章节 1：Claude Code 简介与安装

📝 本节摘要：

本章节作为开场，Boris 首先介绍了 Claude Code 的获取方式（通过 npm 安装）。随后，他定义了 Claude Code 的核心属性：它并非传统的代码补全工具，而是一款能够构建功能、编写文件甚至修复错误的代理式（Agentic）AI 助手。Boris 强调了其跨平台兼容性（支持所有 IDE 和终端），并给出了初次使用时的四项关键配置建议：终端设置、主题调整、安装 GitHub 应用以及自定义工具权限。

[原文] [Speaker]: Hello. Hey everyone. Uh, I'm Boris. I'm a member of technical staff here at Anthropic and I created quad code and here to talk to you a little bit about some practical tips and tricks for using quad code.

[译文] [Speaker]: 大家好。嘿，各位。呃，我是 Boris。我是 Anthropic 的一名技术人员，我创建了 Claude Code，今天来这里跟大家分享一些使用 Claude Code 的实用技巧和窍门。

[原文] [Speaker]: It's going to be very practical. I'm not going to go too much into the history or the theory or anything like this.

[译文] [Speaker]: 内容会非常实用。我不会过多涉猎历史、理论或诸如此类的东西。

[原文] [Speaker]: Before we start, can we get a quick show of hands? Who has used quad code before?

[译文] [Speaker]: 在我们开始之前，能快速举手表决一下吗？谁之前用过 Claude Code？

[原文] [Speaker]: For everyone that didn't raise your hands, I know you're not supposed to do this while people are talking, but if you can open your laptop and type npm install -g claude-code and this will help you install quad code.

[译文] [Speaker]: 对于那些没举手的人，我知道你们不应该在别人说话时做这个，但如果你们能打开笔记本电脑并输入 npm install -g claude-code，这将帮你安装 Claude Code。

[原文] [Speaker]: All you need is NodeJS. If you have it, this should work. If you don't have it yet, this is your chance to install it so you can follow along.

[译文] [Speaker]: 你只需要有 NodeJS。如果你有它，这应该就能行。如果你还没有，现在是安装它的机会，这样你就能跟上进度了。

[原文] [Speaker]: So, what is quad code? Quad code is a new kind of AI assistant.

[译文] [Speaker]: 那么，什么是 Claude Code？Claude Code 是一种新型的 AI 助手。

[原文] [Speaker]: There's been different generations of AI assistants for coding. Most of them have been about completing a line at a time, completing a few lines of code at a time. Claude Code is not for that.

[译文] [Speaker]: 编程 AI 助手已经经历了几代发展。大多数都是关于一次补全一行，或者一次补全几行代码。Claude Code 不是为了做这个的。

[原文] [Speaker]: It's fully agentic, so it's meant for building features, for writing entire functions, entire files, fixing entire bugs at the same time.

[译文] [Speaker]: 它是完全代理式的（agentic），所以它的目的是构建功能、编写完整的函数、完整的文件，以及同时修复整个 Bug。

[原文] [Speaker]: What's kind of cool about Claude Code is it works with all of your tools and you don't have to change out your workflow, you don't have to swap everything to start using it.

[译文] [Speaker]: Claude Code 很酷的一点是，它能与你所有的工具配合使用，你不需要改变你的工作流，也不需要为了开始使用它而替换掉所有东西。

[原文] [Speaker]: So, whatever IDE you use, if you use VS Code, or if you use Xcode, or if you use Jet Brains IDEs, there's some people at Anthropic that you can't pry Vim from their cold dead hands, but they use quad code because quad code works with every single ID, every terminal out there.

[译文] [Speaker]: 所以，无论你使用什么集成开发环境（IDE），如果你用 VS Code，或者你用 Xcode，或者你用 JetBrains IDEs——在 Anthropic 有些人你就算从他们冰冷的尸体上也没法把 Vim 拿走，但他们也用 Claude Code，因为 Claude Code 适用于每一个 IDE，每一个终端。

[原文] [Speaker]: It'll work locally over remote SSH, over T-Max, whatever environment you're in, you can run it.

[译文] [Speaker]: 它可以在本地运行，可以通过远程 SSH 运行，可以通过 Tmux（原文误作 T-Max）运行，无论你在什么环境中，你都可以运行它。

[原文] [Speaker]: It's general purpose. If you haven't used these kind of free-form coding assistants in the past, it can be hard to figure out how to get started because you open it up and you just see a prompt bar and you might wonder what do I do with this? What do I type in?

[译文] [Speaker]: 它是通用的。如果你过去没有使用过这种自由形式的编程助手，可能很难搞清楚如何上手，因为你打开它，只看到一个提示栏，你可能会想我该拿这个做什么？我该输入什么？

[原文] [Speaker]: It's a power tool so you can use it for a lot of things. Because it can do so much, we don't try to guide you towards a particular workflow because you should be able to use it however you want as an engineer.

[译文] [Speaker]: 它是一个强大的工具（power tool），所以你可以用它做很多事情。因为它能做的太多了，我们不会试图引导你使用特定的工作流，因为作为一名工程师，你应该能够按你想要的方式使用它。

[原文] [Speaker]: As you open up Claude Code for the first time, there's a few things that we recommend doing to get your environment set up. These are pretty straightforward:

[译文] [Speaker]: 当你第一次打开 Claude Code 时，为了设置好你的环境，我们要推荐做几件事。这些都非常简单直接：

[原文] [Speaker]: Run terminal setup: This will give you shift enter for new lines, making it a little nicer to use.

[译文] [Speaker]: 运行 terminal setup：这会让你能用 Shift+Enter 换行，让使用体验稍微好一点。

[原文] [Speaker]: Run theme: To set light mode or dark mode or doltonize themes.

[译文] [Speaker]: 运行 theme：设置浅色模式、深色模式或色盲友好模式（daltonized themes）。

[原文] [Speaker]: Run slash install github app: Today we announced a GitHub app where you can at-mention Claude on any GitHub issue or pull request.

[译文] [Speaker]: 运行 /install github app：今天我们发布了一个 GitHub 应用，你可以在任何 GitHub Issue 或 Pull Request 中 @Claude。

[原文] [Speaker]: Customize allowed tools: You can customize the set of allowed tools that you can use so you're not prompted for it every time.

