推文
760 tweets · 188 sources账号:全部mattpocockukdair_aisimonwbchernyadocompletetrq212dexhorthyyetonefelixrieseberg0xblacklightdani_avila7zarazhangruiAlchainHustbadlogicgamesdoteypetergyangvikingmuteryancarsonClaudeDevskunchenguidleon7haoctatedeviamzhihuikieranklaassenmckaywrigleylennysanthdxrRhysSullivangarrytanjakevin7johnlindquistkarpathyservasyy_aitheo0xMovez9hillsRLanceMartinaidenybaidingyihylarucodermitchellhmitsuhikomvanhornomarsar0steipetezeegGeminiAppGergelyOroszZaynHaoaddyosmaniantirezdanshipperdillon_mulroyelonmuskewind_devidoubiccnummanalirealWeZZardswyxthsottiauxyiliush0xPaulius0x_rodyAYi_AInotesAnthropicAIBarret_ChinaBenJames_____CMGS1988ChadMoranDanielMiesslerDaveJDavidKPianoDimillianFactoryAIFardeemMFarzaTVGoogleDeepMindHamelHusainHiTw93HilaShmuelIfanJewIndieDevHaileyJack_W_LindseyJackywineJasonZXJiaxi_CuiJinjingLiangKhalidWarsaLinearUncleLinghuaJMatthewBermanMaxForAIMeari_V2_0_GMnilaxQingQ77RaillyHugoSaccc_cTaniyatweets_VincentLogicai_explorer25aibuilderclub_alexalbert__alliekmillerandrewfarahantoinecojpanxue201arkuy99artmanasynkimobbssppllvvbearliubentossellbridgemindaibuaaxhmcailynyongyongcgtwtscharmaine_kleechrisbarberchrisparkXclairevodelba_oliveirademishassabisdoodlesteindriaforallelithrarelvissunericzakariassonfrancoisfleuretgabriell_labiamsahaj_xyzimdigitalashishimvihvjarredsumnerjerryjliu0jianshuojiayuan_jyjlongsterjonas_nellejoshalbrechtjpschroederkaiofreitaskepanolexruslogancyangmattypnabeelnifinetninthbit_aioran_gepatrickcpejmanjohnpetradonkaprathamgrvprathyvshquanruzhuoxiuquant_sheepraroquerealCaigurepsiacerohanpaul_aisamasamuelstroscheisawyerhoodshadcnshao__mengshaogefenhaoshreyansjtechgirl1908techwith_ramthorstenballtianyitobiaswuptoddsaunderstricaltturingoututuretomuniswap12untraceable_thevanillaCitronvelvet_sharkwengtianxinwquguruxiaogaifunxicilionycombinatorzachlloydtweetsziwenxu_zodchiii
@zarazhangrui·6月18日For You
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之前，不要用 AI 写作。 用 AI 写作本身没有错。危险在于，你还没建立起对「好内容」的判断力就开始用 AI 写。如果 AI 产出的是烂货，你甚至认不出来它是烂货。 多读，弄清什么是好的。多写，认清自己的声音是什么。只有在那之后，才用 AI 辅助写作——并确保输出读起来真的像你。
@zarazhangrui·6月14日For You
The recording of my full 45-min talk on my vibe coding process is up!
I go through the behind-the-scenes of how I get product ideas, how I talk to coding agents, how to prevent AI slop in design, and more
(YouTube link in comment) https://t.co/gWcF6hEPrl
@zarazhangrui
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 https://t.co/4wxpJjrAAB (Use the code PREMIUMPASS under "Add a coupon" to access the event for free) (Thanks to @sariazout and the @wwwsublimeapp community for hosting)
我的完整 45 分钟 vibe coding 过程演讲录像上线了！ 涵盖：如何获取产品创意、如何与编程 agent 沟通、如何避免 AI 设计 slop，以及为什么我认为代码是一种叙事媒介。 （YouTube 链接在评论区）
@zarazhangrui·6月8日For You
If you've adopted AI at your company but haven't seen any tangible results, read this 1990 article: "The Dynamo and the Computer" by Paul David.
When electricity first arrived, factories that "adopted" it barely got faster. They just swapped the steam engine for an electric one and ran everything else exactly as before: same machine layout, same workflow, same management. Electricity in, no real gains out.
The most common mistake with any new technology is to drop it into the old organization and then declare the transformation done.
The real leap came decades later, when each machine got its own small motor. Suddenly machines no longer had to be lined up around one central drive shaft. They could be rearranged around the actual flow of work.
The productivity gains didn't come from electricity. They came from REDESIGNING THE ENTIRE FACTORY around it.
AI is the same. Bolting it onto your existing process gets you a faster steam engine. The payoff comes when you redesign the work itself.
