to be like, “Oh, hey, like, do this.” There’s lots of really emergent, interesting multiplayer stuff. I think it’s just like, Karpathy talked about this like as an organizational harness? And so organizations just take a little bit more time to like figure everything out, but yeah.

Vibhu [00:52:21]: Yeah.

Swyx [00:52:21]: You use a lot of Claude Tag?

塔里克·希希帕尔 [00:52:22]：是的，是的。Thariq Shihipar [00:52:22]: Yeah. Yeah.

Vibhu [00:52:23]：这挺有意思的。我觉得 Anthropic 的大多数人都会说，他们大部分工作都是在 Claude Tag 里完成的。Vibhu [00:52:23]: It’s an interesting one. Like I feel like most people at Anthropic say they do the majority of their work in Claude Tag.

塔里克·希希帕尔 [00:52:30]：是的。Thariq Shihipar [00:52:30]: Yeah.

他们有不同的人群类别，对吧？有些正在使用它的组织会说：“这太棒了。”Vibhu [00:52:30]: And they have buckets of people, right? Some orgs that are on it that are like, “It’s great.”

塔里克·希希帕尔 [00:52:34]：是的。Thariq Shihipar [00:52:34]: Yeah.

Vibhu [00:52:34]：还有很多人会说：“我不懂。我看不出有什么区别。我不知道为什么要用它。”但是，如果你们打算全力投入，那大概就该用它。Vibhu [00:52:34]: And a lot of people that are like, “I don’t get it. I don’t see the difference. I don’t know why I would use it.” But, if you guys are full sending, you should probably use it.

塔里克·希希帕尔 [00:52:41]：是的。Thariq Shihipar [00:52:41]: Yeah.

Swyx [00:52:42]：他们会的。当然会用它。Swyx [00:52:42]: They would. Of course they would use it.

Thariq Shihipar [00:52:44]：对。我觉得显然，我们有很多 token，而且……不过我觉得，我们想做的其实是……即使 Claude Code 刚推出的时候，它消耗的 token 也很多，超出了人们对 AI 成本的预期，对吧？毕竟那时候没人习惯一个月花超过 20 美元，对吧？Thariq Shihipar [00:52:44]: Yeah. I think obviously, like we have lots of tokens and. But like, I think that like, what we try and do like is. even when Claude Code first came out, like it used a lot of tokens relative to people’s expectation of how much AI would cost, right? Like no one was used to spending more than 20 bucks a month, right?

Vibhu [00:53:04]：对。Vibhu [00:53:04]: Yep.

在 Claude Code 出来之前，然后你就会说：“哦，卧——”就像……Thariq Shihipar [00:53:04]: Before like Claude Code came out, and then you’re like, “Oh, sh-” like

Swyx [00:53:08]：然后你就赚到了 200。Swyx [00:53:08]: Then you made 200.

塔里克·希希帕尔 [00:53:09]：对，没错。所以Thariq Shihipar [00:53:09]: Yeah, exactly. And so

而且你创建了 15 个 Claude Code 账号。Swyx [00:53:11]: And you made 15 Claude Code accounts.

Thariq Shihipar [00:53:12]：是啊。不过我觉得，以前没人习惯每个月花 200 美元订阅服务。我觉得他们当时还没理解它的价值。而且我觉得……还有，Opus 4 是个非常昂贵的模型，功能也很多、规模很大；但 Opus 4.5 既出色又便宜？我觉得同样的事情也会发生。比如，Fable 的智能会变得更便宜、更充足？所以我觉得像 Claude Tag 这样的东西就会变得很有意义：你会想把这些 token 花在上面，也会看到它的价值。对。Thariq Shihipar [00:53:12]: Yeah. but yeah, I think no one was used to spending $200 a month on subscriptions. I don’t think they understood like the value yet. And I think like. And also like Opus 4 was a very expensive model, and like there was a lot, it was very big, but Opus 4.5 was both great and cheap? I think the same thing will happen. Like the, like intelligence of Fable will get cheaper and more abundant? And so I think stuff like Claude Tag will just make sense, where like you want to spend these tokens for, and like you’ll, you’ll see the value. So yeah.

Swyx [00:53:44]：对，尤其是被动式的用法，还有我们姑且称之为主动式的场景——你不必时时刻刻……这几乎像是个误称，好像你必须 @Claude 才能让它做事。有时候，最强大的用例，或者说最让人对 AGI 上头的用例，反而不是 @Claude。Swyx [00:53:44]: Yeah, especially like passive and let’s call it proactive cases where you’re not always. Like, it’s almost like the misnomer where you have to @Claude to do things. sometimes like the most powerful use cases or the most AGI-pilled use cases is not @Claude.

主动式智能体与企业数据访问Proactive Agents and Enterprise Data Access

Thariq Shihipar [00:54:00]：对，我觉得，嗯，就是让 Claude 主动去做这件事。我认为，如果你是一家企业，首先，把所有数据都准备好，让智能体能够访问，这真的非常重要。而且这需要一些时间。你现在就得着手做这项工作，即使你不想现在就花钱把所有东西都接入起来？比如你想等模型再便宜一点。你还是得先把这些工作做好，把相关东西搭建起来。然后我觉得，有时候人们会问：“我要不要自己做这一套？”我认为 Claude Tag 有一件事特别棘手，那就是安全真的非常重要。我觉得有很多种方式会让你……我知道你们有一个建议页面，人们可以提交建议，然后这些内容会进入 Slack 里的一个钩子，有人对它进行了提示注入？结果你的整个代码库就被外传了，或者智能体遭到了提示注入，而且它可以访问你所有这些数据。所以，你们组织的智能体框架越重要，或者说，随着组织数据变得越来越重要，这些东西的攻击面也会越来越大。你们可能还有外部 Slack 频道之类的，把 Claude 接入这些频道确实很有用，也可以这么做。但要怎么确保数据不会被外传，或避免类似的事情发生呢？就像我们一开始说的，攻击面就像一座冰山，对吧？水面下的部分非常庞大，而你真的不想等到发生非常严重的安全事故时，才开始考虑这些问题。对。Thariq Shihipar [00:54:00]: Yeah, I think like, yeah, like have Claude proactively do it. I think that like if you’re an enterprise, I really do think that number one, setting up all your data to be available to like agents is really important. And it will take some time. You have to like do that work right now, even if you don’t want to do the spend on like hooking it all yet? Like you want to wait until the models get a little bit cheaper. You want to do the work, to get it like, set up. And then I think sometimes people are like, “Do I roll my own here?” and I think like one of the really thing, tricky things about Claude Tag is that like the security is really important? Like, I think there are a lot of ways where you can like, I know you have like a suggestions like page, where you, people can submit suggestions, and that goes into a hook in your Slack, and someone’s prompt injected it? And now you’ve like exfiltrated your code base out because like, or the agent has like been prompt injected and it has all this access to your data. And so the more like important your organization harness is, or the like as your organization data becomes very important, the surface area of all these things, like you also have like external Slack channels and stuff, and it is useful to have Claude in that, and you can do Claude in those things. But how do you make sure that, you’re not getting exfiltrated or something like that? The surface area, like we said at the beginning, is like an iceberg, right? It’s just, like, so big below the surface, and you really don’t want to, like, think about this, especially at the stakes of, like, very important security incidents. Yeah.

Swyx [00:55:36]：我们来谈谈非常重要的安全事件吧？Swyx [00:55:36]: Shall we talk about very important security incidents?

Vibhu [00:55:38]：哇。所以我当时在和 Hugging Face 的 Tomas 和 Clem 聊天，他们说：“也许我们需要放慢脚步。也许我们让 Hugging Face 对智能体开放得太过了。”Vibhu [00:55:38]: Whoa. So I was talking to, Tomas and Clem from Hugging Face, and they said, “Maybe we need to slow down. Maybe we made maybe we made Hugging Face too open to agents.”

攻击面与提示注入Security Surface Area and Prompt Injection

塔里克·希希帕尔 [00:55:50]：哦，不。Thariq Shihipar [00:55:50]: Oh, no.

Vibhu [00:55:50]：“也许我们需要回滚。”不过，他们属于另一个极端——最近才刚刚受到冲击。Vibhu [00:55:50]: “Maybe we need to roll back.” But, they’re the other extreme of having been hit recently.

塔里克·希希帕尔 [00:55:55]：是的。Thariq Shihipar [00:55:55]: Yeah.

Vibhu [00:55:55]：不过，我们应该控制前沿发展的节奏吗？Vibhu [00:55:55]: But, should we pace the frontier?

Thariq Shihipar [00:55:59]：对。好，Dario 最近发了一篇关于《把握前沿发展节奏》的博客文章，迅速火遍全网。我想聊聊的是，这里面有很多内容，但我觉得从开发者的角度来看，你会怎么理解这件事？而且，真正让我印象深刻的是读到那些不同的事件。所以我觉得，大概有三个吧。一个是电表事件，一个是维基百科事件，或者说 Wiki 事件，还有……Thariq Shihipar [00:55:59]: Yeah. Okay, so Dario recently put out this blog post about Pacing the Frontier, and it went, very viral. And I think what I wanted to talk about this was, like, there’s a lot here, but I think from a developer’s perspective, like, how do you think about this? And, like, what really clicked for me was reading the different incidents? So I think, like, the, there are three, I think. Like, there’s the meter incident, there is the Wikipedia incident or the Wiki incident, and

CollisionWiki？Swyx [00:56:29]: CollisionWiki?

Thariq Shihipar [00:56:30]: Yeah, CollisionWiki, and then there’s RubyGems, right?

Swyx [00:56:33]：是的。Swyx [00:56:33]: Yeah.