[译文] [Speaker]: 自定义允许的工具（Allowed tools）：你可以自定义一套允许使用的工具集，这样你就不会每次都被提示授权了。

章节 2：环境配置与代码库问答 (Setup & Codebase Q&A)

📝 本节摘要：

本章节首先分享了一个 Mac 用户的效率技巧——使用系统自带的语音听写功能来输入 Prompt，以像与真人对话一样自然地交互。随后，Boris 指出上手 Claude Code 的最佳切入点并非写代码，而是代码库问答（Codebase Q&A）。他以 Anthropic 内部流程为例，展示了该功能如何将新员工的技术入职时间从 2-3 周大幅缩短至 2-3 天。最后，章节重点介绍了技术架构：Claude Code 不建立索引，不上传代码，不训练模型，所有操作均在本地即时进行，实现了真正的开箱即用与隐私安全。

[原文] [Speaker]: Use dictation for prompts (Mac OS): For a lot of prompts, you won't need to hand-type them. On Mac OS, you can go into your system settings under accessibility, enable dictation, hit the dictation key twice, and just speak your prompt.

[译文] [Speaker]: 使用听写功能输入提示词（Mac OS）：对于很多提示词，你不需要手动输入。在 Mac OS 上，你可以进入系统设置的“辅助功能”里，启用听写功能，按两次听写键，然后直接说出你的提示词。

[原文] [Speaker]: This helps a lot to have specific prompts and allows you to talk to Claude Code like you would another engineer.

[译文] [Speaker]: 这对于输入具体的提示词非常有帮助，并且让你能像与其他工程师交谈一样与 Claude Code 对话。

[原文] [Speaker]: When you're starting out with Claude Code, it's so free form and it can do everything. What do you start with?

[译文] [Speaker]: 当你刚开始使用 Claude Code 时，它的形式非常自由，无所不能。你应该从哪里开始呢？

[原文] [Speaker]: The thing I recommend above everything else is starting with codebase Q&A – just asking questions to your codebase.

[译文] [Speaker]: 我最为推荐的首要事项就是从“代码库问答”开始——即直接向你的代码库提问。

[原文] [Speaker]: This is something that we teach new hires at Anthropic. On the first day in technical onboarding, you learn about Claude Code, download it, get it set up, and then immediately start asking questions about the codebase.

[译文] [Speaker]: 这是我们在 Anthropic 教给新员工的内容。在技术入职的第一天，你会了解 Claude Code，下载并设置好它，然后立即开始针对代码库提问。

[原文] [Speaker]: In the past, technical onboarding taxed the team a lot; you had to ask other engineers questions, look around the code, and figure out how to use tools, which took a long time.

[译文] [Speaker]: 在过去，技术入职会让团队负担很重；你不得不问其他工程师问题，四处查看代码，弄清楚如何使用工具，这会花费很长时间。

[原文] [Speaker]: With Claude Code, you can just ask it, and it'll explore the codebase and answer these questions.

[译文] [Speaker]: 有了 Claude Code，你只需要问它，它就会自行探索代码库并回答这些问题。

[原文] [Speaker]: At Anthropic, onboarding used to take about two or three weeks for technical hires; it's now about two or three days.

[译文] [Speaker]: 在 Anthropic，技术人员的入职过去大约需要两到三周；现在只需要两到三天。

[原文] [Speaker]: What's also cool about Q&A is we don't do any sort of indexing. There's no remote database with your code, we don't upload it anywhere, your code stays local.

[译文] [Speaker]: 关于问答功能还有一个很酷的点是，我们要不做任何形式的索引。没有存储你代码的远程数据库，我们不把它上传到任何地方，你的代码始终保留在本地。

[原文] [Speaker]: We do not train generative models on the code.

[译文] [Speaker]: 我们不会利用这些代码来训练生成式模型。

[原文] [Speaker]: There are no indices, and what that means is there's no setup. You download it, start it, and can just use it right away.

[译文] [Speaker]: 没有索引，这就意味着不需要设置。你下载它，启动它，然后就能立即使用。

章节 3：深入挖掘信息的技巧 (Advanced Q&A Techniques)

📝 本节摘要：

本章节提供了极具实操性的 Prompt 技巧，旨在挖掘代码库的深层信息。Boris 演示了如何超越简单的文本搜索（Command+F），让 Claude 分析代码的实例化与具体用法；如何利用 Git 历史来追溯奇怪代码（例如“为什么有15个参数”）背后的引入者与决策背景；以及如何结合 GitHub Issues 和提交日志来自动生成“本周工作汇报”。他再次强调，在尝试复杂的代码编辑之前，先掌握这些问答技巧是团队上手的最佳路径。

[原文] [Speaker]: This is a technical talk, so I'm going to show some very specific prompts and code samples that you can use to hopefully improve your Claude Code experience.

[译文] [Speaker]: 这是一场技术讲座，所以我将展示一些非常具体的提示词（prompts）和代码示例，希望这能提升你使用 Claude Code 的体验。

[原文] [Speaker]: Some questions you can ask include: "How is this particular piece of code used?" or "How do I instantiate this thing?"

[译文] [Speaker]: 你可以问的一些问题包括：“这段特定的代码是如何被使用的？”或者“我该如何实例化这个东西？”

[原文] [Speaker]: Claude Code won't just do a text search; it'll often go a level deeper and try to find examples of how a class is instantiated or used, giving you a much deeper answer like you'd get from a wiki or documentation, not just a command F.

[译文] [Speaker]: Claude Code 不会只做文本搜索；它通常会深入一层，尝试找到类（class）是如何被实例化或使用的例子，给你提供一个像从 Wiki 或文档中才能得到的更深度的答案，而不仅仅是 Command+F（查找）。

[原文] [Speaker]: Ask about git history: For example, "Why does this function have 15 arguments and why are the arguments named this weird way?"

[译文] [Speaker]: 询问关于 Git 历史的问题：例如，“为什么这个函数有 15 个参数，而且参数命名得这么奇怪？”

[原文] [Speaker]: Claude Code can look through git history to figure out how arguments were introduced, by whom, the situation, and linked issues, summarizing it all without you having to ask for all these details.

[译文] [Speaker]: Claude Code 可以浏览 Git 历史，弄清楚参数是如何被引入的、被谁引入的、当时的情况以及关联的 Issue，并将这一切总结出来，而不需要你逐一询问这些细节。

[原文] [Speaker]: Just say, "look through git history" and it'll know to do this.

[译文] [Speaker]: 只需要说“查阅 Git 历史”，它就知道该怎么做。

[原文] [Speaker]: The reason it knows this is not because we prompted it to; the model is awesome, and if you tell it to use git, it'll know how to use git.