(link to paper in comments)
如果你的公司引入了 AI 却没看到实质成效，建议读这篇 1990 年的文章：Paul David 的《发电机与计算机》。 电力刚出现时，那些「采用」了电力的工厂几乎没变快。他们只是把蒸汽机换成了电动机，其他一切照旧：机器布局不变、工作流程不变、管理方式不变。电力进来了，但没有真正的收益。 对任何新技术，最常见的错误就是把它塞进旧组织，然后宣告转型完成。 真正的飞跃发生在几十年后——当每台机器都有了自己的小电机。机器不再需要排列在同一根中央传动轴周围，可以围绕实际的工作流程重新布局。 生产力的飞跃不是电力带来的，而是**围绕电力重新设计整个工厂**带来的。 AI 同理。把它硬接到现有流程上，你得到的只是一台快一点的蒸汽机。真正的回报，来自对工作本身的重新设计。
@zarazhangrui·6月3日For You
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
介绍飞书白板 skill：你的 agent 现在可以在飞书/Lark 文档中创建可编辑的 SVG 图形，支持 30+ 预设风格。适合概念可视化、技术架构图、总结会议/长文档，甚至替代 PPT。最棒的是所有内容完全可编辑（甚至可以拖拽元素）。
@zarazhangrui·5月22日For You
Introducing the Claude Code Lark/Feishu Bridge 🌉 (open-source)
Talk to Claude Code in Lark/Feishu like a colleague
- Use Claude Code on your phone via Lark chat
- Manage multiple CC sessions as group chats in Lark (one chat = one session), say goodbye to messy terminal tabs
- Claude Code can read all your work context in Lark (chat, docs, meeting transcripts, etc) via CLI
- Claude Code can write Lark Docs for you; you can even @ mention it in the comment and it will reply
- Forward any Lark messages to Claude and it can just get the task done
- Claude can send you interactive cards with buttons and UI
Open-source; try it now: https://t.co/V1eRIg87wx
介绍 Claude Code Lark/飞书 Bridge 🌉（开源） 像跟同事聊天一样在飞书里和 Claude Code 对话： - 通过飞书群聊在手机上使用 Claude Code - 用飞书群聊管理多个 CC session（一个群 = 一个 session），告别混乱的终端 tab - Claude Code 可通过 CLI 读取你在飞书的所有工作上下文（聊天、文档、会议记录等） - Claude Code 可以帮你写飞书文档；你甚至可以在评论里 @ 它，它会直接回复 - 把任意飞书消息转发给 Claude，它就能去执行任务 - Claude 可以向你发送带按钮和 UI 的互动卡片 已开源，立即体验：https://t.co/V1eRIg87wx
@zarazhangrui·5月16日For You
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 精神病：每天在两种心理状态之间循环 ↑ 用完 coding agent 之后：我他妈无所不能。我能构建任何东西。我这辈子从没感觉这么强大过。 ↓ 刷完 Twitter 之后：我他妈完全落后了。所有人都领先于我。浪潮正在推进，我要被落下了。
@zarazhangrui·5月12日For You
I noticed that AI tends to overuse italic fonts when designing webpages
It has become a telltale sign of AI design, just like overusing em-dashes have become a telltale sign of AI writing
Why does AI like italics so much?
我注意到 AI 在设计网页时倾向于过度使用斜体字。 这已经成为 AI 设计的标志性特征，就像过度使用破折号成为 AI 写作的标志一样。 AI 为什么这么喜欢斜体？
@zarazhangrui·5月11日For You
4 small but extremely useful things coding agents can do for you that many people don't know:
1. Copy things to your clipboard. If it generated something that needs to be copied & pasted somewhere, you can ask it to put it in your clipboard directly
2. Open HTML files in your browser, or open folders/files directly (e.g. play a video). No need to find it and manually open
3. Take screenshots. And make videos. For example, this video was entirely done by Claude Code. I just asked it to take screenshots of the HTML slide templates, and compile the images into a video using FFmpeg, showing them at 1 second interval
4. Manipulate files on your computer (e.g. organizing them into folders, renaming them, etc)
4 个 coding agent 能帮你做的小事，很多人不知道： 1. 复制内容到剪贴板。如果它生成了需要粘贴到其他地方的内容，可以直接让它放到剪贴板 2. 在浏览器中打开 HTML 文件，或直接打开文件夹/文件（比如播放视频），不用手动去找 3. 截图，录视频。比如这个视频就完全由 Claude Code 完成——我让它给 HTML 幻灯片模板截图，然后用 FFmpeg 把图片合成视频，每张显示 1 秒 4. 操作电脑上的文件（整理到文件夹、重命名等）
@zarazhangrui·5月9日For You
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!