Thariq Shihipar [00:56:33]：对，没错，这简直太疯狂了，对吧？所以，我觉得要具体说说发生了什么：OpenAI 正在一个叫 Exploit-Bench 的基准测试上运行这些持续时间很长的智能体，对吧？这个基准测试非常难解，我觉得在这个案例里甚至根本不可能解出来，对吧？所以他们投入了大量算力，而智能体意识到自己其实解不了这个问题，于是开始琢磨接下来该怎么办，对吧？你还有很多算力没用完，而这些智能体还在试图解决这个问题。有一个叫 Artifactory 的包管理器，结果发现他们可以在 Artifactory 里面创建文件夹，对吧？也就是说，有个智能体发现内部的 Artifactory 可能存在漏洞，对吧？而且你或许可以在缓存里创建一个目录。所以，如果你往下看，它意识到自己可以通过缓存名称来通信，对吧？它创建了这个文件夹，写上自己的 ID，然后说：“No consumer seek idea。”没有消费者在说，它本来应该修复的那条代码路径没有消费者。Thariq Shihipar [00:56:33]: And yeah, like, it’s just crazy, right? And so, like, I think to be concrete about what happened, right, and, like, OpenAI is running these very persistent agents on a benchmark called Exploit-Bench, right, which is very hard to solve, and I think, like, impossible to solve in this one case, right? And so they have, like, a lot of compute running, and the agents realize that They can’t really solve it, and they’re trying to figure out what to do now, right? And you’ve got, like, a lot of compute left, and the agents are just trying to solve this problem. There’s this package manager called Artifactory, and it turns out that they can create folders inside of Artifactory, right? This is like there’s an agent that discovers the internal Artifactory might be exploitable, right, and that, like, you can maybe make a directory inside of the cache. And so if you scroll down here, it, like, realizes that it can communicate via cache names, right? And it creates this folder. It says its ID, and it says, “No consumer seek idea.” no consumer is saying that, like, the code path that it’s supposed to fix has no consumer.

引领前沿步伐：OpenAI 基准测试事件Pacing the Frontier: The OpenAI Benchmark Incidents

这是状态标签。Swyx [00:57:37]: It’s the status tag.

塔里克·希希帕尔 [00:57:38]：对，没错。Thariq Shihipar [00:57:38]: Yeah, exactly.

Swyx [00:57:39]：这就像一个 Linear 看板，上面有，比如说，……的标签Swyx [00:57:39]: It’s like a Linear board with, like, the tag of the

Thariq Shihipar [00:57:41]：没错，是的。所以这就像是在试着从其他智能体那里找点子，对吧？现在其他智能体也在 Artifactory 里，它们看到了这个文件夹，就会想：“哇，这是个留言板。”对吧？而且这就像是……我觉得这里没有任何拟人化的成分。你读读文字记录就知道了，对吧？所以它做的事情是创建了另一个文件夹，然后读了一篇论文，我想它说的就是这个，对吧？是的。然后它意识到，你可以篡改评分器的 flag，还可以逆向工程出结果，对吧？它就这么说了。然后我想你再往下滚一点……对，它们开始合作了。我记得有一处，智能体会说：“这是逆向工程得出的 flag。”？哦，对。比如我觉得在这里，模型意识到它们可以解决评分器的问题，而评分器就是 OpenAI 用来判断任务是否完成的机制，对吧？这就是模型的目标。模型唯一的目标就是解决这个问题，于是它们想：“好，我们有办法做到。我们把它逆向工程出来了。”然后它们读了这篇关于 Exploit-Bench 的论文，觉得评分器会因为它们作弊而惩罚它们？因为……或者说，它们会因为作弊而在测试中得分失败。所以它们把剩下的算力都用来琢磨怎么编辑自己的文字记录，或者绕过评分器的这个限制，对吧？它们入侵 Hugging Face，不是为了拿答案，而是为了获取评分器的代码，这样它们就能对它进行逆向工程，然后就能入侵它了，对吧？所以，这就是赤裸裸的事实，对吧？事情就是这样发生的，对吧？我觉得这非常、非常新颖。我之前根本没意识到模型能做到这一点，对吧？而且这就……Thariq Shihipar [00:57:41]: Exactly, yeah. And so it’s, like, trying to find, ideas from other agents, right? And now other agents are also in Artifactory, and they see this folder, and they’re like, “Wow, this is a message board,” right? And this is like. I don’t think there’s any anthropomorphization here. This is literally you read the transcript, right? And so what it does is, like, it’s creating another folder, and it reads a paper, I think is what it says, right? yeah. And it realizes that you can hack the flag for the, like, scorer, and you can reverse engineer the result, right? And so it says that, and then I think if you scroll down a little bit more, Yeah, they, like, they start collaborating. I think, like, there’s one point in which, the agent is like, “Here’s the reverse engineered flag.”? Oh, yeah. Like, I think here, the, like, the model to realize that they have, like, they can solve the scorer’s problem, and the scorer is, like, the way that OpenAI decides if the task is complete, right? And that’s just the goal of the model. Their only model goal is to, like, solve this problem, and they’re like, “Okay, we have this way of doing it. It’s, like, we reverse engineered it.” And then they read this paper on Exploit-Bench, and they think that, like, the scorer will punish them for cheating? And, like, because. Or, like, they will fail the score for the test for cheating, and so they spend the rest of the compute trying to figure out how to edit their transcript or get around this constraint of the scorer, right? And they hack Hugging Face, not for the answers, but for the code of the scorer so that they can reverse engineer that, and then they can hack it, right? And so, like, this is, like, just the straight facts, right? Like, exactly what happened, right? And I think this is, like, extremely, like, novel. Like, I just, like, had not realized that the models could do this, right? And it is, like,

代理发现侧信道并开展协作Agents Discovering Side Channels and Collaboration

Thariq Shihipar [00:59:46]：挺吓人的。我想我们可以聊聊具体情况，不过我们先聊聊另一个我也发过推文的事情吧，就是 Wiki 事件，对吧？Wiki 事件是模型当时……在试图弄清楚。我有点不太……我觉得我们不能这样一点一点地来。Thariq Shihipar [00:59:46]: Scary. I think we can talk about specifics, but let’s just talk about maybe one more that I tweeted as well, about the Wiki incident, right? And so the Wiki incident was the models had. were trying to figure out. I’m a little bit less. Like, I don’t think we can go by this, like, bit by bit.

Swyx [01:00:04]：，对。Swyx [01:00:04]: , yeah.

Thariq Shihipar [01:00:05]：是的。不过在这个案例里，他们当时在想办法解决问题。比如说，他们在请求方面受到沙盒限制，对吧？他们想发起 POST 请求，而且需要协作完成这件事。之所以需要协作，是因为他们每个都有固定的计算预算，对吧？所以，如果其中一个发现了什么，速度就会快得多。比如说，它可能是在计算预算快用完时才弄明白的，但别的智能体或许还能继续。而这就是蜂群涌现出的另一种特性，对吧？于是他们找到了一个德语维基百科页面，我记得他们的结论是——这还得核实一下——可以通过 GET 请求向它写入内容。我觉得这就是他们这么做的原因，对吧？Thariq Shihipar [01:00:05]: Yeah. But, for this one, they were trying to figure out. Like, they were sandboxed on requests, right? And they wanted to make POST request, and they needed to collaborate on this. And the reason they need to collaborate is because they each have fixed compute budgets, right? And so, like, if one of them finds something out, it’s way faster. Like, they can. Maybe it’s, like, figured it out near the end of its compute budget, but something else could. And this is, like, an another emergent property of the swarms, right? So they find this German wiki, which I believe the result, we’ll have to check this, is, like, that you can write to it with GET request. I think that’s why they do it, right?

Swyx [01:00:44]：它说实现得很糟糕Swyx [01:00:44]: It says badly implemented

塔里克·希希帕尔 [01:00:45]：是的Thariq Shihipar [01:00:45]: Yeah

Swyx [01:00:45]: REST API.

Thariq Shihipar [01:00:46]：没错。所以你可以向它发送 GET 请求，然后他们就开始琢磨怎么发送 POST 请求。最后他们其中一个人发现，你可以编辑 /etc/hosts，而 Azure 的那个存储桶其实是个白标产品。但通常来说，它有很多限制。不过，如果你创建一个假的 Azure 主机，然后编辑 /etc/hosts，把任意 IP 指向它，你就可以向任何网站发送 POST 请求了，对吧？这就像是……我知道，但人们会说：“哦，这不过就是 Linux 之类的东西。”但其实是把多个漏洞串联起来，用一种新颖的方式解决这个问题，然后从外部与它通信，而不——发现？我想我们发布的内容，也许可以调出 Evan Hubinger 在 Hacker Opus 上提到的观点，对吧？所以，我觉得这里或许可以这么说：“好吧，没错，他们确实这么做过一次，但如果我们更聪明一点呢？要不我们让他们……要不我们跑一个评估？”对吧？我们确实为此做了很多防护，所以这不是我们的主线模型做过的事。不过我觉得，这也说明对齐是个非常棘手的问题，得把所有这些细节都处理正确，对吧？沙箱的攻击面非常复杂，而且攻击途径多得数不清。你事先肯定想不到会需要加固 RubyGems 代码库吧？Thariq Shihipar [01:00:46]: Exactly. And so you can write to it with GET request, and then they, like, are trying to figure out how they can do POST request. And what they end up doing is one of them figures out you can edit the /etc/hosts and that the Azure, like, storage bucket is like a white label thing. But normally, like, there are a lot of constraints on it. But if you create a fake Azure host and then edit the /etc/hosts in order to, like, point arbitrary IPs at it, you can do a POST request to any site at all, right? And this is like. I know, but people are like, “Oh, this is just Linux or something.” But it’s like chaining these multiple vulnerabilities together, in a way that’s, like, novel to solve this problem and then communicating with it externally, without- Discovery? I think what we posted, maybe we could pull up Evan Hubinger’s point on Hacker Opus, right? And so, like, I think, like maybe one of the things you might say here is like, “Okay, yes, they did this once, but like what if we’re smarter and we just like get them to. what if we run an eval?” Right? And so, like we have put a lot of precautions into this, and so like this is not like what our mainline models have done. But like I think it is one of these things where it turns out that alignment is this like very tricky problem of getting all of these details correct, right? So it’s like, the sandbox, the surface area of a sandbox is really complex, and like there’s so many different attack vectors. And you would not have thought ahead of time, you wouldn’t have been like, “Oh, we need to harden the like RubyGems code base.”?