[译文] [Speaker]: 它之所以知道这一点，并不是因为我们给它预设了提示词；是因为模型本身很棒，如果你告诉它使用 Git，它就知道如何使用 Git。

[原文] [Speaker]: Ask about GitHub issues: It can use web fetch to fetch issues and look up context on issues.

[译文] [Speaker]: 询问关于 GitHub Issues 的问题：它可以使用 web fetch（网络获取）来抓取 Issue 并查找 Issue 的上下文。

[原文] [Speaker]: "What did I ship this week?": This is something the speaker does every Monday in their weekly standup.

[译文] [Speaker]: “我这周发布了什么？”：这是演讲者（我）每周一在周会（weekly standup）上都会做的事情。

[原文] [Speaker]: Claude Code looks at the log, knows their username, and provides a nice readout of everything shipped, which can be copied and pasted into a document.

[译文] [Speaker]: Claude Code 会查看日志，识别用户名，并提供一份很好的已发布内容摘要，你可以直接把它复制粘贴到文档里。

[原文] [Speaker]: So, that's tip number one for people who haven't used Claude Code before, or if you're showing it to someone for the first time, or onboarding your team: start with codebase Q&A.

[译文] [Speaker]: 所以，这就是给那些没用过 Claude Code 的人，或者是你第一次向别人展示它，又或者是给团队做入职培训时的第一条建议：从代码库问答（Codebase Q&A）开始。

[原文] [Speaker]: Don't start by using fancy tools or editing code. Just start by asking questions about the codebase.

[译文] [Speaker]: 不要一上来就使用花哨的工具或编辑代码。先从询问关于代码库的问题开始。

[原文] [Speaker]: This will teach people how to prompt and help them understand what Claude Code can do versus what needs more guidance.

[译文] [Speaker]: 这将教会大家如何编写提示词，并帮助他们理解 Claude Code 能做什么，以及哪些地方需要更多的指导。

章节 4：从问答到代码编辑与提交 (Editing, Planning & Committing)

📝 本节摘要：

当用户熟悉问答功能后，Boris 引导大家进入代码编辑阶段。他解释了 Claude Code 的代理（Agentic）工作原理：仅通过三个基础工具（文件编辑、Bash 命令、文件搜索），模型就能自主串联步骤来完成任务。为了提高成功率，他强烈建议在编码前先要求 Claude “制定计划”或“头脑风暴”，而非直接生成代码。最后，他演示了 "commit push" 指令——Claude 能自动分析 Git 上下文，一次性完成提交、创建分支、推送代码并生成 GitHub Pull Request。

[原文] [Speaker]: Once you're pretty comfortable with Q&A, you can dive into editing code.

[译文] [Speaker]: 一旦你对问答功能相当熟悉了，你就可以深入到编辑代码的环节了。

[原文] [Speaker]: The cool thing about using an LLM in an agentic way is you give it tools, and it magically figures out how to use them.

[译文] [Speaker]: 以代理方式（agentic way）使用大语言模型（LLM）的一件很酷的事情是，你给它工具，它就会神奇地自己弄清楚如何使用它们。

[原文] [Speaker]: With Claude Code, we give it a pretty small set of tools: A tool to edit files. A tool to run bash commands. A tool to search files.

[译文] [Speaker]: 对于 Claude Code，我们给它提供了一套相当小的工具集：一个编辑文件的工具。一个运行 Bash 命令的工具。一个搜索文件的工具。

[原文] [Speaker]: It will string these together to explore code, brainstorm, and finally make edits. You don't have to prompt it specifically to use each tool; you just say, "do this thing," and it'll figure out how to string them together in the right way.

[译文] [Speaker]: 它会将这些工具串联起来去探索代码、头脑风暴，并最终进行编辑。你不需要专门提示它去使用每一个工具；你只需要说，“做这件事”，它就会自己弄清楚如何以正确的方式将它们串联起来。

[原文] [Speaker]: There are a lot of ways to use this. Something the speaker likes to do is, before having Claude jump in to write code, they'll ask it to brainstorm a little bit or make a plan.

[译文] [Speaker]: 有很多种使用它的方式。演讲者（我）喜欢做的一件事是，在让 Claude 直接跳进去写代码之前，会要求它先稍微头脑风暴一下，或者制定一个计划。

[原文] [Speaker]: This is highly recommended.

[译文] [Speaker]: 这一点强烈推荐。

[原文] [Speaker]: Sometimes people ask Claude Code to implement an enormous 30,000-line feature, and while it sometimes gets it right on the first shot, often the result is not what was wanted.

[译文] [Speaker]: 有时候人们要求 Claude Code 去实现一个巨大的 30,000 行的功能，虽然它有时能一次性搞定，但结果往往不是想要的。

[原文] [Speaker]: The easiest way to get the desired result is to ask it to think first.

[译文] [Speaker]: 获得理想结果的最简单方法是先让它思考。

[原文] [Speaker]: So, brainstorm ideas, make a plan, run it by me, ask for approval before you write code. You don't have to use plan mode or any special tools; just ask Claude, and it'll know to do this.

[译文] [Speaker]: 所以，（你可以说）头脑风暴一下想法，制定一个计划，跟我过一遍，在写代码之前先征求我的批准。你不需要使用计划模式（plan mode）或任何特殊工具；只要告诉 Claude，它就知道该这么做。

[原文] [Speaker]: The "commit push" incantation is also a common one used by the speaker.

[译文] [Speaker]: "commit push" （提交并推送）这个咒语也是演讲者（我）常用的一个。

[原文] [Speaker]: There's nothing special about it, but Claude is smart enough to interpret this. It will: Make a commit. Push it to the branch. Make a branch. Make a pull request on GitHub.

[译文] [Speaker]: 这没有什么特别的，但 Claude 足够聪明来解读它。它会：创建一个提交（commit）。把它推送到分支。创建一个分支。在 GitHub 上创建一个拉取请求（Pull Request）。

[原文] [Speaker]: You don't have to explain anything; it'll look through the code, history, and git log by itself to figure out the commit format and make the commit and push it the right way.

[译文] [Speaker]: 你不需要解释任何事情；它会自己查看代码、历史记录和 Git 日志，弄清楚提交格式，并以正确的方式进行提交和推送。

[原文] [Speaker]: Again, it's not system-prompted; the model just knows how to do this.

