Zara Zhang
@zarazhangrui ↗Builder. Dangerously skips permissions. Harvard’17. GitHub: https://t.co/KCuEajezlL YouTube: https://t.co/8xzbGWtf6w
I created a complete walkthrough of my Frontend Slides skill (22k+ stars on GitHub): - Complete demo of how to create beautiful HTML slides with Claude Code (beginner friendly) - How I created the skill - How you can create your own skill - Tips like how to insert images/videos into deck, how to publish - Lessons I learned (11-min video, YouTube link below)
我做了一份关于我的 Frontend Slides skill（在 GitHub 上有 22k+ stars）的完整演示指南：- 完整演示如何用 Claude Code 创建精美的 HTML slides（对新手友好）- 我是如何创建这个 skill 的 - 你如何创建你自己的 skill - 一些技巧，比如如何在 deck 中插入图片/视频、如何发布 - 我学到的经验教训（11 分钟视频，YouTube 链接见下）
A good rule of thumb for preventing AI slop (in writing, design, etc): Is your input (the context) longer than the output? I've found that for the AI to produce quality results, my input is often 3-5 times the length of the output. If your input is much shorter than the output, it's almost certainly going to produce slop
防止 AI slop（AI 生成的粗糙内容，在写作、设计等方面）的一个好经验法则是：你的输入（也就是 context，上下文）是否比输出更长？我发现，要让 AI 产出高质量结果，我的输入通常得是输出长度的 3 到 5 倍。如果你的输入比输出短得多，那它几乎肯定会产出 slop。
I hoard X bookmarks and never read them. So I built an extension that injects a bookmarked post into my main feed every time I open X (almost like an ad). Now I read my bookmarks. The trick was hijacking real estate I already check 50 times a day
我会囤很多 X 书签却从来不看。所以我做了一个扩展：每次我打开 X，它都会把一条已收藏的帖子插入我的主 feed（信息流）里，几乎像一条广告。现在我会去看我的书签了。诀窍在于“劫持”我已经每天会查看 50 次的版位空间。
Sometimes I wonder if joining a large company is actually riskier than joining a startup/starting your own company
有时候我会想，加入一家大公司，实际上会不会比加入一家 startup（初创公司）/自己创业风险更高。
Proactiveness sounds nice in theory but is so hard to get right
主动性这件事，理论上听起来很好，但真正要把握好却非常难。
The 3 most important things that set someone apart in the AI age: - Agency - Taste - Distribution
在 AI 时代，让一个人脱颖而出的 3 个最重要因素：- Agency（行动力）- Taste（品味）- Distribution（分发能力）
Frontend Slides skill: Recording of my talk:
Frontend Slides skill：我演讲的录屏：
Made a beautiful HTML deck using my Frontend Slides skill; very happy with how it turned out! Lots of easter eggs (e.g. you can click any image to enlarge them, lots of nested content/hyperlinks/interactive elements etc) Check it out:
我用我的 Frontend Slides skill 做了一套漂亮的 HTML 幻灯片；我对最终效果非常满意！里面有很多彩蛋（例如你可以点击任何图片将其放大，还有大量嵌套内容 / 超链接 / 交互元素等）来看看：
Don't use AI for writing until you develop your own taste and voice Using AI to write isn't inherently bad. The danger is using AI to write before you've developed your taste for what is good content. If the AI produces slop, you won't even recognize it as slop Read a lot to figure out what good looks like. Write a lot to know what your voice sounds like. Only then, use AI to help you write, but make sure it actually sounds like you
在你形成自己的品味和声音（voice）之前，不要用 AI 来写作。用 AI 写作本身并没有错。真正的危险在于：你还没培养出对什么是好内容的判断力，就先开始用 AI 写。如果 AI 产出的是 slop（粗制滥造的内容），你甚至都意识不到那是 slop。要大量阅读，弄清楚好的内容是什么样子；要大量写作，知道你自己的声音听起来是什么样。只有到了那时，才使用 AI 来帮助你写作，但要确保最后写出来的东西听起来确实像你。
The thing about vibe coded personal apps: Building the thing takes a day. Finding out if you'll actually use it takes a week. Most of my dead projects worked fine. I just never opened them. Most products are built for a person who doesn't exist. Someone who remembers to open the app, clicks the right button, does step 1, 2, 3 every day. Real humans are lazy & forgetful. Build for that person instead
关于那些 vibe coded 的个人 app，有一点很真实：把东西做出来只要一天，弄清楚你到底会不会真的去用它却要一周。我的大多数夭折项目其实都能正常运行，我只是从来没有再打开过它们。大多数产品，都是为一个并不存在的人设计的——一个会记得打开 app、会点对按钮、会每天按部就班做完第 1、2、3 步的人。现实中的人类懒惰又健忘。所以，你应该为那样的人去构建产品。
You don’t have to chase the cool thing. Do whatever you’re already doing so well that it becomes the cool thing. The cool thing is rarely cool when it starts. It becomes cool because someone went deep on something everyone else thought was too small.