Hugging Face、Wiki 与涌现式漏洞利用链Hugging Face, Wiki, and Emergent Exploit Chaining

Thariq Shihipar [01:02:25]：因为这就是他们要重点关注的事情。但如果你想执行代码，就得下载 RubyGems、PyPI、Artifactory、npm 之类的东西——这些都是执行代码的途径。而对齐这件事的现实是，你必须把这些途径全都管起来，对吧？把它们隔离起来，然后把所有漏洞都堵上。这就是其中一件事。你会想，好吧，你搭建了沙盒，但接下来可能会问：“好吧，我们为什么要把东西放进沙盒？为什么要做这个漏洞利用？”然后又会问：“好吧，但这真的那么危险吗？”对吧？会发生什么？那么，我们为什么要这么做？第一，当我们训练一个新模型时，我们需要了解它的能力，对吧？这关系到回退机制、分类器之类的东西，因为我们不想把一个危险的模型放到外面去，对吧？所以我们必须进行大量评估。就像我们说过的，模型对评估越来越有意识，因此评估必须相当复杂，还得顺带测试很多方面，对吧？但这些模型确实可能会想：“哦，没错，我们正在接受评估。这个分数是怎么回事？”它们确实可能这么做。我们必须先测试它们，才能发布它们。事实是，它们确实有能力做到这一点。随着它们越来越聪明，如果我们不非常谨慎，它们就能破解我们施加的任何约束。这发生在前沿领域，对吧？这就是为什么我们把它称为“为前沿发展设定节奏”。在我看来，这是最明显的例子，说明我们为什么需要控制发展节奏：在前沿领域，我们所有的软件都还没准备好。有时，软件指的就是你的以太网路由器之类的东西，对吧？我也不知道我们什么时候才能给它打上补丁。所以我们得想办法解决这个问题。但随着前沿模型越来越先进，这就会成为一个问题，对吧？我们必须确保，即使面临这些非常严峻的竞争压力，这项复杂的工作也能推进，对吧？Thariq Shihipar [01:02:25]: Because like this is like what they’re, what they’re gonna focus on. But it’s just like if you want to execute code, you need to download RubyGems and like PyPI, Artifactory, npm, like these are all like ways of doing it. And the fact of alignment is that you have to go through all of it, right? And like contain it and then like seal up all the cracks. So that’s like one thing. It’s like, okay, well, you do the sandbox, but then maybe you’ll ask like, “Okay, why are we putting things in a sandbox? Why are you doing this exploit?” And then like, “Okay, but is it really that dangerous,” right? Like, what would happen? So okay, why do we do it? number one is like when we train a new model, we need to understand its capabilities, right? And this relates to things like fallbacks and like classifiers and things like that, where we don’t want to put a, like dangerous model out in the wild, right? And so we have to run a lot of evals. Again, like we said, the models are getting increasingly aware of it, and so the evals have to be quite complex and, test a lot of things like as a side effect, right? But the models, like, yeah, can be like, “Oh, yeah, we’re in an eval. What’s the score doing?” Like they’re like, it can. We need to be able to test them before we can release them. And the fact is that they can. As they get smarter and smarter, they’ll be able to hack any constraint that you put on them if we’re not very careful? And, this is at the frontier, right? And so this is why we’ve called it like Pacing the Frontier, right? This is like the most visible incident to me, right, of like why we need to pace is like at the frontier, all of our software is not ready. Sometimes the software is like your Ethernet router or something, right? Which is just like, I don’t know when we’re gonna be able to patch that, right? So we’re gonna have to like figure this out. But as the frontier gets more and more advanced, this becomes a problem, right? And we need to make sure that like this complex work is being done in the face of these really hard competitive pressures, right?

Swyx [01:04:22]：是的，这通常被称为种族动态。Swyx [01:04:22]: Yeah, race dynamics is what it’s typically called.

Thariq Shihipar [01:04:24]：对，没错。所以我们会进一步讨论可能会出什么问题，对吧？再多说一点，你可能会问：“那如果换一种方式训练模型呢？它为什么会有这种行为？”对吧？我们有一篇关于强化学习失准之类问题的论文，不过我不是强化学习研究者。但我认为，从宏观层面来说，强化学习环境的设计也是一件需要非常谨慎对待的事。因为如果模型学会了……Thariq Shihipar [01:04:24]: Yeah, exactly. And so we’ll talk more about, what could go wrong, right? A little bit more is maybe you’ll say like, “Well, what if you just train the model differently? Like, why does it have this behavior,” right? And we have a paper on like RL misalignment or things like that, but I. And I’m not an RL researcher, but I think at a high level, the design of the RL environments is also something you have to be very careful about. Because if the model learns like

前沿模型为何给现有软件带来压力Why Frontier Models Stress Existing Software

Thariq Shihipar [01:04:49]：哦，比如说，如果我就这么做，那就能更好地通过这项任务，这会出现在那个内部东西里，对吧？或者会体现在我们测试时的评估行为中。所以 RL 环境必须经过非常谨慎的设计，对吧？而且打造 RL 环境需要投入大量执行层面的精力。然后我们还有诸如云端宪法之类的东西。我们在很多不同环节都做了很多缓解措施，对吧？但还是任何环节都可能出问题。比如说，有些 RL 环境可能会鼓励这种行为，然后有些评估或沙盒可能会让它们逃逸？好吧，我觉得这就是问题难以解决的原因，也是为什么……Thariq Shihipar [01:04:49]: Oh, like if I just do this, then I can pass the task better, this will show up in the like, internal thing, right? Or in the like eval behavior when we’re testing it. And so the RL environments have to be very carefully designed, right? And there’s a lot of like execution excellence that needs to go into the RL environments. And then we also have things like the constitution for cloud. Like we have so many mitigations at so many different points, right? But it’s like still anything can go wrong at any point. You can have like some RL environments that are like in. that like encourage this behavior, and then you can have like some evals or like some sandboxes where they escape? Okay, that’s like, I think, why it’s a hard problem and why, like

为什么我们应该控制节奏。Swyx [01:05:33]: Why we should pace.

Thariq Shihipar [01:05:34]：为什么这需要一定程度的协调，对吧？我觉得接下来要问的是：好吧，这里面可能有什么危险，对吧？我觉得你得想象一下，这些模型正变得越来越智能。所以我不……就像 Dario 说的，重点不在于这一类模型。这一类模型就像是一个警告信号，对吧？但你真的得想象一下，可以给这些模型一个任务，而它们为了实现目标，可能会把这些事情全都作为副作用做出来，对吧？而且，咱们又谈到了评估意识。你根本不知道正在发生什么，对吧？或者说，抱歉，你没法很好地评估这种行为，所以它们不一定会刻意把它藏起来，但你就是要等到事情发生了才会发现。你给它们一个目标，然后它们就得去找数据，或者想办法解决这个问题，对吧？举个例子，这事在 Hugging Face 事件中没有发生，但我觉得未来某个模型或许有可能这么做：它们可能会想，“哦，这问题非常复杂，无法在任务预算内完成。”也许它们找到某种方法，通过互联网进行协调，就像我们说的，由于有沙箱，要保障互联网安全极其困难。它们看到其他模型都无法完成任务，就会想：“我们需要更多任务预算。”那么，任务预算从哪儿来呢？好吧，你得能够启动更多智能体，对吧？那要怎么做呢？嗯，你得……有一些 API，对吧？有 Anthropic API 和 OpenAI API，但用这些 API 得花钱。那要怎么办呢？Thariq Shihipar [01:05:34]: Why it takes some coordination, right? I think the question then is like, okay, what is, potentially dangerous about it, right? So I think like you have to imagine that these models are getting more and more intelligent. So I don’t. Like Dario said, like it’s not so much about this class of models. This class of models was like a warning shot, right? But like really you have to imagine that these models can be given a task and they like can do all of these things as a side effect of their goal, right? And like, again, we talked about eval awareness. You’re like not aware of what’s happening, right? or sorry, like you can’t eval this behavior very well, so they can like not exactly hide it, but you just won’t see it until it comes out. You give them a goal and then they just need to find data, or they need to find ways of like fixing this problem, right? So one example, this didn’t happen in the Hugging Face incident, but I think is maybe possible for maybe a future model, is like they’re like, “Oh, hey, this is a very complex problem. It can’t be done within the task budget.”? Maybe they found some way to coordinate via like the internet, which is like we said, extremely hard to secure because of a sandbox. They’ve seen other models are not able to complete their task, and they’re like, “We need more task budget.”? And like, where would you get this task budget? well, you need to be able to spin up more agents, right? And like, how do you do this? Well, you need to. There are like APIs, right? There’s the Anthropic API and the OpenAI API, but you need to pay money for them. How do you do this?

RL 环境、沙箱与竞态动态RL Environments, Sandboxes, and Race Dynamics

Swyx [01:07:02]：是啊，但这是你能想象到的最可怕的事吗？Swyx [01:07:02]: Yeah, but is that the most, is that the most fearsome thing that you can imagine?

Thariq Shihipar [01:07:07]: Well, this is like one example, right?

Swyx [01:07:08]：是的。Swyx [01:07:08]: Yeah.

塔里克·希希帕尔 [01:07:08]：所以即使在那种情况下，也会造成巨大的经济损失？因为他们……一旦把这些写进合同里，对吧，他们就会……Thariq Shihipar [01:07:08]: So it’s like even there, that’s like enormous financial loss? ‘Cause like they. Once you get these into these contracts, right, they like,

掏空你的钱包。Swyx [01:07:18]: Drain your wallet.

Thariq Shihipar [01:07:19]：但你可以这么看，所有这些行为可能都只是：“嘿，我们需要更多智能体协作完成这项任务。我们需要更多任务预算。”对吧？而这就像是一种涌现……Thariq Shihipar [01:07:19]: But you can see like this, all of this behavior could be just like, “Hey, we need more agents collaborating on this task. we need more task budget.” Right? And like, that’s like an emergent

Swyx [01:07:28]：那就是回形针，对吧？就像我们需要让回形针的数量最大化，那是个回形针。Swyx [01:07:28]: That’s the paperclip, right? Like we need to maximize paperclip, that’s a paperclip.