[译文] [Speaker]: 再次强调，这不是系统提示词设定的；模型本身就知道如何做这件事。

章节 5：集成团队工具与反馈迭代 (Custom Tools & Iterative Workflows)

📝 本节摘要：

随着使用深入，Boris 建议开发者将团队的专有工具集成到 Claude Code 中。他介绍了两类主要工具：批处理工具（Batch tools）（如自定义 CLI）和 MCP 工具（Model Context Protocol）。后者尤为强大，能让 Claude 直接调用团队现有的工具链。最为关键的是，Boris 展示了 “带反馈的迭代（Iteration with Feedback）” 工作流：通过给 Claude 提供验证手段（如单元测试、Puppeteer 截图或 iOS 模拟器），让 AI 在提交前自主检查并修正错误。这种闭环工作流能让生成结果的准确度从“还不错”提升至“几乎完美”。

[原文] [Speaker]: As you get a little more advanced, you're going to want to start to plug in your team's tools.

[译文] [Speaker]: 当你变得更进阶一点时，你会想要开始接入你团队的工具。

[原文] [Speaker]: This is where Claude Code really starts to shine. There are generally two kinds of tools:

[译文] [Speaker]: 这正是 Claude Code 真正开始大放异彩的地方。通常有两类工具：

[原文] [Speaker]: Batch tools: For example, a custom CLI tool like "barley CLI." You can say "use the CLI to do something," and tell Claude Code about it, even telling it to use --help to figure out how to use it.

[译文] [Speaker]: 批处理工具（Batch tools）：例如，像 "barley CLI" 这样的自定义命令行工具。你可以说“用这个 CLI 做点什么”，并告诉 Claude Code 关于它的信息，甚至告诉它使用 --help 来弄清楚如何使用它。

[原文] [Speaker]: If you use it a lot, you can dump this into your quadmd so Claude can remember it across sessions.

[译文] [Speaker]: 如果你经常使用它，你可以把它丢进你的 CLAUDE.md 文件里，这样 Claude 就能在不同会话间记住它。

[原文] [Speaker]: This is a common pattern at Anthropic and with external customers.

[译文] [Speaker]: 这是 Anthropic 内部以及外部客户中常见的一种模式。

[原文] [Speaker]: MCP tools: Claude Code can use MCP tools. Just tell it about the tools, and you can add the MCP tool and tell it how to use it, and it'll just start using it.

[译文] [Speaker]: MCP 工具：Claude Code 可以使用 MCP 工具（Model Context Protocol）。只需要告诉它有哪些工具，你可以添加 MCP 工具并告诉它如何使用，它就会开始使用它。

[原文] [Speaker]: This is extremely powerful because when you start to use code on a new codebase, you can just give it all of your tools, all the tools your team already uses for this codebase, and Claude Code can use it on your behalf.

[译文] [Speaker]: 这极其强大，因为当你开始在一个新的代码库上写代码时，你可以直接把所有的工具给它，把你团队在这个代码库上已经在用的所有工具都给它，Claude Code 就能代表你使用这些工具。

[原文] [Speaker]: There are a few common workflows: Exploration and Planning: Do a little bit of exploration, a little bit of planning, and ask for confirmation before writing code.

[译文] [Speaker]: 有几种常见的工作流：探索与规划（Exploration and Planning）：做一点探索，做一点规划，并在写代码之前寻求确认。

[原文] [Speaker]: Iteration with Feedback: These other two on the right are extremely powerful when Claude has some way to check its work.

[译文] [Speaker]: 带反馈的迭代（Iteration with Feedback）：右边的这两个工作流非常强大，前提是 Claude 有某种方式可以检查它自己的工作。

[原文] [Speaker]: For example, by writing unit tests, screenshotting with Puppeteer, or screenshotting the iOS simulator, it can iterate.

[译文] [Speaker]: 例如，通过编写单元测试、使用 Puppeteer 截图，或者对 iOS 模拟器进行截图，它就可以进行迭代。

[原文] [Speaker]: If you give it a mock and say "build this web UI," it'll get it pretty good, but if you let it iterate two or three times, it often gets it almost perfect.

[译文] [Speaker]: 如果你给它一个设计稿（mock）并说“构建这个 Web 界面”，它会做得相当不错，但如果你让它迭代两三次，它通常能做得几乎完美。

[原文] [Speaker]: The trick is to give it some sort of tool that it can use for feedback to check its work, and then based on that, it will iterate by itself, and you're going to get a much better result.

[译文] [Speaker]: 诀窍是给它某种它能用来反馈以检查其工作的工具，然后基于此，它会自动迭代，你就会得到一个好得多的结果。

[原文] [Speaker]: Whatever your domain is (unit tests, integration tests, screenshots for apps/web), just give it a way to see its result, and it'll iterate and get better.

[译文] [Speaker]: 无论你的领域是什么（单元测试、集成测试、App 或 Web 的截图），只要给它一种能看到其结果的方式，它就会迭代并变得更好。

[原文] [Speaker]: So these are the next tips: teach Claude how to use your tools and figure out the right workflow.

[译文] [Speaker]: 所以这些就是接下来的建议：教 Claude 如何使用你的工具，并找出正确的工作流。

[原文] [Speaker]: If you want Claude to jump in and code, if you want it to brainstorm a little bit, make a plan, if you want it to iterate, kind of have some sense of that so you know how to prompt Claude to do what you want.

[译文] [Speaker]: 如果你想让 Claude 直接介入写代码，如果你想让它稍微头脑风暴一下、制定计划，如果你想让它迭代，你要对此有某种感知，这样你就知道如何提示 Claude 去做你想要的事。

章节 6：构建上下文语境 (CLAUDE.md)

📝 本节摘要：

为了让 Claude 做出更明智的决策，必须提供充足的上下文。Boris 介绍了最简单的方法：在项目根目录创建一个 `CLAUDE.md` 文件。这个文件会在每次会话开始时自动被读取，内容应包含常用命令、架构决策和重要文件路径。他还区分了共享的 CLAUDE.md（提交到 Git）和个人的 CLAUDE.md.local（仅本地使用）。此外，他还提到了嵌套目录的上下文加载机制，以及通过 /commands 定义自定义斜杠命令的高级用法。

[原文] [Speaker]: As you go deeper beyond tools, you want to start to give Claude more context.

[译文] [Speaker]: 当你超越工具层面深入使用时，你会想要开始给 Claude 提供更多的上下文。

[原文] [Speaker]: The more context, the smarter the decisions will be because as an engineer working in a codebase, you have a ton of context in your head about your systems and all the history.