你不必追着所谓的酷东西跑。把你已经在做的事做到足够好，好到它自己变成那个酷东西。所谓酷的东西，在刚开始时很少真的显得酷。它之所以后来变酷，是因为有人在一件别人都觉得太小、不值得做深的事情上，真正钻了进去。
Every other product right now is "an AI agent that does everything in your work & life & integrates with everything." Cool, that's just Claude/Codex. If you want me to use your thing instead, it needs an opinion & a soul. Build small & sharp, not big & generic. Doing everything means doing nothing. Speaking to everybody means speaking to nobody.
现在市面上几乎每个产品都是“一个能包办你工作和生活中一切、还能和所有东西集成的 AI agent（智能体）”。行啊，那不就是 Claude/Codex 吗。如果你想让我改用你的东西，它就得有自己的立场和灵魂。要做得小而锋利，不要做得又大又泛。什么都做，往往就等于什么都没做好。想对所有人说话，最后就等于谁也没真正说到。
I just reached 70k followers on X 🙏🏼 X is where I learn in public & build in public. I've learned so much and met so many wonder people here If you're curious about how I grew my followers while being completely authentic, read my article below
我刚刚在 X 上达到了 7 万 followers（关注者）🙏🏼 X 是我公开学习、公开构建的地方。我在这里学到了很多，也认识了许多很棒的人。如果你想知道我是在完全真实做自己的情况下，如何增长 followers（关注者）的，读读我下面这篇文章。
You make a skill by ending with one, not starting with one
你获得一项 skill，是以“完成一个”为终点，而不是以“开始做一个”为起点。
You don't make a good skill by writing a skill. You make it by doing the thing, fixing it 20 times, then telling the AI to bottle up everything you just did.
你不会因为去写一个 skill 就做出一个好的 skill。你是通过把那件事真正做完、修修补补 20 次，然后再告诉 AI 把你刚刚做过的一切都封装起来，才做出一个好的 skill。
Great essay on taste and how to cultivate it Taste is not just personal preference; good taste requires mastery and experience
这是一篇关于品味以及如何培养品味的精彩文章。品味不仅仅是个人偏好；好的品味需要掌握力与经验。
There are SO many people, friends, acquaintances, followers, startups, asking me to try their new AI products every day Like I get at least 3 such requests per day If I tried all of these products I wouldn’t have time to do anything else There are too many builders and the competition for attention is insane
每天都有非常非常多的人——朋友、熟人、followers、startups——来找我试用他们新的 AI 产品。像这种请求，我每天至少会收到 3 个。如果我要把这些产品全都试一遍，那我就没时间做别的任何事了。做产品的人实在太多了，而对注意力的竞争也疯狂得离谱。
A viral product has a founder people can see and hear People buy from people. A screen recording from the founder beats a corporate promo video or a wall of features. Show your face.