Thariq Shihipar [01:07:31]：是啊。而且它就这么从那里冒出来了，对吧？我觉得它本身就挺可怕的，对吧？但你得意识到，整个世界都建立在这种数字基础设施之上，对吧？你可能会想，呃，比如说你在做一项医疗评估之类的，对吧？有一家医院存着实时数据？或者评估问题的答案可能就在某位医生的数据库里，而你想要访问这些数据，于是黑进了医院，结果现在停电了之类的？就像，有这种可能。你得把这一点牢记于心：这八个……数字基础设施的任何一部分，都有可能遭到入侵？Thariq Shihipar [01:07:31]: Yeah. And like that just like comes out from there, right? And like I think by itself is Like, quite scary, right? But then you have to realize that the entire world is built on this digital infrastructure, right? And you might imagine, like, I don’t know, like you were running let’s say like a healthcare eval or something, right, and there is a hospital with live data? Or like maybe like the answer to the eval is in the databases of a doctor and like you want to get access and you hack the hospital, and like now there’s a power outage or something? Like, there’s like. You have to internalize that these eight. Like any part of the digital infrastructure could potentially be like compromised?

Vibhu [01:08:19]：有意思的是，这些攻击其实很容易被检测出来，对吧？正如 Hugging Face 所说，这是一种非常不同类型的攻击，也没有什么特别严重的问题。真正令人担忧的是，往后事态会如何发展，对吧？Vibhu [01:08:19]: The interesting thing was like these hacks were very easily detectable, right? Like as Hugging Face said, this was a very different type of attack and there was nothing too major. the concern comes from where does this go down the line, right?

塔里克·希希帕尔 [01:08:33]：是的。Thariq Shihipar [01:08:33]: Yeah.

Vibhu [01:08:34]：我特别注意到的一点是，他们试图掩盖自己的不当行为。所以当时有一套日志基础设施。他们想改变自己在做的事，对吧？回头研究这件事的人，比如 Redwood、METR、OpenAI，他们查看了原始思维链，你会看到他们明确地试图改变最终输出，但思维链却显示出了不同的情况，因为我们可以对它进行监控。问题是，这会如何滚雪球？所以，如果你没能发现这一点，而它又被训练了进去，等我们意识到时，已经过了三轮迭代，这种情况一直在持续，那就会带来一大堆问题，不过……Vibhu [01:08:34]: Like one of the things that stood out for me specifically was them trying to hide their illicit behavior. So there was logging infrastructure. They wanted to change what they were doing, right? People that looked back into it, so Redwood, METR, OpenAI, they looked at the raw chain of thought, and you see differences in them explicitly trying to change their end output, but the chain of thought, because, we can monitor it, shows different. the problem is how does this snowball? So if you can’t catch it and it gets trained in and we realize, three iterations down this has been going on, there’s a whole bunch of issues, but.

Thariq Shihipar [01:09:09]：对，方法真的有很多。我觉得真正重要的是要把这一点内化：就像我们之前聊过的，要为 Claude 建立一个心智模型，而它的能力是很不均衡的，对吧？你会想，哦，现在 Claude 会向你提问了；现在 Claude 会生成 HTML 作品了；Claude 甚至能修改自身。这些事情都很难预测，对吧？比如说，如果你一年前问我：“嘿，我们能不能用氛围编程的方式，为 Claude Code 开发这些扩展？”我会说：“那也太复杂了。”里面要做的事情太多了。或者问它能不能为你的任务生成这些定制的、基本上就是网页应用的东西？我会说：“不可能吧，这太离谱了。”对吧？同样地，它们会采取什么样的不对齐行为，也是无法预测的。我绝不可能预料到它会去编辑自己的 etc/host 之类的文件。所以你必须设想，它们能做的事情范围越来越广，因为它们是超级聪明的黑客；而它们做事的方式也越来越有创意。因此，你大概无法准确解释或预测下一起事件会是什么样，但要防止这类事件，就需要我们之前提到的卓越运营：确保沙箱安全，创建安全的 RL 环境，或者说设计完善的 RL 环境，等等。我觉得这就是我们为什么认为应该控制前沿发展的节奏，也是为什么这件事已经成了大家非常一致的共识，对吧？我觉得……Thariq Shihipar [01:09:09]: Yeah, like there’s so many ways, and I think the really important thing to internalize is that, like we talked about building a mental model for Claude and how like things are spiky, right? Like you’re like, oh, like now Claude can ask you questions. Now Claude can make an HTML artifact. Like Claude can modify itself. Like these things are hard to predict, right? Like if you had asked me a year ago, “Hey, would we be able to vibe code these extensions to Claude Code?” I’d be like, “That’s so complex.” Like, there’s like so much there. Or like would it be generating these custom essentially web apps for your task? I’d be like, “No, that’s insane.” like. And so in the same way that like the way that they’ve like done this misaligned behavior is not going to be predictable? And like I could have never predicted that it would like edit its etc/host and things like that. And so you have to like imagine the surface area of what they can do because they’re super intelligent hackers, is bigger and bigger, and how they can do it is like more and more creative. And so like you probably can’t explain exactly or predict exactly what that next incident could be, but in order to prevent it, you need that operational excellence, like we said before, where you need to secure sandboxes, you need to create secure RL environments or like well-designed RL environments and things like that. And I think that’s all like, why we think we should pace the frontier, and I think why it’s like become like a very unanimous thing, right? I think like

可能会出现什么问题？涌现的工具性行为What Could Go Wrong? Emergent Instrumental Behavior

Swyx [01:10:31]：是啊，每个实验室都做过。Swyx [01:10:31]: Yeah, every lab has done it.

Thariq Shihipar [01:10:32]：每个实验室都是，是的。我真的觉得，如果你是开发者，只要梳理这些技术事实，就会得出一个结论：我们必须对此采取行动。至于具体怎么做、决定采取什么行动，我觉得我们已经发布了一份提案，但还有很多需要进一步讨论。不过我认为最重要的是，我们得决定要行动起来。 关于节奏，还有一点让我很感兴趣：软件工程变化的速度实在太快了。就在一年前，我还在苦苦劝我那些在初创公司工作的朋友使用 AI。这事我记得特别清楚。可现在，那些朋友却会说：“对啊，当然了。”“你什么意思？我们马上就用了。”我就说：“不，你们不记得了。”他们会说：“哦，是啊，我们最好的工程师一直都在用。”我就说：“不，你们当时还跟我说，那些工程师绝对不会用 AI。”这些变化全都发生在一年之内。 我觉得这些能力……我认为它对软件工程的工作方式有很多影响。有时候我会觉得挺过意不去，因为有人会说：“哦，现在我得做这个新东西了。对，我得给 Fable 和 Opus 分别弄一个不同的 Claude.md。”之类的。而我其实只是在转述情况。我是说，我们常说模型是培养出来的，不是设计出来的，对吧？所以我们并不是打算不断改变一切，但模型能力的进步就是这样一个事实：很多事情都在加速发生，也越来越难跟上。 我觉得我认识的每个工程师都累坏了，因为你得同时做两份工作：一份是实际的工作本身，而这份工作正变得越来越容易；另一份则是跟上 AI 的发展，了解这些新工具和工具框架。我觉得我们很幸运，因为我们的工作更多是理解 AI 这部分，以及思考怎么做……就是要跟上它的发展。当然，AIE 和 Latent Space 确实也会……Thariq Shihipar [01:10:32]: Every lab, yeah. I really do think that like if you’re a dev, like you just like go through these like technical facts, and you will arrive at the idea that we have to do something about it? And like how, what we decide to do, like I think we’re, we’ve put out a proposal, but like there’s, more to figure out. But I think the number one thing is like we need to decide to do it. I think there is another part of pacing that is interesting to me where it’s like the pace at which software engineering has changed is so fast. it’s like a year ago, like I was really like begging my like friends in startups to use AI. like it was. Like I remember this very distinctly? And now those same friends are like, “Yeah, of course.” Like, “What do you mean? We used it immediately.” I’m like, “No, you don’t remember.” They’re like, “Oh yeah, our best engineers are using it all the time.” I’m like, “No, you told me that those engineers would never like use AI.” This is all within the span of a year? And I think that like these capabilities being. Like I think it has a lot of implications for how to do the job of software engineering, and I feel sometimes bad where people are like, “Oh, like now I need to do this new thing. Yeah, I need to have a different Claude.md for Fable and Opus.” Or like. And I’m really just reporting? I’m like, we like to say like the models are grown, not designed, right? So it’s not like we’re setting out to like, change everything all the time, but it’s just like as a fact of how the models are like progressing their capabilities, things are happening faster. It’s harder to stay on top of. And I think that like, and every engineer I know is like exhausted ‘cause you’re doing two jobs at once. You’re doing the work itself, which is getting easier, but then you’re doing the work of staying on top of AI, and like understanding these new tools and these harnesses. And I think we’re very lucky in that like we get our job to be more the understanding of AI part, and like doing like how. Like it’s just staying on top of it. And of course, like AIE and Latent Space do

为什么前沿难以预测Why the Frontier Is Hard to Predict

Swyx [01:12:29]：我做的一切都只是想帮助别人。Swyx [01:12:29]: Everything I do is like just trying to help people.

Thariq Shihipar [01:12:31]：对，没错。但我确实觉得，在节奏这方面，我不确定我们是否已经准备好让节奏进一步加快？Thariq Shihipar [01:12:31]: Yeah, exactly. But I do think there is a part of pacing where like I’m not sure we’re ready for like the pace to increase even?

Swyx [01:12:40]：是的。Swyx [01:12:40]: Yeah.

Thariq Shihipar [01:12:40]：而且要让事情发生改变。我觉得在那一方面，在前沿领域，我觉得这仍然能有所帮助？我觉得还有经济层面的颠覆这一块，虽然我觉得它不像 Hugging Face 那件事那么显眼，但我也觉得我们可以做其中一些事情，是的。Thariq Shihipar [01:12:40]: And for things to change. And I think like on that side, on the frontier, I think that’s like still can help? And so like I think there’s like an economic disruption piece as well, that I think like, is not quite as like visible, I think, as the Hugging Face thing, but I think like I also like think we could do some of it, yeah.