[译文] [Speaker]: 上下文越多，决策就会越明智，因为作为一名在代码库中工作的工程师，你的脑海里有关于系统和所有历史的大量上下文。

[原文] [Speaker]: There are different ways to give this to Claude, and as you give Claude more context, it'll do better.

[译文] [Speaker]: 有不同的方式可以把这些提供给 Claude，而且你给 Claude 的上下文越多，它的表现就越好。

[原文] [Speaker]: The simplest way is what we call quad MD (quad.mmd).

[译文] [Speaker]: 最简单的方法就是我们所说的 CLAUDE.md（原文误作 quad MD/quad.mmd）。

[原文] [Speaker]: The simplest place to put it is in the project root, the same directory you start Claude in.

[译文] [Speaker]: 放置它的最简单位置是在项目根目录，也就是你启动 Claude 的那个目录。

[原文] [Speaker]: A quad.mmd file there will get automatically read into context at the start of every session, essentially included in the first user turn.

[译文] [Speaker]: 那里的 CLAUDE.md 文件会在每次会话开始时被自动读入上下文，本质上是包含在第一轮用户对话中。

[原文] [Speaker]: You can also have a local quadmomd.

[译文] [Speaker]: 你也可以有一个本地的 CLAUDE.md.local（原文误作 quadmomd）。

[原文] [Speaker]: This one you usually don't check into source control. quad.mmd should be checked into source control to share with your team, so you write it once and share it. The local one is just for you.

[译文] [Speaker]: 这个通常不提交到源代码版本控制中。CLAUDE.md 应该被提交到源代码控制中以便与团队共享，这样你写一次就能分享给所有人。本地那个则只供你自己使用。

[原文] [Speaker]: The kinds of things you put in quad.mmd are: Common bash commands. Common MCP tools. Architectural decisions. Important files. Anything that you would typically need to know to work in this codebase.

[译文] [Speaker]: 你放在 CLAUDE.md 里的东西包括：常用的 Bash 命令。常用的 MCP 工具。架构决策。重要文件。任何在处理这个代码库时通常需要知道的事情。

[原文] [Speaker]: Try to keep it pretty short because if it gets too long, it's just going to use up a bunch of context and it's usually not that useful. So just try to keep it as short as you can.

[译文] [Speaker]: 尽量保持它简短，因为如果太长，它只会消耗大量的上下文窗口，而且通常没那么有用。所以尽量让它越短越好。

[原文] [Speaker]: For example, in their codebase, they have common batch commands, a style guide, and a few core files.

[译文] [Speaker]: 例如，在他们的代码库中，他们有常用的批处理命令、风格指南（style guide）和几个核心文件。

[原文] [Speaker]: All the other quad.mmd files can be put in other nested child directories, and Claude will pull them in on demand.

[译文] [Speaker]: 所有其他的 CLAUDE.md 文件可以放在其他嵌套的子目录中，Claude 会按需拉取它们。

[原文] [Speaker]: These are the quad.mmd files that will get pulled in automatically. You can also put quad.mmds in nested directories, and those will get automatically pulled when Claude works in those directories.

[译文] [Speaker]: 这些是会被自动拉取的 CLAUDE.md 文件。你也可以把 CLAUDE.md 放在嵌套目录中，当 Claude 在这些目录中工作时，它们会被自动拉取。

[原文] [Speaker]: And, of course, if you're a company, you might want a quadd that's shared across all the different code bases and you want to manage it on behalf of your users, and you can put it in your enterprise root, and that'll get pulled in automatically.

[译文] [Speaker]: 当然，如果你是一家公司，你可能想要一个跨所有不同代码库共享的 CLAUDE.md，并且你想代表你的用户来管理它，你可以把它放在企业根目录下，那样它也会被自动拉取。

[原文] [Speaker]: There's a ton of ways to pull in context. quad.mmd is pulled in automatically.

[译文] [Speaker]: 有很多种拉取上下文的方式。CLAUDE.md 是自动拉取的。

[原文] [Speaker]: You can also use slash commands. So this is quad/comands and this can be in your home directory or it can be checked into your project.

[译文] [Speaker]: 你也可以使用斜杠命令（slash commands）。这可以是 commands 文件，它可以放在你的主目录中，也可以提交到你的项目中。

[原文] [Speaker]: There are a few examples of slash commands that they have in Claude Code itself.

[译文] [Speaker]: 在 Claude Code 本身中有几个斜杠命令的例子。

[原文] [Speaker]: For example, if you're in the Claude Code repo and you see issues getting labeled, that's actually a workflow running there: label GitHub issues.

[译文] [Speaker]: 例如，如果你在 Claude Code 的代码仓库中看到 Issue 被打标签，那实际上是在那里运行的一个工作流：给 GitHub Issues 打标签。

[原文] [Speaker]: They have a GitHub action running where Claude Code will run this command, and it's just a slash command. It'll run and it'll label the issues so humans don't have to. It just saves them a bunch of time.

[译文] [Speaker]: 他们运行着一个 GitHub Action，其中 Claude Code 会运行这个命令，而这只是一个斜杠命令。它运行起来并给 Issue 打标签，这样人类就不必做了。这给他们节省了大量时间。

[原文] [Speaker]: And of course, you can @mention files to pull them into context.

[译文] [Speaker]: 当然，你也可以通过 @提及（@mention）文件把它们拉入上下文。

[原文] [Speaker]: So give Claude more context, and it's definitely worth taking the time to tune context.

[译文] [Speaker]: 所以给 Claude 更多的上下文，花时间去调整上下文绝对是值得的。

[原文] [Speaker]: You can run it through a prompt improver. Consider who the context is for. If you want to pull it in every time, if it's a personal preference, definitely take the time to tune it. This will improve performance dramatically.

[译文] [Speaker]: 你可以通过提示词优化器（prompt improver）来运行它。考虑一下上下文是给谁用的。如果你想每次都拉取它，如果这是个人偏好，一定要花时间去调整它。这将极大地提升性能。

章节 7：配置文件的层级与企业级策略 (Configuration Hierarchy & Enterprise Policies)

📝 本节摘要：

在掌握了基础上下文管理后，Boris 介绍了 Claude Code 的配置层级体系：从针对单个 Git 仓库的“项目级”，到跨项目的“全局级”，再到自动下发给所有员工的“企业级策略（Enterprise Policies）”。他特别强调了企业级策略的双重价值：一方面可以通过自动批准（Auto-approve）常用命令（如测试指令）来以此减少重复确认，提升团队效率；另一方面可以通过阻止（Block）特定命令或 URL 来保障代码库安全。最后，他建议将 MCP 服务器配置（mcp.json）提交到代码库中，实现“一人配置，全员受益”的网络效应。

[原文] [Speaker]: As you get more advanced, you're going to want to think about the hierarchy of different ways to pull in everything: not just quad.mmd, but also config and everything about Claude can be pulled in this hierarchical way.