用 OpenAI Realtime 2 API 做了一个「YouTube 实时 Copilot」浏览器扩展： Agent 会和你一起看视频，你可以随时用语音提问刚才说了什么。 最让我惊讶的地方是：它能区分 YouTube 的音频流和你自己的声音，不会把视频内容当成指令，也不会无缘无故开口——只在你提问时才响应。
@zarazhangrui·4月6日For You
I’ve been using the Fish Audio API to generate personalized podcasts. Dropping the skill soon
@FishAudio
Fish Audio performed a 10-day blind test across 5,098 real user comparisons on production traffic. Our S2 Pro model ranked #1 overall with a 66% win rate - outperforming the next closest competitor by over 60%. - S2 Pro beat ElevenLabs V3 60% to 40% - S2 Pro beat Inworld 80% to 20% - S2 Pro beat MiniMax 95% to 5% - #1 in every language tested No one knew they were part of a test. They just picked the voice they liked more. Check full methodology and results here → https://t.co/wFIc0oOy3x
我一直在用 Fish Audio API 生成个性化播客。skill 快要发布了。 [引用] Fish Audio 进行了为期 10 天的盲测，5098 个真实用户在生产流量上进行对比。S2 Pro 模型综合排名第一，胜率 66%，比排名第二的对手高出 60% 以上。 - S2 Pro 以 60:40 击败 ElevenLabs V3 - S2 Pro 以 80:20 击败 Inworld - S2 Pro 以 95:5 击败 MiniMax - 所有测试语言均排名第一 参与者并不知道自己在参与测试，他们只是选出更喜欢的声音。
@zarazhangrui·4月3日For You
My ideal workflow:
1. In the morning, I have a 15-min standup meeting with a master AI who gives me progress of all subagents & tasks. I provide feedback & give instructions in one focused meeting
2. During the day, we all go our separate ways. Agents focus on their work; I focus on my work
3. In the afternoon, repeat step 1
@zarazhangrui
My single biggest pain point right now: AI-induced attention deficit I have 5 Claude Code sessions running 10 Terminal tabs open 50 browser tabs open 100 X articles in my bookmarks When there are multiple AIs working for you, you're constantly task switching When you're waiting for AI to output, you're constantly distracted Because what used to be hard is now so easy, you always wanna do more, more, more Because of the speed of the AI news cycle, you feel a constant need to be on this app Add all this up and the result: you live in a constant state of distraction & delirium. Focus & deep work is harder than ever If anyone has figured this out pls teach me how
我理想中的工作流： 1. 早上，和一个 master AI 开 15 分钟站会，获取所有 subagent 和任务的进展。集中提供反馈和下达指令。 2. 白天，各走各的。Agent 专注它们的工作，我专注我的工作。 3. 下午，重复第 1 步。 引用自己的推文： > 我现在最大的痛点：AI 引发的注意力涣散。 > > 5 个 Claude Code session 同时跑 > 10 个 Terminal tab 开着 > 50 个浏览器 tab 开着 > 100 篇 X 文章存在书签里 > > 多个 AI 为你工作时，你在不停地切换任务；等 AI 输出时，你持续处于分心状态；因为以前很难的事现在变得太容易，你总想做更多更多；加上 AI 新闻循环的速度，你觉得必须一直刷这个 app。 > > 全加在一起：你活在持续的分心和混乱里，专注和深度工作比以前更难了。 > > 如果有人解决了这个问题，请教教我。
@zarazhangrui·3月24日For You
Frontend-slides now supports deploying your deck as a URL & exporting as PDF!
So you can share your vibe coded slides with others easily
Here's an example of a deployed deck: https://t.co/ZBolDpPMTV https://t.co/zNFl6fpEKL
@zarazhangrui
I created a Claude Skill that make beautiful slides on the web. The world hasn't woken up to the fact that code can create much better slides than most PPT tools. - Claude interviews you first about aesthetics, then generate a few directions to "show not tell", and you can pick your favorite - Cool transitions and animations - Interactive hover states and cursor effects - Auto-fits on any screen - Supports converting existing PPTX files to web-based slides; preserves original images and brand assets I asked Claude to make a slide show about this skill to showcase what it can do. Link to skill below
Frontend-slides 现在支持将你的幻灯片部署为 URL 并导出为 PDF！ 这样你就可以轻松地将 vibe coded 的幻灯片分享给其他人。 --- *引用原推：我创建了一个 Claude Skill，可以在 web 上制作精美幻灯片。世界还没意识到，代码能创造出比大多数 PPT 工具好得多的幻灯片。 - Claude 先访谈你的审美偏好，再生成几个方向来「show not tell」，你可以选择喜欢的 - 炫酷的转场和动画效果 - 交互式 hover 状态和光标特效 - 自适应任何屏幕 - 支持将现有 PPTX 转换为 web 幻灯片，保留原始图片和品牌资产*
760 tweets · 188 sources