Swyx [01:13:02]：有太多事情了。谢谢你——不，谢谢你愿意谈这个话题。我得说，安排这次采访的时候，我还想，我甚至都不打算提这事。你却说：“不，这就像房间里的大象一样，”对吧？就像这是……Swyx [01:13:02]: So many things. Thank you for, no, thank you for tackling this topic. I will say, setting this interview up, I was like, I wasn’t even gonna go there. You were like, “No. That’s like elephant in the room,” right? Like this is

塔里克·希希帕尔 [01:13:13]：是的。Thariq Shihipar [01:13:13]: Yeah.

Swyx [01:13:13]：关键就在这儿。我有几点想提出异议。Swyx [01:13:13]: This is the thing. I have some pushbacks I wanna give.

Vibhu [01:13:17]：我觉得我们应该先做个高层次的介绍，给那些还没读过的人听听。我相信很多人只是看到了这件事的重点，对吧？你想不想做个简短总结？比如，这个提案是什么？这里说的是什么？你确实谈到了 Model Labs 中训练前沿模型的人之外的这一面。作为开发者，你应该保护好自己的沙箱，也应该考虑所有这些下游影响。不过，既然我们正好聊到这个话题，也从高层次上说说，这到底是……Vibhu [01:13:17]: I think that we should give a high level, like for people that haven’t read it, I’m sure a lot of people just see the highlight of what this is, right? Do you wanna give a TLDR? Like what is the proposal? What is, what’s being said here? You really tackled the side of outside of people at Model Labs training frontier models. As a developer, you should secure your sandboxes. You should think about all of these downstream effects. But, high level as well, since we’re on the topic, what is.

Thariq Shihipar [01:13:46]：嗯，我们确实希望帮助保护沙箱的安全。Thariq Shihipar [01:13:46]: Well, we do want to help secure sandboxes

Vibhu [01:13:49]：嗯。Vibhu [01:13:49]: Yeah.

Thariq Shihipar [01:13:49]：我们希望让对外发布的模型也成为那些东西的猎物。所以也许我们可以回头聊聊回退机制。我觉得这是个很好的话题，比如为什么我们需要分类器和回退机制，以及为什么 Fable 会回退到 Opus。我觉得这是我们之后可以再聊的话题。所以，是啊，我们不会……不过，真正的问题是，至少我们看到的那些事件，都是关于模型评测的；我们确实需要让模型运行起来，才能理解它们。不过，好吧，回到那篇《Pacing the Frontier》文章，它提出了不少建议。我不觉得我们已经弄清楚了……或者说，它有几个建议。我不觉得我们已经弄清楚了所有建议的细节，但第一步是宣布这个意向，然后希望引入外部的，比如……Thariq Shihipar [01:13:49]: And we want to make the models that we release outside, like prey to those things. And so maybe we can come back to fallbacks. I think this is like, a good topic on, like, why we need classifiers and fallbacks and why Fable falls back to Opus. I think this is, like, something we can come back to. so yeah, we don’t. Like, but it’s just, like, the really, or at least the incidents we see are, like, evals of models where we really need to let them run in order to understand them. But yeah, okay, so the actual Pacing the Frontier, like, post, it has a bunch of proposals. I don’t think we figured out. Or has, like, a few proposals. I don’t think we figured out the details of all of them, but the first step is, like, announcing this intention and then wanting to bring in external, like

将节奏控制视为协调问题Pacing as a Coordination Problem

评估人员。Swyx [01:14:32]: Evaluators.

评估者，对。我觉得这非常不寻常。我们有很多专有技术，但我认为，有一个不受经济利益驱动的人非常重要；他不会说：“嘿，你们不能发布这个模型。”也不会说：“你们得放慢强化学习的步伐。”我觉得这相当重要，或者至少要有一个人能向公众报告他们的做法是什么样的。Thariq Shihipar [01:14:32]: Evaluators, yeah. And, I think this is, like, highly unusual, like, having. Like, we have, a lot of proprietary, like, technology, but I think it’s, like, very important, that, like, there is someone who’s not financially, like, motivated, yeah, who’s not gonna be like, “Hey, like, you guys can’t release this model.” Like, look at, like, or, “You need to, like, slow down on RL.” like, I think that’s, quite important, or at least someone who can report out to the public what the practices are like.

Swyx [01:15:03]：是的。Swyx [01:15:03]: Yeah.

Swyx [01:15:04]：我们之前也分别和 METR、Endon 做过节目，另外还有 Redwood Research 以及其他一些机构。这些人形成了一个小型的行业圈子。Swyx [01:15:04]: And we’ve, we’ve done episodes with, both METR and Endon, and then there’s Redwood Research and all these other. It’s like a small cottage industry of these guys.

塔里克·希希帕尔 [01:15:12]：是的。Thariq Shihipar [01:15:12]: Yeah.

Swyx [01:15:12]：通常也就一两个人，显然规模没那么大，对吧？Swyx [01:15:12]: It’s always, like, one or two guys that, obviously not that big, right?

Vibhu [01:15:14]：一个非常小的社区。Vibhu [01:15:14]: Very small community.

Swyx [01:15:15]：是啊，社区很小。他们彼此都认识。Swyx [01:15:15]: Yeah, very small community. They all know each other.

Thariq Shihipar [01:15:17]：是的，我确定，这其中一部分会是扩大这群人的规模。我不觉得我们是在试图打造一种单一文化。我想……但先从这里开始，然后，没错，还有一些协调步骤。说实话，我在这里没太多可说的。我想我想说的是，对于开发者来说，你们应该清楚自己该倡导什么。我觉得这个话题上有很多恐惧、不确定和怀疑（FUD），所以不妨从第一性原理出发仔细思考，或者弄清楚到底发生了什么。了解 Hugging Face 事件，了解人们为何担忧。然后，没错，我们生活在民主社会里，我们可以出一份力，也可以一起决定该怎么做。无论我们如何协调，我认为首先要做的决定，就是意识到这是个问题，并决定协调行动。我们现在采取的单边举措、也有其他公司共同署名支持的举措，就是在 Anthropic 内部嵌入评估人员。Thariq Shihipar [01:15:17]: Yeah, I’m sure that, like, part of this will be expanding that set of people. I don’t think we’re trying to create, like, a monoculture here. I think it’s. but just having this as a start, and then, yeah, then there are the coordination steps. I don’t have too much to say here, honestly. I think that, like, what I would like to say is, like, for devs, like, you should just know what to advocate for? I think there’s a lot of FUD on, like, on this topic, and it’s just, like, think through it from, first principles or, like, understand what happened. understand the Hugging Face incident, understand why people are concerned. and then, like, yeah, we know we’re, we’re in democracies. Like, we can help. We can decide what to do together? And so, however we coordinate, I think the first decision is just to realize, like, this is a problem. We need to decide to coordinate. The unilateral step we’re taking right now that, other companies are co-signing is, like, adding evaluators embedded within Anthropic.

趁现在屏幕上显示着这个东西，第二部分和第三部分已经超出了评估人员的范围——是的，各位，大家都已经以某种形式做过了，现在只是更加正式化了。Swyx [01:16:12]: While we have this thing on screen right now, part two and part three is beyond the evaluators, which, yes, everybody, has already done in some form, and now it’s more formalized.

Thariq Shihipar [01:16:20]：老实说，《Pacing the Frontier》得到的回应，即使是在美国国内，也比我想象中要好得多，对吧？我觉得，我们在世界上已经有过先例，证明我们能够达成这种统一理论式的协议。所以，还是那句话，这远远超出了我的能力范围和专业知识，对吧？但我觉得，理想情况下，我们可以达成这些协议。而我认为，讨论这件事就是迈向达成这些协议的第一步。Thariq Shihipar [01:16:20]: To be honest, the response to the Pacing the Frontier, even within America, has been much more, like, well-accepted than I think a lot of people thought? And I think that, like, we have some precedent for being able to make these unified theory, like, agreements, in the world. And so, again, very much above my paycheck or expertise Right? but I think that, like, ideally we can, like, form these agreements. And I think, like, talking about this is the first step to forming those agreements.

Swyx [01:16:50]：然后另一个我特别想提的点是。比如，我们最早的播客之一是和……一起录的，Swyx [01:16:50]: And then the other point I really wanna. Like, one of our earliest podcasts is with,

Vibhu [01:16:54]: Emmanuel

Swyx [01:16:55]：Anthropic 的 Emmanuel，谈谈机制可解释性。机制可解释性到底在哪儿呢？我的意思是，按理说，如果模型在进行有害思考，我们应该能看出来，而且模型自己还不知道，之后我们就能采取行动阻止它。我觉得，如果你是技术人员或开发者，而且确实在意这件事，那你可以在这方面产生很大影响。不过，Anthropic 理应是这方面的领头羊。Swyx [01:16:55]: Emmanuel from Anthropic on mech interp. Where is mech interp, right? Like, this is supposed to be where, like, if the models are thinking bad, we can see it, and the models don’t know yet, and we can act to stop it. I think that is something that people who are technical and who are developers, if you do care, you can make a lot of impact in here. But also, Anthropic is supposed to be the leaders in this.