[译文] [Speaker]: 当你变得更进阶时，你会想要考虑引入所有内容的层级方式：不仅仅是 CLAUDE.md，还有配置以及关于 Claude 的一切都可以通过这种层级方式引入。

[原文] [Speaker]: Projects: Specific to your git repo, and this can be checked in or just for you.

[译文] [Speaker]: 项目级（Projects）：特定于你的 Git 仓库，这可以提交到版本控制中，也可以仅供你个人使用。

[原文] [Speaker]: Global configs: Across all your projects.

[译文] [Speaker]: 全局配置（Global configs）：跨越你所有的项目。

[原文] [Speaker]: Enterprise policies: A global config rolled out for all employees automatically.

[译文] [Speaker]: 企业级策略（Enterprise policies）：一种自动向所有员工推广的全局配置。

[原文] [Speaker]: This applies to a lot of stuff: slash commands, permissions. For example, if you have a batch command that you would run for all your employees, like a test command, you can actually just check it into this enterprise policies file, and then any employee when they run this command, it will be auto-approved, which is pretty convenient.

[译文] [Speaker]: 这适用于很多东西：斜杠命令、权限。例如，如果你有一个希望所有员工都运行的批处理命令，比如一个测试命令，你实际上可以直接把它签入这个企业策略文件中，然后任何员工在运行这个命令时，它都会被自动批准，这非常方便。

[原文] [Speaker]: You can also use this to block commands. For example, if there's a URL that should never be fetched, just add it to this config, and that'll make it so an employee cannot overwrite it, and that URL can never be fetched. This is convenient both to unblock people and to keep your codebase safe.

[译文] [Speaker]: 你也可以用这个来阻止命令。例如，如果有一个 URL 永远不应该被抓取，只需把它添加到这个配置中，这样员工就无法覆盖它，那个 URL 也永远无法被抓取。这既方便了为人们解除限制（指前面的自动批准），又能保持你的代码库安全。

[原文] [Speaker]: And then same thing for MCP servers. Have an MCP JSON file, check it into the codebase. That way anytime someone runs Claude Code in your codebase, they'll be prompted to install the MCP servers and share it with the team.

[译文] [Speaker]: 对于 MCP 服务器也是一样的。创建一个 mcp.json 文件，把它提交到代码库中。这样，任何人在你的代码库中运行 Claude Code 时，都会被提示安装这些 MCP 服务器，并与团队共享。

[原文] [Speaker]: If you're not sure which of these to use, the speaker recommends starting with shared project context.

[译文] [Speaker]: 如果你不确定该用哪一种，演讲者（我）建议从共享的项目上下文开始。

[原文] [Speaker]: You write this once and then you share it with everyone on the team, and you get this kind of network effect where someone does a little bit of work and everyone on the team benefits.

[译文] [Speaker]: 你写一次，然后分享给团队里的每个人，你就会得到这种网络效应：某个人做了一点工作，团队里的每个人都能受益。

章节 8：高效操作：快捷键与控制指令 (Pro Tips: Key Bindings & Controls)

📝 本节摘要：

本章节揭示了终端界面下鲜为人知的高级操作技巧。Boris 介绍了一系列快捷键与控制指令：使用 Shift+Tab 可以进入“自动接受模式”，跳过对代码编辑的逐一确认；使用 !（感叹号）可以直接运行本地 Bash 命令并将其输出喂给 Claude；使用 #（井号）可以快速向 CLAUDE.md 添加记忆。此外，他还讲解了如何通过 Escape 键安全地中断任务，以及使用 Ctrl+R 查看完整的上下文日志。

[原文] [Speaker]: This is a talk about pro tips, so here's a quick reference sheet on key bindings that people may not know.

[译文] [Speaker]: 这是一个关于专业技巧的讲座，所以这里有一份关于大家可能不知道的快捷键的快速参考表。

[原文] [Speaker]: It's very hard to build for terminal, but it's also very fun. It feels like rediscovering a new design language. But something about terminal is it's extremely minimal, so sometimes it's hard to discover these key bindings.

[译文] [Speaker]: 为终端构建应用非常难，但也非常有趣。这感觉就像重新发现了一种新的设计语言。但终端的一个特点是它极其极简，所以有时很难发现这些快捷键。

[原文] [Speaker]: Anytime you can hit shift + tab to accept edits. This switches you into auto-accept edits mode.

[译文] [Speaker]: 任何时候你都可以按 Shift + Tab 来接受编辑。这会让你切换进入“自动接受编辑模式”。

[原文] [Speaker]: Bash commands still need approval, but edits are auto-accepted. You can always ask Claude to undo them later.

[译文] [Speaker]: Bash 命令仍然需要批准，但（代码）编辑会被自动接受。你随时可以要求 Claude 稍后撤销它们。

[原文] [Speaker]: The speaker uses this if they know Claude is on the right track or if it's writing unit tests and iterating on tests, so they don't have to okay every single edit.

[译文] [Speaker]: 如果演讲者（我）知道 Claude 的思路是正确的，或者它正在编写单元测试并进行迭代，为了不必逐个确认每一次编辑，就会使用这个功能。

[原文] [Speaker]: If Claude is not using a tool correctly and you want it to use it correctly from then on, just type the pound sign (#) and then tell it what to remember, and it'll remember it, incorporating it into QuadMD automatically.

[译文] [Speaker]: 如果 Claude 没有正确使用某个工具，而你希望它从今以后能正确使用，只需输入井号键（#），然后告诉它要记住什么，它就会记住，并自动将其合并到 CLAUDE.md 中。

[原文] [Speaker]: If you ever want to drop down to bash mode, just run a bash command. You can hit the exclamation mark (!) and type in your command.

[译文] [Speaker]: 如果你想要切换到 Bash 模式，只需运行一个 Bash 命令。你可以按感叹号键（!）然后输入你的命令。

[原文] [Speaker]: That'll run locally, but that also goes into the context window, so Claude will see it on the next turn.