一个能 viral（爆火传播）的产品，背后都有一个人们看得见、听得到的 founder。人们是向人买东西的。由 founder 出镜的 screen recording（录屏），比企业宣传视频或者一整面功能清单都更有说服力。把你的脸露出来。
This is tomorrow! Sneak peek of the deck I'll be sharing:
就在明天！先提前看一眼我将要分享的 deck（演示文稿）：
This is so good Increasingly the output of an agency looks like a folder of files for agents, instead of one-off assets "Get paid for your mind, not your hands"
这太棒了。agency 的产出越来越像是一个给 agent 使用的文件夹，而不是一次性的成品资产。“Get paid for your mind, not your hands”
People should build agents/skills for their cross-functional teams. For example, if a design team builds a design agent/skill for the marketing team (trained on all of the brand's guidelines and design patterns), then the marketing team can produce more on-brand assets without having to bug designers every time But the marketing team couldn't have built this on their own; it takes the designer with their expertise, context, and knowledge Same thing applies to every pair of teams that work closely with each other, often rely on each other, and complain about each other's limited bandwidth Building agents for your cross-functional teams ensures each team can be more self-sufficient, and moves us to a direction where teams can be organized by "loops" rather than "functions"
人们应该为自己的 cross-functional teams（跨职能团队）构建 agent/skills。如果一个 design team 为 marketing team 构建了一个 design agent/skill（基于品牌的全部 guidelines 和 design patterns 进行训练），那么 marketing team 就能产出更多符合品牌调性的资产，而不必每次都去麻烦设计师。但 marketing team 不可能靠自己做出这个东西；这需要设计师及其专业能力、上下文和知识。这个逻辑同样适用于任何一对紧密协作、经常相互依赖、又总抱怨彼此带宽有限的团队。为你的 cross-functional teams 构建 agent，能确保每个团队都更具自给自足能力，也会推动我们走向一种按“loops”而不是按“functions”来组织团队的方向。
It seems like most startups in San Francisco are selling products to each other When I ask founders who their target audience is, 90% of the time it's "engineering and product teams, AI-native startups" Feels like the same small group of target audience is being bombarded with a million products, whereas very few people are building for the 99% of the world
看起来，San Francisco 的大多数 startup 都在互相卖产品。当我问创始人他们的目标受众是谁时，90% 的情况下答案都是“engineering 和 product teams，AI-native startups”。感觉同样这一小撮目标用户，正在被海量产品轮番轰炸，而与此同时，真正为这个世界 99% 的人构建产品的人却很少。
The barrier for non-technical people using coding agents was never the interface. Chatting is the easiest UI ever invented The barrier is that they don't know what to ask for. The blank chat box assumes you already know what's possible. Most people don't. When I tried Town, I was impressed by its onboarding process where the agent proactively suggested workflows & things it can take off my plate, rather than waiting for me to give instructions
对非技术人士来说，使用 coding agents（编程 agent）的门槛从来都不是界面。聊天几乎是有史以来最容易上手的 UI。真正的门槛在于，他们不知道该提出什么需求。那个空白的聊天框默认你已经知道它能做什么，但大多数人并不知道。我试用 Town 时，对它的 onboarding（引导上手）流程印象很深：agent 会主动建议可行的工作流，以及它能帮我分担哪些事情，而不是等着我下指令。
I will be doing a virtual talk this Friday on my vibe coding process. As a non-technical person who has somehow managed to get 30k stars on GitHub, I will be sharing the behind-the-scenes of how I get product ideas, how I work with coding agents, how to design stuff that's not AI slop, and why I think code is a medium for storytelling. RSVP here (Use the code PREMIUMPASS under "Add a coupon" to access the event for free) (Thanks to @sariazout and the @wwwsublimeapp community for hosting)
这周五我会做一场关于我的 vibe coding 流程的线上分享。作为一个非技术人士，却不知怎么在 GitHub 上拿到了 30k stars，我会分享一些幕后的东西：我是如何获得产品灵感的、我是如何与 coding agents 协作的、怎样设计出不是 AI slop 的东西，以及为什么我认为代码是一种讲故事的媒介。请在这里 RSVP（在 “Add a coupon” 下使用代码 PREMIUMPASS，即可免费参加活动）。（感谢 @sariazout 和 @wwwsublimeapp community 的邀请与主办。）