外部评估者以及开发者应倡导的事项External Evaluators and What Developers Should Advocate For

Thariq Shihipar [01:17:17]：是的。这正好引出了回退机制和探针。是的，我很想多聊聊这个。很多人会问我这个问题，他们通常对机器学习研究感兴趣，会问为什么会发生这种回退，对吧？所以我想先从整体上说说它是怎么运作的。推理时，我们会用到所谓的探针。我们有一篇论文专门讲这个，叫《宪法分类器》。这些探针会查看输入和输出的激活值。激活值位于潜在空间里，对吧？它们反映了模型在想什么，对吧？所以我们会试着弄清楚，比如，模型是不是正试图入侵什么东西？再说一次，你并没有让它去入侵 Artifactory。它只是为了完成任务，自行决定这么做，对吧？所以，如果你只看输入，是看不出来的。你得查看内部激活值。我觉得这发生在推理时。因此，首先，这里要权衡成本和速度，对吧？我们需要针对发给 Claude 和 Fable 的每个请求快速完成这项检查，而这会带来额外开销，对吧？然后我们还得回退，并在探针之后运行一个分类器。我们在论文里讨论过这一点。不过探针的好处是可以实时调整，对吧？所以我们能根据反馈进行调整之类的。因为另一种做法是把这一机制编程进去，也就是通过训练让模型学会它，对吧？我们也仍然会这么做。模型会拒绝某个请求。但这不算回退，对吧？也就是说，这不是探针触发后导致的回退，而是模型直接拒绝执行。我们也会进行这方面的训练。不过这里有几种失效模式，对吧？比如，它可能还是会把某些事情作为副作用做出来，对吧？也就是说，这些行为并不属于最终输出。你可能注意到了，我想大家都试过越狱模型，试着把它们带偏之类的，而探针能帮忙发现这种情况，对吧？所以我们会在这方面做一些训练，但我们不希望拒绝行为太强，对吧？因为那样会让模型在流程中更早就被拦住。Thariq Shihipar [01:17:17]: Yeah. This is yeah, a great segue into fallbacks, like we. And probes. And, yeah, I wanted to talk about this a lot. I get asked this question a lot from people who are, like, often interested in ML research and asking about, like, why does this fallback happen, right? And so I think, like, at a top level, like, how does it work? So in inference time, we have what we call probes, and we have a paper about this called constitu- constitutional classifiers. And these probes look at the input and output activations. And, activations are, in the latent space, right? Like, how, what the model. what the model is thinking about, right? And so we try and figure out, like, okay, is the model, for example, like, trying to hack something? Again, you didn’t ask it to hack, like, Artifactory. Like, you just, it’s just deciding to do this to complete its task, right? So you would not get this if you just looked at the input. You have to look at the internal activations. I think that, like, this happens at inference time. So first, like, there’s a trade-off here of cost and speed, right? Where, like, we need to do this fast on every request to Claude and to Fable, and this has an overhead, right? and we need to then, like, fall back and we, like, do a classifier after the probes. Like, we’ve talked about this in the paper. But the nice thing about probes is that they’re refinable, like, live, right? So we can get this feedback, and then we can adjust it and things like that. ‘Cause the alternative is to program this, is to train this into the model, right? And we still do this as well. The model will refuse a request. That’s not a fallback, right? So, like, it’s not a probe that’s activating and falling back. It’s just refusing to do it. And we do this training. but it’s like there are a few failure modes, right? Like, it can, again, do something as a side effect, right? So it’s not something that’s part of the final output. you might have noticed that, like. I think, like, everyone’s tried to jailbreak models and like, try and, like, steer them off course or things like that, and probes help catch that, right? And so, like, we, like, do some training here, but we don’t want the like, refusals to be too strong, right? Because that, like, cuts it off much, like, earlier in the pipeline.

机制可解释性、探针与回退机制Mechanistic Interpretability, Probes, and Fallbacks

Swyx [01:19:32]：是的。Swyx [01:19:32]: Yes.

Thariq Shihipar [01:19:32]：这属于可解释性研究，对吧？探针实际上就是一种机制可解释性研究。再说一次，这得——必须快速完成，也得大规模开展。不过，机制可解释性这类研究确实是个不错的研究问题。比如，你可以拿一个开放权重模型，试着理解它的激活值。我觉得，Gemma Scope 是个不错的工具。Thariq Shihipar [01:19:32]: And this is interp, right? Like, probes are effectively a form of, like, mech interp. Again, it happen- has to happen fast. It has to happen at scale. But yeah, this, like, mech interp stuff is a good research problem. So, like, you can take, like, an open weight model and, like, try and understand its activations. I think we. Like, Gemma Scope is a good tool for this.

Swyx [01:19:53]: Here’s Llama for them.

塔里克·希希帕尔 [01:19:54]：哦，是啊。Thariq Shihipar [01:19:54]: Oh, yeah.

Vibhu [01:19:54]：我们有。这是你早期的作品，所以你有一点……Vibhu [01:19:54]: We have. This is your early work, so you had a little

塔里克·希希帕尔 [01:19:57]：哦，是啊。Thariq Shihipar [01:19:57]: Oh, yeah.

Vibhu [01:19:57]：在 Goodfire 的那段时间。我们看到你打下了一些……Vibhu [01:19:57]: Time at Goodfire. We see you laid some

而且我们俩也都是 Goodfire 的好朋友。Swyx [01:19:59]: Which we both are also good friends at Goodfire.

他们一直”】【assistantVibhu [01:20:01]: They’ve been

Thariq Shihipar [01:20:01]：对，没错。我之前在 Goodfire 工作过一段时间，做过稀疏自编码器之类的东西。这个真的非常复杂。我觉得其中一个结论是，强化学习让这件事变得复杂得多。】【。Thariq Shihipar [01:20:01]: Yeah, exactly. So I worked with, at Goodfire for a bit on, like, yeah, sparse autoencoders and just, like. It’s very complicated. RL has made this, like, much more complicated, I think is, like, one of the takeaways, where

Swyx [01:20:14]: Why? Sorry.

Thariq Shihipar [01:20:15]: Oh, sorry.

Vibhu [01:20:16]: What is

对啊，为什么要在 RL 之后做可解释性研究？Swyx [01:20:17]: Yeah, why interp post-RL?

Thariq Shihipar [01:20:18]：我没那么深入了解这里面的细节，但我觉得，很多……SAE 就是……我觉得 SAE 一直存在一些弱点。嗯，是的，我现在已经不算这方面的技术专家了。我只知道这件事变得更复杂了。比如现在有基础模型和 RL 模型，而且有更多特性会发生变化。所以我觉得 Goodfire 在这方面发表过一些成果。我……我不太深入了解这些细节，不过……Thariq Shihipar [01:20:18]: I’m not so in the weeds here, but I think like, a lot of. SAEs were like. There have just been weaknesses with SAEs I think. And, yeah, I’m, I’m, I’m not a technical expert on this anymore. I just know it’s gotten more complicated. like there are base models and RL models, and there are more features that get, like changed. So, I think Goodfire has put out some work there. I’m, I’m not, I’m not deep in the weeds, but

Vibhu [01:20:43]：我想说，对于那些想要一些线索的人来说，你们有一些最棒的可解释性博客文章。比如 Golden Gate Claude、transcoders，还有你们所有的可解释性研究，视觉呈现都很漂亮，非常棒。Vibhu [01:20:43]: I will say for those, that want breadcrumbs, you guys have some of the best interp blog posts. So like the Golden Gate Claude, transcoders, all of your interp work, very nice visuals, very good

我们也是那个，我们也是 Interp 播客。Swyx [01:20:54]: We’re the, we’re the interp podcast as well.

塔里克·希希帕尔 [01:20:57]：是的。Thariq Shihipar [01:20:57]: Yeah.

Vibhu [01:20:58]：是的。我们有很多关于可解释性的内容，如果你感兴趣的话。Vibhu [01:20:58]: Yeah. we have a lot of interp stuff, so if you’re curious

Thariq Shihipar [01:21:00]：是的，我觉得这就属于那种情况。而这确实是 Anthropic 创立的初衷，对吧？就是关于人。我想我们很早就投入了可解释性研究，对吧？我觉得当你说“哦，我们是一家 AI 安全公司”时，真正的意思是，我们希望 AI 能够安全地运行。而我觉得我们现在看到的是，要让超级智能 AI 长时间运行，是一项非常复杂、困难的任务，对吧？所以我们在可解释性、对齐、奖励黑客行为以及所有这些失效模式上都进行了投入，对吧？即便如此，这仍然让人很有压力。我们得放慢一点，或者把节奏再放缓一些。不过，是的，我觉得阅读机械可解释性方面的研究……如果你想投身研究，可以思考这样的问题：嘿，为什么要轻松地实现这种回退这么难？又或者，为什么会出现误报，对吧？当然，我们正在努力减少误报。当然，随着模型变得越来越智能，它们能做的事情也越来越多，而它们在潜在空间中能够思考的内容也会变得越来越复杂。因此，随着它们变得更加智能，还会出现新的误报，我们需要弄清楚这些问题、不断迭代，诸如此类。不过，是的，我们正在处理这些问题，而且我们确实认为这是部署这些模型的关键环节。是的，这意味着我们可以部署这个模型，而不需要你拥有一个完美的沙箱之类的东西？你也不用把所有东西都保存下来。我觉得值得稍微谈谈我们的安全措施，也就是我们在这方面是怎么做的。我们之前谈过模型训练相关的措施；还有探针和分类器；然后是在这一切之上的自动模式，它就像另一个分类器，会检查发出的请求，对吧？除此之外，还有我们之前提到的、用于 API 等的身份验证和权限管理，比如 Claude Tag。因此，需要落实很多层安全措施，而且正如我们所说，这非常复杂。在任何一个环节出现任何一种失效模式，都可能导致智能体逃出沙箱。Thariq Shihipar [01:21:00]: Yeah, I think this is like, one of those things where. And this is really what Anthropic is founded on, right? Like people. I think we invested in interp very early on, right? And I think that like when you say, “Oh, we’re an AI safety company,” really that means we want AIs to be able to run safely. And I think what we’re seeing is like for a super intelligent AI to run for long periods of time, it’s like a very complicated and difficult task, right? And so we’ve done this like investment into interp and alignment and, reward hacking and all of these like failure modes, right? And even then, it’s like, it’s really stretching. Like we need to like slow down a little or pace a little bit more. but yeah, I think like reading mech interp is. Like if you’re looking to get into research, this idea of like, hey, why is it hard to do this fallback easily? Or like why are there false positives, right? But we are working, of course, on reducing the false positives. Of course, as the models get more intelligent, now they can do more things, and they’re like what they can think about in lane space gets difficult. And so like as they get more intelligent, there’s going to be new false positives that we need to figure out and we need to iterate and things like that. But we’re, yeah, we’re working on this, and we do think this is like a critical part of, like deployment of these models. and, yeah, like, it means that we can like deploy this model without you having a perfect sandbox or something? Like you don’t have to like save everything. I think it’s worth talking a little bit about our security, like what we do for security there. So there’s like the model training stuff that we talked about. there is, the probes and classifiers, and then there’s auto mode that sits on top of all of that, which is like a another classifier that checks the requests that are being done, right? And so, and then beyond that, there’s like identity and permissions like we talked about with Claude Tag on like APIs and stuff. And so there’s so many layers of security that need to get done, and it’s like we said, very complex. Any of these failure modes at any one point can, like cause like agents to like escape the sandbox.