[译文] [Speaker]: 那将在本地运行，但结果也会进入上下文窗口，所以 Claude 会在下一轮对话中看到它。

[原文] [Speaker]: This is good for long-running commands if you know exactly what you want to do or any command that you want to get into context where Claude will see the command and the output.

[译文] [Speaker]: 如果你知道自己确切想做什么，或者你想把某个命令及其输出放入上下文让 Claude 看到，这对于长时间运行的命令非常有用。

[原文] [Speaker]: Anytime you can hit escape to stop what Claude is doing. No matter what Claude is doing, you can always safely hit escape; it's not going to corrupt the session or mess anything up.

[译文] [Speaker]: 任何时候你都可以按 Escape 键来停止 Claude 正在做的事情。无论 Claude 在做什么，你总是可以安全地按 Escape；它不会损坏会话或搞乱任何东西。

[原文] [Speaker]: For example, if Claude is doing a file edit, hit escape, tell it what to do differently.

[译文] [Speaker]: 例如，如果 Claude 正在编辑文件，按 Escape，告诉它换种方式做。

[原文] [Speaker]: After you're done with a session, you can start Claude with resume to resume that session if you want, or --continue.

[译文] [Speaker]: 当你结束一个会话后，如果你想恢复那个会话，你可以用 resume 或 --continue 来启动 Claude。

[原文] [Speaker]: Anytime if you want to see more output, hit control + R, and that'll show you the entire output, the same thing that Claude sees in its context window.

[译文] [Speaker]: 任何时候如果你想看到更多输出，按 Ctrl + R，那会显示完整的输出，也就是 Claude 在它的上下文窗口中看到的一模一样的内容。

章节 9：开发者SDK与高阶管道操作 (The SDK & Advanced Piping)

📝 本节摘要：

在掌握了交互式用法后，Boris 展示了 Claude Code 的另一面：SDK 与 CLI 管道模式（通常通过 -p 标志调用，转录为 claw-p）。他将 Claude 比作一个“智能 Unix 工具”，能够接收标准输入（如 git status 或服务器日志），并输出结构化的 JSON 数据。这种能力使得 Claude 可以被集成到 CI/CD、故障响应或 Sentry 报错分析中。最后，他描述了“Power User”的终极形态：通过 SSH 和 Tmux（转录为 T-Max）在多个 Git 工作树（worktrees）中并行运行多个 Claude 会话，实现多任务并发处理。

[原文] [Speaker]: The next thing the speaker wants to talk about is the Claude Code SDK.

[译文] [Speaker]: 演讲者（我）想谈的下一件事是 Claude Code SDK。

[原文] [Speaker]: Sid is doing a session just across the hallway that will go super deep on the SDK. If you haven't played around with this, if you used the -p flag in Claude, this is what the SDK is. They've been learning a bunch of features over the last few weeks to make it even better.

[译文] [Speaker]: Sid 正在走廊对面的分会场做一个关于 SDK 的深度分享。如果你还没玩过这个，如果你在 Claude 中使用过 -p 标志，这就是 SDK。过去几周他们一直在学习（开发）一系列功能，以使其变得更好。

[原文] [Speaker]: You can build on top of this. This is exactly the thing that Claude Code uses; it's exactly the same SDK.

[译文] [Speaker]: 你可以在此基础上进行构建。这正是 Claude Code 所使用的东西；完全是同一个 SDK。

[原文] [Speaker]: For example, something you can do is claw-p. This is the CLI SDK. You can pass a prompt, you can pass some allowed tools (which could include specific batch commands), and you can tell it which format you want (JSON or streaming JSON if you want to process this somehow).

[译文] [Speaker]: 例如，你可以做的一件事是 claw-p（注：指调用 CLI SDK 的命令）。这就是 CLI SDK。你可以传递一个提示词，你可以传递一些允许的工具（其中可以包括特定的批处理命令），你还可以告诉它你想要的格式（如果你想以某种方式处理它，可以选择 JSON 或流式 JSON）。

[原文] [Speaker]: This is awesome for building on. They use this in CI all the time, for incident response, and in all sorts of pipelines.

[译文] [Speaker]: 这非常适合用来构建上层应用。他们一直在持续集成（CI）中使用它，用于故障响应，以及各种管道（pipelines）中。

[原文] [Speaker]: Think of it as a Unix utility: you give it a prompt, it gives you JSON. You can use this in any way; you can pipe into it, you can pipe out of it.

[译文] [Speaker]: 把它想象成一个 Unix 实用工具：你给它一个提示词，它给你 JSON。你可以以任何方式使用它；你可以管道输入给它，也可以从它管道输出。

[原文] [Speaker]: The piping is also pretty cool. You can use, for example, git status and pipe this in and use jq to select the result. The combinations are endless, and it's sort of this new idea; it's like a super intelligent Unix utility, and they've barely scratched the surface of how to use this.

[译文] [Speaker]: 管道操作也非常酷。例如，你可以使用 git status 并将其通过管道传入，然后使用 jq 来筛选结果。组合是无穷无尽的，这算是一种新理念；它就像一个超级智能的 Unix 实用工具，而关于如何使用它，他们才刚刚触及皮毛。

[原文] [Speaker]: You can read from a GCP bucket, read a giant log, and pipe it in and tell Claude to figure out what's interesting about this log.

[译文] [Speaker]: 你可以从 GCP（谷歌云）存储桶读取数据，读取巨大的日志，然后将其通过管道传入，告诉 Claude 去找出这个日志里有什么有趣的地方。

[原文] [Speaker]: You can fetch data from the Sentry CLI and pipe it in and have Claude do something with it.

[译文] [Speaker]: 你可以从 Sentry CLI 获取数据并通过管道传入，让 Claude 用它做点什么。

[原文] [Speaker]: The final thing, and probably the most advanced use cases they see, is for power users at Anthropic. The speaker is sort of a "Claude normy" and usually has one Claude running at a time and maybe a few terminal tabs for a few different repos running at a time.

[译文] [Speaker]: 最后一件事，也是他们看到的可能是最高级的用例，是针对 Anthropic 内部的高级用户（Power users）。演讲者（我）算是一个“Claude 普通用户”，通常一次只运行一个 Claude，也许会为了几个不同的代码库同时开几个终端标签页。

[原文] [Speaker]: When they look at power users, almost always they're going to have SSH sessions, T-Max tunnels into their Claude sessions. They're going to have a bunch of checkouts of the same repo so that they can run a bunch of Claudes in parallel in that repo, or they're using git work trees to have some kind of isolation.