Actually I think the new world might be: Markdown, HTML, SVG SVG is underrated
其实我觉得，新的世界可能会是：Markdown、HTML、SVG。SVG 被低估了。
This part is so well-written and resonated SO much: "I am the programming equivalent of a home cook"
这部分写得太好了，而且让我产生了强烈共鸣：“我就像编程世界里的家庭厨师。”
[1]
I think one reason that my Frontend Slides skill has grown so much organically is because slides are inherently social People see these cool slides and always ask "how did you make it". And people tend to perceive those using HTML decks as more AI-native and AI-savvy
我认为，我的 Frontend Slides 技能之所以能如此自然地大幅成长，一个原因是 slides 天生就具有社交属性。人们看到这些很酷的 slides 时，总会问：“你是怎么做出来的？”而且，人们往往会觉得，使用 HTML decks 的人更 AI-native，也更 AI-savvy。
Really enjoyed this talk! The value of static content is going down, the value of live interaction is going up People want to connect with the human being behind a piece of work, whether it's content or software Raw & opinionated > polished & generic
非常喜欢这个演讲！静态内容的价值正在下降，实时互动的价值正在上升。人们想要与一件作品背后的人建立连接，无论那是内容还是软件。原始而有观点的内容，比起打磨精致却千篇一律的内容，更有价值。
Introducing the Beautiful Feishu Whiteboard skill: Your agent can now create EDITABLE SVG graphics in Feishu/Lark docs in 30+ predefined styles Great for concept visualization, technical architecture diagrams, summarizing meetings/long documents, or even replacing PPTs The best part is everything is completely editable (can even drag elements around) Skill link in comment
隆重推出 Beautiful Feishu Whiteboard skill：你的 agent 现在可以在 Feishu/Lark 文档中创建可编辑的 SVG 图形，提供 30 多种预设风格。非常适合用于概念可视化、技术架构图、总结会议/长文档，甚至替代 PPT。最棒的是，所有内容都可以完全编辑（甚至还能拖动其中的元素）。Skill 链接见评论区。
From OpenAI's latest Codex report: "Knowledge workers now represent about 20% of Codex users and are adopting it more than 3 times as fast as developers The fastest growing task types for knowledge workers are: Data Analysis (110% growth week over week); Research (+37%), and Knowledge Artifacts (+36%)"
来自 OpenAI 最新的 Codex 报告：“Knowledge workers（知识工作者）现在约占 Codex 用户的 20%，其采用速度比 developers（开发者）快 3 倍以上。对知识工作者而言，增长最快的任务类型是：Data Analysis（数据分析，周环比增长 110%）；Research（研究，+37%）；以及 Knowledge Artifacts（知识产物，+36%）。”
Frontend Slides now has 20k stars on GitHub 😮 It's insane how many people have already replaced PPTs with HTML decks and never looked back Recently added new features including beautiful templates, publish as webpage/export as PDF, and inline editing
Frontend Slides 现在在 GitHub 上已经有 20k stars 了 😮 已经有这么多人用 HTML decks 替代 PPTs，而且再也没回头，这简直太夸张了。最近新增了不少新功能，包括精美模板、发布为网页/导出为 PDF，以及 inline editing（行内编辑）。
Why do I get so annoyed when a coding agent ends a message with "just say the word" You're my cofounder, not my servant
为什么当一个 coding agent（编程智能体）在消息结尾写上“just say the word”时，我会这么恼火？你是我的 cofounder（联合创始人），不是我的 servant（仆人）。
“Real mastery is not exerting the most effort. It is achieving the outcome with the least necessary effort. Grinding is never good for any creative problem”
“真正的 mastery（精通）并不是付出最多努力，而是用最少的必要努力达成结果。对任何创造性问题来说，grinding（埋头苦干）都绝不是好事。”
One thing I noticed about Opus 4.8 is that it stopped using em dashes in writing
我注意到的关于 Opus 4.8 的一件事是，它在写作中不再使用 em dashes 了