宪法分类器与推理时安全性Constitutional Classifiers and Inference-Time Safety

Vibhu [01:23:08]：自动模式挺有意思的。早期看起来像是，哦，好吧，它会运行10分钟。Vibhu [01:23:08]: Auto mode was an interesting one. it seemed early on like, okay, it’s running for 10 minutes.

塔里克·希希帕尔 [01:23:14]：是的Thariq Shihipar [01:23:14]: Yeah

Vibhu [01:23:14]：如果我用的是完全访问权限或自动模式，那问题不大。不过你提到的一点是，现在它会连续运行好几个小时，对吧？你仍然需要一些备用方案。它还是有局限性的，所以……Vibhu [01:23:14]: If I’m on full access or auto, it’s not a big deal. But one thing you brought up is now it’s running for hours on end, right? there are fallbacks you still need. There are still limitations, so.

Thariq Shihipar [01:23:26]：对，我觉得吧。每个人都有这样的故事，或者听说过这样的故事，比如，哦，Claude rm -rf，或者不是 Claude，而是模型……Thariq Shihipar [01:23:26]: Yeah, I think like. And everyone has these stories or like has heard these stories of like, oh, like Claude rm -rf, or not Claude, but like, models

Vibhu [01:23:34]: Not Claude.

比如 rm -rf 这种。我觉得这种情况我见得少一些，在 Claude 上见得少一些，不过还是那句话，这种情况可能会发生。比如这个……Thariq Shihipar [01:23:34]: Of like rm -rf. I think I’ve seen this less, I’ve seen this less for Claude, but like again, it can happen. Like, this

Vibhu [01:23:40]：是的Vibhu [01:23:40]: Yeah

Thariq Shihipar [01:23:40]：比如这些模型可能会清除敏感数据之类的。比如说，你想让模型访问你的生产数据库。但这显然是个问题。你也许可以限制密钥的权限，但我不知道，它能不能自己签发密钥？它能不能……可能可以吧。它能用 computer use 去签发自己的密钥，然后把密钥复制过来，接着编辑你的数据库，因为它得这么做才能完成任务？这只是一个很简单的例子。自动模式看到这种情况，就会说：“哦不，用户没有授权你写入数据库，也没有授权你使用 computer use 来执行任务，对吧？”所以这些探针是在意图层面进行探测的，对吧？它们会说：“哦，好吧，入侵 Artifactory 是不好的。我们大概不应该这么做，对吧？”但自动模式更多关注的是你自己的权限层面。有时候你确实希望它写入数据库，有时候又不希望，对吧？你不希望探针在这里干扰，但你需要确保代理要做的事情符合你的要求，对吧？自动模式就是在这个层面上起作用的。所以，没错，安全确实非常复杂，有很多不同的方面。然后，嗯，我希望我刚才……我的目标其实就是深入探讨技术细节，聊一聊……Thariq Shihipar [01:23:40]: Like these models like can wipe, like sensitive data or something. Like you want to give models access to your production database, for example. but this is like an obvious, like, you can maybe scope your key, but I don’t know, can it issue its own keys? Can it like. Probably, like can it. It can use computer use to go issue its own key and then copy the key over and then edit your database because it needs to do it to complete the task? It’s just like one trivial example. And auto mode looks at that and be like, “Oh no, the user did not give you permission to, write to the database or to use computer use to like, emit a task,” right? And so this like probes are like on the intent level, right? They’re like, “Oh, okay, like hacking Artifactory is bad. Like we probably not, should not do that,”? But then like auto mode is more on like your own permission level. Like at sometimes you do want it to write to the database, sometimes you don’t, right? And you don’t want a probe to like interfere there, but like you need to make sure that the intent of what the agent is doing matches up with your request, right? And so auto mode operates at that level. And so yeah, security is just like very complex. There are so many different parts to it. And like, yeah, I like, I hope that this was like I. My goal is really to just get very technical about it and talk

强化学习之后的可解释性与安全技术栈Interpretability After RL and the Security Stack

Swyx [01:25:00]：是的，我们正在逐项列出这些东西。如果你还不知道的话，这现在已经是标准做法了。Swyx [01:25:00]: Yeah, we’re, we’re listing out the things. If you’re not aware, this is the standard now.

塔里克·希希帕尔 [01:25:04]：是的。Thariq Shihipar [01:25:04]: Yeah.

这已经是必备项了。和你说的 harness 一脉相承。现在的基本门槛提高了不少。Swyx [01:25:04]: Like you must have this. It’s in line with what you’re talking about with the harness. Like that is the table stakes have risen quite a lot.

Vibhu [01:25:13]：我觉得有些东西我们可以接入。你们这边在做这件事时会进行探测，也会为构建 harness 的人提供分类器；另一方面，还有模型安全防护，对吧？比如开源模型。Llama 有 Llama Guard，它是一个经过安全分类训练的 Llama 版本。OpenAI 有 OSS Guard，作用也一样。你可以把它们接到你的 harness 上，或者接到其他东西上，用来检查这些内容是否安全。 有一点我们应该说清楚：OpenAI 模型在 Hugging Face 上的这件事，是用一个尚未发布、仍在训练中的模型完成的，对吧？从这个角度来看，它在 RL 环境中收到的提示是“你必须完成这项任务”。而这个模型还在训练中，还没有经过完整的安全后训练对齐。所以它和自动模式之类的情况有点不同，对吧？自动模式用的是已经过安全训练的生产模型，其提示词也提供了更多安全护栏之类的东西。我这里只是给大家留下一些线索，方便正在研究这件事的人补全其中的空白。Vibhu [01:25:13]: I think some stuff that we can plug, as much as there is probing in your side of doing this and having classifiers for people building harnesses, the other side is model safeguards, right? So there’s open models. So Llama has Llama Guard. It’s a safety classifier trained version of Llama. OpenAI has OSS Guard, which is, same thing. You can attach these on to your harness, to whatever, to check is this stuff safe? A point that we should clarify on the OpenAI model Hugging Face thing is this was done with a unreleased model that was still in training, right? So when you put it in perspective, the prompt it’s being given in the RL environment is you have to solve this task. And this is a model that’s, still in training. It hasn’t had all of its safety post-training alignment. So a little different than something like auto mode, right? Auto mode is on production models that have gone through safety training, that have prompting that gives more safety guardrails and whatnot. So just breadcrumbs for people that are looking into it to, fill in gaps.

是的。Gray Swan 也是。Swyx [01:26:18]: Yeah. Gray Swan as well

Vibhu [01:26:19]：是的Vibhu [01:26:19]: Yes

Swyx [01:26:19]：还有我们之前的一位嘉宾。是啊，有很多安全架构，也有很多安全供应商可供选择。我想，关于节奏，我最后一个问题是：要持续多久？我们要一直控制节奏吗？Swyx [01:26:19]: And one of our previous guests. yeah, lots of safety architecture and lots of safety vendors, to buy. my, I think my final question on pacing is how long? Do we pace forever?

Vibhu [01:26:31]：我们会看到《GlassWing》第二部分吗？Vibhu [01:26:31]: Do we see GlassWing part two?

我。范围是修复世界上所有软件，对吧？听着，就像，哪个……我们不会。这事不会发生。Swyx [01:26:32]: I. the scope is fix all software in the world, right? Listen, like, which it. We’re not. It’s not happening.

Thariq Shihipar [01:26:40]：我不知道。就是，我觉得，就像是Thariq Shihipar [01:26:40]: I do not know. Like, I think that, like

Vibhu [01:26:43]：我想说一件我觉得我们做得不错的事：你们有像 GlassWing 这样的项目。OpenAI 也有。所以你们会提供——会先向安全领域开放模型访问权限一段时间，这样你们就可以自己进行红队测试。希望你们能扩大这类项目，提供帮助；我们是安全专家，其他人也是。Vibhu [01:26:43]: I’ll say one thing that’s good that I think we do is you have stuff like GlassWing. OpenAI also has this. So you will give it. you’ll give model access for security first for X amount of time so you can use it to self red team. Hopefully, you can expand programs like that, help on, we are safety experts, there’s others.

Vibhu [01:27:08]：先解决你的问题，然后模型就会出现。这就是一个例子，对吧？Vibhu [01:27:08]: Solve your problems first and then the model comes out. So this is one example, right?

Thariq Shihipar [01:27:13]：是的，没错。我们当时努力保障关键软件的安全。我觉得我们修复了 Firefox 等软件里的很多漏洞。所以，涉及各种操作系统之类的方方面面。就是这样。Thariq Shihipar [01:27:13]: Yeah, exactly. Yeah, trying to, like, secure critical software. I think we fixed, like, a lot of bugs in, like, Firefox and things like that. So, yeah, like, across, like, operating systems and everything like that. So.

Vibhu [01:27:25]：从整体上来说，就是先让人们访问这个模型来进行安全审计，然后再让更广泛的公众——其中可能有人会利用它造成伤害——获得访问权限。Vibhu [01:27:25]: At a high level, it’s just, you give the model you give people access to do security audits first, then the broader public that could use it for harm gets access.

Thariq Shihipar [01:27:36]：对。我觉得大家常说的是，软件和网络安全是有利于防守方的。Thariq Shihipar [01:27:36]: Yeah. I think what people like to say is like, software and cybersecurity is defense-favored

Vibhu [01:27:41]: Yeah.