[译文] [Speaker]: 当他们观察高级用户时，发现这些人几乎总是拥有 SSH 会话，以及进入 Claude 会话的 T-Max（注：应为 Tmux）隧道。他们会对同一个代码库进行多次检出（checkouts），这样他们就可以在那个代码库中并行运行一堆 Claude，或者他们使用 Git 工作树（git worktrees）来进行某种隔离。

[原文] [Speaker]: They are actively working on making this easier to use. But for now, these are some ideas for how to do more work in parallel with Claude. You can run as many sessions as you want.

[译文] [Speaker]: 他们正在积极努力使这变得更易于使用。但就目前而言，这些是如何用 Claude 并行做更多工作的一些思路。你想运行多少个会话都可以。

[原文] [Speaker]: There's a lot that you can get done in parallel.

[译文] [Speaker]: 并行处理可以完成很多工作。

章节 10：现场问答环节 (Audience Q&A)

📝 本节摘要：

在最后的问答环节中，Boris 回答了四个关键问题。首先，他坦言开发 Claude Code 最难的部分是确保 Bash 命令的安全性，团队为此构建了一套复杂的分层权限系统。其次，他确认了 Claude Code 具备完全的多模态能力，支持拖拽图片进行 UI 开发。第三，关于“为何做 CLI 而不做 IDE 插件”，他给出了两个理由：一是通过终端兼容所有开发者的工具链（无论是 Vim 还是 Xcode），二是他预判随着模型进化，传统的 IDE 界面可能在年底前就会过时。最后，他透露 Anthropic 内部 80% 的技术人员（包括研究员）每天都在使用 Claude Code，甚至用于机器学习建模和 Notebook 编辑。

[原文] [Speaker]: That's it. The speaker wanted to leave some time for Q&A.

[译文] [Speaker]: 就是这些了。演讲者（我）想留一些时间来进行问答环节。

[原文] [Question]: What was the hardest part of building Claude Code?

[译文] [Question]: 构建 Claude Code 最难的部分是什么？

[原文] [Answer]: Boris mentions that a lot of parts were tricky, but especially making bash commands safe.

[译文] [Answer]: Boris 提到很多部分都很棘手，但特别是确保 Bash 命令的安全性。

[原文] [Answer]: Bash is inherently dangerous and can change system state unexpectedly. Manually approving every command is annoying and hinders productivity.

[译文] [Answer]: Bash 本质上是危险的，可能会意外地改变系统状态。手动批准每一个命令很烦人，而且会阻碍生产力。

[原文] [Answer]: Navigating how to do this safely and at scale across different codebases (since not everyone runs code in Docker) was difficult.

[译文] [Answer]: 摸索如何安全地、且在大规模的不同代码库中做到这一点（因为不是每个人都在 Docker 里运行代码）是很难的。

[原文] [Answer]: They landed on having some commands be read-only, using static analysis to figure out which commands can be combined safely, and a pretty complex tiered permission system for allow-listing and block-listing commands at different levels.

[译文] [Answer]: 他们最终的方案是让某些命令设为只读，使用静态分析来弄清楚哪些命令可以安全地组合，以及建立一套相当复杂的分层权限系统，用于在不同层级对命令进行白名单和黑名单管理。

[原文] [Question]: Is there multimodal functionality for images, or is Claude Code just pointing at an image on the file system?

[译文] [Question]: 有针对图像的多模态功能吗？还是说 Claude Code 只是指向文件系统上的一个图片？

[原文] [Answer]: Claude Code is fully multimodal and has been from the start.

[译文] [Answer]: Claude Code 是完全多模态的，从一开始就是。

[原文] [Answer]: Although it's in a terminal, which makes it a little hard to discover, you can drag and drop an image in, give it a file path, or copy and paste the image in.

[译文] [Answer]: 虽然它是在终端里，这让功能有点难以被发现，但你可以把图片拖放进去，或者给它一个文件路径，或者直接复制粘贴图片进去。

[原文] [Answer]: Boris often uses this for mocks: he'll drag and drop a mock, tell Claude to implement it, and give it a Puppeteer server so it can iterate against it, making it fully automated.

[译文] [Answer]: Boris 经常用这个来处理设计稿（mocks）：他会拖入一张设计稿，告诉 Claude 去实现它，并给它一个 Puppeteer 服务器让它能对照着进行迭代，从而实现完全自动化。

[原文] [Question]: Why did you build a CLI tool instead of an IDE?

[译文] [Question]: 你们为什么构建一个命令行（CLI）工具而不是一个 IDE？

[原文] [Answer]: Two main reasons: At Anthropic, people use a broad range of IDEs (VS Code, Zed, Xcode, Vim, Emacs), so it was hard to build something that works for everyone. The terminal is the common denominator.

[译文] [Answer]: 两个主要原因：在 Anthropic，大家使用各种各样的 IDE（VS Code, Zed, Xcode, Vim, Emacs），所以很难构建出一个对所有人都适用的东西。终端是共同的基础（common denominator）。

[原文] [Answer]: At Anthropic, they see how fast the model is getting better, and Boris believes there's a good chance that by the end of the year, people won't be using IDEs anymore.

[译文] [Answer]: 在 Anthropic，他们看到了模型变强的速度有多快，Boris 认为很有可能到今年年底，人们就不再使用 IDE 了。

[原文] [Answer]: They want to get ready for this future and avoid over-investing in UI and other layers on top, given how quickly the models are progressing.

[译文] [Answer]: 鉴于模型进步如此之快，他们希望为这个未来做好准备，避免在 UI 和其他上层架构上过度投资。

[原文] [Question]: How much have you used Claude Code for machine learning modeling and an autoML experience?

[译文] [Question]: 你们在多大程度上使用 Claude Code 进行机器学习建模和自动化机器学习（AutoML）体验？

[原文] [Answer]: They use it for this a bunch. Both engineers and researchers at Anthropic use Claude Code every day. About 80% of technical people at Anthropic use Claude Code every day.

[译文] [Answer]: 他们大量使用它做这个。Anthropic 的工程师和研究员每天都在使用 Claude Code。在 Anthropic，大约 80% 的技术人员每天都使用 Claude Code。

[原文] [Answer]: This includes researchers who use tools like the notebook tool to edit and run notebooks.

[译文] [Answer]: 这包括了那些使用工具（如 Notebook 工具）来编辑和运行 Notebooks 的研究员。