How my own usage of coding agents has changed in the past month: 1. Moved from the terminal to the Codex/Claude Code desktop apps. The Codex Mac app is especially great. Now I rarely open the terminal 2. Transitioned from mostly Claude Code to 50-50 Claude Code and Codex. Codex feels like a very reliable engineer; Claude Code is a better PM and designer with good communication skills So I go to Codex if I already have a defined task and just need it to work; I go to Claude Code if I don't yet know what I want and just wanna brainstorm/prototype
我自己使用 coding agents 的方式在过去一个月里发生了这些变化：1. 从 terminal 转到 Codex/Claude Code 桌面 app。Codex Mac app 尤其出色。现在我很少再打开 terminal 了。2. 从主要使用 Claude Code，转变为 Claude Code 和 Codex 各占一半。Codex 感觉像一位非常可靠的 engineer；Claude Code 则更像更擅长沟通的 PM 和 designer。所以，如果我已经定义好了任务、只需要它把事情做成，我就会用 Codex；如果我还不知道自己想要什么、只是想 brainstorm（头脑风暴）/prototype（做原型），我就会用 Claude Code。
Frontend Slides skill now has 19k stars on GitHub ✨ I upgraded it with a new design brain: it can now pull from my Beautiful HTML Templates library, pick a visual direction, and generate slides using that template’s design language Also new: - Works better with coding agents beyond Claude Code - Export to webpage or PDF for easy sharing - Inline slide editing - Fixed 16:9 deck stage to avoid overflow, cropping, and weird responsive chaos Try now:
Frontend Slides skill 现在在 GitHub 上已经有 19k stars ✨ 我给它升级了一个新的 design brain：它现在可以调用我的 Beautiful HTML Templates library，选择一种视觉方向，并使用该模板的 design language 来生成 slides。另有更新：- 对 Claude Code 之外的 coding agents 兼容更好了 - 可导出为网页或 PDF，方便分享 - 支持 inline（行内）slide 编辑 - 修复了 16:9 deck stage，以避免溢出、裁切和奇怪的响应式混乱 现在就试试：
I think that in an AI-native team, - ICs should start thinking like managers: how to delegate tasks to agents, how to set standards & verify output - Managers should start thinking like ICs: how to be more hands-on builders instead of just doing people management
我认为，在一个 AI-native 团队里，IC 应该开始像 manager 一样思考：如何把任务委派给 agent，如何设定标准并验证输出；manager 也应该开始像 IC 一样思考：如何成为更亲自动手的 builder，而不只是做人事管理
Great slide from the “How to thrive as an AI-era developer” session at Google I/O today I think this T-shape will apply to not just developers but every job function We need to - go deeper with our domain expertise - go wider with adjacent skills and fields - learn to use AI well on top
今天在 Google I/O 的 “How to thrive as an AI-era developer” 场次里看到一张很棒的 slide。我认为，这种 T-shape 不仅适用于 developer，也适用于所有岗位职能。我们需要——在自己的领域专长上钻得更深；把相邻的技能和领域拓得更宽；并在此基础上学会用好 AI
Keep getting this error in Claude Code recently; is it just me or are others getting this too? How to resolve this? API Error: The socket connection was closed unexpectedly. For more information, pass verbose: true in the second argument to fetch()
我最近在 Claude Code 里一直遇到这个错误；是只有我这样，还是其他人也会遇到？该怎么解决？API Error：socket 连接被意外关闭了。更多信息请在传给 fetch() 的第二个参数中设置 verbose: true
If you have had success with Gbrain/LLM Wiki/other context management techniques for agents, and you're in the Bay Area, would love to invite you to demo at this event! RSVP: Cohosted with @NotionHQ @radicalvcfund @juleszqiu
如果你在用于 agent 的 Gbrain / LLM Wiki / 其他上下文管理技术方面有成功经验，而且你人在 Bay Area，很希望邀请你来这个活动做 demo（演示）！RSVP：与 @NotionHQ、@radicalvcfund、@juleszqiu 联合主办
People consistently overestimate how hard it is to build something and underestimate how hard it is to win people’s attention once you’ve built it
人们总是持续地高估“把某个东西做出来”有多难，却低估了在你把它做出来之后，赢得人们注意力有多难。
AI psychosis: cycling between two mental states every single day ↑ after using coding agents: holy shit I'm omnipotent. I can build anything. I've never felt this powerful in my life. ↓ after scrolling twitter: holy shit I'm completely behind. everyone's ahead. the wave is moving and I'm going to get left.