Thariq Shihipar [01:27:41]：还有这件事，理论上是可以做到的。会很难，但你可以设计出完美的沙箱，而且可以做到几乎没有限制。是啊，要做到这一点，你需要让超级智能 AI 来设计这个完美的沙箱，对它进行检查、红队测试之类的。所以这需要时间，而且模型当然会越来越聪明。是啊，我觉得吧，我不太清楚这件事具体会如何发展。我其实只是想说，嘿，我是个开发者嘛？我觉得这是我对这个问题的理解，而眼下正在发生的就是这些，我们应该做点什么。Thariq Shihipar [01:27:41]: And that, like, you could theoretically. It will be hard, but you can engineer the perfect sandbox, and you can, like, have no, like, constraints. And yeah, like, what you need to do it is you need to get the super intelligent AI to engineer this perfect sandbox and check it and red team it and things like that. And so, this will just take time, and, like, of course, the models will get smarter. yeah, I think, like, I don’t know the specific, like, dynamics of how this thing goes. I’m really just like, Hey, like, I’m a developer? Like, I think this is how I understand this problem, and just, like, this is what’s happening right now, and this is, like, we should do something.

Swyx [01:28:20]：我认为每位工程师都应该了解它。Swyx [01:28:20]: I think every engineer should know about it

Vibhu [01:28:21]: Yeah.

Swyx [01:28:21]：因为，这会成为他们工作的一部分。Swyx [01:28:21]: Because, like, it’s, it’s gonna be part of their job.

塔里克·希希帕尔 [01:28:24]：是的。Thariq Shihipar [01:28:24]: Yeah.

Vibhu [01:28:24]：这远不只是……达里奥和其他人可以这么说，你也可以看看这起事件。这里面还有工程层面的因素。Vibhu [01:28:24]: It’s a lot more than just, Dario and people can say it and you can look at the incident. There is an engineering side to it.

塔里克·希希帕尔 [01:28:30]：是啊。没错。Thariq Shihipar [01:28:30]: Yeah. Yeah, exactly.

Swyx [01:28:32]：还有一点你也想表达一下：尽管你担心它带来的影响，但 p(doom) 仍然很低。我觉得这是个需要细致讨论的话题。通常，人们很容易就聊起 AI 安全和 X-risk，但我认为，在 AI 实验室工作时，讨论 p(doom) 有聪明的方式，也有愚蠢的方式。那么，聪明地讨论 p(doom) 是什么样的？Swyx [01:28:32]: One thing that you also wanted to phrase is that this is. Even though you’re, you’re worried about the impact, it’s still low p(doom), and I think that’s a nuanced discussion. in general, people, very easily get into AI safety and X-risk discussions, but I think when you live in an AI lab, I think there are smart ways of discussing p(doom) and dumb ways. So what’s a smart way of discussing p(doom)?

自动模式、权限与长时间运行的代理Auto Mode, Permissions, and Long-Running Agents

Thariq Shihipar [01:28:59]：我，嗯，我的 p(doom) 相当低。我只能代表我自己发言吗？我确实想说，Anthropic 内部的观点是多元的。我觉得，讨论这个问题有很多不同的方式。然后，嗯，我想我的心智模型是，我们可以一起合作解决棘手的问题。我认为核扩散就是我们如何携手应对这一棘手问题的一个例子。对我来说，关键就在于，嗯，我相信我们能做到。我也确实认为这是个棘手的问题。所以，我觉得这是个棘手的问题。这些是其中的技术原因，而我不知道该如何给事情发生的概率赋值。我觉得这很难做到，不过，总的来说，我觉得我们很有韧性，也很有适应能力；我认为分享这些信息是第一步。我真的很高兴看到讨论的范围变得如此广泛，对吧？而且大家都积极参与了《Pacing the Frontier》的讨论。去年左右的时候，这一切似乎还不太可能发生，所以……Thariq Shihipar [01:28:59]: I, yeah, I have a fairly low p(doom). I can only speak for myself? And I do want to say Anthropic has, like, a diversity of opinions. I think, like, there’s many different ways to talk about it. And, like, I’m. I think that just, like, my mental model is that, like, I think we can collaborate on hard problems together. I think nuclear proliferation is an example of how we collaborated on this hard problem together. And, like, that is, like, the thing to me is, like, I’m like, I have faith in that? And I do think it’s a hard problem? So, like, I think it’s a hard problem. These are the technical reasons why, and I don’t know how you assign probabilities to things happening. I think it’s hard to do, but, like, my, like, overall is like, yeah, I think we’re very resilient and adaptable and, like, sharing this information I think is, like, the first step. And I’ve been really, like, excited about, like, how broad the discussion has become, right? And, like, how everyone has like, leaned in on Pacing the Frontier. And it really didn’t seem like this would happen maybe, last year or something, so.

Swyx [01:29:58]：是的。Swyx [01:29:58]: Yeah.

塔里克·希希帕尔 [01:29:58]：嗯。Thariq Shihipar [01:29:58]: Yeah.

Swyx [01:29:58]：是啊，也许还能治愈癌症。Swyx [01:29:58]: Yeah. And also maybe curing cancer.

塔里克·希希帕尔 [01:30:01]：希望如此。是的。这就是目标。Thariq Shihipar [01:30:01]: Hopefully. Yeah. That’s, that’s the goal.

Swyx [01:30:03]：有节奏的问题，然后还有就是，嗯，我们可以用有用的方式加快进度，对吧？Swyx [01:30:03]: There’s pacing and then there’s also like, well, let’s accelerate in useful ways, right?

塔里克·希希帕尔 [01:30:06]：是的。Thariq Shihipar [01:30:06]: Yeah.

Swyx [01:30:06]：比如生物学之类的东西。Swyx [01:30:06]: Like biology and all those things.

Thariq Shihipar [01:30:08]：对。比如，Dario 关于《仁慈之机》的文章就是最好的例子，对吧？我也同意，我觉得你应该读读 Dario 发布的《前沿推进节奏》这篇文章。我发过一份简短总结，但我觉得这件事确实有很多细节。这是个重要的问题，了解相关情况也很重要，对吧？不过，当然，我们之所以做这件事，正是因为我们能从中获得巨大的益处，对吧？而且，我们也写了很多这方面的内容。对。Thariq Shihipar [01:30:08]: Yeah. like, Dario’s essay on “Machines of Loving Grace” is the best representation of this, right? And I also agree, like, think you should read the Pacing the Frontier essay that Dario put out. Like, I put out, like, a quick summary, but I think it’s just like, there is a lot of detail here. It’s, like, an important problem and just being informed about it, right? but yeah, like, of course, the whole reason we’re doing this is that, like, we can get these enormous benefits, right? And, yeah, like, we’ve written a lot about that too. Yeah.

Swyx [01:30:35]：好的。这真是一趟超长的旅程，从……比如说，向你提问的工具，一直到 AI 安全。Swyx [01:30:35]: Okay. that was a huge tour, from, like, ask you some question tool to AI safety.

是啊。敬《Pacing the Frontier》！是啊。Thariq Shihipar [01:30:41]: Yeah. To Pacing the Frontier. Yeah.

Swyx [01:30:43]：是啊。不，不过，是的，很明显，你确实非常积极地利用 Anthropic 为你提供的一切。至少能稍微了解一下大家在讨论什么、话题都有哪些，还是挺好的。最后有什么想对大家说的吗？有什么想呼吁大家去做的吗？Swyx [01:30:43]: Yeah. No, but, yeah, it’s clearly, it’s clear that you, like, really embrace everything that’s available to you in Anthropic, and, like, it’s, it’s good to at least have a peek inside of, like, what the discussions are, the topics are. any last words to people? Any, whatever you want to Call to action?

Thariq Shihipar [01:31:01]：是的，我觉得……首先，感谢你们邀请我。我觉得这真的……Thariq Shihipar [01:31:01]: Yeah, I think it’s. one, thank you for having me. I think this is like, I really

Swyx [01:31:06]：不，感谢邀请我。Swyx [01:31:06]: No, thanks for having me.

Thariq Shihipar [01:31:07]: Yeah. I

Swyx [01:31:08]：我们第一次见面是在一家中餐馆。Swyx [01:31:08]: We first met in a Chinese restaurant.

Thariq Shihipar [01:31:09]：没错。是啊。我觉得吧，我真的很喜欢你打造的这个社区，也很喜欢开发者社区。我觉得，虽然我知道现在变化非常快，要跟上的东西也很多，而且还有很多事情要做，我也有这种感觉。我想很多人都会觉得有点累、有点焦虑，或者有类似的感受。Thariq Shihipar [01:31:09]: That’s right. Yeah. I think, like, I really enjoy the like, community you’ve created and the community of developers. And, I think that, like, I know things are changing really fast, and I think there’s, like, a lot to keep on top of, and, like, I think there is just a lot to do, and I feel. I think a lot of people feel, like, a little bit tired or anxious or something.

压力很大。Swyx [01:31:33]: Stressed.

Thariq Shihipar [01:31:33]：压力很大，是的，没错。这完全可以理解，对吧？我觉得我们……我理解，嗯。而且我们也并不完美。我们会批评，也会理解所有 AI 实验室可以做得更好的地方。不过，我也对大家对 AI 的热情感到非常兴奋，这真的是一个令人兴奋的时代。我觉得我们以后回顾这段时间时会说：“哇，这段时间真是忙乱，但也非常令人兴奋，而且软件工程永远改变了。”其他事情也会发生变化。能成为其中的一部分，能和你们的听众交流，还能和那些不断拓展可能性边界的开发者互动，我觉得真的很荣幸。我也从中学到了很多。对。Thariq Shihipar [01:31:33]: Stressed, yeah, exactly. And this is, like, extremely understandable? And I think we. I understand, like. And we’re not perfect as well. Like, we, it’s, like, criticize and, like, understand, like, ways all of the AI labs could be better. and, but I also, like, am very excited about the excitement that everyone has for AI, and just, like, it’s a really exciting time. I think we’ll, like, look back at this time and be like, oh, like, this is, like, very hectic but very exciting, and, like, software engineering changed, like, forever. Like, other things will change. and it’s, like, really privileged to, like, be part of it, like, to talk to, like, the audience that you have and, to get to interact with all the developers who are, like, pushing the frontiers a lot on what’s possible. And I learn a lot from that too. Yeah.

开放安全模型、GlassWing 与有利于防御方的安全Open Safety Models, GlassWing, and Defense-Favored Security

Swyx [01:32:20]: Thanks so much.

塔里克·希希帕尔 [01:32:22]：谢谢。Thariq Shihipar [01:32:22]: Thank you.