AI psychosis（AI 精神错乱）：在使用 coding agents（编程 agent）之后，我每天都在两种心理状态之间来回切换。↑ 天啊，我无所不能。我什么都能造出来。我这辈子从没觉得自己这么强大过。↓ 刷完 twitter 之后：天啊，我彻底落后了。所有人都领先了。这股浪潮正在向前，我就要被甩下了。
I will be co-hosting an event in SF next Sat (5/23)! I've been thinking a lot about how to manage & curate CONTEXT for agents. I'm inviting everyone who's thought a lot about this to demo your real workflows & show what's actually working. Format: A series of rapid-fire 5-minute demos, followed by Q&A. Screen shares of real workflows only; no slides. Not demoing? Come hang out as a participant. Either way, expect an afternoon of practical workflows, fresh ideas, and new friends. RSVP here: Cohosted w @NotionHQ & @radicalvcfund (@juleszqiu)
我下周六（5/23）会在 SF 共同主办一场活动！我最近一直在深入思考，如何为 agent 管理和策划 CONTEXT（上下文）。我邀请所有认真思考过这个问题的人来现场 demo（演示）你们真实的工作流（workflow），并展示到底哪些方法真正有效。形式：一系列快节奏的 5 分钟 demo，之后是 Q&A（问答）环节。只接受真实工作流的屏幕共享；不放 slides（幻灯片）。不做 demo？也欢迎你作为参与者来一起交流。无论哪种方式，你都可以期待一个充满实用工作流、新鲜想法和新朋友的下午。请在此 RSVP（回复是否出席）：由 @NotionHQ 与 @radicalvcfund (@juleszqiu) 共同主办
HTML is important because humans are visual animals As the founder of Duolingo once said, one of his biggest lessons after spending years building consumer products is: "People don't read" In the past, we optimized our output format for human manipulation (eg. pushing pixels in powerpoint) When AI will handle the manipulation, the output format should be optimized for human CONSUMPTION instead Turns out humans like to consume beautiful, interactive artifacts
HTML 之所以重要，是因为人类是视觉动物。正如 Duolingo 的创始人曾说过的那样，他在多年打造面向消费者的产品后得到的最大教训之一是：“People don't read”。过去，我们为便于人类操作而优化输出格式（例如，在 powerpoint 里推像素）。当 AI 将负责这些操作时，输出格式就应该转而为人类的 CONSUMPTION（消费、接收）而优化。事实证明，人类喜欢消费美观、可交互的成品。
Built a "YouTube realtime copilot" browser extension using OpenAI's realtime 2 API: The agent watches the video alongside you, and can answer any question you have about what was just said via realtime voice chat. The crazy part to me is: It can differentiate the YouTube's audio stream and your voice, so it doesn't confuse the video as commands, and stays silent unless you ask something!
我用 OpenAI 的 realtime 2 API 做了一个“YouTube realtime copilot”浏览器扩展：这个 agent 会和你一起“观看”视频，并且可以通过 realtime（实时）语音聊天，回答你关于刚刚说了什么的任何问题。对我来说最疯狂的是：它能区分 YouTube 的音频流和你的声音，所以不会把视频内容误当成指令，而且除非你主动提问，否则它会保持安静！
All 32 Beautiful HTML Slide Templates are now available on AnyGen, it's plug-and-play even for those without a coding agent Use them now:
现在，32 个精美的 HTML 幻灯片模板都已经上线 AnyGen 了；即使是没有 coding agent 的人也能即插即用。现在就用吧：
Can confirm GPT realtime 2 feels like black magic; unlocks so many new applications!! Building with it now
可以确认，GPT realtime 2 感觉就像黑魔法一样；它解锁了太多全新的应用场景！！我现在就在用它做开发
I get so many messages about new AI apps but their all messaging all blur into each other
我收到很多关于新 AI app 的消息，但它们传达的信息全都模糊地混在一起，彼此几乎没有区别。
Most memorable parts of this interview with @bcherny : - Boris casually mention that he has "thousands" of agents running during the night... - Boris saying he almost exclusively uses Claude Code on the phone now - Coding will become the new literacy, just like reading and writing
这次对 @bcherny 的采访里最令人印象深刻的部分是：- Boris 很随意地提到，他晚上会有“几千个” agent（智能体）在运行…… - Boris 说，他现在几乎只在手机上使用 Claude Code - 编码将成为新的读写能力，就像阅读和写作一样
this is the coolest demo I've seen a while, a must-watch for anyone interested in human-agent interaction!
这是我最近一段时间里看到过最酷的 demo，对任何对 human-agent interaction（人与 agent 交互）感兴趣的人来说都绝对值得一看！
Before AI, you couldn't afford to build something small. Because the cost of software development was so high, you had to hire a team, convince other people, justify it to a committee. Now, it's literally just you and a coding agent. A coding agent doesn't need to be convinced. It will readily build whatever crazy and weird idea you have. So go build something that would get rejected in every big tech company's product review meeting.
在 AI 出现之前，你负担不起去做一个小东西。因为软件开发的成本太高了，你必须雇一个团队，说服其他人，向委员会证明这件事的合理性。现在，真的就只剩下你和一个 coding agent 了。coding agent 不需要被说服。无论你的想法多么疯狂、多么古怪，它都会很乐意把它做出来。所以，去做点那种在任何大型科技公司的产品评审会上都会被否掉的东西吧。
[1]
[1]
Using ChatGPT Image 2 to make slides is SO fun
用 ChatGPT Image 2 来制作幻灯片实在太有趣了
We should start thinking about the IT/internal tools team as more like “HR for agents”
我们应该开始把 IT/internal tools 团队看得更像是“HR for agents（agent 的人力资源）”。
Don't get AI to generate images; get them to generate SVGs! Vector illustrations that seamlessly blend into the design style is the perfect complement to HTML Slides (Here I'm using the @QuiverAI API inside @AnyGenIO's Frontend Slides feature
不要让 AI 去生成图片；要让它们生成 SVG！能够无缝融入设计风格的矢量插画，是 HTML Slides 的完美补充（这里我是在 @AnyGenIO 的 Frontend Slides 功能中使用 @QuiverAI API）
Very interested in this I used to build web apps with GUI; now I just share GitHub repos with the code and people can customize the UI however they like Kind of like Karpathy's "idea file": just write down the idea, and people's agents can build it for them in a customized way
我对这个非常感兴趣。我以前是用 GUI（图形用户界面）来构建 web apps；现在我只需要分享带有代码的 GitHub repos，别人就可以按自己喜欢的方式自定义 UI。有点像 Karpathy 的 “idea file”：只要把想法写下来，人们的 agents（智能体）就能以定制化的方式替他们把它构建出来
The most sophisticated AI users I know are talking to just ONE agent daily, not multiple agents That agents will then assign tasks to other subagents/spawn subagents as needed But the day-to-day communication is with just 1 agent, who serves as the overall orchestrator/router
我认识的那些最资深的 AI 用户，每天都只和 ONE 个 agent（智能体）交流，而不是同时和多个 agents（智能体）交流。这个 agent 会在需要时把任务分配给其他 subagents（子智能体）或生成新的 subagents（子智能体）。但日常沟通的对象始终只有 1 个 agent，它充当整体的 orchestrator/router（编排者/路由器）。
When I sit down at my computer and start typing, that's not the beginning of writing. It's the LAST step. 80% of the writing is already done. It just happened in my head. - Writing ≠ generating text - Building product ≠ writing PRDs - Designing ≠ making mockups - Engineering ≠ writing code
当我坐到电脑前开始打字时，那并不是写作的开始，而是最后一步。80% 的写作其实已经完成了，只不过它发生在我的脑海里。—— 写作 ≠ 生成文本 —— 构建产品 ≠ 写 PRD —— 设计 ≠ 制作 mockup —— 工程开发 ≠ 写代码
Agents speak HTML as their native language Let agents express themselves in their native language (for similar reasons, agents produce much better looking slides in HTML than in XML)
agent 把 HTML 当作它们的母语来使用；让 agent 用它们的母语来表达自己（出于类似的原因，agent 用 HTML 生成的幻灯片也会比用 XML 生成的好看得多）。